xref: /illumos-gate/usr/src/uts/common/inet/ip.h (revision 63360950109af2ce85a962ca61f40b8782f11100)
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 /*
23  * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 /* Copyright (c) 1990 Mentat Inc. */
27 
28 #ifndef	_INET_IP_H
29 #define	_INET_IP_H
30 
31 #pragma ident	"%Z%%M%	%I%	%E% SMI"
32 
33 #ifdef	__cplusplus
34 extern "C" {
35 #endif
36 
37 #include <sys/isa_defs.h>
38 #include <sys/types.h>
39 #include <inet/mib2.h>
40 #include <inet/nd.h>
41 #include <sys/atomic.h>
42 #include <net/if_dl.h>
43 #include <net/if.h>
44 #include <netinet/ip.h>
45 #include <netinet/igmp.h>
46 #include <sys/neti.h>
47 #include <sys/hook.h>
48 #include <sys/hook_event.h>
49 #include <sys/hook_impl.h>
50 #include <inet/ip_stack.h>
51 
52 #ifdef _KERNEL
53 #include <netinet/ip6.h>
54 #include <sys/avl.h>
55 #include <sys/vmem.h>
56 #include <sys/squeue.h>
57 #include <net/route.h>
58 #include <sys/systm.h>
59 #include <sys/multidata.h>
60 #include <net/radix.h>
61 
62 #ifdef DEBUG
63 #define	ILL_DEBUG
64 #define	IRE_DEBUG
65 #define	NCE_DEBUG
66 #define	CONN_DEBUG
67 #endif
68 
69 #define	IP_DEBUG
70 /*
71  * The mt-streams(9F) flags for the IP module; put here so that other
72  * "drivers" that are actually IP (e.g., ICMP, UDP) can use the same set
73  * of flags.
74  */
75 #define	IP_DEVMTFLAGS D_MP
76 #endif	/* _KERNEL */
77 
78 #define	IP_MOD_NAME	"ip"
79 #define	IP_DEV_NAME	"/dev/ip"
80 #define	IP6_DEV_NAME	"/dev/ip6"
81 
82 #define	UDP_MOD_NAME	"udp"
83 #define	UDP_DEV_NAME	"/dev/udp"
84 #define	UDP6_DEV_NAME	"/dev/udp6"
85 
86 #define	TCP_MOD_NAME	"tcp"
87 #define	TCP_DEV_NAME	"/dev/tcp"
88 #define	TCP6_DEV_NAME	"/dev/tcp6"
89 
90 #define	SCTP_MOD_NAME	"sctp"
91 
92 /* Minor numbers */
93 #define	IPV4_MINOR	0
94 #define	IPV6_MINOR	1
95 #define	TCP_MINOR	2
96 #define	TCP_MINOR6	3
97 
98 #ifndef	_IPADDR_T
99 #define	_IPADDR_T
100 typedef uint32_t ipaddr_t;
101 #endif
102 
103 /* Number of bits in an address */
104 #define	IP_ABITS		32
105 #define	IPV6_ABITS		128
106 
107 #define	IP_HOST_MASK		(ipaddr_t)0xffffffffU
108 
109 #define	IP_CSUM(mp, off, sum)		(~ip_cksum(mp, off, sum) & 0xFFFF)
110 #define	IP_CSUM_PARTIAL(mp, off, sum)	ip_cksum(mp, off, sum)
111 #define	IP_BCSUM_PARTIAL(bp, len, sum)	bcksum(bp, len, sum)
112 #define	IP_MD_CSUM(pd, off, sum)	(~ip_md_cksum(pd, off, sum) & 0xffff)
113 #define	IP_MD_CSUM_PARTIAL(pd, off, sum) ip_md_cksum(pd, off, sum)
114 
115 /*
116  * Flag to IP write side to indicate that the appln has sent in a pre-built
117  * IP header. Stored in ipha_ident (which is otherwise zero).
118  */
119 #define	IP_HDR_INCLUDED			0xFFFF
120 
121 #define	ILL_FRAG_HASH_TBL_COUNT	((unsigned int)64)
122 #define	ILL_FRAG_HASH_TBL_SIZE	(ILL_FRAG_HASH_TBL_COUNT * sizeof (ipfb_t))
123 
124 #define	IPV4_ADDR_LEN			4
125 #define	IP_ADDR_LEN			IPV4_ADDR_LEN
126 #define	IP_ARP_PROTO_TYPE		0x0800
127 
128 #define	IPV4_VERSION			4
129 #define	IP_VERSION			IPV4_VERSION
130 #define	IP_SIMPLE_HDR_LENGTH_IN_WORDS	5
131 #define	IP_SIMPLE_HDR_LENGTH		20
132 #define	IP_MAX_HDR_LENGTH		60
133 
134 #define	IP_MAX_OPT_LENGTH (IP_MAX_HDR_LENGTH-IP_SIMPLE_HDR_LENGTH)
135 
136 #define	IP_MIN_MTU			(IP_MAX_HDR_LENGTH + 8)	/* 68 bytes */
137 
138 /*
139  * XXX IP_MAXPACKET is defined in <netinet/ip.h> as well. At some point the
140  * 2 files should be cleaned up to remove all redundant definitions.
141  */
142 #define	IP_MAXPACKET			65535
143 #define	IP_SIMPLE_HDR_VERSION \
144 	((IP_VERSION << 4) | IP_SIMPLE_HDR_LENGTH_IN_WORDS)
145 
146 #define	UDPH_SIZE			8
147 
148 /* Leave room for ip_newroute to tack on the src and target addresses */
149 #define	OK_RESOLVER_MP(mp)						\
150 	((mp) && ((mp)->b_wptr - (mp)->b_rptr) >= (2 * IP_ADDR_LEN))
151 
152 /*
153  * Constants and type definitions to support IP IOCTL commands
154  */
155 #define	IP_IOCTL			(('i'<<8)|'p')
156 #define	IP_IOC_IRE_DELETE		4
157 #define	IP_IOC_IRE_DELETE_NO_REPLY	5
158 #define	IP_IOC_IRE_ADVISE_NO_REPLY	6
159 #define	IP_IOC_RTS_REQUEST		7
160 
161 /* Common definitions used by IP IOCTL data structures */
162 typedef struct ipllcmd_s {
163 	uint_t	ipllc_cmd;
164 	uint_t	ipllc_name_offset;
165 	uint_t	ipllc_name_length;
166 } ipllc_t;
167 
168 /* IP IRE Change Command Structure. */
169 typedef struct ipic_s {
170 	ipllc_t	ipic_ipllc;
171 	uint_t	ipic_ire_type;
172 	uint_t	ipic_max_frag;
173 	uint_t	ipic_addr_offset;
174 	uint_t	ipic_addr_length;
175 	uint_t	ipic_mask_offset;
176 	uint_t	ipic_mask_length;
177 	uint_t	ipic_src_addr_offset;
178 	uint_t	ipic_src_addr_length;
179 	uint_t	ipic_ll_hdr_offset;
180 	uint_t	ipic_ll_hdr_length;
181 	uint_t	ipic_gateway_addr_offset;
182 	uint_t	ipic_gateway_addr_length;
183 	clock_t	ipic_rtt;
184 	uint32_t ipic_ssthresh;
185 	clock_t	ipic_rtt_sd;
186 	uchar_t ipic_ire_marks;
187 } ipic_t;
188 
189 #define	ipic_cmd		ipic_ipllc.ipllc_cmd
190 #define	ipic_ll_name_length	ipic_ipllc.ipllc_name_length
191 #define	ipic_ll_name_offset	ipic_ipllc.ipllc_name_offset
192 
193 /* IP IRE Delete Command Structure. */
194 typedef struct ipid_s {
195 	ipllc_t	ipid_ipllc;
196 	uint_t	ipid_ire_type;
197 	uint_t	ipid_addr_offset;
198 	uint_t	ipid_addr_length;
199 	uint_t	ipid_mask_offset;
200 	uint_t	ipid_mask_length;
201 } ipid_t;
202 
203 #define	ipid_cmd		ipid_ipllc.ipllc_cmd
204 
205 #ifdef _KERNEL
206 /*
207  * Temporary state for ip options parser.
208  */
209 typedef struct ipoptp_s
210 {
211 	uint8_t		*ipoptp_next;	/* next option to look at */
212 	uint8_t		*ipoptp_end;	/* end of options */
213 	uint8_t		*ipoptp_cur;	/* start of current option */
214 	uint8_t		ipoptp_len;	/* length of current option */
215 	uint32_t	ipoptp_flags;
216 } ipoptp_t;
217 
218 /*
219  * Flag(s) for ipoptp_flags
220  */
221 #define	IPOPTP_ERROR	0x00000001
222 #endif	/* _KERNEL */
223 
224 
225 /* Controls forwarding of IP packets, set via ndd */
226 #define	IP_FORWARD_NEVER	0
227 #define	IP_FORWARD_ALWAYS	1
228 
229 #define	WE_ARE_FORWARDING(ipst)	((ipst)->ips_ip_g_forward == IP_FORWARD_ALWAYS)
230 
231 #define	IPH_HDR_LENGTH(ipha)						\
232 	((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length & 0xF) << 2)
233 
234 #define	IPH_HDR_VERSION(ipha)						\
235 	((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length) >> 4)
236 
237 #ifdef _KERNEL
238 /*
239  * IP reassembly macros.  We hide starting and ending offsets in b_next and
240  * b_prev of messages on the reassembly queue.	The messages are chained using
241  * b_cont.  These macros are used in ip_reassemble() so we don't have to see
242  * the ugly casts and assignments.
243  * Note that the offsets are <= 64k i.e. a uint_t is sufficient to represent
244  * them.
245  */
246 #define	IP_REASS_START(mp)		((uint_t)(uintptr_t)((mp)->b_next))
247 #define	IP_REASS_SET_START(mp, u)	\
248 	((mp)->b_next = (mblk_t *)(uintptr_t)(u))
249 #define	IP_REASS_END(mp)		((uint_t)(uintptr_t)((mp)->b_prev))
250 #define	IP_REASS_SET_END(mp, u)		\
251 	((mp)->b_prev = (mblk_t *)(uintptr_t)(u))
252 
253 #define	IP_REASS_COMPLETE	0x1
254 #define	IP_REASS_PARTIAL	0x2
255 #define	IP_REASS_FAILED		0x4
256 
257 /*
258  * Test to determine whether this is a module instance of IP or a
259  * driver instance of IP.
260  */
261 #define	CONN_Q(q)	(WR(q)->q_next == NULL)
262 
263 #define	Q_TO_CONN(q)	((conn_t *)(q)->q_ptr)
264 #define	Q_TO_TCP(q)	(Q_TO_CONN((q))->conn_tcp)
265 #define	Q_TO_UDP(q)	(Q_TO_CONN((q))->conn_udp)
266 
267 /*
268  * The following two macros are used by IP to get the appropriate
269  * wq and rq for a conn. If it is a TCP conn, then we need
270  * tcp_wq/tcp_rq else, conn_wq/conn_rq. IP can use conn_wq and conn_rq
271  * from a conn directly if it knows that the conn is not TCP.
272  */
273 #define	CONNP_TO_WQ(connp)	\
274 	(IPCL_IS_TCP(connp) ? (connp)->conn_tcp->tcp_wq : (connp)->conn_wq)
275 
276 #define	CONNP_TO_RQ(connp)	RD(CONNP_TO_WQ(connp))
277 
278 #define	GRAB_CONN_LOCK(q)	{				\
279 	if (q != NULL && CONN_Q(q))				\
280 		mutex_enter(&(Q_TO_CONN(q))->conn_lock);	\
281 }
282 
283 #define	RELEASE_CONN_LOCK(q)	{				\
284 	if (q != NULL && CONN_Q(q))				\
285 		mutex_exit(&(Q_TO_CONN(q))->conn_lock);		\
286 }
287 
288 /* "Congestion controlled" protocol */
289 #define	IP_FLOW_CONTROLLED_ULP(p)   ((p) == IPPROTO_TCP || (p) == IPPROTO_SCTP)
290 
291 /*
292  * Complete the pending operation. Usually an ioctl. Can also
293  * be a bind or option management request that got enqueued
294  * in an ipsq_t. Called on completion of the operation.
295  */
296 #define	CONN_OPER_PENDING_DONE(connp)	{			\
297 	mutex_enter(&(connp)->conn_lock);			\
298 	(connp)->conn_oper_pending_ill = NULL;			\
299 	cv_broadcast(&(connp)->conn_refcv);			\
300 	mutex_exit(&(connp)->conn_lock);			\
301 	CONN_DEC_REF(connp);					\
302 }
303 
304 /* Get the credential of an IP queue of unknown type */
305 #define	GET_QUEUE_CRED(wq)						\
306 	((wq)->q_next ? (((ill_t *)(wq)->q_ptr)->ill_credp) \
307 	    : ((Q_TO_CONN((wq)))->conn_cred))
308 
309 /*
310  * Flags for the various ip_fanout_* routines.
311  */
312 #define	IP_FF_SEND_ICMP		0x01	/* Send an ICMP error */
313 #define	IP_FF_HDR_COMPLETE	0x02	/* Call ip_hdr_complete if error */
314 #define	IP_FF_CKSUM		0x04	/* Recompute ipha_cksum if error */
315 #define	IP_FF_RAWIP		0x08	/* Use rawip mib variable */
316 #define	IP_FF_SRC_QUENCH	0x10	/* OK to send ICMP_SOURCE_QUENCH */
317 #define	IP_FF_SYN_ADDIRE	0x20	/* Add IRE if TCP syn packet */
318 #define	IP_FF_IPINFO		0x80	/* Used for both V4 and V6 */
319 #define	IP_FF_SEND_SLLA		0x100	/* Send source link layer info ? */
320 #define	IPV6_REACHABILITY_CONFIRMATION	0x200	/* Flags for ip_xmit_v6 */
321 #define	IP_FF_NO_MCAST_LOOP	0x400	/* No multicasts for sending zone */
322 
323 /*
324  * Following flags are used by IPQoS to determine if policy processing is
325  * required.
326  */
327 #define	IP6_NO_IPPOLICY		0x800	/* Don't do IPQoS processing */
328 #define	IP6_IN_LLMCAST		0x1000	/* Multicast */
329 
330 #define	IP_FF_LOOPBACK		0x2000	/* Loopback fanout */
331 
332 #ifndef	IRE_DB_TYPE
333 #define	IRE_DB_TYPE	M_SIG
334 #endif
335 
336 #ifndef	IRE_DB_REQ_TYPE
337 #define	IRE_DB_REQ_TYPE	M_PCSIG
338 #endif
339 
340 #ifndef	IRE_ARPRESOLVE_TYPE
341 #define	IRE_ARPRESOLVE_TYPE	M_EVENT
342 #endif
343 
344 /*
345  * Values for squeue switch:
346  */
347 
348 #define	IP_SQUEUE_ENTER_NODRAIN	1
349 #define	IP_SQUEUE_ENTER	2
350 /*
351  * This is part of the interface between Transport provider and
352  * IP which can be used to set policy information. This is usually
353  * accompanied with O_T_BIND_REQ/T_BIND_REQ.ip_bind assumes that
354  * only IPSEC_POLICY_SET is there when it is found in the chain.
355  * The information contained is an struct ipsec_req_t. On success
356  * or failure, either the T_BIND_ACK or the T_ERROR_ACK is returned.
357  * IPSEC_POLICY_SET is never returned.
358  */
359 #define	IPSEC_POLICY_SET	M_SETOPTS
360 
361 #define	IRE_IS_LOCAL(ire)	((ire != NULL) && \
362 				((ire)->ire_type & (IRE_LOCAL | IRE_LOOPBACK)))
363 
364 #define	IRE_IS_TARGET(ire)	((ire != NULL) && \
365 				((ire)->ire_type != IRE_BROADCAST))
366 
367 /* IP Fragmentation Reassembly Header */
368 typedef struct ipf_s {
369 	struct ipf_s	*ipf_hash_next;
370 	struct ipf_s	**ipf_ptphn;	/* Pointer to previous hash next. */
371 	uint32_t	ipf_ident;	/* Ident to match. */
372 	uint8_t		ipf_protocol;	/* Protocol to match. */
373 	uchar_t		ipf_last_frag_seen : 1;	/* Last fragment seen ? */
374 	time_t		ipf_timestamp;	/* Reassembly start time. */
375 	mblk_t		*ipf_mp;	/* mblk we live in. */
376 	mblk_t		*ipf_tail_mp;	/* Frag queue tail pointer. */
377 	int		ipf_hole_cnt;	/* Number of holes (hard-case). */
378 	int		ipf_end;	/* Tail end offset (0 -> hard-case). */
379 	uint_t		ipf_gen;	/* Frag queue generation */
380 	size_t		ipf_count;	/* Count of bytes used by frag */
381 	uint_t		ipf_nf_hdr_len; /* Length of nonfragmented header */
382 	in6_addr_t	ipf_v6src;	/* IPv6 source address */
383 	in6_addr_t	ipf_v6dst;	/* IPv6 dest address */
384 	uint_t		ipf_prev_nexthdr_offset; /* Offset for nexthdr value */
385 	uint8_t		ipf_ecn;	/* ECN info for the fragments */
386 	uint8_t		ipf_num_dups;	/* Number of times dup frags recvd */
387 	uint16_t	ipf_checksum_flags; /* Hardware checksum flags */
388 	uint32_t	ipf_checksum;	/* Partial checksum of fragment data */
389 } ipf_t;
390 
391 #define	ipf_src	V4_PART_OF_V6(ipf_v6src)
392 #define	ipf_dst	V4_PART_OF_V6(ipf_v6dst)
393 
394 typedef enum {
395 	IB_PKT =  0x01,
396 	OB_PKT = 0x02
397 } ip_pkt_t;
398 
399 #define	UPDATE_IB_PKT_COUNT(ire)\
400 	{ \
401 	(ire)->ire_ib_pkt_count++; \
402 	if ((ire)->ire_ipif != NULL) { \
403 		/* \
404 		 * forwarding packet \
405 		 */ \
406 		if ((ire)->ire_type & (IRE_LOCAL|IRE_BROADCAST)) \
407 			atomic_add_32(&(ire)->ire_ipif->ipif_ib_pkt_count, 1);\
408 		else \
409 			atomic_add_32(&(ire)->ire_ipif->ipif_fo_pkt_count, 1);\
410 	} \
411 	}
412 #define	UPDATE_OB_PKT_COUNT(ire)\
413 	{ \
414 	(ire)->ire_ob_pkt_count++;\
415 	if ((ire)->ire_ipif != NULL) { \
416 		atomic_add_32(&(ire)->ire_ipif->ipif_ob_pkt_count, 1); \
417 	} \
418 	}
419 
420 
421 #define	IP_RPUT_LOCAL(q, mp, ipha, ire, recv_ill) \
422 { \
423 	switch (ipha->ipha_protocol) { \
424 		case IPPROTO_UDP: \
425 			ip_udp_input(q, mp, ipha, ire, recv_ill); \
426 			break; \
427 		default: \
428 			ip_proto_input(q, mp, ipha, ire, recv_ill); \
429 			break; \
430 	} \
431 }
432 
433 /*
434  * NCE_EXPIRED is TRUE when we have a non-permanent nce that was
435  * found to be REACHABLE more than ip_ire_arp_interval ms ago.
436  * This macro is used to trigger re-arp of existing nce_t entries
437  * so that they can be updated with the latest information.
438  * nce's will get cleaned up (or updated with new info) in the following
439  * circumstances:
440  * - ip_ire_trash_reclaim will free nce's using ndp_cache_reclaim
441  *    when memory is low,
442  * - ip_arp_news, when updates are received.
443  * - if the nce is NCE_EXPIRED(), it will be re-initialized to ND_INITIAL,
444  *   so that a new arp request will be triggered.
445  * nce_last is the timestamp that indicates when the nce_res_mp in the
446  * nce_t was last updated to a valid link-layer address.  nce_last gets
447  * modified/updated :
448  *  - when the nce is created  or reinit-ed
449  *  - every time we get a sane arp response for the nce.
450  */
451 #define	NCE_EXPIRED(nce, ipst)	(nce->nce_last > 0 &&	\
452 	    ((nce->nce_flags & NCE_F_PERMANENT) == 0) &&	\
453 	    ((TICK_TO_MSEC(lbolt64) - nce->nce_last) > 		\
454 		(ipst)->ips_ip_ire_arp_interval))
455 
456 #endif /* _KERNEL */
457 
458 /* ICMP types */
459 #define	ICMP_ECHO_REPLY			0
460 #define	ICMP_DEST_UNREACHABLE		3
461 #define	ICMP_SOURCE_QUENCH		4
462 #define	ICMP_REDIRECT			5
463 #define	ICMP_ECHO_REQUEST		8
464 #define	ICMP_ROUTER_ADVERTISEMENT	9
465 #define	ICMP_ROUTER_SOLICITATION	10
466 #define	ICMP_TIME_EXCEEDED		11
467 #define	ICMP_PARAM_PROBLEM		12
468 #define	ICMP_TIME_STAMP_REQUEST		13
469 #define	ICMP_TIME_STAMP_REPLY		14
470 #define	ICMP_INFO_REQUEST		15
471 #define	ICMP_INFO_REPLY			16
472 #define	ICMP_ADDRESS_MASK_REQUEST	17
473 #define	ICMP_ADDRESS_MASK_REPLY		18
474 
475 /* ICMP_TIME_EXCEEDED codes */
476 #define	ICMP_TTL_EXCEEDED		0
477 #define	ICMP_REASSEMBLY_TIME_EXCEEDED	1
478 
479 /* ICMP_DEST_UNREACHABLE codes */
480 #define	ICMP_NET_UNREACHABLE		0
481 #define	ICMP_HOST_UNREACHABLE		1
482 #define	ICMP_PROTOCOL_UNREACHABLE	2
483 #define	ICMP_PORT_UNREACHABLE		3
484 #define	ICMP_FRAGMENTATION_NEEDED	4
485 #define	ICMP_SOURCE_ROUTE_FAILED	5
486 #define	ICMP_DEST_NET_UNKNOWN		6
487 #define	ICMP_DEST_HOST_UNKNOWN		7
488 #define	ICMP_SRC_HOST_ISOLATED		8
489 #define	ICMP_DEST_NET_UNREACH_ADMIN	9
490 #define	ICMP_DEST_HOST_UNREACH_ADMIN	10
491 #define	ICMP_DEST_NET_UNREACH_TOS	11
492 #define	ICMP_DEST_HOST_UNREACH_TOS	12
493 
494 /* ICMP Header Structure */
495 typedef struct icmph_s {
496 	uint8_t		icmph_type;
497 	uint8_t		icmph_code;
498 	uint16_t	icmph_checksum;
499 	union {
500 		struct { /* ECHO request/response structure */
501 			uint16_t	u_echo_ident;
502 			uint16_t	u_echo_seqnum;
503 		} u_echo;
504 		struct { /* Destination unreachable structure */
505 			uint16_t	u_du_zero;
506 			uint16_t	u_du_mtu;
507 		} u_du;
508 		struct { /* Parameter problem structure */
509 			uint8_t		u_pp_ptr;
510 			uint8_t		u_pp_rsvd[3];
511 		} u_pp;
512 		struct { /* Redirect structure */
513 			ipaddr_t	u_rd_gateway;
514 		} u_rd;
515 	} icmph_u;
516 } icmph_t;
517 
518 #define	icmph_echo_ident	icmph_u.u_echo.u_echo_ident
519 #define	icmph_echo_seqnum	icmph_u.u_echo.u_echo_seqnum
520 #define	icmph_du_zero		icmph_u.u_du.u_du_zero
521 #define	icmph_du_mtu		icmph_u.u_du.u_du_mtu
522 #define	icmph_pp_ptr		icmph_u.u_pp.u_pp_ptr
523 #define	icmph_rd_gateway	icmph_u.u_rd.u_rd_gateway
524 
525 #define	ICMPH_SIZE	8
526 
527 /*
528  * Minimum length of transport layer header included in an ICMP error
529  * message for it to be considered valid.
530  */
531 #define	ICMP_MIN_TP_HDR_LEN	8
532 
533 /* Aligned IP header */
534 typedef struct ipha_s {
535 	uint8_t		ipha_version_and_hdr_length;
536 	uint8_t		ipha_type_of_service;
537 	uint16_t	ipha_length;
538 	uint16_t	ipha_ident;
539 	uint16_t	ipha_fragment_offset_and_flags;
540 	uint8_t		ipha_ttl;
541 	uint8_t		ipha_protocol;
542 	uint16_t	ipha_hdr_checksum;
543 	ipaddr_t	ipha_src;
544 	ipaddr_t	ipha_dst;
545 } ipha_t;
546 
547 #define	IPH_DF		0x4000	/* Don't fragment */
548 #define	IPH_MF		0x2000	/* More fragments to come */
549 #define	IPH_OFFSET	0x1FFF	/* Where the offset lives */
550 #define	IPH_FRAG_HDR	0x8000	/* IPv6 don't fragment bit */
551 
552 /* ECN code points for IPv4 TOS byte and IPv6 traffic class octet. */
553 #define	IPH_ECN_NECT	0x0	/* Not ECN-Capabable Transport */
554 #define	IPH_ECN_ECT1	0x1	/* ECN-Capable Transport, ECT(1) */
555 #define	IPH_ECN_ECT0	0x2	/* ECN-Capable Transport, ECT(0) */
556 #define	IPH_ECN_CE	0x3	/* ECN-Congestion Experienced (CE) */
557 
558 /* IP Mac info structure */
559 typedef struct ip_m_s {
560 	t_uscalar_t	ip_m_mac_type;	/* From <sys/dlpi.h> */
561 	int		ip_m_type;	/* From <net/if_types.h> */
562 	boolean_t	(*ip_m_v4mapinfo)(uint_t, uint8_t *, uint8_t *,
563 			    uint32_t *, ipaddr_t *);
564 	boolean_t	(*ip_m_v6mapinfo)(uint_t, uint8_t *, uint8_t *,
565 			    uint32_t *, in6_addr_t *);
566 	boolean_t	(*ip_m_v6intfid)(uint_t, uint8_t *, in6_addr_t *);
567 } ip_m_t;
568 
569 /*
570  * The following functions attempt to reduce the link layer dependency
571  * of the IP stack. The current set of link specific operations are:
572  * a. map from IPv4 class D (224.0/4) multicast address range to the link
573  * layer multicast address range.
574  * b. map from IPv6 multicast address range (ff00::/8) to the link
575  * layer multicast address range.
576  * c. derive the default IPv6 interface identifier from the link layer
577  * address.
578  */
579 #define	MEDIA_V4MINFO(ip_m, plen, bphys, maddr, hwxp, v4ptr) \
580 	(((ip_m)->ip_m_v4mapinfo != NULL) && \
581 	(*(ip_m)->ip_m_v4mapinfo)(plen, bphys, maddr, hwxp, v4ptr))
582 #define	MEDIA_V6INTFID(ip_m, plen, phys, v6ptr) \
583 	(((ip_m)->ip_m_v6intfid != NULL) && \
584 	(*(ip_m)->ip_m_v6intfid)(plen, phys, v6ptr))
585 #define	MEDIA_V6MINFO(ip_m, plen, bphys, maddr, hwxp, v6ptr) \
586 	(((ip_m)->ip_m_v6mapinfo != NULL) && \
587 	(*(ip_m)->ip_m_v6mapinfo)(plen, bphys, maddr, hwxp, v6ptr))
588 
589 /* Router entry types */
590 #define	IRE_BROADCAST		0x0001	/* Route entry for broadcast address */
591 #define	IRE_DEFAULT		0x0002	/* Route entry for default gateway */
592 #define	IRE_LOCAL		0x0004	/* Route entry for local address */
593 #define	IRE_LOOPBACK		0x0008	/* Route entry for loopback address */
594 #define	IRE_PREFIX		0x0010	/* Route entry for prefix routes */
595 #define	IRE_CACHE		0x0020	/* Cached Route entry */
596 #define	IRE_IF_NORESOLVER	0x0040	/* Route entry for local interface */
597 					/* net without any address mapping. */
598 #define	IRE_IF_RESOLVER		0x0080	/* Route entry for local interface */
599 					/* net with resolver. */
600 #define	IRE_HOST		0x0100	/* Host route entry */
601 #define	IRE_HOST_REDIRECT	0x0200	/* only used for T_SVR4_OPTMGMT_REQ */
602 
603 /*
604  * IRE_MIPRTUN is only set on the ires in the ip_mrtun_table.
605  * This ire_type must not be set for ftable and ctable routing entries.
606  */
607 #define	IRE_MIPRTUN		0x0400	/* Reverse tunnel route entry */
608 
609 #define	IRE_INTERFACE		(IRE_IF_NORESOLVER | IRE_IF_RESOLVER)
610 #define	IRE_OFFSUBNET		(IRE_DEFAULT | IRE_PREFIX | IRE_HOST)
611 #define	IRE_CACHETABLE		(IRE_CACHE | IRE_BROADCAST | IRE_LOCAL | \
612 				IRE_LOOPBACK)
613 #define	IRE_FORWARDTABLE	(IRE_INTERFACE | IRE_OFFSUBNET)
614 
615 /*
616  * If an IRE is marked with IRE_MARK_CONDEMNED, the last walker of
617  * the bucket should delete this IRE from this bucket.
618  */
619 #define	IRE_MARK_CONDEMNED	0x0001
620 /*
621  * If a broadcast IRE is marked with IRE_MARK_NORECV, ip_rput will drop the
622  * broadcast packets received on that interface. This is marked only
623  * on broadcast ires. Employed by IPMP, where we have multiple NICs on the
624  * same subnet receiving the same broadcast packet.
625  */
626 #define	IRE_MARK_NORECV		0x0002
627 /*
628  * IRE_CACHE marked this way won't be returned by ire_cache_lookup. Need
629  * to look specifically using MATCH_IRE_MARK_HIDDEN. Used by IPMP.
630  */
631 #define	IRE_MARK_HIDDEN		0x0004	/* Typically Used by in.mpathd */
632 
633 /*
634  * ire with IRE_MARK_NOADD is  created in ip_newroute_ipif, when outgoing
635  * interface is specified by IP_XMIT_IF socket option. This ire is not
636  * added in IRE_CACHE.  For example, this is used by mipagent to prevent
637  * any entry to be added in the cache table. We do not want to add any
638  * entry for a mobile-node in the routing table for foreign agent originated
639  * packets. Adding routes in cache table in this case, may run the risks of
640  * incorrect routing path in case of private overlapping addresses.
641  */
642 #define	IRE_MARK_NOADD		0x0008	/* Mark not to add ire in cache */
643 
644 /*
645  * IRE marked with IRE_MARK_TEMPORARY means that this IRE has been used
646  * either for forwarding a packet or has not been used for sending
647  * traffic on TCP connections terminated on this system.  In both
648  * cases, this IRE is the first to go when IRE is being cleaned up.
649  */
650 #define	IRE_MARK_TEMPORARY	0x0010
651 
652 /*
653  * IRE marked with IRE_MARK_USESRC_CHECK means that while adding an IRE with
654  * this mark, additional atomic checks need to be performed. For eg: by the
655  * time an IRE_CACHE is created, sent up to ARP and then comes back to IP; the
656  * usesrc grouping could have changed in which case we want to fail adding
657  * the IRE_CACHE entry
658  */
659 #define	IRE_MARK_USESRC_CHECK	0x0020
660 
661 /*
662  * IRE_MARK_PRIVATE_ADDR is used for IP_NEXTHOP. When IP_NEXTHOP is set, the
663  * routing table lookup for the destination is bypassed and the packet is
664  * sent directly to the specified nexthop. The associated IRE_CACHE entries
665  * should be marked with IRE_MARK_PRIVATE_ADDR flag so that they don't show up
666  * in regular ire cache lookups.
667  */
668 #define	IRE_MARK_PRIVATE_ADDR	0x0040
669 
670 /*
671  * When we send an ARP resolution query for the nexthop gateway's ire,
672  * we use esballoc to create the ire_t in the AR_ENTRY_QUERY mblk
673  * chain, and mark its ire_marks with IRE_MARK_UNCACHED. This flag
674  * indicates that information from ARP has not been transferred to a
675  * permanent IRE_CACHE entry. The flag is reset only when the
676  * information is successfully transferred to an ire_cache entry (in
677  * ire_add()). Attempting to free the AR_ENTRY_QUERY mblk chain prior
678  * to ire_add (e.g., from arp, or from ip`ip_wput_nondata) will
679  * require that the resources (incomplete ire_cache and/or nce) must
680  * be cleaned up. The free callback routine (ire_freemblk()) checks
681  * for IRE_MARK_UNCACHED to see if any resources that are pinned down
682  * will need to be cleaned up or not.
683  */
684 
685 #define	IRE_MARK_UNCACHED	0x0080
686 
687 /*
688  * The comment below (and for other netstack_t references) refers
689  * to the fact that we only do netstack_hold in particular cases,
690  * such as the references from open streams (ill_t and conn_t's
691  * pointers). Internally within IP we rely on IP's ability to cleanup e.g.
692  * ire_t's when an ill goes away.
693  */
694 typedef struct ire_expire_arg_s {
695 	int		iea_flush_flag;
696 	ip_stack_t	*iea_ipst;	/* Does not have a netstack_hold */
697 } ire_expire_arg_t;
698 
699 /* Flags with ire_expire routine */
700 #define	FLUSH_ARP_TIME		0x0001	/* ARP info potentially stale timer */
701 #define	FLUSH_REDIRECT_TIME	0x0002	/* Redirects potentially stale */
702 #define	FLUSH_MTU_TIME		0x0004	/* Include path MTU per RFC 1191 */
703 
704 /* Arguments to ire_flush_cache() */
705 #define	IRE_FLUSH_DELETE	0
706 #define	IRE_FLUSH_ADD		1
707 
708 /*
709  * Open/close synchronization flags.
710  * These are kept in a separate field in the conn and the synchronization
711  * depends on the atomic 32 bit access to that field.
712  */
713 #define	CONN_CLOSING		0x01	/* ip_close waiting for ip_wsrv */
714 #define	CONN_IPSEC_LOAD_WAIT	0x02	/* waiting for load */
715 #define	CONN_CONDEMNED		0x04	/* conn is closing, no more refs */
716 #define	CONN_INCIPIENT		0x08	/* conn not yet visible, no refs */
717 #define	CONN_QUIESCED		0x10	/* conn is now quiescent */
718 
719 /*
720  * Parameter to ip_output giving the identity of the caller.
721  * IP_WSRV means the packet was enqueued in the STREAMS queue
722  * due to flow control and is now being reprocessed in the context of
723  * the STREAMS service procedure, consequent to flow control relief.
724  * IRE_SEND means the packet is being reprocessed consequent to an
725  * ire cache creation and addition and this may or may not be happening
726  * in the service procedure context. Anything other than the above 2
727  * cases is identified as IP_WPUT. Most commonly this is the case of
728  * packets coming down from the application.
729  */
730 #ifdef _KERNEL
731 #define	IP_WSRV			1	/* Called from ip_wsrv */
732 #define	IP_WPUT			2	/* Called from ip_wput */
733 #define	IRE_SEND		3	/* Called from ire_send */
734 
735 /*
736  * Extra structures need for per-src-addr filtering (IGMPv3/MLDv2)
737  */
738 #define	MAX_FILTER_SIZE	64
739 
740 typedef struct slist_s {
741 	int		sl_numsrc;
742 	in6_addr_t	sl_addr[MAX_FILTER_SIZE];
743 } slist_t;
744 
745 /*
746  * Following struct is used to maintain retransmission state for
747  * a multicast group.  One rtx_state_t struct is an in-line field
748  * of the ilm_t struct; the slist_ts in the rtx_state_t struct are
749  * alloc'd as needed.
750  */
751 typedef struct rtx_state_s {
752 	uint_t		rtx_timer;	/* retrans timer */
753 	int		rtx_cnt;	/* retrans count */
754 	int		rtx_fmode_cnt;	/* retrans count for fmode change */
755 	slist_t		*rtx_allow;
756 	slist_t		*rtx_block;
757 } rtx_state_t;
758 
759 /*
760  * Used to construct list of multicast address records that will be
761  * sent in a single listener report.
762  */
763 typedef struct mrec_s {
764 	struct mrec_s	*mrec_next;
765 	uint8_t		mrec_type;
766 	uint8_t		mrec_auxlen;	/* currently unused */
767 	in6_addr_t	mrec_group;
768 	slist_t		mrec_srcs;
769 } mrec_t;
770 
771 /* Group membership list per upper conn */
772 /*
773  * XXX add ilg info for ifaddr/ifindex.
774  * XXX can we make ilg survive an ifconfig unplumb + plumb
775  * by setting the ipif/ill to NULL and recover that later?
776  *
777  * ilg_ipif is used by IPv4 as multicast groups are joined using an interface
778  * address (ipif).
779  * ilg_ill is used by IPv6 as multicast groups are joined using an interface
780  * index (phyint->phyint_ifindex).
781  * ilg_ill is NULL for IPv4 and ilg_ipif is NULL for IPv6.
782  *
783  * ilg records the state of multicast memberships of a socket end point.
784  * ilm records the state of multicast memberships with the driver and is
785  * maintained per interface.
786  *
787  * Notes :
788  *
789  * 1) There is no direct link between a given ilg and ilm. If the
790  *    application has joined a group G with ifindex I, we will have
791  *    an ilg with ilg_v6group and ilg_ill. There will be a corresponding
792  *    ilm with ilm_ill/ilm_v6addr recording the multicast membership.
793  *    To delete the membership,
794  *
795  *		a) Search for ilg matching on G and I with ilg_v6group
796  *		   and ilg_ill. Delete ilg_ill.
797  *		b) Search the corresponding ilm matching on G and I with
798  *		   ilm_v6addr and ilm_ill. Delete ilm.
799  *
800  *    In IPv4, the only difference is, we look using ipifs instead of
801  *    ills.
802  *
803  * 2) With IP multipathing, we want to keep receiving even after the
804  *    interface has failed. We do this by moving multicast memberships
805  *    to a new_ill within the group. This is acheived by sending
806  *    DL_DISABMULTI_REQS on ilg_ill/ilm_ill and sending DL_ENABMULTIREQS
807  *    on the new_ill and changing ilg_ill/ilm_ill to new_ill. But, we
808  *    need to be able to delete memberships which will still come down
809  *    with the ifindex of the old ill which is what the application
810  *    knows of. Thus we store the ilm_/ilg_orig_ifindex to keep track
811  *    of where we joined initially so that we can lookup even after we
812  *    moved the membership. It is also used for moving back the membership
813  *    when the old ill has been repaired. This is done by looking up for
814  *    ilms with ilm_orig_ifindex matching on the old ill's ifindex. Only
815  *    ilms actually move from old ill to new ill. ilgs don't move (just
816  *    the ilg_ill is changed when it moves) as it just records the state
817  *    of the application that has joined a group G where as ilm records
818  *    the state joined with the driver. Thus when we send DL_XXXMULTI_REQs
819  *    we also need to keep the ilm in the right ill.
820  *
821  *    In IPv4, as ipifs move from old ill to new_ill, ilgs and ilms move
822  *    implicitly as we use only ipifs in IPv4. Thus, one can always lookup
823  *    a given ilm/ilg even after it fails without the support of
824  *    orig_ifindex. We move ilms still to record the driver state as
825  *    mentioned above.
826  */
827 
828 /*
829  * The ilg_t and ilm_t members are protected by ipsq. They can be changed only
830  * by a thread executing in the ipsq. In other words add/delete of a
831  * multicast group has to execute in the ipsq.
832  */
833 #define	ILG_DELETED	0x1		/* ilg_flags */
834 typedef struct ilg_s {
835 	in6_addr_t	ilg_v6group;
836 	struct ipif_s	*ilg_ipif;	/* Logical interface we are member on */
837 	struct ill_s	*ilg_ill;	/* Used by IPv6 */
838 	int		ilg_orig_ifindex; /* Interface originally joined on */
839 	uint_t		ilg_flags;
840 	mcast_record_t	ilg_fmode;	/* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */
841 	slist_t		*ilg_filter;
842 } ilg_t;
843 
844 
845 /*
846  * Multicast address list entry for lower ill.
847  * ilm_ipif is used by IPv4 as multicast groups are joined using ipif.
848  * ilm_ill is used by IPv6 as multicast groups are joined using ill.
849  * ilm_ill is NULL for IPv4 and ilm_ipif is NULL for IPv6.
850  *
851  * The comment below (and for other netstack_t references) refers
852  * to the fact that we only do netstack_hold in particular cases,
853  * such as the references from open streams (ill_t and conn_t's
854  * pointers). Internally within IP we rely on IP's ability to cleanup e.g.
855  * ire_t's when an ill goes away.
856  */
857 #define	ILM_DELETED	0x1		/* ilm_flags */
858 typedef struct ilm_s {
859 	in6_addr_t	ilm_v6addr;
860 	int		ilm_refcnt;
861 	uint_t		ilm_timer;	/* IGMP/MLD query resp timer, in msec */
862 	struct ipif_s	*ilm_ipif;	/* Back pointer to ipif for IPv4 */
863 	struct ilm_s	*ilm_next;	/* Linked list for each ill */
864 	uint_t		ilm_state;	/* state of the membership */
865 	struct ill_s	*ilm_ill;	/* Back pointer to ill for IPv6 */
866 	int		ilm_orig_ifindex;  /* V6_MULTICAST_IF/ilm_ipif index */
867 	uint_t		ilm_flags;
868 	boolean_t	ilm_is_new;	/* new ilm */
869 	boolean_t	ilm_notify_driver; /* Need to notify the driver */
870 	zoneid_t	ilm_zoneid;
871 	int		ilm_no_ilg_cnt;	/* number of joins w/ no ilg */
872 	mcast_record_t	ilm_fmode;	/* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */
873 	slist_t		*ilm_filter;	/* source filter list */
874 	slist_t		*ilm_pendsrcs;	/* relevant src addrs for pending req */
875 	rtx_state_t	ilm_rtx;	/* SCR retransmission state */
876 	ip_stack_t	*ilm_ipst;	/* Does not have a netstack_hold */
877 } ilm_t;
878 
879 #define	ilm_addr	V4_PART_OF_V6(ilm_v6addr)
880 
881 /*
882  * ilm_walker_cleanup needs to execute when the ilm_walker_cnt goes down to
883  * zero. In addition it needs to block new walkers while it is unlinking ilm's
884  * from the list. Thus simple atomics for the ill_ilm_walker_cnt don't suffice.
885  */
886 #define	ILM_WALKER_HOLD(ill)    {               \
887 	mutex_enter(&(ill)->ill_lock);          \
888 	ill->ill_ilm_walker_cnt++;              \
889 	mutex_exit(&(ill)->ill_lock);           \
890 }
891 
892 #define	ILM_WALKER_RELE(ill)	{ 		\
893 	mutex_enter(&(ill)->ill_lock);		\
894 	(ill)->ill_ilm_walker_cnt--;		\
895 	if ((ill)->ill_ilm_walker_cnt == 0 && (ill)->ill_ilm_cleanup_reqd) \
896 		ilm_walker_cleanup(ill);	\
897 	mutex_exit(&(ill)->ill_lock);		\
898 }
899 
900 /*
901  * Soft reference to an IPsec SA.
902  *
903  * On relative terms, conn's can be persistant (living as long as the
904  * processes which create them), while SA's are ephemeral (dying when
905  * they hit their time-based or byte-based lifetimes).
906  *
907  * We could hold a hard reference to an SA from an ipsec_latch_t,
908  * but this would cause expired SA's to linger for a potentially
909  * unbounded time.
910  *
911  * Instead, we remember the hash bucket number and bucket generation
912  * in addition to the pointer.  The bucket generation is incremented on
913  * each deletion.
914  */
915 typedef struct ipsa_ref_s
916 {
917 	struct ipsa_s	*ipsr_sa;
918 	struct isaf_s	*ipsr_bucket;
919 	uint64_t	ipsr_gen;
920 } ipsa_ref_t;
921 
922 /*
923  * IPsec "latching" state.
924  *
925  * In the presence of IPsec policy, fully-bound conn's bind a connection
926  * to more than just the 5-tuple, but also a specific IPsec action and
927  * identity-pair.
928  *
929  * As an optimization, we also cache soft references to IPsec SA's
930  * here so that we can fast-path around most of the work needed for
931  * outbound IPsec SA selection.
932  *
933  * Were it not for TCP's detached connections, this state would be
934  * in-line in conn_t; instead, this is in a separate structure so it
935  * can be handed off to TCP when a connection is detached.
936  */
937 typedef struct ipsec_latch_s
938 {
939 	kmutex_t	ipl_lock;
940 	uint32_t	ipl_refcnt;
941 
942 	uint64_t	ipl_unique;
943 	struct ipsec_policy_s	*ipl_in_policy; /* latched policy (in) */
944 	struct ipsec_policy_s	*ipl_out_policy; /* latched policy (out) */
945 	struct ipsec_action_s	*ipl_in_action;	/* latched action (in) */
946 	struct ipsec_action_s	*ipl_out_action; /* latched action (out) */
947 	cred_t		*ipl_local_id;
948 	struct ipsid_s	*ipl_local_cid;
949 	struct ipsid_s	*ipl_remote_cid;
950 	unsigned int
951 			ipl_out_action_latched : 1,
952 			ipl_in_action_latched : 1,
953 			ipl_out_policy_latched : 1,
954 			ipl_in_policy_latched : 1,
955 
956 			ipl_ids_latched : 1,
957 
958 			ipl_pad_to_bit_31 : 27;
959 
960 	ipsa_ref_t	ipl_ref[2]; /* 0: ESP, 1: AH */
961 
962 } ipsec_latch_t;
963 
964 #define	IPLATCH_REFHOLD(ipl) { \
965 	atomic_add_32(&(ipl)->ipl_refcnt, 1);		\
966 	ASSERT((ipl)->ipl_refcnt != 0);			\
967 }
968 
969 #define	IPLATCH_REFRELE(ipl, ns) {				\
970 	ASSERT((ipl)->ipl_refcnt != 0);				\
971 	membar_exit();						\
972 	if (atomic_add_32_nv(&(ipl)->ipl_refcnt, -1) == 0)	\
973 		iplatch_free(ipl, ns);			\
974 }
975 
976 /*
977  * peer identity structure.
978  */
979 
980 typedef struct conn_s conn_t;
981 
982 /*
983  * The old IP client structure "ipc_t" is gone. All the data is stored in the
984  * connection structure "conn_t" now. The mapping of old and new fields looks
985  * like this:
986  *
987  * ipc_ulp			conn_ulp
988  * ipc_rq			conn_rq
989  * ipc_wq			conn_wq
990  *
991  * ipc_laddr			conn_src
992  * ipc_faddr			conn_rem
993  * ipc_v6laddr			conn_srcv6
994  * ipc_v6faddr			conn_remv6
995  *
996  * ipc_lport			conn_lport
997  * ipc_fport			conn_fport
998  * ipc_ports			conn_ports
999  *
1000  * ipc_policy			conn_policy
1001  * ipc_latch			conn_latch
1002  *
1003  * ipc_irc_lock			conn_lock
1004  * ipc_ire_cache		conn_ire_cache
1005  *
1006  * ipc_state_flags		conn_state_flags
1007  * ipc_outgoing_ill		conn_outgoing_ill
1008  *
1009  * ipc_dontroute 		conn_dontroute
1010  * ipc_loopback 		conn_loopback
1011  * ipc_broadcast		conn_broadcast
1012  * ipc_reuseaddr		conn_reuseaddr
1013  *
1014  * ipc_multicast_loop		conn_multicast_loop
1015  * ipc_multi_router		conn_multi_router
1016  * ipc_priv_stream 		conn_priv_stream
1017  * ipc_draining 		conn_draining
1018  *
1019  * ipc_did_putbq		conn_did_putbq
1020  * ipc_unspec_src		conn_unspec_src
1021  * ipc_policy_cached		conn_policy_cached
1022  *
1023  * ipc_in_enforce_policy 	conn_in_enforce_policy
1024  * ipc_out_enforce_policy 	conn_out_enforce_policy
1025  * ipc_af_isv6			conn_af_isv6
1026  * ipc_pkt_isv6			conn_pkt_isv6
1027  *
1028  * ipc_ipv6_recvpktinfo		conn_ipv6_recvpktinfo
1029  *
1030  * ipc_ipv6_recvhoplimit	conn_ipv6_recvhoplimit
1031  * ipc_ipv6_recvhopopts		conn_ipv6_recvhopopts
1032  * ipc_ipv6_recvdstopts		conn_ipv6_recvdstopts
1033  *
1034  * ipc_ipv6_recvrthdr 		conn_ipv6_recvrthdr
1035  * ipc_ipv6_recvrtdstopts	conn_ipv6_recvrtdstopts
1036  * ipc_fully_bound		conn_fully_bound
1037  *
1038  * ipc_recvif			conn_recvif
1039  *
1040  * ipc_recvslla 		conn_recvslla
1041  * ipc_acking_unbind 		conn_acking_unbind
1042  * ipc_pad_to_bit_31 		conn_pad_to_bit_31
1043  *
1044  * ipc_xmit_if_ill		conn_xmit_if_ill
1045  * ipc_nofailover_ill		conn_nofailover_ill
1046  *
1047  * ipc_proto			conn_proto
1048  * ipc_incoming_ill		conn_incoming_ill
1049  * ipc_outgoing_pill		conn_outgoing_pill
1050  * ipc_pending_ill		conn_pending_ill
1051  * ipc_unbind_mp		conn_unbind_mp
1052  * ipc_ilg			conn_ilg
1053  * ipc_ilg_allocated		conn_ilg_allocated
1054  * ipc_ilg_inuse		conn_ilg_inuse
1055  * ipc_ilg_walker_cnt		conn_ilg_walker_cnt
1056  * ipc_refcv			conn_refcv
1057  * ipc_multicast_ipif		conn_multicast_ipif
1058  * ipc_multicast_ill		conn_multicast_ill
1059  * ipc_orig_bound_ifindex	conn_orig_bound_ifindex
1060  * ipc_orig_multicast_ifindex	conn_orig_multicast_ifindex
1061  * ipc_orig_xmit_ifindex	conn_orig_xmit_ifindex
1062  * ipc_drain_next		conn_drain_next
1063  * ipc_drain_prev		conn_drain_prev
1064  * ipc_idl			conn_idl
1065  */
1066 
1067 /*
1068  * This is used to match an inbound/outbound datagram with
1069  * policy.
1070  */
1071 
1072 typedef	struct ipsec_selector {
1073 	in6_addr_t	ips_local_addr_v6;
1074 	in6_addr_t	ips_remote_addr_v6;
1075 	uint16_t	ips_local_port;
1076 	uint16_t	ips_remote_port;
1077 	uint8_t		ips_icmp_type;
1078 	uint8_t		ips_icmp_code;
1079 	uint8_t		ips_protocol;
1080 	uint8_t		ips_isv4 : 1,
1081 			ips_is_icmp_inv_acq: 1;
1082 } ipsec_selector_t;
1083 
1084 /*
1085  * Note that we put v4 addresses in the *first* 32-bit word of the
1086  * selector rather than the last to simplify the prefix match/mask code
1087  * in spd.c
1088  */
1089 #define	ips_local_addr_v4 ips_local_addr_v6.s6_addr32[0]
1090 #define	ips_remote_addr_v4 ips_remote_addr_v6.s6_addr32[0]
1091 
1092 /* Values used in IP by IPSEC Code */
1093 #define		IPSEC_OUTBOUND		B_TRUE
1094 #define		IPSEC_INBOUND		B_FALSE
1095 
1096 /*
1097  * There are two variants in policy failures. The packet may come in
1098  * secure when not needed (IPSEC_POLICY_???_NOT_NEEDED) or it may not
1099  * have the desired level of protection (IPSEC_POLICY_MISMATCH).
1100  */
1101 #define	IPSEC_POLICY_NOT_NEEDED		0
1102 #define	IPSEC_POLICY_MISMATCH		1
1103 #define	IPSEC_POLICY_AUTH_NOT_NEEDED	2
1104 #define	IPSEC_POLICY_ENCR_NOT_NEEDED	3
1105 #define	IPSEC_POLICY_SE_NOT_NEEDED	4
1106 #define	IPSEC_POLICY_MAX		5	/* Always max + 1. */
1107 
1108 /*
1109  * Folowing macro is used whenever the code does not know whether there
1110  * is a M_CTL present in the front and it needs to examine the actual mp
1111  * i.e the IP header. As a M_CTL message could be in the front, this
1112  * extracts the packet into mp and the M_CTL mp into first_mp. If M_CTL
1113  * mp is not present, both first_mp and mp point to the same message.
1114  */
1115 #define	EXTRACT_PKT_MP(mp, first_mp, mctl_present)	\
1116 	(first_mp) = (mp);				\
1117 	if ((mp)->b_datap->db_type == M_CTL) {		\
1118 		(mp) = (mp)->b_cont;			\
1119 		(mctl_present) = B_TRUE;		\
1120 	} else {					\
1121 		(mctl_present) = B_FALSE;		\
1122 	}
1123 
1124 /*
1125  * Check with IPSEC inbound policy if
1126  *
1127  * 1) per-socket policy is present - indicated by conn_in_enforce_policy.
1128  * 2) Or if we have not cached policy on the conn and the global policy is
1129  *    non-empty.
1130  */
1131 #define	CONN_INBOUND_POLICY_PRESENT(connp, ipss)	\
1132 	((connp)->conn_in_enforce_policy ||		\
1133 	(!((connp)->conn_policy_cached) && 		\
1134 	(ipss)->ipsec_inbound_v4_policy_present))
1135 
1136 #define	CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)	\
1137 	((connp)->conn_in_enforce_policy ||		\
1138 	(!(connp)->conn_policy_cached &&		\
1139 	(ipss)->ipsec_inbound_v6_policy_present))
1140 
1141 #define	CONN_OUTBOUND_POLICY_PRESENT(connp, ipss)	\
1142 	((connp)->conn_out_enforce_policy ||		\
1143 	(!((connp)->conn_policy_cached) &&		\
1144 	(ipss)->ipsec_outbound_v4_policy_present))
1145 
1146 #define	CONN_OUTBOUND_POLICY_PRESENT_V6(connp, ipss)	\
1147 	((connp)->conn_out_enforce_policy ||		\
1148 	(!(connp)->conn_policy_cached &&		\
1149 	(ipss)->ipsec_outbound_v6_policy_present))
1150 
1151 /*
1152  * Information cached in IRE for upper layer protocol (ULP).
1153  *
1154  * Notice that ire_max_frag is not included in the iulp_t structure, which
1155  * it may seem that it should.  But ire_max_frag cannot really be cached.  It
1156  * is fixed for each interface.  For MTU found by PMTUd, we may want to cache
1157  * it.  But currently, we do not do that.
1158  */
1159 typedef struct iulp_s {
1160 	boolean_t	iulp_set;	/* Is any metric set? */
1161 	uint32_t	iulp_ssthresh;	/* Slow start threshold (TCP). */
1162 	clock_t		iulp_rtt;	/* Guestimate in millisecs. */
1163 	clock_t		iulp_rtt_sd;	/* Cached value of RTT variance. */
1164 	uint32_t	iulp_spipe;	/* Send pipe size. */
1165 	uint32_t	iulp_rpipe;	/* Receive pipe size. */
1166 	uint32_t	iulp_rtomax;	/* Max round trip timeout. */
1167 	uint32_t	iulp_sack;	/* Use SACK option (TCP)? */
1168 	uint32_t
1169 		iulp_tstamp_ok : 1,	/* Use timestamp option (TCP)? */
1170 		iulp_wscale_ok : 1,	/* Use window scale option (TCP)? */
1171 		iulp_ecn_ok : 1,	/* Enable ECN (for TCP)? */
1172 		iulp_pmtud_ok : 1,	/* Enable PMTUd? */
1173 
1174 		iulp_not_used : 28;
1175 } iulp_t;
1176 
1177 /* Zero iulp_t. */
1178 extern const iulp_t ire_uinfo_null;
1179 
1180 /*
1181  * The conn drain list structure.
1182  * The list is protected by idl_lock. Each conn_t inserted in the list
1183  * points back at this idl_t using conn_idl. IP primes the draining of the
1184  * conns queued in these lists, by qenabling the 1st conn of each list. This
1185  * occurs when STREAMS backenables ip_wsrv on the IP module. Each conn instance
1186  * of ip_wsrv successively qenables the next conn in the list.
1187  * idl_lock protects all other members of idl_t and conn_drain_next
1188  * and conn_drain_prev of conn_t. The conn_lock protects IPCF_DRAIN_DISABLED
1189  * flag of the conn_t and conn_idl.
1190  */
1191 typedef struct idl_s {
1192 	conn_t		*idl_conn;		/* Head of drain list */
1193 	kmutex_t	idl_lock;		/* Lock for this list */
1194 	conn_t		*idl_conn_draining;	/* conn that is draining */
1195 	uint32_t
1196 		idl_repeat : 1,			/* Last conn must re-enable */
1197 						/* drain list again */
1198 		idl_unused : 31;
1199 } idl_t;
1200 
1201 #define	CONN_DRAIN_LIST_LOCK(connp)	(&((connp)->conn_idl->idl_lock))
1202 /*
1203  * Interface route structure which holds the necessary information to recreate
1204  * routes that are tied to an interface (namely where ire_ipif != NULL).
1205  * These routes which were initially created via a routing socket or via the
1206  * SIOCADDRT ioctl may be gateway routes (RTF_GATEWAY being set) or may be
1207  * traditional interface routes.  When an interface comes back up after being
1208  * marked down, this information will be used to recreate the routes.  These
1209  * are part of an mblk_t chain that hangs off of the IPIF (ipif_saved_ire_mp).
1210  */
1211 typedef struct ifrt_s {
1212 	ushort_t	ifrt_type;		/* Type of IRE */
1213 	in6_addr_t	ifrt_v6addr;		/* Address IRE represents. */
1214 	in6_addr_t	ifrt_v6gateway_addr;	/* Gateway if IRE_OFFSUBNET */
1215 	in6_addr_t	ifrt_v6src_addr;	/* Src addr if RTF_SETSRC */
1216 	in6_addr_t	ifrt_v6mask;		/* Mask for matching IRE. */
1217 	uint32_t	ifrt_flags;		/* flags related to route */
1218 	uint_t		ifrt_max_frag;		/* MTU (next hop or path). */
1219 	iulp_t		ifrt_iulp_info;		/* Cached IRE ULP info. */
1220 } ifrt_t;
1221 
1222 #define	ifrt_addr		V4_PART_OF_V6(ifrt_v6addr)
1223 #define	ifrt_gateway_addr	V4_PART_OF_V6(ifrt_v6gateway_addr)
1224 #define	ifrt_src_addr		V4_PART_OF_V6(ifrt_v6src_addr)
1225 #define	ifrt_mask		V4_PART_OF_V6(ifrt_v6mask)
1226 
1227 /* Number of IP addresses that can be hosted on a physical interface */
1228 #define	MAX_ADDRS_PER_IF	8192
1229 /*
1230  * Number of Source addresses to be considered for source address
1231  * selection. Used by ipif_select_source[_v6].
1232  */
1233 #define	MAX_IPIF_SELECT_SOURCE	50
1234 
1235 #ifdef IP_DEBUG
1236 /*
1237  * Tracing refholds and refreleases for debugging. Existing tracing mechanisms
1238  * do not allow the granularity need to trace refrences to ipif/ill/ire's. This
1239  * mechanism should be revisited once dtrace is available.
1240  */
1241 #define	IP_STACK_DEPTH	15
1242 typedef struct tr_buf_s {
1243 	int	tr_depth;
1244 	pc_t	tr_stack[IP_STACK_DEPTH];
1245 } tr_buf_t;
1246 
1247 typedef struct th_trace_s {
1248 	struct	th_trace_s *th_next;
1249 	struct	th_trace_s **th_prev;
1250 	kthread_t	*th_id;
1251 	int	th_refcnt;
1252 	uint_t	th_trace_lastref;
1253 #define	TR_BUF_MAX	38
1254 	tr_buf_t th_trbuf[TR_BUF_MAX];
1255 } th_trace_t;
1256 #endif
1257 
1258 /* The following are ipif_state_flags */
1259 #define	IPIF_CONDEMNED		0x1	/* The ipif is being removed */
1260 #define	IPIF_CHANGING		0x2	/* A critcal ipif field is changing */
1261 #define	IPIF_MOVING		0x8	/* The ipif is being moved */
1262 #define	IPIF_SET_LINKLOCAL	0x10	/* transient flag during bringup */
1263 #define	IPIF_ZERO_SOURCE	0x20	/* transient flag during bringup */
1264 
1265 /* IP interface structure, one per local address */
1266 typedef struct ipif_s {
1267 	struct	ipif_s	*ipif_next;
1268 	struct	ill_s	*ipif_ill;	/* Back pointer to our ill */
1269 	int	ipif_id;		/* Logical unit number */
1270 	uint_t	ipif_mtu;		/* Starts at ipif_ill->ill_max_frag */
1271 	uint_t	ipif_saved_mtu;		/* Save of mtu during ipif_move() */
1272 	in6_addr_t ipif_v6lcl_addr;	/* Local IP address for this if. */
1273 	in6_addr_t ipif_v6src_addr;	/* Source IP address for this if. */
1274 	in6_addr_t ipif_v6subnet;	/* Subnet prefix for this if. */
1275 	in6_addr_t ipif_v6net_mask;	/* Net mask for this interface. */
1276 	in6_addr_t ipif_v6brd_addr;	/* Broadcast addr for this interface. */
1277 	in6_addr_t ipif_v6pp_dst_addr;	/* Point-to-point dest address. */
1278 	uint64_t ipif_flags;		/* Interface flags. */
1279 	uint_t	ipif_metric;		/* BSD if metric, for compatibility. */
1280 	uint_t	ipif_ire_type;		/* IRE_LOCAL or IRE_LOOPBACK */
1281 	mblk_t	*ipif_arp_del_mp;	/* Allocated at time arp comes up, to */
1282 					/* prevent awkward out of mem */
1283 					/* condition later */
1284 	mblk_t	*ipif_saved_ire_mp;	/* Allocated for each extra */
1285 					/* IRE_IF_NORESOLVER/IRE_IF_RESOLVER */
1286 					/* on this interface so that they */
1287 					/* can survive ifconfig down. */
1288 	kmutex_t ipif_saved_ire_lock;	/* Protects ipif_saved_ire_mp */
1289 
1290 	mrec_t	*ipif_igmp_rpt;		/* List of group memberships which */
1291 					/* will be reported on.  Used when */
1292 					/* handling an igmp timeout.	   */
1293 
1294 	/*
1295 	 * The packet counts in the ipif contain the sum of the
1296 	 * packet counts in dead IREs that were affiliated with
1297 	 * this ipif.
1298 	 */
1299 	uint_t	ipif_fo_pkt_count;	/* Forwarded thru our dead IREs */
1300 	uint_t	ipif_ib_pkt_count;	/* Inbound packets for our dead IREs */
1301 	uint_t	ipif_ob_pkt_count;	/* Outbound packets to our dead IREs */
1302 	/* Exclusive bit fields, protected by ipsq_t */
1303 	unsigned int
1304 		ipif_multicast_up : 1,	/* We have joined the allhosts group */
1305 		ipif_solmcast_up : 1,	/* We joined solicited node mcast */
1306 		ipif_replace_zero : 1,	/* Replacement for zero */
1307 		ipif_was_up : 1,	/* ipif was up before */
1308 
1309 		ipif_addr_ready : 1,	/* DAD is done */
1310 		ipif_was_dup : 1,	/* DAD had failed */
1311 		ipif_pad_to_31 : 26;
1312 
1313 	int	ipif_orig_ifindex;	/* ifindex before SLIFFAILOVER */
1314 	uint_t	ipif_seqid;		/* unique index across all ills */
1315 	uint_t	ipif_orig_ipifid;	/* ipif_id before SLIFFAILOVER */
1316 	uint_t	ipif_state_flags;	/* See IPIF_* flag defs above */
1317 	uint_t	ipif_refcnt;		/* active consistent reader cnt */
1318 	uint_t	ipif_ire_cnt;		/* Number of ire's referencing ipif */
1319 	uint_t	ipif_saved_ire_cnt;
1320 	zoneid_t
1321 		ipif_zoneid;		/* zone ID number */
1322 	timeout_id_t ipif_recovery_id;	/* Timer for DAD recovery */
1323 #ifdef ILL_DEBUG
1324 #define	IP_TR_HASH_MAX	64
1325 	th_trace_t *ipif_trace[IP_TR_HASH_MAX];
1326 	boolean_t	ipif_trace_disable;	/* True when alloc fails */
1327 #endif
1328 } ipif_t;
1329 
1330 /*
1331  * The following table lists the protection levels of the various members
1332  * of the ipif_t. The following notation is used.
1333  *
1334  * Write once - Written to only once at the time of bringing up
1335  * the interface and can be safely read after the bringup without any lock.
1336  *
1337  * ipsq - Need to execute in the ipsq to perform the indicated access.
1338  *
1339  * ill_lock - Need to hold this mutex to perform the indicated access.
1340  *
1341  * ill_g_lock - Need to hold this rw lock as reader/writer for read access or
1342  * write access respectively.
1343  *
1344  * down ill - Written to only when the ill is down (i.e all ipifs are down)
1345  * up ill - Read only when the ill is up (i.e. at least 1 ipif is up)
1346  *
1347  *		 Table of ipif_t members and their protection
1348  *
1349  * ipif_next		ill_g_lock		ill_g_lock
1350  * ipif_ill		ipsq + down ipif	write once
1351  * ipif_id		ipsq + down ipif	write once
1352  * ipif_mtu		ipsq
1353  * ipif_v6lcl_addr	ipsq + down ipif	up ipif
1354  * ipif_v6src_addr	ipsq + down ipif	up ipif
1355  * ipif_v6subnet	ipsq + down ipif	up ipif
1356  * ipif_v6net_mask	ipsq + down ipif	up ipif
1357  *
1358  * ipif_v6brd_addr
1359  * ipif_v6pp_dst_addr
1360  * ipif_flags		ill_lock		ill_lock
1361  * ipif_metric
1362  * ipif_ire_type	ipsq + down ill		up ill
1363  *
1364  * ipif_arp_del_mp	ipsq			ipsq
1365  * ipif_saved_ire_mp	ipif_saved_ire_lock	ipif_saved_ire_lock
1366  * ipif_igmp_rpt	ipsq			ipsq
1367  *
1368  * ipif_fo_pkt_count	Approx
1369  * ipif_ib_pkt_count	Approx
1370  * ipif_ob_pkt_count	Approx
1371  *
1372  * bit fields		ill_lock		ill_lock
1373  *
1374  * ipif_orig_ifindex	ipsq			None
1375  * ipif_orig_ipifid	ipsq			None
1376  * ipif_seqid		ipsq			Write once
1377  *
1378  * ipif_state_flags	ill_lock		ill_lock
1379  * ipif_refcnt		ill_lock		ill_lock
1380  * ipif_ire_cnt		ill_lock		ill_lock
1381  * ipif_saved_ire_cnt
1382  */
1383 
1384 #define	IP_TR_HASH(tid)	((((uintptr_t)tid) >> 6) & (IP_TR_HASH_MAX - 1))
1385 
1386 #ifdef ILL_DEBUG
1387 #define	IPIF_TRACE_REF(ipif)	ipif_trace_ref(ipif)
1388 #define	ILL_TRACE_REF(ill)	ill_trace_ref(ill)
1389 #define	IPIF_UNTRACE_REF(ipif)	ipif_untrace_ref(ipif)
1390 #define	ILL_UNTRACE_REF(ill)	ill_untrace_ref(ill)
1391 #define	ILL_TRACE_CLEANUP(ill)	ill_trace_cleanup(ill)
1392 #define	IPIF_TRACE_CLEANUP(ipif)	ipif_trace_cleanup(ipif)
1393 #else
1394 #define	IPIF_TRACE_REF(ipif)
1395 #define	ILL_TRACE_REF(ill)
1396 #define	IPIF_UNTRACE_REF(ipif)
1397 #define	ILL_UNTRACE_REF(ill)
1398 #define	ILL_TRACE_CLEANUP(ill)
1399 #define	IPIF_TRACE_CLEANUP(ipif)
1400 #endif
1401 
1402 /* IPv4 compatability macros */
1403 #define	ipif_lcl_addr		V4_PART_OF_V6(ipif_v6lcl_addr)
1404 #define	ipif_src_addr		V4_PART_OF_V6(ipif_v6src_addr)
1405 #define	ipif_subnet		V4_PART_OF_V6(ipif_v6subnet)
1406 #define	ipif_net_mask		V4_PART_OF_V6(ipif_v6net_mask)
1407 #define	ipif_brd_addr		V4_PART_OF_V6(ipif_v6brd_addr)
1408 #define	ipif_pp_dst_addr	V4_PART_OF_V6(ipif_v6pp_dst_addr)
1409 
1410 /* Macros for easy backreferences to the ill. */
1411 #define	ipif_wq			ipif_ill->ill_wq
1412 #define	ipif_rq			ipif_ill->ill_rq
1413 #define	ipif_net_type		ipif_ill->ill_net_type
1414 #define	ipif_resolver_mp	ipif_ill->ill_resolver_mp
1415 #define	ipif_ipif_up_count	ipif_ill->ill_ipif_up_count
1416 #define	ipif_bcast_mp		ipif_ill->ill_bcast_mp
1417 #define	ipif_type		ipif_ill->ill_type
1418 #define	ipif_isv6		ipif_ill->ill_isv6
1419 
1420 #define	SIOCLIFADDR_NDX 112	/* ndx of SIOCLIFADDR in the ndx ioctl table */
1421 
1422 /*
1423  * mode value for ip_ioctl_finish for finishing an ioctl
1424  */
1425 #define	CONN_CLOSE	1		/* No mi_copy */
1426 #define	COPYOUT		2		/* do an mi_copyout if needed */
1427 #define	NO_COPYOUT	3		/* do an mi_copy_done */
1428 #define	IPI2MODE(ipi)	((ipi)->ipi_flags & IPI_GET_CMD ? COPYOUT : NO_COPYOUT)
1429 
1430 /*
1431  * The IP-MT design revolves around the serialization object ipsq_t.
1432  * It is associated with an IPMP group. If IPMP is not enabled, there is
1433  * 1 ipsq_t per phyint. Eg. an ipsq_t would cover both hme0's IPv4 stream
1434  *
1435  * ipsq_lock protects
1436  *	ipsq_reentry_cnt, ipsq_writer, ipsq_xopq_mphead, ipsq_xopq_mptail,
1437  *	ipsq_mphead, ipsq_mptail, ipsq_split
1438  *
1439  *	ipsq_pending_ipif, ipsq_current_ipif, ipsq_pending_mp, ipsq_flags,
1440  *	ipsq_waitfor
1441  *
1442  * The fields in the last line above below are set mostly by a writer thread
1443  * But there is an exception in the last call to ipif_ill_refrele_tail which
1444  * could also race with a conn close which could be cleaning up the
1445  * fields. So we choose to protect using ipsq_lock instead of depending on
1446  * the property of the writer.
1447  * ill_g_lock protects
1448  *	ipsq_refs, ipsq_phyint_list
1449  */
1450 typedef struct ipsq_s {
1451 	kmutex_t ipsq_lock;
1452 	int	ipsq_reentry_cnt;
1453 	kthread_t *ipsq_writer;		/* current owner (thread id) */
1454 	int	ipsq_flags;
1455 	mblk_t	*ipsq_xopq_mphead;	/* list of excl ops mostly ioctls */
1456 	mblk_t	*ipsq_xopq_mptail;
1457 	mblk_t	*ipsq_mphead;		/* msgs on ipsq linked thru b_next */
1458 	mblk_t	*ipsq_mptail;		/* msgs on ipsq linked thru b_next */
1459 	int	ipsq_current_ioctl;	/* current ioctl, or 0 if no ioctl */
1460 	ipif_t	*ipsq_current_ipif;	/* ipif associated with current op */
1461 	ipif_t	*ipsq_pending_ipif;	/* ipif associated w. ipsq_pending_mp */
1462 	mblk_t	*ipsq_pending_mp;	/* current ioctl mp while waiting for */
1463 					/* response from another module */
1464 	struct	ipsq_s	*ipsq_next;	/* list of all syncq's (ipsq_g_list) */
1465 	uint_t		ipsq_refs;	/* Number of phyints on this ipsq */
1466 	struct phyint	*ipsq_phyint_list; /* List of phyints on this ipsq */
1467 	boolean_t	ipsq_split;	/* ipsq may need to be split */
1468 	int		ipsq_waitfor;	/* Values encoded below */
1469 	char		ipsq_name[LIFNAMSIZ+1];	/* same as phyint_groupname */
1470 
1471 #ifdef ILL_DEBUG
1472 	int		ipsq_depth;	/* debugging aid */
1473 	pc_t		ipsq_stack[IP_STACK_DEPTH];	/* debugging aid */
1474 #endif
1475 	ip_stack_t	*ipsq_ipst;	/* Does not have a netstack_hold */
1476 } ipsq_t;
1477 
1478 /* ipsq_flags */
1479 #define	IPSQ_GROUP	0x1	/* This ipsq belongs to an IPMP group */
1480 
1481 /*
1482  * ipsq_waitfor:
1483  *
1484  * IPIF_DOWN	1	ipif_down waiting for refcnts to drop
1485  * ILL_DOWN	2	ill_down waiting for refcnts to drop
1486  * IPIF_FREE	3	ipif_free waiting for refcnts to drop
1487  * ILL_FREE	4	ill unplumb waiting for refcnts to drop
1488  * ILL_MOVE_OK	5	failover waiting for refcnts to drop
1489  */
1490 
1491 enum { IPIF_DOWN = 1, ILL_DOWN, IPIF_FREE, ILL_FREE, ILL_MOVE_OK };
1492 
1493 /* Flags passed to ipsq_try_enter */
1494 #define	CUR_OP 0		/* Current ioctl continuing again */
1495 #define	NEW_OP 1		/* New ioctl starting afresh */
1496 
1497 /*
1498  * phyint represents state that is common to both IPv4 and IPv6 interfaces.
1499  * There is a separate ill_t representing IPv4 and IPv6 which has a
1500  * backpointer to the phyint structure for acessing common state.
1501  *
1502  * NOTE : It just stores the group name as there is only one name for
1503  *	  IPv4 and IPv6 i.e it is a underlying link property. Actually
1504  *        IPv4 and IPv6 ill are grouped together when their phyints have
1505  *        the same name.
1506  */
1507 typedef struct phyint {
1508 	struct ill_s	*phyint_illv4;
1509 	struct ill_s	*phyint_illv6;
1510 	uint_t		phyint_ifindex;		/* SIOCLSLIFINDEX */
1511 	uint_t		phyint_notify_delay;	/* SIOCSLIFNOTIFYDELAY */
1512 	char		*phyint_groupname;	/* SIOCSLIFGROUPNAME */
1513 	uint_t		phyint_groupname_len;
1514 	uint64_t	phyint_flags;
1515 	avl_node_t	phyint_avl_by_index;	/* avl tree by index */
1516 	avl_node_t	phyint_avl_by_name;	/* avl tree by name */
1517 	kmutex_t	phyint_lock;
1518 	struct ipsq_s	*phyint_ipsq;		/* back pointer to ipsq */
1519 	struct phyint	*phyint_ipsq_next;	/* phyint list on this ipsq */
1520 } phyint_t;
1521 
1522 #define	CACHE_ALIGN_SIZE 64
1523 
1524 #define	CACHE_ALIGN(align_struct)	P2ROUNDUP(sizeof (struct align_struct),\
1525 							CACHE_ALIGN_SIZE)
1526 struct _phyint_list_s_ {
1527 	avl_tree_t	phyint_list_avl_by_index;	/* avl tree by index */
1528 	avl_tree_t	phyint_list_avl_by_name;	/* avl tree by name */
1529 };
1530 
1531 typedef union phyint_list_u {
1532 	struct	_phyint_list_s_ phyint_list_s;
1533 	char	phyint_list_filler[CACHE_ALIGN(_phyint_list_s_)];
1534 } phyint_list_t;
1535 
1536 #define	phyint_list_avl_by_index	phyint_list_s.phyint_list_avl_by_index
1537 #define	phyint_list_avl_by_name		phyint_list_s.phyint_list_avl_by_name
1538 /*
1539  * ILL groups. We group ills,
1540  *
1541  * - if the ills have the same group name. (New way)
1542  *
1543  * ill_group locking notes:
1544  *
1545  * illgrp_lock protects ill_grp_ill_schednext.
1546  *
1547  * ill_g_lock protects ill_grp_next, illgrp_ill, illgrp_ill_count.
1548  * Holding ill_g_lock freezes the memberships of ills in IPMP groups.
1549  * It also freezes the global list of ills and all ipifs in all ills.
1550  *
1551  * To remove an ipif from the linked list of ipifs of that ill ipif_free_tail
1552  * holds both ill_g_lock, and ill_lock. Similarly to remove an ill from the
1553  * global list of ills, ill_delete_glist holds ill_g_lock as writer.
1554  * This simplifies things for ipif_select_source, illgrp_scheduler etc.
1555  * that need to walk the members of an illgrp. They just hold ill_g_lock
1556  * as reader to do the walk.
1557  *
1558  */
1559 typedef	struct ill_group {
1560 	kmutex_t	illgrp_lock;
1561 	struct ill_group *illgrp_next;		/* Next ill_group */
1562 	struct ill_s	*illgrp_ill_schednext;	/* Next ill to be scheduled */
1563 	struct ill_s	*illgrp_ill;		/* First ill in the group */
1564 	int		illgrp_ill_count;
1565 } ill_group_t;
1566 
1567 /*
1568  * Fragmentation hash bucket
1569  */
1570 typedef struct ipfb_s {
1571 	struct ipf_s	*ipfb_ipf;	/* List of ... */
1572 	size_t		ipfb_count;	/* Count of bytes used by frag(s) */
1573 	kmutex_t	ipfb_lock;	/* Protect all ipf in list */
1574 	uint_t		ipfb_frag_pkts; /* num of distinct fragmented pkts */
1575 } ipfb_t;
1576 
1577 /*
1578  * IRE bucket structure. Usually there is an array of such structures,
1579  * each pointing to a linked list of ires. irb_refcnt counts the number
1580  * of walkers of a given hash bucket. Usually the reference count is
1581  * bumped up if the walker wants no IRES to be DELETED while walking the
1582  * list. Bumping up does not PREVENT ADDITION. This allows walking a given
1583  * hash bucket without stumbling up on a free pointer.
1584  *
1585  * irb_t structures in ip_ftable are dynamically allocated and freed.
1586  * In order to identify the irb_t structures that can be safely kmem_free'd
1587  * we need to ensure that
1588  *  - the irb_refcnt is quiescent, indicating no other walkers,
1589  *  - no other threads or ire's are holding references to the irb,
1590  *	i.e., irb_nire == 0,
1591  *  - there are no active ire's in the bucket, i.e., irb_ire_cnt == 0
1592  */
1593 typedef struct irb {
1594 	struct ire_s	*irb_ire;	/* First ire in this bucket */
1595 					/* Should be first in this struct */
1596 	krwlock_t	irb_lock;	/* Protect this bucket */
1597 	uint_t		irb_refcnt;	/* Protected by irb_lock */
1598 	uchar_t		irb_marks;	/* CONDEMNED ires in this bucket ? */
1599 #define	IRB_MARK_CONDEMNED	0x0001
1600 #define	IRB_MARK_FTABLE		0x0002
1601 	uint_t		irb_ire_cnt;	/* Num of active IRE in this bucket */
1602 	uint_t		irb_tmp_ire_cnt; /* Num of temporary IRE */
1603 	struct ire_s	*irb_rr_origin;	/* origin for round-robin */
1604 	int		irb_nire;	/* Num of ftable ire's that ref irb */
1605 	ip_stack_t	*irb_ipst;	/* Does not have a netstack_hold */
1606 } irb_t;
1607 
1608 #define	IRB2RT(irb)	(rt_t *)((caddr_t)(irb) - offsetof(rt_t, rt_irb))
1609 
1610 /* The following are return values of ip_xmit_v4() */
1611 typedef enum {
1612 	SEND_PASSED = 0,	 /* sent packet out on wire */
1613 	SEND_FAILED,	 /* sending of packet failed */
1614 	LOOKUP_IN_PROGRESS, /* ire cache found, ARP resolution in progress */
1615 	LLHDR_RESLV_FAILED  /* macaddr resl of onlink dst or nexthop failed */
1616 } ipxmit_state_t;
1617 
1618 #define	IP_V4_G_HEAD	0
1619 #define	IP_V6_G_HEAD	1
1620 
1621 #define	MAX_G_HEADS	2
1622 
1623 /*
1624  * unpadded ill_if structure
1625  */
1626 struct 	_ill_if_s_ {
1627 	union ill_if_u	*illif_next;
1628 	union ill_if_u	*illif_prev;
1629 	avl_tree_t	illif_avl_by_ppa;	/* AVL tree sorted on ppa */
1630 	vmem_t		*illif_ppa_arena;	/* ppa index space */
1631 	uint16_t	illif_mcast_v1;		/* hints for		  */
1632 	uint16_t	illif_mcast_v2;		/* [igmp|mld]_slowtimo	  */
1633 	int		illif_name_len;		/* name length */
1634 	char		illif_name[LIFNAMSIZ];	/* name of interface type */
1635 };
1636 
1637 /* cache aligned ill_if structure */
1638 typedef union 	ill_if_u {
1639 	struct  _ill_if_s_ ill_if_s;
1640 	char 	illif_filler[CACHE_ALIGN(_ill_if_s_)];
1641 } ill_if_t;
1642 
1643 #define	illif_next		ill_if_s.illif_next
1644 #define	illif_prev		ill_if_s.illif_prev
1645 #define	illif_avl_by_ppa	ill_if_s.illif_avl_by_ppa
1646 #define	illif_ppa_arena		ill_if_s.illif_ppa_arena
1647 #define	illif_mcast_v1		ill_if_s.illif_mcast_v1
1648 #define	illif_mcast_v2		ill_if_s.illif_mcast_v2
1649 #define	illif_name		ill_if_s.illif_name
1650 #define	illif_name_len		ill_if_s.illif_name_len
1651 
1652 typedef struct ill_walk_context_s {
1653 	int	ctx_current_list; /* current list being searched */
1654 	int	ctx_last_list;	 /* last list to search */
1655 } ill_walk_context_t;
1656 
1657 /*
1658  * ill_g_heads structure, one for IPV4 and one for IPV6
1659  */
1660 struct _ill_g_head_s_ {
1661 	ill_if_t	*ill_g_list_head;
1662 	ill_if_t	*ill_g_list_tail;
1663 };
1664 
1665 typedef union ill_g_head_u {
1666 	struct _ill_g_head_s_ ill_g_head_s;
1667 	char	ill_g_head_filler[CACHE_ALIGN(_ill_g_head_s_)];
1668 } ill_g_head_t;
1669 
1670 #define	ill_g_list_head	ill_g_head_s.ill_g_list_head
1671 #define	ill_g_list_tail	ill_g_head_s.ill_g_list_tail
1672 
1673 #define	IP_V4_ILL_G_LIST(ipst)	\
1674 	(ipst)->ips_ill_g_heads[IP_V4_G_HEAD].ill_g_list_head
1675 #define	IP_V6_ILL_G_LIST(ipst)	\
1676 	(ipst)->ips_ill_g_heads[IP_V6_G_HEAD].ill_g_list_head
1677 #define	IP_VX_ILL_G_LIST(i, ipst)	\
1678 	(ipst)->ips_ill_g_heads[i].ill_g_list_head
1679 
1680 #define	ILL_START_WALK_V4(ctx_ptr, ipst)	\
1681 	ill_first(IP_V4_G_HEAD, IP_V4_G_HEAD, ctx_ptr, ipst)
1682 #define	ILL_START_WALK_V6(ctx_ptr, ipst)	\
1683 	ill_first(IP_V6_G_HEAD, IP_V6_G_HEAD, ctx_ptr, ipst)
1684 #define	ILL_START_WALK_ALL(ctx_ptr, ipst)	\
1685 	ill_first(MAX_G_HEADS, MAX_G_HEADS, ctx_ptr, ipst)
1686 
1687 /*
1688  * Capabilities, possible flags for ill_capabilities.
1689  */
1690 
1691 #define	ILL_CAPAB_AH		0x01		/* IPsec AH acceleration */
1692 #define	ILL_CAPAB_ESP		0x02		/* IPsec ESP acceleration */
1693 #define	ILL_CAPAB_MDT		0x04		/* Multidata Transmit */
1694 #define	ILL_CAPAB_HCKSUM	0x08		/* Hardware checksumming */
1695 #define	ILL_CAPAB_ZEROCOPY	0x10		/* Zero-copy */
1696 #define	ILL_CAPAB_POLL		0x20		/* Polling Toggle */
1697 #define	ILL_CAPAB_SOFT_RING	0x40		/* Soft_Ring capability */
1698 #define	ILL_CAPAB_LSO		0x80		/* Large Segment Offload */
1699 
1700 /*
1701  * Per-ill Multidata Transmit capabilities.
1702  */
1703 typedef struct ill_mdt_capab_s ill_mdt_capab_t;
1704 
1705 /*
1706  * Per-ill IPsec capabilities.
1707  */
1708 typedef struct ill_ipsec_capab_s ill_ipsec_capab_t;
1709 
1710 /*
1711  * Per-ill Hardware Checksumming capbilities.
1712  */
1713 typedef struct ill_hcksum_capab_s ill_hcksum_capab_t;
1714 
1715 /*
1716  * Per-ill Zero-copy capabilities.
1717  */
1718 typedef struct ill_zerocopy_capab_s ill_zerocopy_capab_t;
1719 
1720 /*
1721  * Per-ill Polling/soft ring capbilities.
1722  */
1723 typedef struct ill_dls_capab_s ill_dls_capab_t;
1724 
1725 /*
1726  * Per-ill polling resource map.
1727  */
1728 typedef struct ill_rx_ring ill_rx_ring_t;
1729 
1730 /*
1731  * Per-ill Large Segment Offload capabilities.
1732  */
1733 typedef struct ill_lso_capab_s ill_lso_capab_t;
1734 
1735 /* The following are ill_state_flags */
1736 #define	ILL_LL_SUBNET_PENDING	0x01	/* Waiting for DL_INFO_ACK from drv */
1737 #define	ILL_CONDEMNED		0x02	/* No more new ref's to the ILL */
1738 #define	ILL_CHANGING		0x04	/* ILL not globally visible */
1739 #define	ILL_DL_UNBIND_IN_PROGRESS	0x08	/* UNBIND_REQ is sent */
1740 #define	ILL_SOFT_RING_ASSIGN	0x10	/* Makeing soft ring assigment */
1741 
1742 /* Is this an ILL whose source address is used by other ILL's ? */
1743 #define	IS_USESRC_ILL(ill)			\
1744 	(((ill)->ill_usesrc_ifindex == 0) &&	\
1745 	((ill)->ill_usesrc_grp_next != NULL))
1746 
1747 /* Is this a client/consumer of the usesrc ILL ? */
1748 #define	IS_USESRC_CLI_ILL(ill)			\
1749 	(((ill)->ill_usesrc_ifindex != 0) &&	\
1750 	((ill)->ill_usesrc_grp_next != NULL))
1751 
1752 /* Is this an virtual network interface (vni) ILL ? */
1753 #define	IS_VNI(ill)							     \
1754 	(((ill) != NULL) &&						     \
1755 	(((ill)->ill_phyint->phyint_flags & (PHYI_LOOPBACK|PHYI_VIRTUAL)) == \
1756 	PHYI_VIRTUAL))
1757 
1758 /*
1759  * IP Lower level Structure.
1760  * Instance data structure in ip_open when there is a device below us.
1761  */
1762 typedef struct ill_s {
1763 	ill_if_t *ill_ifptr;		/* pointer to interface type */
1764 	queue_t	*ill_rq;		/* Read queue. */
1765 	queue_t	*ill_wq;		/* Write queue. */
1766 
1767 	int	ill_error;		/* Error value sent up by device. */
1768 
1769 	ipif_t	*ill_ipif;		/* Interface chain for this ILL. */
1770 
1771 	uint_t	ill_ipif_up_count;	/* Number of IPIFs currently up. */
1772 	uint_t	ill_max_frag;		/* Max IDU from DLPI. */
1773 	char	*ill_name;		/* Our name. */
1774 	uint_t	ill_ipif_dup_count;	/* Number of duplicate addresses. */
1775 	uint_t	ill_name_length;	/* Name length, incl. terminator. */
1776 	char	*ill_ndd_name;		/* Name + ":ip?_forwarding" for NDD. */
1777 	uint_t	ill_net_type;		/* IRE_IF_RESOLVER/IRE_IF_NORESOLVER. */
1778 	/*
1779 	 * Physical Point of Attachment num.  If DLPI style 1 provider
1780 	 * then this is derived from the devname.
1781 	 */
1782 	uint_t	ill_ppa;
1783 	t_uscalar_t	ill_sap;
1784 	t_scalar_t	ill_sap_length;	/* Including sign (for position) */
1785 	uint_t	ill_phys_addr_length;	/* Excluding the sap. */
1786 	uint_t	ill_bcast_addr_length;	/* Only set when the DL provider */
1787 					/* supports broadcast. */
1788 	t_uscalar_t	ill_mactype;
1789 	uint8_t	*ill_frag_ptr;		/* Reassembly state. */
1790 	timeout_id_t ill_frag_timer_id; /* timeout id for the frag timer */
1791 	ipfb_t	*ill_frag_hash_tbl;	/* Fragment hash list head. */
1792 	ipif_t	*ill_pending_ipif;	/* IPIF waiting for DL operation. */
1793 
1794 	ilm_t	*ill_ilm;		/* Multicast mebership for lower ill */
1795 	uint_t	ill_global_timer;	/* for IGMPv3/MLDv2 general queries */
1796 	int	ill_mcast_type;		/* type of router which is querier */
1797 					/* on this interface */
1798 	uint16_t ill_mcast_v1_time;	/* # slow timeouts since last v1 qry */
1799 	uint16_t ill_mcast_v2_time;	/* # slow timeouts since last v2 qry */
1800 	uint8_t	ill_mcast_v1_tset;	/* 1 => timer is set; 0 => not set */
1801 	uint8_t	ill_mcast_v2_tset;	/* 1 => timer is set; 0 => not set */
1802 
1803 	uint8_t	ill_mcast_rv;		/* IGMPv3/MLDv2 robustness variable */
1804 	int	ill_mcast_qi;		/* IGMPv3/MLDv2 query interval var */
1805 
1806 	mblk_t	*ill_pending_mp;	/* IOCTL/DLPI awaiting completion. */
1807 	/*
1808 	 * All non-NULL cells between 'ill_first_mp_to_free' and
1809 	 * 'ill_last_mp_to_free' are freed in ill_delete.
1810 	 */
1811 #define	ill_first_mp_to_free	ill_bcast_mp
1812 	mblk_t	*ill_bcast_mp;		/* DLPI header for broadcasts. */
1813 	mblk_t	*ill_resolver_mp;	/* Resolver template. */
1814 	mblk_t	*ill_detach_mp;		/* detach mp, or NULL if style1 */
1815 	mblk_t	*ill_unbind_mp;		/* unbind mp from ill_dl_up() */
1816 	mblk_t	*ill_dlpi_deferred;	/* b_next chain of control messages */
1817 	mblk_t	*ill_phys_addr_mp;	/* mblk which holds ill_phys_addr */
1818 #define	ill_last_mp_to_free	ill_phys_addr_mp
1819 
1820 	cred_t	*ill_credp;		/* opener's credentials */
1821 	uint8_t	*ill_phys_addr;		/* ill_phys_addr_mp->b_rptr + off */
1822 
1823 	uint_t	ill_state_flags;	/* see ILL_* flags above */
1824 
1825 	/* Following bit fields protected by ipsq_t */
1826 	uint_t
1827 		ill_needs_attach : 1,
1828 		ill_reserved : 1,
1829 		ill_isv6 : 1,
1830 		ill_dlpi_style_set : 1,
1831 
1832 		ill_ifname_pending : 1,
1833 		ill_move_in_progress : 1, /* FAILOVER/FAILBACK in progress */
1834 		ill_join_allmulti : 1,
1835 		ill_logical_down : 1,
1836 
1837 		ill_is_6to4tun : 1,	/* Interface is a 6to4 tunnel */
1838 		ill_promisc_on_phys : 1, /* phys interface in promisc mode */
1839 		ill_dl_up : 1,
1840 		ill_up_ipifs : 1,
1841 
1842 		ill_note_link : 1,	/* supports link-up notification */
1843 
1844 		ill_pad_to_bit_31 : 19;
1845 
1846 	/* Following bit fields protected by ill_lock */
1847 	uint_t
1848 		ill_fragtimer_executing : 1,
1849 		ill_fragtimer_needrestart : 1,
1850 		ill_ilm_cleanup_reqd : 1,
1851 		ill_arp_closing : 1,
1852 
1853 		ill_arp_bringup_pending : 1,
1854 		ill_mtu_userspecified : 1, /* SIOCSLNKINFO has set the mtu */
1855 		ill_arp_extend : 1,	/* ARP has DAD extensions */
1856 		ill_pad_bit_31 : 25;
1857 
1858 	/*
1859 	 * Used in SIOCSIFMUXID and SIOCGIFMUXID for 'ifconfig unplumb'.
1860 	 */
1861 	int	ill_arp_muxid;		/* muxid returned from plink for arp */
1862 	int	ill_ip_muxid;		/* muxid returned from plink for ip */
1863 
1864 	/*
1865 	 * NIC event information attached, to be used by nic event hooks.
1866 	 */
1867 	hook_nic_event_t	*ill_nic_event_info;
1868 
1869 	/*
1870 	 * Used for IP frag reassembly throttling on a per ILL basis.
1871 	 *
1872 	 * Note: frag_count is approximate, its added to and subtracted from
1873 	 *	 without any locking, so simultaneous load/modify/stores can
1874 	 *	 collide, also ill_frag_purge() recalculates its value by
1875 	 *	 summing all the ipfb_count's without locking out updates
1876 	 *	 to the ipfb's.
1877 	 */
1878 	uint_t	ill_ipf_gen;		/* Generation of next fragment queue */
1879 	uint_t	ill_frag_count;		/* Approx count of all mblk bytes */
1880 	uint_t	ill_frag_free_num_pkts;	 /* num of fragmented packets to free */
1881 	clock_t	ill_last_frag_clean_time; /* time when frag's were pruned */
1882 	int	ill_type;		/* From <net/if_types.h> */
1883 	uint_t	ill_dlpi_multicast_state;	/* See below IDS_* */
1884 	uint_t	ill_dlpi_fastpath_state;	/* See below IDS_* */
1885 
1886 	/*
1887 	 * Capabilities related fields.
1888 	 */
1889 	uint_t  ill_dlpi_capab_state;	/* State of capability query, IDS_* */
1890 	uint64_t ill_capabilities;	/* Enabled capabilities, ILL_CAPAB_* */
1891 	ill_mdt_capab_t	*ill_mdt_capab;	/* Multidata Transmit capabilities */
1892 	ill_ipsec_capab_t *ill_ipsec_capab_ah;	/* IPsec AH capabilities */
1893 	ill_ipsec_capab_t *ill_ipsec_capab_esp;	/* IPsec ESP capabilities */
1894 	ill_hcksum_capab_t *ill_hcksum_capab; /* H/W cksumming capabilities */
1895 	ill_zerocopy_capab_t *ill_zerocopy_capab; /* Zero-copy capabilities */
1896 	ill_dls_capab_t *ill_dls_capab; /* Polling, soft ring capabilities */
1897 	ill_lso_capab_t *ill_lso_capab; /* Large Segment Offload capabilities */
1898 
1899 	/*
1900 	 * New fields for IPv6
1901 	 */
1902 	uint8_t	ill_max_hops;	/* Maximum hops for any logical interface */
1903 	uint_t	ill_max_mtu;	/* Maximum MTU for any logical interface */
1904 	uint32_t ill_reachable_time;	/* Value for ND algorithm in msec */
1905 	uint32_t ill_reachable_retrans_time; /* Value for ND algorithm msec */
1906 	uint_t	ill_max_buf;		/* Max # of req to buffer for ND */
1907 	in6_addr_t	ill_token;
1908 	uint_t		ill_token_length;
1909 	uint32_t	ill_xmit_count;		/* ndp max multicast xmits */
1910 	mib2_ipIfStatsEntry_t	*ill_ip_mib;	/* ver indep. interface mib */
1911 	mib2_ipv6IfIcmpEntry_t	*ill_icmp6_mib;	/* Per interface mib */
1912 	/*
1913 	 * Following two mblks are allocated common to all
1914 	 * the ipifs when the first interface is coming up.
1915 	 * It is sent up to arp when the last ipif is coming
1916 	 * down.
1917 	 */
1918 	mblk_t			*ill_arp_down_mp;
1919 	mblk_t			*ill_arp_del_mapping_mp;
1920 	/*
1921 	 * Used for implementing IFF_NOARP. As IFF_NOARP is used
1922 	 * to turn off for all the logicals, it is here instead
1923 	 * of the ipif.
1924 	 */
1925 	mblk_t			*ill_arp_on_mp;
1926 	/* Peer ill of an IPMP move operation */
1927 	struct ill_s		*ill_move_peer;
1928 
1929 	phyint_t		*ill_phyint;
1930 	uint64_t		ill_flags;
1931 	ill_group_t		*ill_group;
1932 	struct ill_s		*ill_group_next;
1933 	/*
1934 	 * Reverse tunnel related count. This count
1935 	 * determines how many mobile nodes are using this
1936 	 * ill to send packet to reverse tunnel via foreign
1937 	 * agent. A non-zero count specifies presence of
1938 	 * mobile node(s) using reverse tunnel through this
1939 	 * interface.
1940 	 */
1941 	uint32_t		ill_mrtun_refcnt;
1942 
1943 	/*
1944 	 * This count is bumped up when a route is added with
1945 	 * RTA_SRCIFP bit flag using routing socket.
1946 	 */
1947 	uint32_t		ill_srcif_refcnt;
1948 	/*
1949 	 * Pointer to the special interface based routing table.
1950 	 * This routing table is created dynamically when RTA_SRCIFP
1951 	 * is set by the routing socket.
1952 	 */
1953 	irb_t			*ill_srcif_table;
1954 	kmutex_t	ill_lock;	/* Please see table below */
1955 	/*
1956 	 * The ill_nd_lla* fields handle the link layer address option
1957 	 * from neighbor discovery. This is used for external IPv6
1958 	 * address resolution.
1959 	 */
1960 	mblk_t		*ill_nd_lla_mp;	/* mblk which holds ill_nd_lla */
1961 	uint8_t		*ill_nd_lla;	/* Link Layer Address */
1962 	uint_t		ill_nd_lla_len;	/* Link Layer Address length */
1963 	/*
1964 	 * We now have 3 phys_addr_req's sent down. This field keeps track
1965 	 * of which one is pending.
1966 	 */
1967 	t_uscalar_t	ill_phys_addr_pend; /* which dl_phys_addr_req pending */
1968 	/*
1969 	 * Used to save errors that occur during plumbing
1970 	 */
1971 	uint_t		ill_ifname_pending_err;
1972 	avl_node_t	ill_avl_byppa; /* avl node based on ppa */
1973 	void		*ill_fastpath_list; /* both ire and nce hang off this */
1974 	uint_t		ill_refcnt;	/* active refcnt by threads */
1975 	uint_t		ill_ire_cnt;	/* ires associated with this ill */
1976 	kcondvar_t	ill_cv;
1977 	uint_t		ill_ilm_walker_cnt;	/* snmp ilm walkers */
1978 	uint_t		ill_nce_cnt;	/* nces associated with this ill */
1979 	uint_t		ill_waiters;	/* threads waiting in ipsq_enter */
1980 	/*
1981 	 * Contains the upper read queue pointer of the module immediately
1982 	 * beneath IP.  This field allows IP to validate sub-capability
1983 	 * acknowledgments coming up from downstream.
1984 	 */
1985 	queue_t		*ill_lmod_rq;	/* read queue pointer of module below */
1986 	uint_t		ill_lmod_cnt;	/* number of modules beneath IP */
1987 	ip_m_t		*ill_media;	/* media specific params/functions */
1988 	t_uscalar_t	ill_dlpi_pending; /* Last DLPI primitive issued */
1989 	uint_t		ill_usesrc_ifindex; /* use src addr from this ILL */
1990 	struct ill_s	*ill_usesrc_grp_next; /* Next ILL in the usesrc group */
1991 #ifdef ILL_DEBUG
1992 	th_trace_t	*ill_trace[IP_TR_HASH_MAX];
1993 	boolean_t	ill_trace_disable;	/* True when alloc fails */
1994 #endif
1995 	zoneid_t	ill_zoneid;
1996 	ip_stack_t	*ill_ipst;	/* Corresponds to a netstack_hold */
1997 } ill_t;
1998 
1999 extern	void	ill_delete_glist(ill_t *);
2000 
2001 /*
2002  * The following table lists the protection levels of the various members
2003  * of the ill_t. Same notation as that used for ipif_t above is used.
2004  *
2005  *				Write			Read
2006  *
2007  * ill_ifptr			ill_g_lock + s		Write once
2008  * ill_rq			ipsq			Write once
2009  * ill_wq			ipsq			Write once
2010  *
2011  * ill_error			ipsq			None
2012  * ill_ipif			ill_g_lock + ipsq	ill_g_lock OR ipsq
2013  * ill_ipif_up_count		ill_lock + ipsq		ill_lock
2014  * ill_max_frag			ipsq			Write once
2015  *
2016  * ill_name			ill_g_lock + ipsq	Write once
2017  * ill_name_length		ill_g_lock + ipsq	Write once
2018  * ill_ndd_name			ipsq			Write once
2019  * ill_net_type			ipsq			Write once
2020  * ill_ppa			ill_g_lock + ipsq	Write once
2021  * ill_sap			ipsq + down ill		Write once
2022  * ill_sap_length		ipsq + down ill		Write once
2023  * ill_phys_addr_length		ipsq + down ill		Write once
2024  *
2025  * ill_bcast_addr_length	ipsq			ipsq
2026  * ill_mactype			ipsq			ipsq
2027  * ill_frag_ptr			ipsq			ipsq
2028  *
2029  * ill_frag_timer_id		ill_lock		ill_lock
2030  * ill_frag_hash_tbl		ipsq			up ill
2031  * ill_ilm			ipsq + ill_lock		ill_lock
2032  * ill_mcast_type		ill_lock		ill_lock
2033  * ill_mcast_v1_time		ill_lock		ill_lock
2034  * ill_mcast_v2_time		ill_lock		ill_lock
2035  * ill_mcast_v1_tset		ill_lock		ill_lock
2036  * ill_mcast_v2_tset		ill_lock		ill_lock
2037  * ill_mcast_rv			ill_lock		ill_lock
2038  * ill_mcast_qi			ill_lock		ill_lock
2039  * ill_pending_mp		ill_lock		ill_lock
2040  *
2041  * ill_bcast_mp			ipsq			ipsq
2042  * ill_resolver_mp		ipsq			only when ill is up
2043  * ill_down_mp			ipsq			ipsq
2044  * ill_dlpi_deferred		ipsq			ipsq
2045  * ill_dlpi_pending		ipsq and ill_lock	ipsq or ill_lock
2046  * ill_phys_addr_mp		ipsq + down ill		only when ill is up
2047  * ill_phys_addr		ipsq + down ill		only when ill is up
2048  *
2049  * ill_state_flags		ill_lock		ill_lock
2050  * exclusive bit flags		ipsq_t			ipsq_t
2051  * shared bit flags		ill_lock		ill_lock
2052  *
2053  * ill_arp_muxid		ipsq			Not atomic
2054  * ill_ip_muxid			ipsq			Not atomic
2055  *
2056  * ill_ipf_gen			Not atomic
2057  * ill_frag_count		Approx. not protected
2058  * ill_type			ipsq + down ill		only when ill is up
2059  * ill_dlpi_multicast_state	ill_lock		ill_lock
2060  * ill_dlpi_fastpath_state	ill_lock		ill_lock
2061  * ill_max_hops			ipsq			Not atomic
2062  *
2063  * ill_max_mtu
2064  *
2065  * ill_reachable_time		ipsq + ill_lock		ill_lock
2066  * ill_reachable_retrans_time	ipsq + ill_lock		ill_lock
2067  * ill_max_buf			ipsq + ill_lock		ill_lock
2068  *
2069  * Next 2 fields need ill_lock because of the get ioctls. They should not
2070  * report partially updated results without executing in the ipsq.
2071  * ill_token			ipsq + ill_lock		ill_lock
2072  * ill_token_length		ipsq + ill_lock		ill_lock
2073  * ill_xmit_count		ipsq + down ill		write once
2074  * ill_ip6_mib			ipsq + down ill		only when ill is up
2075  * ill_icmp6_mib		ipsq + down ill		only when ill is up
2076  * ill_arp_down_mp		ipsq			ipsq
2077  * ill_arp_del_mapping_mp	ipsq			ipsq
2078  * ill_arp_on_mp		ipsq			ipsq
2079  * ill_move_peer		ipsq			ipsq
2080  *
2081  * ill_phyint			ipsq, ill_g_lock, ill_lock	Any of them
2082  * ill_flags			ill_lock		ill_lock
2083  * ill_group			ipsq, ill_g_lock, ill_lock	Any of them
2084  * ill_group_next		ipsq, ill_g_lock, ill_lock	Any of them
2085  * ill_mrtun_refcnt		ill_lock		ill_lock
2086  * ill_srcif_refcnt		ill_lock		ill_lock
2087  * ill_srcif_table		ill_lock		ill_lock
2088  * ill_nd_lla_mp		ipsq + down ill		only when ill is up
2089  * ill_nd_lla			ipsq + down ill		only when ill is up
2090  * ill_nd_lla_len		ipsq + down ill		only when ill is up
2091  * ill_phys_addr_pend		ipsq + down ill		only when ill is up
2092  * ill_ifname_pending_err	ipsq			ipsq
2093  * ill_avl_byppa		ipsq, ill_g_lock	write once
2094  *
2095  * ill_fastpath_list		ill_lock		ill_lock
2096  * ill_refcnt			ill_lock		ill_lock
2097  * ill_ire_cnt			ill_lock		ill_lock
2098  * ill_cv			ill_lock		ill_lock
2099  * ill_ilm_walker_cnt		ill_lock		ill_lock
2100  * ill_nce_cnt			ill_lock		ill_lock
2101  * ill_trace			ill_lock		ill_lock
2102  * ill_usesrc_grp_next		ill_g_usesrc_lock	ill_g_usesrc_lock
2103  */
2104 
2105 /*
2106  * For ioctl restart mechanism see ip_reprocess_ioctl()
2107  */
2108 struct ip_ioctl_cmd_s;
2109 
2110 typedef	int (*ifunc_t)(ipif_t *, struct sockaddr_in *, queue_t *, mblk_t *,
2111     struct ip_ioctl_cmd_s *, void *);
2112 
2113 typedef struct ip_ioctl_cmd_s {
2114 	int	ipi_cmd;
2115 	size_t	ipi_copyin_size;
2116 	uint_t	ipi_flags;
2117 	uint_t	ipi_cmd_type;
2118 	ifunc_t	ipi_func;
2119 	ifunc_t	ipi_func_restart;
2120 } ip_ioctl_cmd_t;
2121 
2122 /*
2123  * ipi_cmd_type:
2124  *
2125  * IF_CMD		1	old style ifreq cmd
2126  * LIF_CMD		2	new style lifreq cmd
2127  * MISC_CMD		3	Misc. (non [l]ifreq, tun) cmds
2128  * TUN_CMD		4	tunnel related
2129  */
2130 
2131 enum { IF_CMD = 1, LIF_CMD, MISC_CMD, TUN_CMD };
2132 
2133 #define	IPI_DONTCARE	0	/* For ioctl encoded values that don't matter */
2134 
2135 /* Flag values in ipi_flags */
2136 #define	IPI_PRIV	0x1		/* Root only command */
2137 #define	IPI_MODOK	0x2		/* Permitted on mod instance of IP */
2138 #define	IPI_WR		0x4		/* Need to grab writer access */
2139 #define	IPI_GET_CMD	0x8		/* branch to mi_copyout on success */
2140 #define	IPI_REPL	0x10	/* valid for replacement ipif created in MOVE */
2141 #define	IPI_NULL_BCONT	0x20	/* ioctl has not data and hence no b_cont */
2142 #define	IPI_PASS_DOWN	0x40	/* pass this ioctl down when a module only */
2143 
2144 extern ip_ioctl_cmd_t	ip_ndx_ioctl_table[];
2145 extern ip_ioctl_cmd_t	ip_misc_ioctl_table[];
2146 extern int ip_ndx_ioctl_count;
2147 extern int ip_misc_ioctl_count;
2148 
2149 #define	ILL_CLEAR_MOVE(ill) {				\
2150 	ill_t *peer_ill;				\
2151 							\
2152 	peer_ill = (ill)->ill_move_peer;		\
2153 	ASSERT(peer_ill != NULL);			\
2154 	(ill)->ill_move_in_progress = B_FALSE;		\
2155 	peer_ill->ill_move_in_progress = B_FALSE;	\
2156 	(ill)->ill_move_peer = NULL;			\
2157 	peer_ill->ill_move_peer = NULL;			\
2158 }
2159 
2160 /* Passed down by ARP to IP during I_PLINK/I_PUNLINK */
2161 typedef struct ipmx_s {
2162 	char	ipmx_name[LIFNAMSIZ];		/* if name */
2163 	uint_t
2164 		ipmx_arpdev_stream : 1,		/* This is the arp stream */
2165 		ipmx_notused : 31;
2166 } ipmx_t;
2167 
2168 /*
2169  * State for detecting if a driver supports certain features.
2170  * Support for DL_ENABMULTI_REQ uses ill_dlpi_multicast_state.
2171  * Support for DLPI M_DATA fastpath uses ill_dlpi_fastpath_state.
2172  * Support for DL_CAPABILITY_REQ uses ill_dlpi_capab_state.
2173  */
2174 #define	IDS_UNKNOWN	0	/* No DLPI request sent */
2175 #define	IDS_INPROGRESS	1	/* DLPI request sent */
2176 #define	IDS_OK		2	/* DLPI request completed successfully */
2177 #define	IDS_FAILED	3	/* DLPI request failed */
2178 #define	IDS_RENEG	4	/* Driver asked for a renegotiation */
2179 
2180 /* Named Dispatch Parameter Management Structure */
2181 typedef struct ipparam_s {
2182 	uint_t	ip_param_min;
2183 	uint_t	ip_param_max;
2184 	uint_t	ip_param_value;
2185 	char	*ip_param_name;
2186 } ipparam_t;
2187 
2188 /* Extended NDP Management Structure */
2189 typedef struct ipndp_s {
2190 	ndgetf_t	ip_ndp_getf;
2191 	ndsetf_t	ip_ndp_setf;
2192 	caddr_t		ip_ndp_data;
2193 	char		*ip_ndp_name;
2194 } ipndp_t;
2195 
2196 /*
2197  * The kernel stores security attributes of all gateways in a database made
2198  * up of one or more tsol_gcdb_t elements.  Each tsol_gcdb_t contains the
2199  * security-related credentials of the gateway.  More than one gateways may
2200  * share entries in the database.
2201  *
2202  * The tsol_gc_t structure represents the gateway to credential association,
2203  * and refers to an entry in the database.  One or more tsol_gc_t entities are
2204  * grouped together to form one or more tsol_gcgrp_t, each representing the
2205  * list of security attributes specific to the gateway.  A gateway may be
2206  * associated with at most one credentials group.
2207  */
2208 struct tsol_gcgrp_s;
2209 
2210 extern uchar_t	ip6opt_ls;	/* TX IPv6 enabler */
2211 
2212 /*
2213  * Gateway security credential record.
2214  */
2215 typedef struct tsol_gcdb_s {
2216 	uint_t		gcdb_refcnt;	/* reference count */
2217 	struct rtsa_s	gcdb_attr;	/* security attributes */
2218 #define	gcdb_mask	gcdb_attr.rtsa_mask
2219 #define	gcdb_doi	gcdb_attr.rtsa_doi
2220 #define	gcdb_slrange	gcdb_attr.rtsa_slrange
2221 } tsol_gcdb_t;
2222 
2223 /*
2224  * Gateway to credential association.
2225  */
2226 typedef struct tsol_gc_s {
2227 	uint_t		gc_refcnt;	/* reference count */
2228 	struct tsol_gcgrp_s *gc_grp;	/* pointer to group */
2229 	struct tsol_gc_s *gc_prev;	/* previous in list */
2230 	struct tsol_gc_s *gc_next;	/* next in list */
2231 	tsol_gcdb_t	*gc_db;		/* pointer to actual credentials */
2232 } tsol_gc_t;
2233 
2234 /*
2235  * Gateway credentials group address.
2236  */
2237 typedef struct tsol_gcgrp_addr_s {
2238 	int		ga_af;		/* address family */
2239 	in6_addr_t	ga_addr;	/* IPv4 mapped or IPv6 address */
2240 } tsol_gcgrp_addr_t;
2241 
2242 /*
2243  * Gateway credentials group.
2244  */
2245 typedef struct tsol_gcgrp_s {
2246 	uint_t		gcgrp_refcnt;	/* reference count */
2247 	krwlock_t	gcgrp_rwlock;	/* lock to protect following */
2248 	uint_t		gcgrp_count;	/* number of credentials */
2249 	tsol_gc_t	*gcgrp_head;	/* first credential in list */
2250 	tsol_gc_t	*gcgrp_tail;	/* last credential in list */
2251 	tsol_gcgrp_addr_t gcgrp_addr;	/* next-hop gateway address */
2252 } tsol_gcgrp_t;
2253 
2254 extern kmutex_t gcgrp_lock;
2255 
2256 #define	GC_REFRELE(p) {				\
2257 	ASSERT((p)->gc_grp != NULL);		\
2258 	rw_enter(&(p)->gc_grp->gcgrp_rwlock, RW_WRITER); \
2259 	ASSERT((p)->gc_refcnt > 0);		\
2260 	if (--((p)->gc_refcnt) == 0)		\
2261 		gc_inactive(p);			\
2262 	else					\
2263 		rw_exit(&(p)->gc_grp->gcgrp_rwlock); \
2264 }
2265 
2266 #define	GCGRP_REFHOLD(p) {			\
2267 	mutex_enter(&gcgrp_lock);		\
2268 	++((p)->gcgrp_refcnt);			\
2269 	ASSERT((p)->gcgrp_refcnt != 0);		\
2270 	mutex_exit(&gcgrp_lock);		\
2271 }
2272 
2273 #define	GCGRP_REFRELE(p) {			\
2274 	mutex_enter(&gcgrp_lock);		\
2275 	ASSERT((p)->gcgrp_refcnt > 0);		\
2276 	if (--((p)->gcgrp_refcnt) == 0)		\
2277 		gcgrp_inactive(p);		\
2278 	ASSERT(MUTEX_HELD(&gcgrp_lock));	\
2279 	mutex_exit(&gcgrp_lock);		\
2280 }
2281 
2282 /*
2283  * IRE gateway security attributes structure, pointed to by tsol_ire_gw_secattr
2284  */
2285 struct tsol_tnrhc;
2286 
2287 typedef struct tsol_ire_gw_secattr_s {
2288 	kmutex_t	igsa_lock;	/* lock to protect following */
2289 	struct tsol_tnrhc *igsa_rhc;	/* host entry for gateway */
2290 	tsol_gc_t	*igsa_gc;	/* for prefix IREs */
2291 	tsol_gcgrp_t	*igsa_gcgrp;	/* for cache IREs */
2292 } tsol_ire_gw_secattr_t;
2293 
2294 /*
2295  * Following are the macros to increment/decrement the reference
2296  * count of the IREs and IRBs (ire bucket).
2297  *
2298  * 1) We bump up the reference count of an IRE to make sure that
2299  *    it does not get deleted and freed while we are using it.
2300  *    Typically all the lookup functions hold the bucket lock,
2301  *    and look for the IRE. If it finds an IRE, it bumps up the
2302  *    reference count before dropping the lock. Sometimes we *may* want
2303  *    to bump up the reference count after we *looked* up i.e without
2304  *    holding the bucket lock. So, the IRE_REFHOLD macro does not assert
2305  *    on the bucket lock being held. Any thread trying to delete from
2306  *    the hash bucket can still do so but cannot free the IRE if
2307  *    ire_refcnt is not 0.
2308  *
2309  * 2) We bump up the reference count on the bucket where the IRE resides
2310  *    (IRB), when we want to prevent the IREs getting deleted from a given
2311  *    hash bucket. This makes life easier for ire_walk type functions which
2312  *    wants to walk the IRE list, call a function, but needs to drop
2313  *    the bucket lock to prevent recursive rw_enters. While the
2314  *    lock is dropped, the list could be changed by other threads or
2315  *    the same thread could end up deleting the ire or the ire pointed by
2316  *    ire_next. IRE_REFHOLDing the ire or ire_next is not sufficient as
2317  *    a delete will still remove the ire from the bucket while we have
2318  *    dropped the lock and hence the ire_next would be NULL. Thus, we
2319  *    need a mechanism to prevent deletions from a given bucket.
2320  *
2321  *    To prevent deletions, we bump up the reference count on the
2322  *    bucket. If the bucket is held, ire_delete just marks IRE_MARK_CONDEMNED
2323  *    both on the ire's ire_marks and the bucket's irb_marks. When the
2324  *    reference count on the bucket drops to zero, all the CONDEMNED ires
2325  *    are deleted. We don't have to bump up the reference count on the
2326  *    bucket if we are walking the bucket and never have to drop the bucket
2327  *    lock. Note that IRB_REFHOLD does not prevent addition of new ires
2328  *    in the list. It is okay because addition of new ires will not cause
2329  *    ire_next to point to freed memory. We do IRB_REFHOLD only when
2330  *    all of the 3 conditions are true :
2331  *
2332  *    1) The code needs to walk the IRE bucket from start to end.
2333  *    2) It may have to drop the bucket lock sometimes while doing (1)
2334  *    3) It does not want any ires to be deleted meanwhile.
2335  */
2336 
2337 /*
2338  * Bump up the reference count on the IRE. We cannot assert that the
2339  * bucket lock is being held as it is legal to bump up the reference
2340  * count after the first lookup has returned the IRE without
2341  * holding the lock. Currently ip_wput does this for caching IRE_CACHEs.
2342  */
2343 
2344 #ifndef IRE_DEBUG
2345 
2346 #define	IRE_REFHOLD_NOTR(ire)	IRE_REFHOLD(ire)
2347 #define	IRE_UNTRACE_REF(ire)
2348 #define	IRE_TRACE_REF(ire)
2349 
2350 #else
2351 
2352 #define	IRE_REFHOLD_NOTR(ire) {				\
2353 	atomic_add_32(&(ire)->ire_refcnt, 1);		\
2354 	ASSERT((ire)->ire_refcnt != 0);			\
2355 }
2356 
2357 #define	IRE_UNTRACE_REF(ire)	ire_untrace_ref(ire);
2358 #define	IRE_TRACE_REF(ire)	ire_trace_ref(ire);
2359 #endif
2360 
2361 #define	IRE_REFHOLD(ire) {				\
2362 	atomic_add_32(&(ire)->ire_refcnt, 1);		\
2363 	ASSERT((ire)->ire_refcnt != 0);			\
2364 	IRE_TRACE_REF(ire);				\
2365 }
2366 
2367 #define	IRE_REFHOLD_LOCKED(ire)	{			\
2368 	IRE_TRACE_REF(ire);				\
2369 	(ire)->ire_refcnt++;				\
2370 }
2371 
2372 /*
2373  * Decrement the reference count on the IRE.
2374  * In architectures e.g sun4u, where atomic_add_32_nv is just
2375  * a cas, we need to maintain the right memory barrier semantics
2376  * as that of mutex_exit i.e all the loads and stores should complete
2377  * before the cas is executed. membar_exit() does that here.
2378  *
2379  * NOTE : This macro is used only in places where we want performance.
2380  *	  To avoid bloating the code, we use the function "ire_refrele"
2381  *	  which essentially calls the macro.
2382  */
2383 #ifndef IRE_DEBUG
2384 #define	IRE_REFRELE(ire) {					\
2385 	ASSERT((ire)->ire_refcnt != 0);				\
2386 	membar_exit();						\
2387 	if (atomic_add_32_nv(&(ire)->ire_refcnt, -1) == 0)	\
2388 		ire_inactive(ire);				\
2389 }
2390 #define	IRE_REFRELE_NOTR(ire)	IRE_REFRELE(ire)
2391 #else
2392 #define	IRE_REFRELE(ire) {					\
2393 	if (ire->ire_bucket != NULL)				\
2394 		ire_untrace_ref(ire);				\
2395 	ASSERT((ire)->ire_refcnt != 0);				\
2396 	membar_exit();						\
2397 	if (atomic_add_32_nv(&(ire)->ire_refcnt, -1) == 0)	\
2398 		ire_inactive(ire);				\
2399 }
2400 #define	IRE_REFRELE_NOTR(ire) {				\
2401 	ASSERT((ire)->ire_refcnt != 0);				\
2402 	membar_exit();						\
2403 	if (atomic_add_32_nv(&(ire)->ire_refcnt, -1) == 0)	\
2404 		ire_inactive(ire);				\
2405 }
2406 #endif
2407 
2408 /*
2409  * Bump up the reference count on the hash bucket - IRB to
2410  * prevent ires from being deleted in this bucket.
2411  */
2412 #define	IRB_REFHOLD(irb) {				\
2413 	rw_enter(&(irb)->irb_lock, RW_WRITER);		\
2414 	(irb)->irb_refcnt++;				\
2415 	ASSERT((irb)->irb_refcnt != 0);			\
2416 	rw_exit(&(irb)->irb_lock);			\
2417 }
2418 #define	IRB_REFHOLD_LOCKED(irb) {			\
2419 	ASSERT(RW_WRITE_HELD(&(irb)->irb_lock));	\
2420 	(irb)->irb_refcnt++;				\
2421 	ASSERT((irb)->irb_refcnt != 0);			\
2422 }
2423 
2424 void irb_refrele_ftable(irb_t *);
2425 /*
2426  * Note: when IRB_MARK_FTABLE (i.e., IRE_CACHETABLE entry), the irb_t
2427  * is statically allocated, so that when the irb_refcnt goes to 0,
2428  * we simply clean up the ire list and continue.
2429  */
2430 #define	IRB_REFRELE(irb) {				\
2431 	if ((irb)->irb_marks & IRB_MARK_FTABLE) {	\
2432 		irb_refrele_ftable((irb));		\
2433 	} else {					\
2434 		rw_enter(&(irb)->irb_lock, RW_WRITER);		\
2435 		ASSERT((irb)->irb_refcnt != 0);			\
2436 		if (--(irb)->irb_refcnt	== 0 &&			\
2437 		    ((irb)->irb_marks & IRE_MARK_CONDEMNED)) {	\
2438 			ire_t *ire_list;			\
2439 								\
2440 			ire_list = ire_unlink(irb);		\
2441 			rw_exit(&(irb)->irb_lock);		\
2442 			ASSERT(ire_list != NULL);		\
2443 			ire_cleanup(ire_list);			\
2444 		} else {					\
2445 			rw_exit(&(irb)->irb_lock);		\
2446 		}						\
2447 	}							\
2448 }
2449 
2450 extern struct kmem_cache *rt_entry_cache;
2451 
2452 /*
2453  * Lock the fast path mp for access, since the fp_mp can be deleted
2454  * due a DL_NOTE_FASTPATH_FLUSH in the case of IRE_BROADCAST and IRE_MIPRTUN
2455  */
2456 
2457 #define	LOCK_IRE_FP_MP(ire) {				\
2458 		if ((ire)->ire_type == IRE_BROADCAST ||	\
2459 		    (ire)->ire_type == IRE_MIPRTUN)	\
2460 			mutex_enter(&ire->ire_nce->nce_lock);	\
2461 	}
2462 #define	UNLOCK_IRE_FP_MP(ire) {				\
2463 		if ((ire)->ire_type == IRE_BROADCAST ||	\
2464 		    (ire)->ire_type == IRE_MIPRTUN)	\
2465 			mutex_exit(&ire->ire_nce->nce_lock);	\
2466 	}
2467 
2468 typedef struct ire4 {
2469 	ipaddr_t ire4_src_addr;		/* Source address to use. */
2470 	ipaddr_t ire4_mask;		/* Mask for matching this IRE. */
2471 	ipaddr_t ire4_addr;		/* Address this IRE represents. */
2472 	ipaddr_t ire4_gateway_addr;	/* Gateway if IRE_CACHE/IRE_OFFSUBNET */
2473 	ipaddr_t ire4_cmask;		/* Mask from parent prefix route */
2474 } ire4_t;
2475 
2476 typedef struct ire6 {
2477 	in6_addr_t ire6_src_addr;	/* Source address to use. */
2478 	in6_addr_t ire6_mask;		/* Mask for matching this IRE. */
2479 	in6_addr_t ire6_addr;		/* Address this IRE represents. */
2480 	in6_addr_t ire6_gateway_addr;	/* Gateway if IRE_CACHE/IRE_OFFSUBNET */
2481 	in6_addr_t ire6_cmask;		/* Mask from parent prefix route */
2482 } ire6_t;
2483 
2484 typedef union ire_addr {
2485 	ire6_t	ire6_u;
2486 	ire4_t	ire4_u;
2487 } ire_addr_u_t;
2488 
2489 /* Internet Routing Entry */
2490 typedef struct ire_s {
2491 	struct	ire_s	*ire_next;	/* The hash chain must be first. */
2492 	struct	ire_s	**ire_ptpn;	/* Pointer to previous next. */
2493 	uint32_t	ire_refcnt;	/* Number of references */
2494 	mblk_t		*ire_mp;	/* Non-null if allocated as mblk */
2495 	queue_t		*ire_rfq;	/* recv from this queue */
2496 	queue_t		*ire_stq;	/* send to this queue */
2497 	union {
2498 		uint_t	*max_fragp;	/* Used only during ire creation */
2499 		uint_t	max_frag;	/* MTU (next hop or path). */
2500 	} imf_u;
2501 #define	ire_max_frag	imf_u.max_frag
2502 #define	ire_max_fragp	imf_u.max_fragp
2503 	uint32_t	ire_frag_flag;	/* IPH_DF or zero. */
2504 	uint32_t	ire_ident;	/* Per IRE IP ident. */
2505 	uint32_t	ire_tire_mark;	/* Used for reclaim of unused. */
2506 	uchar_t		ire_ipversion;	/* IPv4/IPv6 version */
2507 	uchar_t		ire_marks;	/* IRE_MARK_CONDEMNED etc. */
2508 	ushort_t	ire_type;	/* Type of IRE */
2509 	uint_t	ire_ib_pkt_count;	/* Inbound packets for ire_addr */
2510 	uint_t	ire_ob_pkt_count;	/* Outbound packets to ire_addr */
2511 	uint_t	ire_ll_hdr_length;	/* Non-zero if we do M_DATA prepends */
2512 	time_t	ire_create_time;	/* Time (in secs) IRE was created. */
2513 	uint32_t	ire_phandle;	/* Associate prefix IREs to cache */
2514 	uint32_t	ire_ihandle;	/* Associate interface IREs to cache */
2515 	ipif_t		*ire_ipif;	/* the interface that this ire uses */
2516 	uint32_t	ire_flags;	/* flags related to route (RTF_*) */
2517 	uint_t	ire_ipsec_overhead;	/* IPSEC overhead */
2518 	/*
2519 	 * Neighbor Cache Entry for IPv6; arp info for IPv4
2520 	 */
2521 	struct	nce_s	*ire_nce;
2522 	uint_t		ire_masklen;	/* # bits in ire_mask{,_v6} */
2523 	ire_addr_u_t	ire_u;		/* IPv4/IPv6 address info. */
2524 
2525 	irb_t		*ire_bucket;	/* Hash bucket when ire_ptphn is set */
2526 	iulp_t		ire_uinfo;	/* Upper layer protocol info. */
2527 	/*
2528 	 * Protects ire_uinfo, ire_max_frag, and ire_frag_flag.
2529 	 */
2530 	kmutex_t	ire_lock;
2531 	uint_t		ire_ipif_seqid; /* ipif_seqid of ire_ipif */
2532 	/*
2533 	 * For regular routes in forwarding table and cache table the
2534 	 * the following entries are NULL/zero. Only reverse tunnel
2535 	 * table and interface based forwarding table use these fields.
2536 	 * Routes added with RTA_SRCIFP and RTA_SRC respectively have
2537 	 * non-zero values for the following fields.
2538 	 */
2539 	ill_t		*ire_in_ill;	/* Incoming ill interface */
2540 	ipaddr_t	ire_in_src_addr;
2541 					/* source ip-addr of incoming packet */
2542 	clock_t		ire_last_used_time;	/* Last used time */
2543 	tsol_ire_gw_secattr_t *ire_gw_secattr; /* gateway security attributes */
2544 	zoneid_t	ire_zoneid;	/* for local address discrimination */
2545 	/*
2546 	 * ire's that are embedded inside mblk_t and sent to the external
2547 	 * resolver use the ire_stq_ifindex to track the ifindex of the
2548 	 * ire_stq, so that the ill (if it exists) can be correctly recovered
2549 	 * for cleanup in the esbfree routine when arp failure occurs
2550 	 */
2551 	uint_t	ire_stq_ifindex;
2552 	uint_t		ire_defense_count;	/* number of ARP conflicts */
2553 	uint_t		ire_defense_time;	/* last time defended (secs) */
2554 	ip_stack_t	*ire_ipst;	/* Does not have a netstack_hold */
2555 #ifdef IRE_DEBUG
2556 	th_trace_t	*ire_trace[IP_TR_HASH_MAX];
2557 	boolean_t	ire_trace_disable;	/* True when alloc fails */
2558 #endif
2559 } ire_t;
2560 
2561 /* IPv4 compatiblity macros */
2562 #define	ire_src_addr		ire_u.ire4_u.ire4_src_addr
2563 #define	ire_mask		ire_u.ire4_u.ire4_mask
2564 #define	ire_addr		ire_u.ire4_u.ire4_addr
2565 #define	ire_gateway_addr	ire_u.ire4_u.ire4_gateway_addr
2566 #define	ire_cmask		ire_u.ire4_u.ire4_cmask
2567 
2568 #define	ire_src_addr_v6		ire_u.ire6_u.ire6_src_addr
2569 #define	ire_mask_v6		ire_u.ire6_u.ire6_mask
2570 #define	ire_addr_v6		ire_u.ire6_u.ire6_addr
2571 #define	ire_gateway_addr_v6	ire_u.ire6_u.ire6_gateway_addr
2572 #define	ire_cmask_v6		ire_u.ire6_u.ire6_cmask
2573 
2574 /* Convenient typedefs for sockaddrs */
2575 typedef	struct sockaddr_in	sin_t;
2576 typedef	struct sockaddr_in6	sin6_t;
2577 
2578 /* Address structure used for internal bind with IP */
2579 typedef struct ipa_conn_s {
2580 	ipaddr_t	ac_laddr;
2581 	ipaddr_t	ac_faddr;
2582 	uint16_t	ac_fport;
2583 	uint16_t	ac_lport;
2584 } ipa_conn_t;
2585 
2586 typedef struct ipa6_conn_s {
2587 	in6_addr_t	ac6_laddr;
2588 	in6_addr_t	ac6_faddr;
2589 	uint16_t	ac6_fport;
2590 	uint16_t	ac6_lport;
2591 } ipa6_conn_t;
2592 
2593 /*
2594  * Using ipa_conn_x_t or ipa6_conn_x_t allows us to modify the behavior of IP's
2595  * bind handler.
2596  */
2597 typedef struct ipa_conn_extended_s {
2598 	uint64_t	acx_flags;
2599 	ipa_conn_t	acx_conn;
2600 } ipa_conn_x_t;
2601 
2602 typedef struct ipa6_conn_extended_s {
2603 	uint64_t	ac6x_flags;
2604 	ipa6_conn_t	ac6x_conn;
2605 } ipa6_conn_x_t;
2606 
2607 /* flag values for ipa_conn_x_t and ipa6_conn_x_t. */
2608 #define	ACX_VERIFY_DST	0x1ULL	/* verify destination address is reachable */
2609 
2610 /* Name/Value Descriptor. */
2611 typedef struct nv_s {
2612 	uint64_t nv_value;
2613 	char	*nv_name;
2614 } nv_t;
2615 
2616 /* IP Forwarding Ticket */
2617 typedef	struct ipftk_s {
2618 	queue_t	*ipftk_queue;
2619 	ipaddr_t ipftk_dst;
2620 } ipftk_t;
2621 
2622 typedef struct ipt_s {
2623 	pfv_t	func;		/* Routine to call */
2624 	uchar_t	*arg;		/* ire or nce passed in */
2625 } ipt_t;
2626 
2627 #define	ILL_FRAG_HASH(s, i) \
2628 	((ntohl(s) ^ ((i) ^ ((i) >> 8))) % ILL_FRAG_HASH_TBL_COUNT)
2629 
2630 /*
2631  * The MAX number of allowed fragmented packets per hash bucket
2632  * calculation is based on the most common mtu size of 1500. This limit
2633  * will work well for other mtu sizes as well.
2634  */
2635 #define	COMMON_IP_MTU 1500
2636 #define	MAX_FRAG_MIN 10
2637 #define	MAX_FRAG_PKTS(ipst)	\
2638 	MAX(MAX_FRAG_MIN, (2 * (ipst->ips_ip_reass_queue_bytes / \
2639 	    (COMMON_IP_MTU * ILL_FRAG_HASH_TBL_COUNT))))
2640 
2641 /*
2642  * Maximum dups allowed per packet.
2643  */
2644 extern uint_t ip_max_frag_dups;
2645 
2646 /*
2647  * Per-packet information for received packets and transmitted.
2648  * Used by the transport protocols when converting between the packet
2649  * and ancillary data and socket options.
2650  *
2651  * Note: This private data structure and related IPPF_* constant
2652  * definitions are exposed to enable compilation of some debugging tools
2653  * like lsof which use struct tcp_t in <inet/tcp.h>. This is intended to be
2654  * a temporary hack and long term alternate interfaces should be defined
2655  * to support the needs of such tools and private definitions moved to
2656  * private headers.
2657  */
2658 struct ip6_pkt_s {
2659 	uint_t		ipp_fields;		/* Which fields are valid */
2660 	uint_t		ipp_sticky_ignored;	/* sticky fields to ignore */
2661 	uint_t		ipp_ifindex;		/* pktinfo ifindex */
2662 	in6_addr_t	ipp_addr;		/* pktinfo src/dst addr */
2663 	uint_t		ipp_unicast_hops;	/* IPV6_UNICAST_HOPS */
2664 	uint_t		ipp_multicast_hops;	/* IPV6_MULTICAST_HOPS */
2665 	uint_t		ipp_hoplimit;		/* IPV6_HOPLIMIT */
2666 	uint_t		ipp_hopoptslen;
2667 	uint_t		ipp_rtdstoptslen;
2668 	uint_t		ipp_rthdrlen;
2669 	uint_t		ipp_dstoptslen;
2670 	uint_t		ipp_pathmtulen;
2671 	uint_t		ipp_fraghdrlen;
2672 	ip6_hbh_t	*ipp_hopopts;
2673 	ip6_dest_t	*ipp_rtdstopts;
2674 	ip6_rthdr_t	*ipp_rthdr;
2675 	ip6_dest_t	*ipp_dstopts;
2676 	ip6_frag_t	*ipp_fraghdr;
2677 	struct ip6_mtuinfo *ipp_pathmtu;
2678 	in6_addr_t	ipp_nexthop;		/* Transmit only */
2679 	uint8_t		ipp_tclass;
2680 	int8_t		ipp_use_min_mtu;
2681 };
2682 typedef struct ip6_pkt_s ip6_pkt_t;
2683 
2684 extern void ip6_pkt_free(ip6_pkt_t *);	/* free storage inside ip6_pkt_t */
2685 
2686 /*
2687  * This struct is used by ULP_opt_set() functions to return value of IPv4
2688  * ancillary options. Currently this is only used by udp and icmp and only
2689  * IP_PKTINFO option is supported.
2690  */
2691 typedef struct ip4_pkt_s {
2692 	uint_t		ip4_ill_index;	/* interface index */
2693 	ipaddr_t	ip4_addr;	/* source address */
2694 } ip4_pkt_t;
2695 
2696 /*
2697  * Used by ULP's to pass options info to ip_output
2698  * currently only IP_PKTINFO is supported.
2699  */
2700 typedef struct ip_opt_info_s {
2701 	uint_t ip_opt_ill_index;
2702 	uint_t ip_opt_flags;
2703 } ip_opt_info_t;
2704 
2705 /*
2706  * value for ip_opt_flags
2707  */
2708 #define	IP_VERIFY_SRC	0x1
2709 
2710 /*
2711  * This structure is used to convey information from IP and the ULP.
2712  * Currently used for the IP_RECVSLLA, IP_RECVIF and IP_RECVPKTINFO options.
2713  * The type of information field is set to IN_PKTINFO (i.e inbound pkt info)
2714  */
2715 typedef struct ip_pktinfo {
2716 	uint32_t		ip_pkt_ulp_type;	/* type of info sent */
2717 	uint32_t		ip_pkt_flags;	/* what is sent up by IP */
2718 	uint32_t		ip_pkt_ifindex;	/* inbound interface index */
2719 	struct sockaddr_dl	ip_pkt_slla;	/* has source link layer addr */
2720 	struct in_addr		ip_pkt_match_addr; /* matched address */
2721 } ip_pktinfo_t;
2722 
2723 /*
2724  * flags to tell UDP what IP is sending; in_pkt_flags
2725  */
2726 #define	IPF_RECVIF	0x01	/* inbound interface index */
2727 #define	IPF_RECVSLLA	0x02	/* source link layer address */
2728 /*
2729  * Inbound interface index + matched address.
2730  * Used only by IPV4.
2731  */
2732 #define	IPF_RECVADDR	0x04
2733 
2734 /* ipp_fields values */
2735 #define	IPPF_IFINDEX	0x0001	/* Part of in6_pktinfo: ifindex */
2736 #define	IPPF_ADDR	0x0002	/* Part of in6_pktinfo: src/dst addr */
2737 #define	IPPF_SCOPE_ID	0x0004	/* Add xmit ip6i_t for sin6_scope_id */
2738 #define	IPPF_NO_CKSUM	0x0008	/* Add xmit ip6i_t for IP6I_NO_*_CKSUM */
2739 
2740 #define	IPPF_RAW_CKSUM	0x0010	/* Add xmit ip6i_t for IP6I_RAW_CHECKSUM */
2741 #define	IPPF_HOPLIMIT	0x0020
2742 #define	IPPF_HOPOPTS	0x0040
2743 #define	IPPF_RTHDR	0x0080
2744 
2745 #define	IPPF_RTDSTOPTS	0x0100
2746 #define	IPPF_DSTOPTS	0x0200
2747 #define	IPPF_NEXTHOP	0x0400
2748 #define	IPPF_PATHMTU	0x0800
2749 
2750 #define	IPPF_TCLASS	0x1000
2751 #define	IPPF_DONTFRAG	0x2000
2752 #define	IPPF_USE_MIN_MTU	0x04000
2753 #define	IPPF_MULTICAST_HOPS	0x08000
2754 
2755 #define	IPPF_UNICAST_HOPS	0x10000
2756 #define	IPPF_FRAGHDR		0x20000
2757 
2758 #define	IPPF_HAS_IP6I \
2759 	(IPPF_IFINDEX|IPPF_ADDR|IPPF_NEXTHOP|IPPF_SCOPE_ID| \
2760 	IPPF_NO_CKSUM|IPPF_RAW_CKSUM|IPPF_HOPLIMIT|IPPF_DONTFRAG| \
2761 	IPPF_USE_MIN_MTU|IPPF_MULTICAST_HOPS|IPPF_UNICAST_HOPS)
2762 
2763 #define	TCP_PORTS_OFFSET	0
2764 #define	UDP_PORTS_OFFSET	0
2765 
2766 /*
2767  * lookups return the ill/ipif only if the flags are clear OR Iam writer.
2768  * ill / ipif lookup functions increment the refcnt on the ill / ipif only
2769  * after calling these macros. This ensures that the refcnt on the ipif or
2770  * ill will eventually drop down to zero.
2771  */
2772 #define	ILL_LOOKUP_FAILED	1	/* Used as error code */
2773 #define	IPIF_LOOKUP_FAILED	2	/* Used as error code */
2774 
2775 #define	ILL_CAN_LOOKUP(ill) 						\
2776 	(!((ill)->ill_state_flags & (ILL_CONDEMNED | ILL_CHANGING)) ||	\
2777 	IAM_WRITER_ILL(ill))
2778 
2779 #define	ILL_CAN_WAIT(ill, q)	\
2780 	(((q) != NULL) && !((ill)->ill_state_flags & (ILL_CONDEMNED)))
2781 
2782 #define	ILL_CAN_LOOKUP_WALKER(ill)	\
2783 	(!((ill)->ill_state_flags & ILL_CONDEMNED))
2784 
2785 #define	IPIF_CAN_LOOKUP(ipif)						\
2786 	(!((ipif)->ipif_state_flags & (IPIF_CONDEMNED | IPIF_CHANGING)) || \
2787 	IAM_WRITER_IPIF(ipif))
2788 
2789 /*
2790  * If the parameter 'q' is NULL, the caller is not interested in wait and
2791  * restart of the operation if the ILL or IPIF cannot be looked up when it is
2792  * marked as 'CHANGING'. Typically a thread that tries to send out data  will
2793  * end up passing NULLs as the last 4 parameters to ill_lookup_on_ifindex and
2794  * in this case 'q' is NULL
2795  */
2796 #define	IPIF_CAN_WAIT(ipif, q)	\
2797 	(((q) != NULL) && !((ipif)->ipif_state_flags & (IPIF_CONDEMNED)))
2798 
2799 #define	IPIF_CAN_LOOKUP_WALKER(ipif)					\
2800 	(!((ipif)->ipif_state_flags & (IPIF_CONDEMNED)) ||		\
2801 	IAM_WRITER_IPIF(ipif))
2802 
2803 /*
2804  * These macros are used by critical set ioctls and failover ioctls to
2805  * mark the ipif appropriately before starting the operation and to clear the
2806  * marks after completing the operation.
2807  */
2808 #define	IPIF_UNMARK_MOVING(ipif)                                \
2809 	(ipif)->ipif_state_flags &= ~IPIF_MOVING & ~IPIF_CHANGING;
2810 
2811 #define	ILL_UNMARK_CHANGING(ill)                                \
2812 	(ill)->ill_state_flags &= ~ILL_CHANGING;
2813 
2814 /* Macros used to assert that this thread is a writer  */
2815 #define	IAM_WRITER_IPSQ(ipsq)	((ipsq)->ipsq_writer == curthread)
2816 #define	IAM_WRITER_ILL(ill)					\
2817 	((ill)->ill_phyint->phyint_ipsq->ipsq_writer == curthread)
2818 #define	IAM_WRITER_IPIF(ipif)					\
2819 	((ipif)->ipif_ill->ill_phyint->phyint_ipsq->ipsq_writer == curthread)
2820 
2821 /*
2822  * Grab ill locks in the proper order. The order is highest addressed
2823  * ill is locked first.
2824  */
2825 #define	GRAB_ILL_LOCKS(ill_1, ill_2)				\
2826 {								\
2827 	if ((ill_1) > (ill_2)) {				\
2828 		if (ill_1 != NULL)				\
2829 			mutex_enter(&(ill_1)->ill_lock);	\
2830 		if (ill_2 != NULL)				\
2831 			mutex_enter(&(ill_2)->ill_lock);	\
2832 	} else {						\
2833 		if (ill_2 != NULL)				\
2834 			mutex_enter(&(ill_2)->ill_lock);	\
2835 		if (ill_1 != NULL && ill_1 != ill_2)		\
2836 			mutex_enter(&(ill_1)->ill_lock);	\
2837 	}							\
2838 }
2839 
2840 #define	RELEASE_ILL_LOCKS(ill_1, ill_2)		\
2841 {						\
2842 	if (ill_1 != NULL)			\
2843 		mutex_exit(&(ill_1)->ill_lock);	\
2844 	if (ill_2 != NULL && ill_2 != ill_1)	\
2845 		mutex_exit(&(ill_2)->ill_lock);	\
2846 }
2847 
2848 /* Get the other protocol instance ill */
2849 #define	ILL_OTHER(ill)						\
2850 	((ill)->ill_isv6 ? (ill)->ill_phyint->phyint_illv4 :	\
2851 	    (ill)->ill_phyint->phyint_illv6)
2852 
2853 #define	MATCH_V4_ONLY	0x1
2854 #define	MATCH_V6_ONLY	0x2
2855 #define	MATCH_ILL_ONLY	0x4
2856 
2857 /* ioctl command info: Ioctl properties extracted and stored in here */
2858 typedef struct cmd_info_s
2859 {
2860 	char    ci_groupname[LIFNAMSIZ + 1];	/* SIOCSLIFGROUPNAME */
2861 	ipif_t  *ci_ipif;	/* ipif associated with [l]ifreq ioctl's */
2862 	sin_t	*ci_sin;	/* the sin struct passed down */
2863 	sin6_t	*ci_sin6;	/* the sin6_t struct passed down */
2864 	struct lifreq *ci_lifr;	/* the lifreq struct passed down */
2865 } cmd_info_t;
2866 
2867 /*
2868  * List of AH and ESP IPsec acceleration capable ills
2869  */
2870 typedef struct ipsec_capab_ill_s {
2871 	uint_t ill_index;
2872 	boolean_t ill_isv6;
2873 	struct ipsec_capab_ill_s *next;
2874 } ipsec_capab_ill_t;
2875 
2876 extern struct kmem_cache *ire_cache;
2877 
2878 extern ipaddr_t	ip_g_all_ones;
2879 
2880 extern	uint_t	ip_loopback_mtu;	/* /etc/system */
2881 
2882 extern vmem_t *ip_minor_arena;
2883 
2884 /*
2885  * ip_g_forward controls IP forwarding.  It takes two values:
2886  *	0: IP_FORWARD_NEVER	Don't forward packets ever.
2887  *	1: IP_FORWARD_ALWAYS	Forward packets for elsewhere.
2888  *
2889  * RFC1122 says there must be a configuration switch to control forwarding,
2890  * but that the default MUST be to not forward packets ever.  Implicit
2891  * control based on configuration of multiple interfaces MUST NOT be
2892  * implemented (Section 3.1).  SunOS 4.1 did provide the "automatic" capability
2893  * and, in fact, it was the default.  That capability is now provided in the
2894  * /etc/rc2.d/S69inet script.
2895  */
2896 
2897 #define	ips_ip_respond_to_address_mask_broadcast ips_param_arr[0].ip_param_value
2898 #define	ips_ip_g_resp_to_echo_bcast	ips_param_arr[1].ip_param_value
2899 #define	ips_ip_g_resp_to_echo_mcast	ips_param_arr[2].ip_param_value
2900 #define	ips_ip_g_resp_to_timestamp	ips_param_arr[3].ip_param_value
2901 #define	ips_ip_g_resp_to_timestamp_bcast ips_param_arr[4].ip_param_value
2902 #define	ips_ip_g_send_redirects		ips_param_arr[5].ip_param_value
2903 #define	ips_ip_g_forward_directed_bcast	ips_param_arr[6].ip_param_value
2904 #define	ips_ip_debug			ips_param_arr[7].ip_param_value
2905 #define	ips_ip_mrtdebug			ips_param_arr[8].ip_param_value
2906 #define	ips_ip_timer_interval		ips_param_arr[9].ip_param_value
2907 #define	ips_ip_ire_arp_interval		ips_param_arr[10].ip_param_value
2908 #define	ips_ip_ire_redir_interval	ips_param_arr[11].ip_param_value
2909 #define	ips_ip_def_ttl			ips_param_arr[12].ip_param_value
2910 #define	ips_ip_forward_src_routed	ips_param_arr[13].ip_param_value
2911 #define	ips_ip_wroff_extra		ips_param_arr[14].ip_param_value
2912 #define	ips_ip_ire_pathmtu_interval	ips_param_arr[15].ip_param_value
2913 #define	ips_ip_icmp_return		ips_param_arr[16].ip_param_value
2914 #define	ips_ip_path_mtu_discovery	ips_param_arr[17].ip_param_value
2915 #define	ips_ip_ignore_delete_time	ips_param_arr[18].ip_param_value
2916 #define	ips_ip_ignore_redirect		ips_param_arr[19].ip_param_value
2917 #define	ips_ip_output_queue		ips_param_arr[20].ip_param_value
2918 #define	ips_ip_broadcast_ttl		ips_param_arr[21].ip_param_value
2919 #define	ips_ip_icmp_err_interval	ips_param_arr[22].ip_param_value
2920 #define	ips_ip_icmp_err_burst		ips_param_arr[23].ip_param_value
2921 #define	ips_ip_reass_queue_bytes	ips_param_arr[24].ip_param_value
2922 #define	ips_ip_strict_dst_multihoming	ips_param_arr[25].ip_param_value
2923 #define	ips_ip_addrs_per_if		ips_param_arr[26].ip_param_value
2924 #define	ips_ipsec_override_persocket_policy ips_param_arr[27].ip_param_value
2925 #define	ips_icmp_accept_clear_messages	ips_param_arr[28].ip_param_value
2926 #define	ips_igmp_accept_clear_messages	ips_param_arr[29].ip_param_value
2927 
2928 /* IPv6 configuration knobs */
2929 #define	ips_delay_first_probe_time	ips_param_arr[30].ip_param_value
2930 #define	ips_max_unicast_solicit		ips_param_arr[31].ip_param_value
2931 #define	ips_ipv6_def_hops		ips_param_arr[32].ip_param_value
2932 #define	ips_ipv6_icmp_return		ips_param_arr[33].ip_param_value
2933 #define	ips_ipv6_forward_src_routed	ips_param_arr[34].ip_param_value
2934 #define	ips_ipv6_resp_echo_mcast	ips_param_arr[35].ip_param_value
2935 #define	ips_ipv6_send_redirects		ips_param_arr[36].ip_param_value
2936 #define	ips_ipv6_ignore_redirect	ips_param_arr[37].ip_param_value
2937 #define	ips_ipv6_strict_dst_multihoming	ips_param_arr[38].ip_param_value
2938 #define	ips_ip_ire_reclaim_fraction	ips_param_arr[39].ip_param_value
2939 #define	ips_ipsec_policy_log_interval	ips_param_arr[40].ip_param_value
2940 #define	ips_pim_accept_clear_messages	ips_param_arr[41].ip_param_value
2941 #define	ips_ip_ndp_unsolicit_interval	ips_param_arr[42].ip_param_value
2942 #define	ips_ip_ndp_unsolicit_count	ips_param_arr[43].ip_param_value
2943 #define	ips_ipv6_ignore_home_address_opt ips_param_arr[44].ip_param_value
2944 #define	ips_ip_policy_mask		ips_param_arr[45].ip_param_value
2945 #define	ips_ip_multirt_resolution_interval ips_param_arr[46].ip_param_value
2946 #define	ips_ip_multirt_ttl  		ips_param_arr[47].ip_param_value
2947 #define	ips_ip_multidata_outbound	ips_param_arr[48].ip_param_value
2948 #define	ips_ip_ndp_defense_interval	ips_param_arr[49].ip_param_value
2949 #define	ips_ip_max_temp_idle		ips_param_arr[50].ip_param_value
2950 #define	ips_ip_max_temp_defend		ips_param_arr[51].ip_param_value
2951 #define	ips_ip_max_defend		ips_param_arr[52].ip_param_value
2952 #define	ips_ip_defend_interval		ips_param_arr[53].ip_param_value
2953 #define	ips_ip_dup_recovery		ips_param_arr[54].ip_param_value
2954 #define	ips_ip_restrict_interzone_loopback ips_param_arr[55].ip_param_value
2955 #define	ips_ip_lso_outbound		ips_param_arr[56].ip_param_value
2956 #define	ips_ipv6_drop_inbound_icmpv6	ips_param_arr[57].ip_param_value
2957 
2958 extern int	dohwcksum;	/* use h/w cksum if supported by the h/w */
2959 #ifdef ZC_TEST
2960 extern int	noswcksum;
2961 #endif
2962 
2963 extern char	ipif_loopback_name[];
2964 
2965 extern nv_t	*ire_nv_tbl;
2966 
2967 extern struct module_info ip_mod_info;
2968 
2969 #define	HOOKS4_INTERESTED_PHYSICAL_IN(ipst)	\
2970 	((ipst)->ips_ip4_physical_in_event.he_interested)
2971 #define	HOOKS6_INTERESTED_PHYSICAL_IN(ipst)	\
2972 	((ipst)->ips_ip6_physical_in_event.he_interested)
2973 #define	HOOKS4_INTERESTED_PHYSICAL_OUT(ipst)	\
2974 	((ipst)->ips_ip4_physical_out_event.he_interested)
2975 #define	HOOKS6_INTERESTED_PHYSICAL_OUT(ipst)	\
2976 	((ipst)->ips_ip6_physical_out_event.he_interested)
2977 #define	HOOKS4_INTERESTED_FORWARDING(ipst)	\
2978 	((ipst)->ips_ip4_forwarding_event.he_interested)
2979 #define	HOOKS6_INTERESTED_FORWARDING(ipst)	\
2980 	((ipst)->ips_ip6_forwarding_event.he_interested)
2981 #define	HOOKS4_INTERESTED_LOOPBACK_IN(ipst)	\
2982 	((ipst)->ips_ip4_loopback_in_event.he_interested)
2983 #define	HOOKS6_INTERESTED_LOOPBACK_IN(ipst)	\
2984 	((ipst)->ips_ip6_loopback_in_event.he_interested)
2985 #define	HOOKS4_INTERESTED_LOOPBACK_OUT(ipst)	\
2986 	((ipst)->ips_ip4_loopback_out_event.he_interested)
2987 #define	HOOKS6_INTERESTED_LOOPBACK_OUT(ipst)	\
2988 	((ipst)->ips_ip6_loopback_out_event.he_interested)
2989 
2990 /*
2991  * Hooks marcos used inside of ip
2992  */
2993 #define	IPHA_VHL	ipha_version_and_hdr_length
2994 
2995 #define	FW_HOOKS(_hook, _event, _ilp, _olp, _iph, _fm, _m, ipst)	\
2996 									\
2997 	if ((_hook).he_interested) {	\
2998 		hook_pkt_event_t info;					\
2999 									\
3000 		_NOTE(CONSTCOND)					\
3001 		ASSERT((_ilp != NULL) || (_olp != NULL));		\
3002 									\
3003 		_NOTE(CONSTCOND)					\
3004 		if ((_ilp != NULL) &&					\
3005 		    (((ill_t *)(_ilp))->ill_phyint != NULL))		\
3006 			info.hpe_ifp = (phy_if_t)((ill_t *)		\
3007 			    (_ilp))->ill_phyint->phyint_ifindex;	\
3008 		else							\
3009 			info.hpe_ifp = 0;				\
3010 									\
3011 		_NOTE(CONSTCOND)					\
3012 		if ((_olp != NULL) &&					\
3013 		    (((ill_t *)(_olp))->ill_phyint != NULL))		\
3014 			info.hpe_ofp = (phy_if_t)((ill_t *)		\
3015 			    (_olp))->ill_phyint->phyint_ifindex;	\
3016 		else							\
3017 			info.hpe_ofp = 0;				\
3018 		info.hpe_hdr = _iph;					\
3019 		info.hpe_mp = &(_fm);					\
3020 		info.hpe_mb = _m;					\
3021 		if (hook_run(_event, (hook_data_t)&info,		\
3022 		    ipst->ips_netstack) != 0) {				\
3023 			ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
3024 			    (_hook).he_name, (void *)_fm, (void *)_m));	\
3025 			if (_fm != NULL) {				\
3026 				freemsg(_fm);				\
3027 				_fm = NULL;				\
3028 			}						\
3029 			_iph = NULL;					\
3030 			_m = NULL;					\
3031 		} else {						\
3032 			_iph = info.hpe_hdr;				\
3033 			_m = info.hpe_mb;				\
3034 		}							\
3035 	}
3036 
3037 #define	FW_HOOKS6(_hook, _event, _ilp, _olp, _iph, _fm, _m, ipst)	\
3038 									\
3039 	if ((_hook).he_interested) {	\
3040 		hook_pkt_event_t info;					\
3041 									\
3042 		_NOTE(CONSTCOND)					\
3043 		ASSERT((_ilp != NULL) || (_olp != NULL));		\
3044 									\
3045 		_NOTE(CONSTCOND)					\
3046 		if ((_ilp != NULL) &&					\
3047 		    (((ill_t *)(_ilp))->ill_phyint != NULL))		\
3048 			info.hpe_ifp = (phy_if_t)((ill_t *)		\
3049 			    (_ilp))->ill_phyint->phyint_ifindex;	\
3050 		else							\
3051 			info.hpe_ifp = 0;				\
3052 									\
3053 		_NOTE(CONSTCOND)					\
3054 		if ((_olp != NULL) &&					\
3055 		    (((ill_t *)(_olp))->ill_phyint != NULL))		\
3056 			info.hpe_ofp = (phy_if_t)((ill_t *)		\
3057 			    (_olp))->ill_phyint->phyint_ifindex;	\
3058 		else							\
3059 			info.hpe_ofp = 0;				\
3060 		info.hpe_hdr = _iph;					\
3061 		info.hpe_mp = &(_fm);					\
3062 		info.hpe_mb = _m;					\
3063 		if (hook_run(_event, (hook_data_t)&info,		\
3064 		    ipst->ips_netstack) != 0) {				\
3065 			ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
3066 			    (_hook).he_name, (void *)_fm, (void *)_m));	\
3067 			if (_fm != NULL) {				\
3068 				freemsg(_fm);				\
3069 				_fm = NULL;				\
3070 			}						\
3071 			_iph = NULL;					\
3072 			_m = NULL;					\
3073 		} else {						\
3074 			_iph = info.hpe_hdr;				\
3075 			_m = info.hpe_mb;				\
3076 		}							\
3077 	}
3078 
3079 /*
3080  * Network byte order macros
3081  */
3082 #ifdef	_BIG_ENDIAN
3083 #define	N_IN_CLASSD_NET		IN_CLASSD_NET
3084 #define	N_INADDR_UNSPEC_GROUP	INADDR_UNSPEC_GROUP
3085 #else /* _BIG_ENDIAN */
3086 #define	N_IN_CLASSD_NET		(ipaddr_t)0x000000f0U
3087 #define	N_INADDR_UNSPEC_GROUP	(ipaddr_t)0x000000e0U
3088 #endif /* _BIG_ENDIAN */
3089 #define	CLASSD(addr)	(((addr) & N_IN_CLASSD_NET) == N_INADDR_UNSPEC_GROUP)
3090 
3091 #define	IP_LOOPBACK_ADDR(addr)			\
3092 	((ntohl(addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
3093 
3094 #ifdef DEBUG
3095 /* IPsec HW acceleration debugging support */
3096 
3097 #define	IPSECHW_CAPAB		0x0001	/* capability negotiation */
3098 #define	IPSECHW_SADB		0x0002	/* SADB exchange */
3099 #define	IPSECHW_PKT		0x0004	/* general packet flow */
3100 #define	IPSECHW_PKTIN		0x0008	/* driver in pkt processing details */
3101 #define	IPSECHW_PKTOUT		0x0010	/* driver out pkt processing details */
3102 
3103 #define	IPSECHW_DEBUG(f, x)	if (ipsechw_debug & (f)) { (void) printf x; }
3104 #define	IPSECHW_CALL(f, r, x)	if (ipsechw_debug & (f)) { (void) r x; }
3105 
3106 extern uint32_t ipsechw_debug;
3107 #else
3108 #define	IPSECHW_DEBUG(f, x)	{}
3109 #define	IPSECHW_CALL(f, r, x)	{}
3110 #endif
3111 
3112 extern int	ip_debug;
3113 
3114 #ifdef IP_DEBUG
3115 #include <sys/debug.h>
3116 #include <sys/promif.h>
3117 
3118 #define	ip0dbg(a)	printf a
3119 #define	ip1dbg(a)	if (ip_debug > 2) printf a
3120 #define	ip2dbg(a)	if (ip_debug > 3) printf a
3121 #define	ip3dbg(a)	if (ip_debug > 4) printf a
3122 #else
3123 #define	ip0dbg(a)	/* */
3124 #define	ip1dbg(a)	/* */
3125 #define	ip2dbg(a)	/* */
3126 #define	ip3dbg(a)	/* */
3127 #endif	/* IP_DEBUG */
3128 
3129 /* Default MAC-layer address string length for mac_colon_addr */
3130 #define	MAC_STR_LEN	128
3131 
3132 struct	ipsec_out_s;
3133 
3134 struct	mac_header_info_s;
3135 
3136 extern const char *dlpi_prim_str(int);
3137 extern const char *dlpi_err_str(int);
3138 extern void	ill_frag_timer(void *);
3139 extern ill_t	*ill_first(int, int, ill_walk_context_t *, ip_stack_t *);
3140 extern ill_t	*ill_next(ill_walk_context_t *, ill_t *);
3141 extern void	ill_frag_timer_start(ill_t *);
3142 extern mblk_t	*ip_carve_mp(mblk_t **, ssize_t);
3143 extern mblk_t	*ip_dlpi_alloc(size_t, t_uscalar_t);
3144 extern char	*ip_dot_addr(ipaddr_t, char *);
3145 extern const char *mac_colon_addr(const uint8_t *, size_t, char *, size_t);
3146 extern void	ip_lwput(queue_t *, mblk_t *);
3147 extern boolean_t icmp_err_rate_limit(ip_stack_t *);
3148 extern void	icmp_time_exceeded(queue_t *, mblk_t *, uint8_t, zoneid_t,
3149     ip_stack_t *);
3150 extern void	icmp_unreachable(queue_t *, mblk_t *, uint8_t, zoneid_t,
3151     ip_stack_t *);
3152 extern mblk_t	*ip_add_info(mblk_t *, ill_t *, uint_t, zoneid_t, ip_stack_t *);
3153 extern mblk_t	*ip_bind_v4(queue_t *, mblk_t *, conn_t *);
3154 extern int	ip_bind_connected(conn_t *, mblk_t *, ipaddr_t *, uint16_t,
3155     ipaddr_t, uint16_t, boolean_t, boolean_t, boolean_t,
3156     boolean_t);
3157 extern boolean_t ip_bind_ipsec_policy_set(conn_t *, mblk_t *);
3158 extern int	ip_bind_laddr(conn_t *, mblk_t *, ipaddr_t, uint16_t,
3159     boolean_t, boolean_t, boolean_t);
3160 extern uint_t	ip_cksum(mblk_t *, int, uint32_t);
3161 extern int	ip_close(queue_t *, int);
3162 extern uint16_t	ip_csum_hdr(ipha_t *);
3163 extern void	ip_proto_not_sup(queue_t *, mblk_t *, uint_t, zoneid_t,
3164     ip_stack_t *);
3165 extern void	ip_ire_g_fini(void);
3166 extern void	ip_ire_g_init(void);
3167 extern void	ip_ire_fini(ip_stack_t *);
3168 extern void	ip_ire_init(ip_stack_t *);
3169 extern int	ip_open(queue_t *, dev_t *, int, int, cred_t *);
3170 extern int	ip_reassemble(mblk_t *, ipf_t *, uint_t, boolean_t, ill_t *,
3171     size_t);
3172 extern int	ip_opt_set_ill(conn_t *, int, boolean_t, boolean_t,
3173     int, int, mblk_t *);
3174 extern void	ip_rput(queue_t *, mblk_t *);
3175 extern void	ip_input(ill_t *, ill_rx_ring_t *, mblk_t *,
3176     struct mac_header_info_s *);
3177 extern void	ip_rput_dlpi(queue_t *, mblk_t *);
3178 extern void	ip_rput_forward(ire_t *, ipha_t *, mblk_t *, ill_t *);
3179 extern void	ip_rput_forward_multicast(ipaddr_t, mblk_t *, ipif_t *);
3180 
3181 extern int	ip_snmpmod_close(queue_t *);
3182 extern void	ip_snmpmod_wput(queue_t *, mblk_t *);
3183 extern void	ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *,
3184 		    mib2_ipIfStatsEntry_t *);
3185 extern void	ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *,
3186 		    mib2_ipv6IfIcmpEntry_t *);
3187 extern void	ip_udp_input(queue_t *, mblk_t *, ipha_t *, ire_t *, ill_t *);
3188 extern void	ip_proto_input(queue_t *, mblk_t *, ipha_t *, ire_t *, ill_t *);
3189 extern void	ip_rput_other(ipsq_t *, queue_t *, mblk_t *, void *);
3190 extern void	ip_setqinfo(queue_t *, minor_t, boolean_t, ip_stack_t *);
3191 extern void	ip_trash_ire_reclaim(void *);
3192 extern void	ip_trash_timer_expire(void *);
3193 extern void	ip_wput(queue_t *, mblk_t *);
3194 extern void	ip_output(void *, mblk_t *, void *, int);
3195 extern void	ip_output_options(void *, mblk_t *, void *, int,
3196     ip_opt_info_t *);
3197 extern void	ip_wput_md(queue_t *, mblk_t *, conn_t *);
3198 
3199 extern void	ip_wput_ire(queue_t *, mblk_t *, ire_t *, conn_t *, int,
3200 		    zoneid_t);
3201 extern void	ip_wput_local(queue_t *, ill_t *, ipha_t *, mblk_t *, ire_t *,
3202 		    int, zoneid_t);
3203 extern void	ip_wput_multicast(queue_t *, mblk_t *, ipif_t *, zoneid_t);
3204 extern void	ip_wput_nondata(ipsq_t *, queue_t *, mblk_t *, void *);
3205 extern void	ip_wsrv(queue_t *);
3206 extern char	*ip_nv_lookup(nv_t *, int);
3207 extern boolean_t ip_local_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3208 extern boolean_t ip_remote_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3209 extern ipaddr_t ip_massage_options(ipha_t *, netstack_t *);
3210 extern ipaddr_t ip_net_mask(ipaddr_t);
3211 extern void	ip_newroute(queue_t *, mblk_t *, ipaddr_t, ill_t *, conn_t *,
3212 		    zoneid_t, ip_stack_t *);
3213 extern ipxmit_state_t	ip_xmit_v4(mblk_t *, ire_t *, struct ipsec_out_s *,
3214     boolean_t);
3215 extern int	ip_hdr_complete(ipha_t *, zoneid_t, ip_stack_t *);
3216 
3217 extern struct qinit rinit_ipv6;
3218 extern struct qinit winit_ipv6;
3219 extern struct qinit rinit_tcp;
3220 extern struct qinit rinit_tcp6;
3221 extern struct qinit winit_tcp;
3222 extern struct qinit rinit_acceptor_tcp;
3223 extern struct qinit winit_acceptor_tcp;
3224 
3225 extern void	conn_drain_insert(conn_t *connp);
3226 extern	int	conn_ipsec_length(conn_t *connp);
3227 extern void	ip_wput_ipsec_out(queue_t *, mblk_t *, ipha_t *, ill_t *,
3228     ire_t *);
3229 extern ipaddr_t	ip_get_dst(ipha_t *);
3230 extern int	ipsec_out_extra_length(mblk_t *);
3231 extern int	ipsec_in_extra_length(mblk_t *);
3232 extern mblk_t	*ipsec_in_alloc(boolean_t, netstack_t *);
3233 extern boolean_t ipsec_in_is_secure(mblk_t *);
3234 extern void	ipsec_out_process(queue_t *, mblk_t *, ire_t *, uint_t);
3235 extern void	ipsec_out_to_in(mblk_t *);
3236 extern int	ill_forward_set(queue_t *, mblk_t *, boolean_t, caddr_t);
3237 extern void	ip_fanout_proto_again(mblk_t *, ill_t *, ill_t *, ire_t *);
3238 
3239 extern void	ire_cleanup(ire_t *);
3240 extern void	ire_inactive(ire_t *);
3241 extern boolean_t irb_inactive(irb_t *);
3242 extern ire_t	*ire_unlink(irb_t *);
3243 #ifdef IRE_DEBUG
3244 extern	void	ire_trace_ref(ire_t *ire);
3245 extern	void	ire_untrace_ref(ire_t *ire);
3246 extern	void	ire_thread_exit(ire_t *ire, caddr_t);
3247 #endif
3248 #ifdef ILL_DEBUG
3249 extern	void	ill_trace_cleanup(ill_t *);
3250 extern	void	ipif_trace_cleanup(ipif_t *);
3251 #endif
3252 
3253 extern int	ip_srcid_insert(const in6_addr_t *, zoneid_t, ip_stack_t *);
3254 extern int	ip_srcid_remove(const in6_addr_t *, zoneid_t, ip_stack_t *);
3255 extern void	ip_srcid_find_id(uint_t, in6_addr_t *, zoneid_t, netstack_t *);
3256 extern uint_t	ip_srcid_find_addr(const in6_addr_t *, zoneid_t, netstack_t *);
3257 extern int	ip_srcid_report(queue_t *, mblk_t *, caddr_t, cred_t *);
3258 
3259 extern uint8_t	ipoptp_next(ipoptp_t *);
3260 extern uint8_t	ipoptp_first(ipoptp_t *, ipha_t *);
3261 extern int	ip_opt_get_user(const ipha_t *, uchar_t *);
3262 extern ill_t	*ip_grab_attach_ill(ill_t *, mblk_t *, int, boolean_t,
3263     ip_stack_t *);
3264 extern ire_t	*conn_set_outgoing_ill(conn_t *, ire_t *, ill_t **);
3265 extern int	ipsec_req_from_conn(conn_t *, ipsec_req_t *, int);
3266 extern int	ip_snmp_get(queue_t *q, mblk_t *mctl);
3267 extern int	ip_snmp_set(queue_t *q, int, int, uchar_t *, int);
3268 extern void	ip_process_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3269 extern void	ip_quiesce_conn(conn_t *);
3270 extern  void    ip_reprocess_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3271 extern void	ip_restart_optmgmt(ipsq_t *, queue_t *, mblk_t *, void *);
3272 extern void	ip_ioctl_finish(queue_t *, mblk_t *, int, int, ipsq_t *);
3273 
3274 extern boolean_t ip_cmpbuf(const void *, uint_t, boolean_t, const void *,
3275     uint_t);
3276 extern boolean_t ip_allocbuf(void **, uint_t *, boolean_t, const void *,
3277     uint_t);
3278 extern void	ip_savebuf(void **, uint_t *, boolean_t, const void *, uint_t);
3279 
3280 extern boolean_t	ipsq_pending_mp_cleanup(ill_t *, conn_t *);
3281 extern void	conn_ioctl_cleanup(conn_t *);
3282 extern ill_t	*conn_get_held_ill(conn_t *, ill_t **, int *);
3283 extern ill_t	*ip_newroute_get_dst_ill(ill_t *);
3284 
3285 struct multidata_s;
3286 struct pdesc_s;
3287 
3288 extern mblk_t	*ip_mdinfo_alloc(ill_mdt_capab_t *);
3289 extern mblk_t	*ip_mdinfo_return(ire_t *, conn_t *, char *, ill_mdt_capab_t *);
3290 extern mblk_t	*ip_lsoinfo_alloc(ill_lso_capab_t *);
3291 extern mblk_t	*ip_lsoinfo_return(ire_t *, conn_t *, char *,
3292     ill_lso_capab_t *);
3293 extern uint_t	ip_md_cksum(struct pdesc_s *, int, uint_t);
3294 extern boolean_t ip_md_addr_attr(struct multidata_s *, struct pdesc_s *,
3295 			const mblk_t *);
3296 extern boolean_t ip_md_hcksum_attr(struct multidata_s *, struct pdesc_s *,
3297 			uint32_t, uint32_t, uint32_t, uint32_t);
3298 extern boolean_t ip_md_zcopy_attr(struct multidata_s *, struct pdesc_s *,
3299 			uint_t);
3300 extern mblk_t	*ip_unbind(queue_t *, mblk_t *);
3301 
3302 extern void tnet_init(void);
3303 extern void tnet_fini(void);
3304 
3305 /* Hooks for CGTP (multirt routes) filtering module */
3306 #define	CGTP_FILTER_REV_1	1
3307 #define	CGTP_FILTER_REV_2	2
3308 #define	CGTP_FILTER_REV		CGTP_FILTER_REV_2
3309 
3310 /* cfo_filter, cfo_filter_fp, cfo_filter_v6 hooks return values */
3311 #define	CGTP_IP_PKT_NOT_CGTP	0
3312 #define	CGTP_IP_PKT_PREMIUM	1
3313 #define	CGTP_IP_PKT_DUPLICATE	2
3314 
3315 typedef struct cgtp_filter_ops {
3316 	int	cfo_filter_rev;
3317 	int	(*cfo_change_state)(int);
3318 	int	(*cfo_add_dest_v4)(ipaddr_t, ipaddr_t, ipaddr_t, ipaddr_t);
3319 	int	(*cfo_del_dest_v4)(ipaddr_t, ipaddr_t);
3320 	int	(*cfo_add_dest_v6)(in6_addr_t *, in6_addr_t *, in6_addr_t *,
3321 		    in6_addr_t *);
3322 	int	(*cfo_del_dest_v6)(in6_addr_t *, in6_addr_t *);
3323 	int	(*cfo_filter)(queue_t *, mblk_t *);
3324 	int	(*cfo_filter_fp)(queue_t *, mblk_t *);
3325 	int	(*cfo_filter_v6)(queue_t *, ip6_t *, ip6_frag_t *);
3326 } cgtp_filter_ops_t;
3327 
3328 #define	CGTP_MCAST_SUCCESS	1
3329 
3330 /*
3331  * The separate CGTP module needs these as globals. It uses the first
3332  * to unregister (since there is no ip_cgtp_filter_unregister() function)
3333  * and it uses the second one to verify that the filter has been
3334  * turned off (a ip_cgtp_filter_active() function would be good for that.)
3335  */
3336 extern cgtp_filter_ops_t *ip_cgtp_filter_ops;
3337 extern boolean_t ip_cgtp_filter;
3338 
3339 extern int	ip_cgtp_filter_supported(void);
3340 extern int	ip_cgtp_filter_register(cgtp_filter_ops_t *);
3341 
3342 /* Flags for ire_multirt_lookup() */
3343 
3344 #define	MULTIRT_USESTAMP	0x0001
3345 #define	MULTIRT_SETSTAMP	0x0002
3346 #define	MULTIRT_CACHEGW		0x0004
3347 
3348 /* Debug stuff for multirt route resolution. */
3349 #if defined(DEBUG) && !defined(__lint)
3350 /* Our "don't send, rather drop" flag. */
3351 #define	MULTIRT_DEBUG_FLAG	0x8000
3352 
3353 #define	MULTIRT_TRACE(x)	ip2dbg(x)
3354 
3355 #define	MULTIRT_DEBUG_TAG(mblk)	\
3356 	do { \
3357 		ASSERT(mblk != NULL); \
3358 		MULTIRT_TRACE(("%s[%d]: tagging mblk %p, tag was %d\n", \
3359 		__FILE__, __LINE__, \
3360 		(void *)(mblk), (mblk)->b_flag & MULTIRT_DEBUG_FLAG)); \
3361 		(mblk)->b_flag |= MULTIRT_DEBUG_FLAG; \
3362 	} while (0)
3363 
3364 #define	MULTIRT_DEBUG_UNTAG(mblk) \
3365 	do { \
3366 		ASSERT(mblk != NULL); \
3367 		MULTIRT_TRACE(("%s[%d]: untagging mblk %p, tag was %d\n", \
3368 		__FILE__, __LINE__, \
3369 		(void *)(mblk), (mblk)->b_flag & MULTIRT_DEBUG_FLAG)); \
3370 		(mblk)->b_flag &= ~MULTIRT_DEBUG_FLAG; \
3371 	} while (0)
3372 
3373 #define	MULTIRT_DEBUG_TAGGED(mblk) \
3374 	(((mblk)->b_flag & MULTIRT_DEBUG_FLAG) ? B_TRUE : B_FALSE)
3375 #else
3376 #define	MULTIRT_DEBUG_TAG(mblk)		ASSERT(mblk != NULL)
3377 #define	MULTIRT_DEBUG_UNTAG(mblk)	ASSERT(mblk != NULL)
3378 #define	MULTIRT_DEBUG_TAGGED(mblk)	B_FALSE
3379 #endif
3380 
3381 /*
3382  * Per-ILL Multidata Transmit capabilities.
3383  */
3384 struct ill_mdt_capab_s {
3385 	uint_t ill_mdt_version;  /* interface version */
3386 	uint_t ill_mdt_on;	 /* on/off switch for MDT on this ILL */
3387 	uint_t ill_mdt_hdr_head; /* leading header fragment extra space */
3388 	uint_t ill_mdt_hdr_tail; /* trailing header fragment extra space */
3389 	uint_t ill_mdt_max_pld;	 /* maximum payload buffers per Multidata */
3390 	uint_t ill_mdt_span_limit; /* maximum payload span per packet */
3391 };
3392 
3393 struct ill_hcksum_capab_s {
3394 	uint_t	ill_hcksum_version;	/* interface version */
3395 	uint_t	ill_hcksum_txflags;	/* capabilities on transmit */
3396 };
3397 
3398 struct ill_zerocopy_capab_s {
3399 	uint_t	ill_zerocopy_version;	/* interface version */
3400 	uint_t	ill_zerocopy_flags;	/* capabilities */
3401 };
3402 
3403 struct ill_lso_capab_s {
3404 	uint_t	ill_lso_version;	/* interface version */
3405 	uint_t	ill_lso_on;		/* on/off switch for LSO on this ILL */
3406 	uint_t	ill_lso_flags;		/* capabilities */
3407 	uint_t	ill_lso_max;		/* maximum size of payload */
3408 };
3409 
3410 /* Possible ill_states */
3411 #define	ILL_RING_INPROC		3	/* Being assigned to squeue */
3412 #define	ILL_RING_INUSE		2	/* Already Assigned to Rx Ring */
3413 #define	ILL_RING_BEING_FREED	1	/* Being Unassigned */
3414 #define	ILL_RING_FREE		0	/* Available to be assigned to Ring */
3415 
3416 #define	ILL_MAX_RINGS		256	/* Max num of rx rings we can manage */
3417 #define	ILL_POLLING		0x01	/* Polling in use */
3418 
3419 /*
3420  * These functions pointer types are exported by the mac/dls layer.
3421  * we need to duplicate the definitions here because we cannot
3422  * include mac/dls header files here.
3423  */
3424 typedef void	(*ip_mac_blank_t)(void *, time_t, uint_t);
3425 typedef void	(*ip_dld_tx_t)(void *, mblk_t *);
3426 
3427 typedef void	(*ip_dls_chg_soft_ring_t)(void *, int);
3428 typedef void	(*ip_dls_bind_t)(void *, processorid_t);
3429 typedef void	(*ip_dls_unbind_t)(void *);
3430 
3431 struct ill_rx_ring {
3432 	ip_mac_blank_t		rr_blank; /* Driver interrupt blanking func */
3433 	void			*rr_handle; /* Handle for Rx ring */
3434 	squeue_t		*rr_sqp; /* Squeue the ring is bound to */
3435 	ill_t			*rr_ill; /* back pointer to ill */
3436 	clock_t			rr_poll_time; /* Last lbolt polling was used */
3437 	uint32_t		rr_poll_state; /* polling state flags */
3438 	uint32_t		rr_max_blank_time; /* Max interrupt blank */
3439 	uint32_t		rr_min_blank_time; /* Min interrupt blank */
3440 	uint32_t		rr_max_pkt_cnt; /* Max pkts before interrupt */
3441 	uint32_t		rr_min_pkt_cnt; /* Mix pkts before interrupt */
3442 	uint32_t		rr_normal_blank_time; /* Normal intr freq */
3443 	uint32_t		rr_normal_pkt_cnt; /* Normal intr pkt cnt */
3444 	uint32_t		rr_ring_state; /* State of this ring */
3445 };
3446 
3447 struct ill_dls_capab_s {
3448 	ip_dld_tx_t		ill_tx;		/* Driver Tx routine */
3449 	void			*ill_tx_handle;	/* Driver Tx handle */
3450 	ip_dls_chg_soft_ring_t	ill_dls_change_status;
3451 						/* change soft ring fanout */
3452 	ip_dls_bind_t		ill_dls_bind;	/* to add CPU affinity */
3453 	ip_dls_unbind_t		ill_dls_unbind;	/* remove CPU affinity */
3454 	ill_rx_ring_t		*ill_ring_tbl; /* Ring to Sqp mapping table */
3455 	uint_t			ill_dls_soft_ring_cnt; /* Number of soft ring */
3456 	conn_t			*ill_unbind_conn; /* Conn used during unplumb */
3457 };
3458 
3459 /*
3460  * This message is sent by an upper-layer protocol to tell IP that it knows all
3461  * about labels and will construct them itself.  IP takes the slow path and
3462  * recomputes the label on every packet when this isn't true.
3463  */
3464 #define	IP_ULP_OUT_LABELED		(('O' << 8) + 'L')
3465 typedef struct out_labeled_s {
3466 	uint32_t	out_labeled_type;	/* OUT_LABELED */
3467 	queue_t		*out_qnext;		/* intermediate detection */
3468 } out_labeled_t;
3469 
3470 /*
3471  * IP squeues exports
3472  */
3473 extern int 		ip_squeue_profile;
3474 extern int 		ip_squeue_bind;
3475 extern boolean_t 	ip_squeue_fanout;
3476 extern boolean_t	ip_squeue_soft_ring;
3477 extern uint_t		ip_threads_per_cpu;
3478 extern uint_t		ip_squeues_per_cpu;
3479 extern uint_t		ip_soft_rings_cnt;
3480 
3481 typedef struct squeue_set_s {
3482 	kmutex_t	sqs_lock;
3483 	struct squeue_s	**sqs_list;
3484 	int		sqs_size;
3485 	int		sqs_max_size;
3486 	processorid_t	sqs_bind;
3487 } squeue_set_t;
3488 
3489 #define	IP_SQUEUE_GET(hint) 						\
3490 	((!ip_squeue_fanout) ?	(CPU->cpu_squeue_set->sqs_list[0]) :	\
3491 		ip_squeue_random(hint))
3492 
3493 typedef void (*squeue_func_t)(squeue_t *, mblk_t *, sqproc_t, void *, uint8_t);
3494 
3495 extern void ip_squeue_init(void (*)(squeue_t *));
3496 extern squeue_t	*ip_squeue_random(uint_t);
3497 extern squeue_t *ip_squeue_get(ill_rx_ring_t *);
3498 extern void ip_squeue_get_pkts(squeue_t *);
3499 extern int ip_squeue_bind_set(queue_t *, mblk_t *, char *, caddr_t, cred_t *);
3500 extern int ip_squeue_bind_get(queue_t *, mblk_t *, caddr_t, cred_t *);
3501 extern void ip_squeue_clean(void *, mblk_t *, void *);
3502 extern void ip_resume_tcp_bind(void *, mblk_t *, void *);
3503 extern void	ip_soft_ring_assignment(ill_t *, ill_rx_ring_t *,
3504     mblk_t *, struct mac_header_info_s *);
3505 
3506 extern void tcp_wput(queue_t *, mblk_t *);
3507 
3508 extern int	ip_fill_mtuinfo(struct in6_addr *, in_port_t,
3509 	struct ip6_mtuinfo *, netstack_t *);
3510 extern	ipif_t *conn_get_held_ipif(conn_t *, ipif_t **, int *);
3511 
3512 typedef void    (*ipsq_func_t)(ipsq_t *, queue_t *, mblk_t *, void *);
3513 
3514 /*
3515  * Squeue tags. Tags only need to be unique when the callback function is the
3516  * same to distinguish between different calls, but we use unique tags for
3517  * convenience anyway.
3518  */
3519 #define	SQTAG_IP_INPUT			1
3520 #define	SQTAG_TCP_INPUT_ICMP_ERR	2
3521 #define	SQTAG_TCP6_INPUT_ICMP_ERR	3
3522 #define	SQTAG_IP_TCP_INPUT		4
3523 #define	SQTAG_IP6_TCP_INPUT		5
3524 #define	SQTAG_IP_TCP_CLOSE		6
3525 #define	SQTAG_TCP_OUTPUT		7
3526 #define	SQTAG_TCP_TIMER			8
3527 #define	SQTAG_TCP_TIMEWAIT		9
3528 #define	SQTAG_TCP_ACCEPT_FINISH		10
3529 #define	SQTAG_TCP_ACCEPT_FINISH_Q0	11
3530 #define	SQTAG_TCP_ACCEPT_PENDING	12
3531 #define	SQTAG_TCP_LISTEN_DISCON		13
3532 #define	SQTAG_TCP_CONN_REQ		14
3533 #define	SQTAG_TCP_EAGER_BLOWOFF		15
3534 #define	SQTAG_TCP_EAGER_CLEANUP		16
3535 #define	SQTAG_TCP_EAGER_CLEANUP_Q0	17
3536 #define	SQTAG_TCP_CONN_IND		18
3537 #define	SQTAG_TCP_RSRV			19
3538 #define	SQTAG_TCP_ABORT_BUCKET		20
3539 #define	SQTAG_TCP_REINPUT		21
3540 #define	SQTAG_TCP_REINPUT_EAGER		22
3541 #define	SQTAG_TCP_INPUT_MCTL		23
3542 #define	SQTAG_TCP_RPUTOTHER		24
3543 #define	SQTAG_IP_PROTO_AGAIN		25
3544 #define	SQTAG_IP_FANOUT_TCP		26
3545 #define	SQTAG_IPSQ_CLEAN_RING		27
3546 #define	SQTAG_TCP_WPUT_OTHER		28
3547 #define	SQTAG_TCP_CONN_REQ_UNBOUND	29
3548 #define	SQTAG_TCP_SEND_PENDING		30
3549 #define	SQTAG_BIND_RETRY		31
3550 #define	SQTAG_UDP_FANOUT		32
3551 #define	SQTAG_UDP_INPUT			33
3552 #define	SQTAG_UDP_WPUT			34
3553 #define	SQTAG_UDP_OUTPUT		35
3554 #define	SQTAG_TCP_KSSL_INPUT		36
3555 #define	SQTAG_TCP_DROP_Q0		37
3556 
3557 #define	NOT_OVER_IP(ip_wq)	\
3558 	(ip_wq->q_next != NULL ||	\
3559 	    (ip_wq->q_qinfo->qi_minfo->mi_idname) == NULL ||	\
3560 	    strcmp(ip_wq->q_qinfo->qi_minfo->mi_idname,	\
3561 	    IP_MOD_NAME) != 0 ||	\
3562 	    ip_wq->q_qinfo->qi_minfo->mi_idnum != IP_MOD_ID)
3563 
3564 #endif	/* _KERNEL */
3565 
3566 #ifdef	__cplusplus
3567 }
3568 #endif
3569 
3570 #endif	/* _INET_IP_H */
3571