xref: /titanic_41/usr/src/uts/common/inet/ipclassifier.h (revision d24234c24aeaca4ca56ee3ac2794507968f274c4)
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 2009 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 #ifdef	__cplusplus
30 extern "C" {
31 #endif
32 
33 #include <inet/common.h>
34 #include <inet/ip.h>
35 #include <inet/mi.h>
36 #include <inet/tcp.h>
37 #include <inet/ip6.h>
38 #include <netinet/in.h>		/* for IPPROTO_* constants */
39 #include <sys/sdt.h>
40 #include <sys/socket_proto.h>
41 #include <sys/sunddi.h>
42 #include <sys/sunldi.h>
43 
44 typedef void (*edesc_rpf)(void *, mblk_t *, void *, ip_recv_attr_t *);
45 struct icmph_s;
46 struct icmp6_hdr;
47 typedef boolean_t (*edesc_vpf)(conn_t *, void *, struct icmph_s *,
48     struct icmp6_hdr *, ip_recv_attr_t *);
49 
50 /*
51  * ==============================
52  * =	The CONNECTION		=
53  * ==============================
54  */
55 
56 /*
57  * The connection structure contains the common information/flags/ref needed.
58  * Implementation will keep the connection struct, the layers (with their
59  * respective data for event i.e. tcp_t if event was tcp_input_data) all in one
60  * contiguous memory location.
61  */
62 
63 /* Conn Flags */
64 /* Unused			0x00020000 */
65 /* Unused			0x00040000 */
66 #define	IPCL_FULLY_BOUND	0x00080000	/* Bound to correct squeue */
67 /* Unused			0x00100000 */
68 /* Unused 			0x00200000 */
69 /* Unused			0x00400000 */
70 #define	IPCL_CL_LISTENER	0x00800000	/* Cluster listener */
71 /* Unused			0x01000000 */
72 /* Unused			0x02000000 */
73 /* Unused			0x04000000 */
74 /* Unused			0x08000000 */
75 /* Unused			0x10000000 */
76 /* Unused			0x20000000 */
77 #define	IPCL_CONNECTED		0x40000000	/* Conn in connected table */
78 #define	IPCL_BOUND		0x80000000	/* Conn in bind table */
79 
80 /* Flags identifying the type of conn */
81 #define	IPCL_TCPCONN		0x00000001	/* From tcp_conn_cache */
82 #define	IPCL_SCTPCONN		0x00000002	/* From sctp_conn_cache */
83 #define	IPCL_IPCCONN		0x00000004	/* From ip_conn_cache */
84 #define	IPCL_UDPCONN		0x00000008	/* From udp_conn_cache */
85 #define	IPCL_RAWIPCONN		0x00000010	/* From rawip_conn_cache */
86 #define	IPCL_RTSCONN		0x00000020	/* From rts_conn_cache */
87 /* Unused			0x00000040 */
88 #define	IPCL_IPTUN		0x00000080	/* iptun module above us */
89 
90 #define	IPCL_NONSTR		0x00001000	/* A non-STREAMS socket */
91 /* Unused			0x10000000 */
92 
93 #define	IPCL_REMOVED		0x00000100
94 #define	IPCL_REUSED		0x00000200
95 
96 #define	IPCL_IS_CONNECTED(connp)					\
97 	((connp)->conn_flags & IPCL_CONNECTED)
98 
99 #define	IPCL_IS_BOUND(connp)						\
100 	((connp)->conn_flags & IPCL_BOUND)
101 
102 /*
103  * Can't use conn_proto since we need to tell difference
104  * between a real TCP socket and a SOCK_RAW, IPPROTO_TCP.
105  */
106 #define	IPCL_IS_TCP(connp)						\
107 	((connp)->conn_flags & IPCL_TCPCONN)
108 
109 #define	IPCL_IS_SCTP(connp)						\
110 	((connp)->conn_flags & IPCL_SCTPCONN)
111 
112 #define	IPCL_IS_UDP(connp)						\
113 	((connp)->conn_flags & IPCL_UDPCONN)
114 
115 #define	IPCL_IS_RAWIP(connp)						\
116 	((connp)->conn_flags & IPCL_RAWIPCONN)
117 
118 #define	IPCL_IS_RTS(connp)						\
119 	((connp)->conn_flags & IPCL_RTSCONN)
120 
121 #define	IPCL_IS_IPTUN(connp)						\
122 	((connp)->conn_flags & IPCL_IPTUN)
123 
124 #define	IPCL_IS_NONSTR(connp)	((connp)->conn_flags & IPCL_NONSTR)
125 
126 typedef struct connf_s connf_t;
127 
128 typedef struct
129 {
130 	int	ctb_depth;
131 #define	CONN_STACK_DEPTH	15
132 	pc_t	ctb_stack[CONN_STACK_DEPTH];
133 } conn_trace_t;
134 
135 typedef struct ip_helper_minor_info_s {
136 	dev_t	ip_minfo_dev;		/* Device */
137 	vmem_t	*ip_minfo_arena;	/* Arena */
138 } ip_helper_minfo_t;
139 
140 /*
141  * ip helper stream info
142  */
143 typedef struct ip_helper_stream_info_s {
144 	ldi_handle_t		iphs_handle;
145 	queue_t 		*iphs_rq;
146 	queue_t 		*iphs_wq;
147 	ip_helper_minfo_t	*iphs_minfo;
148 } ip_helper_stream_info_t;
149 
150 /*
151  * Mandatory Access Control mode, in conn_t's conn_mac_mode field.
152  * 	CONN_MAC_DEFAULT: strict enforcement of MAC.
153  * 	CONN_MAC_AWARE:   allows communications between unlabeled systems
154  *			  and privileged daemons
155  *	CONN_MAC_IMPLICIT: allows communications without explicit labels
156  *		           on the wire with privileged daemons.
157  *
158  * CONN_MAC_IMPLICIT is intended specifically for labeled IPsec key management
159  * in networks which don't pass CIPSO-labeled packets.
160  */
161 #define	CONN_MAC_DEFAULT 0
162 #define	CONN_MAC_AWARE 1
163 #define	CONN_MAC_IMPLICIT 2
164 
165 /*
166  * conn receive ancillary definition.
167  *
168  * These are the set of socket options that make the receive side
169  * potentially pass up ancillary data items.
170  * We have a union with an integer so that we can quickly check whether
171  * any ancillary data items need to be added.
172  */
173 typedef struct crb_s {
174 	union {
175 		uint32_t	crbu_all;
176 		struct {
177 			uint32_t
178 	crbb_recvdstaddr : 1,		/* IP_RECVDSTADDR option */
179 	crbb_recvopts : 1,		/* IP_RECVOPTS option */
180 	crbb_recvif : 1,		/* IP_RECVIF option */
181 	crbb_recvslla : 1,		/* IP_RECVSLLA option */
182 
183 	crbb_recvttl : 1,		/* IP_RECVTTL option */
184 	crbb_ip_recvpktinfo : 1,	/* IP*_RECVPKTINFO option  */
185 	crbb_ipv6_recvhoplimit : 1,	/* IPV6_RECVHOPLIMIT option */
186 	crbb_ipv6_recvhopopts : 1,	/* IPV6_RECVHOPOPTS option */
187 
188 	crbb_ipv6_recvdstopts : 1,	/* IPV6_RECVDSTOPTS option */
189 	crbb_ipv6_recvrthdr : 1,	/* IPV6_RECVRTHDR option */
190 	crbb_old_ipv6_recvdstopts : 1,	/* old form of IPV6_DSTOPTS */
191 	crbb_ipv6_recvrthdrdstopts : 1,	/* IPV6_RECVRTHDRDSTOPTS */
192 
193 	crbb_ipv6_recvtclass : 1,	/* IPV6_RECVTCLASS */
194 	crbb_recvucred : 1,		/* IP_RECVUCRED option */
195 	crbb_timestamp : 1;		/* SO_TIMESTAMP "socket" option */
196 
197 		} crbb;
198 	} crbu;
199 } crb_t;
200 
201 #define	crb_all				crbu.crbu_all
202 #define	crb_recvdstaddr			crbu.crbb.crbb_recvdstaddr
203 #define	crb_recvopts			crbu.crbb.crbb_recvopts
204 #define	crb_recvif			crbu.crbb.crbb_recvif
205 #define	crb_recvslla			crbu.crbb.crbb_recvslla
206 #define	crb_recvttl			crbu.crbb.crbb_recvttl
207 #define	crb_ip_recvpktinfo		crbu.crbb.crbb_ip_recvpktinfo
208 #define	crb_ipv6_recvhoplimit		crbu.crbb.crbb_ipv6_recvhoplimit
209 #define	crb_ipv6_recvhopopts		crbu.crbb.crbb_ipv6_recvhopopts
210 #define	crb_ipv6_recvdstopts		crbu.crbb.crbb_ipv6_recvdstopts
211 #define	crb_ipv6_recvrthdr		crbu.crbb.crbb_ipv6_recvrthdr
212 #define	crb_old_ipv6_recvdstopts	crbu.crbb.crbb_old_ipv6_recvdstopts
213 #define	crb_ipv6_recvrthdrdstopts	crbu.crbb.crbb_ipv6_recvrthdrdstopts
214 #define	crb_ipv6_recvtclass		crbu.crbb.crbb_ipv6_recvtclass
215 #define	crb_recvucred			crbu.crbb.crbb_recvucred
216 #define	crb_timestamp			crbu.crbb.crbb_timestamp
217 
218 /*
219  * The initial fields in the conn_t are setup by the kmem_cache constructor,
220  * and are preserved when it is freed. Fields after that are bzero'ed when
221  * the conn_t is freed.
222  *
223  * Much of the conn_t is protected by conn_lock.
224  *
225  * conn_lock is also used by some ULPs (like UDP and RAWIP) to protect
226  * their state.
227  */
228 struct conn_s {
229 	kmutex_t	conn_lock;
230 	uint32_t	conn_ref;		/* Reference counter */
231 	uint32_t	conn_flags;		/* Conn Flags */
232 
233 	union {
234 		tcp_t		*cp_tcp;	/* Pointer to the tcp struct */
235 		struct udp_s	*cp_udp;	/* Pointer to the udp struct */
236 		struct icmp_s	*cp_icmp;	/* Pointer to rawip struct */
237 		struct rts_s	*cp_rts;	/* Pointer to rts struct */
238 		struct iptun_s	*cp_iptun;	/* Pointer to iptun_t */
239 		struct sctp_s	*cp_sctp;	/* For IPCL_SCTPCONN */
240 		void		*cp_priv;
241 	} conn_proto_priv;
242 #define	conn_tcp	conn_proto_priv.cp_tcp
243 #define	conn_udp	conn_proto_priv.cp_udp
244 #define	conn_icmp	conn_proto_priv.cp_icmp
245 #define	conn_rts	conn_proto_priv.cp_rts
246 #define	conn_iptun	conn_proto_priv.cp_iptun
247 #define	conn_sctp	conn_proto_priv.cp_sctp
248 #define	conn_priv	conn_proto_priv.cp_priv
249 
250 	kcondvar_t	conn_cv;
251 	uint8_t		conn_proto;		/* protocol type */
252 
253 	edesc_rpf	conn_recv;		/* Pointer to recv routine */
254 	edesc_rpf	conn_recvicmp;		/* For ICMP error */
255 	edesc_vpf	conn_verifyicmp;	/* Verify ICMP error */
256 
257 	ip_xmit_attr_t	*conn_ixa;		/* Options if no ancil data */
258 
259 	/* Fields after this are bzero'ed when the conn_t is freed. */
260 #define	conn_start_clr	conn_recv_ancillary
261 
262 	/* Options for receive-side ancillary data */
263 	crb_t		conn_recv_ancillary;
264 
265 	squeue_t	*conn_sqp;		/* Squeue for processing */
266 	uint_t		conn_state_flags;	/* IP state flags */
267 
268 	int		conn_lingertime;	/* linger time (in seconds) */
269 
270 	unsigned int
271 		conn_on_sqp : 1,		/* Conn is being processed */
272 		conn_linger : 1,		/* SO_LINGER state */
273 		conn_useloopback : 1,		/* SO_USELOOPBACK state */
274 		conn_broadcast : 1,		/* SO_BROADCAST state */
275 
276 		conn_reuseaddr : 1,		/* SO_REUSEADDR state */
277 		conn_keepalive : 1,		/* SO_KEEPALIVE state */
278 		conn_multi_router : 1,		/* Wants all multicast pkts */
279 		conn_did_putbq : 1,		/* ip_wput did a putbq */
280 
281 		conn_unspec_src : 1,		/* IP_UNSPEC_SRC */
282 		conn_policy_cached : 1,		/* Is policy cached/latched ? */
283 		conn_in_enforce_policy : 1,	/* Enforce Policy on inbound */
284 		conn_out_enforce_policy : 1,	/* Enforce Policy on outbound */
285 
286 		conn_debug : 1,			/* SO_DEBUG */
287 		conn_ipv6_v6only : 1,		/* IPV6_V6ONLY */
288 		conn_oobinline : 1, 		/* SO_OOBINLINE state */
289 		conn_dgram_errind : 1,		/* SO_DGRAM_ERRIND state */
290 
291 		conn_exclbind : 1,		/* SO_EXCLBIND state */
292 		conn_mdt_ok : 1,		/* MDT is permitted */
293 		conn_allzones : 1,		/* SO_ALLZONES */
294 		conn_ipv6_recvpathmtu : 1,	/* IPV6_RECVPATHMTU */
295 
296 		conn_mcbc_bind : 1,		/* Bound to multi/broadcast */
297 
298 		conn_pad_to_bit_31 : 11;
299 
300 	boolean_t conn_direct_blocked;		/* conn is flow-controlled */
301 
302 	squeue_t	*conn_initial_sqp;	/* Squeue at open time */
303 	squeue_t	*conn_final_sqp;	/* Squeue after connect */
304 	ill_t		*conn_dhcpinit_ill;	/* IP_DHCPINIT_IF */
305 	ipsec_latch_t	*conn_latch;		/* latched IDS */
306 	struct ipsec_policy_s	*conn_latch_in_policy; /* latched policy (in) */
307 	struct ipsec_action_s	*conn_latch_in_action; /* latched action (in) */
308 	uint_t		conn_bound_if;		/* IP*_BOUND_IF */
309 	queue_t		*conn_rq;		/* Read queue */
310 	queue_t		*conn_wq;		/* Write queue */
311 	dev_t		conn_dev;		/* Minor number */
312 	vmem_t		*conn_minor_arena;	/* Minor arena */
313 	ip_helper_stream_info_t *conn_helper_info;
314 
315 	cred_t		*conn_cred;		/* Credentials */
316 	pid_t		conn_cpid;		/* pid from open/connect */
317 	uint64_t	conn_open_time;		/* time when this was opened */
318 
319 	connf_t		*conn_g_fanout;		/* Global Hash bucket head */
320 	struct conn_s	*conn_g_next;		/* Global Hash chain next */
321 	struct conn_s	*conn_g_prev;		/* Global Hash chain prev */
322 	struct ipsec_policy_head_s *conn_policy; /* Configured policy */
323 	in6_addr_t	conn_bound_addr_v6;	/* Address in bind() */
324 #define	conn_bound_addr_v4	V4_PART_OF_V6(conn_bound_addr_v6)
325 	connf_t		*conn_fanout;		/* Hash bucket we're part of */
326 	struct conn_s	*conn_next;		/* Hash chain next */
327 	struct conn_s	*conn_prev;		/* Hash chain prev */
328 
329 	struct {
330 		in6_addr_t connua_laddr;	/* Local address - match */
331 		in6_addr_t connua_faddr;	/* Remote address */
332 	} connua_v6addr;
333 #define	conn_laddr_v4	V4_PART_OF_V6(connua_v6addr.connua_laddr)
334 #define	conn_faddr_v4	V4_PART_OF_V6(connua_v6addr.connua_faddr)
335 #define	conn_laddr_v6	connua_v6addr.connua_laddr
336 #define	conn_faddr_v6	connua_v6addr.connua_faddr
337 	in6_addr_t	conn_saddr_v6;		/* Local address - source */
338 #define	conn_saddr_v4	V4_PART_OF_V6(conn_saddr_v6)
339 
340 	union {
341 		/* Used for classifier match performance */
342 		uint32_t		connu_ports2;
343 		struct {
344 			in_port_t	connu_fport;	/* Remote port */
345 			in_port_t	connu_lport;	/* Local port */
346 		} connu_ports;
347 	} u_port;
348 #define	conn_fport	u_port.connu_ports.connu_fport
349 #define	conn_lport	u_port.connu_ports.connu_lport
350 #define	conn_ports	u_port.connu_ports2
351 
352 	uint_t		conn_incoming_ifindex;	/* IP{,V6}_BOUND_IF, scopeid */
353 	ill_t		*conn_oper_pending_ill; /* pending shared ioctl */
354 
355 	krwlock_t	conn_ilg_lock;		/* Protects conn_ilg_* */
356 	ilg_t		*conn_ilg;		/* Group memberships */
357 
358 	kcondvar_t	conn_refcv;		/* For conn_oper_pending_ill */
359 
360 	struct conn_s 	*conn_drain_next;	/* Next conn in drain list */
361 	struct conn_s	*conn_drain_prev;	/* Prev conn in drain list */
362 	idl_t		*conn_idl;		/* Ptr to the drain list head */
363 	mblk_t		*conn_ipsec_opt_mp;	/* ipsec option mblk */
364 	zoneid_t	conn_zoneid;		/* zone connection is in */
365 	int		conn_rtaware; 		/* RT_AWARE sockopt value */
366 	kcondvar_t	conn_sq_cv;		/* For non-STREAMS socket IO */
367 	sock_upcalls_t	*conn_upcalls;		/* Upcalls to sockfs */
368 	sock_upper_handle_t conn_upper_handle;	/* Upper handle: sonode * */
369 
370 	unsigned int
371 		conn_mlp_type : 2,		/* mlp_type_t; tsol/tndb.h */
372 		conn_anon_mlp : 1,		/* user wants anon MLP */
373 		conn_anon_port : 1,		/* user bound anonymously */
374 
375 		conn_mac_mode : 2,		/* normal/loose/implicit MAC */
376 		conn_anon_priv_bind : 1,	/* *_ANON_PRIV_BIND state */
377 		conn_zone_is_global : 1,	/* GLOBAL_ZONEID */
378 		conn_isvrrp : 1,		/* VRRP control socket */
379 		conn_spare : 23;
380 
381 	boolean_t	conn_flow_cntrld;
382 	netstack_t	*conn_netstack;	/* Corresponds to a netstack_hold */
383 
384 	/*
385 	 * IP format that packets received for this struct should use.
386 	 * Value can be IP4_VERSION or IPV6_VERSION.
387 	 * The sending version is encoded using IXAF_IS_IPV4.
388 	 */
389 	ushort_t	conn_ipversion;
390 
391 	/* Written to only once at the time of opening the endpoint */
392 	sa_family_t	conn_family;		/* Family from socket() call */
393 	uint_t		conn_so_type;		/* Type from socket() call */
394 
395 	uint_t		conn_sndbuf;		/* SO_SNDBUF state */
396 	uint_t		conn_rcvbuf;		/* SO_RCVBUF state */
397 	uint_t		conn_wroff;		/* Current write offset */
398 
399 	uint_t		conn_sndlowat;		/* Send buffer low water mark */
400 	uint_t		conn_rcvlowat;		/* Recv buffer low water mark */
401 
402 	uint8_t		conn_default_ttl;	/* Default TTL/hoplimit */
403 
404 	uint32_t	conn_flowinfo;	/* Connected flow id and tclass */
405 
406 	/*
407 	 * The most recent address for sendto. Initially set to zero
408 	 * which is always different than then the destination address
409 	 * since the send interprets zero as the loopback address.
410 	 */
411 	in6_addr_t	conn_v6lastdst;
412 #define	conn_v4lastdst	V4_PART_OF_V6(conn_v6lastdst)
413 	ushort_t	conn_lastipversion;
414 	in_port_t	conn_lastdstport;
415 	uint32_t	conn_lastflowinfo;	/* IPv6-only */
416 	uint_t		conn_lastscopeid;	/* IPv6-only */
417 	uint_t		conn_lastsrcid;		/* Only for AF_INET6 */
418 	/*
419 	 * When we are not connected conn_saddr might be unspecified.
420 	 * We track the source that was used with conn_v6lastdst here.
421 	 */
422 	in6_addr_t	conn_v6lastsrc;
423 #define	conn_v4lastsrc	V4_PART_OF_V6(conn_v6lastsrc)
424 
425 	/* Templates for transmitting packets */
426 	ip_pkt_t	conn_xmit_ipp;		/* Options if no ancil data */
427 
428 	/*
429 	 * Header template - conn_ht_ulp is a pointer into conn_ht_iphc.
430 	 * Note that ixa_ip_hdr_length indicates the offset of ht_ulp in
431 	 * ht_iphc
432 	 *
433 	 * The header template is maintained for connected endpoints (and
434 	 * updated when sticky options are changed) and also for the lastdst.
435 	 * There is no conflict between those usages since SOCK_DGRAM and
436 	 * SOCK_RAW can not be used to specify a destination address (with
437 	 * sendto/sendmsg) if the socket has been connected.
438 	 */
439 	uint8_t		*conn_ht_iphc;		/* Start of IP header */
440 	uint_t		conn_ht_iphc_allocated;	/* Allocated buffer size */
441 	uint_t		conn_ht_iphc_len;	/* IP+ULP size */
442 	uint8_t		*conn_ht_ulp;		/* Upper-layer header */
443 	uint_t		conn_ht_ulp_len;	/* ULP header len */
444 
445 	/* Checksum to compensate for source routed packets. Host byte order */
446 	uint32_t	conn_sum;
447 
448 	uint32_t	conn_ioctlref;		/* ioctl ref count */
449 #ifdef CONN_DEBUG
450 #define	CONN_TRACE_MAX	10
451 	int		conn_trace_last;	/* ndx of last used tracebuf */
452 	conn_trace_t	conn_trace_buf[CONN_TRACE_MAX];
453 #endif
454 };
455 
456 /*
457  * connf_t - connection fanout data.
458  *
459  * The hash tables and their linkage (conn_t.{hashnextp, hashprevp} are
460  * protected by the per-bucket lock. Each conn_t inserted in the list
461  * points back at the connf_t that heads the bucket.
462  */
463 struct connf_s {
464 	struct conn_s	*connf_head;
465 	kmutex_t	connf_lock;
466 };
467 
468 #define	CONN_INC_REF(connp)	{				\
469 	mutex_enter(&(connp)->conn_lock);			\
470 	DTRACE_PROBE1(conn__inc__ref, conn_t *, connp);		\
471 	ASSERT(conn_trace_ref(connp));				\
472 	(connp)->conn_ref++;					\
473 	ASSERT((connp)->conn_ref != 0);				\
474 	mutex_exit(&(connp)->conn_lock);			\
475 }
476 
477 #define	CONN_INC_REF_LOCKED(connp)	{			\
478 	DTRACE_PROBE1(conn__inc__ref, conn_t *, connp);		\
479 	ASSERT(MUTEX_HELD(&(connp)->conn_lock));	 	\
480 	ASSERT(conn_trace_ref(connp));				\
481 	(connp)->conn_ref++;					\
482 	ASSERT((connp)->conn_ref != 0);				\
483 }
484 
485 #define	CONN_DEC_REF(connp)	{					\
486 	mutex_enter(&(connp)->conn_lock);				\
487 	DTRACE_PROBE1(conn__dec__ref, conn_t *, connp);			\
488 	/*								\
489 	 * The squeue framework always does a CONN_DEC_REF after return	\
490 	 * from TCP. Hence the refcnt must be at least 2 if conn_on_sqp	\
491 	 * is B_TRUE and conn_ref is being decremented. This is to	\
492 	 * account for the mblk being currently processed.		\
493 	 */								\
494 	if ((connp)->conn_ref == 0 ||					\
495 	    ((connp)->conn_ref == 1 && (connp)->conn_on_sqp))		\
496 		cmn_err(CE_PANIC, "CONN_DEC_REF: connp(%p) has ref "	\
497 			"= %d\n", (void *)(connp), (connp)->conn_ref);	\
498 	ASSERT(conn_untrace_ref(connp));				\
499 	(connp)->conn_ref--;						\
500 	if ((connp)->conn_ref == 0) {					\
501 		/* Refcnt can't increase again, safe to drop lock */	\
502 		mutex_exit(&(connp)->conn_lock);			\
503 		ipcl_conn_destroy(connp);				\
504 	} else {							\
505 		cv_broadcast(&(connp)->conn_cv);			\
506 		mutex_exit(&(connp)->conn_lock);			\
507 	}								\
508 }
509 
510 /*
511  * For use with subsystems within ip which use ALL_ZONES as a wildcard
512  */
513 #define	IPCL_ZONEID(connp)						\
514 	((connp)->conn_allzones ? ALL_ZONES : (connp)->conn_zoneid)
515 
516 /*
517  * For matching between a conn_t and a zoneid.
518  */
519 #define	IPCL_ZONE_MATCH(connp, zoneid) 					\
520 	(((connp)->conn_allzones) ||					\
521 	    ((zoneid) == ALL_ZONES) ||					\
522 	    (connp)->conn_zoneid == (zoneid))
523 
524 /*
525  * On a labeled system, we must treat bindings to ports
526  * on shared IP addresses by sockets with MAC exemption
527  * privilege as being in all zones, as there's
528  * otherwise no way to identify the right receiver.
529  */
530 
531 #define	IPCL_CONNS_MAC(conn1, conn2)					\
532 	(((conn1)->conn_mac_mode != CONN_MAC_DEFAULT) ||		\
533 	((conn2)->conn_mac_mode != CONN_MAC_DEFAULT))
534 
535 #define	IPCL_BIND_ZONE_MATCH(conn1, conn2)				\
536 	(IPCL_CONNS_MAC(conn1, conn2) ||				\
537 	IPCL_ZONE_MATCH(conn1, conn2->conn_zoneid) ||			\
538 	IPCL_ZONE_MATCH(conn2, conn1->conn_zoneid))
539 
540 
541 #define	_IPCL_V4_MATCH(v6addr, v4addr)	\
542 	(V4_PART_OF_V6((v6addr)) == (v4addr) && IN6_IS_ADDR_V4MAPPED(&(v6addr)))
543 
544 #define	_IPCL_V4_MATCH_ANY(addr)	\
545 	(IN6_IS_ADDR_V4MAPPED_ANY(&(addr)) || IN6_IS_ADDR_UNSPECIFIED(&(addr)))
546 
547 
548 /*
549  * IPCL_PROTO_MATCH() and IPCL_PROTO_MATCH_V6() only matches conns with
550  * the specified ira_zoneid or conn_allzones by calling conn_wantpacket.
551  */
552 #define	IPCL_PROTO_MATCH(connp, ira, ipha)				\
553 	((((connp)->conn_laddr_v4 == INADDR_ANY) ||			\
554 	(((connp)->conn_laddr_v4 == ((ipha)->ipha_dst)) &&		\
555 	    (((connp)->conn_faddr_v4 == INADDR_ANY) ||			\
556 	((connp)->conn_faddr_v4 == ((ipha)->ipha_src))))) &&		\
557 	conn_wantpacket((connp), (ira), (ipha)))
558 
559 #define	IPCL_PROTO_MATCH_V6(connp, ira, ip6h)				\
560 	((IN6_IS_ADDR_UNSPECIFIED(&(connp)->conn_laddr_v6) ||		\
561 	(IN6_ARE_ADDR_EQUAL(&(connp)->conn_laddr_v6, &((ip6h)->ip6_dst)) &&   \
562 	(IN6_IS_ADDR_UNSPECIFIED(&(connp)->conn_faddr_v6) ||		      \
563 	IN6_ARE_ADDR_EQUAL(&(connp)->conn_faddr_v6, &((ip6h)->ip6_src))))) && \
564 	(conn_wantpacket_v6((connp), (ira), (ip6h))))
565 
566 #define	IPCL_CONN_HASH(src, ports, ipst)				\
567 	((unsigned)(ntohl((src)) ^ ((ports) >> 24) ^ ((ports) >> 16) ^	\
568 	((ports) >> 8) ^ (ports)) % (ipst)->ips_ipcl_conn_fanout_size)
569 
570 #define	IPCL_CONN_HASH_V6(src, ports, ipst)				\
571 	IPCL_CONN_HASH(V4_PART_OF_V6((src)), (ports), (ipst))
572 
573 #define	IPCL_CONN_MATCH(connp, proto, src, dst, ports)			\
574 	((connp)->conn_proto == (proto) &&				\
575 		(connp)->conn_ports == (ports) &&      			\
576 		_IPCL_V4_MATCH((connp)->conn_faddr_v6, (src)) &&	\
577 		_IPCL_V4_MATCH((connp)->conn_laddr_v6, (dst)) &&	\
578 		!(connp)->conn_ipv6_v6only)
579 
580 #define	IPCL_CONN_MATCH_V6(connp, proto, src, dst, ports)		\
581 	((connp)->conn_proto == (proto) &&				\
582 		(connp)->conn_ports == (ports) &&      			\
583 		IN6_ARE_ADDR_EQUAL(&(connp)->conn_faddr_v6, &(src)) &&	\
584 		IN6_ARE_ADDR_EQUAL(&(connp)->conn_laddr_v6, &(dst)))
585 
586 #define	IPCL_PORT_HASH(port, size) \
587 	((((port) >> 8) ^ (port)) & ((size) - 1))
588 
589 #define	IPCL_BIND_HASH(lport, ipst)					\
590 	((unsigned)(((lport) >> 8) ^ (lport)) % \
591 	    (ipst)->ips_ipcl_bind_fanout_size)
592 
593 #define	IPCL_BIND_MATCH(connp, proto, laddr, lport)			\
594 	((connp)->conn_proto == (proto) &&				\
595 		(connp)->conn_lport == (lport) &&			\
596 		(_IPCL_V4_MATCH_ANY((connp)->conn_laddr_v6) ||		\
597 		_IPCL_V4_MATCH((connp)->conn_laddr_v6, (laddr))) &&	\
598 		!(connp)->conn_ipv6_v6only)
599 
600 #define	IPCL_BIND_MATCH_V6(connp, proto, laddr, lport)			\
601 	((connp)->conn_proto == (proto) &&				\
602 		(connp)->conn_lport == (lport) &&			\
603 		(IN6_ARE_ADDR_EQUAL(&(connp)->conn_laddr_v6, &(laddr)) || \
604 		IN6_IS_ADDR_UNSPECIFIED(&(connp)->conn_laddr_v6)))
605 
606 /*
607  * We compare conn_laddr since it captures both connected and a bind to
608  * a multicast or broadcast address.
609  * The caller needs to match the zoneid and also call conn_wantpacket
610  * for multicast, broadcast, or when conn_incoming_ifindex is set.
611  */
612 #define	IPCL_UDP_MATCH(connp, lport, laddr, fport, faddr)		\
613 	(((connp)->conn_lport == (lport)) &&				\
614 	((_IPCL_V4_MATCH_ANY((connp)->conn_laddr_v6) ||			\
615 	(_IPCL_V4_MATCH((connp)->conn_laddr_v6, (laddr)) &&		\
616 	(_IPCL_V4_MATCH_ANY((connp)->conn_faddr_v6) ||			\
617 	(_IPCL_V4_MATCH((connp)->conn_faddr_v6, (faddr)) &&		\
618 	(connp)->conn_fport == (fport)))))) &&				\
619 	!(connp)->conn_ipv6_v6only)
620 
621 /*
622  * We compare conn_laddr since it captures both connected and a bind to
623  * a multicast or broadcast address.
624  * The caller needs to match the zoneid and also call conn_wantpacket_v6
625  * for multicast or when conn_incoming_ifindex is set.
626  */
627 #define	IPCL_UDP_MATCH_V6(connp, lport, laddr, fport, faddr)	\
628 	(((connp)->conn_lport == (lport)) &&			\
629 	(IN6_IS_ADDR_UNSPECIFIED(&(connp)->conn_laddr_v6) ||	\
630 	(IN6_ARE_ADDR_EQUAL(&(connp)->conn_laddr_v6, &(laddr)) &&	\
631 	(IN6_IS_ADDR_UNSPECIFIED(&(connp)->conn_faddr_v6) ||	\
632 	(IN6_ARE_ADDR_EQUAL(&(connp)->conn_faddr_v6, &(faddr)) &&	\
633 	(connp)->conn_fport == (fport))))))
634 
635 #define	IPCL_IPTUN_HASH(laddr, faddr)					\
636 	((ntohl(laddr) ^ ((ntohl(faddr) << 24) | (ntohl(faddr) >> 8))) % \
637 	ipcl_iptun_fanout_size)
638 
639 #define	IPCL_IPTUN_HASH_V6(laddr, faddr)				\
640 	IPCL_IPTUN_HASH((laddr)->s6_addr32[0] ^ (laddr)->s6_addr32[1] ^	\
641 	    (faddr)->s6_addr32[2] ^ (faddr)->s6_addr32[3],		\
642 	    (faddr)->s6_addr32[0] ^ (faddr)->s6_addr32[1] ^		\
643 	    (laddr)->s6_addr32[2] ^ (laddr)->s6_addr32[3])
644 
645 #define	IPCL_IPTUN_MATCH(connp, laddr, faddr)			\
646 	(_IPCL_V4_MATCH((connp)->conn_laddr_v6, (laddr)) &&	\
647 	_IPCL_V4_MATCH((connp)->conn_faddr_v6, (faddr)))
648 
649 #define	IPCL_IPTUN_MATCH_V6(connp, laddr, faddr)		\
650 	(IN6_ARE_ADDR_EQUAL(&(connp)->conn_laddr_v6, (laddr)) &&	\
651 	IN6_ARE_ADDR_EQUAL(&(connp)->conn_faddr_v6, (faddr)))
652 
653 #define	IPCL_UDP_HASH(lport, ipst)	\
654 	IPCL_PORT_HASH(lport, (ipst)->ips_ipcl_udp_fanout_size)
655 
656 #define	CONN_G_HASH_SIZE	1024
657 
658 /* Raw socket hash function. */
659 #define	IPCL_RAW_HASH(lport, ipst)	\
660 	IPCL_PORT_HASH(lport, (ipst)->ips_ipcl_raw_fanout_size)
661 
662 /*
663  * This is similar to IPCL_BIND_MATCH except that the local port check
664  * is changed to a wildcard port check.
665  * We compare conn_laddr since it captures both connected and a bind to
666  * a multicast or broadcast address.
667  */
668 #define	IPCL_RAW_MATCH(connp, proto, laddr)			\
669 	((connp)->conn_proto == (proto) &&			\
670 	(connp)->conn_lport == 0 &&				\
671 	(_IPCL_V4_MATCH_ANY((connp)->conn_laddr_v6) ||		\
672 	_IPCL_V4_MATCH((connp)->conn_laddr_v6, (laddr))))
673 
674 #define	IPCL_RAW_MATCH_V6(connp, proto, laddr)			\
675 	((connp)->conn_proto == (proto) &&			\
676 	(connp)->conn_lport == 0 &&				\
677 	(IN6_IS_ADDR_UNSPECIFIED(&(connp)->conn_laddr_v6) ||	\
678 	IN6_ARE_ADDR_EQUAL(&(connp)->conn_laddr_v6, &(laddr))))
679 
680 /* Function prototypes */
681 extern void ipcl_g_init(void);
682 extern void ipcl_init(ip_stack_t *);
683 extern void ipcl_g_destroy(void);
684 extern void ipcl_destroy(ip_stack_t *);
685 extern conn_t *ipcl_conn_create(uint32_t, int, netstack_t *);
686 extern void ipcl_conn_destroy(conn_t *);
687 
688 void ipcl_hash_insert_wildcard(connf_t *, conn_t *);
689 void ipcl_hash_remove(conn_t *);
690 void ipcl_hash_remove_locked(conn_t *connp, connf_t *connfp);
691 
692 extern int	ipcl_bind_insert(conn_t *);
693 extern int	ipcl_bind_insert_v4(conn_t *);
694 extern int	ipcl_bind_insert_v6(conn_t *);
695 extern int	ipcl_conn_insert(conn_t *);
696 extern int	ipcl_conn_insert_v4(conn_t *);
697 extern int	ipcl_conn_insert_v6(conn_t *);
698 extern conn_t	*ipcl_get_next_conn(connf_t *, conn_t *, uint32_t);
699 
700 conn_t *ipcl_classify_v4(mblk_t *, uint8_t, uint_t, ip_recv_attr_t *,
701 	    ip_stack_t *);
702 conn_t *ipcl_classify_v6(mblk_t *, uint8_t, uint_t, ip_recv_attr_t *,
703 	    ip_stack_t *);
704 conn_t *ipcl_classify(mblk_t *, ip_recv_attr_t *, ip_stack_t *);
705 conn_t *ipcl_classify_raw(mblk_t *, uint8_t, uint32_t, ipha_t *,
706     ip6_t *, ip_recv_attr_t *, ip_stack_t *);
707 conn_t *ipcl_iptun_classify_v4(ipaddr_t *, ipaddr_t *, ip_stack_t *);
708 conn_t *ipcl_iptun_classify_v6(in6_addr_t *, in6_addr_t *, ip_stack_t *);
709 void	ipcl_globalhash_insert(conn_t *);
710 void	ipcl_globalhash_remove(conn_t *);
711 void	ipcl_walk(pfv_t, void *, ip_stack_t *);
712 conn_t	*ipcl_tcp_lookup_reversed_ipv4(ipha_t *, tcpha_t *, int, ip_stack_t *);
713 conn_t	*ipcl_tcp_lookup_reversed_ipv6(ip6_t *, tcpha_t *, int, uint_t,
714 	    ip_stack_t *);
715 conn_t	*ipcl_lookup_listener_v4(uint16_t, ipaddr_t, zoneid_t, ip_stack_t *);
716 conn_t	*ipcl_lookup_listener_v6(uint16_t, in6_addr_t *, uint_t, zoneid_t,
717 	    ip_stack_t *);
718 int	conn_trace_ref(conn_t *);
719 int	conn_untrace_ref(conn_t *);
720 void	ipcl_conn_cleanup(conn_t *);
721 extern uint_t	conn_recvancillary_size(conn_t *, crb_t, ip_recv_attr_t *,
722     mblk_t *, ip_pkt_t *);
723 extern void	conn_recvancillary_add(conn_t *, crb_t, ip_recv_attr_t *,
724     ip_pkt_t *, uchar_t *, uint_t);
725 conn_t *ipcl_conn_tcp_lookup_reversed_ipv4(conn_t *, ipha_t *, tcpha_t *,
726 	    ip_stack_t *);
727 conn_t *ipcl_conn_tcp_lookup_reversed_ipv6(conn_t *, ip6_t *, tcpha_t *,
728 	    ip_stack_t *);
729 
730 extern int ip_create_helper_stream(conn_t *, ldi_ident_t);
731 extern void ip_free_helper_stream(conn_t *);
732 extern int	ip_helper_stream_setup(queue_t *, dev_t *, int, int,
733     cred_t *, boolean_t);
734 
735 #ifdef	__cplusplus
736 }
737 #endif
738 
739 #endif	/* _INET_IPCLASSIFIER_H */
740