xref: /illumos-gate/usr/src/uts/common/inet/ip.h (revision 9c3e527f75ba61da54edb0d2417f87f78edbe80f)
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 2008 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 #include <sys/modhash.h>
62 
63 #ifdef DEBUG
64 #define	CONN_DEBUG
65 #endif
66 
67 #define	IP_DEBUG
68 /*
69  * The mt-streams(9F) flags for the IP module; put here so that other
70  * "drivers" that are actually IP (e.g., ICMP, UDP) can use the same set
71  * of flags.
72  */
73 #define	IP_DEVMTFLAGS D_MP
74 #endif	/* _KERNEL */
75 
76 #define	IP_MOD_NAME	"ip"
77 #define	IP_DEV_NAME	"/dev/ip"
78 #define	IP6_DEV_NAME	"/dev/ip6"
79 
80 #define	UDP_MOD_NAME	"udp"
81 #define	UDP_DEV_NAME	"/dev/udp"
82 #define	UDP6_DEV_NAME	"/dev/udp6"
83 
84 #define	TCP_MOD_NAME	"tcp"
85 #define	TCP_DEV_NAME	"/dev/tcp"
86 #define	TCP6_DEV_NAME	"/dev/tcp6"
87 
88 #define	SCTP_MOD_NAME	"sctp"
89 
90 #ifndef	_IPADDR_T
91 #define	_IPADDR_T
92 typedef uint32_t ipaddr_t;
93 #endif
94 
95 /* Number of bits in an address */
96 #define	IP_ABITS		32
97 #define	IPV6_ABITS		128
98 
99 #define	IP_HOST_MASK		(ipaddr_t)0xffffffffU
100 
101 #define	IP_CSUM(mp, off, sum)		(~ip_cksum(mp, off, sum) & 0xFFFF)
102 #define	IP_CSUM_PARTIAL(mp, off, sum)	ip_cksum(mp, off, sum)
103 #define	IP_BCSUM_PARTIAL(bp, len, sum)	bcksum(bp, len, sum)
104 #define	IP_MD_CSUM(pd, off, sum)	(~ip_md_cksum(pd, off, sum) & 0xffff)
105 #define	IP_MD_CSUM_PARTIAL(pd, off, sum) ip_md_cksum(pd, off, sum)
106 
107 /*
108  * Flag to IP write side to indicate that the appln has sent in a pre-built
109  * IP header. Stored in ipha_ident (which is otherwise zero).
110  */
111 #define	IP_HDR_INCLUDED			0xFFFF
112 
113 #define	ILL_FRAG_HASH_TBL_COUNT	((unsigned int)64)
114 #define	ILL_FRAG_HASH_TBL_SIZE	(ILL_FRAG_HASH_TBL_COUNT * sizeof (ipfb_t))
115 
116 #define	IPV4_ADDR_LEN			4
117 #define	IP_ADDR_LEN			IPV4_ADDR_LEN
118 #define	IP_ARP_PROTO_TYPE		0x0800
119 
120 #define	IPV4_VERSION			4
121 #define	IP_VERSION			IPV4_VERSION
122 #define	IP_SIMPLE_HDR_LENGTH_IN_WORDS	5
123 #define	IP_SIMPLE_HDR_LENGTH		20
124 #define	IP_MAX_HDR_LENGTH		60
125 
126 #define	IP_MAX_OPT_LENGTH (IP_MAX_HDR_LENGTH-IP_SIMPLE_HDR_LENGTH)
127 
128 #define	IP_MIN_MTU			(IP_MAX_HDR_LENGTH + 8)	/* 68 bytes */
129 
130 /*
131  * XXX IP_MAXPACKET is defined in <netinet/ip.h> as well. At some point the
132  * 2 files should be cleaned up to remove all redundant definitions.
133  */
134 #define	IP_MAXPACKET			65535
135 #define	IP_SIMPLE_HDR_VERSION \
136 	((IP_VERSION << 4) | IP_SIMPLE_HDR_LENGTH_IN_WORDS)
137 
138 #define	UDPH_SIZE			8
139 
140 /* Leave room for ip_newroute to tack on the src and target addresses */
141 #define	OK_RESOLVER_MP(mp)						\
142 	((mp) && ((mp)->b_wptr - (mp)->b_rptr) >= (2 * IP_ADDR_LEN))
143 
144 /*
145  * Constants and type definitions to support IP IOCTL commands
146  */
147 #define	IP_IOCTL			(('i'<<8)|'p')
148 #define	IP_IOC_IRE_DELETE		4
149 #define	IP_IOC_IRE_DELETE_NO_REPLY	5
150 #define	IP_IOC_IRE_ADVISE_NO_REPLY	6
151 #define	IP_IOC_RTS_REQUEST		7
152 
153 /* Common definitions used by IP IOCTL data structures */
154 typedef struct ipllcmd_s {
155 	uint_t	ipllc_cmd;
156 	uint_t	ipllc_name_offset;
157 	uint_t	ipllc_name_length;
158 } ipllc_t;
159 
160 /* IP IRE Change Command Structure. */
161 typedef struct ipic_s {
162 	ipllc_t	ipic_ipllc;
163 	uint_t	ipic_ire_type;
164 	uint_t	ipic_max_frag;
165 	uint_t	ipic_addr_offset;
166 	uint_t	ipic_addr_length;
167 	uint_t	ipic_mask_offset;
168 	uint_t	ipic_mask_length;
169 	uint_t	ipic_src_addr_offset;
170 	uint_t	ipic_src_addr_length;
171 	uint_t	ipic_ll_hdr_offset;
172 	uint_t	ipic_ll_hdr_length;
173 	uint_t	ipic_gateway_addr_offset;
174 	uint_t	ipic_gateway_addr_length;
175 	clock_t	ipic_rtt;
176 	uint32_t ipic_ssthresh;
177 	clock_t	ipic_rtt_sd;
178 	uchar_t ipic_ire_marks;
179 } ipic_t;
180 
181 #define	ipic_cmd		ipic_ipllc.ipllc_cmd
182 #define	ipic_ll_name_length	ipic_ipllc.ipllc_name_length
183 #define	ipic_ll_name_offset	ipic_ipllc.ipllc_name_offset
184 
185 /* IP IRE Delete Command Structure. */
186 typedef struct ipid_s {
187 	ipllc_t	ipid_ipllc;
188 	uint_t	ipid_ire_type;
189 	uint_t	ipid_addr_offset;
190 	uint_t	ipid_addr_length;
191 	uint_t	ipid_mask_offset;
192 	uint_t	ipid_mask_length;
193 } ipid_t;
194 
195 #define	ipid_cmd		ipid_ipllc.ipllc_cmd
196 
197 #ifdef _KERNEL
198 /*
199  * Temporary state for ip options parser.
200  */
201 typedef struct ipoptp_s
202 {
203 	uint8_t		*ipoptp_next;	/* next option to look at */
204 	uint8_t		*ipoptp_end;	/* end of options */
205 	uint8_t		*ipoptp_cur;	/* start of current option */
206 	uint8_t		ipoptp_len;	/* length of current option */
207 	uint32_t	ipoptp_flags;
208 } ipoptp_t;
209 
210 /*
211  * Flag(s) for ipoptp_flags
212  */
213 #define	IPOPTP_ERROR	0x00000001
214 #endif	/* _KERNEL */
215 
216 
217 /* Controls forwarding of IP packets, set via ndd */
218 #define	IP_FORWARD_NEVER	0
219 #define	IP_FORWARD_ALWAYS	1
220 
221 #define	WE_ARE_FORWARDING(ipst)	((ipst)->ips_ip_g_forward == IP_FORWARD_ALWAYS)
222 
223 #define	IPH_HDR_LENGTH(ipha)						\
224 	((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length & 0xF) << 2)
225 
226 #define	IPH_HDR_VERSION(ipha)						\
227 	((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length) >> 4)
228 
229 #ifdef _KERNEL
230 /*
231  * IP reassembly macros.  We hide starting and ending offsets in b_next and
232  * b_prev of messages on the reassembly queue.	The messages are chained using
233  * b_cont.  These macros are used in ip_reassemble() so we don't have to see
234  * the ugly casts and assignments.
235  * Note that the offsets are <= 64k i.e. a uint_t is sufficient to represent
236  * them.
237  */
238 #define	IP_REASS_START(mp)		((uint_t)(uintptr_t)((mp)->b_next))
239 #define	IP_REASS_SET_START(mp, u)	\
240 	((mp)->b_next = (mblk_t *)(uintptr_t)(u))
241 #define	IP_REASS_END(mp)		((uint_t)(uintptr_t)((mp)->b_prev))
242 #define	IP_REASS_SET_END(mp, u)		\
243 	((mp)->b_prev = (mblk_t *)(uintptr_t)(u))
244 
245 #define	IP_REASS_COMPLETE	0x1
246 #define	IP_REASS_PARTIAL	0x2
247 #define	IP_REASS_FAILED		0x4
248 
249 /*
250  * Test to determine whether this is a module instance of IP or a
251  * driver instance of IP.
252  */
253 #define	CONN_Q(q)	(WR(q)->q_next == NULL)
254 
255 #define	Q_TO_CONN(q)	((conn_t *)(q)->q_ptr)
256 #define	Q_TO_TCP(q)	(Q_TO_CONN((q))->conn_tcp)
257 #define	Q_TO_UDP(q)	(Q_TO_CONN((q))->conn_udp)
258 #define	Q_TO_ICMP(q)	(Q_TO_CONN((q))->conn_icmp)
259 #define	Q_TO_RTS(q)	(Q_TO_CONN((q))->conn_rts)
260 
261 /*
262  * The following two macros are used by IP to get the appropriate
263  * wq and rq for a conn. If it is a TCP conn, then we need
264  * tcp_wq/tcp_rq else, conn_wq/conn_rq. IP can use conn_wq and conn_rq
265  * from a conn directly if it knows that the conn is not TCP.
266  */
267 #define	CONNP_TO_WQ(connp)	\
268 	(IPCL_IS_TCP(connp) ? (connp)->conn_tcp->tcp_wq : (connp)->conn_wq)
269 
270 #define	CONNP_TO_RQ(connp)	RD(CONNP_TO_WQ(connp))
271 
272 #define	GRAB_CONN_LOCK(q)	{				\
273 	if (q != NULL && CONN_Q(q))				\
274 		mutex_enter(&(Q_TO_CONN(q))->conn_lock);	\
275 }
276 
277 #define	RELEASE_CONN_LOCK(q)	{				\
278 	if (q != NULL && CONN_Q(q))				\
279 		mutex_exit(&(Q_TO_CONN(q))->conn_lock);		\
280 }
281 
282 /* "Congestion controlled" protocol */
283 #define	IP_FLOW_CONTROLLED_ULP(p)   ((p) == IPPROTO_TCP || (p) == IPPROTO_SCTP)
284 
285 /*
286  * Complete the pending operation. Usually an ioctl. Can also
287  * be a bind or option management request that got enqueued
288  * in an ipsq_t. Called on completion of the operation.
289  */
290 #define	CONN_OPER_PENDING_DONE(connp)	{			\
291 	mutex_enter(&(connp)->conn_lock);			\
292 	(connp)->conn_oper_pending_ill = NULL;			\
293 	cv_broadcast(&(connp)->conn_refcv);			\
294 	mutex_exit(&(connp)->conn_lock);			\
295 	CONN_DEC_REF(connp);					\
296 }
297 
298 /* Get the credential of an IP queue of unknown type */
299 #define	GET_QUEUE_CRED(wq)						\
300 	((wq)->q_next ? (((ill_t *)(wq)->q_ptr)->ill_credp) \
301 	    : ((Q_TO_CONN((wq)))->conn_cred))
302 
303 /*
304  * Flags for the various ip_fanout_* routines.
305  */
306 #define	IP_FF_SEND_ICMP		0x01	/* Send an ICMP error */
307 #define	IP_FF_HDR_COMPLETE	0x02	/* Call ip_hdr_complete if error */
308 #define	IP_FF_CKSUM		0x04	/* Recompute ipha_cksum if error */
309 #define	IP_FF_RAWIP		0x08	/* Use rawip mib variable */
310 #define	IP_FF_SRC_QUENCH	0x10	/* OK to send ICMP_SOURCE_QUENCH */
311 #define	IP_FF_SYN_ADDIRE	0x20	/* Add IRE if TCP syn packet */
312 #define	IP_FF_IPINFO		0x80	/* Used for both V4 and V6 */
313 #define	IP_FF_SEND_SLLA		0x100	/* Send source link layer info ? */
314 #define	IPV6_REACHABILITY_CONFIRMATION	0x200	/* Flags for ip_xmit_v6 */
315 #define	IP_FF_NO_MCAST_LOOP	0x400	/* No multicasts for sending zone */
316 
317 /*
318  * Following flags are used by IPQoS to determine if policy processing is
319  * required.
320  */
321 #define	IP6_NO_IPPOLICY		0x800	/* Don't do IPQoS processing */
322 #define	IP6_IN_LLMCAST		0x1000	/* Multicast */
323 
324 #define	IP_FF_LOOPBACK		0x2000	/* Loopback fanout */
325 #define	IP_FF_SCTP_CSUM_ERR	0x4000	/* sctp pkt has failed chksum */
326 
327 #ifndef	IRE_DB_TYPE
328 #define	IRE_DB_TYPE	M_SIG
329 #endif
330 
331 #ifndef	IRE_DB_REQ_TYPE
332 #define	IRE_DB_REQ_TYPE	M_PCSIG
333 #endif
334 
335 #ifndef	IRE_ARPRESOLVE_TYPE
336 #define	IRE_ARPRESOLVE_TYPE	M_EVENT
337 #endif
338 
339 /*
340  * Values for squeue switch:
341  */
342 
343 #define	IP_SQUEUE_ENTER_NODRAIN	1
344 #define	IP_SQUEUE_ENTER	2
345 /*
346  * This is part of the interface between Transport provider and
347  * IP which can be used to set policy information. This is usually
348  * accompanied with O_T_BIND_REQ/T_BIND_REQ.ip_bind assumes that
349  * only IPSEC_POLICY_SET is there when it is found in the chain.
350  * The information contained is an struct ipsec_req_t. On success
351  * or failure, either the T_BIND_ACK or the T_ERROR_ACK is returned.
352  * IPSEC_POLICY_SET is never returned.
353  */
354 #define	IPSEC_POLICY_SET	M_SETOPTS
355 
356 #define	IRE_IS_LOCAL(ire)	((ire != NULL) && \
357 				((ire)->ire_type & (IRE_LOCAL | IRE_LOOPBACK)))
358 
359 #define	IRE_IS_TARGET(ire)	((ire != NULL) && \
360 				((ire)->ire_type != IRE_BROADCAST))
361 
362 /* IP Fragmentation Reassembly Header */
363 typedef struct ipf_s {
364 	struct ipf_s	*ipf_hash_next;
365 	struct ipf_s	**ipf_ptphn;	/* Pointer to previous hash next. */
366 	uint32_t	ipf_ident;	/* Ident to match. */
367 	uint8_t		ipf_protocol;	/* Protocol to match. */
368 	uchar_t		ipf_last_frag_seen : 1;	/* Last fragment seen ? */
369 	time_t		ipf_timestamp;	/* Reassembly start time. */
370 	mblk_t		*ipf_mp;	/* mblk we live in. */
371 	mblk_t		*ipf_tail_mp;	/* Frag queue tail pointer. */
372 	int		ipf_hole_cnt;	/* Number of holes (hard-case). */
373 	int		ipf_end;	/* Tail end offset (0 -> hard-case). */
374 	uint_t		ipf_gen;	/* Frag queue generation */
375 	size_t		ipf_count;	/* Count of bytes used by frag */
376 	uint_t		ipf_nf_hdr_len; /* Length of nonfragmented header */
377 	in6_addr_t	ipf_v6src;	/* IPv6 source address */
378 	in6_addr_t	ipf_v6dst;	/* IPv6 dest address */
379 	uint_t		ipf_prev_nexthdr_offset; /* Offset for nexthdr value */
380 	uint8_t		ipf_ecn;	/* ECN info for the fragments */
381 	uint8_t		ipf_num_dups;	/* Number of times dup frags recvd */
382 	uint16_t	ipf_checksum_flags; /* Hardware checksum flags */
383 	uint32_t	ipf_checksum;	/* Partial checksum of fragment data */
384 } ipf_t;
385 
386 #define	ipf_src	V4_PART_OF_V6(ipf_v6src)
387 #define	ipf_dst	V4_PART_OF_V6(ipf_v6dst)
388 
389 typedef enum {
390 	IB_PKT =  0x01,
391 	OB_PKT = 0x02
392 } ip_pkt_t;
393 
394 #define	UPDATE_IB_PKT_COUNT(ire)\
395 	{ \
396 	(ire)->ire_ib_pkt_count++; \
397 	if ((ire)->ire_ipif != NULL) { \
398 		/* \
399 		 * forwarding packet \
400 		 */ \
401 		if ((ire)->ire_type & (IRE_LOCAL|IRE_BROADCAST)) \
402 			atomic_add_32(&(ire)->ire_ipif->ipif_ib_pkt_count, 1);\
403 		else \
404 			atomic_add_32(&(ire)->ire_ipif->ipif_fo_pkt_count, 1);\
405 	} \
406 	}
407 #define	UPDATE_OB_PKT_COUNT(ire)\
408 	{ \
409 	(ire)->ire_ob_pkt_count++;\
410 	if ((ire)->ire_ipif != NULL) { \
411 		atomic_add_32(&(ire)->ire_ipif->ipif_ob_pkt_count, 1); \
412 	} \
413 	}
414 
415 
416 #define	IP_RPUT_LOCAL(q, mp, ipha, ire, recv_ill) \
417 { \
418 	switch (ipha->ipha_protocol) { \
419 		case IPPROTO_UDP: \
420 			ip_udp_input(q, mp, ipha, ire, recv_ill); \
421 			break; \
422 		default: \
423 			ip_proto_input(q, mp, ipha, ire, recv_ill, B_FALSE); \
424 			break; \
425 	} \
426 }
427 
428 /*
429  * NCE_EXPIRED is TRUE when we have a non-permanent nce that was
430  * found to be REACHABLE more than ip_ire_arp_interval ms ago.
431  * This macro is used to age existing nce_t entries. The
432  * nce's will get cleaned up in the following circumstances:
433  * - ip_ire_trash_reclaim will free nce's using ndp_cache_reclaim
434  *    when memory is low,
435  * - ip_arp_news, when updates are received.
436  * - if the nce is NCE_EXPIRED(), it will deleted, so that a new
437  *   arp request will need to be triggered from an ND_INITIAL nce.
438  *
439  * Note that the nce state transition follows the pattern:
440  *	ND_INITIAL -> ND_INCOMPLETE -> ND_REACHABLE
441  * after which the nce is deleted when it has expired.
442  *
443  * nce_last is the timestamp that indicates when the nce_res_mp in the
444  * nce_t was last updated to a valid link-layer address.  nce_last gets
445  * modified/updated :
446  *  - when the nce is created
447  *  - every time we get a sane arp response for the nce.
448  */
449 #define	NCE_EXPIRED(nce, ipst)	(nce->nce_last > 0 &&	\
450 	    ((nce->nce_flags & NCE_F_PERMANENT) == 0) &&	\
451 	    ((TICK_TO_MSEC(lbolt64) - nce->nce_last) > 		\
452 		(ipst)->ips_ip_ire_arp_interval))
453 
454 #endif /* _KERNEL */
455 
456 /* ICMP types */
457 #define	ICMP_ECHO_REPLY			0
458 #define	ICMP_DEST_UNREACHABLE		3
459 #define	ICMP_SOURCE_QUENCH		4
460 #define	ICMP_REDIRECT			5
461 #define	ICMP_ECHO_REQUEST		8
462 #define	ICMP_ROUTER_ADVERTISEMENT	9
463 #define	ICMP_ROUTER_SOLICITATION	10
464 #define	ICMP_TIME_EXCEEDED		11
465 #define	ICMP_PARAM_PROBLEM		12
466 #define	ICMP_TIME_STAMP_REQUEST		13
467 #define	ICMP_TIME_STAMP_REPLY		14
468 #define	ICMP_INFO_REQUEST		15
469 #define	ICMP_INFO_REPLY			16
470 #define	ICMP_ADDRESS_MASK_REQUEST	17
471 #define	ICMP_ADDRESS_MASK_REPLY		18
472 
473 /* ICMP_TIME_EXCEEDED codes */
474 #define	ICMP_TTL_EXCEEDED		0
475 #define	ICMP_REASSEMBLY_TIME_EXCEEDED	1
476 
477 /* ICMP_DEST_UNREACHABLE codes */
478 #define	ICMP_NET_UNREACHABLE		0
479 #define	ICMP_HOST_UNREACHABLE		1
480 #define	ICMP_PROTOCOL_UNREACHABLE	2
481 #define	ICMP_PORT_UNREACHABLE		3
482 #define	ICMP_FRAGMENTATION_NEEDED	4
483 #define	ICMP_SOURCE_ROUTE_FAILED	5
484 #define	ICMP_DEST_NET_UNKNOWN		6
485 #define	ICMP_DEST_HOST_UNKNOWN		7
486 #define	ICMP_SRC_HOST_ISOLATED		8
487 #define	ICMP_DEST_NET_UNREACH_ADMIN	9
488 #define	ICMP_DEST_HOST_UNREACH_ADMIN	10
489 #define	ICMP_DEST_NET_UNREACH_TOS	11
490 #define	ICMP_DEST_HOST_UNREACH_TOS	12
491 
492 /* ICMP Header Structure */
493 typedef struct icmph_s {
494 	uint8_t		icmph_type;
495 	uint8_t		icmph_code;
496 	uint16_t	icmph_checksum;
497 	union {
498 		struct { /* ECHO request/response structure */
499 			uint16_t	u_echo_ident;
500 			uint16_t	u_echo_seqnum;
501 		} u_echo;
502 		struct { /* Destination unreachable structure */
503 			uint16_t	u_du_zero;
504 			uint16_t	u_du_mtu;
505 		} u_du;
506 		struct { /* Parameter problem structure */
507 			uint8_t		u_pp_ptr;
508 			uint8_t		u_pp_rsvd[3];
509 		} u_pp;
510 		struct { /* Redirect structure */
511 			ipaddr_t	u_rd_gateway;
512 		} u_rd;
513 	} icmph_u;
514 } icmph_t;
515 
516 #define	icmph_echo_ident	icmph_u.u_echo.u_echo_ident
517 #define	icmph_echo_seqnum	icmph_u.u_echo.u_echo_seqnum
518 #define	icmph_du_zero		icmph_u.u_du.u_du_zero
519 #define	icmph_du_mtu		icmph_u.u_du.u_du_mtu
520 #define	icmph_pp_ptr		icmph_u.u_pp.u_pp_ptr
521 #define	icmph_rd_gateway	icmph_u.u_rd.u_rd_gateway
522 
523 #define	ICMPH_SIZE	8
524 
525 /*
526  * Minimum length of transport layer header included in an ICMP error
527  * message for it to be considered valid.
528  */
529 #define	ICMP_MIN_TP_HDR_LEN	8
530 
531 /* Aligned IP header */
532 typedef struct ipha_s {
533 	uint8_t		ipha_version_and_hdr_length;
534 	uint8_t		ipha_type_of_service;
535 	uint16_t	ipha_length;
536 	uint16_t	ipha_ident;
537 	uint16_t	ipha_fragment_offset_and_flags;
538 	uint8_t		ipha_ttl;
539 	uint8_t		ipha_protocol;
540 	uint16_t	ipha_hdr_checksum;
541 	ipaddr_t	ipha_src;
542 	ipaddr_t	ipha_dst;
543 } ipha_t;
544 
545 #define	IPH_DF		0x4000	/* Don't fragment */
546 #define	IPH_MF		0x2000	/* More fragments to come */
547 #define	IPH_OFFSET	0x1FFF	/* Where the offset lives */
548 #define	IPH_FRAG_HDR	0x8000	/* IPv6 don't fragment bit */
549 
550 /* ECN code points for IPv4 TOS byte and IPv6 traffic class octet. */
551 #define	IPH_ECN_NECT	0x0	/* Not ECN-Capabable Transport */
552 #define	IPH_ECN_ECT1	0x1	/* ECN-Capable Transport, ECT(1) */
553 #define	IPH_ECN_ECT0	0x2	/* ECN-Capable Transport, ECT(0) */
554 #define	IPH_ECN_CE	0x3	/* ECN-Congestion Experienced (CE) */
555 
556 /* IP Mac info structure */
557 typedef struct ip_m_s {
558 	t_uscalar_t	ip_m_mac_type;	/* From <sys/dlpi.h> */
559 	int		ip_m_type;	/* From <net/if_types.h> */
560 	boolean_t	(*ip_m_v4mapinfo)(uint_t, uint8_t *, uint8_t *,
561 			    uint32_t *, ipaddr_t *);
562 	boolean_t	(*ip_m_v6mapinfo)(uint_t, uint8_t *, uint8_t *,
563 			    uint32_t *, in6_addr_t *);
564 	boolean_t	(*ip_m_v6intfid)(uint_t, uint8_t *, in6_addr_t *);
565 } ip_m_t;
566 
567 /*
568  * The following functions attempt to reduce the link layer dependency
569  * of the IP stack. The current set of link specific operations are:
570  * a. map from IPv4 class D (224.0/4) multicast address range to the link
571  * layer multicast address range.
572  * b. map from IPv6 multicast address range (ff00::/8) to the link
573  * layer multicast address range.
574  * c. derive the default IPv6 interface identifier from the link layer
575  * address.
576  */
577 #define	MEDIA_V4MINFO(ip_m, plen, bphys, maddr, hwxp, v4ptr) \
578 	(((ip_m)->ip_m_v4mapinfo != NULL) && \
579 	(*(ip_m)->ip_m_v4mapinfo)(plen, bphys, maddr, hwxp, v4ptr))
580 #define	MEDIA_V6INTFID(ip_m, plen, phys, v6ptr) \
581 	(((ip_m)->ip_m_v6intfid != NULL) && \
582 	(*(ip_m)->ip_m_v6intfid)(plen, phys, v6ptr))
583 #define	MEDIA_V6MINFO(ip_m, plen, bphys, maddr, hwxp, v6ptr) \
584 	(((ip_m)->ip_m_v6mapinfo != NULL) && \
585 	(*(ip_m)->ip_m_v6mapinfo)(plen, bphys, maddr, hwxp, v6ptr))
586 
587 /* Router entry types */
588 #define	IRE_BROADCAST		0x0001	/* Route entry for broadcast address */
589 #define	IRE_DEFAULT		0x0002	/* Route entry for default gateway */
590 #define	IRE_LOCAL		0x0004	/* Route entry for local address */
591 #define	IRE_LOOPBACK		0x0008	/* Route entry for loopback address */
592 #define	IRE_PREFIX		0x0010	/* Route entry for prefix routes */
593 #define	IRE_CACHE		0x0020	/* Cached Route entry */
594 #define	IRE_IF_NORESOLVER	0x0040	/* Route entry for local interface */
595 					/* net without any address mapping. */
596 #define	IRE_IF_RESOLVER		0x0080	/* Route entry for local interface */
597 					/* net with resolver. */
598 #define	IRE_HOST		0x0100	/* Host route entry */
599 #define	IRE_HOST_REDIRECT	0x0200	/* only used for T_SVR4_OPTMGMT_REQ */
600 
601 #define	IRE_INTERFACE		(IRE_IF_NORESOLVER | IRE_IF_RESOLVER)
602 #define	IRE_OFFSUBNET		(IRE_DEFAULT | IRE_PREFIX | IRE_HOST)
603 #define	IRE_CACHETABLE		(IRE_CACHE | IRE_BROADCAST | IRE_LOCAL | \
604 				IRE_LOOPBACK)
605 #define	IRE_FORWARDTABLE	(IRE_INTERFACE | IRE_OFFSUBNET)
606 
607 /*
608  * If an IRE is marked with IRE_MARK_CONDEMNED, the last walker of
609  * the bucket should delete this IRE from this bucket.
610  */
611 #define	IRE_MARK_CONDEMNED	0x0001
612 /*
613  * If a broadcast IRE is marked with IRE_MARK_NORECV, ip_rput will drop the
614  * broadcast packets received on that interface. This is marked only
615  * on broadcast ires. Employed by IPMP, where we have multiple NICs on the
616  * same subnet receiving the same broadcast packet.
617  */
618 #define	IRE_MARK_NORECV		0x0002
619 /*
620  * IRE_CACHE marked this way won't be returned by ire_cache_lookup. Need
621  * to look specifically using MATCH_IRE_MARK_HIDDEN. Used by IPMP.
622  */
623 #define	IRE_MARK_HIDDEN		0x0004	/* Typically Used by in.mpathd */
624 
625 /*
626  * An IRE with IRE_MARK_NOADD is created in ip_newroute_ipif when the outgoing
627  * interface is specified by e.g. IP_PKTINFO.  The IRE is not added to the IRE
628  * cache table.
629  */
630 #define	IRE_MARK_NOADD		0x0008	/* Mark not to add ire in cache */
631 
632 /*
633  * IRE marked with IRE_MARK_TEMPORARY means that this IRE has been used
634  * either for forwarding a packet or has not been used for sending
635  * traffic on TCP connections terminated on this system.  In both
636  * cases, this IRE is the first to go when IRE is being cleaned up.
637  */
638 #define	IRE_MARK_TEMPORARY	0x0010
639 
640 /*
641  * IRE marked with IRE_MARK_USESRC_CHECK means that while adding an IRE with
642  * this mark, additional atomic checks need to be performed. For eg: by the
643  * time an IRE_CACHE is created, sent up to ARP and then comes back to IP; the
644  * usesrc grouping could have changed in which case we want to fail adding
645  * the IRE_CACHE entry
646  */
647 #define	IRE_MARK_USESRC_CHECK	0x0020
648 
649 /*
650  * IRE_MARK_PRIVATE_ADDR is used for IP_NEXTHOP. When IP_NEXTHOP is set, the
651  * routing table lookup for the destination is bypassed and the packet is
652  * sent directly to the specified nexthop. The associated IRE_CACHE entries
653  * should be marked with IRE_MARK_PRIVATE_ADDR flag so that they don't show up
654  * in regular ire cache lookups.
655  */
656 #define	IRE_MARK_PRIVATE_ADDR	0x0040
657 
658 /*
659  * When we send an ARP resolution query for the nexthop gateway's ire,
660  * we use esballoc to create the ire_t in the AR_ENTRY_QUERY mblk
661  * chain, and mark its ire_marks with IRE_MARK_UNCACHED. This flag
662  * indicates that information from ARP has not been transferred to a
663  * permanent IRE_CACHE entry. The flag is reset only when the
664  * information is successfully transferred to an ire_cache entry (in
665  * ire_add()). Attempting to free the AR_ENTRY_QUERY mblk chain prior
666  * to ire_add (e.g., from arp, or from ip`ip_wput_nondata) will
667  * require that the resources (incomplete ire_cache and/or nce) must
668  * be cleaned up. The free callback routine (ire_freemblk()) checks
669  * for IRE_MARK_UNCACHED to see if any resources that are pinned down
670  * will need to be cleaned up or not.
671  */
672 
673 #define	IRE_MARK_UNCACHED	0x0080
674 
675 /*
676  * The comment below (and for other netstack_t references) refers
677  * to the fact that we only do netstack_hold in particular cases,
678  * such as the references from open streams (ill_t and conn_t's
679  * pointers). Internally within IP we rely on IP's ability to cleanup e.g.
680  * ire_t's when an ill goes away.
681  */
682 typedef struct ire_expire_arg_s {
683 	int		iea_flush_flag;
684 	ip_stack_t	*iea_ipst;	/* Does not have a netstack_hold */
685 } ire_expire_arg_t;
686 
687 /* Flags with ire_expire routine */
688 #define	FLUSH_ARP_TIME		0x0001	/* ARP info potentially stale timer */
689 #define	FLUSH_REDIRECT_TIME	0x0002	/* Redirects potentially stale */
690 #define	FLUSH_MTU_TIME		0x0004	/* Include path MTU per RFC 1191 */
691 
692 /* Arguments to ire_flush_cache() */
693 #define	IRE_FLUSH_DELETE	0
694 #define	IRE_FLUSH_ADD		1
695 
696 /*
697  * Open/close synchronization flags.
698  * These are kept in a separate field in the conn and the synchronization
699  * depends on the atomic 32 bit access to that field.
700  */
701 #define	CONN_CLOSING		0x01	/* ip_close waiting for ip_wsrv */
702 #define	CONN_IPSEC_LOAD_WAIT	0x02	/* waiting for load */
703 #define	CONN_CONDEMNED		0x04	/* conn is closing, no more refs */
704 #define	CONN_INCIPIENT		0x08	/* conn not yet visible, no refs */
705 #define	CONN_QUIESCED		0x10	/* conn is now quiescent */
706 
707 /* Used to check connection state flags before caching the IRE */
708 #define	CONN_CACHE_IRE(connp)	\
709 	(!((connp)->conn_state_flags & (CONN_CLOSING|CONN_CONDEMNED)))
710 
711 /*
712  * Parameter to ip_output giving the identity of the caller.
713  * IP_WSRV means the packet was enqueued in the STREAMS queue
714  * due to flow control and is now being reprocessed in the context of
715  * the STREAMS service procedure, consequent to flow control relief.
716  * IRE_SEND means the packet is being reprocessed consequent to an
717  * ire cache creation and addition and this may or may not be happening
718  * in the service procedure context. Anything other than the above 2
719  * cases is identified as IP_WPUT. Most commonly this is the case of
720  * packets coming down from the application.
721  */
722 #ifdef _KERNEL
723 #define	IP_WSRV			1	/* Called from ip_wsrv */
724 #define	IP_WPUT			2	/* Called from ip_wput */
725 #define	IRE_SEND		3	/* Called from ire_send */
726 
727 /*
728  * Extra structures need for per-src-addr filtering (IGMPv3/MLDv2)
729  */
730 #define	MAX_FILTER_SIZE	64
731 
732 typedef struct slist_s {
733 	int		sl_numsrc;
734 	in6_addr_t	sl_addr[MAX_FILTER_SIZE];
735 } slist_t;
736 
737 /*
738  * Following struct is used to maintain retransmission state for
739  * a multicast group.  One rtx_state_t struct is an in-line field
740  * of the ilm_t struct; the slist_ts in the rtx_state_t struct are
741  * alloc'd as needed.
742  */
743 typedef struct rtx_state_s {
744 	uint_t		rtx_timer;	/* retrans timer */
745 	int		rtx_cnt;	/* retrans count */
746 	int		rtx_fmode_cnt;	/* retrans count for fmode change */
747 	slist_t		*rtx_allow;
748 	slist_t		*rtx_block;
749 } rtx_state_t;
750 
751 /*
752  * Used to construct list of multicast address records that will be
753  * sent in a single listener report.
754  */
755 typedef struct mrec_s {
756 	struct mrec_s	*mrec_next;
757 	uint8_t		mrec_type;
758 	uint8_t		mrec_auxlen;	/* currently unused */
759 	in6_addr_t	mrec_group;
760 	slist_t		mrec_srcs;
761 } mrec_t;
762 
763 /* Group membership list per upper conn */
764 /*
765  * XXX add ilg info for ifaddr/ifindex.
766  * XXX can we make ilg survive an ifconfig unplumb + plumb
767  * by setting the ipif/ill to NULL and recover that later?
768  *
769  * ilg_ipif is used by IPv4 as multicast groups are joined using an interface
770  * address (ipif).
771  * ilg_ill is used by IPv6 as multicast groups are joined using an interface
772  * index (phyint->phyint_ifindex).
773  * ilg_ill is NULL for IPv4 and ilg_ipif is NULL for IPv6.
774  *
775  * ilg records the state of multicast memberships of a socket end point.
776  * ilm records the state of multicast memberships with the driver and is
777  * maintained per interface.
778  *
779  * Notes :
780  *
781  * 1) There is no direct link between a given ilg and ilm. If the
782  *    application has joined a group G with ifindex I, we will have
783  *    an ilg with ilg_v6group and ilg_ill. There will be a corresponding
784  *    ilm with ilm_ill/ilm_v6addr recording the multicast membership.
785  *    To delete the membership,
786  *
787  *		a) Search for ilg matching on G and I with ilg_v6group
788  *		   and ilg_ill. Delete ilg_ill.
789  *		b) Search the corresponding ilm matching on G and I with
790  *		   ilm_v6addr and ilm_ill. Delete ilm.
791  *
792  *    In IPv4, the only difference is, we look using ipifs instead of
793  *    ills.
794  *
795  * 2) With IP multipathing, we want to keep receiving even after the
796  *    interface has failed. We do this by moving multicast memberships
797  *    to a new_ill within the group. This is acheived by sending
798  *    DL_DISABMULTI_REQS on ilg_ill/ilm_ill and sending DL_ENABMULTIREQS
799  *    on the new_ill and changing ilg_ill/ilm_ill to new_ill. But, we
800  *    need to be able to delete memberships which will still come down
801  *    with the ifindex of the old ill which is what the application
802  *    knows of. Thus we store the ilm_/ilg_orig_ifindex to keep track
803  *    of where we joined initially so that we can lookup even after we
804  *    moved the membership. It is also used for moving back the membership
805  *    when the old ill has been repaired. This is done by looking up for
806  *    ilms with ilm_orig_ifindex matching on the old ill's ifindex. Only
807  *    ilms actually move from old ill to new ill. ilgs don't move (just
808  *    the ilg_ill is changed when it moves) as it just records the state
809  *    of the application that has joined a group G where as ilm records
810  *    the state joined with the driver. Thus when we send DL_XXXMULTI_REQs
811  *    we also need to keep the ilm in the right ill.
812  *
813  *    In IPv4, as ipifs move from old ill to new_ill, ilgs and ilms move
814  *    implicitly as we use only ipifs in IPv4. Thus, one can always lookup
815  *    a given ilm/ilg even after it fails without the support of
816  *    orig_ifindex. We move ilms still to record the driver state as
817  *    mentioned above.
818  */
819 
820 /*
821  * The ilg_t and ilm_t members are protected by ipsq. They can be changed only
822  * by a thread executing in the ipsq. In other words add/delete of a
823  * multicast group has to execute in the ipsq.
824  */
825 #define	ILG_DELETED	0x1		/* ilg_flags */
826 typedef struct ilg_s {
827 	in6_addr_t	ilg_v6group;
828 	struct ipif_s	*ilg_ipif;	/* Logical interface we are member on */
829 	struct ill_s	*ilg_ill;	/* Used by IPv6 */
830 	int		ilg_orig_ifindex; /* Interface originally joined on */
831 	uint_t		ilg_flags;
832 	mcast_record_t	ilg_fmode;	/* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */
833 	slist_t		*ilg_filter;
834 } ilg_t;
835 
836 
837 /*
838  * Multicast address list entry for lower ill.
839  * ilm_ipif is used by IPv4 as multicast groups are joined using ipif.
840  * ilm_ill is used by IPv6 as multicast groups are joined using ill.
841  * ilm_ill is NULL for IPv4 and ilm_ipif is NULL for IPv6.
842  *
843  * The comment below (and for other netstack_t references) refers
844  * to the fact that we only do netstack_hold in particular cases,
845  * such as the references from open streams (ill_t and conn_t's
846  * pointers). Internally within IP we rely on IP's ability to cleanup e.g.
847  * ire_t's when an ill goes away.
848  */
849 #define	ILM_DELETED	0x1		/* ilm_flags */
850 typedef struct ilm_s {
851 	in6_addr_t	ilm_v6addr;
852 	int		ilm_refcnt;
853 	uint_t		ilm_timer;	/* IGMP/MLD query resp timer, in msec */
854 	struct ipif_s	*ilm_ipif;	/* Back pointer to ipif for IPv4 */
855 	struct ilm_s	*ilm_next;	/* Linked list for each ill */
856 	uint_t		ilm_state;	/* state of the membership */
857 	struct ill_s	*ilm_ill;	/* Back pointer to ill for IPv6 */
858 	int		ilm_orig_ifindex;  /* V6_MULTICAST_IF/ilm_ipif index */
859 	uint_t		ilm_flags;
860 	boolean_t	ilm_is_new;	/* new ilm */
861 	boolean_t	ilm_notify_driver; /* Need to notify the driver */
862 	zoneid_t	ilm_zoneid;
863 	int		ilm_no_ilg_cnt;	/* number of joins w/ no ilg */
864 	mcast_record_t	ilm_fmode;	/* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */
865 	slist_t		*ilm_filter;	/* source filter list */
866 	slist_t		*ilm_pendsrcs;	/* relevant src addrs for pending req */
867 	rtx_state_t	ilm_rtx;	/* SCR retransmission state */
868 	ip_stack_t	*ilm_ipst;	/* Does not have a netstack_hold */
869 } ilm_t;
870 
871 #define	ilm_addr	V4_PART_OF_V6(ilm_v6addr)
872 
873 /*
874  * ilm_walker_cleanup needs to execute when the ilm_walker_cnt goes down to
875  * zero. In addition it needs to block new walkers while it is unlinking ilm's
876  * from the list. Thus simple atomics for the ill_ilm_walker_cnt don't suffice.
877  */
878 #define	ILM_WALKER_HOLD(ill)    {               \
879 	mutex_enter(&(ill)->ill_lock);          \
880 	ill->ill_ilm_walker_cnt++;              \
881 	mutex_exit(&(ill)->ill_lock);           \
882 }
883 
884 /*
885  * ilm_walker_cleanup releases ill_lock
886  */
887 #define	ILM_WALKER_RELE(ill)	{ 		\
888 	mutex_enter(&(ill)->ill_lock);		\
889 	(ill)->ill_ilm_walker_cnt--;		\
890 	if ((ill)->ill_ilm_walker_cnt == 0 && (ill)->ill_ilm_cleanup_reqd) \
891 		ilm_walker_cleanup(ill);	\
892 	else 					\
893 		mutex_exit(&(ill)->ill_lock);	\
894 }
895 
896 /*
897  * Soft reference to an IPsec SA.
898  *
899  * On relative terms, conn's can be persistant (living as long as the
900  * processes which create them), while SA's are ephemeral (dying when
901  * they hit their time-based or byte-based lifetimes).
902  *
903  * We could hold a hard reference to an SA from an ipsec_latch_t,
904  * but this would cause expired SA's to linger for a potentially
905  * unbounded time.
906  *
907  * Instead, we remember the hash bucket number and bucket generation
908  * in addition to the pointer.  The bucket generation is incremented on
909  * each deletion.
910  */
911 typedef struct ipsa_ref_s
912 {
913 	struct ipsa_s	*ipsr_sa;
914 	struct isaf_s	*ipsr_bucket;
915 	uint64_t	ipsr_gen;
916 } ipsa_ref_t;
917 
918 /*
919  * IPsec "latching" state.
920  *
921  * In the presence of IPsec policy, fully-bound conn's bind a connection
922  * to more than just the 5-tuple, but also a specific IPsec action and
923  * identity-pair.
924  *
925  * As an optimization, we also cache soft references to IPsec SA's
926  * here so that we can fast-path around most of the work needed for
927  * outbound IPsec SA selection.
928  *
929  * Were it not for TCP's detached connections, this state would be
930  * in-line in conn_t; instead, this is in a separate structure so it
931  * can be handed off to TCP when a connection is detached.
932  */
933 typedef struct ipsec_latch_s
934 {
935 	kmutex_t	ipl_lock;
936 	uint32_t	ipl_refcnt;
937 
938 	uint64_t	ipl_unique;
939 	struct ipsec_policy_s	*ipl_in_policy; /* latched policy (in) */
940 	struct ipsec_policy_s	*ipl_out_policy; /* latched policy (out) */
941 	struct ipsec_action_s	*ipl_in_action;	/* latched action (in) */
942 	struct ipsec_action_s	*ipl_out_action; /* latched action (out) */
943 	cred_t		*ipl_local_id;
944 	struct ipsid_s	*ipl_local_cid;
945 	struct ipsid_s	*ipl_remote_cid;
946 	unsigned int
947 			ipl_out_action_latched : 1,
948 			ipl_in_action_latched : 1,
949 			ipl_out_policy_latched : 1,
950 			ipl_in_policy_latched : 1,
951 
952 			ipl_ids_latched : 1,
953 
954 			ipl_pad_to_bit_31 : 27;
955 
956 	ipsa_ref_t	ipl_ref[2]; /* 0: ESP, 1: AH */
957 
958 } ipsec_latch_t;
959 
960 #define	IPLATCH_REFHOLD(ipl) { \
961 	atomic_add_32(&(ipl)->ipl_refcnt, 1);		\
962 	ASSERT((ipl)->ipl_refcnt != 0);			\
963 }
964 
965 #define	IPLATCH_REFRELE(ipl, ns) {				\
966 	ASSERT((ipl)->ipl_refcnt != 0);				\
967 	membar_exit();						\
968 	if (atomic_add_32_nv(&(ipl)->ipl_refcnt, -1) == 0)	\
969 		iplatch_free(ipl, ns);			\
970 }
971 
972 /*
973  * peer identity structure.
974  */
975 
976 typedef struct conn_s conn_t;
977 
978 /*
979  * The old IP client structure "ipc_t" is gone. All the data is stored in the
980  * connection structure "conn_t" now. The mapping of old and new fields looks
981  * like this:
982  *
983  * ipc_ulp			conn_ulp
984  * ipc_rq			conn_rq
985  * ipc_wq			conn_wq
986  *
987  * ipc_laddr			conn_src
988  * ipc_faddr			conn_rem
989  * ipc_v6laddr			conn_srcv6
990  * ipc_v6faddr			conn_remv6
991  *
992  * ipc_lport			conn_lport
993  * ipc_fport			conn_fport
994  * ipc_ports			conn_ports
995  *
996  * ipc_policy			conn_policy
997  * ipc_latch			conn_latch
998  *
999  * ipc_irc_lock			conn_lock
1000  * ipc_ire_cache		conn_ire_cache
1001  *
1002  * ipc_state_flags		conn_state_flags
1003  * ipc_outgoing_ill		conn_outgoing_ill
1004  *
1005  * ipc_dontroute 		conn_dontroute
1006  * ipc_loopback 		conn_loopback
1007  * ipc_broadcast		conn_broadcast
1008  * ipc_reuseaddr		conn_reuseaddr
1009  *
1010  * ipc_multicast_loop		conn_multicast_loop
1011  * ipc_multi_router		conn_multi_router
1012  * ipc_draining 		conn_draining
1013  *
1014  * ipc_did_putbq		conn_did_putbq
1015  * ipc_unspec_src		conn_unspec_src
1016  * ipc_policy_cached		conn_policy_cached
1017  *
1018  * ipc_in_enforce_policy 	conn_in_enforce_policy
1019  * ipc_out_enforce_policy 	conn_out_enforce_policy
1020  * ipc_af_isv6			conn_af_isv6
1021  * ipc_pkt_isv6			conn_pkt_isv6
1022  *
1023  * ipc_ipv6_recvpktinfo		conn_ipv6_recvpktinfo
1024  *
1025  * ipc_ipv6_recvhoplimit	conn_ipv6_recvhoplimit
1026  * ipc_ipv6_recvhopopts		conn_ipv6_recvhopopts
1027  * ipc_ipv6_recvdstopts		conn_ipv6_recvdstopts
1028  *
1029  * ipc_ipv6_recvrthdr 		conn_ipv6_recvrthdr
1030  * ipc_ipv6_recvrtdstopts	conn_ipv6_recvrtdstopts
1031  * ipc_fully_bound		conn_fully_bound
1032  *
1033  * ipc_recvif			conn_recvif
1034  *
1035  * ipc_recvslla 		conn_recvslla
1036  * ipc_acking_unbind 		conn_acking_unbind
1037  * ipc_pad_to_bit_31 		conn_pad_to_bit_31
1038  *
1039  * ipc_nofailover_ill		conn_nofailover_ill
1040  *
1041  * ipc_proto			conn_proto
1042  * ipc_incoming_ill		conn_incoming_ill
1043  * ipc_outgoing_pill		conn_outgoing_pill
1044  * ipc_pending_ill		conn_pending_ill
1045  * ipc_unbind_mp		conn_unbind_mp
1046  * ipc_ilg			conn_ilg
1047  * ipc_ilg_allocated		conn_ilg_allocated
1048  * ipc_ilg_inuse		conn_ilg_inuse
1049  * ipc_ilg_walker_cnt		conn_ilg_walker_cnt
1050  * ipc_refcv			conn_refcv
1051  * ipc_multicast_ipif		conn_multicast_ipif
1052  * ipc_multicast_ill		conn_multicast_ill
1053  * ipc_orig_bound_ifindex	conn_orig_bound_ifindex
1054  * ipc_orig_multicast_ifindex	conn_orig_multicast_ifindex
1055  * ipc_drain_next		conn_drain_next
1056  * ipc_drain_prev		conn_drain_prev
1057  * ipc_idl			conn_idl
1058  */
1059 
1060 /*
1061  * This is used to match an inbound/outbound datagram with
1062  * policy.
1063  */
1064 
1065 typedef	struct ipsec_selector {
1066 	in6_addr_t	ips_local_addr_v6;
1067 	in6_addr_t	ips_remote_addr_v6;
1068 	uint16_t	ips_local_port;
1069 	uint16_t	ips_remote_port;
1070 	uint8_t		ips_icmp_type;
1071 	uint8_t		ips_icmp_code;
1072 	uint8_t		ips_protocol;
1073 	uint8_t		ips_isv4 : 1,
1074 			ips_is_icmp_inv_acq: 1;
1075 } ipsec_selector_t;
1076 
1077 /*
1078  * Note that we put v4 addresses in the *first* 32-bit word of the
1079  * selector rather than the last to simplify the prefix match/mask code
1080  * in spd.c
1081  */
1082 #define	ips_local_addr_v4 ips_local_addr_v6.s6_addr32[0]
1083 #define	ips_remote_addr_v4 ips_remote_addr_v6.s6_addr32[0]
1084 
1085 /* Values used in IP by IPSEC Code */
1086 #define		IPSEC_OUTBOUND		B_TRUE
1087 #define		IPSEC_INBOUND		B_FALSE
1088 
1089 /*
1090  * There are two variants in policy failures. The packet may come in
1091  * secure when not needed (IPSEC_POLICY_???_NOT_NEEDED) or it may not
1092  * have the desired level of protection (IPSEC_POLICY_MISMATCH).
1093  */
1094 #define	IPSEC_POLICY_NOT_NEEDED		0
1095 #define	IPSEC_POLICY_MISMATCH		1
1096 #define	IPSEC_POLICY_AUTH_NOT_NEEDED	2
1097 #define	IPSEC_POLICY_ENCR_NOT_NEEDED	3
1098 #define	IPSEC_POLICY_SE_NOT_NEEDED	4
1099 #define	IPSEC_POLICY_MAX		5	/* Always max + 1. */
1100 
1101 /*
1102  * Folowing macro is used whenever the code does not know whether there
1103  * is a M_CTL present in the front and it needs to examine the actual mp
1104  * i.e the IP header. As a M_CTL message could be in the front, this
1105  * extracts the packet into mp and the M_CTL mp into first_mp. If M_CTL
1106  * mp is not present, both first_mp and mp point to the same message.
1107  */
1108 #define	EXTRACT_PKT_MP(mp, first_mp, mctl_present)	\
1109 	(first_mp) = (mp);				\
1110 	if ((mp)->b_datap->db_type == M_CTL) {		\
1111 		(mp) = (mp)->b_cont;			\
1112 		(mctl_present) = B_TRUE;		\
1113 	} else {					\
1114 		(mctl_present) = B_FALSE;		\
1115 	}
1116 
1117 /*
1118  * Check with IPSEC inbound policy if
1119  *
1120  * 1) per-socket policy is present - indicated by conn_in_enforce_policy.
1121  * 2) Or if we have not cached policy on the conn and the global policy is
1122  *    non-empty.
1123  */
1124 #define	CONN_INBOUND_POLICY_PRESENT(connp, ipss)	\
1125 	((connp)->conn_in_enforce_policy ||		\
1126 	(!((connp)->conn_policy_cached) && 		\
1127 	(ipss)->ipsec_inbound_v4_policy_present))
1128 
1129 #define	CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)	\
1130 	((connp)->conn_in_enforce_policy ||		\
1131 	(!(connp)->conn_policy_cached &&		\
1132 	(ipss)->ipsec_inbound_v6_policy_present))
1133 
1134 #define	CONN_OUTBOUND_POLICY_PRESENT(connp, ipss)	\
1135 	((connp)->conn_out_enforce_policy ||		\
1136 	(!((connp)->conn_policy_cached) &&		\
1137 	(ipss)->ipsec_outbound_v4_policy_present))
1138 
1139 #define	CONN_OUTBOUND_POLICY_PRESENT_V6(connp, ipss)	\
1140 	((connp)->conn_out_enforce_policy ||		\
1141 	(!(connp)->conn_policy_cached &&		\
1142 	(ipss)->ipsec_outbound_v6_policy_present))
1143 
1144 /*
1145  * Information cached in IRE for upper layer protocol (ULP).
1146  *
1147  * Notice that ire_max_frag is not included in the iulp_t structure, which
1148  * it may seem that it should.  But ire_max_frag cannot really be cached.  It
1149  * is fixed for each interface.  For MTU found by PMTUd, we may want to cache
1150  * it.  But currently, we do not do that.
1151  */
1152 typedef struct iulp_s {
1153 	boolean_t	iulp_set;	/* Is any metric set? */
1154 	uint32_t	iulp_ssthresh;	/* Slow start threshold (TCP). */
1155 	clock_t		iulp_rtt;	/* Guestimate in millisecs. */
1156 	clock_t		iulp_rtt_sd;	/* Cached value of RTT variance. */
1157 	uint32_t	iulp_spipe;	/* Send pipe size. */
1158 	uint32_t	iulp_rpipe;	/* Receive pipe size. */
1159 	uint32_t	iulp_rtomax;	/* Max round trip timeout. */
1160 	uint32_t	iulp_sack;	/* Use SACK option (TCP)? */
1161 	uint32_t
1162 		iulp_tstamp_ok : 1,	/* Use timestamp option (TCP)? */
1163 		iulp_wscale_ok : 1,	/* Use window scale option (TCP)? */
1164 		iulp_ecn_ok : 1,	/* Enable ECN (for TCP)? */
1165 		iulp_pmtud_ok : 1,	/* Enable PMTUd? */
1166 
1167 		iulp_not_used : 28;
1168 } iulp_t;
1169 
1170 /* Zero iulp_t. */
1171 extern const iulp_t ire_uinfo_null;
1172 
1173 /*
1174  * The conn drain list structure.
1175  * The list is protected by idl_lock. Each conn_t inserted in the list
1176  * points back at this idl_t using conn_idl. IP primes the draining of the
1177  * conns queued in these lists, by qenabling the 1st conn of each list. This
1178  * occurs when STREAMS backenables ip_wsrv on the IP module. Each conn instance
1179  * of ip_wsrv successively qenables the next conn in the list.
1180  * idl_lock protects all other members of idl_t and conn_drain_next
1181  * and conn_drain_prev of conn_t. The conn_lock protects IPCF_DRAIN_DISABLED
1182  * flag of the conn_t and conn_idl.
1183  */
1184 typedef struct idl_s {
1185 	conn_t		*idl_conn;		/* Head of drain list */
1186 	kmutex_t	idl_lock;		/* Lock for this list */
1187 	conn_t		*idl_conn_draining;	/* conn that is draining */
1188 	uint32_t
1189 		idl_repeat : 1,			/* Last conn must re-enable */
1190 						/* drain list again */
1191 		idl_unused : 31;
1192 } idl_t;
1193 
1194 #define	CONN_DRAIN_LIST_LOCK(connp)	(&((connp)->conn_idl->idl_lock))
1195 /*
1196  * Interface route structure which holds the necessary information to recreate
1197  * routes that are tied to an interface (namely where ire_ipif != NULL).
1198  * These routes which were initially created via a routing socket or via the
1199  * SIOCADDRT ioctl may be gateway routes (RTF_GATEWAY being set) or may be
1200  * traditional interface routes.  When an interface comes back up after being
1201  * marked down, this information will be used to recreate the routes.  These
1202  * are part of an mblk_t chain that hangs off of the IPIF (ipif_saved_ire_mp).
1203  */
1204 typedef struct ifrt_s {
1205 	ushort_t	ifrt_type;		/* Type of IRE */
1206 	in6_addr_t	ifrt_v6addr;		/* Address IRE represents. */
1207 	in6_addr_t	ifrt_v6gateway_addr;	/* Gateway if IRE_OFFSUBNET */
1208 	in6_addr_t	ifrt_v6src_addr;	/* Src addr if RTF_SETSRC */
1209 	in6_addr_t	ifrt_v6mask;		/* Mask for matching IRE. */
1210 	uint32_t	ifrt_flags;		/* flags related to route */
1211 	uint_t		ifrt_max_frag;		/* MTU (next hop or path). */
1212 	iulp_t		ifrt_iulp_info;		/* Cached IRE ULP info. */
1213 } ifrt_t;
1214 
1215 #define	ifrt_addr		V4_PART_OF_V6(ifrt_v6addr)
1216 #define	ifrt_gateway_addr	V4_PART_OF_V6(ifrt_v6gateway_addr)
1217 #define	ifrt_src_addr		V4_PART_OF_V6(ifrt_v6src_addr)
1218 #define	ifrt_mask		V4_PART_OF_V6(ifrt_v6mask)
1219 
1220 /* Number of IP addresses that can be hosted on a physical interface */
1221 #define	MAX_ADDRS_PER_IF	8192
1222 /*
1223  * Number of Source addresses to be considered for source address
1224  * selection. Used by ipif_select_source[_v6].
1225  */
1226 #define	MAX_IPIF_SELECT_SOURCE	50
1227 
1228 #ifdef IP_DEBUG
1229 /*
1230  * Tracing refholds and refreleases for debugging. Existing tracing mechanisms
1231  * do not allow the granularity need to trace refrences to ipif/ill/ire's. This
1232  * mechanism should be revisited once dtrace is available.
1233  */
1234 #define	TR_STACK_DEPTH	14
1235 typedef struct tr_buf_s {
1236 	int	tr_depth;
1237 	clock_t	tr_time;
1238 	pc_t	tr_stack[TR_STACK_DEPTH];
1239 } tr_buf_t;
1240 
1241 typedef struct th_trace_s {
1242 	int		th_refcnt;
1243 	uint_t		th_trace_lastref;
1244 	kthread_t	*th_id;
1245 #define	TR_BUF_MAX	38
1246 	tr_buf_t	th_trbuf[TR_BUF_MAX];
1247 } th_trace_t;
1248 
1249 typedef struct th_hash_s {
1250 	list_node_t	thh_link;
1251 	mod_hash_t	*thh_hash;
1252 	ip_stack_t	*thh_ipst;
1253 } th_hash_t;
1254 #endif
1255 
1256 /* The following are ipif_state_flags */
1257 #define	IPIF_CONDEMNED		0x1	/* The ipif is being removed */
1258 #define	IPIF_CHANGING		0x2	/* A critcal ipif field is changing */
1259 #define	IPIF_MOVING		0x8	/* The ipif is being moved */
1260 #define	IPIF_SET_LINKLOCAL	0x10	/* transient flag during bringup */
1261 #define	IPIF_ZERO_SOURCE	0x20	/* transient flag during bringup */
1262 
1263 /* IP interface structure, one per local address */
1264 typedef struct ipif_s {
1265 	struct	ipif_s	*ipif_next;
1266 	struct	ill_s	*ipif_ill;	/* Back pointer to our ill */
1267 	int	ipif_id;		/* Logical unit number */
1268 	uint_t	ipif_mtu;		/* Starts at ipif_ill->ill_max_frag */
1269 	uint_t	ipif_saved_mtu;		/* Save of mtu during ipif_move() */
1270 	in6_addr_t ipif_v6lcl_addr;	/* Local IP address for this if. */
1271 	in6_addr_t ipif_v6src_addr;	/* Source IP address for this if. */
1272 	in6_addr_t ipif_v6subnet;	/* Subnet prefix for this if. */
1273 	in6_addr_t ipif_v6net_mask;	/* Net mask for this interface. */
1274 	in6_addr_t ipif_v6brd_addr;	/* Broadcast addr for this interface. */
1275 	in6_addr_t ipif_v6pp_dst_addr;	/* Point-to-point dest address. */
1276 	uint64_t ipif_flags;		/* Interface flags. */
1277 	uint_t	ipif_metric;		/* BSD if metric, for compatibility. */
1278 	uint_t	ipif_ire_type;		/* IRE_LOCAL or IRE_LOOPBACK */
1279 	mblk_t	*ipif_arp_del_mp;	/* Allocated at time arp comes up, to */
1280 					/* prevent awkward out of mem */
1281 					/* condition later */
1282 	mblk_t	*ipif_saved_ire_mp;	/* Allocated for each extra */
1283 					/* IRE_IF_NORESOLVER/IRE_IF_RESOLVER */
1284 					/* on this interface so that they */
1285 					/* can survive ifconfig down. */
1286 	kmutex_t ipif_saved_ire_lock;	/* Protects ipif_saved_ire_mp */
1287 
1288 	mrec_t	*ipif_igmp_rpt;		/* List of group memberships which */
1289 					/* will be reported on.  Used when */
1290 					/* handling an igmp timeout.	   */
1291 
1292 	/*
1293 	 * The packet counts in the ipif contain the sum of the
1294 	 * packet counts in dead IREs that were affiliated with
1295 	 * this ipif.
1296 	 */
1297 	uint_t	ipif_fo_pkt_count;	/* Forwarded thru our dead IREs */
1298 	uint_t	ipif_ib_pkt_count;	/* Inbound packets for our dead IREs */
1299 	uint_t	ipif_ob_pkt_count;	/* Outbound packets to our dead IREs */
1300 	/* Exclusive bit fields, protected by ipsq_t */
1301 	unsigned int
1302 		ipif_multicast_up : 1,	/* We have joined the allhosts group */
1303 		ipif_replace_zero : 1,	/* Replacement for zero */
1304 		ipif_was_up : 1,	/* ipif was up before */
1305 		ipif_addr_ready : 1,	/* DAD is done */
1306 
1307 		ipif_was_dup : 1,	/* DAD had failed */
1308 		ipif_pad_to_31 : 27;
1309 
1310 	int	ipif_orig_ifindex;	/* ifindex before SLIFFAILOVER */
1311 	uint_t	ipif_seqid;		/* unique index across all ills */
1312 	uint_t	ipif_orig_ipifid;	/* ipif_id before SLIFFAILOVER */
1313 	uint_t	ipif_state_flags;	/* See IPIF_* flag defs above */
1314 	uint_t	ipif_refcnt;		/* active consistent reader cnt */
1315 
1316 	/* Number of ire's and ilm's referencing this ipif */
1317 	uint_t	ipif_ire_cnt;
1318 	uint_t	ipif_ilm_cnt;
1319 
1320 	uint_t  ipif_saved_ire_cnt;
1321 	zoneid_t ipif_zoneid;		/* zone ID number */
1322 	timeout_id_t ipif_recovery_id;	/* Timer for DAD recovery */
1323 	boolean_t ipif_trace_disable;	/* True when alloc fails */
1324 } ipif_t;
1325 
1326 /*
1327  * IPIF_FREE_OK() means that there are no incoming references
1328  * to the ipif. Incoming refs would prevent the ipif from being freed.
1329  */
1330 #define	IPIF_FREE_OK(ipif)	\
1331 	((ipif)->ipif_ire_cnt == 0 && (ipif)->ipif_ilm_cnt == 0)
1332 /*
1333  * IPIF_DOWN_OK() determines whether the incoming pointer reference counts
1334  * would permit the ipif to be considered quiescent. In order for
1335  * an ipif or ill to be considered quiescent, the ire and nce references
1336  * to that ipif/ill must be zero.
1337  *
1338  * We do not require the ilm references to go to zero for quiescence
1339  * because the quiescence checks are done to ensure that
1340  * outgoing packets do not use addresses from the ipif/ill after it
1341  * has been marked down, and incoming packets to addresses on a
1342  * queiscent interface are rejected. This implies that all the
1343  * ire/nce's using that source address need to be deleted and future
1344  * creation of any ires using that source address must be prevented.
1345  * Similarly incoming unicast packets destined to the 'down' address
1346  * will not be accepted once that ire is gone. However incoming
1347  * multicast packets are not destined to the downed address.
1348  * They are only related to the ill in question. Furthermore
1349  * the current API behavior allows applications to join or leave
1350  * multicast groups, i.e., IP_ADD_MEMBERSHIP / LEAVE_MEMBERSHIP, using a
1351  * down address. Therefore the ilm references are not included in
1352  * the _DOWN_OK macros.
1353  */
1354 #define	IPIF_DOWN_OK(ipif)		((ipif)->ipif_ire_cnt == 0)
1355 
1356 /*
1357  * The following table lists the protection levels of the various members
1358  * of the ipif_t. The following notation is used.
1359  *
1360  * Write once - Written to only once at the time of bringing up
1361  * the interface and can be safely read after the bringup without any lock.
1362  *
1363  * ipsq - Need to execute in the ipsq to perform the indicated access.
1364  *
1365  * ill_lock - Need to hold this mutex to perform the indicated access.
1366  *
1367  * ill_g_lock - Need to hold this rw lock as reader/writer for read access or
1368  * write access respectively.
1369  *
1370  * down ill - Written to only when the ill is down (i.e all ipifs are down)
1371  * up ill - Read only when the ill is up (i.e. at least 1 ipif is up)
1372  *
1373  *		 Table of ipif_t members and their protection
1374  *
1375  * ipif_next		ipsq + ill_lock +	ipsq OR ill_lock OR
1376  *			ill_g_lock		ill_g_lock
1377  * ipif_ill		ipsq + down ipif	write once
1378  * ipif_id		ipsq + down ipif	write once
1379  * ipif_mtu		ipsq
1380  * ipif_v6lcl_addr	ipsq + down ipif	up ipif
1381  * ipif_v6src_addr	ipsq + down ipif	up ipif
1382  * ipif_v6subnet	ipsq + down ipif	up ipif
1383  * ipif_v6net_mask	ipsq + down ipif	up ipif
1384  *
1385  * ipif_v6brd_addr
1386  * ipif_v6pp_dst_addr
1387  * ipif_flags		ill_lock		ill_lock
1388  * ipif_metric
1389  * ipif_ire_type	ipsq + down ill		up ill
1390  *
1391  * ipif_arp_del_mp	ipsq			ipsq
1392  * ipif_saved_ire_mp	ipif_saved_ire_lock	ipif_saved_ire_lock
1393  * ipif_igmp_rpt	ipsq			ipsq
1394  *
1395  * ipif_fo_pkt_count	Approx
1396  * ipif_ib_pkt_count	Approx
1397  * ipif_ob_pkt_count	Approx
1398  *
1399  * bit fields		ill_lock		ill_lock
1400  *
1401  * ipif_orig_ifindex	ipsq			None
1402  * ipif_orig_ipifid	ipsq			None
1403  * ipif_seqid		ipsq			Write once
1404  *
1405  * ipif_state_flags	ill_lock		ill_lock
1406  * ipif_refcnt		ill_lock		ill_lock
1407  * ipif_ire_cnt		ill_lock		ill_lock
1408  * ipif_ilm_cnt		ill_lock		ill_lock
1409  * ipif_saved_ire_cnt
1410  */
1411 
1412 #define	IP_TR_HASH(tid)	((((uintptr_t)tid) >> 6) & (IP_TR_HASH_MAX - 1))
1413 
1414 #ifdef DEBUG
1415 #define	IPIF_TRACE_REF(ipif)	ipif_trace_ref(ipif)
1416 #define	ILL_TRACE_REF(ill)	ill_trace_ref(ill)
1417 #define	IPIF_UNTRACE_REF(ipif)	ipif_untrace_ref(ipif)
1418 #define	ILL_UNTRACE_REF(ill)	ill_untrace_ref(ill)
1419 #else
1420 #define	IPIF_TRACE_REF(ipif)
1421 #define	ILL_TRACE_REF(ill)
1422 #define	IPIF_UNTRACE_REF(ipif)
1423 #define	ILL_UNTRACE_REF(ill)
1424 #endif
1425 
1426 /* IPv4 compatability macros */
1427 #define	ipif_lcl_addr		V4_PART_OF_V6(ipif_v6lcl_addr)
1428 #define	ipif_src_addr		V4_PART_OF_V6(ipif_v6src_addr)
1429 #define	ipif_subnet		V4_PART_OF_V6(ipif_v6subnet)
1430 #define	ipif_net_mask		V4_PART_OF_V6(ipif_v6net_mask)
1431 #define	ipif_brd_addr		V4_PART_OF_V6(ipif_v6brd_addr)
1432 #define	ipif_pp_dst_addr	V4_PART_OF_V6(ipif_v6pp_dst_addr)
1433 
1434 /* Macros for easy backreferences to the ill. */
1435 #define	ipif_wq			ipif_ill->ill_wq
1436 #define	ipif_rq			ipif_ill->ill_rq
1437 #define	ipif_net_type		ipif_ill->ill_net_type
1438 #define	ipif_ipif_up_count	ipif_ill->ill_ipif_up_count
1439 #define	ipif_type		ipif_ill->ill_type
1440 #define	ipif_isv6		ipif_ill->ill_isv6
1441 
1442 #define	SIOCLIFADDR_NDX 112	/* ndx of SIOCLIFADDR in the ndx ioctl table */
1443 
1444 /*
1445  * mode value for ip_ioctl_finish for finishing an ioctl
1446  */
1447 #define	CONN_CLOSE	1		/* No mi_copy */
1448 #define	COPYOUT		2		/* do an mi_copyout if needed */
1449 #define	NO_COPYOUT	3		/* do an mi_copy_done */
1450 #define	IPI2MODE(ipi)	((ipi)->ipi_flags & IPI_GET_CMD ? COPYOUT : NO_COPYOUT)
1451 
1452 /*
1453  * The IP-MT design revolves around the serialization object ipsq_t.
1454  * It is associated with an IPMP group. If IPMP is not enabled, there is
1455  * 1 ipsq_t per phyint. Eg. an ipsq_t would cover both hme0's IPv4 stream
1456  *
1457  * ipsq_lock protects
1458  *	ipsq_reentry_cnt, ipsq_writer, ipsq_xopq_mphead, ipsq_xopq_mptail,
1459  *	ipsq_mphead, ipsq_mptail, ipsq_split
1460  *
1461  *	ipsq_pending_ipif, ipsq_current_ipif, ipsq_pending_mp, ipsq_flags,
1462  *	ipsq_waitfor
1463  *
1464  * The fields in the last line above below are set mostly by a writer thread
1465  * But there is an exception in the last call to ipif_ill_refrele_tail which
1466  * could also race with a conn close which could be cleaning up the
1467  * fields. So we choose to protect using ipsq_lock instead of depending on
1468  * the property of the writer.
1469  * ill_g_lock protects
1470  *	ipsq_refs, ipsq_phyint_list
1471  */
1472 typedef struct ipsq_s {
1473 	kmutex_t ipsq_lock;
1474 	int	ipsq_reentry_cnt;
1475 	kthread_t *ipsq_writer;		/* current owner (thread id) */
1476 	int	ipsq_flags;
1477 	mblk_t	*ipsq_xopq_mphead;	/* list of excl ops mostly ioctls */
1478 	mblk_t	*ipsq_xopq_mptail;
1479 	mblk_t	*ipsq_mphead;		/* msgs on ipsq linked thru b_next */
1480 	mblk_t	*ipsq_mptail;		/* msgs on ipsq linked thru b_next */
1481 	int	ipsq_current_ioctl;	/* current ioctl, or 0 if no ioctl */
1482 	boolean_t ipsq_current_done; 	/* is the current op done? */
1483 	ipif_t	*ipsq_current_ipif;	/* ipif associated with current op */
1484 	ipif_t	*ipsq_pending_ipif;	/* ipif associated w. ipsq_pending_mp */
1485 	mblk_t	*ipsq_pending_mp;	/* current ioctl mp while waiting for */
1486 					/* response from another module */
1487 	struct	ipsq_s	*ipsq_next;	/* list of all syncq's (ipsq_g_list) */
1488 	uint_t		ipsq_refs;	/* Number of phyints on this ipsq */
1489 	struct phyint	*ipsq_phyint_list; /* List of phyints on this ipsq */
1490 	boolean_t	ipsq_split;	/* ipsq may need to be split */
1491 	int		ipsq_waitfor;	/* Values encoded below */
1492 	char		ipsq_name[LIFNAMSIZ+1];	/* same as phyint_groupname */
1493 	ip_stack_t	*ipsq_ipst;	/* Does not have a netstack_hold */
1494 
1495 #ifdef DEBUG
1496 	int		ipsq_depth;	/* debugging aid */
1497 #define	IPSQ_STACK_DEPTH	15
1498 	pc_t		ipsq_stack[IPSQ_STACK_DEPTH];	/* debugging aid */
1499 #endif
1500 } ipsq_t;
1501 
1502 /* ipsq_flags */
1503 #define	IPSQ_GROUP	0x1	/* This ipsq belongs to an IPMP group */
1504 
1505 /*
1506  * ipsq_waitfor:
1507  *
1508  * IPIF_DOWN	1	ipif_down waiting for refcnts to drop
1509  * ILL_DOWN	2	ill_down waiting for refcnts to drop
1510  * IPIF_FREE	3	ipif_free waiting for refcnts to drop
1511  * ILL_FREE	4	ill unplumb waiting for refcnts to drop
1512  * ILL_MOVE_OK	5	failover waiting for refcnts to drop
1513  */
1514 
1515 enum { IPIF_DOWN = 1, ILL_DOWN, IPIF_FREE, ILL_FREE, ILL_MOVE_OK };
1516 
1517 /* Flags passed to ipsq_try_enter */
1518 #define	CUR_OP 0		/* Current ioctl continuing again */
1519 #define	NEW_OP 1		/* New ioctl starting afresh */
1520 
1521 /*
1522  * phyint represents state that is common to both IPv4 and IPv6 interfaces.
1523  * There is a separate ill_t representing IPv4 and IPv6 which has a
1524  * backpointer to the phyint structure for acessing common state.
1525  *
1526  * NOTE : It just stores the group name as there is only one name for
1527  *	  IPv4 and IPv6 i.e it is a underlying link property. Actually
1528  *        IPv4 and IPv6 ill are grouped together when their phyints have
1529  *        the same name.
1530  */
1531 typedef struct phyint {
1532 	struct ill_s	*phyint_illv4;
1533 	struct ill_s	*phyint_illv6;
1534 	uint_t		phyint_ifindex;		/* SIOCLSLIFINDEX */
1535 	char		*phyint_groupname;	/* SIOCSLIFGROUPNAME */
1536 	uint_t		phyint_groupname_len;
1537 	uint64_t	phyint_flags;
1538 	avl_node_t	phyint_avl_by_index;	/* avl tree by index */
1539 	avl_node_t	phyint_avl_by_name;	/* avl tree by name */
1540 	kmutex_t	phyint_lock;
1541 	struct ipsq_s	*phyint_ipsq;		/* back pointer to ipsq */
1542 	struct phyint	*phyint_ipsq_next;	/* phyint list on this ipsq */
1543 	/* Once Clearview IPMP is added the follow two fields can be removed */
1544 	uint_t		phyint_group_ifindex;	/* index assigned to group */
1545 	uint_t		phyint_hook_ifindex;	/* index used with neti/hook */
1546 } phyint_t;
1547 
1548 #define	CACHE_ALIGN_SIZE 64
1549 
1550 #define	CACHE_ALIGN(align_struct)	P2ROUNDUP(sizeof (struct align_struct),\
1551 							CACHE_ALIGN_SIZE)
1552 struct _phyint_list_s_ {
1553 	avl_tree_t	phyint_list_avl_by_index;	/* avl tree by index */
1554 	avl_tree_t	phyint_list_avl_by_name;	/* avl tree by name */
1555 };
1556 
1557 typedef union phyint_list_u {
1558 	struct	_phyint_list_s_ phyint_list_s;
1559 	char	phyint_list_filler[CACHE_ALIGN(_phyint_list_s_)];
1560 } phyint_list_t;
1561 
1562 #define	phyint_list_avl_by_index	phyint_list_s.phyint_list_avl_by_index
1563 #define	phyint_list_avl_by_name		phyint_list_s.phyint_list_avl_by_name
1564 /*
1565  * ILL groups. We group ills,
1566  *
1567  * - if the ills have the same group name. (New way)
1568  *
1569  * ill_group locking notes:
1570  *
1571  * illgrp_lock protects ill_grp_ill_schednext.
1572  *
1573  * ill_g_lock protects ill_grp_next, illgrp_ill, illgrp_ill_count.
1574  * Holding ill_g_lock freezes the memberships of ills in IPMP groups.
1575  * It also freezes the global list of ills and all ipifs in all ills.
1576  *
1577  * To remove an ipif from the linked list of ipifs of that ill ipif_free_tail
1578  * holds both ill_g_lock, and ill_lock. Similarly to remove an ill from the
1579  * global list of ills, ill_glist_delete() holds ill_g_lock as writer.
1580  * This simplifies things for ipif_select_source, illgrp_scheduler etc.
1581  * that need to walk the members of an illgrp. They just hold ill_g_lock
1582  * as reader to do the walk.
1583  *
1584  */
1585 typedef	struct ill_group {
1586 	kmutex_t	illgrp_lock;
1587 	struct ill_group *illgrp_next;		/* Next ill_group */
1588 	struct ill_s	*illgrp_ill_schednext;	/* Next ill to be scheduled */
1589 	struct ill_s	*illgrp_ill;		/* First ill in the group */
1590 	int		illgrp_ill_count;
1591 } ill_group_t;
1592 
1593 /*
1594  * Fragmentation hash bucket
1595  */
1596 typedef struct ipfb_s {
1597 	struct ipf_s	*ipfb_ipf;	/* List of ... */
1598 	size_t		ipfb_count;	/* Count of bytes used by frag(s) */
1599 	kmutex_t	ipfb_lock;	/* Protect all ipf in list */
1600 	uint_t		ipfb_frag_pkts; /* num of distinct fragmented pkts */
1601 } ipfb_t;
1602 
1603 /*
1604  * IRE bucket structure. Usually there is an array of such structures,
1605  * each pointing to a linked list of ires. irb_refcnt counts the number
1606  * of walkers of a given hash bucket. Usually the reference count is
1607  * bumped up if the walker wants no IRES to be DELETED while walking the
1608  * list. Bumping up does not PREVENT ADDITION. This allows walking a given
1609  * hash bucket without stumbling up on a free pointer.
1610  *
1611  * irb_t structures in ip_ftable are dynamically allocated and freed.
1612  * In order to identify the irb_t structures that can be safely kmem_free'd
1613  * we need to ensure that
1614  *  - the irb_refcnt is quiescent, indicating no other walkers,
1615  *  - no other threads or ire's are holding references to the irb,
1616  *	i.e., irb_nire == 0,
1617  *  - there are no active ire's in the bucket, i.e., irb_ire_cnt == 0
1618  */
1619 typedef struct irb {
1620 	struct ire_s	*irb_ire;	/* First ire in this bucket */
1621 					/* Should be first in this struct */
1622 	krwlock_t	irb_lock;	/* Protect this bucket */
1623 	uint_t		irb_refcnt;	/* Protected by irb_lock */
1624 	uchar_t		irb_marks;	/* CONDEMNED ires in this bucket ? */
1625 #define	IRB_MARK_CONDEMNED	0x0001
1626 #define	IRB_MARK_FTABLE		0x0002
1627 	uint_t		irb_ire_cnt;	/* Num of active IRE in this bucket */
1628 	uint_t		irb_tmp_ire_cnt; /* Num of temporary IRE */
1629 	struct ire_s	*irb_rr_origin;	/* origin for round-robin */
1630 	int		irb_nire;	/* Num of ftable ire's that ref irb */
1631 	ip_stack_t	*irb_ipst;	/* Does not have a netstack_hold */
1632 } irb_t;
1633 
1634 #define	IRB2RT(irb)	(rt_t *)((caddr_t)(irb) - offsetof(rt_t, rt_irb))
1635 
1636 /* The following are return values of ip_xmit_v4() */
1637 typedef enum {
1638 	SEND_PASSED = 0,	 /* sent packet out on wire */
1639 	SEND_FAILED,	 /* sending of packet failed */
1640 	LOOKUP_IN_PROGRESS, /* ire cache found, ARP resolution in progress */
1641 	LLHDR_RESLV_FAILED  /* macaddr resl of onlink dst or nexthop failed */
1642 } ipxmit_state_t;
1643 
1644 #define	IP_V4_G_HEAD	0
1645 #define	IP_V6_G_HEAD	1
1646 
1647 #define	MAX_G_HEADS	2
1648 
1649 /*
1650  * unpadded ill_if structure
1651  */
1652 struct 	_ill_if_s_ {
1653 	union ill_if_u	*illif_next;
1654 	union ill_if_u	*illif_prev;
1655 	avl_tree_t	illif_avl_by_ppa;	/* AVL tree sorted on ppa */
1656 	vmem_t		*illif_ppa_arena;	/* ppa index space */
1657 	uint16_t	illif_mcast_v1;		/* hints for		  */
1658 	uint16_t	illif_mcast_v2;		/* [igmp|mld]_slowtimo	  */
1659 	int		illif_name_len;		/* name length */
1660 	char		illif_name[LIFNAMSIZ];	/* name of interface type */
1661 };
1662 
1663 /* cache aligned ill_if structure */
1664 typedef union 	ill_if_u {
1665 	struct  _ill_if_s_ ill_if_s;
1666 	char 	illif_filler[CACHE_ALIGN(_ill_if_s_)];
1667 } ill_if_t;
1668 
1669 #define	illif_next		ill_if_s.illif_next
1670 #define	illif_prev		ill_if_s.illif_prev
1671 #define	illif_avl_by_ppa	ill_if_s.illif_avl_by_ppa
1672 #define	illif_ppa_arena		ill_if_s.illif_ppa_arena
1673 #define	illif_mcast_v1		ill_if_s.illif_mcast_v1
1674 #define	illif_mcast_v2		ill_if_s.illif_mcast_v2
1675 #define	illif_name		ill_if_s.illif_name
1676 #define	illif_name_len		ill_if_s.illif_name_len
1677 
1678 typedef struct ill_walk_context_s {
1679 	int	ctx_current_list; /* current list being searched */
1680 	int	ctx_last_list;	 /* last list to search */
1681 } ill_walk_context_t;
1682 
1683 /*
1684  * ill_g_heads structure, one for IPV4 and one for IPV6
1685  */
1686 struct _ill_g_head_s_ {
1687 	ill_if_t	*ill_g_list_head;
1688 	ill_if_t	*ill_g_list_tail;
1689 };
1690 
1691 typedef union ill_g_head_u {
1692 	struct _ill_g_head_s_ ill_g_head_s;
1693 	char	ill_g_head_filler[CACHE_ALIGN(_ill_g_head_s_)];
1694 } ill_g_head_t;
1695 
1696 #define	ill_g_list_head	ill_g_head_s.ill_g_list_head
1697 #define	ill_g_list_tail	ill_g_head_s.ill_g_list_tail
1698 
1699 #define	IP_V4_ILL_G_LIST(ipst)	\
1700 	(ipst)->ips_ill_g_heads[IP_V4_G_HEAD].ill_g_list_head
1701 #define	IP_V6_ILL_G_LIST(ipst)	\
1702 	(ipst)->ips_ill_g_heads[IP_V6_G_HEAD].ill_g_list_head
1703 #define	IP_VX_ILL_G_LIST(i, ipst)	\
1704 	(ipst)->ips_ill_g_heads[i].ill_g_list_head
1705 
1706 #define	ILL_START_WALK_V4(ctx_ptr, ipst)	\
1707 	ill_first(IP_V4_G_HEAD, IP_V4_G_HEAD, ctx_ptr, ipst)
1708 #define	ILL_START_WALK_V6(ctx_ptr, ipst)	\
1709 	ill_first(IP_V6_G_HEAD, IP_V6_G_HEAD, ctx_ptr, ipst)
1710 #define	ILL_START_WALK_ALL(ctx_ptr, ipst)	\
1711 	ill_first(MAX_G_HEADS, MAX_G_HEADS, ctx_ptr, ipst)
1712 
1713 /*
1714  * Capabilities, possible flags for ill_capabilities.
1715  */
1716 
1717 #define	ILL_CAPAB_AH		0x01		/* IPsec AH acceleration */
1718 #define	ILL_CAPAB_ESP		0x02		/* IPsec ESP acceleration */
1719 #define	ILL_CAPAB_MDT		0x04		/* Multidata Transmit */
1720 #define	ILL_CAPAB_HCKSUM	0x08		/* Hardware checksumming */
1721 #define	ILL_CAPAB_ZEROCOPY	0x10		/* Zero-copy */
1722 #define	ILL_CAPAB_POLL		0x20		/* Polling Toggle */
1723 #define	ILL_CAPAB_SOFT_RING	0x40		/* Soft_Ring capability */
1724 #define	ILL_CAPAB_LSO		0x80		/* Large Segment Offload */
1725 
1726 /*
1727  * Per-ill Multidata Transmit capabilities.
1728  */
1729 typedef struct ill_mdt_capab_s ill_mdt_capab_t;
1730 
1731 /*
1732  * Per-ill IPsec capabilities.
1733  */
1734 typedef struct ill_ipsec_capab_s ill_ipsec_capab_t;
1735 
1736 /*
1737  * Per-ill Hardware Checksumming capbilities.
1738  */
1739 typedef struct ill_hcksum_capab_s ill_hcksum_capab_t;
1740 
1741 /*
1742  * Per-ill Zero-copy capabilities.
1743  */
1744 typedef struct ill_zerocopy_capab_s ill_zerocopy_capab_t;
1745 
1746 /*
1747  * Per-ill Polling/soft ring capbilities.
1748  */
1749 typedef struct ill_dls_capab_s ill_dls_capab_t;
1750 
1751 /*
1752  * Per-ill polling resource map.
1753  */
1754 typedef struct ill_rx_ring ill_rx_ring_t;
1755 
1756 /*
1757  * Per-ill Large Segment Offload capabilities.
1758  */
1759 typedef struct ill_lso_capab_s ill_lso_capab_t;
1760 
1761 /* The following are ill_state_flags */
1762 #define	ILL_LL_SUBNET_PENDING	0x01	/* Waiting for DL_INFO_ACK from drv */
1763 #define	ILL_CONDEMNED		0x02	/* No more new ref's to the ILL */
1764 #define	ILL_CHANGING		0x04	/* ILL not globally visible */
1765 #define	ILL_DL_UNBIND_IN_PROGRESS	0x08	/* UNBIND_REQ is sent */
1766 #define	ILL_SOFT_RING_ASSIGN	0x10	/* Makeing soft ring assigment */
1767 
1768 /* Is this an ILL whose source address is used by other ILL's ? */
1769 #define	IS_USESRC_ILL(ill)			\
1770 	(((ill)->ill_usesrc_ifindex == 0) &&	\
1771 	((ill)->ill_usesrc_grp_next != NULL))
1772 
1773 /* Is this a client/consumer of the usesrc ILL ? */
1774 #define	IS_USESRC_CLI_ILL(ill)			\
1775 	(((ill)->ill_usesrc_ifindex != 0) &&	\
1776 	((ill)->ill_usesrc_grp_next != NULL))
1777 
1778 /* Is this an virtual network interface (vni) ILL ? */
1779 #define	IS_VNI(ill)							     \
1780 	(((ill) != NULL) &&						     \
1781 	(((ill)->ill_phyint->phyint_flags & (PHYI_LOOPBACK|PHYI_VIRTUAL)) == \
1782 	PHYI_VIRTUAL))
1783 
1784 /* Is this a loopback ILL? */
1785 #define	IS_LOOPBACK(ill) \
1786 	((ill)->ill_phyint->phyint_flags & PHYI_LOOPBACK)
1787 
1788 /*
1789  * IP Lower level Structure.
1790  * Instance data structure in ip_open when there is a device below us.
1791  */
1792 typedef struct ill_s {
1793 	ill_if_t *ill_ifptr;		/* pointer to interface type */
1794 	queue_t	*ill_rq;		/* Read queue. */
1795 	queue_t	*ill_wq;		/* Write queue. */
1796 
1797 	int	ill_error;		/* Error value sent up by device. */
1798 
1799 	ipif_t	*ill_ipif;		/* Interface chain for this ILL. */
1800 
1801 	uint_t	ill_ipif_up_count;	/* Number of IPIFs currently up. */
1802 	uint_t	ill_max_frag;		/* Max IDU from DLPI. */
1803 	char	*ill_name;		/* Our name. */
1804 	uint_t	ill_ipif_dup_count;	/* Number of duplicate addresses. */
1805 	uint_t	ill_name_length;	/* Name length, incl. terminator. */
1806 	char	*ill_ndd_name;		/* Name + ":ip?_forwarding" for NDD. */
1807 	uint_t	ill_net_type;		/* IRE_IF_RESOLVER/IRE_IF_NORESOLVER. */
1808 	/*
1809 	 * Physical Point of Attachment num.  If DLPI style 1 provider
1810 	 * then this is derived from the devname.
1811 	 */
1812 	uint_t	ill_ppa;
1813 	t_uscalar_t	ill_sap;
1814 	t_scalar_t	ill_sap_length;	/* Including sign (for position) */
1815 	uint_t	ill_phys_addr_length;	/* Excluding the sap. */
1816 	uint_t	ill_bcast_addr_length;	/* Only set when the DL provider */
1817 					/* supports broadcast. */
1818 	t_uscalar_t	ill_mactype;
1819 	uint8_t	*ill_frag_ptr;		/* Reassembly state. */
1820 	timeout_id_t ill_frag_timer_id; /* timeout id for the frag timer */
1821 	ipfb_t	*ill_frag_hash_tbl;	/* Fragment hash list head. */
1822 	ipif_t	*ill_pending_ipif;	/* IPIF waiting for DL operation. */
1823 
1824 	ilm_t	*ill_ilm;		/* Multicast mebership for lower ill */
1825 	uint_t	ill_global_timer;	/* for IGMPv3/MLDv2 general queries */
1826 	int	ill_mcast_type;		/* type of router which is querier */
1827 					/* on this interface */
1828 	uint16_t ill_mcast_v1_time;	/* # slow timeouts since last v1 qry */
1829 	uint16_t ill_mcast_v2_time;	/* # slow timeouts since last v2 qry */
1830 	uint8_t	ill_mcast_v1_tset;	/* 1 => timer is set; 0 => not set */
1831 	uint8_t	ill_mcast_v2_tset;	/* 1 => timer is set; 0 => not set */
1832 
1833 	uint8_t	ill_mcast_rv;		/* IGMPv3/MLDv2 robustness variable */
1834 	int	ill_mcast_qi;		/* IGMPv3/MLDv2 query interval var */
1835 
1836 	mblk_t	*ill_pending_mp;	/* IOCTL/DLPI awaiting completion. */
1837 	/*
1838 	 * All non-NULL cells between 'ill_first_mp_to_free' and
1839 	 * 'ill_last_mp_to_free' are freed in ill_delete.
1840 	 */
1841 #define	ill_first_mp_to_free	ill_bcast_mp
1842 	mblk_t	*ill_bcast_mp;		/* DLPI header for broadcasts. */
1843 	mblk_t	*ill_resolver_mp;	/* Resolver template. */
1844 	mblk_t	*ill_unbind_mp;		/* unbind mp from ill_dl_up() */
1845 	mblk_t	*ill_dlpi_deferred;	/* b_next chain of control messages */
1846 	mblk_t	*ill_phys_addr_mp;	/* mblk which holds ill_phys_addr */
1847 #define	ill_last_mp_to_free	ill_phys_addr_mp
1848 
1849 	cred_t	*ill_credp;		/* opener's credentials */
1850 	uint8_t	*ill_phys_addr;		/* ill_phys_addr_mp->b_rptr + off */
1851 
1852 	uint_t	ill_state_flags;	/* see ILL_* flags above */
1853 
1854 	/* Following bit fields protected by ipsq_t */
1855 	uint_t
1856 		ill_needs_attach : 1,
1857 		ill_reserved : 1,
1858 		ill_isv6 : 1,
1859 		ill_dlpi_style_set : 1,
1860 
1861 		ill_ifname_pending : 1,
1862 		ill_move_in_progress : 1, /* FAILOVER/FAILBACK in progress */
1863 		ill_join_allmulti : 1,
1864 		ill_logical_down : 1,
1865 
1866 		ill_is_6to4tun : 1,	/* Interface is a 6to4 tunnel */
1867 		ill_promisc_on_phys : 1, /* phys interface in promisc mode */
1868 		ill_dl_up : 1,
1869 		ill_up_ipifs : 1,
1870 
1871 		ill_note_link : 1,	/* supports link-up notification */
1872 		ill_capab_reneg : 1, /* capability renegotiation to be done */
1873 		ill_pad_to_bit_31 : 18;
1874 
1875 	/* Following bit fields protected by ill_lock */
1876 	uint_t
1877 		ill_fragtimer_executing : 1,
1878 		ill_fragtimer_needrestart : 1,
1879 		ill_ilm_cleanup_reqd : 1,
1880 		ill_arp_closing : 1,
1881 
1882 		ill_arp_bringup_pending : 1,
1883 		ill_mtu_userspecified : 1, /* SIOCSLIFLNKINFO has set the mtu */
1884 		ill_arp_extend : 1,	/* ARP has DAD extensions */
1885 		ill_pad_bit_31 : 25;
1886 
1887 	/*
1888 	 * Used in SIOCSIFMUXID and SIOCGIFMUXID for 'ifconfig unplumb'.
1889 	 */
1890 	int	ill_arp_muxid;		/* muxid returned from plink for arp */
1891 	int	ill_ip_muxid;		/* muxid returned from plink for ip */
1892 
1893 	/*
1894 	 * NIC event information attached, to be used by nic event hooks.
1895 	 */
1896 	hook_nic_event_t	*ill_nic_event_info;
1897 
1898 	/*
1899 	 * Used for IP frag reassembly throttling on a per ILL basis.
1900 	 *
1901 	 * Note: frag_count is approximate, its added to and subtracted from
1902 	 *	 without any locking, so simultaneous load/modify/stores can
1903 	 *	 collide, also ill_frag_purge() recalculates its value by
1904 	 *	 summing all the ipfb_count's without locking out updates
1905 	 *	 to the ipfb's.
1906 	 */
1907 	uint_t	ill_ipf_gen;		/* Generation of next fragment queue */
1908 	uint_t	ill_frag_count;		/* Approx count of all mblk bytes */
1909 	uint_t	ill_frag_free_num_pkts;	 /* num of fragmented packets to free */
1910 	clock_t	ill_last_frag_clean_time; /* time when frag's were pruned */
1911 	int	ill_type;		/* From <net/if_types.h> */
1912 	uint_t	ill_dlpi_multicast_state;	/* See below IDS_* */
1913 	uint_t	ill_dlpi_fastpath_state;	/* See below IDS_* */
1914 
1915 	/*
1916 	 * Capabilities related fields.
1917 	 */
1918 	uint_t  ill_dlpi_capab_state;	/* State of capability query, IDS_* */
1919 	uint64_t ill_capabilities;	/* Enabled capabilities, ILL_CAPAB_* */
1920 	ill_mdt_capab_t	*ill_mdt_capab;	/* Multidata Transmit capabilities */
1921 	ill_ipsec_capab_t *ill_ipsec_capab_ah;	/* IPsec AH capabilities */
1922 	ill_ipsec_capab_t *ill_ipsec_capab_esp;	/* IPsec ESP capabilities */
1923 	ill_hcksum_capab_t *ill_hcksum_capab; /* H/W cksumming capabilities */
1924 	ill_zerocopy_capab_t *ill_zerocopy_capab; /* Zero-copy capabilities */
1925 	ill_dls_capab_t *ill_dls_capab; /* Polling, soft ring capabilities */
1926 	ill_lso_capab_t *ill_lso_capab; /* Large Segment Offload capabilities */
1927 
1928 	/*
1929 	 * New fields for IPv6
1930 	 */
1931 	uint8_t	ill_max_hops;	/* Maximum hops for any logical interface */
1932 	uint_t	ill_max_mtu;	/* Maximum MTU for any logical interface */
1933 	uint32_t ill_reachable_time;	/* Value for ND algorithm in msec */
1934 	uint32_t ill_reachable_retrans_time; /* Value for ND algorithm msec */
1935 	uint_t	ill_max_buf;		/* Max # of req to buffer for ND */
1936 	in6_addr_t	ill_token;
1937 	uint_t		ill_token_length;
1938 	uint32_t	ill_xmit_count;		/* ndp max multicast xmits */
1939 	mib2_ipIfStatsEntry_t	*ill_ip_mib;	/* ver indep. interface mib */
1940 	mib2_ipv6IfIcmpEntry_t	*ill_icmp6_mib;	/* Per interface mib */
1941 	/*
1942 	 * Following two mblks are allocated common to all
1943 	 * the ipifs when the first interface is coming up.
1944 	 * It is sent up to arp when the last ipif is coming
1945 	 * down.
1946 	 */
1947 	mblk_t			*ill_arp_down_mp;
1948 	mblk_t			*ill_arp_del_mapping_mp;
1949 	/*
1950 	 * Used for implementing IFF_NOARP. As IFF_NOARP is used
1951 	 * to turn off for all the logicals, it is here instead
1952 	 * of the ipif.
1953 	 */
1954 	mblk_t			*ill_arp_on_mp;
1955 	/* Peer ill of an IPMP move operation */
1956 	struct ill_s		*ill_move_peer;
1957 
1958 	phyint_t		*ill_phyint;
1959 	uint64_t		ill_flags;
1960 	ill_group_t		*ill_group;
1961 	struct ill_s		*ill_group_next;
1962 
1963 	kmutex_t	ill_lock;	/* Please see table below */
1964 	/*
1965 	 * The ill_nd_lla* fields handle the link layer address option
1966 	 * from neighbor discovery. This is used for external IPv6
1967 	 * address resolution.
1968 	 */
1969 	mblk_t		*ill_nd_lla_mp;	/* mblk which holds ill_nd_lla */
1970 	uint8_t		*ill_nd_lla;	/* Link Layer Address */
1971 	uint_t		ill_nd_lla_len;	/* Link Layer Address length */
1972 	/*
1973 	 * We now have 3 phys_addr_req's sent down. This field keeps track
1974 	 * of which one is pending.
1975 	 */
1976 	t_uscalar_t	ill_phys_addr_pend; /* which dl_phys_addr_req pending */
1977 	/*
1978 	 * Used to save errors that occur during plumbing
1979 	 */
1980 	uint_t		ill_ifname_pending_err;
1981 	avl_node_t	ill_avl_byppa; /* avl node based on ppa */
1982 	void		*ill_fastpath_list; /* both ire and nce hang off this */
1983 	uint_t		ill_refcnt;	/* active refcnt by threads */
1984 	uint_t		ill_ire_cnt;	/* ires associated with this ill */
1985 	kcondvar_t	ill_cv;
1986 	uint_t		ill_ilm_walker_cnt;	/* snmp ilm walkers */
1987 	uint_t		ill_nce_cnt;	/* nces associated with this ill */
1988 	uint_t		ill_waiters;	/* threads waiting in ipsq_enter */
1989 	/*
1990 	 * Contains the upper read queue pointer of the module immediately
1991 	 * beneath IP.  This field allows IP to validate sub-capability
1992 	 * acknowledgments coming up from downstream.
1993 	 */
1994 	queue_t		*ill_lmod_rq;	/* read queue pointer of module below */
1995 	uint_t		ill_lmod_cnt;	/* number of modules beneath IP */
1996 	ip_m_t		*ill_media;	/* media specific params/functions */
1997 	t_uscalar_t	ill_dlpi_pending; /* Last DLPI primitive issued */
1998 	uint_t		ill_usesrc_ifindex; /* use src addr from this ILL */
1999 	struct ill_s	*ill_usesrc_grp_next; /* Next ILL in the usesrc group */
2000 	boolean_t	ill_trace_disable;	/* True when alloc fails */
2001 	zoneid_t	ill_zoneid;
2002 	ip_stack_t	*ill_ipst;	/* Corresponds to a netstack_hold */
2003 	uint32_t	ill_dhcpinit;	/* IP_DHCPINIT_IFs for ill */
2004 	uint_t		ill_ilm_cnt;    /* ilms referencing this ill */
2005 } ill_t;
2006 
2007 /*
2008  * ILL_FREE_OK() means that there are no incoming pointer references
2009  * to the ill.
2010  */
2011 #define	ILL_FREE_OK(ill)					\
2012 	((ill)->ill_ire_cnt == 0 && (ill)->ill_ilm_cnt == 0 &&	\
2013 	(ill)->ill_nce_cnt == 0)
2014 
2015 /*
2016  * An ipif/ill can be marked down only when the ire and nce references
2017  * to that ipif/ill goes to zero. ILL_DOWN_OK() is a necessary condition
2018  * quiescence checks. See comments above IPIF_DOWN_OK for details
2019  * on why ires and nces are selectively considered for this macro.
2020  */
2021 #define	ILL_DOWN_OK(ill)	(ill->ill_ire_cnt == 0 && ill->ill_nce_cnt == 0)
2022 
2023 /*
2024  * The following table lists the protection levels of the various members
2025  * of the ill_t. Same notation as that used for ipif_t above is used.
2026  *
2027  *				Write			Read
2028  *
2029  * ill_ifptr			ill_g_lock + s		Write once
2030  * ill_rq			ipsq			Write once
2031  * ill_wq			ipsq			Write once
2032  *
2033  * ill_error			ipsq			None
2034  * ill_ipif			ill_g_lock + ipsq	ill_g_lock OR ipsq
2035  * ill_ipif_up_count		ill_lock + ipsq		ill_lock OR ipsq
2036  * ill_max_frag			ipsq			Write once
2037  *
2038  * ill_name			ill_g_lock + ipsq	Write once
2039  * ill_name_length		ill_g_lock + ipsq	Write once
2040  * ill_ndd_name			ipsq			Write once
2041  * ill_net_type			ipsq			Write once
2042  * ill_ppa			ill_g_lock + ipsq	Write once
2043  * ill_sap			ipsq + down ill		Write once
2044  * ill_sap_length		ipsq + down ill		Write once
2045  * ill_phys_addr_length		ipsq + down ill		Write once
2046  *
2047  * ill_bcast_addr_length	ipsq			ipsq
2048  * ill_mactype			ipsq			ipsq
2049  * ill_frag_ptr			ipsq			ipsq
2050  *
2051  * ill_frag_timer_id		ill_lock		ill_lock
2052  * ill_frag_hash_tbl		ipsq			up ill
2053  * ill_ilm			ipsq + ill_lock		ill_lock
2054  * ill_mcast_type		ill_lock		ill_lock
2055  * ill_mcast_v1_time		ill_lock		ill_lock
2056  * ill_mcast_v2_time		ill_lock		ill_lock
2057  * ill_mcast_v1_tset		ill_lock		ill_lock
2058  * ill_mcast_v2_tset		ill_lock		ill_lock
2059  * ill_mcast_rv			ill_lock		ill_lock
2060  * ill_mcast_qi			ill_lock		ill_lock
2061  * ill_pending_mp		ill_lock		ill_lock
2062  *
2063  * ill_bcast_mp			ipsq			ipsq
2064  * ill_resolver_mp		ipsq			only when ill is up
2065  * ill_down_mp			ipsq			ipsq
2066  * ill_dlpi_deferred		ill_lock		ill_lock
2067  * ill_dlpi_pending		ill_lock		ill_lock
2068  * ill_phys_addr_mp		ipsq + down ill		only when ill is up
2069  * ill_phys_addr		ipsq + down ill		only when ill is up
2070  *
2071  * ill_state_flags		ill_lock		ill_lock
2072  * exclusive bit flags		ipsq_t			ipsq_t
2073  * shared bit flags		ill_lock		ill_lock
2074  *
2075  * ill_arp_muxid		ipsq			Not atomic
2076  * ill_ip_muxid			ipsq			Not atomic
2077  *
2078  * ill_ipf_gen			Not atomic
2079  * ill_frag_count		Approx. not protected
2080  * ill_type			ipsq + down ill		only when ill is up
2081  * ill_dlpi_multicast_state	ill_lock		ill_lock
2082  * ill_dlpi_fastpath_state	ill_lock		ill_lock
2083  * ill_max_hops			ipsq			Not atomic
2084  *
2085  * ill_max_mtu
2086  *
2087  * ill_reachable_time		ipsq + ill_lock		ill_lock
2088  * ill_reachable_retrans_time	ipsq + ill_lock		ill_lock
2089  * ill_max_buf			ipsq + ill_lock		ill_lock
2090  *
2091  * Next 2 fields need ill_lock because of the get ioctls. They should not
2092  * report partially updated results without executing in the ipsq.
2093  * ill_token			ipsq + ill_lock		ill_lock
2094  * ill_token_length		ipsq + ill_lock		ill_lock
2095  * ill_xmit_count		ipsq + down ill		write once
2096  * ill_ip6_mib			ipsq + down ill		only when ill is up
2097  * ill_icmp6_mib		ipsq + down ill		only when ill is up
2098  * ill_arp_down_mp		ipsq			ipsq
2099  * ill_arp_del_mapping_mp	ipsq			ipsq
2100  * ill_arp_on_mp		ipsq			ipsq
2101  * ill_move_peer		ipsq			ipsq
2102  *
2103  * ill_phyint			ipsq, ill_g_lock, ill_lock	Any of them
2104  * ill_flags			ill_lock		ill_lock
2105  * ill_group			ipsq, ill_g_lock, ill_lock	Any of them
2106  * ill_group_next		ipsq, ill_g_lock, ill_lock	Any of them
2107  * ill_nd_lla_mp		ipsq + down ill		only when ill is up
2108  * ill_nd_lla			ipsq + down ill		only when ill is up
2109  * ill_nd_lla_len		ipsq + down ill		only when ill is up
2110  * ill_phys_addr_pend		ipsq + down ill		only when ill is up
2111  * ill_ifname_pending_err	ipsq			ipsq
2112  * ill_avl_byppa		ipsq, ill_g_lock	write once
2113  *
2114  * ill_fastpath_list		ill_lock		ill_lock
2115  * ill_refcnt			ill_lock		ill_lock
2116  * ill_ire_cnt			ill_lock		ill_lock
2117  * ill_cv			ill_lock		ill_lock
2118  * ill_ilm_walker_cnt		ill_lock		ill_lock
2119  * ill_nce_cnt			ill_lock		ill_lock
2120  * ill_ilm_cnt			ill_lock		ill_lock
2121  * ill_trace			ill_lock		ill_lock
2122  * ill_usesrc_grp_next		ill_g_usesrc_lock	ill_g_usesrc_lock
2123  * ill_dhcpinit			atomics			atomics
2124  */
2125 
2126 /*
2127  * For ioctl restart mechanism see ip_reprocess_ioctl()
2128  */
2129 struct ip_ioctl_cmd_s;
2130 
2131 typedef	int (*ifunc_t)(ipif_t *, struct sockaddr_in *, queue_t *, mblk_t *,
2132     struct ip_ioctl_cmd_s *, void *);
2133 
2134 typedef struct ip_ioctl_cmd_s {
2135 	int	ipi_cmd;
2136 	size_t	ipi_copyin_size;
2137 	uint_t	ipi_flags;
2138 	uint_t	ipi_cmd_type;
2139 	ifunc_t	ipi_func;
2140 	ifunc_t	ipi_func_restart;
2141 } ip_ioctl_cmd_t;
2142 
2143 /*
2144  * ipi_cmd_type:
2145  *
2146  * IF_CMD		1	old style ifreq cmd
2147  * LIF_CMD		2	new style lifreq cmd
2148  * TUN_CMD		3	tunnel related
2149  * ARP_CMD		4	arpreq cmd
2150  * XARP_CMD		5	xarpreq cmd
2151  * MSFILT_CMD		6	multicast source filter cmd
2152  * MISC_CMD		7	misc cmd (not a more specific one above)
2153  */
2154 
2155 enum { IF_CMD = 1, LIF_CMD, TUN_CMD, ARP_CMD, XARP_CMD, MSFILT_CMD, MISC_CMD };
2156 
2157 #define	IPI_DONTCARE	0	/* For ioctl encoded values that don't matter */
2158 
2159 /* Flag values in ipi_flags */
2160 #define	IPI_PRIV	0x1	/* Root only command */
2161 #define	IPI_MODOK	0x2	/* Permitted on mod instance of IP */
2162 #define	IPI_WR		0x4	/* Need to grab writer access */
2163 #define	IPI_GET_CMD	0x8	/* branch to mi_copyout on success */
2164 #define	IPI_REPL	0x10	/* valid for replacement ipif created in MOVE */
2165 #define	IPI_NULL_BCONT	0x20	/* ioctl has not data and hence no b_cont */
2166 #define	IPI_PASS_DOWN	0x40	/* pass this ioctl down when a module only */
2167 
2168 extern ip_ioctl_cmd_t	ip_ndx_ioctl_table[];
2169 extern ip_ioctl_cmd_t	ip_misc_ioctl_table[];
2170 extern int ip_ndx_ioctl_count;
2171 extern int ip_misc_ioctl_count;
2172 
2173 #define	ILL_CLEAR_MOVE(ill) {				\
2174 	ill_t *peer_ill;				\
2175 							\
2176 	peer_ill = (ill)->ill_move_peer;		\
2177 	ASSERT(peer_ill != NULL);			\
2178 	(ill)->ill_move_in_progress = B_FALSE;		\
2179 	peer_ill->ill_move_in_progress = B_FALSE;	\
2180 	(ill)->ill_move_peer = NULL;			\
2181 	peer_ill->ill_move_peer = NULL;			\
2182 }
2183 
2184 /* Passed down by ARP to IP during I_PLINK/I_PUNLINK */
2185 typedef struct ipmx_s {
2186 	char	ipmx_name[LIFNAMSIZ];		/* if name */
2187 	uint_t
2188 		ipmx_arpdev_stream : 1,		/* This is the arp stream */
2189 		ipmx_notused : 31;
2190 } ipmx_t;
2191 
2192 /*
2193  * State for detecting if a driver supports certain features.
2194  * Support for DL_ENABMULTI_REQ uses ill_dlpi_multicast_state.
2195  * Support for DLPI M_DATA fastpath uses ill_dlpi_fastpath_state.
2196  * Support for DL_CAPABILITY_REQ uses ill_dlpi_capab_state.
2197  */
2198 #define	IDS_UNKNOWN	0	/* No DLPI request sent */
2199 #define	IDS_INPROGRESS	1	/* DLPI request sent */
2200 #define	IDS_OK		2	/* DLPI request completed successfully */
2201 #define	IDS_FAILED	3	/* DLPI request failed */
2202 
2203 /* Named Dispatch Parameter Management Structure */
2204 typedef struct ipparam_s {
2205 	uint_t	ip_param_min;
2206 	uint_t	ip_param_max;
2207 	uint_t	ip_param_value;
2208 	char	*ip_param_name;
2209 } ipparam_t;
2210 
2211 /* Extended NDP Management Structure */
2212 typedef struct ipndp_s {
2213 	ndgetf_t	ip_ndp_getf;
2214 	ndsetf_t	ip_ndp_setf;
2215 	caddr_t		ip_ndp_data;
2216 	char		*ip_ndp_name;
2217 } ipndp_t;
2218 
2219 /*
2220  * The kernel stores security attributes of all gateways in a database made
2221  * up of one or more tsol_gcdb_t elements.  Each tsol_gcdb_t contains the
2222  * security-related credentials of the gateway.  More than one gateways may
2223  * share entries in the database.
2224  *
2225  * The tsol_gc_t structure represents the gateway to credential association,
2226  * and refers to an entry in the database.  One or more tsol_gc_t entities are
2227  * grouped together to form one or more tsol_gcgrp_t, each representing the
2228  * list of security attributes specific to the gateway.  A gateway may be
2229  * associated with at most one credentials group.
2230  */
2231 struct tsol_gcgrp_s;
2232 
2233 extern uchar_t	ip6opt_ls;	/* TX IPv6 enabler */
2234 
2235 /*
2236  * Gateway security credential record.
2237  */
2238 typedef struct tsol_gcdb_s {
2239 	uint_t		gcdb_refcnt;	/* reference count */
2240 	struct rtsa_s	gcdb_attr;	/* security attributes */
2241 #define	gcdb_mask	gcdb_attr.rtsa_mask
2242 #define	gcdb_doi	gcdb_attr.rtsa_doi
2243 #define	gcdb_slrange	gcdb_attr.rtsa_slrange
2244 } tsol_gcdb_t;
2245 
2246 /*
2247  * Gateway to credential association.
2248  */
2249 typedef struct tsol_gc_s {
2250 	uint_t		gc_refcnt;	/* reference count */
2251 	struct tsol_gcgrp_s *gc_grp;	/* pointer to group */
2252 	struct tsol_gc_s *gc_prev;	/* previous in list */
2253 	struct tsol_gc_s *gc_next;	/* next in list */
2254 	tsol_gcdb_t	*gc_db;		/* pointer to actual credentials */
2255 } tsol_gc_t;
2256 
2257 /*
2258  * Gateway credentials group address.
2259  */
2260 typedef struct tsol_gcgrp_addr_s {
2261 	int		ga_af;		/* address family */
2262 	in6_addr_t	ga_addr;	/* IPv4 mapped or IPv6 address */
2263 } tsol_gcgrp_addr_t;
2264 
2265 /*
2266  * Gateway credentials group.
2267  */
2268 typedef struct tsol_gcgrp_s {
2269 	uint_t		gcgrp_refcnt;	/* reference count */
2270 	krwlock_t	gcgrp_rwlock;	/* lock to protect following */
2271 	uint_t		gcgrp_count;	/* number of credentials */
2272 	tsol_gc_t	*gcgrp_head;	/* first credential in list */
2273 	tsol_gc_t	*gcgrp_tail;	/* last credential in list */
2274 	tsol_gcgrp_addr_t gcgrp_addr;	/* next-hop gateway address */
2275 } tsol_gcgrp_t;
2276 
2277 extern kmutex_t gcgrp_lock;
2278 
2279 #define	GC_REFRELE(p) {				\
2280 	ASSERT((p)->gc_grp != NULL);		\
2281 	rw_enter(&(p)->gc_grp->gcgrp_rwlock, RW_WRITER); \
2282 	ASSERT((p)->gc_refcnt > 0);		\
2283 	if (--((p)->gc_refcnt) == 0)		\
2284 		gc_inactive(p);			\
2285 	else					\
2286 		rw_exit(&(p)->gc_grp->gcgrp_rwlock); \
2287 }
2288 
2289 #define	GCGRP_REFHOLD(p) {			\
2290 	mutex_enter(&gcgrp_lock);		\
2291 	++((p)->gcgrp_refcnt);			\
2292 	ASSERT((p)->gcgrp_refcnt != 0);		\
2293 	mutex_exit(&gcgrp_lock);		\
2294 }
2295 
2296 #define	GCGRP_REFRELE(p) {			\
2297 	mutex_enter(&gcgrp_lock);		\
2298 	ASSERT((p)->gcgrp_refcnt > 0);		\
2299 	if (--((p)->gcgrp_refcnt) == 0)		\
2300 		gcgrp_inactive(p);		\
2301 	ASSERT(MUTEX_HELD(&gcgrp_lock));	\
2302 	mutex_exit(&gcgrp_lock);		\
2303 }
2304 
2305 /*
2306  * IRE gateway security attributes structure, pointed to by tsol_ire_gw_secattr
2307  */
2308 struct tsol_tnrhc;
2309 
2310 typedef struct tsol_ire_gw_secattr_s {
2311 	kmutex_t	igsa_lock;	/* lock to protect following */
2312 	struct tsol_tnrhc *igsa_rhc;	/* host entry for gateway */
2313 	tsol_gc_t	*igsa_gc;	/* for prefix IREs */
2314 	tsol_gcgrp_t	*igsa_gcgrp;	/* for cache IREs */
2315 } tsol_ire_gw_secattr_t;
2316 
2317 /*
2318  * Following are the macros to increment/decrement the reference
2319  * count of the IREs and IRBs (ire bucket).
2320  *
2321  * 1) We bump up the reference count of an IRE to make sure that
2322  *    it does not get deleted and freed while we are using it.
2323  *    Typically all the lookup functions hold the bucket lock,
2324  *    and look for the IRE. If it finds an IRE, it bumps up the
2325  *    reference count before dropping the lock. Sometimes we *may* want
2326  *    to bump up the reference count after we *looked* up i.e without
2327  *    holding the bucket lock. So, the IRE_REFHOLD macro does not assert
2328  *    on the bucket lock being held. Any thread trying to delete from
2329  *    the hash bucket can still do so but cannot free the IRE if
2330  *    ire_refcnt is not 0.
2331  *
2332  * 2) We bump up the reference count on the bucket where the IRE resides
2333  *    (IRB), when we want to prevent the IREs getting deleted from a given
2334  *    hash bucket. This makes life easier for ire_walk type functions which
2335  *    wants to walk the IRE list, call a function, but needs to drop
2336  *    the bucket lock to prevent recursive rw_enters. While the
2337  *    lock is dropped, the list could be changed by other threads or
2338  *    the same thread could end up deleting the ire or the ire pointed by
2339  *    ire_next. IRE_REFHOLDing the ire or ire_next is not sufficient as
2340  *    a delete will still remove the ire from the bucket while we have
2341  *    dropped the lock and hence the ire_next would be NULL. Thus, we
2342  *    need a mechanism to prevent deletions from a given bucket.
2343  *
2344  *    To prevent deletions, we bump up the reference count on the
2345  *    bucket. If the bucket is held, ire_delete just marks IRE_MARK_CONDEMNED
2346  *    both on the ire's ire_marks and the bucket's irb_marks. When the
2347  *    reference count on the bucket drops to zero, all the CONDEMNED ires
2348  *    are deleted. We don't have to bump up the reference count on the
2349  *    bucket if we are walking the bucket and never have to drop the bucket
2350  *    lock. Note that IRB_REFHOLD does not prevent addition of new ires
2351  *    in the list. It is okay because addition of new ires will not cause
2352  *    ire_next to point to freed memory. We do IRB_REFHOLD only when
2353  *    all of the 3 conditions are true :
2354  *
2355  *    1) The code needs to walk the IRE bucket from start to end.
2356  *    2) It may have to drop the bucket lock sometimes while doing (1)
2357  *    3) It does not want any ires to be deleted meanwhile.
2358  */
2359 
2360 /*
2361  * Bump up the reference count on the IRE. We cannot assert that the
2362  * bucket lock is being held as it is legal to bump up the reference
2363  * count after the first lookup has returned the IRE without
2364  * holding the lock. Currently ip_wput does this for caching IRE_CACHEs.
2365  */
2366 
2367 #ifdef DEBUG
2368 #define	IRE_UNTRACE_REF(ire)	ire_untrace_ref(ire);
2369 #define	IRE_TRACE_REF(ire)	ire_trace_ref(ire);
2370 #else
2371 #define	IRE_UNTRACE_REF(ire)
2372 #define	IRE_TRACE_REF(ire)
2373 #endif
2374 
2375 #define	IRE_REFHOLD_NOTR(ire) {				\
2376 	atomic_add_32(&(ire)->ire_refcnt, 1);		\
2377 	ASSERT((ire)->ire_refcnt != 0);			\
2378 }
2379 
2380 #define	IRE_REFHOLD(ire) {				\
2381 	IRE_REFHOLD_NOTR(ire);				\
2382 	IRE_TRACE_REF(ire);				\
2383 }
2384 
2385 #define	IRE_REFHOLD_LOCKED(ire)	{			\
2386 	IRE_TRACE_REF(ire);				\
2387 	(ire)->ire_refcnt++;				\
2388 }
2389 
2390 /*
2391  * Decrement the reference count on the IRE.
2392  * In architectures e.g sun4u, where atomic_add_32_nv is just
2393  * a cas, we need to maintain the right memory barrier semantics
2394  * as that of mutex_exit i.e all the loads and stores should complete
2395  * before the cas is executed. membar_exit() does that here.
2396  *
2397  * NOTE : This macro is used only in places where we want performance.
2398  *	  To avoid bloating the code, we use the function "ire_refrele"
2399  *	  which essentially calls the macro.
2400  */
2401 #define	IRE_REFRELE_NOTR(ire) {					\
2402 	ASSERT((ire)->ire_refcnt != 0);				\
2403 	membar_exit();						\
2404 	if (atomic_add_32_nv(&(ire)->ire_refcnt, -1) == 0)	\
2405 		ire_inactive(ire);				\
2406 }
2407 
2408 #define	IRE_REFRELE(ire) {					\
2409 	if (ire->ire_bucket != NULL) {				\
2410 		IRE_UNTRACE_REF(ire);				\
2411 	}							\
2412 	IRE_REFRELE_NOTR(ire);					\
2413 }
2414 
2415 /*
2416  * Bump up the reference count on the hash bucket - IRB to
2417  * prevent ires from being deleted in this bucket.
2418  */
2419 #define	IRB_REFHOLD(irb) {				\
2420 	rw_enter(&(irb)->irb_lock, RW_WRITER);		\
2421 	(irb)->irb_refcnt++;				\
2422 	ASSERT((irb)->irb_refcnt != 0);			\
2423 	rw_exit(&(irb)->irb_lock);			\
2424 }
2425 #define	IRB_REFHOLD_LOCKED(irb) {			\
2426 	ASSERT(RW_WRITE_HELD(&(irb)->irb_lock));	\
2427 	(irb)->irb_refcnt++;				\
2428 	ASSERT((irb)->irb_refcnt != 0);			\
2429 }
2430 
2431 void irb_refrele_ftable(irb_t *);
2432 /*
2433  * Note: when IRB_MARK_FTABLE (i.e., IRE_CACHETABLE entry), the irb_t
2434  * is statically allocated, so that when the irb_refcnt goes to 0,
2435  * we simply clean up the ire list and continue.
2436  */
2437 #define	IRB_REFRELE(irb) {				\
2438 	if ((irb)->irb_marks & IRB_MARK_FTABLE) {	\
2439 		irb_refrele_ftable((irb));		\
2440 	} else {					\
2441 		rw_enter(&(irb)->irb_lock, RW_WRITER);		\
2442 		ASSERT((irb)->irb_refcnt != 0);			\
2443 		if (--(irb)->irb_refcnt	== 0 &&			\
2444 		    ((irb)->irb_marks & IRE_MARK_CONDEMNED)) {	\
2445 			ire_t *ire_list;			\
2446 								\
2447 			ire_list = ire_unlink(irb);		\
2448 			rw_exit(&(irb)->irb_lock);		\
2449 			ASSERT(ire_list != NULL);		\
2450 			ire_cleanup(ire_list);			\
2451 		} else {					\
2452 			rw_exit(&(irb)->irb_lock);		\
2453 		}						\
2454 	}							\
2455 }
2456 
2457 extern struct kmem_cache *rt_entry_cache;
2458 
2459 /*
2460  * Lock the fast path mp for access, since the fp_mp can be deleted
2461  * due a DL_NOTE_FASTPATH_FLUSH in the case of IRE_BROADCAST
2462  */
2463 
2464 #define	LOCK_IRE_FP_MP(ire) {				\
2465 		if ((ire)->ire_type == IRE_BROADCAST)	\
2466 			mutex_enter(&ire->ire_nce->nce_lock);	\
2467 	}
2468 #define	UNLOCK_IRE_FP_MP(ire) {				\
2469 		if ((ire)->ire_type == IRE_BROADCAST)	\
2470 			mutex_exit(&ire->ire_nce->nce_lock);	\
2471 	}
2472 
2473 typedef struct ire4 {
2474 	ipaddr_t ire4_src_addr;		/* Source address to use. */
2475 	ipaddr_t ire4_mask;		/* Mask for matching this IRE. */
2476 	ipaddr_t ire4_addr;		/* Address this IRE represents. */
2477 	ipaddr_t ire4_gateway_addr;	/* Gateway if IRE_CACHE/IRE_OFFSUBNET */
2478 	ipaddr_t ire4_cmask;		/* Mask from parent prefix route */
2479 } ire4_t;
2480 
2481 typedef struct ire6 {
2482 	in6_addr_t ire6_src_addr;	/* Source address to use. */
2483 	in6_addr_t ire6_mask;		/* Mask for matching this IRE. */
2484 	in6_addr_t ire6_addr;		/* Address this IRE represents. */
2485 	in6_addr_t ire6_gateway_addr;	/* Gateway if IRE_CACHE/IRE_OFFSUBNET */
2486 	in6_addr_t ire6_cmask;		/* Mask from parent prefix route */
2487 } ire6_t;
2488 
2489 typedef union ire_addr {
2490 	ire6_t	ire6_u;
2491 	ire4_t	ire4_u;
2492 } ire_addr_u_t;
2493 
2494 /* Internet Routing Entry */
2495 typedef struct ire_s {
2496 	struct	ire_s	*ire_next;	/* The hash chain must be first. */
2497 	struct	ire_s	**ire_ptpn;	/* Pointer to previous next. */
2498 	uint32_t	ire_refcnt;	/* Number of references */
2499 	mblk_t		*ire_mp;	/* Non-null if allocated as mblk */
2500 	queue_t		*ire_rfq;	/* recv from this queue */
2501 	queue_t		*ire_stq;	/* send to this queue */
2502 	union {
2503 		uint_t	*max_fragp;	/* Used only during ire creation */
2504 		uint_t	max_frag;	/* MTU (next hop or path). */
2505 	} imf_u;
2506 #define	ire_max_frag	imf_u.max_frag
2507 #define	ire_max_fragp	imf_u.max_fragp
2508 	uint32_t	ire_frag_flag;	/* IPH_DF or zero. */
2509 	uint32_t	ire_ident;	/* Per IRE IP ident. */
2510 	uint32_t	ire_tire_mark;	/* Used for reclaim of unused. */
2511 	uchar_t		ire_ipversion;	/* IPv4/IPv6 version */
2512 	uchar_t		ire_marks;	/* IRE_MARK_CONDEMNED etc. */
2513 	ushort_t	ire_type;	/* Type of IRE */
2514 	uint_t	ire_ib_pkt_count;	/* Inbound packets for ire_addr */
2515 	uint_t	ire_ob_pkt_count;	/* Outbound packets to ire_addr */
2516 	uint_t	ire_ll_hdr_length;	/* Non-zero if we do M_DATA prepends */
2517 	time_t	ire_create_time;	/* Time (in secs) IRE was created. */
2518 	uint32_t	ire_phandle;	/* Associate prefix IREs to cache */
2519 	uint32_t	ire_ihandle;	/* Associate interface IREs to cache */
2520 	ipif_t		*ire_ipif;	/* the interface that this ire uses */
2521 	uint32_t	ire_flags;	/* flags related to route (RTF_*) */
2522 	uint_t	ire_ipsec_overhead;	/* IPSEC overhead */
2523 	/*
2524 	 * Neighbor Cache Entry for IPv6; arp info for IPv4
2525 	 */
2526 	struct	nce_s	*ire_nce;
2527 	uint_t		ire_masklen;	/* # bits in ire_mask{,_v6} */
2528 	ire_addr_u_t	ire_u;		/* IPv4/IPv6 address info. */
2529 
2530 	irb_t		*ire_bucket;	/* Hash bucket when ire_ptphn is set */
2531 	iulp_t		ire_uinfo;	/* Upper layer protocol info. */
2532 	/*
2533 	 * Protects ire_uinfo, ire_max_frag, and ire_frag_flag.
2534 	 */
2535 	kmutex_t	ire_lock;
2536 	uint_t		ire_ipif_seqid; /* ipif_seqid of ire_ipif */
2537 	clock_t		ire_last_used_time;	/* Last used time */
2538 	tsol_ire_gw_secattr_t *ire_gw_secattr; /* gateway security attributes */
2539 	zoneid_t	ire_zoneid;	/* for local address discrimination */
2540 	/*
2541 	 * ire's that are embedded inside mblk_t and sent to the external
2542 	 * resolver use the ire_stq_ifindex to track the ifindex of the
2543 	 * ire_stq, so that the ill (if it exists) can be correctly recovered
2544 	 * for cleanup in the esbfree routine when arp failure occurs
2545 	 */
2546 	uint_t	ire_stq_ifindex;
2547 	uint_t		ire_defense_count;	/* number of ARP conflicts */
2548 	uint_t		ire_defense_time;	/* last time defended (secs) */
2549 	boolean_t	ire_trace_disable;	/* True when alloc fails */
2550 	ip_stack_t	*ire_ipst;	/* Does not have a netstack_hold */
2551 } ire_t;
2552 
2553 /* IPv4 compatiblity macros */
2554 #define	ire_src_addr		ire_u.ire4_u.ire4_src_addr
2555 #define	ire_mask		ire_u.ire4_u.ire4_mask
2556 #define	ire_addr		ire_u.ire4_u.ire4_addr
2557 #define	ire_gateway_addr	ire_u.ire4_u.ire4_gateway_addr
2558 #define	ire_cmask		ire_u.ire4_u.ire4_cmask
2559 
2560 #define	ire_src_addr_v6		ire_u.ire6_u.ire6_src_addr
2561 #define	ire_mask_v6		ire_u.ire6_u.ire6_mask
2562 #define	ire_addr_v6		ire_u.ire6_u.ire6_addr
2563 #define	ire_gateway_addr_v6	ire_u.ire6_u.ire6_gateway_addr
2564 #define	ire_cmask_v6		ire_u.ire6_u.ire6_cmask
2565 
2566 /* Convenient typedefs for sockaddrs */
2567 typedef	struct sockaddr_in	sin_t;
2568 typedef	struct sockaddr_in6	sin6_t;
2569 
2570 /* Address structure used for internal bind with IP */
2571 typedef struct ipa_conn_s {
2572 	ipaddr_t	ac_laddr;
2573 	ipaddr_t	ac_faddr;
2574 	uint16_t	ac_fport;
2575 	uint16_t	ac_lport;
2576 } ipa_conn_t;
2577 
2578 typedef struct ipa6_conn_s {
2579 	in6_addr_t	ac6_laddr;
2580 	in6_addr_t	ac6_faddr;
2581 	uint16_t	ac6_fport;
2582 	uint16_t	ac6_lport;
2583 } ipa6_conn_t;
2584 
2585 /*
2586  * Using ipa_conn_x_t or ipa6_conn_x_t allows us to modify the behavior of IP's
2587  * bind handler.
2588  */
2589 typedef struct ipa_conn_extended_s {
2590 	uint64_t	acx_flags;
2591 	ipa_conn_t	acx_conn;
2592 } ipa_conn_x_t;
2593 
2594 typedef struct ipa6_conn_extended_s {
2595 	uint64_t	ac6x_flags;
2596 	ipa6_conn_t	ac6x_conn;
2597 } ipa6_conn_x_t;
2598 
2599 /* flag values for ipa_conn_x_t and ipa6_conn_x_t. */
2600 #define	ACX_VERIFY_DST	0x1ULL	/* verify destination address is reachable */
2601 
2602 /* Name/Value Descriptor. */
2603 typedef struct nv_s {
2604 	uint64_t nv_value;
2605 	char	*nv_name;
2606 } nv_t;
2607 
2608 #define	ILL_FRAG_HASH(s, i) \
2609 	((ntohl(s) ^ ((i) ^ ((i) >> 8))) % ILL_FRAG_HASH_TBL_COUNT)
2610 
2611 /*
2612  * The MAX number of allowed fragmented packets per hash bucket
2613  * calculation is based on the most common mtu size of 1500. This limit
2614  * will work well for other mtu sizes as well.
2615  */
2616 #define	COMMON_IP_MTU 1500
2617 #define	MAX_FRAG_MIN 10
2618 #define	MAX_FRAG_PKTS(ipst)	\
2619 	MAX(MAX_FRAG_MIN, (2 * (ipst->ips_ip_reass_queue_bytes / \
2620 	    (COMMON_IP_MTU * ILL_FRAG_HASH_TBL_COUNT))))
2621 
2622 /*
2623  * Maximum dups allowed per packet.
2624  */
2625 extern uint_t ip_max_frag_dups;
2626 
2627 /*
2628  * Per-packet information for received packets and transmitted.
2629  * Used by the transport protocols when converting between the packet
2630  * and ancillary data and socket options.
2631  *
2632  * Note: This private data structure and related IPPF_* constant
2633  * definitions are exposed to enable compilation of some debugging tools
2634  * like lsof which use struct tcp_t in <inet/tcp.h>. This is intended to be
2635  * a temporary hack and long term alternate interfaces should be defined
2636  * to support the needs of such tools and private definitions moved to
2637  * private headers.
2638  */
2639 struct ip6_pkt_s {
2640 	uint_t		ipp_fields;		/* Which fields are valid */
2641 	uint_t		ipp_sticky_ignored;	/* sticky fields to ignore */
2642 	uint_t		ipp_ifindex;		/* pktinfo ifindex */
2643 	in6_addr_t	ipp_addr;		/* pktinfo src/dst addr */
2644 	uint_t		ipp_unicast_hops;	/* IPV6_UNICAST_HOPS */
2645 	uint_t		ipp_multicast_hops;	/* IPV6_MULTICAST_HOPS */
2646 	uint_t		ipp_hoplimit;		/* IPV6_HOPLIMIT */
2647 	uint_t		ipp_hopoptslen;
2648 	uint_t		ipp_rtdstoptslen;
2649 	uint_t		ipp_rthdrlen;
2650 	uint_t		ipp_dstoptslen;
2651 	uint_t		ipp_pathmtulen;
2652 	uint_t		ipp_fraghdrlen;
2653 	ip6_hbh_t	*ipp_hopopts;
2654 	ip6_dest_t	*ipp_rtdstopts;
2655 	ip6_rthdr_t	*ipp_rthdr;
2656 	ip6_dest_t	*ipp_dstopts;
2657 	ip6_frag_t	*ipp_fraghdr;
2658 	struct ip6_mtuinfo *ipp_pathmtu;
2659 	in6_addr_t	ipp_nexthop;		/* Transmit only */
2660 	uint8_t		ipp_tclass;
2661 	int8_t		ipp_use_min_mtu;
2662 };
2663 typedef struct ip6_pkt_s ip6_pkt_t;
2664 
2665 extern void ip6_pkt_free(ip6_pkt_t *);	/* free storage inside ip6_pkt_t */
2666 
2667 /*
2668  * This struct is used by ULP_opt_set() functions to return value of IPv4
2669  * ancillary options. Currently this is only used by udp and icmp and only
2670  * IP_PKTINFO option is supported.
2671  */
2672 typedef struct ip4_pkt_s {
2673 	uint_t		ip4_ill_index;	/* interface index */
2674 	ipaddr_t	ip4_addr;	/* source address */
2675 } ip4_pkt_t;
2676 
2677 /*
2678  * Used by ULP's to pass options info to ip_output
2679  * currently only IP_PKTINFO is supported.
2680  */
2681 typedef struct ip_opt_info_s {
2682 	uint_t ip_opt_ill_index;
2683 	uint_t ip_opt_flags;
2684 } ip_opt_info_t;
2685 
2686 /*
2687  * value for ip_opt_flags
2688  */
2689 #define	IP_VERIFY_SRC	0x1
2690 
2691 /*
2692  * This structure is used to convey information from IP and the ULP.
2693  * Currently used for the IP_RECVSLLA, IP_RECVIF and IP_RECVPKTINFO options.
2694  * The type of information field is set to IN_PKTINFO (i.e inbound pkt info)
2695  */
2696 typedef struct ip_pktinfo {
2697 	uint32_t		ip_pkt_ulp_type;	/* type of info sent */
2698 	uint32_t		ip_pkt_flags;	/* what is sent up by IP */
2699 	uint32_t		ip_pkt_ifindex;	/* inbound interface index */
2700 	struct sockaddr_dl	ip_pkt_slla;	/* has source link layer addr */
2701 	struct in_addr		ip_pkt_match_addr; /* matched address */
2702 } ip_pktinfo_t;
2703 
2704 /*
2705  * flags to tell UDP what IP is sending; in_pkt_flags
2706  */
2707 #define	IPF_RECVIF	0x01	/* inbound interface index */
2708 #define	IPF_RECVSLLA	0x02	/* source link layer address */
2709 /*
2710  * Inbound interface index + matched address.
2711  * Used only by IPV4.
2712  */
2713 #define	IPF_RECVADDR	0x04
2714 
2715 /* ipp_fields values */
2716 #define	IPPF_IFINDEX	0x0001	/* Part of in6_pktinfo: ifindex */
2717 #define	IPPF_ADDR	0x0002	/* Part of in6_pktinfo: src/dst addr */
2718 #define	IPPF_SCOPE_ID	0x0004	/* Add xmit ip6i_t for sin6_scope_id */
2719 #define	IPPF_NO_CKSUM	0x0008	/* Add xmit ip6i_t for IP6I_NO_*_CKSUM */
2720 
2721 #define	IPPF_RAW_CKSUM	0x0010	/* Add xmit ip6i_t for IP6I_RAW_CHECKSUM */
2722 #define	IPPF_HOPLIMIT	0x0020
2723 #define	IPPF_HOPOPTS	0x0040
2724 #define	IPPF_RTHDR	0x0080
2725 
2726 #define	IPPF_RTDSTOPTS	0x0100
2727 #define	IPPF_DSTOPTS	0x0200
2728 #define	IPPF_NEXTHOP	0x0400
2729 #define	IPPF_PATHMTU	0x0800
2730 
2731 #define	IPPF_TCLASS	0x1000
2732 #define	IPPF_DONTFRAG	0x2000
2733 #define	IPPF_USE_MIN_MTU	0x04000
2734 #define	IPPF_MULTICAST_HOPS	0x08000
2735 
2736 #define	IPPF_UNICAST_HOPS	0x10000
2737 #define	IPPF_FRAGHDR		0x20000
2738 
2739 #define	IPPF_HAS_IP6I \
2740 	(IPPF_IFINDEX|IPPF_ADDR|IPPF_NEXTHOP|IPPF_SCOPE_ID| \
2741 	IPPF_NO_CKSUM|IPPF_RAW_CKSUM|IPPF_HOPLIMIT|IPPF_DONTFRAG| \
2742 	IPPF_USE_MIN_MTU|IPPF_MULTICAST_HOPS|IPPF_UNICAST_HOPS)
2743 
2744 #define	TCP_PORTS_OFFSET	0
2745 #define	UDP_PORTS_OFFSET	0
2746 
2747 /*
2748  * lookups return the ill/ipif only if the flags are clear OR Iam writer.
2749  * ill / ipif lookup functions increment the refcnt on the ill / ipif only
2750  * after calling these macros. This ensures that the refcnt on the ipif or
2751  * ill will eventually drop down to zero.
2752  */
2753 #define	ILL_LOOKUP_FAILED	1	/* Used as error code */
2754 #define	IPIF_LOOKUP_FAILED	2	/* Used as error code */
2755 
2756 #define	ILL_CAN_LOOKUP(ill) 						\
2757 	(!((ill)->ill_state_flags & (ILL_CONDEMNED | ILL_CHANGING)) ||	\
2758 	IAM_WRITER_ILL(ill))
2759 
2760 #define	ILL_CAN_WAIT(ill, q)	\
2761 	(((q) != NULL) && !((ill)->ill_state_flags & (ILL_CONDEMNED)))
2762 
2763 #define	ILL_CAN_LOOKUP_WALKER(ill)	\
2764 	(!((ill)->ill_state_flags & ILL_CONDEMNED))
2765 
2766 #define	IPIF_CAN_LOOKUP(ipif)						\
2767 	(!((ipif)->ipif_state_flags & (IPIF_CONDEMNED | IPIF_CHANGING)) || \
2768 	IAM_WRITER_IPIF(ipif))
2769 
2770 /*
2771  * If the parameter 'q' is NULL, the caller is not interested in wait and
2772  * restart of the operation if the ILL or IPIF cannot be looked up when it is
2773  * marked as 'CHANGING'. Typically a thread that tries to send out data  will
2774  * end up passing NULLs as the last 4 parameters to ill_lookup_on_ifindex and
2775  * in this case 'q' is NULL
2776  */
2777 #define	IPIF_CAN_WAIT(ipif, q)	\
2778 	(((q) != NULL) && !((ipif)->ipif_state_flags & (IPIF_CONDEMNED)))
2779 
2780 #define	IPIF_CAN_LOOKUP_WALKER(ipif)					\
2781 	(!((ipif)->ipif_state_flags & (IPIF_CONDEMNED)) ||		\
2782 	IAM_WRITER_IPIF(ipif))
2783 
2784 /*
2785  * These macros are used by critical set ioctls and failover ioctls to
2786  * mark the ipif appropriately before starting the operation and to clear the
2787  * marks after completing the operation.
2788  */
2789 #define	IPIF_UNMARK_MOVING(ipif)                                \
2790 	(ipif)->ipif_state_flags &= ~IPIF_MOVING & ~IPIF_CHANGING;
2791 
2792 #define	ILL_UNMARK_CHANGING(ill)                                \
2793 	(ill)->ill_state_flags &= ~ILL_CHANGING;
2794 
2795 /* Macros used to assert that this thread is a writer  */
2796 #define	IAM_WRITER_IPSQ(ipsq)	((ipsq)->ipsq_writer == curthread)
2797 #define	IAM_WRITER_ILL(ill)					\
2798 	((ill)->ill_phyint->phyint_ipsq->ipsq_writer == curthread)
2799 #define	IAM_WRITER_IPIF(ipif)					\
2800 	((ipif)->ipif_ill->ill_phyint->phyint_ipsq->ipsq_writer == curthread)
2801 
2802 /*
2803  * Grab ill locks in the proper order. The order is highest addressed
2804  * ill is locked first.
2805  */
2806 #define	GRAB_ILL_LOCKS(ill_1, ill_2)				\
2807 {								\
2808 	if ((ill_1) > (ill_2)) {				\
2809 		if (ill_1 != NULL)				\
2810 			mutex_enter(&(ill_1)->ill_lock);	\
2811 		if (ill_2 != NULL)				\
2812 			mutex_enter(&(ill_2)->ill_lock);	\
2813 	} else {						\
2814 		if (ill_2 != NULL)				\
2815 			mutex_enter(&(ill_2)->ill_lock);	\
2816 		if (ill_1 != NULL && ill_1 != ill_2)		\
2817 			mutex_enter(&(ill_1)->ill_lock);	\
2818 	}							\
2819 }
2820 
2821 #define	RELEASE_ILL_LOCKS(ill_1, ill_2)		\
2822 {						\
2823 	if (ill_1 != NULL)			\
2824 		mutex_exit(&(ill_1)->ill_lock);	\
2825 	if (ill_2 != NULL && ill_2 != ill_1)	\
2826 		mutex_exit(&(ill_2)->ill_lock);	\
2827 }
2828 
2829 /* Get the other protocol instance ill */
2830 #define	ILL_OTHER(ill)						\
2831 	((ill)->ill_isv6 ? (ill)->ill_phyint->phyint_illv4 :	\
2832 	    (ill)->ill_phyint->phyint_illv6)
2833 
2834 #define	MATCH_V4_ONLY	0x1
2835 #define	MATCH_V6_ONLY	0x2
2836 #define	MATCH_ILL_ONLY	0x4
2837 
2838 /* ioctl command info: Ioctl properties extracted and stored in here */
2839 typedef struct cmd_info_s
2840 {
2841 	char    ci_groupname[LIFNAMSIZ + 1];	/* SIOCSLIFGROUPNAME */
2842 	ipif_t  *ci_ipif;	/* ipif associated with [l]ifreq ioctl's */
2843 	sin_t	*ci_sin;	/* the sin struct passed down */
2844 	sin6_t	*ci_sin6;	/* the sin6_t struct passed down */
2845 	struct lifreq *ci_lifr;	/* the lifreq struct passed down */
2846 } cmd_info_t;
2847 
2848 /*
2849  * List of AH and ESP IPsec acceleration capable ills
2850  */
2851 typedef struct ipsec_capab_ill_s {
2852 	uint_t ill_index;
2853 	boolean_t ill_isv6;
2854 	struct ipsec_capab_ill_s *next;
2855 } ipsec_capab_ill_t;
2856 
2857 extern struct kmem_cache *ire_cache;
2858 
2859 extern ipaddr_t	ip_g_all_ones;
2860 
2861 extern	uint_t	ip_loopback_mtu;	/* /etc/system */
2862 
2863 extern vmem_t *ip_minor_arena_sa;
2864 extern vmem_t *ip_minor_arena_la;
2865 
2866 /*
2867  * ip_g_forward controls IP forwarding.  It takes two values:
2868  *	0: IP_FORWARD_NEVER	Don't forward packets ever.
2869  *	1: IP_FORWARD_ALWAYS	Forward packets for elsewhere.
2870  *
2871  * RFC1122 says there must be a configuration switch to control forwarding,
2872  * but that the default MUST be to not forward packets ever.  Implicit
2873  * control based on configuration of multiple interfaces MUST NOT be
2874  * implemented (Section 3.1).  SunOS 4.1 did provide the "automatic" capability
2875  * and, in fact, it was the default.  That capability is now provided in the
2876  * /etc/rc2.d/S69inet script.
2877  */
2878 
2879 #define	ips_ip_respond_to_address_mask_broadcast ips_param_arr[0].ip_param_value
2880 #define	ips_ip_g_resp_to_echo_bcast	ips_param_arr[1].ip_param_value
2881 #define	ips_ip_g_resp_to_echo_mcast	ips_param_arr[2].ip_param_value
2882 #define	ips_ip_g_resp_to_timestamp	ips_param_arr[3].ip_param_value
2883 #define	ips_ip_g_resp_to_timestamp_bcast ips_param_arr[4].ip_param_value
2884 #define	ips_ip_g_send_redirects		ips_param_arr[5].ip_param_value
2885 #define	ips_ip_g_forward_directed_bcast	ips_param_arr[6].ip_param_value
2886 #define	ips_ip_mrtdebug			ips_param_arr[7].ip_param_value
2887 #define	ips_ip_timer_interval		ips_param_arr[8].ip_param_value
2888 #define	ips_ip_ire_arp_interval		ips_param_arr[9].ip_param_value
2889 #define	ips_ip_ire_redir_interval	ips_param_arr[10].ip_param_value
2890 #define	ips_ip_def_ttl			ips_param_arr[11].ip_param_value
2891 #define	ips_ip_forward_src_routed	ips_param_arr[12].ip_param_value
2892 #define	ips_ip_wroff_extra		ips_param_arr[13].ip_param_value
2893 #define	ips_ip_ire_pathmtu_interval	ips_param_arr[14].ip_param_value
2894 #define	ips_ip_icmp_return		ips_param_arr[15].ip_param_value
2895 #define	ips_ip_path_mtu_discovery	ips_param_arr[16].ip_param_value
2896 #define	ips_ip_ignore_delete_time	ips_param_arr[17].ip_param_value
2897 #define	ips_ip_ignore_redirect		ips_param_arr[18].ip_param_value
2898 #define	ips_ip_output_queue		ips_param_arr[19].ip_param_value
2899 #define	ips_ip_broadcast_ttl		ips_param_arr[20].ip_param_value
2900 #define	ips_ip_icmp_err_interval	ips_param_arr[21].ip_param_value
2901 #define	ips_ip_icmp_err_burst		ips_param_arr[22].ip_param_value
2902 #define	ips_ip_reass_queue_bytes	ips_param_arr[23].ip_param_value
2903 #define	ips_ip_strict_dst_multihoming	ips_param_arr[24].ip_param_value
2904 #define	ips_ip_addrs_per_if		ips_param_arr[25].ip_param_value
2905 #define	ips_ipsec_override_persocket_policy ips_param_arr[26].ip_param_value
2906 #define	ips_icmp_accept_clear_messages	ips_param_arr[27].ip_param_value
2907 #define	ips_igmp_accept_clear_messages	ips_param_arr[28].ip_param_value
2908 
2909 /* IPv6 configuration knobs */
2910 #define	ips_delay_first_probe_time	ips_param_arr[29].ip_param_value
2911 #define	ips_max_unicast_solicit		ips_param_arr[30].ip_param_value
2912 #define	ips_ipv6_def_hops		ips_param_arr[31].ip_param_value
2913 #define	ips_ipv6_icmp_return		ips_param_arr[32].ip_param_value
2914 #define	ips_ipv6_forward_src_routed	ips_param_arr[33].ip_param_value
2915 #define	ips_ipv6_resp_echo_mcast	ips_param_arr[34].ip_param_value
2916 #define	ips_ipv6_send_redirects		ips_param_arr[35].ip_param_value
2917 #define	ips_ipv6_ignore_redirect	ips_param_arr[36].ip_param_value
2918 #define	ips_ipv6_strict_dst_multihoming	ips_param_arr[37].ip_param_value
2919 #define	ips_ip_ire_reclaim_fraction	ips_param_arr[38].ip_param_value
2920 #define	ips_ipsec_policy_log_interval	ips_param_arr[39].ip_param_value
2921 #define	ips_pim_accept_clear_messages	ips_param_arr[40].ip_param_value
2922 #define	ips_ip_ndp_unsolicit_interval	ips_param_arr[41].ip_param_value
2923 #define	ips_ip_ndp_unsolicit_count	ips_param_arr[42].ip_param_value
2924 #define	ips_ipv6_ignore_home_address_opt ips_param_arr[43].ip_param_value
2925 #define	ips_ip_policy_mask		ips_param_arr[44].ip_param_value
2926 #define	ips_ip_multirt_resolution_interval ips_param_arr[45].ip_param_value
2927 #define	ips_ip_multirt_ttl  		ips_param_arr[46].ip_param_value
2928 #define	ips_ip_multidata_outbound	ips_param_arr[47].ip_param_value
2929 #define	ips_ip_ndp_defense_interval	ips_param_arr[48].ip_param_value
2930 #define	ips_ip_max_temp_idle		ips_param_arr[49].ip_param_value
2931 #define	ips_ip_max_temp_defend		ips_param_arr[50].ip_param_value
2932 #define	ips_ip_max_defend		ips_param_arr[51].ip_param_value
2933 #define	ips_ip_defend_interval		ips_param_arr[52].ip_param_value
2934 #define	ips_ip_dup_recovery		ips_param_arr[53].ip_param_value
2935 #define	ips_ip_restrict_interzone_loopback ips_param_arr[54].ip_param_value
2936 #define	ips_ip_lso_outbound		ips_param_arr[55].ip_param_value
2937 #define	ips_igmp_max_version		ips_param_arr[56].ip_param_value
2938 #define	ips_mld_max_version		ips_param_arr[57].ip_param_value
2939 #define	ips_ipv6_drop_inbound_icmpv6	ips_param_arr[58].ip_param_value
2940 
2941 extern int	dohwcksum;	/* use h/w cksum if supported by the h/w */
2942 #ifdef ZC_TEST
2943 extern int	noswcksum;
2944 #endif
2945 
2946 extern char	ipif_loopback_name[];
2947 
2948 extern nv_t	*ire_nv_tbl;
2949 
2950 extern struct module_info ip_mod_info;
2951 
2952 #define	HOOKS4_INTERESTED_PHYSICAL_IN(ipst)	\
2953 	((ipst)->ips_ip4_physical_in_event.he_interested)
2954 #define	HOOKS6_INTERESTED_PHYSICAL_IN(ipst)	\
2955 	((ipst)->ips_ip6_physical_in_event.he_interested)
2956 #define	HOOKS4_INTERESTED_PHYSICAL_OUT(ipst)	\
2957 	((ipst)->ips_ip4_physical_out_event.he_interested)
2958 #define	HOOKS6_INTERESTED_PHYSICAL_OUT(ipst)	\
2959 	((ipst)->ips_ip6_physical_out_event.he_interested)
2960 #define	HOOKS4_INTERESTED_FORWARDING(ipst)	\
2961 	((ipst)->ips_ip4_forwarding_event.he_interested)
2962 #define	HOOKS6_INTERESTED_FORWARDING(ipst)	\
2963 	((ipst)->ips_ip6_forwarding_event.he_interested)
2964 #define	HOOKS4_INTERESTED_LOOPBACK_IN(ipst)	\
2965 	((ipst)->ips_ip4_loopback_in_event.he_interested)
2966 #define	HOOKS6_INTERESTED_LOOPBACK_IN(ipst)	\
2967 	((ipst)->ips_ip6_loopback_in_event.he_interested)
2968 #define	HOOKS4_INTERESTED_LOOPBACK_OUT(ipst)	\
2969 	((ipst)->ips_ip4_loopback_out_event.he_interested)
2970 #define	HOOKS6_INTERESTED_LOOPBACK_OUT(ipst)	\
2971 	((ipst)->ips_ip6_loopback_out_event.he_interested)
2972 
2973 /*
2974  * Hooks marcos used inside of ip
2975  */
2976 #define	IPHA_VHL	ipha_version_and_hdr_length
2977 
2978 #define	FW_HOOKS(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst)	\
2979 									\
2980 	if ((_hook).he_interested) {	\
2981 		hook_pkt_event_t info;					\
2982 									\
2983 		_NOTE(CONSTCOND)					\
2984 		ASSERT((_ilp != NULL) || (_olp != NULL));		\
2985 									\
2986 		_NOTE(CONSTCOND)					\
2987 		if ((_ilp != NULL) &&					\
2988 		    (((ill_t *)(_ilp))->ill_phyint != NULL))		\
2989 			info.hpe_ifp = (phy_if_t)((ill_t *)		\
2990 			    (_ilp))->ill_phyint->phyint_hook_ifindex;	\
2991 		else							\
2992 			info.hpe_ifp = 0;				\
2993 									\
2994 		_NOTE(CONSTCOND)					\
2995 		if ((_olp != NULL) &&					\
2996 		    (((ill_t *)(_olp))->ill_phyint != NULL))		\
2997 			info.hpe_ofp = (phy_if_t)((ill_t *)		\
2998 			    (_olp))->ill_phyint->phyint_hook_ifindex;	\
2999 		else							\
3000 			info.hpe_ofp = 0;				\
3001 		info.hpe_hdr = _iph;					\
3002 		info.hpe_mp = &(_fm);					\
3003 		info.hpe_mb = _m;					\
3004 		info.hpe_flags = _llm;					\
3005 		if (hook_run(_event, (hook_data_t)&info,		\
3006 		    ipst->ips_netstack) != 0) {				\
3007 			ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
3008 			    (_hook).he_name, (void *)_fm, (void *)_m));	\
3009 			if (_fm != NULL) {				\
3010 				freemsg(_fm);				\
3011 				_fm = NULL;				\
3012 			}						\
3013 			_iph = NULL;					\
3014 			_m = NULL;					\
3015 		} else {						\
3016 			_iph = info.hpe_hdr;				\
3017 			_m = info.hpe_mb;				\
3018 		}							\
3019 	}
3020 
3021 #define	FW_HOOKS6(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst)	\
3022 									\
3023 	if ((_hook).he_interested) {	\
3024 		hook_pkt_event_t info;					\
3025 									\
3026 		_NOTE(CONSTCOND)					\
3027 		ASSERT((_ilp != NULL) || (_olp != NULL));		\
3028 									\
3029 		_NOTE(CONSTCOND)					\
3030 		if ((_ilp != NULL) &&					\
3031 		    (((ill_t *)(_ilp))->ill_phyint != NULL))		\
3032 			info.hpe_ifp = (phy_if_t)((ill_t *)		\
3033 			    (_ilp))->ill_phyint->phyint_hook_ifindex;	\
3034 		else							\
3035 			info.hpe_ifp = 0;				\
3036 									\
3037 		_NOTE(CONSTCOND)					\
3038 		if ((_olp != NULL) &&					\
3039 		    (((ill_t *)(_olp))->ill_phyint != NULL))		\
3040 			info.hpe_ofp = (phy_if_t)((ill_t *)		\
3041 			    (_olp))->ill_phyint->phyint_hook_ifindex;	\
3042 		else							\
3043 			info.hpe_ofp = 0;				\
3044 		info.hpe_hdr = _iph;					\
3045 		info.hpe_mp = &(_fm);					\
3046 		info.hpe_mb = _m;					\
3047 		info.hpe_flags = _llm;					\
3048 		if (hook_run(_event, (hook_data_t)&info,		\
3049 		    ipst->ips_netstack) != 0) {				\
3050 			ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
3051 			    (_hook).he_name, (void *)_fm, (void *)_m));	\
3052 			if (_fm != NULL) {				\
3053 				freemsg(_fm);				\
3054 				_fm = NULL;				\
3055 			}						\
3056 			_iph = NULL;					\
3057 			_m = NULL;					\
3058 		} else {						\
3059 			_iph = info.hpe_hdr;				\
3060 			_m = info.hpe_mb;				\
3061 		}							\
3062 	}
3063 
3064 /*
3065  * Network byte order macros
3066  */
3067 #ifdef	_BIG_ENDIAN
3068 #define	N_IN_CLASSA_NET		IN_CLASSA_NET
3069 #define	N_IN_CLASSD_NET		IN_CLASSD_NET
3070 #define	N_INADDR_UNSPEC_GROUP	INADDR_UNSPEC_GROUP
3071 #define	N_IN_LOOPBACK_NET	(ipaddr_t)0x7f000000U
3072 #else /* _BIG_ENDIAN */
3073 #define	N_IN_CLASSA_NET		(ipaddr_t)0x000000ffU
3074 #define	N_IN_CLASSD_NET		(ipaddr_t)0x000000f0U
3075 #define	N_INADDR_UNSPEC_GROUP	(ipaddr_t)0x000000e0U
3076 #define	N_IN_LOOPBACK_NET	(ipaddr_t)0x0000007fU
3077 #endif /* _BIG_ENDIAN */
3078 #define	CLASSD(addr)	(((addr) & N_IN_CLASSD_NET) == N_INADDR_UNSPEC_GROUP)
3079 #define	CLASSE(addr)	(((addr) & N_IN_CLASSD_NET) == N_IN_CLASSD_NET)
3080 #define	IP_LOOPBACK_ADDR(addr)			\
3081 	(((addr) & N_IN_CLASSA_NET == N_IN_LOOPBACK_NET))
3082 
3083 #ifdef DEBUG
3084 /* IPsec HW acceleration debugging support */
3085 
3086 #define	IPSECHW_CAPAB		0x0001	/* capability negotiation */
3087 #define	IPSECHW_SADB		0x0002	/* SADB exchange */
3088 #define	IPSECHW_PKT		0x0004	/* general packet flow */
3089 #define	IPSECHW_PKTIN		0x0008	/* driver in pkt processing details */
3090 #define	IPSECHW_PKTOUT		0x0010	/* driver out pkt processing details */
3091 
3092 #define	IPSECHW_DEBUG(f, x)	if (ipsechw_debug & (f)) { (void) printf x; }
3093 #define	IPSECHW_CALL(f, r, x)	if (ipsechw_debug & (f)) { (void) r x; }
3094 
3095 extern uint32_t ipsechw_debug;
3096 #else
3097 #define	IPSECHW_DEBUG(f, x)	{}
3098 #define	IPSECHW_CALL(f, r, x)	{}
3099 #endif
3100 
3101 extern int	ip_debug;
3102 extern uint_t	ip_thread_data;
3103 extern krwlock_t ip_thread_rwlock;
3104 extern list_t	ip_thread_list;
3105 
3106 #ifdef IP_DEBUG
3107 #include <sys/debug.h>
3108 #include <sys/promif.h>
3109 
3110 #define	ip0dbg(a)	printf a
3111 #define	ip1dbg(a)	if (ip_debug > 2) printf a
3112 #define	ip2dbg(a)	if (ip_debug > 3) printf a
3113 #define	ip3dbg(a)	if (ip_debug > 4) printf a
3114 #else
3115 #define	ip0dbg(a)	/* */
3116 #define	ip1dbg(a)	/* */
3117 #define	ip2dbg(a)	/* */
3118 #define	ip3dbg(a)	/* */
3119 #endif	/* IP_DEBUG */
3120 
3121 /* Default MAC-layer address string length for mac_colon_addr */
3122 #define	MAC_STR_LEN	128
3123 
3124 struct	ipsec_out_s;
3125 
3126 struct	mac_header_info_s;
3127 
3128 extern boolean_t ip_assign_ifindex(uint_t *, ip_stack_t *);
3129 extern void	ill_frag_timer(void *);
3130 extern ill_t	*ill_first(int, int, ill_walk_context_t *, ip_stack_t *);
3131 extern ill_t	*ill_next(ill_walk_context_t *, ill_t *);
3132 extern void	ill_frag_timer_start(ill_t *);
3133 extern void	ill_nic_info_dispatch(ill_t *);
3134 extern void	ill_nic_info_plumb(ill_t *, boolean_t);
3135 extern mblk_t	*ip_carve_mp(mblk_t **, ssize_t);
3136 extern mblk_t	*ip_dlpi_alloc(size_t, t_uscalar_t);
3137 extern char	*ip_dot_addr(ipaddr_t, char *);
3138 extern const char *mac_colon_addr(const uint8_t *, size_t, char *, size_t);
3139 extern void	ip_lwput(queue_t *, mblk_t *);
3140 extern boolean_t icmp_err_rate_limit(ip_stack_t *);
3141 extern void	icmp_time_exceeded(queue_t *, mblk_t *, uint8_t, zoneid_t,
3142     ip_stack_t *);
3143 extern void	icmp_unreachable(queue_t *, mblk_t *, uint8_t, zoneid_t,
3144     ip_stack_t *);
3145 extern mblk_t	*ip_add_info(mblk_t *, ill_t *, uint_t, zoneid_t, ip_stack_t *);
3146 extern mblk_t	*ip_bind_v4(queue_t *, mblk_t *, conn_t *);
3147 extern int	ip_bind_connected(conn_t *, mblk_t *, ipaddr_t *, uint16_t,
3148     ipaddr_t, uint16_t, boolean_t, boolean_t, boolean_t,
3149     boolean_t);
3150 extern boolean_t ip_bind_ipsec_policy_set(conn_t *, mblk_t *);
3151 extern int	ip_bind_laddr(conn_t *, mblk_t *, ipaddr_t, uint16_t,
3152     boolean_t, boolean_t, boolean_t);
3153 extern uint_t	ip_cksum(mblk_t *, int, uint32_t);
3154 extern int	ip_close(queue_t *, int);
3155 extern uint16_t	ip_csum_hdr(ipha_t *);
3156 extern void	ip_proto_not_sup(queue_t *, mblk_t *, uint_t, zoneid_t,
3157     ip_stack_t *);
3158 extern void	ip_ire_g_fini(void);
3159 extern void	ip_ire_g_init(void);
3160 extern void	ip_ire_fini(ip_stack_t *);
3161 extern void	ip_ire_init(ip_stack_t *);
3162 extern int	ip_openv4(queue_t *q, dev_t *devp, int flag, int sflag,
3163 		    cred_t *credp);
3164 extern int	ip_openv6(queue_t *q, dev_t *devp, int flag, int sflag,
3165 		    cred_t *credp);
3166 extern int	ip_reassemble(mblk_t *, ipf_t *, uint_t, boolean_t, ill_t *,
3167     size_t);
3168 extern int	ip_opt_set_ill(conn_t *, int, boolean_t, boolean_t,
3169     int, int, mblk_t *);
3170 extern void	ip_rput(queue_t *, mblk_t *);
3171 extern void	ip_input(ill_t *, ill_rx_ring_t *, mblk_t *,
3172     struct mac_header_info_s *);
3173 extern void	ip_rput_dlpi(queue_t *, mblk_t *);
3174 extern void	ip_rput_forward(ire_t *, ipha_t *, mblk_t *, ill_t *);
3175 extern void	ip_rput_forward_multicast(ipaddr_t, mblk_t *, ipif_t *);
3176 
3177 extern void	ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *,
3178 		    mib2_ipIfStatsEntry_t *);
3179 extern void	ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *,
3180 		    mib2_ipv6IfIcmpEntry_t *);
3181 extern void	ip_udp_input(queue_t *, mblk_t *, ipha_t *, ire_t *, ill_t *);
3182 extern void	ip_proto_input(queue_t *, mblk_t *, ipha_t *, ire_t *, ill_t *,
3183     boolean_t);
3184 extern void	ip_rput_other(ipsq_t *, queue_t *, mblk_t *, void *);
3185 extern ire_t	*ip_check_multihome(void *, ire_t *, ill_t *);
3186 extern void	ip_setpktversion(conn_t *, boolean_t, boolean_t, ip_stack_t *);
3187 extern void	ip_trash_ire_reclaim(void *);
3188 extern void	ip_trash_timer_expire(void *);
3189 extern void	ip_wput(queue_t *, mblk_t *);
3190 extern void	ip_output(void *, mblk_t *, void *, int);
3191 extern void	ip_output_options(void *, mblk_t *, void *, int,
3192     ip_opt_info_t *);
3193 
3194 extern void	ip_wput_ire(queue_t *, mblk_t *, ire_t *, conn_t *, int,
3195 		    zoneid_t);
3196 extern void	ip_wput_local(queue_t *, ill_t *, ipha_t *, mblk_t *, ire_t *,
3197 		    int, zoneid_t);
3198 extern void	ip_wput_multicast(queue_t *, mblk_t *, ipif_t *, zoneid_t);
3199 extern void	ip_wput_nondata(ipsq_t *, queue_t *, mblk_t *, void *);
3200 extern void	ip_wsrv(queue_t *);
3201 extern char	*ip_nv_lookup(nv_t *, int);
3202 extern boolean_t ip_local_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3203 extern boolean_t ip_remote_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3204 extern ipaddr_t ip_massage_options(ipha_t *, netstack_t *);
3205 extern ipaddr_t ip_net_mask(ipaddr_t);
3206 extern void	ip_newroute(queue_t *, mblk_t *, ipaddr_t, conn_t *, zoneid_t,
3207 		    ip_stack_t *);
3208 extern ipxmit_state_t	ip_xmit_v4(mblk_t *, ire_t *, struct ipsec_out_s *,
3209     boolean_t);
3210 extern int	ip_hdr_complete(ipha_t *, zoneid_t, ip_stack_t *);
3211 
3212 extern struct qinit iprinitv6;
3213 extern struct qinit ipwinitv6;
3214 
3215 extern void	conn_drain_insert(conn_t *connp);
3216 extern	int	conn_ipsec_length(conn_t *connp);
3217 extern void	ip_wput_ipsec_out(queue_t *, mblk_t *, ipha_t *, ill_t *,
3218     ire_t *);
3219 extern ipaddr_t	ip_get_dst(ipha_t *);
3220 extern int	ipsec_out_extra_length(mblk_t *);
3221 extern int	ipsec_in_extra_length(mblk_t *);
3222 extern mblk_t	*ipsec_in_alloc(boolean_t, netstack_t *);
3223 extern boolean_t ipsec_in_is_secure(mblk_t *);
3224 extern void	ipsec_out_process(queue_t *, mblk_t *, ire_t *, uint_t);
3225 extern void	ipsec_out_to_in(mblk_t *);
3226 extern void	ip_fanout_proto_again(mblk_t *, ill_t *, ill_t *, ire_t *);
3227 
3228 extern void	ire_cleanup(ire_t *);
3229 extern void	ire_inactive(ire_t *);
3230 extern boolean_t irb_inactive(irb_t *);
3231 extern ire_t	*ire_unlink(irb_t *);
3232 
3233 #ifdef DEBUG
3234 extern	boolean_t th_trace_ref(const void *, ip_stack_t *);
3235 extern	void	th_trace_unref(const void *);
3236 extern	void	th_trace_cleanup(const void *, boolean_t);
3237 extern	void	ire_trace_ref(ire_t *);
3238 extern	void	ire_untrace_ref(ire_t *);
3239 #endif
3240 
3241 extern int	ip_srcid_insert(const in6_addr_t *, zoneid_t, ip_stack_t *);
3242 extern int	ip_srcid_remove(const in6_addr_t *, zoneid_t, ip_stack_t *);
3243 extern void	ip_srcid_find_id(uint_t, in6_addr_t *, zoneid_t, netstack_t *);
3244 extern uint_t	ip_srcid_find_addr(const in6_addr_t *, zoneid_t, netstack_t *);
3245 extern int	ip_srcid_report(queue_t *, mblk_t *, caddr_t, cred_t *);
3246 
3247 extern uint8_t	ipoptp_next(ipoptp_t *);
3248 extern uint8_t	ipoptp_first(ipoptp_t *, ipha_t *);
3249 extern int	ip_opt_get_user(const ipha_t *, uchar_t *);
3250 extern ill_t	*ip_grab_attach_ill(ill_t *, mblk_t *, int, boolean_t,
3251     ip_stack_t *);
3252 extern ire_t	*conn_set_outgoing_ill(conn_t *, ire_t *, ill_t **);
3253 extern int	ipsec_req_from_conn(conn_t *, ipsec_req_t *, int);
3254 extern int	ip_snmp_get(queue_t *q, mblk_t *mctl, int level);
3255 extern int	ip_snmp_set(queue_t *q, int, int, uchar_t *, int);
3256 extern void	ip_process_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3257 extern void	ip_quiesce_conn(conn_t *);
3258 extern  void    ip_reprocess_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3259 extern void	ip_restart_optmgmt(ipsq_t *, queue_t *, mblk_t *, void *);
3260 extern void	ip_ioctl_finish(queue_t *, mblk_t *, int, int, ipsq_t *);
3261 
3262 extern boolean_t ip_cmpbuf(const void *, uint_t, boolean_t, const void *,
3263     uint_t);
3264 extern boolean_t ip_allocbuf(void **, uint_t *, boolean_t, const void *,
3265     uint_t);
3266 extern void	ip_savebuf(void **, uint_t *, boolean_t, const void *, uint_t);
3267 
3268 extern boolean_t	ipsq_pending_mp_cleanup(ill_t *, conn_t *);
3269 extern void	conn_ioctl_cleanup(conn_t *);
3270 extern ill_t	*conn_get_held_ill(conn_t *, ill_t **, int *);
3271 extern ill_t	*ip_newroute_get_dst_ill(ill_t *);
3272 
3273 struct multidata_s;
3274 struct pdesc_s;
3275 
3276 extern mblk_t	*ip_mdinfo_alloc(ill_mdt_capab_t *);
3277 extern mblk_t	*ip_mdinfo_return(ire_t *, conn_t *, char *, ill_mdt_capab_t *);
3278 extern mblk_t	*ip_lsoinfo_alloc(ill_lso_capab_t *);
3279 extern mblk_t	*ip_lsoinfo_return(ire_t *, conn_t *, char *,
3280     ill_lso_capab_t *);
3281 extern uint_t	ip_md_cksum(struct pdesc_s *, int, uint_t);
3282 extern boolean_t ip_md_addr_attr(struct multidata_s *, struct pdesc_s *,
3283 			const mblk_t *);
3284 extern boolean_t ip_md_hcksum_attr(struct multidata_s *, struct pdesc_s *,
3285 			uint32_t, uint32_t, uint32_t, uint32_t);
3286 extern boolean_t ip_md_zcopy_attr(struct multidata_s *, struct pdesc_s *,
3287 			uint_t);
3288 extern mblk_t	*ip_unbind(queue_t *, mblk_t *);
3289 
3290 extern phyint_t *phyint_lookup_group(char *, boolean_t, ip_stack_t *);
3291 extern phyint_t *phyint_lookup_group_ifindex(uint_t, ip_stack_t *);
3292 
3293 extern void tnet_init(void);
3294 extern void tnet_fini(void);
3295 
3296 /* Hooks for CGTP (multirt routes) filtering module */
3297 #define	CGTP_FILTER_REV_1	1
3298 #define	CGTP_FILTER_REV_2	2
3299 #define	CGTP_FILTER_REV_3	3
3300 #define	CGTP_FILTER_REV		CGTP_FILTER_REV_3
3301 
3302 /* cfo_filter and cfo_filter_v6 hooks return values */
3303 #define	CGTP_IP_PKT_NOT_CGTP	0
3304 #define	CGTP_IP_PKT_PREMIUM	1
3305 #define	CGTP_IP_PKT_DUPLICATE	2
3306 
3307 /* Version 3 of the filter interface */
3308 typedef struct cgtp_filter_ops {
3309 	int	cfo_filter_rev;			/* CGTP_FILTER_REV_3 */
3310 	int	(*cfo_change_state)(netstackid_t, int);
3311 	int	(*cfo_add_dest_v4)(netstackid_t, ipaddr_t, ipaddr_t,
3312 		    ipaddr_t, ipaddr_t);
3313 	int	(*cfo_del_dest_v4)(netstackid_t, ipaddr_t, ipaddr_t);
3314 	int	(*cfo_add_dest_v6)(netstackid_t, in6_addr_t *, in6_addr_t *,
3315 		    in6_addr_t *, in6_addr_t *);
3316 	int	(*cfo_del_dest_v6)(netstackid_t, in6_addr_t *, in6_addr_t *);
3317 	int	(*cfo_filter)(netstackid_t, uint_t, mblk_t *);
3318 	int	(*cfo_filter_v6)(netstackid_t, uint_t, ip6_t *,
3319 		    ip6_frag_t *);
3320 } cgtp_filter_ops_t;
3321 
3322 #define	CGTP_MCAST_SUCCESS	1
3323 
3324 /*
3325  * The separate CGTP module needs this global symbol so that it
3326  * can check the version and determine whether to use the old or the new
3327  * version of the filtering interface.
3328  */
3329 extern int	ip_cgtp_filter_rev;
3330 
3331 extern int	ip_cgtp_filter_supported(void);
3332 extern int	ip_cgtp_filter_register(netstackid_t, cgtp_filter_ops_t *);
3333 extern int	ip_cgtp_filter_unregister(netstackid_t);
3334 extern int	ip_cgtp_filter_is_registered(netstackid_t);
3335 
3336 /* Flags for ire_multirt_lookup() */
3337 
3338 #define	MULTIRT_USESTAMP	0x0001
3339 #define	MULTIRT_SETSTAMP	0x0002
3340 #define	MULTIRT_CACHEGW		0x0004
3341 
3342 /* Debug stuff for multirt route resolution. */
3343 #if defined(DEBUG) && !defined(__lint)
3344 /* Our "don't send, rather drop" flag. */
3345 #define	MULTIRT_DEBUG_FLAG	0x8000
3346 
3347 #define	MULTIRT_TRACE(x)	ip2dbg(x)
3348 
3349 #define	MULTIRT_DEBUG_TAG(mblk)	\
3350 	do { \
3351 		ASSERT(mblk != NULL); \
3352 		MULTIRT_TRACE(("%s[%d]: tagging mblk %p, tag was %d\n", \
3353 		__FILE__, __LINE__, \
3354 		(void *)(mblk), (mblk)->b_flag & MULTIRT_DEBUG_FLAG)); \
3355 		(mblk)->b_flag |= MULTIRT_DEBUG_FLAG; \
3356 	} while (0)
3357 
3358 #define	MULTIRT_DEBUG_UNTAG(mblk) \
3359 	do { \
3360 		ASSERT(mblk != NULL); \
3361 		MULTIRT_TRACE(("%s[%d]: untagging mblk %p, tag was %d\n", \
3362 		__FILE__, __LINE__, \
3363 		(void *)(mblk), (mblk)->b_flag & MULTIRT_DEBUG_FLAG)); \
3364 		(mblk)->b_flag &= ~MULTIRT_DEBUG_FLAG; \
3365 	} while (0)
3366 
3367 #define	MULTIRT_DEBUG_TAGGED(mblk) \
3368 	(((mblk)->b_flag & MULTIRT_DEBUG_FLAG) ? B_TRUE : B_FALSE)
3369 #else
3370 #define	MULTIRT_DEBUG_TAG(mblk)		ASSERT(mblk != NULL)
3371 #define	MULTIRT_DEBUG_UNTAG(mblk)	ASSERT(mblk != NULL)
3372 #define	MULTIRT_DEBUG_TAGGED(mblk)	B_FALSE
3373 #endif
3374 
3375 /*
3376  * Per-ILL Multidata Transmit capabilities.
3377  */
3378 struct ill_mdt_capab_s {
3379 	uint_t ill_mdt_version;  /* interface version */
3380 	uint_t ill_mdt_on;	 /* on/off switch for MDT on this ILL */
3381 	uint_t ill_mdt_hdr_head; /* leading header fragment extra space */
3382 	uint_t ill_mdt_hdr_tail; /* trailing header fragment extra space */
3383 	uint_t ill_mdt_max_pld;	 /* maximum payload buffers per Multidata */
3384 	uint_t ill_mdt_span_limit; /* maximum payload span per packet */
3385 };
3386 
3387 struct ill_hcksum_capab_s {
3388 	uint_t	ill_hcksum_version;	/* interface version */
3389 	uint_t	ill_hcksum_txflags;	/* capabilities on transmit */
3390 };
3391 
3392 struct ill_zerocopy_capab_s {
3393 	uint_t	ill_zerocopy_version;	/* interface version */
3394 	uint_t	ill_zerocopy_flags;	/* capabilities */
3395 };
3396 
3397 struct ill_lso_capab_s {
3398 	uint_t	ill_lso_version;	/* interface version */
3399 	uint_t	ill_lso_on;		/* on/off switch for LSO on this ILL */
3400 	uint_t	ill_lso_flags;		/* capabilities */
3401 	uint_t	ill_lso_max;		/* maximum size of payload */
3402 };
3403 
3404 /* Possible ill_states */
3405 #define	ILL_RING_INPROC		3	/* Being assigned to squeue */
3406 #define	ILL_RING_INUSE		2	/* Already Assigned to Rx Ring */
3407 #define	ILL_RING_BEING_FREED	1	/* Being Unassigned */
3408 #define	ILL_RING_FREE		0	/* Available to be assigned to Ring */
3409 
3410 #define	ILL_MAX_RINGS		256	/* Max num of rx rings we can manage */
3411 #define	ILL_POLLING		0x01	/* Polling in use */
3412 
3413 /*
3414  * These functions pointer types are exported by the mac/dls layer.
3415  * we need to duplicate the definitions here because we cannot
3416  * include mac/dls header files here.
3417  */
3418 typedef void	(*ip_mac_blank_t)(void *, time_t, uint_t);
3419 typedef void	(*ip_dld_tx_t)(void *, mblk_t *);
3420 
3421 typedef void	(*ip_dls_chg_soft_ring_t)(void *, int);
3422 typedef void	(*ip_dls_bind_t)(void *, processorid_t);
3423 typedef void	(*ip_dls_unbind_t)(void *);
3424 
3425 struct ill_rx_ring {
3426 	ip_mac_blank_t		rr_blank; /* Driver interrupt blanking func */
3427 	void			*rr_handle; /* Handle for Rx ring */
3428 	squeue_t		*rr_sqp; /* Squeue the ring is bound to */
3429 	ill_t			*rr_ill; /* back pointer to ill */
3430 	clock_t			rr_poll_time; /* Last lbolt polling was used */
3431 	uint32_t		rr_poll_state; /* polling state flags */
3432 	uint32_t		rr_max_blank_time; /* Max interrupt blank */
3433 	uint32_t		rr_min_blank_time; /* Min interrupt blank */
3434 	uint32_t		rr_max_pkt_cnt; /* Max pkts before interrupt */
3435 	uint32_t		rr_min_pkt_cnt; /* Mix pkts before interrupt */
3436 	uint32_t		rr_normal_blank_time; /* Normal intr freq */
3437 	uint32_t		rr_normal_pkt_cnt; /* Normal intr pkt cnt */
3438 	uint32_t		rr_ring_state; /* State of this ring */
3439 };
3440 
3441 struct ill_dls_capab_s {
3442 	ip_dld_tx_t		ill_tx;		/* Driver Tx routine */
3443 	void			*ill_tx_handle;	/* Driver Tx handle */
3444 	ip_dls_chg_soft_ring_t	ill_dls_change_status;
3445 						/* change soft ring fanout */
3446 	ip_dls_bind_t		ill_dls_bind;	/* to add CPU affinity */
3447 	ip_dls_unbind_t		ill_dls_unbind;	/* remove CPU affinity */
3448 	ill_rx_ring_t		*ill_ring_tbl; /* Ring to Sqp mapping table */
3449 	uint_t			ill_dls_soft_ring_cnt; /* Number of soft ring */
3450 	conn_t			*ill_unbind_conn; /* Conn used during unplumb */
3451 };
3452 
3453 /*
3454  * IP squeues exports
3455  */
3456 extern int 		ip_squeue_profile;
3457 extern int 		ip_squeue_bind;
3458 extern boolean_t 	ip_squeue_fanout;
3459 extern boolean_t	ip_squeue_soft_ring;
3460 extern uint_t		ip_threads_per_cpu;
3461 extern uint_t		ip_squeues_per_cpu;
3462 extern uint_t		ip_soft_rings_cnt;
3463 
3464 typedef struct squeue_set_s {
3465 	kmutex_t	sqs_lock;
3466 	struct squeue_s	**sqs_list;
3467 	int		sqs_size;
3468 	int		sqs_max_size;
3469 	processorid_t	sqs_bind;
3470 } squeue_set_t;
3471 
3472 #define	IP_SQUEUE_GET(hint) 						\
3473 	((!ip_squeue_fanout) ?	(CPU->cpu_squeue_set->sqs_list[0]) :	\
3474 		ip_squeue_random(hint))
3475 
3476 typedef void (*squeue_func_t)(squeue_t *, mblk_t *, sqproc_t, void *, uint8_t);
3477 
3478 extern void ip_squeue_init(void (*)(squeue_t *));
3479 extern squeue_t	*ip_squeue_random(uint_t);
3480 extern squeue_t *ip_squeue_get(ill_rx_ring_t *);
3481 extern int ip_squeue_bind_set(queue_t *, mblk_t *, char *, caddr_t, cred_t *);
3482 extern void ip_squeue_clean_all(ill_t *);
3483 extern void ip_soft_ring_assignment(ill_t *, ill_rx_ring_t *,
3484     mblk_t *, struct mac_header_info_s *);
3485 
3486 extern void ip_resume_tcp_bind(void *, mblk_t *, void *);
3487 extern void tcp_wput(queue_t *, mblk_t *);
3488 
3489 extern int	ip_fill_mtuinfo(struct in6_addr *, in_port_t,
3490 	struct ip6_mtuinfo *, netstack_t *);
3491 extern	ipif_t *conn_get_held_ipif(conn_t *, ipif_t **, int *);
3492 
3493 typedef void    (*ipsq_func_t)(ipsq_t *, queue_t *, mblk_t *, void *);
3494 
3495 /*
3496  * Squeue tags. Tags only need to be unique when the callback function is the
3497  * same to distinguish between different calls, but we use unique tags for
3498  * convenience anyway.
3499  */
3500 #define	SQTAG_IP_INPUT			1
3501 #define	SQTAG_TCP_INPUT_ICMP_ERR	2
3502 #define	SQTAG_TCP6_INPUT_ICMP_ERR	3
3503 #define	SQTAG_IP_TCP_INPUT		4
3504 #define	SQTAG_IP6_TCP_INPUT		5
3505 #define	SQTAG_IP_TCP_CLOSE		6
3506 #define	SQTAG_TCP_OUTPUT		7
3507 #define	SQTAG_TCP_TIMER			8
3508 #define	SQTAG_TCP_TIMEWAIT		9
3509 #define	SQTAG_TCP_ACCEPT_FINISH		10
3510 #define	SQTAG_TCP_ACCEPT_FINISH_Q0	11
3511 #define	SQTAG_TCP_ACCEPT_PENDING	12
3512 #define	SQTAG_TCP_LISTEN_DISCON		13
3513 #define	SQTAG_TCP_CONN_REQ_1		14
3514 #define	SQTAG_TCP_EAGER_BLOWOFF		15
3515 #define	SQTAG_TCP_EAGER_CLEANUP		16
3516 #define	SQTAG_TCP_EAGER_CLEANUP_Q0	17
3517 #define	SQTAG_TCP_CONN_IND		18
3518 #define	SQTAG_TCP_RSRV			19
3519 #define	SQTAG_TCP_ABORT_BUCKET		20
3520 #define	SQTAG_TCP_REINPUT		21
3521 #define	SQTAG_TCP_REINPUT_EAGER		22
3522 #define	SQTAG_TCP_INPUT_MCTL		23
3523 #define	SQTAG_TCP_RPUTOTHER		24
3524 #define	SQTAG_IP_PROTO_AGAIN		25
3525 #define	SQTAG_IP_FANOUT_TCP		26
3526 #define	SQTAG_IPSQ_CLEAN_RING		27
3527 #define	SQTAG_TCP_WPUT_OTHER		28
3528 #define	SQTAG_TCP_CONN_REQ_UNBOUND	29
3529 #define	SQTAG_TCP_SEND_PENDING		30
3530 #define	SQTAG_BIND_RETRY		31
3531 #define	SQTAG_UDP_FANOUT		32
3532 #define	SQTAG_UDP_INPUT			33
3533 #define	SQTAG_UDP_WPUT			34
3534 #define	SQTAG_UDP_OUTPUT		35
3535 #define	SQTAG_TCP_KSSL_INPUT		36
3536 #define	SQTAG_TCP_DROP_Q0		37
3537 #define	SQTAG_TCP_CONN_REQ_2		38
3538 
3539 #define	NOT_OVER_IP(ip_wq)	\
3540 	(ip_wq->q_next != NULL ||	\
3541 	    (ip_wq->q_qinfo->qi_minfo->mi_idname) == NULL ||	\
3542 	    strcmp(ip_wq->q_qinfo->qi_minfo->mi_idname,	\
3543 	    IP_MOD_NAME) != 0 ||	\
3544 	    ip_wq->q_qinfo->qi_minfo->mi_idnum != IP_MOD_ID)
3545 
3546 #endif	/* _KERNEL */
3547 
3548 #ifdef	__cplusplus
3549 }
3550 #endif
3551 
3552 #endif	/* _INET_IP_H */
3553