xref: /freebsd/sys/netinet/in_pcb.h (revision d1a0d267b78b542fbd7e6553af2493760f49bfa8)
1 /*-
2  * Copyright (c) 1982, 1986, 1990, 1993
3  *	The Regents of the University of California.
4  * Copyright (c) 2010-2011 Juniper Networks, Inc.
5  * All rights reserved.
6  *
7  * Portions of this software were developed by Robert N. M. Watson under
8  * contract to Juniper Networks, Inc.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)in_pcb.h	8.1 (Berkeley) 6/10/93
35  * $FreeBSD$
36  */
37 
38 #ifndef _NETINET_IN_PCB_H_
39 #define _NETINET_IN_PCB_H_
40 
41 #include <sys/queue.h>
42 #include <sys/_lock.h>
43 #include <sys/_mutex.h>
44 #include <sys/_rwlock.h>
45 
46 #ifdef _KERNEL
47 #include <sys/lock.h>
48 #include <sys/rwlock.h>
49 #include <net/vnet.h>
50 #include <vm/uma.h>
51 #endif
52 
53 #define	in6pcb		inpcb	/* for KAME src sync over BSD*'s */
54 #define	in6p_sp		inp_sp	/* for KAME src sync over BSD*'s */
55 struct inpcbpolicy;
56 
57 /*
58  * struct inpcb is the common protocol control block structure used in most
59  * IP transport protocols.
60  *
61  * Pointers to local and foreign host table entries, local and foreign socket
62  * numbers, and pointers up (to a socket structure) and down (to a
63  * protocol-specific control block) are stored here.
64  */
65 LIST_HEAD(inpcbhead, inpcb);
66 LIST_HEAD(inpcbporthead, inpcbport);
67 typedef	u_quad_t	inp_gen_t;
68 
69 /*
70  * PCB with AF_INET6 null bind'ed laddr can receive AF_INET input packet.
71  * So, AF_INET6 null laddr is also used as AF_INET null laddr, by utilizing
72  * the following structure.
73  */
74 struct in_addr_4in6 {
75 	u_int32_t	ia46_pad32[3];
76 	struct	in_addr	ia46_addr4;
77 };
78 
79 /*
80  * NOTE: ipv6 addrs should be 64-bit aligned, per RFC 2553.  in_conninfo has
81  * some extra padding to accomplish this.
82  */
83 struct in_endpoints {
84 	u_int16_t	ie_fport;		/* foreign port */
85 	u_int16_t	ie_lport;		/* local port */
86 	/* protocol dependent part, local and foreign addr */
87 	union {
88 		/* foreign host table entry */
89 		struct	in_addr_4in6 ie46_foreign;
90 		struct	in6_addr ie6_foreign;
91 	} ie_dependfaddr;
92 	union {
93 		/* local host table entry */
94 		struct	in_addr_4in6 ie46_local;
95 		struct	in6_addr ie6_local;
96 	} ie_dependladdr;
97 	u_int32_t	ie6_zoneid;		/* scope zone id */
98 };
99 #define	ie_faddr	ie_dependfaddr.ie46_foreign.ia46_addr4
100 #define	ie_laddr	ie_dependladdr.ie46_local.ia46_addr4
101 #define	ie6_faddr	ie_dependfaddr.ie6_foreign
102 #define	ie6_laddr	ie_dependladdr.ie6_local
103 
104 /*
105  * XXX The defines for inc_* are hacks and should be changed to direct
106  * references.
107  */
108 struct in_conninfo {
109 	u_int8_t	inc_flags;
110 	u_int8_t	inc_len;
111 	u_int16_t	inc_fibnum;	/* XXX was pad, 16 bits is plenty */
112 	/* protocol dependent part */
113 	struct	in_endpoints inc_ie;
114 };
115 
116 /*
117  * Flags for inc_flags.
118  */
119 #define	INC_ISIPV6	0x01
120 
121 #define inc_isipv6	inc_flags	/* temp compatability */
122 #define	inc_fport	inc_ie.ie_fport
123 #define	inc_lport	inc_ie.ie_lport
124 #define	inc_faddr	inc_ie.ie_faddr
125 #define	inc_laddr	inc_ie.ie_laddr
126 #define	inc6_faddr	inc_ie.ie6_faddr
127 #define	inc6_laddr	inc_ie.ie6_laddr
128 #define	inc6_zoneid	inc_ie.ie6_zoneid
129 
130 struct	icmp6_filter;
131 
132 /*-
133  * struct inpcb captures the network layer state for TCP, UDP, and raw IPv4 and
134  * IPv6 sockets.  In the case of TCP and UDP, further per-connection state is
135  * hung off of inp_ppcb most of the time.  Almost all fields of struct inpcb
136  * are static after creation or protected by a per-inpcb rwlock, inp_lock.  A
137  * few fields are protected by multiple locks as indicated in the locking notes
138  * below.  For these fields, all of the listed locks must be write-locked for
139  * any modifications.  However, these fields can be safely read while any one of
140  * the listed locks are read-locked.  This model can permit greater concurrency
141  * for read operations.  For example, connections can be looked up while only
142  * holding a read lock on the global pcblist lock.  This is important for
143  * performance when attempting to find the connection for a packet given its IP
144  * and port tuple.
145  *
146  * One noteworthy exception is that the global pcbinfo lock follows a different
147  * set of rules in relation to the inp_list field.  Rather than being
148  * write-locked for modifications and read-locked for list iterations, it must
149  * be read-locked during modifications and write-locked during list iterations.
150  * This ensures that the relatively rare global list iterations safely walk a
151  * stable snapshot of connections while allowing more common list modifications
152  * to safely grab the pcblist lock just while adding or removing a connection
153  * from the global list.
154  *
155  * Key:
156  * (c) - Constant after initialization
157  * (g) - Protected by the pcbgroup lock
158  * (i) - Protected by the inpcb lock
159  * (p) - Protected by the pcbinfo lock for the inpcb
160  * (l) - Protected by the pcblist lock for the inpcb
161  * (h) - Protected by the pcbhash lock for the inpcb
162  * (s) - Protected by another subsystem's locks
163  * (x) - Undefined locking
164  *
165  * A few other notes:
166  *
167  * When a read lock is held, stability of the field is guaranteed; to write
168  * to a field, a write lock must generally be held.
169  *
170  * netinet/netinet6-layer code should not assume that the inp_socket pointer
171  * is safe to dereference without inp_lock being held, even for protocols
172  * other than TCP (where the inpcb persists during TIMEWAIT even after the
173  * socket has been freed), or there may be close(2)-related races.
174  *
175  * The inp_vflag field is overloaded, and would otherwise ideally be (c).
176  *
177  * TODO:  Currently only the TCP stack is leveraging the global pcbinfo lock
178  * read-lock usage during modification, this model can be applied to other
179  * protocols (especially SCTP).
180  */
181 struct inpcb {
182 	LIST_ENTRY(inpcb) inp_hash;	/* (h/i) hash list */
183 	LIST_ENTRY(inpcb) inp_pcbgrouphash;	/* (g/i) hash list */
184 	LIST_ENTRY(inpcb) inp_list;	/* (p/l) list for all PCBs for proto */
185 	                                /* (p[w]) for list iteration */
186 	                                /* (p[r]/l) for addition/removal */
187 	void	*inp_ppcb;		/* (i) pointer to per-protocol pcb */
188 	struct	inpcbinfo *inp_pcbinfo;	/* (c) PCB list info */
189 	struct	inpcbgroup *inp_pcbgroup; /* (g/i) PCB group list */
190 	LIST_ENTRY(inpcb) inp_pcbgroup_wild; /* (g/i/h) group wildcard entry */
191 	struct	socket *inp_socket;	/* (i) back pointer to socket */
192 	struct	ucred	*inp_cred;	/* (c) cache of socket cred */
193 	u_int32_t inp_flow;		/* (i) IPv6 flow information */
194 	int	inp_flags;		/* (i) generic IP/datagram flags */
195 	int	inp_flags2;		/* (i) generic IP/datagram flags #2*/
196 	u_char	inp_vflag;		/* (i) IP version flag (v4/v6) */
197 	u_char	inp_ip_ttl;		/* (i) time to live proto */
198 	u_char	inp_ip_p;		/* (c) protocol proto */
199 	u_char	inp_ip_minttl;		/* (i) minimum TTL or drop */
200 	uint32_t inp_flowid;		/* (x) flow id / queue id */
201 	u_int	inp_refcount;		/* (i) refcount */
202 	void	*inp_pspare[5];		/* (x) route caching / general use */
203 	uint32_t inp_flowtype;		/* (x) M_HASHTYPE value */
204 	uint32_t inp_rss_listen_bucket;	/* (x) overridden RSS listen bucket */
205 	u_int	inp_ispare[4];		/* (x) route caching / user cookie /
206 					 *     general use */
207 
208 	/* Local and foreign ports, local and foreign addr. */
209 	struct	in_conninfo inp_inc;	/* (i) list for PCB's local port */
210 
211 	/* MAC and IPSEC policy information. */
212 	struct	label *inp_label;	/* (i) MAC label */
213 	struct	inpcbpolicy *inp_sp;    /* (s) for IPSEC */
214 
215 	/* Protocol-dependent part; options. */
216 	struct {
217 		u_char	inp4_ip_tos;		/* (i) type of service proto */
218 		struct	mbuf *inp4_options;	/* (i) IP options */
219 		struct	ip_moptions *inp4_moptions; /* (i) IP mcast options */
220 	} inp_depend4;
221 	struct {
222 		/* (i) IP options */
223 		struct	mbuf *inp6_options;
224 		/* (i) IP6 options for outgoing packets */
225 		struct	ip6_pktopts *inp6_outputopts;
226 		/* (i) IP multicast options */
227 		struct	ip6_moptions *inp6_moptions;
228 		/* (i) ICMPv6 code type filter */
229 		struct	icmp6_filter *inp6_icmp6filt;
230 		/* (i) IPV6_CHECKSUM setsockopt */
231 		int	inp6_cksum;
232 		short	inp6_hops;
233 	} inp_depend6;
234 	LIST_ENTRY(inpcb) inp_portlist;	/* (i/h) */
235 	struct	inpcbport *inp_phd;	/* (i/h) head of this list */
236 #define inp_zero_size offsetof(struct inpcb, inp_gencnt)
237 	inp_gen_t	inp_gencnt;	/* (c) generation count */
238 	struct llentry	*inp_lle;	/* cached L2 information */
239 	struct rtentry	*inp_rt;	/* cached L3 information */
240 	struct rwlock	inp_lock;
241 };
242 #define	inp_fport	inp_inc.inc_fport
243 #define	inp_lport	inp_inc.inc_lport
244 #define	inp_faddr	inp_inc.inc_faddr
245 #define	inp_laddr	inp_inc.inc_laddr
246 #define	inp_ip_tos	inp_depend4.inp4_ip_tos
247 #define	inp_options	inp_depend4.inp4_options
248 #define	inp_moptions	inp_depend4.inp4_moptions
249 
250 #define	in6p_faddr	inp_inc.inc6_faddr
251 #define	in6p_laddr	inp_inc.inc6_laddr
252 #define	in6p_zoneid	inp_inc.inc6_zoneid
253 #define	in6p_hops	inp_depend6.inp6_hops	/* default hop limit */
254 #define	in6p_flowinfo	inp_flow
255 #define	in6p_options	inp_depend6.inp6_options
256 #define	in6p_outputopts	inp_depend6.inp6_outputopts
257 #define	in6p_moptions	inp_depend6.inp6_moptions
258 #define	in6p_icmp6filt	inp_depend6.inp6_icmp6filt
259 #define	in6p_cksum	inp_depend6.inp6_cksum
260 
261 #define	inp_vnet	inp_pcbinfo->ipi_vnet
262 
263 /*
264  * The range of the generation count, as used in this implementation, is 9e19.
265  * We would have to create 300 billion connections per second for this number
266  * to roll over in a year.  This seems sufficiently unlikely that we simply
267  * don't concern ourselves with that possibility.
268  */
269 
270 /*
271  * Interface exported to userland by various protocols which use inpcbs.  Hack
272  * alert -- only define if struct xsocket is in scope.
273  */
274 #ifdef _SYS_SOCKETVAR_H_
275 struct	xinpcb {
276 	size_t	xi_len;		/* length of this structure */
277 	struct	inpcb xi_inp;
278 	struct	xsocket xi_socket;
279 	u_quad_t	xi_alignment_hack;
280 };
281 
282 struct	xinpgen {
283 	size_t	xig_len;	/* length of this structure */
284 	u_int	xig_count;	/* number of PCBs at this time */
285 	inp_gen_t xig_gen;	/* generation count at this time */
286 	so_gen_t xig_sogen;	/* socket generation count at this time */
287 };
288 #endif /* _SYS_SOCKETVAR_H_ */
289 
290 struct inpcbport {
291 	LIST_ENTRY(inpcbport) phd_hash;
292 	struct inpcbhead phd_pcblist;
293 	u_short phd_port;
294 };
295 
296 /*-
297  * Global data structure for each high-level protocol (UDP, TCP, ...) in both
298  * IPv4 and IPv6.  Holds inpcb lists and information for managing them.
299  *
300  * Each pcbinfo is protected by three locks: ipi_lock, ipi_hash_lock and
301  * ipi_list_lock:
302  *  - ipi_lock covering the global pcb list stability during loop iteration,
303  *  - ipi_hash_lock covering the hashed lookup tables,
304  *  - ipi_list_lock covering mutable global fields (such as the global
305  *    pcb list)
306  *
307  * The lock order is:
308  *
309  *    ipi_lock (before)
310  *        inpcb locks (before)
311  *            ipi_list locks (before)
312  *                {ipi_hash_lock, pcbgroup locks}
313  *
314  * Locking key:
315  *
316  * (c) Constant or nearly constant after initialisation
317  * (g) Locked by ipi_lock
318  * (l) Locked by ipi_list_lock
319  * (h) Read using either ipi_hash_lock or inpcb lock; write requires both
320  * (p) Protected by one or more pcbgroup locks
321  * (x) Synchronisation properties poorly defined
322  */
323 struct inpcbinfo {
324 	/*
325 	 * Global lock protecting full inpcb list traversal
326 	 */
327 	struct rwlock		 ipi_lock;
328 
329 	/*
330 	 * Global list of inpcbs on the protocol.
331 	 */
332 	struct inpcbhead	*ipi_listhead;		/* (g/l) */
333 	u_int			 ipi_count;		/* (l) */
334 
335 	/*
336 	 * Generation count -- incremented each time a connection is allocated
337 	 * or freed.
338 	 */
339 	u_quad_t		 ipi_gencnt;		/* (l) */
340 
341 	/*
342 	 * Fields associated with port lookup and allocation.
343 	 */
344 	u_short			 ipi_lastport;		/* (x) */
345 	u_short			 ipi_lastlow;		/* (x) */
346 	u_short			 ipi_lasthi;		/* (x) */
347 
348 	/*
349 	 * UMA zone from which inpcbs are allocated for this protocol.
350 	 */
351 	struct	uma_zone	*ipi_zone;		/* (c) */
352 
353 	/*
354 	 * Connection groups associated with this protocol.  These fields are
355 	 * constant, but pcbgroup structures themselves are protected by
356 	 * per-pcbgroup locks.
357 	 */
358 	struct inpcbgroup	*ipi_pcbgroups;		/* (c) */
359 	u_int			 ipi_npcbgroups;	/* (c) */
360 	u_int			 ipi_hashfields;	/* (c) */
361 
362 	/*
363 	 * Global lock protecting non-pcbgroup hash lookup tables.
364 	 */
365 	struct rwlock		 ipi_hash_lock;
366 
367 	/*
368 	 * Global hash of inpcbs, hashed by local and foreign addresses and
369 	 * port numbers.
370 	 */
371 	struct inpcbhead	*ipi_hashbase;		/* (h) */
372 	u_long			 ipi_hashmask;		/* (h) */
373 
374 	/*
375 	 * Global hash of inpcbs, hashed by only local port number.
376 	 */
377 	struct inpcbporthead	*ipi_porthashbase;	/* (h) */
378 	u_long			 ipi_porthashmask;	/* (h) */
379 
380 	/*
381 	 * List of wildcard inpcbs for use with pcbgroups.  In the past, was
382 	 * per-pcbgroup but is now global.  All pcbgroup locks must be held
383 	 * to modify the list, so any is sufficient to read it.
384 	 */
385 	struct inpcbhead	*ipi_wildbase;		/* (p) */
386 	u_long			 ipi_wildmask;		/* (p) */
387 
388 	/*
389 	 * Pointer to network stack instance
390 	 */
391 	struct vnet		*ipi_vnet;		/* (c) */
392 
393 	/*
394 	 * general use 2
395 	 */
396 	void 			*ipi_pspare[2];
397 
398 	/*
399 	 * Global lock protecting global inpcb list, inpcb count, etc.
400 	 */
401 	struct rwlock		 ipi_list_lock;
402 };
403 
404 #ifdef _KERNEL
405 /*
406  * Connection groups hold sets of connections that have similar CPU/thread
407  * affinity.  Each connection belongs to exactly one connection group.
408  */
409 struct inpcbgroup {
410 	/*
411 	 * Per-connection group hash of inpcbs, hashed by local and foreign
412 	 * addresses and port numbers.
413 	 */
414 	struct inpcbhead	*ipg_hashbase;		/* (c) */
415 	u_long			 ipg_hashmask;		/* (c) */
416 
417 	/*
418 	 * Notional affinity of this pcbgroup.
419 	 */
420 	u_int			 ipg_cpu;		/* (p) */
421 
422 	/*
423 	 * Per-connection group lock, not to be confused with ipi_lock.
424 	 * Protects the hash table hung off the group, but also the global
425 	 * wildcard list in inpcbinfo.
426 	 */
427 	struct mtx		 ipg_lock;
428 } __aligned(CACHE_LINE_SIZE);
429 
430 #define INP_LOCK_INIT(inp, d, t) \
431 	rw_init_flags(&(inp)->inp_lock, (t), RW_RECURSE |  RW_DUPOK)
432 #define INP_LOCK_DESTROY(inp)	rw_destroy(&(inp)->inp_lock)
433 #define INP_RLOCK(inp)		rw_rlock(&(inp)->inp_lock)
434 #define INP_WLOCK(inp)		rw_wlock(&(inp)->inp_lock)
435 #define INP_TRY_RLOCK(inp)	rw_try_rlock(&(inp)->inp_lock)
436 #define INP_TRY_WLOCK(inp)	rw_try_wlock(&(inp)->inp_lock)
437 #define INP_RUNLOCK(inp)	rw_runlock(&(inp)->inp_lock)
438 #define INP_WUNLOCK(inp)	rw_wunlock(&(inp)->inp_lock)
439 #define	INP_TRY_UPGRADE(inp)	rw_try_upgrade(&(inp)->inp_lock)
440 #define	INP_DOWNGRADE(inp)	rw_downgrade(&(inp)->inp_lock)
441 #define	INP_WLOCKED(inp)	rw_wowned(&(inp)->inp_lock)
442 #define	INP_LOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_LOCKED)
443 #define	INP_RLOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_RLOCKED)
444 #define	INP_WLOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_WLOCKED)
445 #define	INP_UNLOCK_ASSERT(inp)	rw_assert(&(inp)->inp_lock, RA_UNLOCKED)
446 
447 /*
448  * These locking functions are for inpcb consumers outside of sys/netinet,
449  * more specifically, they were added for the benefit of TOE drivers. The
450  * macros are reserved for use by the stack.
451  */
452 void inp_wlock(struct inpcb *);
453 void inp_wunlock(struct inpcb *);
454 void inp_rlock(struct inpcb *);
455 void inp_runlock(struct inpcb *);
456 
457 #ifdef INVARIANTS
458 void inp_lock_assert(struct inpcb *);
459 void inp_unlock_assert(struct inpcb *);
460 #else
461 static __inline void
462 inp_lock_assert(struct inpcb *inp __unused)
463 {
464 }
465 
466 static __inline void
467 inp_unlock_assert(struct inpcb *inp __unused)
468 {
469 }
470 
471 #endif
472 
473 void	inp_apply_all(void (*func)(struct inpcb *, void *), void *arg);
474 int 	inp_ip_tos_get(const struct inpcb *inp);
475 void 	inp_ip_tos_set(struct inpcb *inp, int val);
476 struct socket *
477 	inp_inpcbtosocket(struct inpcb *inp);
478 struct tcpcb *
479 	inp_inpcbtotcpcb(struct inpcb *inp);
480 void 	inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp,
481 		uint32_t *faddr, uint16_t *fp);
482 short	inp_so_options(const struct inpcb *inp);
483 
484 #endif /* _KERNEL */
485 
486 #define INP_INFO_LOCK_INIT(ipi, d) \
487 	rw_init_flags(&(ipi)->ipi_lock, (d), RW_RECURSE)
488 #define INP_INFO_LOCK_DESTROY(ipi)  rw_destroy(&(ipi)->ipi_lock)
489 #define INP_INFO_RLOCK(ipi)	rw_rlock(&(ipi)->ipi_lock)
490 #define INP_INFO_WLOCK(ipi)	rw_wlock(&(ipi)->ipi_lock)
491 #define INP_INFO_TRY_RLOCK(ipi)	rw_try_rlock(&(ipi)->ipi_lock)
492 #define INP_INFO_TRY_WLOCK(ipi)	rw_try_wlock(&(ipi)->ipi_lock)
493 #define INP_INFO_TRY_UPGRADE(ipi)	rw_try_upgrade(&(ipi)->ipi_lock)
494 #define INP_INFO_WLOCKED(ipi)	rw_wowned(&(ipi)->ipi_lock)
495 #define INP_INFO_RUNLOCK(ipi)	rw_runlock(&(ipi)->ipi_lock)
496 #define INP_INFO_WUNLOCK(ipi)	rw_wunlock(&(ipi)->ipi_lock)
497 #define	INP_INFO_LOCK_ASSERT(ipi)	rw_assert(&(ipi)->ipi_lock, RA_LOCKED)
498 #define INP_INFO_RLOCK_ASSERT(ipi)	rw_assert(&(ipi)->ipi_lock, RA_RLOCKED)
499 #define INP_INFO_WLOCK_ASSERT(ipi)	rw_assert(&(ipi)->ipi_lock, RA_WLOCKED)
500 #define INP_INFO_UNLOCK_ASSERT(ipi)	rw_assert(&(ipi)->ipi_lock, RA_UNLOCKED)
501 
502 #define INP_LIST_LOCK_INIT(ipi, d) \
503         rw_init_flags(&(ipi)->ipi_list_lock, (d), 0)
504 #define INP_LIST_LOCK_DESTROY(ipi)  rw_destroy(&(ipi)->ipi_list_lock)
505 #define INP_LIST_RLOCK(ipi)     rw_rlock(&(ipi)->ipi_list_lock)
506 #define INP_LIST_WLOCK(ipi)     rw_wlock(&(ipi)->ipi_list_lock)
507 #define INP_LIST_TRY_RLOCK(ipi) rw_try_rlock(&(ipi)->ipi_list_lock)
508 #define INP_LIST_TRY_WLOCK(ipi) rw_try_wlock(&(ipi)->ipi_list_lock)
509 #define INP_LIST_TRY_UPGRADE(ipi)       rw_try_upgrade(&(ipi)->ipi_list_lock)
510 #define INP_LIST_RUNLOCK(ipi)   rw_runlock(&(ipi)->ipi_list_lock)
511 #define INP_LIST_WUNLOCK(ipi)   rw_wunlock(&(ipi)->ipi_list_lock)
512 #define INP_LIST_LOCK_ASSERT(ipi) \
513 	rw_assert(&(ipi)->ipi_list_lock, RA_LOCKED)
514 #define INP_LIST_RLOCK_ASSERT(ipi) \
515 	rw_assert(&(ipi)->ipi_list_lock, RA_RLOCKED)
516 #define INP_LIST_WLOCK_ASSERT(ipi) \
517 	rw_assert(&(ipi)->ipi_list_lock, RA_WLOCKED)
518 #define INP_LIST_UNLOCK_ASSERT(ipi) \
519 	rw_assert(&(ipi)->ipi_list_lock, RA_UNLOCKED)
520 
521 #define	INP_HASH_LOCK_INIT(ipi, d) \
522 	rw_init_flags(&(ipi)->ipi_hash_lock, (d), 0)
523 #define	INP_HASH_LOCK_DESTROY(ipi)	rw_destroy(&(ipi)->ipi_hash_lock)
524 #define	INP_HASH_RLOCK(ipi)		rw_rlock(&(ipi)->ipi_hash_lock)
525 #define	INP_HASH_WLOCK(ipi)		rw_wlock(&(ipi)->ipi_hash_lock)
526 #define	INP_HASH_RUNLOCK(ipi)		rw_runlock(&(ipi)->ipi_hash_lock)
527 #define	INP_HASH_WUNLOCK(ipi)		rw_wunlock(&(ipi)->ipi_hash_lock)
528 #define	INP_HASH_LOCK_ASSERT(ipi)	rw_assert(&(ipi)->ipi_hash_lock, \
529 					    RA_LOCKED)
530 #define	INP_HASH_WLOCK_ASSERT(ipi)	rw_assert(&(ipi)->ipi_hash_lock, \
531 					    RA_WLOCKED)
532 
533 #define	INP_GROUP_LOCK_INIT(ipg, d)	mtx_init(&(ipg)->ipg_lock, (d), NULL, \
534 					    MTX_DEF | MTX_DUPOK)
535 #define	INP_GROUP_LOCK_DESTROY(ipg)	mtx_destroy(&(ipg)->ipg_lock)
536 
537 #define	INP_GROUP_LOCK(ipg)		mtx_lock(&(ipg)->ipg_lock)
538 #define	INP_GROUP_LOCK_ASSERT(ipg)	mtx_assert(&(ipg)->ipg_lock, MA_OWNED)
539 #define	INP_GROUP_UNLOCK(ipg)		mtx_unlock(&(ipg)->ipg_lock)
540 
541 #define INP_PCBHASH(faddr, lport, fport, mask) \
542 	(((faddr) ^ ((faddr) >> 16) ^ ntohs((lport) ^ (fport))) & (mask))
543 #define INP_PCBPORTHASH(lport, mask) \
544 	(ntohs((lport)) & (mask))
545 #define	INP6_PCBHASHKEY(faddr)	((faddr)->s6_addr32[3])
546 
547 /*
548  * Flags for inp_vflags -- historically version flags only
549  */
550 #define	INP_IPV4	0x1
551 #define	INP_IPV6	0x2
552 #define	INP_IPV6PROTO	0x4		/* opened under IPv6 protocol */
553 
554 /*
555  * Flags for inp_flags.
556  */
557 #define	INP_RECVOPTS		0x00000001 /* receive incoming IP options */
558 #define	INP_RECVRETOPTS		0x00000002 /* receive IP options for reply */
559 #define	INP_RECVDSTADDR		0x00000004 /* receive IP dst address */
560 #define	INP_HDRINCL		0x00000008 /* user supplies entire IP header */
561 #define	INP_HIGHPORT		0x00000010 /* user wants "high" port binding */
562 #define	INP_LOWPORT		0x00000020 /* user wants "low" port binding */
563 #define	INP_ANONPORT		0x00000040 /* port chosen for user */
564 #define	INP_RECVIF		0x00000080 /* receive incoming interface */
565 #define	INP_MTUDISC		0x00000100 /* user can do MTU discovery */
566 				   	   /* 0x000200 unused: was INP_FAITH */
567 #define	INP_RECVTTL		0x00000400 /* receive incoming IP TTL */
568 #define	INP_DONTFRAG		0x00000800 /* don't fragment packet */
569 #define	INP_BINDANY		0x00001000 /* allow bind to any address */
570 #define	INP_INHASHLIST		0x00002000 /* in_pcbinshash() has been called */
571 #define	INP_RECVTOS		0x00004000 /* receive incoming IP TOS */
572 #define	IN6P_IPV6_V6ONLY	0x00008000 /* restrict AF_INET6 socket for v6 */
573 #define	IN6P_PKTINFO		0x00010000 /* receive IP6 dst and I/F */
574 #define	IN6P_HOPLIMIT		0x00020000 /* receive hoplimit */
575 #define	IN6P_HOPOPTS		0x00040000 /* receive hop-by-hop options */
576 #define	IN6P_DSTOPTS		0x00080000 /* receive dst options after rthdr */
577 #define	IN6P_RTHDR		0x00100000 /* receive routing header */
578 #define	IN6P_RTHDRDSTOPTS	0x00200000 /* receive dstoptions before rthdr */
579 #define	IN6P_TCLASS		0x00400000 /* receive traffic class value */
580 #define	IN6P_AUTOFLOWLABEL	0x00800000 /* attach flowlabel automatically */
581 #define	INP_TIMEWAIT		0x01000000 /* in TIMEWAIT, ppcb is tcptw */
582 #define	INP_ONESBCAST		0x02000000 /* send all-ones broadcast */
583 #define	INP_DROPPED		0x04000000 /* protocol drop flag */
584 #define	INP_SOCKREF		0x08000000 /* strong socket reference */
585 #define	INP_RESERVED_0          0x10000000 /* reserved field */
586 #define	INP_RESERVED_1          0x20000000 /* reserved field */
587 #define	IN6P_RFC2292		0x40000000 /* used RFC2292 API on the socket */
588 #define	IN6P_MTU		0x80000000 /* receive path MTU */
589 
590 #define	INP_CONTROLOPTS		(INP_RECVOPTS|INP_RECVRETOPTS|INP_RECVDSTADDR|\
591 				 INP_RECVIF|INP_RECVTTL|INP_RECVTOS|\
592 				 IN6P_PKTINFO|IN6P_HOPLIMIT|IN6P_HOPOPTS|\
593 				 IN6P_DSTOPTS|IN6P_RTHDR|IN6P_RTHDRDSTOPTS|\
594 				 IN6P_TCLASS|IN6P_AUTOFLOWLABEL|IN6P_RFC2292|\
595 				 IN6P_MTU)
596 
597 /*
598  * Flags for inp_flags2.
599  */
600 #define	INP_LLE_VALID		0x00000001 /* cached lle is valid */
601 #define	INP_RT_VALID		0x00000002 /* cached rtentry is valid */
602 #define	INP_PCBGROUPWILD	0x00000004 /* in pcbgroup wildcard list */
603 #define	INP_REUSEPORT		0x00000008 /* SO_REUSEPORT option is set */
604 #define	INP_FREED		0x00000010 /* inp itself is not valid */
605 #define	INP_REUSEADDR		0x00000020 /* SO_REUSEADDR option is set */
606 #define	INP_BINDMULTI		0x00000040 /* IP_BINDMULTI option is set */
607 #define	INP_RSS_BUCKET_SET	0x00000080 /* IP_RSS_LISTEN_BUCKET is set */
608 #define	INP_RECVFLOWID		0x00000100 /* populate recv datagram with flow info */
609 #define	INP_RECVRSSBUCKETID	0x00000200 /* populate recv datagram with bucket id */
610 
611 /*
612  * Flags passed to in_pcblookup*() functions.
613  */
614 #define	INPLOOKUP_WILDCARD	0x00000001	/* Allow wildcard sockets. */
615 #define	INPLOOKUP_RLOCKPCB	0x00000002	/* Return inpcb read-locked. */
616 #define	INPLOOKUP_WLOCKPCB	0x00000004	/* Return inpcb write-locked. */
617 
618 #define	INPLOOKUP_MASK	(INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB | \
619 			    INPLOOKUP_WLOCKPCB)
620 
621 #define	sotoinpcb(so)	((struct inpcb *)(so)->so_pcb)
622 #define	sotoin6pcb(so)	sotoinpcb(so) /* for KAME src sync over BSD*'s */
623 
624 #define	INP_SOCKAF(so) so->so_proto->pr_domain->dom_family
625 
626 #define	INP_CHECK_SOCKAF(so, af)	(INP_SOCKAF(so) == af)
627 
628 /*
629  * Constants for pcbinfo.ipi_hashfields.
630  */
631 #define	IPI_HASHFIELDS_NONE	0
632 #define	IPI_HASHFIELDS_2TUPLE	1
633 #define	IPI_HASHFIELDS_4TUPLE	2
634 
635 #ifdef _KERNEL
636 VNET_DECLARE(int, ipport_reservedhigh);
637 VNET_DECLARE(int, ipport_reservedlow);
638 VNET_DECLARE(int, ipport_lowfirstauto);
639 VNET_DECLARE(int, ipport_lowlastauto);
640 VNET_DECLARE(int, ipport_firstauto);
641 VNET_DECLARE(int, ipport_lastauto);
642 VNET_DECLARE(int, ipport_hifirstauto);
643 VNET_DECLARE(int, ipport_hilastauto);
644 VNET_DECLARE(int, ipport_randomized);
645 VNET_DECLARE(int, ipport_randomcps);
646 VNET_DECLARE(int, ipport_randomtime);
647 VNET_DECLARE(int, ipport_stoprandom);
648 VNET_DECLARE(int, ipport_tcpallocs);
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 #define	V_ipport_randomcps	VNET(ipport_randomcps)
660 #define	V_ipport_randomtime	VNET(ipport_randomtime)
661 #define	V_ipport_stoprandom	VNET(ipport_stoprandom)
662 #define	V_ipport_tcpallocs	VNET(ipport_tcpallocs)
663 
664 void	in_pcbinfo_destroy(struct inpcbinfo *);
665 void	in_pcbinfo_init(struct inpcbinfo *, const char *, struct inpcbhead *,
666 	    int, int, char *, uma_init, uma_fini, uint32_t, u_int);
667 
668 int	in_pcbbind_check_bindmulti(const struct inpcb *ni,
669 	    const struct inpcb *oi);
670 
671 struct inpcbgroup *
672 	in_pcbgroup_byhash(struct inpcbinfo *, u_int, uint32_t);
673 struct inpcbgroup *
674 	in_pcbgroup_byinpcb(struct inpcb *);
675 struct inpcbgroup *
676 	in_pcbgroup_bytuple(struct inpcbinfo *, struct in_addr, u_short,
677 	    struct in_addr, u_short);
678 void	in_pcbgroup_destroy(struct inpcbinfo *);
679 int	in_pcbgroup_enabled(struct inpcbinfo *);
680 void	in_pcbgroup_init(struct inpcbinfo *, u_int, int);
681 void	in_pcbgroup_remove(struct inpcb *);
682 void	in_pcbgroup_update(struct inpcb *);
683 void	in_pcbgroup_update_mbuf(struct inpcb *, struct mbuf *);
684 
685 void	in_pcbpurgeif0(struct inpcbinfo *, struct ifnet *);
686 int	in_pcballoc(struct socket *, struct inpcbinfo *);
687 int	in_pcbbind(struct inpcb *, struct sockaddr *, struct ucred *);
688 int	in_pcb_lport(struct inpcb *, struct in_addr *, u_short *,
689 	    struct ucred *, int);
690 int	in_pcbbind_setup(struct inpcb *, struct sockaddr *, in_addr_t *,
691 	    u_short *, struct ucred *);
692 int	in_pcbconnect(struct inpcb *, struct sockaddr *, struct ucred *);
693 int	in_pcbconnect_mbuf(struct inpcb *, struct sockaddr *, struct ucred *,
694 	    struct mbuf *);
695 int	in_pcbconnect_setup(struct inpcb *, struct sockaddr *, in_addr_t *,
696 	    u_short *, in_addr_t *, u_short *, struct inpcb **,
697 	    struct ucred *);
698 void	in_pcbdetach(struct inpcb *);
699 void	in_pcbdisconnect(struct inpcb *);
700 void	in_pcbdrop(struct inpcb *);
701 void	in_pcbfree(struct inpcb *);
702 int	in_pcbinshash(struct inpcb *);
703 int	in_pcbinshash_nopcbgroup(struct inpcb *);
704 int	in_pcbladdr(struct inpcb *, struct in_addr *, struct in_addr *,
705 	    struct ucred *);
706 struct inpcb *
707 	in_pcblookup_local(struct inpcbinfo *,
708 	    struct in_addr, u_short, int, struct ucred *);
709 struct inpcb *
710 	in_pcblookup(struct inpcbinfo *, struct in_addr, u_int,
711 	    struct in_addr, u_int, int, struct ifnet *);
712 struct inpcb *
713 	in_pcblookup_mbuf(struct inpcbinfo *, struct in_addr, u_int,
714 	    struct in_addr, u_int, int, struct ifnet *, struct mbuf *);
715 void	in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr,
716 	    int, struct inpcb *(*)(struct inpcb *, int));
717 void	in_pcbref(struct inpcb *);
718 void	in_pcbrehash(struct inpcb *);
719 void	in_pcbrehash_mbuf(struct inpcb *, struct mbuf *);
720 int	in_pcbrele(struct inpcb *);
721 int	in_pcbrele_rlocked(struct inpcb *);
722 int	in_pcbrele_wlocked(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 #endif /* _KERNEL */
730 
731 #endif /* !_NETINET_IN_PCB_H_ */
732