xref: /titanic_50/usr/src/uts/common/inet/ipclassifier.h (revision c0889d7a91fa87e1cb7ef4457629b0cb51d47b50)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #ifndef	_INET_IPCLASSIFIER_H
27 #define	_INET_IPCLASSIFIER_H
28 
29 #pragma ident	"%Z%%M%	%I%	%E% SMI"
30 
31 #ifdef	__cplusplus
32 extern "C" {
33 #endif
34 
35 #include <inet/common.h>
36 #include <inet/ip.h>
37 #include <inet/mi.h>
38 #include <inet/tcp.h>
39 #include <inet/ip6.h>
40 #include <netinet/in.h>		/* for IPPROTO_* constants */
41 #include <sys/sdt.h>
42 
43 typedef void (*edesc_spf)(void *, mblk_t *, void *, int);
44 typedef void (*edesc_rpf)(void *, mblk_t *, void *);
45 
46 /*
47  * ==============================
48  * =	The CONNECTION		=
49  * ==============================
50  */
51 
52 /*
53  * The connection structure contains the common information/flags/ref needed.
54  * Implementation will keep the connection struct, the layers (with their
55  * respective data for event i.e. tcp_t if event was tcp_input) all in one
56  * contiguous memory location.
57  */
58 
59 /* Conn Flags */
60 /* Unused			0x00020000 */
61 /* Unused			0x00040000 */
62 #define	IPCL_FULLY_BOUND	0x00080000	/* Bound to correct squeue */
63 #define	IPCL_CHECK_POLICY	0x00100000	/* Needs policy checking */
64 #define	IPCL_SOCKET		0x00200000	/* Sockfs connection */
65 #define	IPCL_ACCEPTOR		0x00400000	/* Sockfs priv acceptor */
66 #define	IPCL_CL_LISTENER	0x00800000	/* Cluster listener */
67 #define	IPCL_EAGER		0x01000000	/* Incoming connection */
68 /* Unused			0x02000000 */
69 #define	IPCL_TCP6		0x04000000	/* AF_INET6 TCP */
70 #define	IPCL_TCP4		0x08000000	/* IPv4 packet format TCP */
71 /* Unused			0x10000000 */
72 /* Unused			0x20000000 */
73 #define	IPCL_CONNECTED		0x40000000	/* Conn in connected table */
74 #define	IPCL_BOUND		0x80000000	/* Conn in bind table */
75 
76 /* Flags identifying the type of conn */
77 #define	IPCL_TCPCONN		0x00000001	/* From tcp_conn_cache */
78 #define	IPCL_SCTPCONN		0x00000002	/* From sctp_conn_cache */
79 #define	IPCL_IPCCONN		0x00000004	/* From ip_conn_cache */
80 #define	IPCL_UDPCONN		0x00000008	/* From udp_conn_cache */
81 #define	IPCL_RAWIPCONN		0x00000010	/* From rawip_conn_cache */
82 #define	IPCL_RTSCONN		0x00000020	/* From rts_conn_cache */
83 #define	IPCL_ISV6		0x00000040	/* AF_INET6 */
84 #define	IPCL_IPTUN		0x00000080	/* Has "tun" plumbed above it */
85 
86 /* Conn Masks */
87 #define	IPCL_TCP		(IPCL_TCP4|IPCL_TCP6)
88 #define	IPCL_REMOVED		0x00000100
89 #define	IPCL_REUSED		0x00000200
90 
91 /* The packet format is IPv4; could be an AF_INET or AF_INET6 socket */
92 #define	IPCL_IS_TCP4(connp)						\
93 	(((connp)->conn_flags & IPCL_TCP4))
94 
95 /* Connected AF_INET with no IPsec policy */
96 #define	IPCL_IS_TCP4_CONNECTED_NO_POLICY(connp)				\
97 	(((connp)->conn_flags &						\
98 		(IPCL_TCP4|IPCL_CONNECTED|IPCL_CHECK_POLICY|IPCL_TCP6))	\
99 		== (IPCL_TCP4|IPCL_CONNECTED))
100 
101 #define	IPCL_IS_CONNECTED(connp)					\
102 	((connp)->conn_flags & IPCL_CONNECTED)
103 
104 #define	IPCL_IS_BOUND(connp)						\
105 	((connp)->conn_flags & IPCL_BOUND)
106 
107 /* AF_INET TCP that is bound */
108 #define	IPCL_IS_TCP4_BOUND(connp)					\
109 	(((connp)->conn_flags &						\
110 		(IPCL_TCP4|IPCL_BOUND|IPCL_TCP6)) ==			\
111 		(IPCL_TCP4|IPCL_BOUND))
112 
113 #define	IPCL_IS_FULLY_BOUND(connp)					\
114 	((connp)->conn_flags & IPCL_FULLY_BOUND)
115 
116 /*
117  * Can't use conn_protocol since we need to tell difference
118  * between a real TCP socket and a SOCK_RAW, IPPROTO_TCP.
119  */
120 #define	IPCL_IS_TCP(connp)						\
121 	((connp)->conn_flags & IPCL_TCPCONN)
122 
123 #define	IPCL_IS_SCTP(connp)						\
124 	((connp)->conn_flags & IPCL_SCTPCONN)
125 
126 #define	IPCL_IS_UDP(connp)						\
127 	((connp)->conn_flags & IPCL_UDPCONN)
128 
129 #define	IPCL_IS_RAWIP(connp)						\
130 	((connp)->conn_flags & IPCL_RAWIPCONN)
131 
132 #define	IPCL_IS_RTS(connp)						\
133 	((connp)->conn_flags & IPCL_RTSCONN)
134 
135 /* FIXME: Isn't it sufficient to check IPCL_IPTUN? */
136 #define	IPCL_IS_IPTUN(connp)						\
137 	(((connp)->conn_ulp == IPPROTO_ENCAP ||				\
138 	(connp)->conn_ulp == IPPROTO_IPV6) &&				\
139 	((connp)->conn_flags & IPCL_IPTUN))
140 
141 typedef struct connf_s connf_t;
142 
143 typedef struct
144 {
145 	int	ctb_depth;
146 #define	CONN_STACK_DEPTH	15
147 	pc_t	ctb_stack[CONN_STACK_DEPTH];
148 } conn_trace_t;
149 
150 /*
151  * The initial fields in the conn_t are setup by the kmem_cache constructor,
152  * and are preserved when it is freed. Fields after that are bzero'ed when
153  * the conn_t is freed.
154  */
155 struct conn_s {
156 	kmutex_t	conn_lock;
157 	uint32_t	conn_ref;		/* Reference counter */
158 	uint32_t	conn_flags;		/* Conn Flags */
159 
160 
161 	union {
162 		tcp_t		*cp_tcp;	/* Pointer to the tcp struct */
163 		struct udp_s	*cp_udp;	/* Pointer to the udp struct */
164 		struct icmp_s	*cp_icmp;	/* Pointer to rawip struct */
165 		struct rts_s	*cp_rts;	/* Pointer to rts struct */
166 		void		*cp_priv;
167 	} conn_proto_priv;
168 #define	conn_tcp	conn_proto_priv.cp_tcp
169 #define	conn_udp	conn_proto_priv.cp_udp
170 #define	conn_icmp	conn_proto_priv.cp_icmp
171 #define	conn_rts	conn_proto_priv.cp_rts
172 #define	conn_priv	conn_proto_priv.cp_priv
173 
174 	kcondvar_t	conn_cv;
175 	uint8_t		conn_ulp;		/* protocol type */
176 
177 	edesc_rpf	conn_recv;		/* Pointer to recv routine */
178 
179 	/* Fields after this are bzero'ed when the conn_t is freed. */
180 
181 	squeue_t	*conn_sqp;		/* Squeue for processing */
182 	uint_t		conn_state_flags;	/* IP state flags */
183 #define	conn_start_clr	conn_state_flags
184 
185 	ire_t		*conn_ire_cache; 	/* outbound ire cache */
186 	unsigned int
187 		conn_on_sqp : 1,		/* Conn is being processed */
188 		conn_dontroute : 1,		/* SO_DONTROUTE state */
189 		conn_loopback : 1,		/* SO_LOOPBACK state */
190 		conn_broadcast : 1,		/* SO_BROADCAST state */
191 
192 		conn_reuseaddr : 1,		/* SO_REUSEADDR state */
193 		conn_multicast_loop : 1,	/* IP_MULTICAST_LOOP */
194 		conn_multi_router : 1,		/* Wants all multicast pkts */
195 		conn_draining : 1,		/* ip_wsrv running */
196 
197 		conn_did_putbq : 1,		/* ip_wput did a putbq */
198 		conn_unspec_src : 1,		/* IP_UNSPEC_SRC */
199 		conn_policy_cached : 1,		/* Is policy cached/latched ? */
200 		conn_in_enforce_policy : 1,	/* Enforce Policy on inbound */
201 
202 		conn_out_enforce_policy : 1,	/* Enforce Policy on outbound */
203 		conn_af_isv6 : 1,		/* ip address family ver 6 */
204 		conn_pkt_isv6 : 1,		/* ip packet format ver 6 */
205 		conn_ip_recvpktinfo : 1,	/* IPV*_RECVPKTINFO option */
206 
207 		conn_ipv6_recvhoplimit : 1,	/* IPV6_RECVHOPLIMIT option */
208 		conn_ipv6_recvhopopts : 1,	/* IPV6_RECVHOPOPTS option */
209 		conn_ipv6_recvdstopts : 1,	/* IPV6_RECVDSTOPTS option */
210 		conn_ipv6_recvrthdr : 1,	/* IPV6_RECVRTHDR option */
211 
212 		conn_ipv6_recvrtdstopts : 1,	/* IPV6_RECVRTHDRDSTOPTS */
213 		conn_ipv6_v6only : 1,		/* IPV6_V6ONLY */
214 		conn_ipv6_recvtclass : 1,	/* IPV6_RECVTCLASS */
215 		conn_ipv6_recvpathmtu : 1,	/* IPV6_RECVPATHMTU */
216 
217 		conn_pathmtu_valid : 1,		/* The cached mtu is valid. */
218 		conn_ipv6_dontfrag : 1,		/* IPV6_DONTFRAG */
219 		conn_fully_bound : 1,		/* Fully bound connection */
220 		conn_recvif : 1,		/* IP_RECVIF option */
221 
222 		conn_recvslla : 1,		/* IP_RECVSLLA option */
223 		conn_mdt_ok : 1,		/* MDT is permitted */
224 		conn_nexthop_set : 1,
225 		conn_allzones : 1,		/* SO_ALLZONES */
226 
227 		conn_lso_ok : 1;		/* LSO is usable */
228 
229 	ill_t		*conn_nofailover_ill;	/* Failover ill */
230 	ill_t		*conn_dhcpinit_ill;	/* IP_DHCPINIT_IF */
231 	ipsec_latch_t	*conn_latch;		/* latched state */
232 	ill_t		*conn_outgoing_ill;	/* IP{,V6}_BOUND_IF */
233 	edesc_spf	conn_send;		/* Pointer to send routine */
234 	queue_t		*conn_rq;		/* Read queue */
235 	queue_t		*conn_wq;		/* Write queue */
236 	dev_t		conn_dev;		/* Minor number */
237 	vmem_t		*conn_minor_arena;	/* Minor arena */
238 
239 	cred_t		*conn_cred;		/* Credentials */
240 	connf_t		*conn_g_fanout;		/* Global Hash bucket head */
241 	struct conn_s	*conn_g_next;		/* Global Hash chain next */
242 	struct conn_s	*conn_g_prev;		/* Global Hash chain prev */
243 	struct ipsec_policy_head_s *conn_policy; /* Configured policy */
244 	in6_addr_t	conn_bound_source_v6;
245 #define	conn_bound_source	V4_PART_OF_V6(conn_bound_source_v6)
246 
247 	connf_t		*conn_fanout;		/* Hash bucket we're part of */
248 	struct conn_s	*conn_next;		/* Hash chain next */
249 	struct conn_s	*conn_prev;		/* Hash chain prev */
250 	struct {
251 		in6_addr_t connua_laddr;	/* Local address */
252 		in6_addr_t connua_faddr;	/* Remote address */
253 	} connua_v6addr;
254 #define	conn_src	V4_PART_OF_V6(connua_v6addr.connua_laddr)
255 #define	conn_rem	V4_PART_OF_V6(connua_v6addr.connua_faddr)
256 #define	conn_srcv6	connua_v6addr.connua_laddr
257 #define	conn_remv6	connua_v6addr.connua_faddr
258 	union {
259 		/* Used for classifier match performance */
260 		uint32_t		conn_ports2;
261 		struct {
262 			in_port_t	tcpu_fport;	/* Remote port */
263 			in_port_t	tcpu_lport;	/* Local port */
264 		} tcpu_ports;
265 	} u_port;
266 #define	conn_fport	u_port.tcpu_ports.tcpu_fport
267 #define	conn_lport	u_port.tcpu_ports.tcpu_lport
268 #define	conn_ports	u_port.conn_ports2
269 #define	conn_upq	conn_rq
270 	uint8_t		conn_unused_byte;
271 
272 	uint_t		conn_proto;		/* SO_PROTOTYPE state */
273 	ill_t		*conn_incoming_ill;	/* IP{,V6}_BOUND_IF */
274 	ill_t		*conn_outgoing_pill;	/* IP{,V6}_BOUND_PIF */
275 	ill_t		*conn_oper_pending_ill; /* pending shared ioctl */
276 
277 	ilg_t	*conn_ilg;		/* Group memberships */
278 	int	conn_ilg_allocated;	/* Number allocated */
279 	int	conn_ilg_inuse;		/* Number currently used */
280 	int	conn_ilg_walker_cnt;	/* No of ilg walkers */
281 	/* XXXX get rid of this, once ilg_delete_all is fixed */
282 	kcondvar_t	conn_refcv;
283 
284 	struct ipif_s	*conn_multicast_ipif;	/* IP_MULTICAST_IF */
285 	ill_t		*conn_multicast_ill;	/* IPV6_MULTICAST_IF */
286 	int		conn_orig_bound_ifindex; /* BOUND_IF before MOVE */
287 	int		conn_orig_multicast_ifindex;
288 						/* IPv6 MC IF before MOVE */
289 	struct conn_s 	*conn_drain_next;	/* Next conn in drain list */
290 	struct conn_s	*conn_drain_prev;	/* Prev conn in drain list */
291 	idl_t		*conn_idl;		/* Ptr to the drain list head */
292 	mblk_t		*conn_ipsec_opt_mp;	/* ipsec option mblk */
293 	uint32_t	conn_src_preferences;	/* prefs for src addr select */
294 	/* mtuinfo from IPV6_PACKET_TOO_BIG conditional on conn_pathmtu_valid */
295 	struct ip6_mtuinfo mtuinfo;
296 	zoneid_t	conn_zoneid;		/* zone connection is in */
297 	in6_addr_t	conn_nexthop_v6;	/* nexthop IP address */
298 	uchar_t		conn_broadcast_ttl; 	/* IP_BROADCAST_TTL */
299 #define	conn_nexthop_v4	V4_PART_OF_V6(conn_nexthop_v6)
300 	cred_t		*conn_peercred;		/* Peer credentials, if any */
301 
302 	unsigned int
303 		conn_ulp_labeled : 1,		/* ULP label is synced */
304 		conn_mlp_type : 2,		/* mlp_type_t; tsol/tndb.h */
305 		conn_anon_mlp : 1,		/* user wants anon MLP */
306 
307 		conn_anon_port : 1,		/* user bound anonymously */
308 		conn_mac_exempt : 1,		/* unlabeled with loose MAC */
309 		conn_spare : 26;
310 	netstack_t	*conn_netstack;	/* Corresponds to a netstack_hold */
311 #ifdef CONN_DEBUG
312 #define	CONN_TRACE_MAX	10
313 	int		conn_trace_last;	/* ndx of last used tracebuf */
314 	conn_trace_t	conn_trace_buf[CONN_TRACE_MAX];
315 #endif
316 };
317 
318 #define	CONN_CRED(connp) ((connp)->conn_peercred == NULL ? \
319 	(connp)->conn_cred : (connp)->conn_peercred)
320 #define	BEST_CRED(mp, connp) ((DB_CRED(mp) != NULL &&	\
321 	crgetlabel(DB_CRED(mp)) != NULL) ? DB_CRED(mp) : CONN_CRED(connp))
322 
323 /*
324  * connf_t - connection fanout data.
325  *
326  * The hash tables and their linkage (conn_t.{hashnextp, hashprevp} are
327  * protected by the per-bucket lock. Each conn_t inserted in the list
328  * points back at the connf_t that heads the bucket.
329  */
330 struct connf_s {
331 	struct conn_s	*connf_head;
332 	kmutex_t	connf_lock;
333 };
334 
335 #define	CONN_INC_REF(connp)	{				\
336 	mutex_enter(&(connp)->conn_lock);			\
337 	DTRACE_PROBE1(conn__inc__ref, conn_t *, connp);		\
338 	ASSERT(conn_trace_ref(connp));				\
339 	(connp)->conn_ref++;					\
340 	ASSERT((connp)->conn_ref != 0);				\
341 	mutex_exit(&(connp)->conn_lock);			\
342 }
343 
344 #define	CONN_INC_REF_LOCKED(connp)	{			\
345 	DTRACE_PROBE1(conn__inc__ref, conn_t *, connp);		\
346 	ASSERT(MUTEX_HELD(&(connp)->conn_lock));	 	\
347 	ASSERT(conn_trace_ref(connp));				\
348 	(connp)->conn_ref++;					\
349 	ASSERT((connp)->conn_ref != 0);				\
350 }
351 
352 #define	CONN_DEC_REF(connp)	{					\
353 	mutex_enter(&(connp)->conn_lock);				\
354 	DTRACE_PROBE1(conn__dec__ref, conn_t *, connp);			\
355 	/*								\
356 	 * The squeue framework always does a CONN_DEC_REF after return	\
357 	 * from TCP. Hence the refcnt must be at least 2 if conn_on_sqp	\
358 	 * is B_TRUE and conn_ref is being decremented. This is to	\
359 	 * account for the mblk being currently processed.		\
360 	 */								\
361 	if ((connp)->conn_ref <= 0 ||					\
362 	    ((connp)->conn_ref == 1 && (connp)->conn_on_sqp))		\
363 		cmn_err(CE_PANIC, "CONN_DEC_REF: connp(%p) has ref "	\
364 			"= %d\n", (void *)(connp), (connp)->conn_ref);	\
365 	ASSERT(conn_untrace_ref(connp));				\
366 	(connp)->conn_ref--;						\
367 	if ((connp)->conn_ref == 0) {					\
368 		/* Refcnt can't increase again, safe to drop lock */	\
369 		mutex_exit(&(connp)->conn_lock);			\
370 		ipcl_conn_destroy(connp);				\
371 	} else {							\
372 		cv_broadcast(&(connp)->conn_cv);			\
373 		mutex_exit(&(connp)->conn_lock);			\
374 	}								\
375 }
376 
377 /*
378  * For use with subsystems within ip which use ALL_ZONES as a wildcard
379  */
380 #define	IPCL_ZONEID(connp)						\
381 	((connp)->conn_allzones ? ALL_ZONES : (connp)->conn_zoneid)
382 
383 /*
384  * For matching between a conn_t and a zoneid.
385  */
386 #define	IPCL_ZONE_MATCH(connp, zoneid) 					\
387 	(((connp)->conn_allzones) ||					\
388 	    ((zoneid) == ALL_ZONES) ||					\
389 	    (connp)->conn_zoneid == (zoneid))
390 
391 
392 #define	_IPCL_V4_MATCH(v6addr, v4addr)	\
393 	(V4_PART_OF_V6((v6addr)) == (v4addr) && IN6_IS_ADDR_V4MAPPED(&(v6addr)))
394 
395 #define	_IPCL_V4_MATCH_ANY(addr)	\
396 	(IN6_IS_ADDR_V4MAPPED_ANY(&(addr)) || IN6_IS_ADDR_UNSPECIFIED(&(addr)))
397 
398 
399 /*
400  * IPCL_PROTO_MATCH() only matches conns with the specified zoneid, while
401  * IPCL_PROTO_MATCH_V6() can match other conns in the multicast case, see
402  * ip_fanout_proto().
403  */
404 #define	IPCL_PROTO_MATCH(connp, protocol, ipha, ill,			\
405     fanout_flags, zoneid)						\
406 	((((connp)->conn_src == INADDR_ANY) ||				\
407 	(((connp)->conn_src == ((ipha)->ipha_dst)) &&			\
408 	    (((connp)->conn_rem == INADDR_ANY) ||			\
409 	((connp)->conn_rem == ((ipha)->ipha_src))))) &&			\
410 	IPCL_ZONE_MATCH(connp, zoneid) &&				\
411 	(conn_wantpacket((connp), (ill), (ipha), (fanout_flags), 	\
412 	    (zoneid)) || ((protocol) == IPPROTO_PIM) ||			\
413 	    ((protocol) == IPPROTO_RSVP)))
414 
415 #define	IPCL_PROTO_MATCH_V6(connp, protocol, ip6h, ill,			   \
416     fanout_flags, zoneid)						   \
417 	((IN6_IS_ADDR_UNSPECIFIED(&(connp)->conn_srcv6) ||		   \
418 	(IN6_ARE_ADDR_EQUAL(&(connp)->conn_srcv6, &((ip6h)->ip6_dst)) &&   \
419 	(IN6_IS_ADDR_UNSPECIFIED(&(connp)->conn_remv6) ||		   \
420 	IN6_ARE_ADDR_EQUAL(&(connp)->conn_remv6, &((ip6h)->ip6_src))))) && \
421 	(conn_wantpacket_v6((connp), (ill), (ip6h),			   \
422 	(fanout_flags), (zoneid)) || ((protocol) == IPPROTO_RSVP)))
423 
424 #define	IPCL_CONN_HASH(src, ports, ipst)				\
425 	((unsigned)(ntohl((src)) ^ ((ports) >> 24) ^ ((ports) >> 16) ^	\
426 	((ports) >> 8) ^ (ports)) % (ipst)->ips_ipcl_conn_fanout_size)
427 
428 #define	IPCL_CONN_HASH_V6(src, ports, ipst)				\
429 	IPCL_CONN_HASH(V4_PART_OF_V6((src)), (ports), (ipst))
430 
431 #define	IPCL_CONN_MATCH(connp, proto, src, dst, ports)			\
432 	((connp)->conn_ulp == (proto) &&				\
433 		(connp)->conn_ports == (ports) &&      			\
434 		_IPCL_V4_MATCH((connp)->conn_remv6, (src)) &&		\
435 		_IPCL_V4_MATCH((connp)->conn_srcv6, (dst)) &&		\
436 		!(connp)->conn_ipv6_v6only)
437 
438 #define	IPCL_CONN_MATCH_V6(connp, proto, src, dst, ports)		\
439 	((connp)->conn_ulp == (proto) &&				\
440 		(connp)->conn_ports == (ports) &&      			\
441 		IN6_ARE_ADDR_EQUAL(&(connp)->conn_remv6, &(src)) &&	\
442 		IN6_ARE_ADDR_EQUAL(&(connp)->conn_srcv6, &(dst)))
443 
444 #define	IPCL_CONN_INIT(connp, protocol, src, rem, ports) {		\
445 	(connp)->conn_ulp = protocol;					\
446 	IN6_IPADDR_TO_V4MAPPED(src, &(connp)->conn_srcv6);		\
447 	IN6_IPADDR_TO_V4MAPPED(rem, &(connp)->conn_remv6);		\
448 	(connp)->conn_ports = ports;					\
449 }
450 
451 #define	IPCL_CONN_INIT_V6(connp, protocol, src, rem, ports) {		\
452 	(connp)->conn_ulp = protocol;					\
453 	(connp)->conn_srcv6 = src;					\
454 	(connp)->conn_remv6 = rem;					\
455 	(connp)->conn_ports = ports;					\
456 }
457 
458 #define	IPCL_PORT_HASH(port, size) \
459 	((((port) >> 8) ^ (port)) & ((size) - 1))
460 
461 #define	IPCL_BIND_HASH(lport, ipst)					\
462 	((unsigned)(((lport) >> 8) ^ (lport)) % \
463 	    (ipst)->ips_ipcl_bind_fanout_size)
464 
465 #define	IPCL_BIND_MATCH(connp, proto, laddr, lport)			\
466 	((connp)->conn_ulp == (proto) &&				\
467 		(connp)->conn_lport == (lport) &&			\
468 		(_IPCL_V4_MATCH_ANY((connp)->conn_srcv6) ||		\
469 		_IPCL_V4_MATCH((connp)->conn_srcv6, (laddr))) &&	\
470 		!(connp)->conn_ipv6_v6only)
471 
472 #define	IPCL_BIND_MATCH_V6(connp, proto, laddr, lport)			\
473 	((connp)->conn_ulp == (proto) &&				\
474 		(connp)->conn_lport == (lport) &&			\
475 		(IN6_ARE_ADDR_EQUAL(&(connp)->conn_srcv6, &(laddr)) ||	\
476 		IN6_IS_ADDR_UNSPECIFIED(&(connp)->conn_srcv6)))
477 
478 #define	IPCL_UDP_MATCH(connp, lport, laddr, fport, faddr)		\
479 	(((connp)->conn_lport == (lport)) &&				\
480 	((_IPCL_V4_MATCH_ANY((connp)->conn_srcv6) ||			\
481 	(_IPCL_V4_MATCH((connp)->conn_srcv6, (laddr)) &&		\
482 	(_IPCL_V4_MATCH_ANY((connp)->conn_remv6) ||			\
483 	(_IPCL_V4_MATCH((connp)->conn_remv6, (faddr)) &&		\
484 	(connp)->conn_fport == (fport)))))) &&				\
485 	!(connp)->conn_ipv6_v6only)
486 
487 #define	IPCL_UDP_MATCH_V6(connp, lport, laddr, fport, faddr)	\
488 	(((connp)->conn_lport == (lport)) &&			\
489 	(IN6_IS_ADDR_UNSPECIFIED(&(connp)->conn_srcv6) ||	\
490 	(IN6_ARE_ADDR_EQUAL(&(connp)->conn_srcv6, &(laddr)) &&	\
491 	(IN6_IS_ADDR_UNSPECIFIED(&(connp)->conn_remv6) ||	\
492 	(IN6_ARE_ADDR_EQUAL(&(connp)->conn_remv6, &(faddr)) &&	\
493 	(connp)->conn_fport == (fport))))))
494 
495 #define	IPCL_TCP_EAGER_INIT(connp, protocol, src, rem, ports) {		\
496 	(connp)->conn_flags |= (IPCL_TCP4|IPCL_EAGER);			\
497 	IN6_IPADDR_TO_V4MAPPED(src, &(connp)->conn_srcv6);		\
498 	IN6_IPADDR_TO_V4MAPPED(rem, &(connp)->conn_remv6);		\
499 	(connp)->conn_ports = ports;					\
500 	(connp)->conn_send = ip_output;					\
501 	(connp)->conn_sqp = IP_SQUEUE_GET(lbolt);			\
502 }
503 
504 #define	IPCL_TCP_EAGER_INIT_V6(connp, protocol, src, rem, ports) {	\
505 	(connp)->conn_flags |= (IPCL_TCP6|IPCL_EAGER|IPCL_ISV6);	\
506 	(connp)->conn_srcv6 = src;					\
507 	(connp)->conn_remv6 = rem;					\
508 	(connp)->conn_ports = ports;					\
509 	(connp)->conn_send = ip_output_v6;				\
510 	(connp)->conn_sqp = IP_SQUEUE_GET(lbolt);			\
511 }
512 
513 #define	IPCL_UDP_HASH(lport, ipst)	\
514 	IPCL_PORT_HASH(lport, (ipst)->ips_ipcl_udp_fanout_size)
515 
516 #define	CONN_G_HASH_SIZE	1024
517 
518 /* Raw socket hash function. */
519 #define	IPCL_RAW_HASH(lport, ipst)	\
520 	IPCL_PORT_HASH(lport, (ipst)->ips_ipcl_raw_fanout_size)
521 
522 /*
523  * This is similar to IPCL_BIND_MATCH except that the local port check
524  * is changed to a wildcard port check.
525  */
526 #define	IPCL_RAW_MATCH(connp, proto, laddr)			\
527 	((connp)->conn_ulp == (proto) &&			\
528 	(connp)->conn_lport == 0 &&				\
529 	(_IPCL_V4_MATCH_ANY((connp)->conn_srcv6) ||		\
530 	_IPCL_V4_MATCH((connp)->conn_srcv6, (laddr))))
531 
532 #define	IPCL_RAW_MATCH_V6(connp, proto, laddr)			\
533 	((connp)->conn_ulp == (proto) &&			\
534 	(connp)->conn_lport == 0 &&				\
535 	(IN6_IS_ADDR_UNSPECIFIED(&(connp)->conn_srcv6) ||	\
536 	IN6_ARE_ADDR_EQUAL(&(connp)->conn_srcv6, &(laddr))))
537 
538 /* Function prototypes */
539 extern void ipcl_g_init(void);
540 extern void ipcl_init(ip_stack_t *);
541 extern void ipcl_g_destroy(void);
542 extern void ipcl_destroy(ip_stack_t *);
543 extern conn_t *ipcl_conn_create(uint32_t, int, netstack_t *);
544 extern void ipcl_conn_destroy(conn_t *);
545 
546 void ipcl_hash_insert_wildcard(connf_t *, conn_t *);
547 void ipcl_hash_remove(conn_t *);
548 void ipcl_hash_remove_locked(conn_t *connp, connf_t *connfp);
549 
550 extern int	ipcl_bind_insert(conn_t *, uint8_t, ipaddr_t, uint16_t);
551 extern int	ipcl_bind_insert_v6(conn_t *, uint8_t, const in6_addr_t *,
552 		    uint16_t);
553 extern int	ipcl_conn_insert(conn_t *, uint8_t, ipaddr_t, ipaddr_t,
554 		    uint32_t);
555 extern int	ipcl_conn_insert_v6(conn_t *, uint8_t, const in6_addr_t *,
556 		    const in6_addr_t *, uint32_t, uint_t);
557 extern conn_t	*ipcl_get_next_conn(connf_t *, conn_t *, uint32_t);
558 
559 void ipcl_proto_insert(conn_t *, uint8_t);
560 void ipcl_proto_insert_v6(conn_t *, uint8_t);
561 conn_t *ipcl_classify_v4(mblk_t *, uint8_t, uint_t, zoneid_t, ip_stack_t *);
562 conn_t *ipcl_classify_v6(mblk_t *, uint8_t, uint_t, zoneid_t, ip_stack_t *);
563 conn_t *ipcl_classify(mblk_t *, zoneid_t, ip_stack_t *);
564 conn_t *ipcl_classify_raw(mblk_t *, uint8_t, zoneid_t, uint32_t, ipha_t *,
565 	    ip_stack_t *);
566 void	ipcl_globalhash_insert(conn_t *);
567 void	ipcl_globalhash_remove(conn_t *);
568 void	ipcl_walk(pfv_t, void *, ip_stack_t *);
569 conn_t	*ipcl_tcp_lookup_reversed_ipv4(ipha_t *, tcph_t *, int, ip_stack_t *);
570 conn_t	*ipcl_tcp_lookup_reversed_ipv6(ip6_t *, tcpha_t *, int, uint_t,
571 	    ip_stack_t *);
572 conn_t	*ipcl_lookup_listener_v4(uint16_t, ipaddr_t, zoneid_t, ip_stack_t *);
573 conn_t	*ipcl_lookup_listener_v6(uint16_t, in6_addr_t *, uint_t, zoneid_t,
574 	    ip_stack_t *);
575 int	conn_trace_ref(conn_t *);
576 int	conn_untrace_ref(conn_t *);
577 void	ipcl_conn_cleanup(conn_t *);
578 conn_t *ipcl_conn_tcp_lookup_reversed_ipv4(conn_t *, ipha_t *, tcph_t *,
579 	    ip_stack_t *);
580 conn_t *ipcl_conn_tcp_lookup_reversed_ipv6(conn_t *, ip6_t *, tcph_t *,
581 	    ip_stack_t *);
582 #ifdef	__cplusplus
583 }
584 #endif
585 
586 #endif	/* _INET_IPCLASSIFIER_H */
587