xref: /freebsd/sys/netinet/in_pcb.h (revision 8a0a413e)
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  * $FreeBSD$
38  */
39 
40 #ifndef _NETINET_IN_PCB_H_
41 #define _NETINET_IN_PCB_H_
42 
43 #include <sys/queue.h>
44 #include <sys/_lock.h>
45 #include <sys/_mutex.h>
46 #include <sys/_rwlock.h>
47 #include <net/route.h>
48 
49 #ifdef _KERNEL
50 #include <sys/lock.h>
51 #include <sys/rwlock.h>
52 #include <net/vnet.h>
53 #include <vm/uma.h>
54 #endif
55 
56 #define	in6pcb		inpcb	/* for KAME src sync over BSD*'s */
57 #define	in6p_sp		inp_sp	/* for KAME src sync over BSD*'s */
58 
59 /*
60  * struct inpcb is the common protocol control block structure used in most
61  * IP transport protocols.
62  *
63  * Pointers to local and foreign host table entries, local and foreign socket
64  * numbers, and pointers up (to a socket structure) and down (to a
65  * protocol-specific control block) are stored here.
66  */
67 LIST_HEAD(inpcbhead, inpcb);
68 LIST_HEAD(inpcbporthead, inpcbport);
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.
75  */
76 struct in_addr_4in6 {
77 	u_int32_t	ia46_pad32[3];
78 	struct	in_addr	ia46_addr4;
79 };
80 
81 /*
82  * NOTE: ipv6 addrs should be 64-bit aligned, per RFC 2553.  in_conninfo has
83  * some extra padding to accomplish this.
84  * NOTE 2: tcp_syncache.c uses first 5 32-bit words, which identify fport,
85  * lport, faddr to generate hash, so these fields shouldn't be moved.
86  */
87 struct in_endpoints {
88 	u_int16_t	ie_fport;		/* foreign port */
89 	u_int16_t	ie_lport;		/* local port */
90 	/* protocol dependent part, local and foreign addr */
91 	union {
92 		/* foreign host table entry */
93 		struct	in_addr_4in6 ie46_foreign;
94 		struct	in6_addr ie6_foreign;
95 	} ie_dependfaddr;
96 	union {
97 		/* local host table entry */
98 		struct	in_addr_4in6 ie46_local;
99 		struct	in6_addr ie6_local;
100 	} ie_dependladdr;
101 	u_int32_t	ie6_zoneid;		/* scope zone id */
102 };
103 #define	ie_faddr	ie_dependfaddr.ie46_foreign.ia46_addr4
104 #define	ie_laddr	ie_dependladdr.ie46_local.ia46_addr4
105 #define	ie6_faddr	ie_dependfaddr.ie6_foreign
106 #define	ie6_laddr	ie_dependladdr.ie6_local
107 
108 /*
109  * XXX The defines for inc_* are hacks and should be changed to direct
110  * references.
111  */
112 struct 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 
125 #define	inc_isipv6	inc_flags	/* temp compatibility */
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.  A
140  * few fields are protected by multiple locks as indicated in the locking notes
141  * below.  For these fields, all of the listed locks must be write-locked for
142  * any modifications.  However, these fields can be safely read while any one of
143  * the listed locks are read-locked.  This model can permit greater concurrency
144  * for read operations.  For example, connections can be looked up while only
145  * holding a read lock on the global pcblist lock.  This is important for
146  * performance when attempting to find the connection for a packet given its IP
147  * and port tuple.
148  *
149  * One noteworthy exception is that the global pcbinfo lock follows a different
150  * set of rules in relation to the inp_list field.  Rather than being
151  * write-locked for modifications and read-locked for list iterations, it must
152  * be read-locked during modifications and write-locked during list iterations.
153  * This ensures that the relatively rare global list iterations safely walk a
154  * stable snapshot of connections while allowing more common list modifications
155  * to safely grab the pcblist lock just while adding or removing a connection
156  * from the global list.
157  *
158  * Key:
159  * (c) - Constant after initialization
160  * (g) - Protected by the pcbgroup lock
161  * (i) - Protected by the inpcb lock
162  * (p) - Protected by the pcbinfo lock for the inpcb
163  * (l) - Protected by the pcblist lock for the inpcb
164  * (h) - Protected by the pcbhash lock for the inpcb
165  * (s) - Protected by another subsystem's locks
166  * (x) - Undefined locking
167  *
168  * A few other notes:
169  *
170  * When a read lock is held, stability of the field is guaranteed; to write
171  * to a field, a write lock must generally be held.
172  *
173  * netinet/netinet6-layer code should not assume that the inp_socket pointer
174  * is safe to dereference without inp_lock being held, even for protocols
175  * other than TCP (where the inpcb persists during TIMEWAIT even after the
176  * socket has been freed), or there may be close(2)-related races.
177  *
178  * The inp_vflag field is overloaded, and would otherwise ideally be (c).
179  *
180  * TODO:  Currently only the TCP stack is leveraging the global pcbinfo lock
181  * read-lock usage during modification, this model can be applied to other
182  * protocols (especially SCTP).
183  */
184 struct icmp6_filter;
185 struct inpcbpolicy;
186 struct m_snd_tag;
187 struct inpcb {
188 	/* Cache line #1 (amd64) */
189 	LIST_ENTRY(inpcb) inp_hash;	/* (h/i) hash list */
190 	LIST_ENTRY(inpcb) inp_pcbgrouphash;	/* (g/i) hash list */
191 	struct rwlock	inp_lock;
192 	/* Cache line #2 (amd64) */
193 #define	inp_start_zero	inp_refcount
194 #define	inp_zero_size	(sizeof(struct inpcb) - \
195 			    offsetof(struct inpcb, inp_start_zero))
196 	u_int	inp_refcount;		/* (i) refcount */
197 	int	inp_flags;		/* (i) generic IP/datagram flags */
198 	int	inp_flags2;		/* (i) generic IP/datagram flags #2*/
199 	void	*inp_ppcb;		/* (i) pointer to per-protocol pcb */
200 	struct	socket *inp_socket;	/* (i) back pointer to socket */
201 	struct	inpcbinfo *inp_pcbinfo;	/* (c) PCB list info */
202 	struct	inpcbgroup *inp_pcbgroup; /* (g/i) PCB group list */
203 	LIST_ENTRY(inpcb) inp_pcbgroup_wild; /* (g/i/h) group wildcard entry */
204 	struct	ucred	*inp_cred;	/* (c) cache of socket cred */
205 	u_int32_t inp_flow;		/* (i) IPv6 flow information */
206 	u_char	inp_vflag;		/* (i) IP version flag (v4/v6) */
207 	u_char	inp_ip_ttl;		/* (i) time to live proto */
208 	u_char	inp_ip_p;		/* (c) protocol proto */
209 	u_char	inp_ip_minttl;		/* (i) minimum TTL or drop */
210 	uint32_t inp_flowid;		/* (x) flow id / queue id */
211 	struct m_snd_tag *inp_snd_tag;	/* (i) send tag for outgoing mbufs */
212 	uint32_t inp_flowtype;		/* (x) M_HASHTYPE value */
213 	uint32_t inp_rss_listen_bucket;	/* (x) overridden RSS listen bucket */
214 
215 	/* Local and foreign ports, local and foreign addr. */
216 	struct	in_conninfo inp_inc;	/* (i) list for PCB's local port */
217 
218 	/* MAC and IPSEC policy information. */
219 	struct	label *inp_label;	/* (i) MAC label */
220 	struct	inpcbpolicy *inp_sp;    /* (s) for IPSEC */
221 
222 	/* Protocol-dependent part; options. */
223 	struct {
224 		u_char	inp_ip_tos;		/* (i) type of service proto */
225 		struct mbuf		*inp_options;	/* (i) IP options */
226 		struct ip_moptions	*inp_moptions;	/* (i) mcast options */
227 	};
228 	struct {
229 		/* (i) IP options */
230 		struct mbuf		*in6p_options;
231 		/* (i) IP6 options for outgoing packets */
232 		struct ip6_pktopts	*in6p_outputopts;
233 		/* (i) IP multicast options */
234 		struct ip6_moptions	*in6p_moptions;
235 		/* (i) ICMPv6 code type filter */
236 		struct icmp6_filter	*in6p_icmp6filt;
237 		/* (i) IPV6_CHECKSUM setsockopt */
238 		int	in6p_cksum;
239 		short	in6p_hops;
240 	};
241 	LIST_ENTRY(inpcb) inp_portlist;	/* (i/h) */
242 	struct	inpcbport *inp_phd;	/* (i/h) head of this list */
243 	inp_gen_t	inp_gencnt;	/* (c) generation count */
244 	struct llentry	*inp_lle;	/* cached L2 information */
245 	rt_gen_t	inp_rt_cookie;	/* generation for route entry */
246 	union {				/* cached L3 information */
247 		struct route inp_route;
248 		struct route_in6 inp_route6;
249 	};
250 	LIST_ENTRY(inpcb) inp_list;	/* (p/l) list for all PCBs for proto */
251 	                                /* (p[w]) for list iteration */
252 	                                /* (p[r]/l) for addition/removal */
253 };
254 #endif	/* _KERNEL */
255 
256 #define	inp_fport	inp_inc.inc_fport
257 #define	inp_lport	inp_inc.inc_lport
258 #define	inp_faddr	inp_inc.inc_faddr
259 #define	inp_laddr	inp_inc.inc_laddr
260 
261 #define	in6p_faddr	inp_inc.inc6_faddr
262 #define	in6p_laddr	inp_inc.inc6_laddr
263 #define	in6p_zoneid	inp_inc.inc6_zoneid
264 #define	in6p_flowinfo	inp_flow
265 
266 #define	inp_vnet	inp_pcbinfo->ipi_vnet
267 
268 /*
269  * The range of the generation count, as used in this implementation, is 9e19.
270  * We would have to create 300 billion connections per second for this number
271  * to roll over in a year.  This seems sufficiently unlikely that we simply
272  * don't concern ourselves with that possibility.
273  */
274 
275 /*
276  * Interface exported to userland by various protocols which use inpcbs.  Hack
277  * alert -- only define if struct xsocket is in scope.
278  * Fields prefixed with "xi_" are unique to this structure, and the rest
279  * match fields in the struct inpcb, to ease coding and porting.
280  *
281  * Legend:
282  * (s) - used by userland utilities in src
283  * (p) - used by utilities in ports
284  * (3) - is known to be used by third party software not in ports
285  * (n) - no known usage
286  */
287 #ifdef _SYS_SOCKETVAR_H_
288 struct xinpcb {
289 	size_t		xi_len;		/* length of this structure */
290 	struct xsocket	xi_socket;		/* (s,p) */
291 	struct in_conninfo inp_inc;		/* (s,p) */
292 	uint64_t	inp_gencnt;		/* (s,p) */
293 	union {
294 		void	*inp_ppcb;		/* (s) netstat(1) */
295 		int64_t	ph_ppcb;
296 	};
297 	int64_t		inp_spare64[4];
298 	uint32_t	inp_flow;		/* (s) */
299 	uint32_t	inp_flowid;		/* (s) */
300 	uint32_t	inp_flowtype;		/* (s) */
301 	int32_t		inp_flags;		/* (s,p) */
302 	int32_t		inp_flags2;		/* (s) */
303 	int32_t		inp_rss_listen_bucket;	/* (n) */
304 	int32_t		in6p_cksum;		/* (n) */
305 	int32_t		inp_spare32[4];
306 	uint16_t	in6p_hops;		/* (n) */
307 	uint8_t		inp_ip_tos;		/* (n) */
308 	int8_t		pad8;
309 	uint8_t		inp_vflag;		/* (s,p) */
310 	uint8_t		inp_ip_ttl;		/* (n) */
311 	uint8_t		inp_ip_p;		/* (n) */
312 	uint8_t		inp_ip_minttl;		/* (n) */
313 	int8_t		inp_spare8[4];
314 } __aligned(8);
315 
316 struct xinpgen {
317 	size_t		xig_len;	/* length of this structure */
318 	u_int		xig_count;	/* number of PCBs at this time */
319 	inp_gen_t	xig_gen;	/* generation count at this time */
320 	so_gen_t	xig_sogen;	/* socket generation count this time */
321 } __aligned(8);
322 #ifdef	_KERNEL
323 void	in_pcbtoxinpcb(const struct inpcb *, struct xinpcb *);
324 #endif
325 #endif /* _SYS_SOCKETVAR_H_ */
326 
327 struct inpcbport {
328 	LIST_ENTRY(inpcbport) phd_hash;
329 	struct inpcbhead phd_pcblist;
330 	u_short phd_port;
331 };
332 
333 /*-
334  * Global data structure for each high-level protocol (UDP, TCP, ...) in both
335  * IPv4 and IPv6.  Holds inpcb lists and information for managing them.
336  *
337  * Each pcbinfo is protected by three locks: ipi_lock, ipi_hash_lock and
338  * ipi_list_lock:
339  *  - ipi_lock covering the global pcb list stability during loop iteration,
340  *  - ipi_hash_lock covering the hashed lookup tables,
341  *  - ipi_list_lock covering mutable global fields (such as the global
342  *    pcb list)
343  *
344  * The lock order is:
345  *
346  *    ipi_lock (before)
347  *        inpcb locks (before)
348  *            ipi_list locks (before)
349  *                {ipi_hash_lock, pcbgroup locks}
350  *
351  * Locking key:
352  *
353  * (c) Constant or nearly constant after initialisation
354  * (g) Locked by ipi_lock
355  * (l) Locked by ipi_list_lock
356  * (h) Read using either ipi_hash_lock or inpcb lock; write requires both
357  * (p) Protected by one or more pcbgroup locks
358  * (x) Synchronisation properties poorly defined
359  */
360 struct inpcbinfo {
361 	/*
362 	 * Global lock protecting full inpcb list traversal
363 	 */
364 	struct rwlock		 ipi_lock;
365 
366 	/*
367 	 * Global list of inpcbs on the protocol.
368 	 */
369 	struct inpcbhead	*ipi_listhead;		/* (g/l) */
370 	u_int			 ipi_count;		/* (l) */
371 
372 	/*
373 	 * Generation count -- incremented each time a connection is allocated
374 	 * or freed.
375 	 */
376 	u_quad_t		 ipi_gencnt;		/* (l) */
377 
378 	/*
379 	 * Fields associated with port lookup and allocation.
380 	 */
381 	u_short			 ipi_lastport;		/* (x) */
382 	u_short			 ipi_lastlow;		/* (x) */
383 	u_short			 ipi_lasthi;		/* (x) */
384 
385 	/*
386 	 * UMA zone from which inpcbs are allocated for this protocol.
387 	 */
388 	struct	uma_zone	*ipi_zone;		/* (c) */
389 
390 	/*
391 	 * Connection groups associated with this protocol.  These fields are
392 	 * constant, but pcbgroup structures themselves are protected by
393 	 * per-pcbgroup locks.
394 	 */
395 	struct inpcbgroup	*ipi_pcbgroups;		/* (c) */
396 	u_int			 ipi_npcbgroups;	/* (c) */
397 	u_int			 ipi_hashfields;	/* (c) */
398 
399 	/*
400 	 * Global lock protecting non-pcbgroup hash lookup tables.
401 	 */
402 	struct rwlock		 ipi_hash_lock;
403 
404 	/*
405 	 * Global hash of inpcbs, hashed by local and foreign addresses and
406 	 * port numbers.
407 	 */
408 	struct inpcbhead	*ipi_hashbase;		/* (h) */
409 	u_long			 ipi_hashmask;		/* (h) */
410 
411 	/*
412 	 * Global hash of inpcbs, hashed by only local port number.
413 	 */
414 	struct inpcbporthead	*ipi_porthashbase;	/* (h) */
415 	u_long			 ipi_porthashmask;	/* (h) */
416 
417 	/*
418 	 * List of wildcard inpcbs for use with pcbgroups.  In the past, was
419 	 * per-pcbgroup but is now global.  All pcbgroup locks must be held
420 	 * to modify the list, so any is sufficient to read it.
421 	 */
422 	struct inpcbhead	*ipi_wildbase;		/* (p) */
423 	u_long			 ipi_wildmask;		/* (p) */
424 
425 	/*
426 	 * Pointer to network stack instance
427 	 */
428 	struct vnet		*ipi_vnet;		/* (c) */
429 
430 	/*
431 	 * general use 2
432 	 */
433 	void 			*ipi_pspare[2];
434 
435 	/*
436 	 * Global lock protecting global inpcb list, inpcb count, etc.
437 	 */
438 	struct rwlock		 ipi_list_lock;
439 };
440 
441 #ifdef _KERNEL
442 /*
443  * Connection groups hold sets of connections that have similar CPU/thread
444  * affinity.  Each connection belongs to exactly one connection group.
445  */
446 struct inpcbgroup {
447 	/*
448 	 * Per-connection group hash of inpcbs, hashed by local and foreign
449 	 * addresses and port numbers.
450 	 */
451 	struct inpcbhead	*ipg_hashbase;		/* (c) */
452 	u_long			 ipg_hashmask;		/* (c) */
453 
454 	/*
455 	 * Notional affinity of this pcbgroup.
456 	 */
457 	u_int			 ipg_cpu;		/* (p) */
458 
459 	/*
460 	 * Per-connection group lock, not to be confused with ipi_lock.
461 	 * Protects the hash table hung off the group, but also the global
462 	 * wildcard list in inpcbinfo.
463 	 */
464 	struct mtx		 ipg_lock;
465 } __aligned(CACHE_LINE_SIZE);
466 
467 #define INP_LOCK_INIT(inp, d, t) \
468 	rw_init_flags(&(inp)->inp_lock, (t), RW_RECURSE |  RW_DUPOK)
469 #define INP_LOCK_DESTROY(inp)	rw_destroy(&(inp)->inp_lock)
470 #define INP_RLOCK(inp)		rw_rlock(&(inp)->inp_lock)
471 #define INP_WLOCK(inp)		rw_wlock(&(inp)->inp_lock)
472 #define INP_TRY_RLOCK(inp)	rw_try_rlock(&(inp)->inp_lock)
473 #define INP_TRY_WLOCK(inp)	rw_try_wlock(&(inp)->inp_lock)
474 #define INP_RUNLOCK(inp)	rw_runlock(&(inp)->inp_lock)
475 #define INP_WUNLOCK(inp)	rw_wunlock(&(inp)->inp_lock)
476 #define	INP_TRY_UPGRADE(inp)	rw_try_upgrade(&(inp)->inp_lock)
477 #define	INP_DOWNGRADE(inp)	rw_downgrade(&(inp)->inp_lock)
478 #define	INP_WLOCKED(inp)	rw_wowned(&(inp)->inp_lock)
479 #define	INP_LOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_LOCKED)
480 #define	INP_RLOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_RLOCKED)
481 #define	INP_WLOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_WLOCKED)
482 #define	INP_UNLOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_UNLOCKED)
483 
484 /*
485  * These locking functions are for inpcb consumers outside of sys/netinet,
486  * more specifically, they were added for the benefit of TOE drivers. The
487  * macros are reserved for use by the stack.
488  */
489 void inp_wlock(struct inpcb *);
490 void inp_wunlock(struct inpcb *);
491 void inp_rlock(struct inpcb *);
492 void inp_runlock(struct inpcb *);
493 
494 #ifdef INVARIANT_SUPPORT
495 void inp_lock_assert(struct inpcb *);
496 void inp_unlock_assert(struct inpcb *);
497 #else
498 #define	inp_lock_assert(inp)	do {} while (0)
499 #define	inp_unlock_assert(inp)	do {} while (0)
500 #endif
501 
502 void	inp_apply_all(void (*func)(struct inpcb *, void *), void *arg);
503 int 	inp_ip_tos_get(const struct inpcb *inp);
504 void 	inp_ip_tos_set(struct inpcb *inp, int val);
505 struct socket *
506 	inp_inpcbtosocket(struct inpcb *inp);
507 struct tcpcb *
508 	inp_inpcbtotcpcb(struct inpcb *inp);
509 void 	inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
510 		uint32_t *faddr, uint16_t *fp);
511 short	inp_so_options(const struct inpcb *inp);
512 
513 #endif /* _KERNEL */
514 
515 #define INP_INFO_LOCK_INIT(ipi, d) \
516 	rw_init_flags(&(ipi)->ipi_lock, (d), RW_RECURSE)
517 #define INP_INFO_LOCK_DESTROY(ipi)  rw_destroy(&(ipi)->ipi_lock)
518 #define INP_INFO_RLOCK(ipi)	rw_rlock(&(ipi)->ipi_lock)
519 #define INP_INFO_WLOCK(ipi)	rw_wlock(&(ipi)->ipi_lock)
520 #define INP_INFO_TRY_RLOCK(ipi)	rw_try_rlock(&(ipi)->ipi_lock)
521 #define INP_INFO_TRY_WLOCK(ipi)	rw_try_wlock(&(ipi)->ipi_lock)
522 #define INP_INFO_TRY_UPGRADE(ipi)	rw_try_upgrade(&(ipi)->ipi_lock)
523 #define INP_INFO_WLOCKED(ipi)	rw_wowned(&(ipi)->ipi_lock)
524 #define INP_INFO_RUNLOCK(ipi)	rw_runlock(&(ipi)->ipi_lock)
525 #define INP_INFO_WUNLOCK(ipi)	rw_wunlock(&(ipi)->ipi_lock)
526 #define	INP_INFO_LOCK_ASSERT(ipi)	rw_assert(&(ipi)->ipi_lock, RA_LOCKED)
527 #define INP_INFO_RLOCK_ASSERT(ipi)	rw_assert(&(ipi)->ipi_lock, RA_RLOCKED)
528 #define INP_INFO_WLOCK_ASSERT(ipi)	rw_assert(&(ipi)->ipi_lock, RA_WLOCKED)
529 #define INP_INFO_UNLOCK_ASSERT(ipi)	rw_assert(&(ipi)->ipi_lock, RA_UNLOCKED)
530 
531 #define INP_LIST_LOCK_INIT(ipi, d) \
532         rw_init_flags(&(ipi)->ipi_list_lock, (d), 0)
533 #define INP_LIST_LOCK_DESTROY(ipi)  rw_destroy(&(ipi)->ipi_list_lock)
534 #define INP_LIST_RLOCK(ipi)     rw_rlock(&(ipi)->ipi_list_lock)
535 #define INP_LIST_WLOCK(ipi)     rw_wlock(&(ipi)->ipi_list_lock)
536 #define INP_LIST_TRY_RLOCK(ipi) rw_try_rlock(&(ipi)->ipi_list_lock)
537 #define INP_LIST_TRY_WLOCK(ipi) rw_try_wlock(&(ipi)->ipi_list_lock)
538 #define INP_LIST_TRY_UPGRADE(ipi)       rw_try_upgrade(&(ipi)->ipi_list_lock)
539 #define INP_LIST_RUNLOCK(ipi)   rw_runlock(&(ipi)->ipi_list_lock)
540 #define INP_LIST_WUNLOCK(ipi)   rw_wunlock(&(ipi)->ipi_list_lock)
541 #define INP_LIST_LOCK_ASSERT(ipi) \
542 	rw_assert(&(ipi)->ipi_list_lock, RA_LOCKED)
543 #define INP_LIST_RLOCK_ASSERT(ipi) \
544 	rw_assert(&(ipi)->ipi_list_lock, RA_RLOCKED)
545 #define INP_LIST_WLOCK_ASSERT(ipi) \
546 	rw_assert(&(ipi)->ipi_list_lock, RA_WLOCKED)
547 #define INP_LIST_UNLOCK_ASSERT(ipi) \
548 	rw_assert(&(ipi)->ipi_list_lock, RA_UNLOCKED)
549 
550 #define	INP_HASH_LOCK_INIT(ipi, d) \
551 	rw_init_flags(&(ipi)->ipi_hash_lock, (d), 0)
552 #define	INP_HASH_LOCK_DESTROY(ipi)	rw_destroy(&(ipi)->ipi_hash_lock)
553 #define	INP_HASH_RLOCK(ipi)		rw_rlock(&(ipi)->ipi_hash_lock)
554 #define	INP_HASH_WLOCK(ipi)		rw_wlock(&(ipi)->ipi_hash_lock)
555 #define	INP_HASH_RUNLOCK(ipi)		rw_runlock(&(ipi)->ipi_hash_lock)
556 #define	INP_HASH_WUNLOCK(ipi)		rw_wunlock(&(ipi)->ipi_hash_lock)
557 #define	INP_HASH_LOCK_ASSERT(ipi)	rw_assert(&(ipi)->ipi_hash_lock, \
558 					    RA_LOCKED)
559 #define	INP_HASH_WLOCK_ASSERT(ipi)	rw_assert(&(ipi)->ipi_hash_lock, \
560 					    RA_WLOCKED)
561 
562 #define	INP_GROUP_LOCK_INIT(ipg, d)	mtx_init(&(ipg)->ipg_lock, (d), NULL, \
563 					    MTX_DEF | MTX_DUPOK)
564 #define	INP_GROUP_LOCK_DESTROY(ipg)	mtx_destroy(&(ipg)->ipg_lock)
565 
566 #define	INP_GROUP_LOCK(ipg)		mtx_lock(&(ipg)->ipg_lock)
567 #define	INP_GROUP_LOCK_ASSERT(ipg)	mtx_assert(&(ipg)->ipg_lock, MA_OWNED)
568 #define	INP_GROUP_UNLOCK(ipg)		mtx_unlock(&(ipg)->ipg_lock)
569 
570 #define INP_PCBHASH(faddr, lport, fport, mask) \
571 	(((faddr) ^ ((faddr) >> 16) ^ ntohs((lport) ^ (fport))) & (mask))
572 #define INP_PCBPORTHASH(lport, mask) \
573 	(ntohs((lport)) & (mask))
574 #define	INP6_PCBHASHKEY(faddr)	((faddr)->s6_addr32[3])
575 
576 /*
577  * Flags for inp_vflags -- historically version flags only
578  */
579 #define	INP_IPV4	0x1
580 #define	INP_IPV6	0x2
581 #define	INP_IPV6PROTO	0x4		/* opened under IPv6 protocol */
582 
583 /*
584  * Flags for inp_flags.
585  */
586 #define	INP_RECVOPTS		0x00000001 /* receive incoming IP options */
587 #define	INP_RECVRETOPTS		0x00000002 /* receive IP options for reply */
588 #define	INP_RECVDSTADDR		0x00000004 /* receive IP dst address */
589 #define	INP_HDRINCL		0x00000008 /* user supplies entire IP header */
590 #define	INP_HIGHPORT		0x00000010 /* user wants "high" port binding */
591 #define	INP_LOWPORT		0x00000020 /* user wants "low" port binding */
592 #define	INP_ANONPORT		0x00000040 /* port chosen for user */
593 #define	INP_RECVIF		0x00000080 /* receive incoming interface */
594 #define	INP_MTUDISC		0x00000100 /* user can do MTU discovery */
595 				   	   /* 0x000200 unused: was INP_FAITH */
596 #define	INP_RECVTTL		0x00000400 /* receive incoming IP TTL */
597 #define	INP_DONTFRAG		0x00000800 /* don't fragment packet */
598 #define	INP_BINDANY		0x00001000 /* allow bind to any address */
599 #define	INP_INHASHLIST		0x00002000 /* in_pcbinshash() has been called */
600 #define	INP_RECVTOS		0x00004000 /* receive incoming IP TOS */
601 #define	IN6P_IPV6_V6ONLY	0x00008000 /* restrict AF_INET6 socket for v6 */
602 #define	IN6P_PKTINFO		0x00010000 /* receive IP6 dst and I/F */
603 #define	IN6P_HOPLIMIT		0x00020000 /* receive hoplimit */
604 #define	IN6P_HOPOPTS		0x00040000 /* receive hop-by-hop options */
605 #define	IN6P_DSTOPTS		0x00080000 /* receive dst options after rthdr */
606 #define	IN6P_RTHDR		0x00100000 /* receive routing header */
607 #define	IN6P_RTHDRDSTOPTS	0x00200000 /* receive dstoptions before rthdr */
608 #define	IN6P_TCLASS		0x00400000 /* receive traffic class value */
609 #define	IN6P_AUTOFLOWLABEL	0x00800000 /* attach flowlabel automatically */
610 #define	INP_TIMEWAIT		0x01000000 /* in TIMEWAIT, ppcb is tcptw */
611 #define	INP_ONESBCAST		0x02000000 /* send all-ones broadcast */
612 #define	INP_DROPPED		0x04000000 /* protocol drop flag */
613 #define	INP_SOCKREF		0x08000000 /* strong socket reference */
614 #define	INP_RESERVED_0          0x10000000 /* reserved field */
615 #define	INP_RESERVED_1          0x20000000 /* reserved field */
616 #define	IN6P_RFC2292		0x40000000 /* used RFC2292 API on the socket */
617 #define	IN6P_MTU		0x80000000 /* receive path MTU */
618 
619 #define	INP_CONTROLOPTS		(INP_RECVOPTS|INP_RECVRETOPTS|INP_RECVDSTADDR|\
620 				 INP_RECVIF|INP_RECVTTL|INP_RECVTOS|\
621 				 IN6P_PKTINFO|IN6P_HOPLIMIT|IN6P_HOPOPTS|\
622 				 IN6P_DSTOPTS|IN6P_RTHDR|IN6P_RTHDRDSTOPTS|\
623 				 IN6P_TCLASS|IN6P_AUTOFLOWLABEL|IN6P_RFC2292|\
624 				 IN6P_MTU)
625 
626 /*
627  * Flags for inp_flags2.
628  */
629 #define	INP_LLE_VALID		0x00000001 /* cached lle is valid */
630 #define	INP_RT_VALID		0x00000002 /* cached rtentry is valid */
631 #define	INP_PCBGROUPWILD	0x00000004 /* in pcbgroup wildcard list */
632 #define	INP_REUSEPORT		0x00000008 /* SO_REUSEPORT option is set */
633 #define	INP_FREED		0x00000010 /* inp itself is not valid */
634 #define	INP_REUSEADDR		0x00000020 /* SO_REUSEADDR option is set */
635 #define	INP_BINDMULTI		0x00000040 /* IP_BINDMULTI option is set */
636 #define	INP_RSS_BUCKET_SET	0x00000080 /* IP_RSS_LISTEN_BUCKET is set */
637 #define	INP_RECVFLOWID		0x00000100 /* populate recv datagram with flow info */
638 #define	INP_RECVRSSBUCKETID	0x00000200 /* populate recv datagram with bucket id */
639 #define	INP_RATE_LIMIT_CHANGED	0x00000400 /* rate limit needs attention */
640 #define	INP_ORIGDSTADDR		0x00000800 /* receive IP dst address/port */
641 
642 /*
643  * Flags passed to in_pcblookup*() functions.
644  */
645 #define	INPLOOKUP_WILDCARD	0x00000001	/* Allow wildcard sockets. */
646 #define	INPLOOKUP_RLOCKPCB	0x00000002	/* Return inpcb read-locked. */
647 #define	INPLOOKUP_WLOCKPCB	0x00000004	/* Return inpcb write-locked. */
648 
649 #define	INPLOOKUP_MASK	(INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB | \
650 			    INPLOOKUP_WLOCKPCB)
651 
652 #define	sotoinpcb(so)	((struct inpcb *)(so)->so_pcb)
653 #define	sotoin6pcb(so)	sotoinpcb(so) /* for KAME src sync over BSD*'s */
654 
655 #define	INP_SOCKAF(so) so->so_proto->pr_domain->dom_family
656 
657 #define	INP_CHECK_SOCKAF(so, af)	(INP_SOCKAF(so) == af)
658 
659 /*
660  * Constants for pcbinfo.ipi_hashfields.
661  */
662 #define	IPI_HASHFIELDS_NONE	0
663 #define	IPI_HASHFIELDS_2TUPLE	1
664 #define	IPI_HASHFIELDS_4TUPLE	2
665 
666 #ifdef _KERNEL
667 VNET_DECLARE(int, ipport_reservedhigh);
668 VNET_DECLARE(int, ipport_reservedlow);
669 VNET_DECLARE(int, ipport_lowfirstauto);
670 VNET_DECLARE(int, ipport_lowlastauto);
671 VNET_DECLARE(int, ipport_firstauto);
672 VNET_DECLARE(int, ipport_lastauto);
673 VNET_DECLARE(int, ipport_hifirstauto);
674 VNET_DECLARE(int, ipport_hilastauto);
675 VNET_DECLARE(int, ipport_randomized);
676 VNET_DECLARE(int, ipport_randomcps);
677 VNET_DECLARE(int, ipport_randomtime);
678 VNET_DECLARE(int, ipport_stoprandom);
679 VNET_DECLARE(int, ipport_tcpallocs);
680 
681 #define	V_ipport_reservedhigh	VNET(ipport_reservedhigh)
682 #define	V_ipport_reservedlow	VNET(ipport_reservedlow)
683 #define	V_ipport_lowfirstauto	VNET(ipport_lowfirstauto)
684 #define	V_ipport_lowlastauto	VNET(ipport_lowlastauto)
685 #define	V_ipport_firstauto	VNET(ipport_firstauto)
686 #define	V_ipport_lastauto	VNET(ipport_lastauto)
687 #define	V_ipport_hifirstauto	VNET(ipport_hifirstauto)
688 #define	V_ipport_hilastauto	VNET(ipport_hilastauto)
689 #define	V_ipport_randomized	VNET(ipport_randomized)
690 #define	V_ipport_randomcps	VNET(ipport_randomcps)
691 #define	V_ipport_randomtime	VNET(ipport_randomtime)
692 #define	V_ipport_stoprandom	VNET(ipport_stoprandom)
693 #define	V_ipport_tcpallocs	VNET(ipport_tcpallocs)
694 
695 void	in_pcbinfo_destroy(struct inpcbinfo *);
696 void	in_pcbinfo_init(struct inpcbinfo *, const char *, struct inpcbhead *,
697 	    int, int, char *, uma_init, u_int);
698 
699 int	in_pcbbind_check_bindmulti(const struct inpcb *ni,
700 	    const struct inpcb *oi);
701 
702 struct inpcbgroup *
703 	in_pcbgroup_byhash(struct inpcbinfo *, u_int, uint32_t);
704 struct inpcbgroup *
705 	in_pcbgroup_byinpcb(struct inpcb *);
706 struct inpcbgroup *
707 	in_pcbgroup_bytuple(struct inpcbinfo *, struct in_addr, u_short,
708 	    struct in_addr, u_short);
709 void	in_pcbgroup_destroy(struct inpcbinfo *);
710 int	in_pcbgroup_enabled(struct inpcbinfo *);
711 void	in_pcbgroup_init(struct inpcbinfo *, u_int, int);
712 void	in_pcbgroup_remove(struct inpcb *);
713 void	in_pcbgroup_update(struct inpcb *);
714 void	in_pcbgroup_update_mbuf(struct inpcb *, struct mbuf *);
715 
716 void	in_pcbpurgeif0(struct inpcbinfo *, struct ifnet *);
717 int	in_pcballoc(struct socket *, struct inpcbinfo *);
718 int	in_pcbbind(struct inpcb *, struct sockaddr *, struct ucred *);
719 int	in_pcb_lport(struct inpcb *, struct in_addr *, u_short *,
720 	    struct ucred *, int);
721 int	in_pcbbind_setup(struct inpcb *, struct sockaddr *, in_addr_t *,
722 	    u_short *, struct ucred *);
723 int	in_pcbconnect(struct inpcb *, struct sockaddr *, struct ucred *);
724 int	in_pcbconnect_mbuf(struct inpcb *, struct sockaddr *, struct ucred *,
725 	    struct mbuf *);
726 int	in_pcbconnect_setup(struct inpcb *, struct sockaddr *, in_addr_t *,
727 	    u_short *, in_addr_t *, u_short *, struct inpcb **,
728 	    struct ucred *);
729 void	in_pcbdetach(struct inpcb *);
730 void	in_pcbdisconnect(struct inpcb *);
731 void	in_pcbdrop(struct inpcb *);
732 void	in_pcbfree(struct inpcb *);
733 int	in_pcbinshash(struct inpcb *);
734 int	in_pcbinshash_nopcbgroup(struct inpcb *);
735 int	in_pcbladdr(struct inpcb *, struct in_addr *, struct in_addr *,
736 	    struct ucred *);
737 struct inpcb *
738 	in_pcblookup_local(struct inpcbinfo *,
739 	    struct in_addr, u_short, int, struct ucred *);
740 struct inpcb *
741 	in_pcblookup(struct inpcbinfo *, struct in_addr, u_int,
742 	    struct in_addr, u_int, int, struct ifnet *);
743 struct inpcb *
744 	in_pcblookup_mbuf(struct inpcbinfo *, struct in_addr, u_int,
745 	    struct in_addr, u_int, int, struct ifnet *, struct mbuf *);
746 void	in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr,
747 	    int, struct inpcb *(*)(struct inpcb *, int));
748 void	in_pcbref(struct inpcb *);
749 void	in_pcbrehash(struct inpcb *);
750 void	in_pcbrehash_mbuf(struct inpcb *, struct mbuf *);
751 int	in_pcbrele(struct inpcb *);
752 int	in_pcbrele_rlocked(struct inpcb *);
753 int	in_pcbrele_wlocked(struct inpcb *);
754 void	in_losing(struct inpcb *);
755 void	in_pcbsetsolabel(struct socket *so);
756 int	in_getpeeraddr(struct socket *so, struct sockaddr **nam);
757 int	in_getsockaddr(struct socket *so, struct sockaddr **nam);
758 struct sockaddr *
759 	in_sockaddr(in_port_t port, struct in_addr *addr);
760 void	in_pcbsosetlabel(struct socket *so);
761 #ifdef RATELIMIT
762 int	in_pcbattach_txrtlmt(struct inpcb *, struct ifnet *, uint32_t, uint32_t, uint32_t);
763 void	in_pcbdetach_txrtlmt(struct inpcb *);
764 int	in_pcbmodify_txrtlmt(struct inpcb *, uint32_t);
765 int	in_pcbquery_txrtlmt(struct inpcb *, uint32_t *);
766 int	in_pcbquery_txrlevel(struct inpcb *, uint32_t *);
767 void	in_pcboutput_txrtlmt(struct inpcb *, struct ifnet *, struct mbuf *);
768 void	in_pcboutput_eagain(struct inpcb *);
769 #endif
770 #endif /* _KERNEL */
771 
772 #endif /* !_NETINET_IN_PCB_H_ */
773