xref: /titanic_51/usr/src/uts/common/inet/ip.h (revision b9ccdc5a0f0a722ae408b257a831b90011369316)
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 /* Controls forwarding of IP packets, set via ndd */
217 #define	IP_FORWARD_NEVER	0
218 #define	IP_FORWARD_ALWAYS	1
219 
220 #define	WE_ARE_FORWARDING(ipst)	((ipst)->ips_ip_g_forward == IP_FORWARD_ALWAYS)
221 
222 #define	IPH_HDR_LENGTH(ipha)						\
223 	((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length & 0xF) << 2)
224 
225 #define	IPH_HDR_VERSION(ipha)						\
226 	((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length) >> 4)
227 
228 #ifdef _KERNEL
229 /*
230  * IP reassembly macros.  We hide starting and ending offsets in b_next and
231  * b_prev of messages on the reassembly queue.	The messages are chained using
232  * b_cont.  These macros are used in ip_reassemble() so we don't have to see
233  * the ugly casts and assignments.
234  * Note that the offsets are <= 64k i.e. a uint_t is sufficient to represent
235  * them.
236  */
237 #define	IP_REASS_START(mp)		((uint_t)(uintptr_t)((mp)->b_next))
238 #define	IP_REASS_SET_START(mp, u)	\
239 	((mp)->b_next = (mblk_t *)(uintptr_t)(u))
240 #define	IP_REASS_END(mp)		((uint_t)(uintptr_t)((mp)->b_prev))
241 #define	IP_REASS_SET_END(mp, u)		\
242 	((mp)->b_prev = (mblk_t *)(uintptr_t)(u))
243 
244 #define	IP_REASS_COMPLETE	0x1
245 #define	IP_REASS_PARTIAL	0x2
246 #define	IP_REASS_FAILED		0x4
247 
248 /*
249  * Test to determine whether this is a module instance of IP or a
250  * driver instance of IP.
251  */
252 #define	CONN_Q(q)	(WR(q)->q_next == NULL)
253 
254 #define	Q_TO_CONN(q)	((conn_t *)(q)->q_ptr)
255 #define	Q_TO_TCP(q)	(Q_TO_CONN((q))->conn_tcp)
256 #define	Q_TO_UDP(q)	(Q_TO_CONN((q))->conn_udp)
257 #define	Q_TO_ICMP(q)	(Q_TO_CONN((q))->conn_icmp)
258 #define	Q_TO_RTS(q)	(Q_TO_CONN((q))->conn_rts)
259 
260 /*
261  * The following two macros are used by IP to get the appropriate
262  * wq and rq for a conn. If it is a TCP conn, then we need
263  * tcp_wq/tcp_rq else, conn_wq/conn_rq. IP can use conn_wq and conn_rq
264  * from a conn directly if it knows that the conn is not TCP.
265  */
266 #define	CONNP_TO_WQ(connp)	\
267 	(IPCL_IS_TCP(connp) ? (connp)->conn_tcp->tcp_wq : (connp)->conn_wq)
268 
269 #define	CONNP_TO_RQ(connp)	RD(CONNP_TO_WQ(connp))
270 
271 #define	GRAB_CONN_LOCK(q)	{				\
272 	if (q != NULL && CONN_Q(q))				\
273 		mutex_enter(&(Q_TO_CONN(q))->conn_lock);	\
274 }
275 
276 #define	RELEASE_CONN_LOCK(q)	{				\
277 	if (q != NULL && CONN_Q(q))				\
278 		mutex_exit(&(Q_TO_CONN(q))->conn_lock);		\
279 }
280 
281 /* "Congestion controlled" protocol */
282 #define	IP_FLOW_CONTROLLED_ULP(p)   ((p) == IPPROTO_TCP || (p) == IPPROTO_SCTP)
283 
284 /*
285  * Complete the pending operation. Usually an ioctl. Can also
286  * be a bind or option management request that got enqueued
287  * in an ipsq_t. Called on completion of the operation.
288  */
289 #define	CONN_OPER_PENDING_DONE(connp)	{			\
290 	mutex_enter(&(connp)->conn_lock);			\
291 	(connp)->conn_oper_pending_ill = NULL;			\
292 	cv_broadcast(&(connp)->conn_refcv);			\
293 	mutex_exit(&(connp)->conn_lock);			\
294 	CONN_DEC_REF(connp);					\
295 }
296 
297 /* Get the credential of an IP queue of unknown type */
298 #define	GET_QUEUE_CRED(wq)						\
299 	((wq)->q_next ? (((ill_t *)(wq)->q_ptr)->ill_credp) \
300 	    : ((Q_TO_CONN((wq)))->conn_cred))
301 
302 /*
303  * Flags for the various ip_fanout_* routines.
304  */
305 #define	IP_FF_SEND_ICMP		0x01	/* Send an ICMP error */
306 #define	IP_FF_HDR_COMPLETE	0x02	/* Call ip_hdr_complete if error */
307 #define	IP_FF_CKSUM		0x04	/* Recompute ipha_cksum if error */
308 #define	IP_FF_RAWIP		0x08	/* Use rawip mib variable */
309 #define	IP_FF_SRC_QUENCH	0x10	/* OK to send ICMP_SOURCE_QUENCH */
310 #define	IP_FF_SYN_ADDIRE	0x20	/* Add IRE if TCP syn packet */
311 #define	IP_FF_IPINFO		0x80	/* Used for both V4 and V6 */
312 #define	IP_FF_SEND_SLLA		0x100	/* Send source link layer info ? */
313 #define	IPV6_REACHABILITY_CONFIRMATION	0x200	/* Flags for ip_xmit_v6 */
314 #define	IP_FF_NO_MCAST_LOOP	0x400	/* No multicasts for sending zone */
315 
316 /*
317  * Following flags are used by IPQoS to determine if policy processing is
318  * required.
319  */
320 #define	IP6_NO_IPPOLICY		0x800	/* Don't do IPQoS processing */
321 #define	IP6_IN_LLMCAST		0x1000	/* Multicast */
322 
323 #define	IP_FF_LOOPBACK		0x2000	/* Loopback fanout */
324 #define	IP_FF_SCTP_CSUM_ERR	0x4000	/* sctp pkt has failed chksum */
325 
326 #ifndef	IRE_DB_TYPE
327 #define	IRE_DB_TYPE	M_SIG
328 #endif
329 
330 #ifndef	IRE_DB_REQ_TYPE
331 #define	IRE_DB_REQ_TYPE	M_PCSIG
332 #endif
333 
334 #ifndef	IRE_ARPRESOLVE_TYPE
335 #define	IRE_ARPRESOLVE_TYPE	M_EVENT
336 #endif
337 
338 /*
339  * Values for squeue switch:
340  */
341 
342 #define	IP_SQUEUE_ENTER_NODRAIN	1
343 #define	IP_SQUEUE_ENTER	2
344 /*
345  * This is part of the interface between Transport provider and
346  * IP which can be used to set policy information. This is usually
347  * accompanied with O_T_BIND_REQ/T_BIND_REQ.ip_bind assumes that
348  * only IPSEC_POLICY_SET is there when it is found in the chain.
349  * The information contained is an struct ipsec_req_t. On success
350  * or failure, either the T_BIND_ACK or the T_ERROR_ACK is returned.
351  * IPSEC_POLICY_SET is never returned.
352  */
353 #define	IPSEC_POLICY_SET	M_SETOPTS
354 
355 #define	IRE_IS_LOCAL(ire)	((ire != NULL) && \
356 				((ire)->ire_type & (IRE_LOCAL | IRE_LOOPBACK)))
357 
358 #define	IRE_IS_TARGET(ire)	((ire != NULL) && \
359 				((ire)->ire_type != IRE_BROADCAST))
360 
361 /* IP Fragmentation Reassembly Header */
362 typedef struct ipf_s {
363 	struct ipf_s	*ipf_hash_next;
364 	struct ipf_s	**ipf_ptphn;	/* Pointer to previous hash next. */
365 	uint32_t	ipf_ident;	/* Ident to match. */
366 	uint8_t		ipf_protocol;	/* Protocol to match. */
367 	uchar_t		ipf_last_frag_seen : 1;	/* Last fragment seen ? */
368 	time_t		ipf_timestamp;	/* Reassembly start time. */
369 	mblk_t		*ipf_mp;	/* mblk we live in. */
370 	mblk_t		*ipf_tail_mp;	/* Frag queue tail pointer. */
371 	int		ipf_hole_cnt;	/* Number of holes (hard-case). */
372 	int		ipf_end;	/* Tail end offset (0 -> hard-case). */
373 	uint_t		ipf_gen;	/* Frag queue generation */
374 	size_t		ipf_count;	/* Count of bytes used by frag */
375 	uint_t		ipf_nf_hdr_len; /* Length of nonfragmented header */
376 	in6_addr_t	ipf_v6src;	/* IPv6 source address */
377 	in6_addr_t	ipf_v6dst;	/* IPv6 dest address */
378 	uint_t		ipf_prev_nexthdr_offset; /* Offset for nexthdr value */
379 	uint8_t		ipf_ecn;	/* ECN info for the fragments */
380 	uint8_t		ipf_num_dups;	/* Number of times dup frags recvd */
381 	uint16_t	ipf_checksum_flags; /* Hardware checksum flags */
382 	uint32_t	ipf_checksum;	/* Partial checksum of fragment data */
383 } ipf_t;
384 
385 #define	ipf_src	V4_PART_OF_V6(ipf_v6src)
386 #define	ipf_dst	V4_PART_OF_V6(ipf_v6dst)
387 
388 typedef enum {
389 	IB_PKT = 0x01,
390 	OB_PKT = 0x02
391 } ip_pkt_t;
392 
393 #define	UPDATE_IB_PKT_COUNT(ire)\
394 	{ \
395 	(ire)->ire_ib_pkt_count++; \
396 	if ((ire)->ire_ipif != NULL) { \
397 		/* \
398 		 * forwarding packet \
399 		 */ \
400 		if ((ire)->ire_type & (IRE_LOCAL|IRE_BROADCAST)) \
401 			atomic_add_32(&(ire)->ire_ipif->ipif_ib_pkt_count, 1);\
402 		else \
403 			atomic_add_32(&(ire)->ire_ipif->ipif_fo_pkt_count, 1);\
404 	} \
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 #define	IP_RPUT_LOCAL(q, mp, ipha, ire, recv_ill) \
416 { \
417 	switch (ipha->ipha_protocol) { \
418 		case IPPROTO_UDP: \
419 			ip_udp_input(q, mp, ipha, ire, recv_ill); \
420 			break; \
421 		default: \
422 			ip_proto_input(q, mp, ipha, ire, recv_ill, 0); \
423 			break; \
424 	} \
425 }
426 
427 /*
428  * NCE_EXPIRED is TRUE when we have a non-permanent nce that was
429  * found to be REACHABLE more than ip_ire_arp_interval ms ago.
430  * This macro is used to age existing nce_t entries. The
431  * nce's will get cleaned up in the following circumstances:
432  * - ip_ire_trash_reclaim will free nce's using ndp_cache_reclaim
433  *    when memory is low,
434  * - ip_arp_news, when updates are received.
435  * - if the nce is NCE_EXPIRED(), it will deleted, so that a new
436  *   arp request will need to be triggered from an ND_INITIAL nce.
437  *
438  * Note that the nce state transition follows the pattern:
439  *	ND_INITIAL -> ND_INCOMPLETE -> ND_REACHABLE
440  * after which the nce is deleted when it has expired.
441  *
442  * nce_last is the timestamp that indicates when the nce_res_mp in the
443  * nce_t was last updated to a valid link-layer address.  nce_last gets
444  * modified/updated :
445  *  - when the nce is created
446  *  - every time we get a sane arp response for the nce.
447  */
448 #define	NCE_EXPIRED(nce, ipst)	(nce->nce_last > 0 &&	\
449 	    ((nce->nce_flags & NCE_F_PERMANENT) == 0) &&	\
450 	    ((TICK_TO_MSEC(lbolt64) - nce->nce_last) > 		\
451 		(ipst)->ips_ip_ire_arp_interval))
452 
453 #endif /* _KERNEL */
454 
455 /* ICMP types */
456 #define	ICMP_ECHO_REPLY			0
457 #define	ICMP_DEST_UNREACHABLE		3
458 #define	ICMP_SOURCE_QUENCH		4
459 #define	ICMP_REDIRECT			5
460 #define	ICMP_ECHO_REQUEST		8
461 #define	ICMP_ROUTER_ADVERTISEMENT	9
462 #define	ICMP_ROUTER_SOLICITATION	10
463 #define	ICMP_TIME_EXCEEDED		11
464 #define	ICMP_PARAM_PROBLEM		12
465 #define	ICMP_TIME_STAMP_REQUEST		13
466 #define	ICMP_TIME_STAMP_REPLY		14
467 #define	ICMP_INFO_REQUEST		15
468 #define	ICMP_INFO_REPLY			16
469 #define	ICMP_ADDRESS_MASK_REQUEST	17
470 #define	ICMP_ADDRESS_MASK_REPLY		18
471 
472 /* ICMP_TIME_EXCEEDED codes */
473 #define	ICMP_TTL_EXCEEDED		0
474 #define	ICMP_REASSEMBLY_TIME_EXCEEDED	1
475 
476 /* ICMP_DEST_UNREACHABLE codes */
477 #define	ICMP_NET_UNREACHABLE		0
478 #define	ICMP_HOST_UNREACHABLE		1
479 #define	ICMP_PROTOCOL_UNREACHABLE	2
480 #define	ICMP_PORT_UNREACHABLE		3
481 #define	ICMP_FRAGMENTATION_NEEDED	4
482 #define	ICMP_SOURCE_ROUTE_FAILED	5
483 #define	ICMP_DEST_NET_UNKNOWN		6
484 #define	ICMP_DEST_HOST_UNKNOWN		7
485 #define	ICMP_SRC_HOST_ISOLATED		8
486 #define	ICMP_DEST_NET_UNREACH_ADMIN	9
487 #define	ICMP_DEST_HOST_UNREACH_ADMIN	10
488 #define	ICMP_DEST_NET_UNREACH_TOS	11
489 #define	ICMP_DEST_HOST_UNREACH_TOS	12
490 
491 /* ICMP Header Structure */
492 typedef struct icmph_s {
493 	uint8_t		icmph_type;
494 	uint8_t		icmph_code;
495 	uint16_t	icmph_checksum;
496 	union {
497 		struct { /* ECHO request/response structure */
498 			uint16_t	u_echo_ident;
499 			uint16_t	u_echo_seqnum;
500 		} u_echo;
501 		struct { /* Destination unreachable structure */
502 			uint16_t	u_du_zero;
503 			uint16_t	u_du_mtu;
504 		} u_du;
505 		struct { /* Parameter problem structure */
506 			uint8_t		u_pp_ptr;
507 			uint8_t		u_pp_rsvd[3];
508 		} u_pp;
509 		struct { /* Redirect structure */
510 			ipaddr_t	u_rd_gateway;
511 		} u_rd;
512 	} icmph_u;
513 } icmph_t;
514 
515 #define	icmph_echo_ident	icmph_u.u_echo.u_echo_ident
516 #define	icmph_echo_seqnum	icmph_u.u_echo.u_echo_seqnum
517 #define	icmph_du_zero		icmph_u.u_du.u_du_zero
518 #define	icmph_du_mtu		icmph_u.u_du.u_du_mtu
519 #define	icmph_pp_ptr		icmph_u.u_pp.u_pp_ptr
520 #define	icmph_rd_gateway	icmph_u.u_rd.u_rd_gateway
521 
522 #define	ICMPH_SIZE	8
523 
524 /*
525  * Minimum length of transport layer header included in an ICMP error
526  * message for it to be considered valid.
527  */
528 #define	ICMP_MIN_TP_HDR_LEN	8
529 
530 /* Aligned IP header */
531 typedef struct ipha_s {
532 	uint8_t		ipha_version_and_hdr_length;
533 	uint8_t		ipha_type_of_service;
534 	uint16_t	ipha_length;
535 	uint16_t	ipha_ident;
536 	uint16_t	ipha_fragment_offset_and_flags;
537 	uint8_t		ipha_ttl;
538 	uint8_t		ipha_protocol;
539 	uint16_t	ipha_hdr_checksum;
540 	ipaddr_t	ipha_src;
541 	ipaddr_t	ipha_dst;
542 } ipha_t;
543 
544 /*
545  * IP Flags
546  *
547  * Some of these constant names are copied for the DTrace IP provider in
548  * usr/src/lib/libdtrace/common/{ip.d.in, ip.sed.in}, which should be kept
549  * in sync.
550  */
551 #define	IPH_DF		0x4000	/* Don't fragment */
552 #define	IPH_MF		0x2000	/* More fragments to come */
553 #define	IPH_OFFSET	0x1FFF	/* Where the offset lives */
554 #define	IPH_FRAG_HDR	0x8000	/* IPv6 don't fragment bit */
555 
556 /* ECN code points for IPv4 TOS byte and IPv6 traffic class octet. */
557 #define	IPH_ECN_NECT	0x0	/* Not ECN-Capable Transport */
558 #define	IPH_ECN_ECT1	0x1	/* ECN-Capable Transport, ECT(1) */
559 #define	IPH_ECN_ECT0	0x2	/* ECN-Capable Transport, ECT(0) */
560 #define	IPH_ECN_CE	0x3	/* ECN-Congestion Experienced (CE) */
561 
562 /* IP Mac info structure */
563 typedef struct ip_m_s {
564 	t_uscalar_t	ip_m_mac_type;	/* From <sys/dlpi.h> */
565 	int		ip_m_type;	/* From <net/if_types.h> */
566 	boolean_t	(*ip_m_v4mapinfo)(uint_t, uint8_t *, uint8_t *,
567 			    uint32_t *, ipaddr_t *);
568 	boolean_t	(*ip_m_v6mapinfo)(uint_t, uint8_t *, uint8_t *,
569 			    uint32_t *, in6_addr_t *);
570 	boolean_t	(*ip_m_v6intfid)(uint_t, uint8_t *, in6_addr_t *);
571 } ip_m_t;
572 
573 /*
574  * The following functions attempt to reduce the link layer dependency
575  * of the IP stack. The current set of link specific operations are:
576  * a. map from IPv4 class D (224.0/4) multicast address range to the link
577  * layer multicast address range.
578  * b. map from IPv6 multicast address range (ff00::/8) to the link
579  * layer multicast address range.
580  * c. derive the default IPv6 interface identifier from the link layer
581  * address.
582  */
583 #define	MEDIA_V4MINFO(ip_m, plen, bphys, maddr, hwxp, v4ptr) \
584 	(((ip_m)->ip_m_v4mapinfo != NULL) && \
585 	(*(ip_m)->ip_m_v4mapinfo)(plen, bphys, maddr, hwxp, v4ptr))
586 #define	MEDIA_V6INTFID(ip_m, plen, phys, v6ptr) \
587 	(((ip_m)->ip_m_v6intfid != NULL) && \
588 	(*(ip_m)->ip_m_v6intfid)(plen, phys, v6ptr))
589 #define	MEDIA_V6MINFO(ip_m, plen, bphys, maddr, hwxp, v6ptr) \
590 	(((ip_m)->ip_m_v6mapinfo != NULL) && \
591 	(*(ip_m)->ip_m_v6mapinfo)(plen, bphys, maddr, hwxp, v6ptr))
592 
593 /* Router entry types */
594 #define	IRE_BROADCAST		0x0001	/* Route entry for broadcast address */
595 #define	IRE_DEFAULT		0x0002	/* Route entry for default gateway */
596 #define	IRE_LOCAL		0x0004	/* Route entry for local address */
597 #define	IRE_LOOPBACK		0x0008	/* Route entry for loopback address */
598 #define	IRE_PREFIX		0x0010	/* Route entry for prefix routes */
599 #define	IRE_CACHE		0x0020	/* Cached Route entry */
600 #define	IRE_IF_NORESOLVER	0x0040	/* Route entry for local interface */
601 					/* net without any address mapping. */
602 #define	IRE_IF_RESOLVER		0x0080	/* Route entry for local interface */
603 					/* net with resolver. */
604 #define	IRE_HOST		0x0100	/* Host route entry */
605 #define	IRE_HOST_REDIRECT	0x0200	/* only used for T_SVR4_OPTMGMT_REQ */
606 
607 #define	IRE_INTERFACE		(IRE_IF_NORESOLVER | IRE_IF_RESOLVER)
608 #define	IRE_OFFSUBNET		(IRE_DEFAULT | IRE_PREFIX | IRE_HOST)
609 #define	IRE_CACHETABLE		(IRE_CACHE | IRE_BROADCAST | IRE_LOCAL | \
610 				IRE_LOOPBACK)
611 #define	IRE_FORWARDTABLE	(IRE_INTERFACE | IRE_OFFSUBNET)
612 
613 /*
614  * If an IRE is marked with IRE_MARK_CONDEMNED, the last walker of
615  * the bucket should delete this IRE from this bucket.
616  */
617 #define	IRE_MARK_CONDEMNED	0x0001
618 /*
619  * If a broadcast IRE is marked with IRE_MARK_NORECV, ip_rput will drop the
620  * broadcast packets received on that interface. This is marked only
621  * on broadcast ires. Employed by IPMP, where we have multiple NICs on the
622  * same subnet receiving the same broadcast packet.
623  */
624 #define	IRE_MARK_NORECV		0x0002
625 /*
626  * IRE_CACHE marked this way won't be returned by ire_cache_lookup. Need
627  * to look specifically using MATCH_IRE_MARK_HIDDEN. Used by IPMP.
628  */
629 #define	IRE_MARK_HIDDEN		0x0004	/* Typically Used by in.mpathd */
630 
631 /*
632  * An IRE with IRE_MARK_NOADD is created in ip_newroute_ipif when the outgoing
633  * interface is specified by e.g. IP_PKTINFO.  The IRE is not added to the IRE
634  * cache table.
635  */
636 #define	IRE_MARK_NOADD		0x0008	/* Mark not to add ire in cache */
637 
638 /*
639  * IRE marked with IRE_MARK_TEMPORARY means that this IRE has been used
640  * either for forwarding a packet or has not been used for sending
641  * traffic on TCP connections terminated on this system.  In both
642  * cases, this IRE is the first to go when IRE is being cleaned up.
643  */
644 #define	IRE_MARK_TEMPORARY	0x0010
645 
646 /*
647  * IRE marked with IRE_MARK_USESRC_CHECK means that while adding an IRE with
648  * this mark, additional atomic checks need to be performed. For eg: by the
649  * time an IRE_CACHE is created, sent up to ARP and then comes back to IP; the
650  * usesrc grouping could have changed in which case we want to fail adding
651  * the IRE_CACHE entry
652  */
653 #define	IRE_MARK_USESRC_CHECK	0x0020
654 
655 /*
656  * IRE_MARK_PRIVATE_ADDR is used for IP_NEXTHOP. When IP_NEXTHOP is set, the
657  * routing table lookup for the destination is bypassed and the packet is
658  * sent directly to the specified nexthop. The associated IRE_CACHE entries
659  * should be marked with IRE_MARK_PRIVATE_ADDR flag so that they don't show up
660  * in regular ire cache lookups.
661  */
662 #define	IRE_MARK_PRIVATE_ADDR	0x0040
663 
664 /*
665  * When we send an ARP resolution query for the nexthop gateway's ire,
666  * we use esballoc to create the ire_t in the AR_ENTRY_QUERY mblk
667  * chain, and mark its ire_marks with IRE_MARK_UNCACHED. This flag
668  * indicates that information from ARP has not been transferred to a
669  * permanent IRE_CACHE entry. The flag is reset only when the
670  * information is successfully transferred to an ire_cache entry (in
671  * ire_add()). Attempting to free the AR_ENTRY_QUERY mblk chain prior
672  * to ire_add (e.g., from arp, or from ip`ip_wput_nondata) will
673  * require that the resources (incomplete ire_cache and/or nce) must
674  * be cleaned up. The free callback routine (ire_freemblk()) checks
675  * for IRE_MARK_UNCACHED to see if any resources that are pinned down
676  * will need to be cleaned up or not.
677  */
678 
679 #define	IRE_MARK_UNCACHED	0x0080
680 
681 /*
682  * The comment below (and for other netstack_t references) refers
683  * to the fact that we only do netstack_hold in particular cases,
684  * such as the references from open streams (ill_t and conn_t's
685  * pointers). Internally within IP we rely on IP's ability to cleanup e.g.
686  * ire_t's when an ill goes away.
687  */
688 typedef struct ire_expire_arg_s {
689 	int		iea_flush_flag;
690 	ip_stack_t	*iea_ipst;	/* Does not have a netstack_hold */
691 } ire_expire_arg_t;
692 
693 /* Flags with ire_expire routine */
694 #define	FLUSH_ARP_TIME		0x0001	/* ARP info potentially stale timer */
695 #define	FLUSH_REDIRECT_TIME	0x0002	/* Redirects potentially stale */
696 #define	FLUSH_MTU_TIME		0x0004	/* Include path MTU per RFC 1191 */
697 
698 /* Arguments to ire_flush_cache() */
699 #define	IRE_FLUSH_DELETE	0
700 #define	IRE_FLUSH_ADD		1
701 
702 /*
703  * Open/close synchronization flags.
704  * These are kept in a separate field in the conn and the synchronization
705  * depends on the atomic 32 bit access to that field.
706  */
707 #define	CONN_CLOSING		0x01	/* ip_close waiting for ip_wsrv */
708 #define	CONN_IPSEC_LOAD_WAIT	0x02	/* waiting for load */
709 #define	CONN_CONDEMNED		0x04	/* conn is closing, no more refs */
710 #define	CONN_INCIPIENT		0x08	/* conn not yet visible, no refs */
711 #define	CONN_QUIESCED		0x10	/* conn is now quiescent */
712 
713 /* Used to check connection state flags before caching the IRE */
714 #define	CONN_CACHE_IRE(connp)	\
715 	(!((connp)->conn_state_flags & (CONN_CLOSING|CONN_CONDEMNED)))
716 
717 /*
718  * Parameter to ip_output giving the identity of the caller.
719  * IP_WSRV means the packet was enqueued in the STREAMS queue
720  * due to flow control and is now being reprocessed in the context of
721  * the STREAMS service procedure, consequent to flow control relief.
722  * IRE_SEND means the packet is being reprocessed consequent to an
723  * ire cache creation and addition and this may or may not be happening
724  * in the service procedure context. Anything other than the above 2
725  * cases is identified as IP_WPUT. Most commonly this is the case of
726  * packets coming down from the application.
727  */
728 #ifdef _KERNEL
729 #define	IP_WSRV			1	/* Called from ip_wsrv */
730 #define	IP_WPUT			2	/* Called from ip_wput */
731 #define	IRE_SEND		3	/* Called from ire_send */
732 
733 /*
734  * Extra structures need for per-src-addr filtering (IGMPv3/MLDv2)
735  */
736 #define	MAX_FILTER_SIZE	64
737 
738 typedef struct slist_s {
739 	int		sl_numsrc;
740 	in6_addr_t	sl_addr[MAX_FILTER_SIZE];
741 } slist_t;
742 
743 /*
744  * Following struct is used to maintain retransmission state for
745  * a multicast group.  One rtx_state_t struct is an in-line field
746  * of the ilm_t struct; the slist_ts in the rtx_state_t struct are
747  * alloc'd as needed.
748  */
749 typedef struct rtx_state_s {
750 	uint_t		rtx_timer;	/* retrans timer */
751 	int		rtx_cnt;	/* retrans count */
752 	int		rtx_fmode_cnt;	/* retrans count for fmode change */
753 	slist_t		*rtx_allow;
754 	slist_t		*rtx_block;
755 } rtx_state_t;
756 
757 /*
758  * Used to construct list of multicast address records that will be
759  * sent in a single listener report.
760  */
761 typedef struct mrec_s {
762 	struct mrec_s	*mrec_next;
763 	uint8_t		mrec_type;
764 	uint8_t		mrec_auxlen;	/* currently unused */
765 	in6_addr_t	mrec_group;
766 	slist_t		mrec_srcs;
767 } mrec_t;
768 
769 /* Group membership list per upper conn */
770 /*
771  * XXX add ilg info for ifaddr/ifindex.
772  * XXX can we make ilg survive an ifconfig unplumb + plumb
773  * by setting the ipif/ill to NULL and recover that later?
774  *
775  * ilg_ipif is used by IPv4 as multicast groups are joined using an interface
776  * address (ipif).
777  * ilg_ill is used by IPv6 as multicast groups are joined using an interface
778  * index (phyint->phyint_ifindex).
779  * ilg_ill is NULL for IPv4 and ilg_ipif is NULL for IPv6.
780  *
781  * ilg records the state of multicast memberships of a socket end point.
782  * ilm records the state of multicast memberships with the driver and is
783  * maintained per interface.
784  *
785  * Notes :
786  *
787  * 1) There is no direct link between a given ilg and ilm. If the
788  *    application has joined a group G with ifindex I, we will have
789  *    an ilg with ilg_v6group and ilg_ill. There will be a corresponding
790  *    ilm with ilm_ill/ilm_v6addr recording the multicast membership.
791  *    To delete the membership,
792  *
793  *		a) Search for ilg matching on G and I with ilg_v6group
794  *		   and ilg_ill. Delete ilg_ill.
795  *		b) Search the corresponding ilm matching on G and I with
796  *		   ilm_v6addr and ilm_ill. Delete ilm.
797  *
798  *    In IPv4, the only difference is, we look using ipifs instead of
799  *    ills.
800  *
801  * 2) With IP multipathing, we want to keep receiving even after the
802  *    interface has failed. We do this by moving multicast memberships
803  *    to a new_ill within the group. This is achieved by sending
804  *    DL_DISABMULTI_REQS on ilg_ill/ilm_ill and sending DL_ENABMULTIREQS
805  *    on the new_ill and changing ilg_ill/ilm_ill to new_ill. But, we
806  *    need to be able to delete memberships which will still come down
807  *    with the ifindex of the old ill which is what the application
808  *    knows of. Thus we store the ilm_/ilg_orig_ifindex to keep track
809  *    of where we joined initially so that we can lookup even after we
810  *    moved the membership. It is also used for moving back the membership
811  *    when the old ill has been repaired. This is done by looking up for
812  *    ilms with ilm_orig_ifindex matching on the old ill's ifindex. Only
813  *    ilms actually move from old ill to new ill. ilgs don't move (just
814  *    the ilg_ill is changed when it moves) as it just records the state
815  *    of the application that has joined a group G where as ilm records
816  *    the state joined with the driver. Thus when we send DL_XXXMULTI_REQs
817  *    we also need to keep the ilm in the right ill.
818  *
819  *    In IPv4, as ipifs move from old ill to new_ill, ilgs and ilms move
820  *    implicitly as we use only ipifs in IPv4. Thus, one can always lookup
821  *    a given ilm/ilg even after it fails without the support of
822  *    orig_ifindex. We move ilms still to record the driver state as
823  *    mentioned above.
824  */
825 
826 /*
827  * The ilg_t and ilm_t members are protected by ipsq. They can be changed only
828  * by a thread executing in the ipsq. In other words add/delete of a
829  * multicast group has to execute in the ipsq.
830  */
831 #define	ILG_DELETED	0x1		/* ilg_flags */
832 typedef struct ilg_s {
833 	in6_addr_t	ilg_v6group;
834 	struct ipif_s	*ilg_ipif;	/* Logical interface we are member on */
835 	struct ill_s	*ilg_ill;	/* Used by IPv6 */
836 	int		ilg_orig_ifindex; /* Interface originally joined on */
837 	uint_t		ilg_flags;
838 	mcast_record_t	ilg_fmode;	/* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */
839 	slist_t		*ilg_filter;
840 } ilg_t;
841 
842 /*
843  * Multicast address list entry for ill.
844  * ilm_ipif is used by IPv4 as multicast groups are joined using ipif.
845  * ilm_ill is used by IPv6 as multicast groups are joined using ill.
846  * ilm_ill is NULL for IPv4 and ilm_ipif is NULL for IPv6.
847  *
848  * The comment below (and for other netstack_t references) refers
849  * to the fact that we only do netstack_hold in particular cases,
850  * such as the references from open streams (ill_t and conn_t's
851  * pointers). Internally within IP we rely on IP's ability to cleanup e.g.
852  * ire_t's when an ill goes away.
853  */
854 #define	ILM_DELETED	0x1		/* ilm_flags */
855 typedef struct ilm_s {
856 	in6_addr_t	ilm_v6addr;
857 	int		ilm_refcnt;
858 	uint_t		ilm_timer;	/* IGMP/MLD query resp timer, in msec */
859 	struct ipif_s	*ilm_ipif;	/* Back pointer to ipif for IPv4 */
860 	struct ilm_s	*ilm_next;	/* Linked list for each ill */
861 	uint_t		ilm_state;	/* state of the membership */
862 	struct ill_s	*ilm_ill;	/* Back pointer to ill for IPv6 */
863 	int		ilm_orig_ifindex;  /* V6_MULTICAST_IF/ilm_ipif index */
864 	uint_t		ilm_flags;
865 	boolean_t	ilm_is_new;	/* new ilm */
866 	boolean_t	ilm_notify_driver; /* Need to notify the driver */
867 	zoneid_t	ilm_zoneid;
868 	int		ilm_no_ilg_cnt;	/* number of joins w/ no ilg */
869 	mcast_record_t	ilm_fmode;	/* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */
870 	slist_t		*ilm_filter;	/* source filter list */
871 	slist_t		*ilm_pendsrcs;	/* relevant src addrs for pending req */
872 	rtx_state_t	ilm_rtx;	/* SCR retransmission state */
873 	ip_stack_t	*ilm_ipst;	/* Does not have a netstack_hold */
874 } ilm_t;
875 
876 #define	ilm_addr	V4_PART_OF_V6(ilm_v6addr)
877 
878 /*
879  * ilm_walker_cleanup needs to execute when the ilm_walker_cnt goes down to
880  * zero. In addition it needs to block new walkers while it is unlinking ilm's
881  * from the list. Thus simple atomics for the ill_ilm_walker_cnt don't suffice.
882  */
883 #define	ILM_WALKER_HOLD(ill)    {               \
884 	mutex_enter(&(ill)->ill_lock);          \
885 	ill->ill_ilm_walker_cnt++;              \
886 	mutex_exit(&(ill)->ill_lock);           \
887 }
888 
889 /*
890  * ilm_walker_cleanup releases ill_lock
891  */
892 #define	ILM_WALKER_RELE(ill)	{ 		\
893 	mutex_enter(&(ill)->ill_lock);		\
894 	(ill)->ill_ilm_walker_cnt--;		\
895 	if ((ill)->ill_ilm_walker_cnt == 0 && (ill)->ill_ilm_cleanup_reqd) \
896 		ilm_walker_cleanup(ill);	\
897 	else 					\
898 		mutex_exit(&(ill)->ill_lock);	\
899 }
900 
901 /*
902  * Soft reference to an IPsec SA.
903  *
904  * On relative terms, conn's can be persistent (living as long as the
905  * processes which create them), while SA's are ephemeral (dying when
906  * they hit their time-based or byte-based lifetimes).
907  *
908  * We could hold a hard reference to an SA from an ipsec_latch_t,
909  * but this would cause expired SA's to linger for a potentially
910  * unbounded time.
911  *
912  * Instead, we remember the hash bucket number and bucket generation
913  * in addition to the pointer.  The bucket generation is incremented on
914  * each deletion.
915  */
916 typedef struct ipsa_ref_s
917 {
918 	struct ipsa_s	*ipsr_sa;
919 	struct isaf_s	*ipsr_bucket;
920 	uint64_t	ipsr_gen;
921 } ipsa_ref_t;
922 
923 /*
924  * IPsec "latching" state.
925  *
926  * In the presence of IPsec policy, fully-bound conn's bind a connection
927  * to more than just the 5-tuple, but also a specific IPsec action and
928  * identity-pair.
929  *
930  * As an optimization, we also cache soft references to IPsec SA's
931  * here so that we can fast-path around most of the work needed for
932  * outbound IPsec SA selection.
933  *
934  * Were it not for TCP's detached connections, this state would be
935  * in-line in conn_t; instead, this is in a separate structure so it
936  * can be handed off to TCP when a connection is detached.
937  */
938 typedef struct ipsec_latch_s
939 {
940 	kmutex_t	ipl_lock;
941 	uint32_t	ipl_refcnt;
942 
943 	uint64_t	ipl_unique;
944 	struct ipsec_policy_s	*ipl_in_policy; /* latched policy (in) */
945 	struct ipsec_policy_s	*ipl_out_policy; /* latched policy (out) */
946 	struct ipsec_action_s	*ipl_in_action;	/* latched action (in) */
947 	struct ipsec_action_s	*ipl_out_action; /* latched action (out) */
948 	cred_t		*ipl_local_id;
949 	struct ipsid_s	*ipl_local_cid;
950 	struct ipsid_s	*ipl_remote_cid;
951 	unsigned int
952 			ipl_out_action_latched : 1,
953 			ipl_in_action_latched : 1,
954 			ipl_out_policy_latched : 1,
955 			ipl_in_policy_latched : 1,
956 
957 			ipl_ids_latched : 1,
958 
959 			ipl_pad_to_bit_31 : 27;
960 
961 	ipsa_ref_t	ipl_ref[2]; /* 0: ESP, 1: AH */
962 
963 } ipsec_latch_t;
964 
965 #define	IPLATCH_REFHOLD(ipl) { \
966 	atomic_add_32(&(ipl)->ipl_refcnt, 1);		\
967 	ASSERT((ipl)->ipl_refcnt != 0);			\
968 }
969 
970 #define	IPLATCH_REFRELE(ipl, ns) {				\
971 	ASSERT((ipl)->ipl_refcnt != 0);				\
972 	membar_exit();						\
973 	if (atomic_add_32_nv(&(ipl)->ipl_refcnt, -1) == 0)	\
974 		iplatch_free(ipl, ns);			\
975 }
976 
977 /*
978  * peer identity structure.
979  */
980 
981 typedef struct conn_s conn_t;
982 
983 /*
984  * The old IP client structure "ipc_t" is gone. All the data is stored in the
985  * connection structure "conn_t" now. The mapping of old and new fields looks
986  * like this:
987  *
988  * ipc_ulp			conn_ulp
989  * ipc_rq			conn_rq
990  * ipc_wq			conn_wq
991  *
992  * ipc_laddr			conn_src
993  * ipc_faddr			conn_rem
994  * ipc_v6laddr			conn_srcv6
995  * ipc_v6faddr			conn_remv6
996  *
997  * ipc_lport			conn_lport
998  * ipc_fport			conn_fport
999  * ipc_ports			conn_ports
1000  *
1001  * ipc_policy			conn_policy
1002  * ipc_latch			conn_latch
1003  *
1004  * ipc_irc_lock			conn_lock
1005  * ipc_ire_cache		conn_ire_cache
1006  *
1007  * ipc_state_flags		conn_state_flags
1008  * ipc_outgoing_ill		conn_outgoing_ill
1009  *
1010  * ipc_dontroute 		conn_dontroute
1011  * ipc_loopback 		conn_loopback
1012  * ipc_broadcast		conn_broadcast
1013  * ipc_reuseaddr		conn_reuseaddr
1014  *
1015  * ipc_multicast_loop		conn_multicast_loop
1016  * ipc_multi_router		conn_multi_router
1017  * ipc_draining 		conn_draining
1018  *
1019  * ipc_did_putbq		conn_did_putbq
1020  * ipc_unspec_src		conn_unspec_src
1021  * ipc_policy_cached		conn_policy_cached
1022  *
1023  * ipc_in_enforce_policy 	conn_in_enforce_policy
1024  * ipc_out_enforce_policy 	conn_out_enforce_policy
1025  * ipc_af_isv6			conn_af_isv6
1026  * ipc_pkt_isv6			conn_pkt_isv6
1027  *
1028  * ipc_ipv6_recvpktinfo		conn_ipv6_recvpktinfo
1029  *
1030  * ipc_ipv6_recvhoplimit	conn_ipv6_recvhoplimit
1031  * ipc_ipv6_recvhopopts		conn_ipv6_recvhopopts
1032  * ipc_ipv6_recvdstopts		conn_ipv6_recvdstopts
1033  *
1034  * ipc_ipv6_recvrthdr 		conn_ipv6_recvrthdr
1035  * ipc_ipv6_recvrtdstopts	conn_ipv6_recvrtdstopts
1036  * ipc_fully_bound		conn_fully_bound
1037  *
1038  * ipc_recvif			conn_recvif
1039  *
1040  * ipc_recvslla 		conn_recvslla
1041  * ipc_acking_unbind 		conn_acking_unbind
1042  * ipc_pad_to_bit_31 		conn_pad_to_bit_31
1043  *
1044  * ipc_nofailover_ill		conn_nofailover_ill
1045  *
1046  * ipc_proto			conn_proto
1047  * ipc_incoming_ill		conn_incoming_ill
1048  * ipc_outgoing_pill		conn_outgoing_pill
1049  * ipc_pending_ill		conn_pending_ill
1050  * ipc_unbind_mp		conn_unbind_mp
1051  * ipc_ilg			conn_ilg
1052  * ipc_ilg_allocated		conn_ilg_allocated
1053  * ipc_ilg_inuse		conn_ilg_inuse
1054  * ipc_ilg_walker_cnt		conn_ilg_walker_cnt
1055  * ipc_refcv			conn_refcv
1056  * ipc_multicast_ipif		conn_multicast_ipif
1057  * ipc_multicast_ill		conn_multicast_ill
1058  * ipc_orig_bound_ifindex	conn_orig_bound_ifindex
1059  * ipc_orig_multicast_ifindex	conn_orig_multicast_ifindex
1060  * ipc_drain_next		conn_drain_next
1061  * ipc_drain_prev		conn_drain_prev
1062  * ipc_idl			conn_idl
1063  */
1064 
1065 /*
1066  * This is used to match an inbound/outbound datagram with policy.
1067  */
1068 typedef	struct ipsec_selector {
1069 	in6_addr_t	ips_local_addr_v6;
1070 	in6_addr_t	ips_remote_addr_v6;
1071 	uint16_t	ips_local_port;
1072 	uint16_t	ips_remote_port;
1073 	uint8_t		ips_icmp_type;
1074 	uint8_t		ips_icmp_code;
1075 	uint8_t		ips_protocol;
1076 	uint8_t		ips_isv4 : 1,
1077 			ips_is_icmp_inv_acq: 1;
1078 } ipsec_selector_t;
1079 
1080 /*
1081  * Note that we put v4 addresses in the *first* 32-bit word of the
1082  * selector rather than the last to simplify the prefix match/mask code
1083  * in spd.c
1084  */
1085 #define	ips_local_addr_v4 ips_local_addr_v6.s6_addr32[0]
1086 #define	ips_remote_addr_v4 ips_remote_addr_v6.s6_addr32[0]
1087 
1088 /* Values used in IP by IPSEC Code */
1089 #define		IPSEC_OUTBOUND		B_TRUE
1090 #define		IPSEC_INBOUND		B_FALSE
1091 
1092 /*
1093  * There are two variants in policy failures. The packet may come in
1094  * secure when not needed (IPSEC_POLICY_???_NOT_NEEDED) or it may not
1095  * have the desired level of protection (IPSEC_POLICY_MISMATCH).
1096  */
1097 #define	IPSEC_POLICY_NOT_NEEDED		0
1098 #define	IPSEC_POLICY_MISMATCH		1
1099 #define	IPSEC_POLICY_AUTH_NOT_NEEDED	2
1100 #define	IPSEC_POLICY_ENCR_NOT_NEEDED	3
1101 #define	IPSEC_POLICY_SE_NOT_NEEDED	4
1102 #define	IPSEC_POLICY_MAX		5	/* Always max + 1. */
1103 
1104 /*
1105  * Folowing macro is used whenever the code does not know whether there
1106  * is a M_CTL present in the front and it needs to examine the actual mp
1107  * i.e the IP header. As a M_CTL message could be in the front, this
1108  * extracts the packet into mp and the M_CTL mp into first_mp. If M_CTL
1109  * mp is not present, both first_mp and mp point to the same message.
1110  */
1111 #define	EXTRACT_PKT_MP(mp, first_mp, mctl_present)	\
1112 	(first_mp) = (mp);				\
1113 	if ((mp)->b_datap->db_type == M_CTL) {		\
1114 		(mp) = (mp)->b_cont;			\
1115 		(mctl_present) = B_TRUE;		\
1116 	} else {					\
1117 		(mctl_present) = B_FALSE;		\
1118 	}
1119 
1120 /*
1121  * Check with IPSEC inbound policy if
1122  *
1123  * 1) per-socket policy is present - indicated by conn_in_enforce_policy.
1124  * 2) Or if we have not cached policy on the conn and the global policy is
1125  *    non-empty.
1126  */
1127 #define	CONN_INBOUND_POLICY_PRESENT(connp, ipss)	\
1128 	((connp)->conn_in_enforce_policy ||		\
1129 	(!((connp)->conn_policy_cached) && 		\
1130 	(ipss)->ipsec_inbound_v4_policy_present))
1131 
1132 #define	CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)	\
1133 	((connp)->conn_in_enforce_policy ||		\
1134 	(!(connp)->conn_policy_cached &&		\
1135 	(ipss)->ipsec_inbound_v6_policy_present))
1136 
1137 #define	CONN_OUTBOUND_POLICY_PRESENT(connp, ipss)	\
1138 	((connp)->conn_out_enforce_policy ||		\
1139 	(!((connp)->conn_policy_cached) &&		\
1140 	(ipss)->ipsec_outbound_v4_policy_present))
1141 
1142 #define	CONN_OUTBOUND_POLICY_PRESENT_V6(connp, ipss)	\
1143 	((connp)->conn_out_enforce_policy ||		\
1144 	(!(connp)->conn_policy_cached &&		\
1145 	(ipss)->ipsec_outbound_v6_policy_present))
1146 
1147 /*
1148  * Information cached in IRE for upper layer protocol (ULP).
1149  *
1150  * Notice that ire_max_frag is not included in the iulp_t structure, which
1151  * it may seem that it should.  But ire_max_frag cannot really be cached.  It
1152  * is fixed for each interface.  For MTU found by PMTUd, we may want to cache
1153  * it.  But currently, we do not do that.
1154  */
1155 typedef struct iulp_s {
1156 	boolean_t	iulp_set;	/* Is any metric set? */
1157 	uint32_t	iulp_ssthresh;	/* Slow start threshold (TCP). */
1158 	clock_t		iulp_rtt;	/* Guestimate in millisecs. */
1159 	clock_t		iulp_rtt_sd;	/* Cached value of RTT variance. */
1160 	uint32_t	iulp_spipe;	/* Send pipe size. */
1161 	uint32_t	iulp_rpipe;	/* Receive pipe size. */
1162 	uint32_t	iulp_rtomax;	/* Max round trip timeout. */
1163 	uint32_t	iulp_sack;	/* Use SACK option (TCP)? */
1164 	uint32_t
1165 		iulp_tstamp_ok : 1,	/* Use timestamp option (TCP)? */
1166 		iulp_wscale_ok : 1,	/* Use window scale option (TCP)? */
1167 		iulp_ecn_ok : 1,	/* Enable ECN (for TCP)? */
1168 		iulp_pmtud_ok : 1,	/* Enable PMTUd? */
1169 
1170 		iulp_not_used : 28;
1171 } iulp_t;
1172 
1173 /* Zero iulp_t. */
1174 extern const iulp_t ire_uinfo_null;
1175 
1176 /*
1177  * The conn drain list structure.
1178  * The list is protected by idl_lock. Each conn_t inserted in the list
1179  * points back at this idl_t using conn_idl. IP primes the draining of the
1180  * conns queued in these lists, by qenabling the 1st conn of each list. This
1181  * occurs when STREAMS backenables ip_wsrv on the IP module. Each conn instance
1182  * of ip_wsrv successively qenables the next conn in the list.
1183  * idl_lock protects all other members of idl_t and conn_drain_next
1184  * and conn_drain_prev of conn_t. The conn_lock protects IPCF_DRAIN_DISABLED
1185  * flag of the conn_t and conn_idl.
1186  */
1187 typedef struct idl_s {
1188 	conn_t		*idl_conn;		/* Head of drain list */
1189 	kmutex_t	idl_lock;		/* Lock for this list */
1190 	conn_t		*idl_conn_draining;	/* conn that is draining */
1191 	uint32_t
1192 		idl_repeat : 1,			/* Last conn must re-enable */
1193 						/* drain list again */
1194 		idl_unused : 31;
1195 } idl_t;
1196 
1197 #define	CONN_DRAIN_LIST_LOCK(connp)	(&((connp)->conn_idl->idl_lock))
1198 /*
1199  * Interface route structure which holds the necessary information to recreate
1200  * routes that are tied to an interface (namely where ire_ipif != NULL).
1201  * These routes which were initially created via a routing socket or via the
1202  * SIOCADDRT ioctl may be gateway routes (RTF_GATEWAY being set) or may be
1203  * traditional interface routes.  When an interface comes back up after being
1204  * marked down, this information will be used to recreate the routes.  These
1205  * are part of an mblk_t chain that hangs off of the IPIF (ipif_saved_ire_mp).
1206  */
1207 typedef struct ifrt_s {
1208 	ushort_t	ifrt_type;		/* Type of IRE */
1209 	in6_addr_t	ifrt_v6addr;		/* Address IRE represents. */
1210 	in6_addr_t	ifrt_v6gateway_addr;	/* Gateway if IRE_OFFSUBNET */
1211 	in6_addr_t	ifrt_v6src_addr;	/* Src addr if RTF_SETSRC */
1212 	in6_addr_t	ifrt_v6mask;		/* Mask for matching IRE. */
1213 	uint32_t	ifrt_flags;		/* flags related to route */
1214 	uint_t		ifrt_max_frag;		/* MTU (next hop or path). */
1215 	iulp_t		ifrt_iulp_info;		/* Cached IRE ULP info. */
1216 } ifrt_t;
1217 
1218 #define	ifrt_addr		V4_PART_OF_V6(ifrt_v6addr)
1219 #define	ifrt_gateway_addr	V4_PART_OF_V6(ifrt_v6gateway_addr)
1220 #define	ifrt_src_addr		V4_PART_OF_V6(ifrt_v6src_addr)
1221 #define	ifrt_mask		V4_PART_OF_V6(ifrt_v6mask)
1222 
1223 /* Number of IP addresses that can be hosted on a physical interface */
1224 #define	MAX_ADDRS_PER_IF	8192
1225 /*
1226  * Number of Source addresses to be considered for source address
1227  * selection. Used by ipif_select_source[_v6].
1228  */
1229 #define	MAX_IPIF_SELECT_SOURCE	50
1230 
1231 #ifdef IP_DEBUG
1232 /*
1233  * Trace refholds and refreles for debugging.
1234  */
1235 #define	TR_STACK_DEPTH	14
1236 typedef struct tr_buf_s {
1237 	int	tr_depth;
1238 	clock_t	tr_time;
1239 	pc_t	tr_stack[TR_STACK_DEPTH];
1240 } tr_buf_t;
1241 
1242 typedef struct th_trace_s {
1243 	int		th_refcnt;
1244 	uint_t		th_trace_lastref;
1245 	kthread_t	*th_id;
1246 #define	TR_BUF_MAX	38
1247 	tr_buf_t	th_trbuf[TR_BUF_MAX];
1248 } th_trace_t;
1249 
1250 typedef struct th_hash_s {
1251 	list_node_t	thh_link;
1252 	mod_hash_t	*thh_hash;
1253 	ip_stack_t	*thh_ipst;
1254 } th_hash_t;
1255 #endif
1256 
1257 /* The following are ipif_state_flags */
1258 #define	IPIF_CONDEMNED		0x1	/* The ipif is being removed */
1259 #define	IPIF_CHANGING		0x2	/* A critcal ipif field is changing */
1260 #define	IPIF_MOVING		0x8	/* The ipif is being moved */
1261 #define	IPIF_SET_LINKLOCAL	0x10	/* transient flag during bringup */
1262 #define	IPIF_ZERO_SOURCE	0x20	/* transient flag during bringup */
1263 
1264 /* IP interface structure, one per local address */
1265 typedef struct ipif_s {
1266 	struct	ipif_s	*ipif_next;
1267 	struct	ill_s	*ipif_ill;	/* Back pointer to our ill */
1268 	int	ipif_id;		/* Logical unit number */
1269 	uint_t	ipif_mtu;		/* Starts at ipif_ill->ill_max_frag */
1270 	uint_t	ipif_saved_mtu;		/* Save of mtu during ipif_move() */
1271 	in6_addr_t ipif_v6lcl_addr;	/* Local IP address for this if. */
1272 	in6_addr_t ipif_v6src_addr;	/* Source IP address for this if. */
1273 	in6_addr_t ipif_v6subnet;	/* Subnet prefix for this if. */
1274 	in6_addr_t ipif_v6net_mask;	/* Net mask for this interface. */
1275 	in6_addr_t ipif_v6brd_addr;	/* Broadcast addr for this interface. */
1276 	in6_addr_t ipif_v6pp_dst_addr;	/* Point-to-point dest address. */
1277 	uint64_t ipif_flags;		/* Interface flags. */
1278 	uint_t	ipif_metric;		/* BSD if metric, for compatibility. */
1279 	uint_t	ipif_ire_type;		/* IRE_LOCAL or IRE_LOOPBACK */
1280 	mblk_t	*ipif_arp_del_mp;	/* Allocated at time arp comes up, to */
1281 					/* prevent awkward out of mem */
1282 					/* condition later */
1283 	mblk_t	*ipif_saved_ire_mp;	/* Allocated for each extra */
1284 					/* IRE_IF_NORESOLVER/IRE_IF_RESOLVER */
1285 					/* on this interface so that they */
1286 					/* can survive ifconfig down. */
1287 	kmutex_t ipif_saved_ire_lock;	/* Protects ipif_saved_ire_mp */
1288 
1289 	mrec_t	*ipif_igmp_rpt;		/* List of group memberships which */
1290 					/* will be reported on.  Used when */
1291 					/* handling an igmp timeout.	   */
1292 
1293 	/*
1294 	 * The packet counts in the ipif contain the sum of the
1295 	 * packet counts in dead IREs that were affiliated with
1296 	 * this ipif.
1297 	 */
1298 	uint_t	ipif_fo_pkt_count;	/* Forwarded thru our dead IREs */
1299 	uint_t	ipif_ib_pkt_count;	/* Inbound packets for our dead IREs */
1300 	uint_t	ipif_ob_pkt_count;	/* Outbound packets to our dead IREs */
1301 	/* Exclusive bit fields, protected by ipsq_t */
1302 	unsigned int
1303 		ipif_multicast_up : 1,	/* We have joined the allhosts group */
1304 		ipif_replace_zero : 1,	/* Replacement for zero */
1305 		ipif_was_up : 1,	/* ipif was up before */
1306 		ipif_addr_ready : 1,	/* DAD is done */
1307 
1308 		ipif_was_dup : 1,	/* DAD had failed */
1309 		ipif_pad_to_31 : 27;
1310 
1311 	int	ipif_orig_ifindex;	/* ifindex before SLIFFAILOVER */
1312 	uint_t	ipif_seqid;		/* unique index across all ills */
1313 	uint_t	ipif_orig_ipifid;	/* ipif_id before SLIFFAILOVER */
1314 	uint_t	ipif_state_flags;	/* See IPIF_* flag defs above */
1315 	uint_t	ipif_refcnt;		/* active consistent reader cnt */
1316 
1317 	/* Number of ire's and ilm's referencing this ipif */
1318 	uint_t	ipif_ire_cnt;
1319 	uint_t	ipif_ilm_cnt;
1320 
1321 	uint_t  ipif_saved_ire_cnt;
1322 	zoneid_t ipif_zoneid;		/* zone ID number */
1323 	timeout_id_t ipif_recovery_id;	/* Timer for DAD recovery */
1324 	boolean_t ipif_trace_disable;	/* True when alloc fails */
1325 } ipif_t;
1326 
1327 /*
1328  * IPIF_FREE_OK() means that there are no incoming references
1329  * to the ipif. Incoming refs would prevent the ipif from being freed.
1330  */
1331 #define	IPIF_FREE_OK(ipif)	\
1332 	((ipif)->ipif_ire_cnt == 0 && (ipif)->ipif_ilm_cnt == 0)
1333 /*
1334  * IPIF_DOWN_OK() determines whether the incoming pointer reference counts
1335  * would permit the ipif to be considered quiescent. In order for
1336  * an ipif or ill to be considered quiescent, the ire and nce references
1337  * to that ipif/ill must be zero.
1338  *
1339  * We do not require the ilm references to go to zero for quiescence
1340  * because the quiescence checks are done to ensure that
1341  * outgoing packets do not use addresses from the ipif/ill after it
1342  * has been marked down, and incoming packets to addresses on a
1343  * queiscent interface are rejected. This implies that all the
1344  * ire/nce's using that source address need to be deleted and future
1345  * creation of any ires using that source address must be prevented.
1346  * Similarly incoming unicast packets destined to the 'down' address
1347  * will not be accepted once that ire is gone. However incoming
1348  * multicast packets are not destined to the downed address.
1349  * They are only related to the ill in question. Furthermore
1350  * the current API behavior allows applications to join or leave
1351  * multicast groups, i.e., IP_ADD_MEMBERSHIP / LEAVE_MEMBERSHIP, using a
1352  * down address. Therefore the ilm references are not included in
1353  * the _DOWN_OK macros.
1354  */
1355 #define	IPIF_DOWN_OK(ipif)		((ipif)->ipif_ire_cnt == 0)
1356 
1357 /*
1358  * The following table lists the protection levels of the various members
1359  * of the ipif_t. The following notation is used.
1360  *
1361  * Write once - Written to only once at the time of bringing up
1362  * the interface and can be safely read after the bringup without any lock.
1363  *
1364  * ipsq - Need to execute in the ipsq to perform the indicated access.
1365  *
1366  * ill_lock - Need to hold this mutex to perform the indicated access.
1367  *
1368  * ill_g_lock - Need to hold this rw lock as reader/writer for read access or
1369  * write access respectively.
1370  *
1371  * down ill - Written to only when the ill is down (i.e all ipifs are down)
1372  * up ill - Read only when the ill is up (i.e. at least 1 ipif is up)
1373  *
1374  *		 Table of ipif_t members and their protection
1375  *
1376  * ipif_next		ipsq + ill_lock +	ipsq OR ill_lock OR
1377  *			ill_g_lock		ill_g_lock
1378  * ipif_ill		ipsq + down ipif	write once
1379  * ipif_id		ipsq + down ipif	write once
1380  * ipif_mtu		ipsq
1381  * ipif_v6lcl_addr	ipsq + down ipif	up ipif
1382  * ipif_v6src_addr	ipsq + down ipif	up ipif
1383  * ipif_v6subnet	ipsq + down ipif	up ipif
1384  * ipif_v6net_mask	ipsq + down ipif	up ipif
1385  *
1386  * ipif_v6brd_addr
1387  * ipif_v6pp_dst_addr
1388  * ipif_flags		ill_lock		ill_lock
1389  * ipif_metric
1390  * ipif_ire_type	ipsq + down ill		up ill
1391  *
1392  * ipif_arp_del_mp	ipsq			ipsq
1393  * ipif_saved_ire_mp	ipif_saved_ire_lock	ipif_saved_ire_lock
1394  * ipif_igmp_rpt	ipsq			ipsq
1395  *
1396  * ipif_fo_pkt_count	Approx
1397  * ipif_ib_pkt_count	Approx
1398  * ipif_ob_pkt_count	Approx
1399  *
1400  * bit fields		ill_lock		ill_lock
1401  *
1402  * ipif_orig_ifindex	ipsq			None
1403  * ipif_orig_ipifid	ipsq			None
1404  * ipif_seqid		ipsq			Write once
1405  *
1406  * ipif_state_flags	ill_lock		ill_lock
1407  * ipif_refcnt		ill_lock		ill_lock
1408  * ipif_ire_cnt		ill_lock		ill_lock
1409  * ipif_ilm_cnt		ill_lock		ill_lock
1410  * ipif_saved_ire_cnt
1411  */
1412 
1413 #define	IP_TR_HASH(tid)	((((uintptr_t)tid) >> 6) & (IP_TR_HASH_MAX - 1))
1414 
1415 #ifdef DEBUG
1416 #define	IPIF_TRACE_REF(ipif)	ipif_trace_ref(ipif)
1417 #define	ILL_TRACE_REF(ill)	ill_trace_ref(ill)
1418 #define	IPIF_UNTRACE_REF(ipif)	ipif_untrace_ref(ipif)
1419 #define	ILL_UNTRACE_REF(ill)	ill_untrace_ref(ill)
1420 #else
1421 #define	IPIF_TRACE_REF(ipif)
1422 #define	ILL_TRACE_REF(ill)
1423 #define	IPIF_UNTRACE_REF(ipif)
1424 #define	ILL_UNTRACE_REF(ill)
1425 #endif
1426 
1427 /* IPv4 compatibility macros */
1428 #define	ipif_lcl_addr		V4_PART_OF_V6(ipif_v6lcl_addr)
1429 #define	ipif_src_addr		V4_PART_OF_V6(ipif_v6src_addr)
1430 #define	ipif_subnet		V4_PART_OF_V6(ipif_v6subnet)
1431 #define	ipif_net_mask		V4_PART_OF_V6(ipif_v6net_mask)
1432 #define	ipif_brd_addr		V4_PART_OF_V6(ipif_v6brd_addr)
1433 #define	ipif_pp_dst_addr	V4_PART_OF_V6(ipif_v6pp_dst_addr)
1434 
1435 /* Macros for easy backreferences to the ill. */
1436 #define	ipif_wq			ipif_ill->ill_wq
1437 #define	ipif_rq			ipif_ill->ill_rq
1438 #define	ipif_net_type		ipif_ill->ill_net_type
1439 #define	ipif_ipif_up_count	ipif_ill->ill_ipif_up_count
1440 #define	ipif_type		ipif_ill->ill_type
1441 #define	ipif_isv6		ipif_ill->ill_isv6
1442 
1443 #define	SIOCLIFADDR_NDX 112	/* ndx of SIOCLIFADDR in the ndx ioctl table */
1444 
1445 /*
1446  * mode value for ip_ioctl_finish for finishing an ioctl
1447  */
1448 #define	CONN_CLOSE	1		/* No mi_copy */
1449 #define	COPYOUT		2		/* do an mi_copyout if needed */
1450 #define	NO_COPYOUT	3		/* do an mi_copy_done */
1451 #define	IPI2MODE(ipi)	((ipi)->ipi_flags & IPI_GET_CMD ? COPYOUT : NO_COPYOUT)
1452 
1453 /*
1454  * The IP-MT design revolves around the serialization object ipsq_t.
1455  * It is associated with an IPMP group. If IPMP is not enabled, there is
1456  * 1 ipsq_t per phyint. Eg. an ipsq_t would cover both hme0's IPv4 stream
1457  *
1458  * ipsq_lock protects
1459  *	ipsq_reentry_cnt, ipsq_writer, ipsq_xopq_mphead, ipsq_xopq_mptail,
1460  *	ipsq_mphead, ipsq_mptail, ipsq_split
1461  *
1462  *	ipsq_pending_ipif, ipsq_current_ipif, ipsq_pending_mp, ipsq_flags,
1463  *	ipsq_waitfor
1464  *
1465  * The fields in the last line above below are set mostly by a writer thread
1466  * But there is an exception in the last call to ipif_ill_refrele_tail which
1467  * could also race with a conn close which could be cleaning up the
1468  * fields. So we choose to protect using ipsq_lock instead of depending on
1469  * the property of the writer.
1470  * ill_g_lock protects
1471  *	ipsq_refs, ipsq_phyint_list
1472  */
1473 typedef struct ipsq_s {
1474 	kmutex_t ipsq_lock;
1475 	int	ipsq_reentry_cnt;
1476 	kthread_t *ipsq_writer;		/* current owner (thread id) */
1477 	int	ipsq_flags;
1478 	mblk_t	*ipsq_xopq_mphead;	/* list of excl ops mostly ioctls */
1479 	mblk_t	*ipsq_xopq_mptail;
1480 	mblk_t	*ipsq_mphead;		/* msgs on ipsq linked thru b_next */
1481 	mblk_t	*ipsq_mptail;		/* msgs on ipsq linked thru b_next */
1482 	int	ipsq_current_ioctl;	/* current ioctl, or 0 if no ioctl */
1483 	boolean_t ipsq_current_done; 	/* is the current op done? */
1484 	ipif_t	*ipsq_current_ipif;	/* ipif associated with current op */
1485 	ipif_t	*ipsq_pending_ipif;	/* ipif associated w. ipsq_pending_mp */
1486 	mblk_t	*ipsq_pending_mp;	/* current ioctl mp while waiting for */
1487 					/* response from another module */
1488 	struct	ipsq_s	*ipsq_next;	/* list of all syncq's (ipsq_g_list) */
1489 	uint_t		ipsq_refs;	/* Number of phyints on this ipsq */
1490 	struct phyint	*ipsq_phyint_list; /* List of phyints on this ipsq */
1491 	boolean_t	ipsq_split;	/* ipsq may need to be split */
1492 	int		ipsq_waitfor;	/* Values encoded below */
1493 	char		ipsq_name[LIFNAMSIZ+1];	/* same as phyint_groupname */
1494 	ip_stack_t	*ipsq_ipst;	/* Does not have a netstack_hold */
1495 
1496 #ifdef DEBUG
1497 	int		ipsq_depth;	/* debugging aid */
1498 #define	IPSQ_STACK_DEPTH	15
1499 	pc_t		ipsq_stack[IPSQ_STACK_DEPTH];	/* debugging aid */
1500 #endif
1501 } ipsq_t;
1502 
1503 /* ipsq_flags */
1504 #define	IPSQ_GROUP	0x1	/* This ipsq belongs to an IPMP group */
1505 
1506 /*
1507  * ipsq_waitfor:
1508  *
1509  * IPIF_DOWN	1	ipif_down waiting for refcnts to drop
1510  * ILL_DOWN	2	ill_down waiting for refcnts to drop
1511  * IPIF_FREE	3	ipif_free waiting for refcnts to drop
1512  * ILL_FREE	4	ill unplumb waiting for refcnts to drop
1513  * ILL_MOVE_OK	5	failover waiting for refcnts to drop
1514  */
1515 
1516 enum { IPIF_DOWN = 1, ILL_DOWN, IPIF_FREE, ILL_FREE, ILL_MOVE_OK };
1517 
1518 /* Flags passed to ipsq_try_enter */
1519 #define	CUR_OP 0		/* Current ioctl continuing again */
1520 #define	NEW_OP 1		/* New ioctl starting afresh */
1521 
1522 /*
1523  * phyint represents state that is common to both IPv4 and IPv6 interfaces.
1524  * There is a separate ill_t representing IPv4 and IPv6 which has a
1525  * backpointer to the phyint structure for accessing common state.
1526  *
1527  * NOTE : It just stores the group name as there is only one name for
1528  *	  IPv4 and IPv6 i.e it is a underlying link property. Actually
1529  *        IPv4 and IPv6 ill are grouped together when their phyints have
1530  *        the same name.
1531  */
1532 typedef struct phyint {
1533 	struct ill_s	*phyint_illv4;
1534 	struct ill_s	*phyint_illv6;
1535 	uint_t		phyint_ifindex;		/* SIOCLSLIFINDEX */
1536 	char		*phyint_groupname;	/* SIOCSLIFGROUPNAME */
1537 	uint_t		phyint_groupname_len;
1538 	uint64_t	phyint_flags;
1539 	avl_node_t	phyint_avl_by_index;	/* avl tree by index */
1540 	avl_node_t	phyint_avl_by_name;	/* avl tree by name */
1541 	kmutex_t	phyint_lock;
1542 	struct ipsq_s	*phyint_ipsq;		/* back pointer to ipsq */
1543 	struct phyint	*phyint_ipsq_next;	/* phyint list on this ipsq */
1544 	/* Once Clearview IPMP is added the follow two fields can be removed */
1545 	uint_t		phyint_group_ifindex;	/* index assigned to group */
1546 	uint_t		phyint_hook_ifindex;	/* index used with neti/hook */
1547 } phyint_t;
1548 
1549 #define	CACHE_ALIGN_SIZE 64
1550 
1551 #define	CACHE_ALIGN(align_struct)	P2ROUNDUP(sizeof (struct align_struct),\
1552 							CACHE_ALIGN_SIZE)
1553 struct _phyint_list_s_ {
1554 	avl_tree_t	phyint_list_avl_by_index;	/* avl tree by index */
1555 	avl_tree_t	phyint_list_avl_by_name;	/* avl tree by name */
1556 };
1557 
1558 typedef union phyint_list_u {
1559 	struct	_phyint_list_s_ phyint_list_s;
1560 	char	phyint_list_filler[CACHE_ALIGN(_phyint_list_s_)];
1561 } phyint_list_t;
1562 
1563 #define	phyint_list_avl_by_index	phyint_list_s.phyint_list_avl_by_index
1564 #define	phyint_list_avl_by_name		phyint_list_s.phyint_list_avl_by_name
1565 /*
1566  * ILL groups. We group ills,
1567  *
1568  * - if the ills have the same group name. (New way)
1569  *
1570  * ill_group locking notes:
1571  *
1572  * illgrp_lock protects ill_grp_ill_schednext.
1573  *
1574  * ill_g_lock protects ill_grp_next, illgrp_ill, illgrp_ill_count.
1575  * Holding ill_g_lock freezes the memberships of ills in IPMP groups.
1576  * It also freezes the global list of ills and all ipifs in all ills.
1577  *
1578  * To remove an ipif from the linked list of ipifs of that ill ipif_free_tail
1579  * holds both ill_g_lock, and ill_lock. Similarly to remove an ill from the
1580  * global list of ills, ill_glist_delete() holds ill_g_lock as writer.
1581  * This simplifies things for ipif_select_source, illgrp_scheduler etc.
1582  * that need to walk the members of an illgrp. They just hold ill_g_lock
1583  * as reader to do the walk.
1584  *
1585  */
1586 typedef	struct ill_group {
1587 	kmutex_t	illgrp_lock;
1588 	struct ill_group *illgrp_next;		/* Next ill_group */
1589 	struct ill_s	*illgrp_ill_schednext;	/* Next ill to be scheduled */
1590 	struct ill_s	*illgrp_ill;		/* First ill in the group */
1591 	int		illgrp_ill_count;
1592 } ill_group_t;
1593 
1594 /*
1595  * Fragmentation hash bucket
1596  */
1597 typedef struct ipfb_s {
1598 	struct ipf_s	*ipfb_ipf;	/* List of ... */
1599 	size_t		ipfb_count;	/* Count of bytes used by frag(s) */
1600 	kmutex_t	ipfb_lock;	/* Protect all ipf in list */
1601 	uint_t		ipfb_frag_pkts; /* num of distinct fragmented pkts */
1602 } ipfb_t;
1603 
1604 /*
1605  * IRE bucket structure. Usually there is an array of such structures,
1606  * each pointing to a linked list of ires. irb_refcnt counts the number
1607  * of walkers of a given hash bucket. Usually the reference count is
1608  * bumped up if the walker wants no IRES to be DELETED while walking the
1609  * list. Bumping up does not PREVENT ADDITION. This allows walking a given
1610  * hash bucket without stumbling up on a free pointer.
1611  *
1612  * irb_t structures in ip_ftable are dynamically allocated and freed.
1613  * In order to identify the irb_t structures that can be safely kmem_free'd
1614  * we need to ensure that
1615  *  - the irb_refcnt is quiescent, indicating no other walkers,
1616  *  - no other threads or ire's are holding references to the irb,
1617  *	i.e., irb_nire == 0,
1618  *  - there are no active ire's in the bucket, i.e., irb_ire_cnt == 0
1619  */
1620 typedef struct irb {
1621 	struct ire_s	*irb_ire;	/* First ire in this bucket */
1622 					/* Should be first in this struct */
1623 	krwlock_t	irb_lock;	/* Protect this bucket */
1624 	uint_t		irb_refcnt;	/* Protected by irb_lock */
1625 	uchar_t		irb_marks;	/* CONDEMNED ires in this bucket ? */
1626 #define	IRB_MARK_CONDEMNED	0x0001
1627 #define	IRB_MARK_FTABLE		0x0002
1628 	uint_t		irb_ire_cnt;	/* Num of active IRE in this bucket */
1629 	uint_t		irb_tmp_ire_cnt; /* Num of temporary IRE */
1630 	struct ire_s	*irb_rr_origin;	/* origin for round-robin */
1631 	int		irb_nire;	/* Num of ftable ire's that ref irb */
1632 	ip_stack_t	*irb_ipst;	/* Does not have a netstack_hold */
1633 } irb_t;
1634 
1635 #define	IRB2RT(irb)	(rt_t *)((caddr_t)(irb) - offsetof(rt_t, rt_irb))
1636 
1637 /* The following are return values of ip_xmit_v4() */
1638 typedef enum {
1639 	SEND_PASSED = 0,	 /* sent packet out on wire */
1640 	SEND_FAILED,	 /* sending of packet failed */
1641 	LOOKUP_IN_PROGRESS, /* ire cache found, ARP resolution in progress */
1642 	LLHDR_RESLV_FAILED  /* macaddr resl of onlink dst or nexthop failed */
1643 } ipxmit_state_t;
1644 
1645 #define	IP_V4_G_HEAD	0
1646 #define	IP_V6_G_HEAD	1
1647 
1648 #define	MAX_G_HEADS	2
1649 
1650 /*
1651  * unpadded ill_if structure
1652  */
1653 struct 	_ill_if_s_ {
1654 	union ill_if_u	*illif_next;
1655 	union ill_if_u	*illif_prev;
1656 	avl_tree_t	illif_avl_by_ppa;	/* AVL tree sorted on ppa */
1657 	vmem_t		*illif_ppa_arena;	/* ppa index space */
1658 	uint16_t	illif_mcast_v1;		/* hints for		  */
1659 	uint16_t	illif_mcast_v2;		/* [igmp|mld]_slowtimo	  */
1660 	int		illif_name_len;		/* name length */
1661 	char		illif_name[LIFNAMSIZ];	/* name of interface type */
1662 };
1663 
1664 /* cache aligned ill_if structure */
1665 typedef union 	ill_if_u {
1666 	struct  _ill_if_s_ ill_if_s;
1667 	char 	illif_filler[CACHE_ALIGN(_ill_if_s_)];
1668 } ill_if_t;
1669 
1670 #define	illif_next		ill_if_s.illif_next
1671 #define	illif_prev		ill_if_s.illif_prev
1672 #define	illif_avl_by_ppa	ill_if_s.illif_avl_by_ppa
1673 #define	illif_ppa_arena		ill_if_s.illif_ppa_arena
1674 #define	illif_mcast_v1		ill_if_s.illif_mcast_v1
1675 #define	illif_mcast_v2		ill_if_s.illif_mcast_v2
1676 #define	illif_name		ill_if_s.illif_name
1677 #define	illif_name_len		ill_if_s.illif_name_len
1678 
1679 typedef struct ill_walk_context_s {
1680 	int	ctx_current_list; /* current list being searched */
1681 	int	ctx_last_list;	 /* last list to search */
1682 } ill_walk_context_t;
1683 
1684 /*
1685  * ill_g_heads structure, one for IPV4 and one for IPV6
1686  */
1687 struct _ill_g_head_s_ {
1688 	ill_if_t	*ill_g_list_head;
1689 	ill_if_t	*ill_g_list_tail;
1690 };
1691 
1692 typedef union ill_g_head_u {
1693 	struct _ill_g_head_s_ ill_g_head_s;
1694 	char	ill_g_head_filler[CACHE_ALIGN(_ill_g_head_s_)];
1695 } ill_g_head_t;
1696 
1697 #define	ill_g_list_head	ill_g_head_s.ill_g_list_head
1698 #define	ill_g_list_tail	ill_g_head_s.ill_g_list_tail
1699 
1700 #define	IP_V4_ILL_G_LIST(ipst)	\
1701 	(ipst)->ips_ill_g_heads[IP_V4_G_HEAD].ill_g_list_head
1702 #define	IP_V6_ILL_G_LIST(ipst)	\
1703 	(ipst)->ips_ill_g_heads[IP_V6_G_HEAD].ill_g_list_head
1704 #define	IP_VX_ILL_G_LIST(i, ipst)	\
1705 	(ipst)->ips_ill_g_heads[i].ill_g_list_head
1706 
1707 #define	ILL_START_WALK_V4(ctx_ptr, ipst)	\
1708 	ill_first(IP_V4_G_HEAD, IP_V4_G_HEAD, ctx_ptr, ipst)
1709 #define	ILL_START_WALK_V6(ctx_ptr, ipst)	\
1710 	ill_first(IP_V6_G_HEAD, IP_V6_G_HEAD, ctx_ptr, ipst)
1711 #define	ILL_START_WALK_ALL(ctx_ptr, ipst)	\
1712 	ill_first(MAX_G_HEADS, MAX_G_HEADS, ctx_ptr, ipst)
1713 
1714 /*
1715  * Capabilities, possible flags for ill_capabilities.
1716  */
1717 
1718 #define	ILL_CAPAB_AH		0x01		/* IPsec AH acceleration */
1719 #define	ILL_CAPAB_ESP		0x02		/* IPsec ESP acceleration */
1720 #define	ILL_CAPAB_MDT		0x04		/* Multidata Transmit */
1721 #define	ILL_CAPAB_HCKSUM	0x08		/* Hardware checksumming */
1722 #define	ILL_CAPAB_ZEROCOPY	0x10		/* Zero-copy */
1723 #define	ILL_CAPAB_POLL		0x20		/* Polling Toggle */
1724 #define	ILL_CAPAB_SOFT_RING	0x40		/* Soft_Ring capability */
1725 #define	ILL_CAPAB_LSO		0x80		/* Large Segment Offload */
1726 
1727 /*
1728  * Per-ill Multidata Transmit capabilities.
1729  */
1730 typedef struct ill_mdt_capab_s ill_mdt_capab_t;
1731 
1732 /*
1733  * Per-ill IPsec capabilities.
1734  */
1735 typedef struct ill_ipsec_capab_s ill_ipsec_capab_t;
1736 
1737 /*
1738  * Per-ill Hardware Checksumming capbilities.
1739  */
1740 typedef struct ill_hcksum_capab_s ill_hcksum_capab_t;
1741 
1742 /*
1743  * Per-ill Zero-copy capabilities.
1744  */
1745 typedef struct ill_zerocopy_capab_s ill_zerocopy_capab_t;
1746 
1747 /*
1748  * Per-ill Polling/soft ring capbilities.
1749  */
1750 typedef struct ill_dls_capab_s ill_dls_capab_t;
1751 
1752 /*
1753  * Per-ill polling resource map.
1754  */
1755 typedef struct ill_rx_ring ill_rx_ring_t;
1756 
1757 /*
1758  * Per-ill Large Segment Offload capabilities.
1759  */
1760 typedef struct ill_lso_capab_s ill_lso_capab_t;
1761 
1762 /* The following are ill_state_flags */
1763 #define	ILL_LL_SUBNET_PENDING	0x01	/* Waiting for DL_INFO_ACK from drv */
1764 #define	ILL_CONDEMNED		0x02	/* No more new ref's to the ILL */
1765 #define	ILL_CHANGING		0x04	/* ILL not globally visible */
1766 #define	ILL_DL_UNBIND_IN_PROGRESS	0x08	/* UNBIND_REQ is sent */
1767 #define	ILL_SOFT_RING_ASSIGN	0x10	/* Making soft ring assignment */
1768 
1769 /* Is this an ILL whose source address is used by other ILL's ? */
1770 #define	IS_USESRC_ILL(ill)			\
1771 	(((ill)->ill_usesrc_ifindex == 0) &&	\
1772 	((ill)->ill_usesrc_grp_next != NULL))
1773 
1774 /* Is this a client/consumer of the usesrc ILL ? */
1775 #define	IS_USESRC_CLI_ILL(ill)			\
1776 	(((ill)->ill_usesrc_ifindex != 0) &&	\
1777 	((ill)->ill_usesrc_grp_next != NULL))
1778 
1779 /* Is this an virtual network interface (vni) ILL ? */
1780 #define	IS_VNI(ill)							     \
1781 	(((ill) != NULL) &&						     \
1782 	(((ill)->ill_phyint->phyint_flags & (PHYI_LOOPBACK|PHYI_VIRTUAL)) == \
1783 	PHYI_VIRTUAL))
1784 
1785 /* Is this a loopback ILL? */
1786 #define	IS_LOOPBACK(ill) \
1787 	((ill)->ill_phyint->phyint_flags & PHYI_LOOPBACK)
1788 
1789 /*
1790  * IP Lower level Structure.
1791  * Instance data structure in ip_open when there is a device below us.
1792  */
1793 typedef struct ill_s {
1794 	ill_if_t *ill_ifptr;		/* pointer to interface type */
1795 	queue_t	*ill_rq;		/* Read queue. */
1796 	queue_t	*ill_wq;		/* Write queue. */
1797 
1798 	int	ill_error;		/* Error value sent up by device. */
1799 
1800 	ipif_t	*ill_ipif;		/* Interface chain for this ILL. */
1801 
1802 	uint_t	ill_ipif_up_count;	/* Number of IPIFs currently up. */
1803 	uint_t	ill_max_frag;		/* Max IDU from DLPI. */
1804 	char	*ill_name;		/* Our name. */
1805 	uint_t	ill_ipif_dup_count;	/* Number of duplicate addresses. */
1806 	uint_t	ill_name_length;	/* Name length, incl. terminator. */
1807 	char	*ill_ndd_name;		/* Name + ":ip?_forwarding" for NDD. */
1808 	uint_t	ill_net_type;		/* IRE_IF_RESOLVER/IRE_IF_NORESOLVER. */
1809 	/*
1810 	 * Physical Point of Attachment num.  If DLPI style 1 provider
1811 	 * then this is derived from the devname.
1812 	 */
1813 	uint_t	ill_ppa;
1814 	t_uscalar_t	ill_sap;
1815 	t_scalar_t	ill_sap_length;	/* Including sign (for position) */
1816 	uint_t	ill_phys_addr_length;	/* Excluding the sap. */
1817 	uint_t	ill_bcast_addr_length;	/* Only set when the DL provider */
1818 					/* supports broadcast. */
1819 	t_uscalar_t	ill_mactype;
1820 	uint8_t	*ill_frag_ptr;		/* Reassembly state. */
1821 	timeout_id_t ill_frag_timer_id; /* timeout id for the frag timer */
1822 	ipfb_t	*ill_frag_hash_tbl;	/* Fragment hash list head. */
1823 	ipif_t	*ill_pending_ipif;	/* IPIF waiting for DL operation. */
1824 
1825 	ilm_t	*ill_ilm;		/* Multicast membership for ill */
1826 	uint_t	ill_global_timer;	/* for IGMPv3/MLDv2 general queries */
1827 	int	ill_mcast_type;		/* type of router which is querier */
1828 					/* on this interface */
1829 	uint16_t ill_mcast_v1_time;	/* # slow timeouts since last v1 qry */
1830 	uint16_t ill_mcast_v2_time;	/* # slow timeouts since last v2 qry */
1831 	uint8_t	ill_mcast_v1_tset;	/* 1 => timer is set; 0 => not set */
1832 	uint8_t	ill_mcast_v2_tset;	/* 1 => timer is set; 0 => not set */
1833 
1834 	uint8_t	ill_mcast_rv;		/* IGMPv3/MLDv2 robustness variable */
1835 	int	ill_mcast_qi;		/* IGMPv3/MLDv2 query interval var */
1836 
1837 	mblk_t	*ill_pending_mp;	/* IOCTL/DLPI awaiting completion. */
1838 	/*
1839 	 * All non-NULL cells between 'ill_first_mp_to_free' and
1840 	 * 'ill_last_mp_to_free' are freed in ill_delete.
1841 	 */
1842 #define	ill_first_mp_to_free	ill_bcast_mp
1843 	mblk_t	*ill_bcast_mp;		/* DLPI header for broadcasts. */
1844 	mblk_t	*ill_resolver_mp;	/* Resolver template. */
1845 	mblk_t	*ill_unbind_mp;		/* unbind mp from ill_dl_up() */
1846 	mblk_t	*ill_dlpi_deferred;	/* b_next chain of control messages */
1847 	mblk_t	*ill_phys_addr_mp;	/* mblk which holds ill_phys_addr */
1848 #define	ill_last_mp_to_free	ill_phys_addr_mp
1849 
1850 	cred_t	*ill_credp;		/* opener's credentials */
1851 	uint8_t	*ill_phys_addr;		/* ill_phys_addr_mp->b_rptr + off */
1852 
1853 	uint_t	ill_state_flags;	/* see ILL_* flags above */
1854 
1855 	/* Following bit fields protected by ipsq_t */
1856 	uint_t
1857 		ill_needs_attach : 1,
1858 		ill_reserved : 1,
1859 		ill_isv6 : 1,
1860 		ill_dlpi_style_set : 1,
1861 
1862 		ill_ifname_pending : 1,
1863 		ill_move_in_progress : 1, /* FAILOVER/FAILBACK in progress */
1864 		ill_join_allmulti : 1,
1865 		ill_logical_down : 1,
1866 
1867 		ill_is_6to4tun : 1,	/* Interface is a 6to4 tunnel */
1868 		ill_promisc_on_phys : 1, /* phys interface in promisc mode */
1869 		ill_dl_up : 1,
1870 		ill_up_ipifs : 1,
1871 
1872 		ill_note_link : 1,	/* supports link-up notification */
1873 		ill_capab_reneg : 1, /* capability renegotiation to be done */
1874 		ill_pad_to_bit_31 : 18;
1875 
1876 	/* Following bit fields protected by ill_lock */
1877 	uint_t
1878 		ill_fragtimer_executing : 1,
1879 		ill_fragtimer_needrestart : 1,
1880 		ill_ilm_cleanup_reqd : 1,
1881 		ill_arp_closing : 1,
1882 
1883 		ill_arp_bringup_pending : 1,
1884 		ill_mtu_userspecified : 1, /* SIOCSLIFLNKINFO has set the mtu */
1885 		ill_arp_extend : 1,	/* ARP has DAD extensions */
1886 		ill_pad_bit_31 : 25;
1887 
1888 	/*
1889 	 * Used in SIOCSIFMUXID and SIOCGIFMUXID for 'ifconfig unplumb'.
1890 	 */
1891 	int	ill_arp_muxid;		/* muxid returned from plink for arp */
1892 	int	ill_ip_muxid;		/* muxid returned from plink for ip */
1893 
1894 	/*
1895 	 * NIC event information attached, to be used by nic event hooks.
1896 	 */
1897 	hook_nic_event_t	*ill_nic_event_info;
1898 
1899 	/* Used for IP frag reassembly throttling on a per ILL basis.  */
1900 	uint_t	ill_ipf_gen;		/* Generation of next fragment queue */
1901 	uint_t	ill_frag_count;		/* Count of all reassembly mblk bytes */
1902 	uint_t	ill_frag_free_num_pkts;	 /* num of fragmented packets to free */
1903 	clock_t	ill_last_frag_clean_time; /* time when frag's were pruned */
1904 	int	ill_type;		/* From <net/if_types.h> */
1905 	uint_t	ill_dlpi_multicast_state;	/* See below IDS_* */
1906 	uint_t	ill_dlpi_fastpath_state;	/* See below IDS_* */
1907 
1908 	/*
1909 	 * Capabilities related fields.
1910 	 */
1911 	uint_t  ill_dlpi_capab_state;	/* State of capability query, IDS_* */
1912 	uint64_t ill_capabilities;	/* Enabled capabilities, ILL_CAPAB_* */
1913 	ill_mdt_capab_t	*ill_mdt_capab;	/* Multidata Transmit capabilities */
1914 	ill_ipsec_capab_t *ill_ipsec_capab_ah;	/* IPsec AH capabilities */
1915 	ill_ipsec_capab_t *ill_ipsec_capab_esp;	/* IPsec ESP capabilities */
1916 	ill_hcksum_capab_t *ill_hcksum_capab; /* H/W cksumming capabilities */
1917 	ill_zerocopy_capab_t *ill_zerocopy_capab; /* Zero-copy capabilities */
1918 	ill_dls_capab_t *ill_dls_capab; /* Polling, soft ring capabilities */
1919 	ill_lso_capab_t *ill_lso_capab; /* Large Segment Offload capabilities */
1920 
1921 	/*
1922 	 * New fields for IPv6
1923 	 */
1924 	uint8_t	ill_max_hops;	/* Maximum hops for any logical interface */
1925 	uint_t	ill_max_mtu;	/* Maximum MTU for any logical interface */
1926 	uint32_t ill_reachable_time;	/* Value for ND algorithm in msec */
1927 	uint32_t ill_reachable_retrans_time; /* Value for ND algorithm msec */
1928 	uint_t	ill_max_buf;		/* Max # of req to buffer for ND */
1929 	in6_addr_t	ill_token;
1930 	uint_t		ill_token_length;
1931 	uint32_t	ill_xmit_count;		/* ndp max multicast xmits */
1932 	mib2_ipIfStatsEntry_t	*ill_ip_mib;	/* ver indep. interface mib */
1933 	mib2_ipv6IfIcmpEntry_t	*ill_icmp6_mib;	/* Per interface mib */
1934 	/*
1935 	 * Following two mblks are allocated common to all
1936 	 * the ipifs when the first interface is coming up.
1937 	 * It is sent up to arp when the last ipif is coming
1938 	 * down.
1939 	 */
1940 	mblk_t			*ill_arp_down_mp;
1941 	mblk_t			*ill_arp_del_mapping_mp;
1942 	/*
1943 	 * Used for implementing IFF_NOARP. As IFF_NOARP is used
1944 	 * to turn off for all the logicals, it is here instead
1945 	 * of the ipif.
1946 	 */
1947 	mblk_t			*ill_arp_on_mp;
1948 	/* Peer ill of an IPMP move operation */
1949 	struct ill_s		*ill_move_peer;
1950 
1951 	phyint_t		*ill_phyint;
1952 	uint64_t		ill_flags;
1953 	ill_group_t		*ill_group;
1954 	struct ill_s		*ill_group_next;
1955 
1956 	kmutex_t	ill_lock;	/* Please see table below */
1957 	/*
1958 	 * The ill_nd_lla* fields handle the link layer address option
1959 	 * from neighbor discovery. This is used for external IPv6
1960 	 * address resolution.
1961 	 */
1962 	mblk_t		*ill_nd_lla_mp;	/* mblk which holds ill_nd_lla */
1963 	uint8_t		*ill_nd_lla;	/* Link Layer Address */
1964 	uint_t		ill_nd_lla_len;	/* Link Layer Address length */
1965 	/*
1966 	 * We now have 3 phys_addr_req's sent down. This field keeps track
1967 	 * of which one is pending.
1968 	 */
1969 	t_uscalar_t	ill_phys_addr_pend; /* which dl_phys_addr_req pending */
1970 	/*
1971 	 * Used to save errors that occur during plumbing
1972 	 */
1973 	uint_t		ill_ifname_pending_err;
1974 	avl_node_t	ill_avl_byppa; /* avl node based on ppa */
1975 	void		*ill_fastpath_list; /* both ire and nce hang off this */
1976 	uint_t		ill_refcnt;	/* active refcnt by threads */
1977 	uint_t		ill_ire_cnt;	/* ires associated with this ill */
1978 	kcondvar_t	ill_cv;
1979 	uint_t		ill_ilm_walker_cnt;	/* snmp ilm walkers */
1980 	uint_t		ill_nce_cnt;	/* nces associated with this ill */
1981 	uint_t		ill_waiters;	/* threads waiting in ipsq_enter */
1982 	/*
1983 	 * Contains the upper read queue pointer of the module immediately
1984 	 * beneath IP.  This field allows IP to validate sub-capability
1985 	 * acknowledgments coming up from downstream.
1986 	 */
1987 	queue_t		*ill_lmod_rq;	/* read queue pointer of module below */
1988 	uint_t		ill_lmod_cnt;	/* number of modules beneath IP */
1989 	ip_m_t		*ill_media;	/* media specific params/functions */
1990 	t_uscalar_t	ill_dlpi_pending; /* Last DLPI primitive issued */
1991 	uint_t		ill_usesrc_ifindex; /* use src addr from this ILL */
1992 	struct ill_s	*ill_usesrc_grp_next; /* Next ILL in the usesrc group */
1993 	boolean_t	ill_trace_disable;	/* True when alloc fails */
1994 	zoneid_t	ill_zoneid;
1995 	ip_stack_t	*ill_ipst;	/* Corresponds to a netstack_hold */
1996 	uint32_t	ill_dhcpinit;	/* IP_DHCPINIT_IFs for ill */
1997 	uint_t		ill_ilm_cnt;    /* ilms referencing this ill */
1998 } ill_t;
1999 
2000 /*
2001  * ILL_FREE_OK() means that there are no incoming pointer references
2002  * to the ill.
2003  */
2004 #define	ILL_FREE_OK(ill)					\
2005 	((ill)->ill_ire_cnt == 0 && (ill)->ill_ilm_cnt == 0 &&	\
2006 	(ill)->ill_nce_cnt == 0)
2007 
2008 /*
2009  * An ipif/ill can be marked down only when the ire and nce references
2010  * to that ipif/ill goes to zero. ILL_DOWN_OK() is a necessary condition
2011  * quiescence checks. See comments above IPIF_DOWN_OK for details
2012  * on why ires and nces are selectively considered for this macro.
2013  */
2014 #define	ILL_DOWN_OK(ill)	(ill->ill_ire_cnt == 0 && ill->ill_nce_cnt == 0)
2015 
2016 /*
2017  * The following table lists the protection levels of the various members
2018  * of the ill_t. Same notation as that used for ipif_t above is used.
2019  *
2020  *				Write			Read
2021  *
2022  * ill_ifptr			ill_g_lock + s		Write once
2023  * ill_rq			ipsq			Write once
2024  * ill_wq			ipsq			Write once
2025  *
2026  * ill_error			ipsq			None
2027  * ill_ipif			ill_g_lock + ipsq	ill_g_lock OR ipsq
2028  * ill_ipif_up_count		ill_lock + ipsq		ill_lock OR ipsq
2029  * ill_max_frag			ipsq			Write once
2030  *
2031  * ill_name			ill_g_lock + ipsq	Write once
2032  * ill_name_length		ill_g_lock + ipsq	Write once
2033  * ill_ndd_name			ipsq			Write once
2034  * ill_net_type			ipsq			Write once
2035  * ill_ppa			ill_g_lock + ipsq	Write once
2036  * ill_sap			ipsq + down ill		Write once
2037  * ill_sap_length		ipsq + down ill		Write once
2038  * ill_phys_addr_length		ipsq + down ill		Write once
2039  *
2040  * ill_bcast_addr_length	ipsq			ipsq
2041  * ill_mactype			ipsq			ipsq
2042  * ill_frag_ptr			ipsq			ipsq
2043  *
2044  * ill_frag_timer_id		ill_lock		ill_lock
2045  * ill_frag_hash_tbl		ipsq			up ill
2046  * ill_ilm			ipsq + ill_lock		ill_lock
2047  * ill_mcast_type		ill_lock		ill_lock
2048  * ill_mcast_v1_time		ill_lock		ill_lock
2049  * ill_mcast_v2_time		ill_lock		ill_lock
2050  * ill_mcast_v1_tset		ill_lock		ill_lock
2051  * ill_mcast_v2_tset		ill_lock		ill_lock
2052  * ill_mcast_rv			ill_lock		ill_lock
2053  * ill_mcast_qi			ill_lock		ill_lock
2054  * ill_pending_mp		ill_lock		ill_lock
2055  *
2056  * ill_bcast_mp			ipsq			ipsq
2057  * ill_resolver_mp		ipsq			only when ill is up
2058  * ill_down_mp			ipsq			ipsq
2059  * ill_dlpi_deferred		ill_lock		ill_lock
2060  * ill_dlpi_pending		ill_lock		ill_lock
2061  * ill_phys_addr_mp		ipsq + down ill		only when ill is up
2062  * ill_phys_addr		ipsq + down ill		only when ill is up
2063  *
2064  * ill_state_flags		ill_lock		ill_lock
2065  * exclusive bit flags		ipsq_t			ipsq_t
2066  * shared bit flags		ill_lock		ill_lock
2067  *
2068  * ill_arp_muxid		ipsq			Not atomic
2069  * ill_ip_muxid			ipsq			Not atomic
2070  *
2071  * ill_ipf_gen			Not atomic
2072  * ill_frag_count		atomics			atomics
2073  * ill_type			ipsq + down ill		only when ill is up
2074  * ill_dlpi_multicast_state	ill_lock		ill_lock
2075  * ill_dlpi_fastpath_state	ill_lock		ill_lock
2076  * ill_max_hops			ipsq			Not atomic
2077  *
2078  * ill_max_mtu
2079  *
2080  * ill_reachable_time		ipsq + ill_lock		ill_lock
2081  * ill_reachable_retrans_time	ipsq + ill_lock		ill_lock
2082  * ill_max_buf			ipsq + ill_lock		ill_lock
2083  *
2084  * Next 2 fields need ill_lock because of the get ioctls. They should not
2085  * report partially updated results without executing in the ipsq.
2086  * ill_token			ipsq + ill_lock		ill_lock
2087  * ill_token_length		ipsq + ill_lock		ill_lock
2088  * ill_xmit_count		ipsq + down ill		write once
2089  * ill_ip6_mib			ipsq + down ill		only when ill is up
2090  * ill_icmp6_mib		ipsq + down ill		only when ill is up
2091  * ill_arp_down_mp		ipsq			ipsq
2092  * ill_arp_del_mapping_mp	ipsq			ipsq
2093  * ill_arp_on_mp		ipsq			ipsq
2094  * ill_move_peer		ipsq			ipsq
2095  *
2096  * ill_phyint			ipsq, ill_g_lock, ill_lock	Any of them
2097  * ill_flags			ill_lock		ill_lock
2098  * ill_group			ipsq, ill_g_lock, ill_lock	Any of them
2099  * ill_group_next		ipsq, ill_g_lock, ill_lock	Any of them
2100  * ill_nd_lla_mp		ipsq + down ill		only when ill is up
2101  * ill_nd_lla			ipsq + down ill		only when ill is up
2102  * ill_nd_lla_len		ipsq + down ill		only when ill is up
2103  * ill_phys_addr_pend		ipsq + down ill		only when ill is up
2104  * ill_ifname_pending_err	ipsq			ipsq
2105  * ill_avl_byppa		ipsq, ill_g_lock	write once
2106  *
2107  * ill_fastpath_list		ill_lock		ill_lock
2108  * ill_refcnt			ill_lock		ill_lock
2109  * ill_ire_cnt			ill_lock		ill_lock
2110  * ill_cv			ill_lock		ill_lock
2111  * ill_ilm_walker_cnt		ill_lock		ill_lock
2112  * ill_nce_cnt			ill_lock		ill_lock
2113  * ill_ilm_cnt			ill_lock		ill_lock
2114  * ill_trace			ill_lock		ill_lock
2115  * ill_usesrc_grp_next		ill_g_usesrc_lock	ill_g_usesrc_lock
2116  * ill_dhcpinit			atomics			atomics
2117  */
2118 
2119 /*
2120  * For ioctl restart mechanism see ip_reprocess_ioctl()
2121  */
2122 struct ip_ioctl_cmd_s;
2123 
2124 typedef	int (*ifunc_t)(ipif_t *, struct sockaddr_in *, queue_t *, mblk_t *,
2125     struct ip_ioctl_cmd_s *, void *);
2126 
2127 typedef struct ip_ioctl_cmd_s {
2128 	int	ipi_cmd;
2129 	size_t	ipi_copyin_size;
2130 	uint_t	ipi_flags;
2131 	uint_t	ipi_cmd_type;
2132 	ifunc_t	ipi_func;
2133 	ifunc_t	ipi_func_restart;
2134 } ip_ioctl_cmd_t;
2135 
2136 /*
2137  * ipi_cmd_type:
2138  *
2139  * IF_CMD		1	old style ifreq cmd
2140  * LIF_CMD		2	new style lifreq cmd
2141  * TUN_CMD		3	tunnel related
2142  * ARP_CMD		4	arpreq cmd
2143  * XARP_CMD		5	xarpreq cmd
2144  * MSFILT_CMD		6	multicast source filter cmd
2145  * MISC_CMD		7	misc cmd (not a more specific one above)
2146  */
2147 
2148 enum { IF_CMD = 1, LIF_CMD, TUN_CMD, ARP_CMD, XARP_CMD, MSFILT_CMD, MISC_CMD };
2149 
2150 #define	IPI_DONTCARE	0	/* For ioctl encoded values that don't matter */
2151 
2152 /* Flag values in ipi_flags */
2153 #define	IPI_PRIV	0x1	/* Root only command */
2154 #define	IPI_MODOK	0x2	/* Permitted on mod instance of IP */
2155 #define	IPI_WR		0x4	/* Need to grab writer access */
2156 #define	IPI_GET_CMD	0x8	/* branch to mi_copyout on success */
2157 #define	IPI_REPL	0x10	/* valid for replacement ipif created in MOVE */
2158 #define	IPI_NULL_BCONT	0x20	/* ioctl has not data and hence no b_cont */
2159 #define	IPI_PASS_DOWN	0x40	/* pass this ioctl down when a module only */
2160 
2161 extern ip_ioctl_cmd_t	ip_ndx_ioctl_table[];
2162 extern ip_ioctl_cmd_t	ip_misc_ioctl_table[];
2163 extern int ip_ndx_ioctl_count;
2164 extern int ip_misc_ioctl_count;
2165 
2166 #define	ILL_CLEAR_MOVE(ill) {				\
2167 	ill_t *peer_ill;				\
2168 							\
2169 	peer_ill = (ill)->ill_move_peer;		\
2170 	ASSERT(peer_ill != NULL);			\
2171 	(ill)->ill_move_in_progress = B_FALSE;		\
2172 	peer_ill->ill_move_in_progress = B_FALSE;	\
2173 	(ill)->ill_move_peer = NULL;			\
2174 	peer_ill->ill_move_peer = NULL;			\
2175 }
2176 
2177 /* Passed down by ARP to IP during I_PLINK/I_PUNLINK */
2178 typedef struct ipmx_s {
2179 	char	ipmx_name[LIFNAMSIZ];		/* if name */
2180 	uint_t
2181 		ipmx_arpdev_stream : 1,		/* This is the arp stream */
2182 		ipmx_notused : 31;
2183 } ipmx_t;
2184 
2185 /*
2186  * State for detecting if a driver supports certain features.
2187  * Support for DL_ENABMULTI_REQ uses ill_dlpi_multicast_state.
2188  * Support for DLPI M_DATA fastpath uses ill_dlpi_fastpath_state.
2189  * Support for DL_CAPABILITY_REQ uses ill_dlpi_capab_state.
2190  */
2191 #define	IDS_UNKNOWN	0	/* No DLPI request sent */
2192 #define	IDS_INPROGRESS	1	/* DLPI request sent */
2193 #define	IDS_OK		2	/* DLPI request completed successfully */
2194 #define	IDS_FAILED	3	/* DLPI request failed */
2195 
2196 /* Named Dispatch Parameter Management Structure */
2197 typedef struct ipparam_s {
2198 	uint_t	ip_param_min;
2199 	uint_t	ip_param_max;
2200 	uint_t	ip_param_value;
2201 	char	*ip_param_name;
2202 } ipparam_t;
2203 
2204 /* Extended NDP Management Structure */
2205 typedef struct ipndp_s {
2206 	ndgetf_t	ip_ndp_getf;
2207 	ndsetf_t	ip_ndp_setf;
2208 	caddr_t		ip_ndp_data;
2209 	char		*ip_ndp_name;
2210 } ipndp_t;
2211 
2212 /*
2213  * The kernel stores security attributes of all gateways in a database made
2214  * up of one or more tsol_gcdb_t elements.  Each tsol_gcdb_t contains the
2215  * security-related credentials of the gateway.  More than one gateways may
2216  * share entries in the database.
2217  *
2218  * The tsol_gc_t structure represents the gateway to credential association,
2219  * and refers to an entry in the database.  One or more tsol_gc_t entities are
2220  * grouped together to form one or more tsol_gcgrp_t, each representing the
2221  * list of security attributes specific to the gateway.  A gateway may be
2222  * associated with at most one credentials group.
2223  */
2224 struct tsol_gcgrp_s;
2225 
2226 extern uchar_t	ip6opt_ls;	/* TX IPv6 enabler */
2227 
2228 /*
2229  * Gateway security credential record.
2230  */
2231 typedef struct tsol_gcdb_s {
2232 	uint_t		gcdb_refcnt;	/* reference count */
2233 	struct rtsa_s	gcdb_attr;	/* security attributes */
2234 #define	gcdb_mask	gcdb_attr.rtsa_mask
2235 #define	gcdb_doi	gcdb_attr.rtsa_doi
2236 #define	gcdb_slrange	gcdb_attr.rtsa_slrange
2237 } tsol_gcdb_t;
2238 
2239 /*
2240  * Gateway to credential association.
2241  */
2242 typedef struct tsol_gc_s {
2243 	uint_t		gc_refcnt;	/* reference count */
2244 	struct tsol_gcgrp_s *gc_grp;	/* pointer to group */
2245 	struct tsol_gc_s *gc_prev;	/* previous in list */
2246 	struct tsol_gc_s *gc_next;	/* next in list */
2247 	tsol_gcdb_t	*gc_db;		/* pointer to actual credentials */
2248 } tsol_gc_t;
2249 
2250 /*
2251  * Gateway credentials group address.
2252  */
2253 typedef struct tsol_gcgrp_addr_s {
2254 	int		ga_af;		/* address family */
2255 	in6_addr_t	ga_addr;	/* IPv4 mapped or IPv6 address */
2256 } tsol_gcgrp_addr_t;
2257 
2258 /*
2259  * Gateway credentials group.
2260  */
2261 typedef struct tsol_gcgrp_s {
2262 	uint_t		gcgrp_refcnt;	/* reference count */
2263 	krwlock_t	gcgrp_rwlock;	/* lock to protect following */
2264 	uint_t		gcgrp_count;	/* number of credentials */
2265 	tsol_gc_t	*gcgrp_head;	/* first credential in list */
2266 	tsol_gc_t	*gcgrp_tail;	/* last credential in list */
2267 	tsol_gcgrp_addr_t gcgrp_addr;	/* next-hop gateway address */
2268 } tsol_gcgrp_t;
2269 
2270 extern kmutex_t gcgrp_lock;
2271 
2272 #define	GC_REFRELE(p) {				\
2273 	ASSERT((p)->gc_grp != NULL);		\
2274 	rw_enter(&(p)->gc_grp->gcgrp_rwlock, RW_WRITER); \
2275 	ASSERT((p)->gc_refcnt > 0);		\
2276 	if (--((p)->gc_refcnt) == 0)		\
2277 		gc_inactive(p);			\
2278 	else					\
2279 		rw_exit(&(p)->gc_grp->gcgrp_rwlock); \
2280 }
2281 
2282 #define	GCGRP_REFHOLD(p) {			\
2283 	mutex_enter(&gcgrp_lock);		\
2284 	++((p)->gcgrp_refcnt);			\
2285 	ASSERT((p)->gcgrp_refcnt != 0);		\
2286 	mutex_exit(&gcgrp_lock);		\
2287 }
2288 
2289 #define	GCGRP_REFRELE(p) {			\
2290 	mutex_enter(&gcgrp_lock);		\
2291 	ASSERT((p)->gcgrp_refcnt > 0);		\
2292 	if (--((p)->gcgrp_refcnt) == 0)		\
2293 		gcgrp_inactive(p);		\
2294 	ASSERT(MUTEX_HELD(&gcgrp_lock));	\
2295 	mutex_exit(&gcgrp_lock);		\
2296 }
2297 
2298 /*
2299  * IRE gateway security attributes structure, pointed to by tsol_ire_gw_secattr
2300  */
2301 struct tsol_tnrhc;
2302 
2303 typedef struct tsol_ire_gw_secattr_s {
2304 	kmutex_t	igsa_lock;	/* lock to protect following */
2305 	struct tsol_tnrhc *igsa_rhc;	/* host entry for gateway */
2306 	tsol_gc_t	*igsa_gc;	/* for prefix IREs */
2307 	tsol_gcgrp_t	*igsa_gcgrp;	/* for cache IREs */
2308 } tsol_ire_gw_secattr_t;
2309 
2310 /*
2311  * Following are the macros to increment/decrement the reference
2312  * count of the IREs and IRBs (ire bucket).
2313  *
2314  * 1) We bump up the reference count of an IRE to make sure that
2315  *    it does not get deleted and freed while we are using it.
2316  *    Typically all the lookup functions hold the bucket lock,
2317  *    and look for the IRE. If it finds an IRE, it bumps up the
2318  *    reference count before dropping the lock. Sometimes we *may* want
2319  *    to bump up the reference count after we *looked* up i.e without
2320  *    holding the bucket lock. So, the IRE_REFHOLD macro does not assert
2321  *    on the bucket lock being held. Any thread trying to delete from
2322  *    the hash bucket can still do so but cannot free the IRE if
2323  *    ire_refcnt is not 0.
2324  *
2325  * 2) We bump up the reference count on the bucket where the IRE resides
2326  *    (IRB), when we want to prevent the IREs getting deleted from a given
2327  *    hash bucket. This makes life easier for ire_walk type functions which
2328  *    wants to walk the IRE list, call a function, but needs to drop
2329  *    the bucket lock to prevent recursive rw_enters. While the
2330  *    lock is dropped, the list could be changed by other threads or
2331  *    the same thread could end up deleting the ire or the ire pointed by
2332  *    ire_next. IRE_REFHOLDing the ire or ire_next is not sufficient as
2333  *    a delete will still remove the ire from the bucket while we have
2334  *    dropped the lock and hence the ire_next would be NULL. Thus, we
2335  *    need a mechanism to prevent deletions from a given bucket.
2336  *
2337  *    To prevent deletions, we bump up the reference count on the
2338  *    bucket. If the bucket is held, ire_delete just marks IRE_MARK_CONDEMNED
2339  *    both on the ire's ire_marks and the bucket's irb_marks. When the
2340  *    reference count on the bucket drops to zero, all the CONDEMNED ires
2341  *    are deleted. We don't have to bump up the reference count on the
2342  *    bucket if we are walking the bucket and never have to drop the bucket
2343  *    lock. Note that IRB_REFHOLD does not prevent addition of new ires
2344  *    in the list. It is okay because addition of new ires will not cause
2345  *    ire_next to point to freed memory. We do IRB_REFHOLD only when
2346  *    all of the 3 conditions are true :
2347  *
2348  *    1) The code needs to walk the IRE bucket from start to end.
2349  *    2) It may have to drop the bucket lock sometimes while doing (1)
2350  *    3) It does not want any ires to be deleted meanwhile.
2351  */
2352 
2353 /*
2354  * Bump up the reference count on the IRE. We cannot assert that the
2355  * bucket lock is being held as it is legal to bump up the reference
2356  * count after the first lookup has returned the IRE without
2357  * holding the lock. Currently ip_wput does this for caching IRE_CACHEs.
2358  */
2359 
2360 #ifdef DEBUG
2361 #define	IRE_UNTRACE_REF(ire)	ire_untrace_ref(ire);
2362 #define	IRE_TRACE_REF(ire)	ire_trace_ref(ire);
2363 #else
2364 #define	IRE_UNTRACE_REF(ire)
2365 #define	IRE_TRACE_REF(ire)
2366 #endif
2367 
2368 #define	IRE_REFHOLD_NOTR(ire) {				\
2369 	atomic_add_32(&(ire)->ire_refcnt, 1);		\
2370 	ASSERT((ire)->ire_refcnt != 0);			\
2371 }
2372 
2373 #define	IRE_REFHOLD(ire) {				\
2374 	IRE_REFHOLD_NOTR(ire);				\
2375 	IRE_TRACE_REF(ire);				\
2376 }
2377 
2378 #define	IRE_REFHOLD_LOCKED(ire)	{			\
2379 	IRE_TRACE_REF(ire);				\
2380 	(ire)->ire_refcnt++;				\
2381 }
2382 
2383 /*
2384  * Decrement the reference count on the IRE.
2385  * In architectures e.g sun4u, where atomic_add_32_nv is just
2386  * a cas, we need to maintain the right memory barrier semantics
2387  * as that of mutex_exit i.e all the loads and stores should complete
2388  * before the cas is executed. membar_exit() does that here.
2389  *
2390  * NOTE : This macro is used only in places where we want performance.
2391  *	  To avoid bloating the code, we use the function "ire_refrele"
2392  *	  which essentially calls the macro.
2393  */
2394 #define	IRE_REFRELE_NOTR(ire) {					\
2395 	ASSERT((ire)->ire_refcnt != 0);				\
2396 	membar_exit();						\
2397 	if (atomic_add_32_nv(&(ire)->ire_refcnt, -1) == 0)	\
2398 		ire_inactive(ire);				\
2399 }
2400 
2401 #define	IRE_REFRELE(ire) {					\
2402 	if (ire->ire_bucket != NULL) {				\
2403 		IRE_UNTRACE_REF(ire);				\
2404 	}							\
2405 	IRE_REFRELE_NOTR(ire);					\
2406 }
2407 
2408 /*
2409  * Bump up the reference count on the hash bucket - IRB to
2410  * prevent ires from being deleted in this bucket.
2411  */
2412 #define	IRB_REFHOLD(irb) {				\
2413 	rw_enter(&(irb)->irb_lock, RW_WRITER);		\
2414 	(irb)->irb_refcnt++;				\
2415 	ASSERT((irb)->irb_refcnt != 0);			\
2416 	rw_exit(&(irb)->irb_lock);			\
2417 }
2418 #define	IRB_REFHOLD_LOCKED(irb) {			\
2419 	ASSERT(RW_WRITE_HELD(&(irb)->irb_lock));	\
2420 	(irb)->irb_refcnt++;				\
2421 	ASSERT((irb)->irb_refcnt != 0);			\
2422 }
2423 
2424 void irb_refrele_ftable(irb_t *);
2425 /*
2426  * Note: when IRB_MARK_FTABLE (i.e., IRE_CACHETABLE entry), the irb_t
2427  * is statically allocated, so that when the irb_refcnt goes to 0,
2428  * we simply clean up the ire list and continue.
2429  */
2430 #define	IRB_REFRELE(irb) {				\
2431 	if ((irb)->irb_marks & IRB_MARK_FTABLE) {	\
2432 		irb_refrele_ftable((irb));		\
2433 	} else {					\
2434 		rw_enter(&(irb)->irb_lock, RW_WRITER);		\
2435 		ASSERT((irb)->irb_refcnt != 0);			\
2436 		if (--(irb)->irb_refcnt	== 0 &&			\
2437 		    ((irb)->irb_marks & IRE_MARK_CONDEMNED)) {	\
2438 			ire_t *ire_list;			\
2439 								\
2440 			ire_list = ire_unlink(irb);		\
2441 			rw_exit(&(irb)->irb_lock);		\
2442 			ASSERT(ire_list != NULL);		\
2443 			ire_cleanup(ire_list);			\
2444 		} else {					\
2445 			rw_exit(&(irb)->irb_lock);		\
2446 		}						\
2447 	}							\
2448 }
2449 
2450 extern struct kmem_cache *rt_entry_cache;
2451 
2452 /*
2453  * Lock the fast path mp for access, since the fp_mp can be deleted
2454  * due a DL_NOTE_FASTPATH_FLUSH in the case of IRE_BROADCAST
2455  */
2456 
2457 #define	LOCK_IRE_FP_MP(ire) {				\
2458 		if ((ire)->ire_type == IRE_BROADCAST)	\
2459 			mutex_enter(&ire->ire_nce->nce_lock);	\
2460 	}
2461 #define	UNLOCK_IRE_FP_MP(ire) {				\
2462 		if ((ire)->ire_type == IRE_BROADCAST)	\
2463 			mutex_exit(&ire->ire_nce->nce_lock);	\
2464 	}
2465 
2466 typedef struct ire4 {
2467 	ipaddr_t ire4_src_addr;		/* Source address to use. */
2468 	ipaddr_t ire4_mask;		/* Mask for matching this IRE. */
2469 	ipaddr_t ire4_addr;		/* Address this IRE represents. */
2470 	ipaddr_t ire4_gateway_addr;	/* Gateway if IRE_CACHE/IRE_OFFSUBNET */
2471 	ipaddr_t ire4_cmask;		/* Mask from parent prefix route */
2472 } ire4_t;
2473 
2474 typedef struct ire6 {
2475 	in6_addr_t ire6_src_addr;	/* Source address to use. */
2476 	in6_addr_t ire6_mask;		/* Mask for matching this IRE. */
2477 	in6_addr_t ire6_addr;		/* Address this IRE represents. */
2478 	in6_addr_t ire6_gateway_addr;	/* Gateway if IRE_CACHE/IRE_OFFSUBNET */
2479 	in6_addr_t ire6_cmask;		/* Mask from parent prefix route */
2480 } ire6_t;
2481 
2482 typedef union ire_addr {
2483 	ire6_t	ire6_u;
2484 	ire4_t	ire4_u;
2485 } ire_addr_u_t;
2486 
2487 /* Internet Routing Entry */
2488 typedef struct ire_s {
2489 	struct	ire_s	*ire_next;	/* The hash chain must be first. */
2490 	struct	ire_s	**ire_ptpn;	/* Pointer to previous next. */
2491 	uint32_t	ire_refcnt;	/* Number of references */
2492 	mblk_t		*ire_mp;	/* Non-null if allocated as mblk */
2493 	queue_t		*ire_rfq;	/* recv from this queue */
2494 	queue_t		*ire_stq;	/* send to this queue */
2495 	union {
2496 		uint_t	*max_fragp;	/* Used only during ire creation */
2497 		uint_t	max_frag;	/* MTU (next hop or path). */
2498 	} imf_u;
2499 #define	ire_max_frag	imf_u.max_frag
2500 #define	ire_max_fragp	imf_u.max_fragp
2501 	uint32_t	ire_frag_flag;	/* IPH_DF or zero. */
2502 	uint32_t	ire_ident;	/* Per IRE IP ident. */
2503 	uint32_t	ire_tire_mark;	/* Used for reclaim of unused. */
2504 	uchar_t		ire_ipversion;	/* IPv4/IPv6 version */
2505 	uchar_t		ire_marks;	/* IRE_MARK_CONDEMNED etc. */
2506 	ushort_t	ire_type;	/* Type of IRE */
2507 	uint_t	ire_ib_pkt_count;	/* Inbound packets for ire_addr */
2508 	uint_t	ire_ob_pkt_count;	/* Outbound packets to ire_addr */
2509 	uint_t	ire_ll_hdr_length;	/* Non-zero if we do M_DATA prepends */
2510 	time_t	ire_create_time;	/* Time (in secs) IRE was created. */
2511 	uint32_t	ire_phandle;	/* Associate prefix IREs to cache */
2512 	uint32_t	ire_ihandle;	/* Associate interface IREs to cache */
2513 	ipif_t		*ire_ipif;	/* the interface that this ire uses */
2514 	uint32_t	ire_flags;	/* flags related to route (RTF_*) */
2515 	uint_t	ire_ipsec_overhead;	/* IPSEC overhead */
2516 	/*
2517 	 * Neighbor Cache Entry for IPv6; arp info for IPv4
2518 	 */
2519 	struct	nce_s	*ire_nce;
2520 	uint_t		ire_masklen;	/* # bits in ire_mask{,_v6} */
2521 	ire_addr_u_t	ire_u;		/* IPv4/IPv6 address info. */
2522 
2523 	irb_t		*ire_bucket;	/* Hash bucket when ire_ptphn is set */
2524 	iulp_t		ire_uinfo;	/* Upper layer protocol info. */
2525 	/*
2526 	 * Protects ire_uinfo, ire_max_frag, and ire_frag_flag.
2527 	 */
2528 	kmutex_t	ire_lock;
2529 	uint_t		ire_ipif_seqid; /* ipif_seqid of ire_ipif */
2530 	clock_t		ire_last_used_time;	/* Last used time */
2531 	tsol_ire_gw_secattr_t *ire_gw_secattr; /* gateway security attributes */
2532 	zoneid_t	ire_zoneid;	/* for local address discrimination */
2533 	/*
2534 	 * ire's that are embedded inside mblk_t and sent to the external
2535 	 * resolver use the ire_stq_ifindex to track the ifindex of the
2536 	 * ire_stq, so that the ill (if it exists) can be correctly recovered
2537 	 * for cleanup in the esbfree routine when arp failure occurs
2538 	 */
2539 	uint_t	ire_stq_ifindex;
2540 	uint_t		ire_defense_count;	/* number of ARP conflicts */
2541 	uint_t		ire_defense_time;	/* last time defended (secs) */
2542 	boolean_t	ire_trace_disable;	/* True when alloc fails */
2543 	ip_stack_t	*ire_ipst;	/* Does not have a netstack_hold */
2544 } ire_t;
2545 
2546 /* IPv4 compatibility macros */
2547 #define	ire_src_addr		ire_u.ire4_u.ire4_src_addr
2548 #define	ire_mask		ire_u.ire4_u.ire4_mask
2549 #define	ire_addr		ire_u.ire4_u.ire4_addr
2550 #define	ire_gateway_addr	ire_u.ire4_u.ire4_gateway_addr
2551 #define	ire_cmask		ire_u.ire4_u.ire4_cmask
2552 
2553 #define	ire_src_addr_v6		ire_u.ire6_u.ire6_src_addr
2554 #define	ire_mask_v6		ire_u.ire6_u.ire6_mask
2555 #define	ire_addr_v6		ire_u.ire6_u.ire6_addr
2556 #define	ire_gateway_addr_v6	ire_u.ire6_u.ire6_gateway_addr
2557 #define	ire_cmask_v6		ire_u.ire6_u.ire6_cmask
2558 
2559 /* Convenient typedefs for sockaddrs */
2560 typedef	struct sockaddr_in	sin_t;
2561 typedef	struct sockaddr_in6	sin6_t;
2562 
2563 /* Address structure used for internal bind with IP */
2564 typedef struct ipa_conn_s {
2565 	ipaddr_t	ac_laddr;
2566 	ipaddr_t	ac_faddr;
2567 	uint16_t	ac_fport;
2568 	uint16_t	ac_lport;
2569 } ipa_conn_t;
2570 
2571 typedef struct ipa6_conn_s {
2572 	in6_addr_t	ac6_laddr;
2573 	in6_addr_t	ac6_faddr;
2574 	uint16_t	ac6_fport;
2575 	uint16_t	ac6_lport;
2576 } ipa6_conn_t;
2577 
2578 /*
2579  * Using ipa_conn_x_t or ipa6_conn_x_t allows us to modify the behavior of IP's
2580  * bind handler.
2581  */
2582 typedef struct ipa_conn_extended_s {
2583 	uint64_t	acx_flags;
2584 	ipa_conn_t	acx_conn;
2585 } ipa_conn_x_t;
2586 
2587 typedef struct ipa6_conn_extended_s {
2588 	uint64_t	ac6x_flags;
2589 	ipa6_conn_t	ac6x_conn;
2590 } ipa6_conn_x_t;
2591 
2592 /* flag values for ipa_conn_x_t and ipa6_conn_x_t. */
2593 #define	ACX_VERIFY_DST	0x1ULL	/* verify destination address is reachable */
2594 
2595 /* Name/Value Descriptor. */
2596 typedef struct nv_s {
2597 	uint64_t nv_value;
2598 	char	*nv_name;
2599 } nv_t;
2600 
2601 #define	ILL_FRAG_HASH(s, i) \
2602 	((ntohl(s) ^ ((i) ^ ((i) >> 8))) % ILL_FRAG_HASH_TBL_COUNT)
2603 
2604 /*
2605  * The MAX number of allowed fragmented packets per hash bucket
2606  * calculation is based on the most common mtu size of 1500. This limit
2607  * will work well for other mtu sizes as well.
2608  */
2609 #define	COMMON_IP_MTU 1500
2610 #define	MAX_FRAG_MIN 10
2611 #define	MAX_FRAG_PKTS(ipst)	\
2612 	MAX(MAX_FRAG_MIN, (2 * (ipst->ips_ip_reass_queue_bytes / \
2613 	    (COMMON_IP_MTU * ILL_FRAG_HASH_TBL_COUNT))))
2614 
2615 /*
2616  * Maximum dups allowed per packet.
2617  */
2618 extern uint_t ip_max_frag_dups;
2619 
2620 /*
2621  * Per-packet information for received packets and transmitted.
2622  * Used by the transport protocols when converting between the packet
2623  * and ancillary data and socket options.
2624  *
2625  * Note: This private data structure and related IPPF_* constant
2626  * definitions are exposed to enable compilation of some debugging tools
2627  * like lsof which use struct tcp_t in <inet/tcp.h>. This is intended to be
2628  * a temporary hack and long term alternate interfaces should be defined
2629  * to support the needs of such tools and private definitions moved to
2630  * private headers.
2631  */
2632 struct ip6_pkt_s {
2633 	uint_t		ipp_fields;		/* Which fields are valid */
2634 	uint_t		ipp_sticky_ignored;	/* sticky fields to ignore */
2635 	uint_t		ipp_ifindex;		/* pktinfo ifindex */
2636 	in6_addr_t	ipp_addr;		/* pktinfo src/dst addr */
2637 	uint_t		ipp_unicast_hops;	/* IPV6_UNICAST_HOPS */
2638 	uint_t		ipp_multicast_hops;	/* IPV6_MULTICAST_HOPS */
2639 	uint_t		ipp_hoplimit;		/* IPV6_HOPLIMIT */
2640 	uint_t		ipp_hopoptslen;
2641 	uint_t		ipp_rtdstoptslen;
2642 	uint_t		ipp_rthdrlen;
2643 	uint_t		ipp_dstoptslen;
2644 	uint_t		ipp_pathmtulen;
2645 	uint_t		ipp_fraghdrlen;
2646 	ip6_hbh_t	*ipp_hopopts;
2647 	ip6_dest_t	*ipp_rtdstopts;
2648 	ip6_rthdr_t	*ipp_rthdr;
2649 	ip6_dest_t	*ipp_dstopts;
2650 	ip6_frag_t	*ipp_fraghdr;
2651 	struct ip6_mtuinfo *ipp_pathmtu;
2652 	in6_addr_t	ipp_nexthop;		/* Transmit only */
2653 	uint8_t		ipp_tclass;
2654 	int8_t		ipp_use_min_mtu;
2655 };
2656 typedef struct ip6_pkt_s ip6_pkt_t;
2657 
2658 extern void ip6_pkt_free(ip6_pkt_t *);	/* free storage inside ip6_pkt_t */
2659 
2660 /*
2661  * This struct is used by ULP_opt_set() functions to return value of IPv4
2662  * ancillary options. Currently this is only used by udp and icmp and only
2663  * IP_PKTINFO option is supported.
2664  */
2665 typedef struct ip4_pkt_s {
2666 	uint_t		ip4_ill_index;	/* interface index */
2667 	ipaddr_t	ip4_addr;	/* source address */
2668 } ip4_pkt_t;
2669 
2670 /*
2671  * Used by ULP's to pass options info to ip_output
2672  * currently only IP_PKTINFO is supported.
2673  */
2674 typedef struct ip_opt_info_s {
2675 	uint_t ip_opt_ill_index;
2676 	uint_t ip_opt_flags;
2677 } ip_opt_info_t;
2678 
2679 /*
2680  * value for ip_opt_flags
2681  */
2682 #define	IP_VERIFY_SRC	0x1
2683 
2684 /*
2685  * This structure is used to convey information from IP and the ULP.
2686  * Currently used for the IP_RECVSLLA, IP_RECVIF and IP_RECVPKTINFO options.
2687  * The type of information field is set to IN_PKTINFO (i.e inbound pkt info)
2688  */
2689 typedef struct ip_pktinfo {
2690 	uint32_t		ip_pkt_ulp_type;	/* type of info sent */
2691 	uint32_t		ip_pkt_flags;	/* what is sent up by IP */
2692 	uint32_t		ip_pkt_ifindex;	/* inbound interface index */
2693 	struct sockaddr_dl	ip_pkt_slla;	/* has source link layer addr */
2694 	struct in_addr		ip_pkt_match_addr; /* matched address */
2695 } ip_pktinfo_t;
2696 
2697 /*
2698  * flags to tell UDP what IP is sending; in_pkt_flags
2699  */
2700 #define	IPF_RECVIF	0x01	/* inbound interface index */
2701 #define	IPF_RECVSLLA	0x02	/* source link layer address */
2702 /*
2703  * Inbound interface index + matched address.
2704  * Used only by IPV4.
2705  */
2706 #define	IPF_RECVADDR	0x04
2707 
2708 /* ipp_fields values */
2709 #define	IPPF_IFINDEX	0x0001	/* Part of in6_pktinfo: ifindex */
2710 #define	IPPF_ADDR	0x0002	/* Part of in6_pktinfo: src/dst addr */
2711 #define	IPPF_SCOPE_ID	0x0004	/* Add xmit ip6i_t for sin6_scope_id */
2712 #define	IPPF_NO_CKSUM	0x0008	/* Add xmit ip6i_t for IP6I_NO_*_CKSUM */
2713 
2714 #define	IPPF_RAW_CKSUM	0x0010	/* Add xmit ip6i_t for IP6I_RAW_CHECKSUM */
2715 #define	IPPF_HOPLIMIT	0x0020
2716 #define	IPPF_HOPOPTS	0x0040
2717 #define	IPPF_RTHDR	0x0080
2718 
2719 #define	IPPF_RTDSTOPTS	0x0100
2720 #define	IPPF_DSTOPTS	0x0200
2721 #define	IPPF_NEXTHOP	0x0400
2722 #define	IPPF_PATHMTU	0x0800
2723 
2724 #define	IPPF_TCLASS	0x1000
2725 #define	IPPF_DONTFRAG	0x2000
2726 #define	IPPF_USE_MIN_MTU	0x04000
2727 #define	IPPF_MULTICAST_HOPS	0x08000
2728 
2729 #define	IPPF_UNICAST_HOPS	0x10000
2730 #define	IPPF_FRAGHDR		0x20000
2731 
2732 #define	IPPF_HAS_IP6I \
2733 	(IPPF_IFINDEX|IPPF_ADDR|IPPF_NEXTHOP|IPPF_SCOPE_ID| \
2734 	IPPF_NO_CKSUM|IPPF_RAW_CKSUM|IPPF_HOPLIMIT|IPPF_DONTFRAG| \
2735 	IPPF_USE_MIN_MTU|IPPF_MULTICAST_HOPS|IPPF_UNICAST_HOPS)
2736 
2737 #define	TCP_PORTS_OFFSET	0
2738 #define	UDP_PORTS_OFFSET	0
2739 
2740 /*
2741  * lookups return the ill/ipif only if the flags are clear OR Iam writer.
2742  * ill / ipif lookup functions increment the refcnt on the ill / ipif only
2743  * after calling these macros. This ensures that the refcnt on the ipif or
2744  * ill will eventually drop down to zero.
2745  */
2746 #define	ILL_LOOKUP_FAILED	1	/* Used as error code */
2747 #define	IPIF_LOOKUP_FAILED	2	/* Used as error code */
2748 
2749 #define	ILL_CAN_LOOKUP(ill) 						\
2750 	(!((ill)->ill_state_flags & (ILL_CONDEMNED | ILL_CHANGING)) ||	\
2751 	IAM_WRITER_ILL(ill))
2752 
2753 #define	ILL_CAN_WAIT(ill, q)	\
2754 	(((q) != NULL) && !((ill)->ill_state_flags & (ILL_CONDEMNED)))
2755 
2756 #define	ILL_CAN_LOOKUP_WALKER(ill)	\
2757 	(!((ill)->ill_state_flags & ILL_CONDEMNED))
2758 
2759 #define	IPIF_CAN_LOOKUP(ipif)						\
2760 	(!((ipif)->ipif_state_flags & (IPIF_CONDEMNED | IPIF_CHANGING)) || \
2761 	IAM_WRITER_IPIF(ipif))
2762 
2763 /*
2764  * If the parameter 'q' is NULL, the caller is not interested in wait and
2765  * restart of the operation if the ILL or IPIF cannot be looked up when it is
2766  * marked as 'CHANGING'. Typically a thread that tries to send out data  will
2767  * end up passing NULLs as the last 4 parameters to ill_lookup_on_ifindex and
2768  * in this case 'q' is NULL
2769  */
2770 #define	IPIF_CAN_WAIT(ipif, q)	\
2771 	(((q) != NULL) && !((ipif)->ipif_state_flags & (IPIF_CONDEMNED)))
2772 
2773 #define	IPIF_CAN_LOOKUP_WALKER(ipif)					\
2774 	(!((ipif)->ipif_state_flags & (IPIF_CONDEMNED)) ||		\
2775 	IAM_WRITER_IPIF(ipif))
2776 
2777 /*
2778  * These macros are used by critical set ioctls and failover ioctls to
2779  * mark the ipif appropriately before starting the operation and to clear the
2780  * marks after completing the operation.
2781  */
2782 #define	IPIF_UNMARK_MOVING(ipif)                                \
2783 	(ipif)->ipif_state_flags &= ~IPIF_MOVING & ~IPIF_CHANGING;
2784 
2785 #define	ILL_UNMARK_CHANGING(ill)                                \
2786 	(ill)->ill_state_flags &= ~ILL_CHANGING;
2787 
2788 /* Macros used to assert that this thread is a writer */
2789 #define	IAM_WRITER_IPSQ(ipsq)	((ipsq)->ipsq_writer == curthread)
2790 #define	IAM_WRITER_ILL(ill)	IAM_WRITER_IPSQ((ill)->ill_phyint->phyint_ipsq)
2791 #define	IAM_WRITER_IPIF(ipif)	IAM_WRITER_ILL((ipif)->ipif_ill)
2792 
2793 /*
2794  * Grab ill locks in the proper order. The order is highest addressed
2795  * ill is locked first.
2796  */
2797 #define	GRAB_ILL_LOCKS(ill_1, ill_2)				\
2798 {								\
2799 	if ((ill_1) > (ill_2)) {				\
2800 		if (ill_1 != NULL)				\
2801 			mutex_enter(&(ill_1)->ill_lock);	\
2802 		if (ill_2 != NULL)				\
2803 			mutex_enter(&(ill_2)->ill_lock);	\
2804 	} else {						\
2805 		if (ill_2 != NULL)				\
2806 			mutex_enter(&(ill_2)->ill_lock);	\
2807 		if (ill_1 != NULL && ill_1 != ill_2)		\
2808 			mutex_enter(&(ill_1)->ill_lock);	\
2809 	}							\
2810 }
2811 
2812 #define	RELEASE_ILL_LOCKS(ill_1, ill_2)		\
2813 {						\
2814 	if (ill_1 != NULL)			\
2815 		mutex_exit(&(ill_1)->ill_lock);	\
2816 	if (ill_2 != NULL && ill_2 != ill_1)	\
2817 		mutex_exit(&(ill_2)->ill_lock);	\
2818 }
2819 
2820 /* Get the other protocol instance ill */
2821 #define	ILL_OTHER(ill)						\
2822 	((ill)->ill_isv6 ? (ill)->ill_phyint->phyint_illv4 :	\
2823 	    (ill)->ill_phyint->phyint_illv6)
2824 
2825 #define	MATCH_V4_ONLY	0x1
2826 #define	MATCH_V6_ONLY	0x2
2827 #define	MATCH_ILL_ONLY	0x4
2828 
2829 /* ioctl command info: Ioctl properties extracted and stored in here */
2830 typedef struct cmd_info_s
2831 {
2832 	char    ci_groupname[LIFNAMSIZ + 1];	/* SIOCSLIFGROUPNAME */
2833 	ipif_t  *ci_ipif;	/* ipif associated with [l]ifreq ioctl's */
2834 	sin_t	*ci_sin;	/* the sin struct passed down */
2835 	sin6_t	*ci_sin6;	/* the sin6_t struct passed down */
2836 	struct lifreq *ci_lifr;	/* the lifreq struct passed down */
2837 } cmd_info_t;
2838 
2839 /*
2840  * List of AH and ESP IPsec acceleration capable ills
2841  */
2842 typedef struct ipsec_capab_ill_s {
2843 	uint_t ill_index;
2844 	boolean_t ill_isv6;
2845 	struct ipsec_capab_ill_s *next;
2846 } ipsec_capab_ill_t;
2847 
2848 extern struct kmem_cache *ire_cache;
2849 
2850 extern ipaddr_t	ip_g_all_ones;
2851 
2852 extern	uint_t	ip_loopback_mtu;	/* /etc/system */
2853 
2854 extern vmem_t *ip_minor_arena_sa;
2855 extern vmem_t *ip_minor_arena_la;
2856 
2857 /*
2858  * ip_g_forward controls IP forwarding.  It takes two values:
2859  *	0: IP_FORWARD_NEVER	Don't forward packets ever.
2860  *	1: IP_FORWARD_ALWAYS	Forward packets for elsewhere.
2861  *
2862  * RFC1122 says there must be a configuration switch to control forwarding,
2863  * but that the default MUST be to not forward packets ever.  Implicit
2864  * control based on configuration of multiple interfaces MUST NOT be
2865  * implemented (Section 3.1).  SunOS 4.1 did provide the "automatic" capability
2866  * and, in fact, it was the default.  That capability is now provided in the
2867  * /etc/rc2.d/S69inet script.
2868  */
2869 
2870 #define	ips_ip_respond_to_address_mask_broadcast ips_param_arr[0].ip_param_value
2871 #define	ips_ip_g_resp_to_echo_bcast	ips_param_arr[1].ip_param_value
2872 #define	ips_ip_g_resp_to_echo_mcast	ips_param_arr[2].ip_param_value
2873 #define	ips_ip_g_resp_to_timestamp	ips_param_arr[3].ip_param_value
2874 #define	ips_ip_g_resp_to_timestamp_bcast ips_param_arr[4].ip_param_value
2875 #define	ips_ip_g_send_redirects		ips_param_arr[5].ip_param_value
2876 #define	ips_ip_g_forward_directed_bcast	ips_param_arr[6].ip_param_value
2877 #define	ips_ip_mrtdebug			ips_param_arr[7].ip_param_value
2878 #define	ips_ip_timer_interval		ips_param_arr[8].ip_param_value
2879 #define	ips_ip_ire_arp_interval		ips_param_arr[9].ip_param_value
2880 #define	ips_ip_ire_redir_interval	ips_param_arr[10].ip_param_value
2881 #define	ips_ip_def_ttl			ips_param_arr[11].ip_param_value
2882 #define	ips_ip_forward_src_routed	ips_param_arr[12].ip_param_value
2883 #define	ips_ip_wroff_extra		ips_param_arr[13].ip_param_value
2884 #define	ips_ip_ire_pathmtu_interval	ips_param_arr[14].ip_param_value
2885 #define	ips_ip_icmp_return		ips_param_arr[15].ip_param_value
2886 #define	ips_ip_path_mtu_discovery	ips_param_arr[16].ip_param_value
2887 #define	ips_ip_ignore_delete_time	ips_param_arr[17].ip_param_value
2888 #define	ips_ip_ignore_redirect		ips_param_arr[18].ip_param_value
2889 #define	ips_ip_output_queue		ips_param_arr[19].ip_param_value
2890 #define	ips_ip_broadcast_ttl		ips_param_arr[20].ip_param_value
2891 #define	ips_ip_icmp_err_interval	ips_param_arr[21].ip_param_value
2892 #define	ips_ip_icmp_err_burst		ips_param_arr[22].ip_param_value
2893 #define	ips_ip_reass_queue_bytes	ips_param_arr[23].ip_param_value
2894 #define	ips_ip_strict_dst_multihoming	ips_param_arr[24].ip_param_value
2895 #define	ips_ip_addrs_per_if		ips_param_arr[25].ip_param_value
2896 #define	ips_ipsec_override_persocket_policy ips_param_arr[26].ip_param_value
2897 #define	ips_icmp_accept_clear_messages	ips_param_arr[27].ip_param_value
2898 #define	ips_igmp_accept_clear_messages	ips_param_arr[28].ip_param_value
2899 
2900 /* IPv6 configuration knobs */
2901 #define	ips_delay_first_probe_time	ips_param_arr[29].ip_param_value
2902 #define	ips_max_unicast_solicit		ips_param_arr[30].ip_param_value
2903 #define	ips_ipv6_def_hops		ips_param_arr[31].ip_param_value
2904 #define	ips_ipv6_icmp_return		ips_param_arr[32].ip_param_value
2905 #define	ips_ipv6_forward_src_routed	ips_param_arr[33].ip_param_value
2906 #define	ips_ipv6_resp_echo_mcast	ips_param_arr[34].ip_param_value
2907 #define	ips_ipv6_send_redirects		ips_param_arr[35].ip_param_value
2908 #define	ips_ipv6_ignore_redirect	ips_param_arr[36].ip_param_value
2909 #define	ips_ipv6_strict_dst_multihoming	ips_param_arr[37].ip_param_value
2910 #define	ips_ip_ire_reclaim_fraction	ips_param_arr[38].ip_param_value
2911 #define	ips_ipsec_policy_log_interval	ips_param_arr[39].ip_param_value
2912 #define	ips_pim_accept_clear_messages	ips_param_arr[40].ip_param_value
2913 #define	ips_ip_ndp_unsolicit_interval	ips_param_arr[41].ip_param_value
2914 #define	ips_ip_ndp_unsolicit_count	ips_param_arr[42].ip_param_value
2915 #define	ips_ipv6_ignore_home_address_opt ips_param_arr[43].ip_param_value
2916 #define	ips_ip_policy_mask		ips_param_arr[44].ip_param_value
2917 #define	ips_ip_multirt_resolution_interval ips_param_arr[45].ip_param_value
2918 #define	ips_ip_multirt_ttl  		ips_param_arr[46].ip_param_value
2919 #define	ips_ip_multidata_outbound	ips_param_arr[47].ip_param_value
2920 #define	ips_ip_ndp_defense_interval	ips_param_arr[48].ip_param_value
2921 #define	ips_ip_max_temp_idle		ips_param_arr[49].ip_param_value
2922 #define	ips_ip_max_temp_defend		ips_param_arr[50].ip_param_value
2923 #define	ips_ip_max_defend		ips_param_arr[51].ip_param_value
2924 #define	ips_ip_defend_interval		ips_param_arr[52].ip_param_value
2925 #define	ips_ip_dup_recovery		ips_param_arr[53].ip_param_value
2926 #define	ips_ip_restrict_interzone_loopback ips_param_arr[54].ip_param_value
2927 #define	ips_ip_lso_outbound		ips_param_arr[55].ip_param_value
2928 #define	ips_igmp_max_version		ips_param_arr[56].ip_param_value
2929 #define	ips_mld_max_version		ips_param_arr[57].ip_param_value
2930 #define	ips_ipv6_drop_inbound_icmpv6	ips_param_arr[58].ip_param_value
2931 
2932 extern int	dohwcksum;	/* use h/w cksum if supported by the h/w */
2933 #ifdef ZC_TEST
2934 extern int	noswcksum;
2935 #endif
2936 
2937 extern char	ipif_loopback_name[];
2938 
2939 extern nv_t	*ire_nv_tbl;
2940 
2941 extern struct module_info ip_mod_info;
2942 
2943 #define	HOOKS4_INTERESTED_PHYSICAL_IN(ipst)	\
2944 	((ipst)->ips_ip4_physical_in_event.he_interested)
2945 #define	HOOKS6_INTERESTED_PHYSICAL_IN(ipst)	\
2946 	((ipst)->ips_ip6_physical_in_event.he_interested)
2947 #define	HOOKS4_INTERESTED_PHYSICAL_OUT(ipst)	\
2948 	((ipst)->ips_ip4_physical_out_event.he_interested)
2949 #define	HOOKS6_INTERESTED_PHYSICAL_OUT(ipst)	\
2950 	((ipst)->ips_ip6_physical_out_event.he_interested)
2951 #define	HOOKS4_INTERESTED_FORWARDING(ipst)	\
2952 	((ipst)->ips_ip4_forwarding_event.he_interested)
2953 #define	HOOKS6_INTERESTED_FORWARDING(ipst)	\
2954 	((ipst)->ips_ip6_forwarding_event.he_interested)
2955 #define	HOOKS4_INTERESTED_LOOPBACK_IN(ipst)	\
2956 	((ipst)->ips_ip4_loopback_in_event.he_interested)
2957 #define	HOOKS6_INTERESTED_LOOPBACK_IN(ipst)	\
2958 	((ipst)->ips_ip6_loopback_in_event.he_interested)
2959 #define	HOOKS4_INTERESTED_LOOPBACK_OUT(ipst)	\
2960 	((ipst)->ips_ip4_loopback_out_event.he_interested)
2961 #define	HOOKS6_INTERESTED_LOOPBACK_OUT(ipst)	\
2962 	((ipst)->ips_ip6_loopback_out_event.he_interested)
2963 
2964 /*
2965  * Hooks marcos used inside of ip
2966  */
2967 #define	IPHA_VHL	ipha_version_and_hdr_length
2968 
2969 #define	FW_HOOKS(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst)	\
2970 									\
2971 	if ((_hook).he_interested) {	\
2972 		hook_pkt_event_t info;					\
2973 									\
2974 		_NOTE(CONSTCOND)					\
2975 		ASSERT((_ilp != NULL) || (_olp != NULL));		\
2976 									\
2977 		_NOTE(CONSTCOND)					\
2978 		if ((_ilp != NULL) &&					\
2979 		    (((ill_t *)(_ilp))->ill_phyint != NULL))		\
2980 			info.hpe_ifp = (phy_if_t)((ill_t *)		\
2981 			    (_ilp))->ill_phyint->phyint_hook_ifindex;	\
2982 		else							\
2983 			info.hpe_ifp = 0;				\
2984 									\
2985 		_NOTE(CONSTCOND)					\
2986 		if ((_olp != NULL) &&					\
2987 		    (((ill_t *)(_olp))->ill_phyint != NULL))		\
2988 			info.hpe_ofp = (phy_if_t)((ill_t *)		\
2989 			    (_olp))->ill_phyint->phyint_hook_ifindex;	\
2990 		else							\
2991 			info.hpe_ofp = 0;				\
2992 		info.hpe_hdr = _iph;					\
2993 		info.hpe_mp = &(_fm);					\
2994 		info.hpe_mb = _m;					\
2995 		info.hpe_flags = _llm;					\
2996 		if (hook_run(_event, (hook_data_t)&info,		\
2997 		    ipst->ips_netstack) != 0) {				\
2998 			ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
2999 			    (_hook).he_name, (void *)_fm, (void *)_m));	\
3000 			if (_fm != NULL) {				\
3001 				freemsg(_fm);				\
3002 				_fm = NULL;				\
3003 			}						\
3004 			_iph = NULL;					\
3005 			_m = NULL;					\
3006 		} else {						\
3007 			_iph = info.hpe_hdr;				\
3008 			_m = info.hpe_mb;				\
3009 		}							\
3010 	}
3011 
3012 #define	FW_HOOKS6(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst)	\
3013 									\
3014 	if ((_hook).he_interested) {	\
3015 		hook_pkt_event_t info;					\
3016 									\
3017 		_NOTE(CONSTCOND)					\
3018 		ASSERT((_ilp != NULL) || (_olp != NULL));		\
3019 									\
3020 		_NOTE(CONSTCOND)					\
3021 		if ((_ilp != NULL) &&					\
3022 		    (((ill_t *)(_ilp))->ill_phyint != NULL))		\
3023 			info.hpe_ifp = (phy_if_t)((ill_t *)		\
3024 			    (_ilp))->ill_phyint->phyint_hook_ifindex;	\
3025 		else							\
3026 			info.hpe_ifp = 0;				\
3027 									\
3028 		_NOTE(CONSTCOND)					\
3029 		if ((_olp != NULL) &&					\
3030 		    (((ill_t *)(_olp))->ill_phyint != NULL))		\
3031 			info.hpe_ofp = (phy_if_t)((ill_t *)		\
3032 			    (_olp))->ill_phyint->phyint_hook_ifindex;	\
3033 		else							\
3034 			info.hpe_ofp = 0;				\
3035 		info.hpe_hdr = _iph;					\
3036 		info.hpe_mp = &(_fm);					\
3037 		info.hpe_mb = _m;					\
3038 		info.hpe_flags = _llm;					\
3039 		if (hook_run(_event, (hook_data_t)&info,		\
3040 		    ipst->ips_netstack) != 0) {				\
3041 			ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
3042 			    (_hook).he_name, (void *)_fm, (void *)_m));	\
3043 			if (_fm != NULL) {				\
3044 				freemsg(_fm);				\
3045 				_fm = NULL;				\
3046 			}						\
3047 			_iph = NULL;					\
3048 			_m = NULL;					\
3049 		} else {						\
3050 			_iph = info.hpe_hdr;				\
3051 			_m = info.hpe_mb;				\
3052 		}							\
3053 	}
3054 
3055 /*
3056  * Network byte order macros
3057  */
3058 #ifdef	_BIG_ENDIAN
3059 #define	N_IN_CLASSA_NET		IN_CLASSA_NET
3060 #define	N_IN_CLASSD_NET		IN_CLASSD_NET
3061 #define	N_INADDR_UNSPEC_GROUP	INADDR_UNSPEC_GROUP
3062 #define	N_IN_LOOPBACK_NET	(ipaddr_t)0x7f000000U
3063 #else /* _BIG_ENDIAN */
3064 #define	N_IN_CLASSA_NET		(ipaddr_t)0x000000ffU
3065 #define	N_IN_CLASSD_NET		(ipaddr_t)0x000000f0U
3066 #define	N_INADDR_UNSPEC_GROUP	(ipaddr_t)0x000000e0U
3067 #define	N_IN_LOOPBACK_NET	(ipaddr_t)0x0000007fU
3068 #endif /* _BIG_ENDIAN */
3069 #define	CLASSD(addr)	(((addr) & N_IN_CLASSD_NET) == N_INADDR_UNSPEC_GROUP)
3070 #define	CLASSE(addr)	(((addr) & N_IN_CLASSD_NET) == N_IN_CLASSD_NET)
3071 #define	IP_LOOPBACK_ADDR(addr)			\
3072 	(((addr) & N_IN_CLASSA_NET == N_IN_LOOPBACK_NET))
3073 
3074 #ifdef DEBUG
3075 /* IPsec HW acceleration debugging support */
3076 
3077 #define	IPSECHW_CAPAB		0x0001	/* capability negotiation */
3078 #define	IPSECHW_SADB		0x0002	/* SADB exchange */
3079 #define	IPSECHW_PKT		0x0004	/* general packet flow */
3080 #define	IPSECHW_PKTIN		0x0008	/* driver in pkt processing details */
3081 #define	IPSECHW_PKTOUT		0x0010	/* driver out pkt processing details */
3082 
3083 #define	IPSECHW_DEBUG(f, x)	if (ipsechw_debug & (f)) { (void) printf x; }
3084 #define	IPSECHW_CALL(f, r, x)	if (ipsechw_debug & (f)) { (void) r x; }
3085 
3086 extern uint32_t ipsechw_debug;
3087 #else
3088 #define	IPSECHW_DEBUG(f, x)	{}
3089 #define	IPSECHW_CALL(f, r, x)	{}
3090 #endif
3091 
3092 extern int	ip_debug;
3093 extern uint_t	ip_thread_data;
3094 extern krwlock_t ip_thread_rwlock;
3095 extern list_t	ip_thread_list;
3096 
3097 #ifdef IP_DEBUG
3098 #include <sys/debug.h>
3099 #include <sys/promif.h>
3100 
3101 #define	ip0dbg(a)	printf a
3102 #define	ip1dbg(a)	if (ip_debug > 2) printf a
3103 #define	ip2dbg(a)	if (ip_debug > 3) printf a
3104 #define	ip3dbg(a)	if (ip_debug > 4) printf a
3105 #else
3106 #define	ip0dbg(a)	/* */
3107 #define	ip1dbg(a)	/* */
3108 #define	ip2dbg(a)	/* */
3109 #define	ip3dbg(a)	/* */
3110 #endif	/* IP_DEBUG */
3111 
3112 /* Default MAC-layer address string length for mac_colon_addr */
3113 #define	MAC_STR_LEN	128
3114 
3115 struct	ipsec_out_s;
3116 
3117 struct	mac_header_info_s;
3118 
3119 extern boolean_t ip_assign_ifindex(uint_t *, ip_stack_t *);
3120 extern void	ill_frag_timer(void *);
3121 extern ill_t	*ill_first(int, int, ill_walk_context_t *, ip_stack_t *);
3122 extern ill_t	*ill_next(ill_walk_context_t *, ill_t *);
3123 extern void	ill_frag_timer_start(ill_t *);
3124 extern void	ill_nic_info_dispatch(ill_t *);
3125 extern void	ill_nic_info_plumb(ill_t *, boolean_t);
3126 extern mblk_t	*ip_carve_mp(mblk_t **, ssize_t);
3127 extern mblk_t	*ip_dlpi_alloc(size_t, t_uscalar_t);
3128 extern char	*ip_dot_addr(ipaddr_t, char *);
3129 extern const char *mac_colon_addr(const uint8_t *, size_t, char *, size_t);
3130 extern void	ip_lwput(queue_t *, mblk_t *);
3131 extern boolean_t icmp_err_rate_limit(ip_stack_t *);
3132 extern void	icmp_time_exceeded(queue_t *, mblk_t *, uint8_t, zoneid_t,
3133     ip_stack_t *);
3134 extern void	icmp_unreachable(queue_t *, mblk_t *, uint8_t, zoneid_t,
3135     ip_stack_t *);
3136 extern mblk_t	*ip_add_info(mblk_t *, ill_t *, uint_t, zoneid_t, ip_stack_t *);
3137 extern mblk_t	*ip_bind_v4(queue_t *, mblk_t *, conn_t *);
3138 extern int	ip_bind_connected(conn_t *, mblk_t *, ipaddr_t *, uint16_t,
3139     ipaddr_t, uint16_t, boolean_t, boolean_t, boolean_t, boolean_t);
3140 extern boolean_t ip_bind_ipsec_policy_set(conn_t *, mblk_t *);
3141 extern int	ip_bind_laddr(conn_t *, mblk_t *, ipaddr_t, uint16_t,
3142     boolean_t, boolean_t, boolean_t);
3143 extern uint_t	ip_cksum(mblk_t *, int, uint32_t);
3144 extern int	ip_close(queue_t *, int);
3145 extern uint16_t	ip_csum_hdr(ipha_t *);
3146 extern void	ip_proto_not_sup(queue_t *, mblk_t *, uint_t, zoneid_t,
3147     ip_stack_t *);
3148 extern void	ip_ire_g_fini(void);
3149 extern void	ip_ire_g_init(void);
3150 extern void	ip_ire_fini(ip_stack_t *);
3151 extern void	ip_ire_init(ip_stack_t *);
3152 extern int	ip_openv4(queue_t *q, dev_t *devp, int flag, int sflag,
3153 		    cred_t *credp);
3154 extern int	ip_openv6(queue_t *q, dev_t *devp, int flag, int sflag,
3155 		    cred_t *credp);
3156 extern int	ip_reassemble(mblk_t *, ipf_t *, uint_t, boolean_t, ill_t *,
3157     size_t);
3158 extern int	ip_opt_set_ill(conn_t *, int, boolean_t, boolean_t,
3159     int, int, mblk_t *);
3160 extern void	ip_rput(queue_t *, mblk_t *);
3161 extern void	ip_input(ill_t *, ill_rx_ring_t *, mblk_t *,
3162     struct mac_header_info_s *);
3163 extern void	ip_rput_dlpi(queue_t *, mblk_t *);
3164 extern void	ip_rput_forward(ire_t *, ipha_t *, mblk_t *, ill_t *);
3165 extern void	ip_rput_forward_multicast(ipaddr_t, mblk_t *, ipif_t *);
3166 
3167 extern void	ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *,
3168 		    mib2_ipIfStatsEntry_t *);
3169 extern void	ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *,
3170 		    mib2_ipv6IfIcmpEntry_t *);
3171 extern void	ip_udp_input(queue_t *, mblk_t *, ipha_t *, ire_t *, ill_t *);
3172 extern void	ip_proto_input(queue_t *, mblk_t *, ipha_t *, ire_t *, ill_t *,
3173     uint32_t);
3174 extern void	ip_rput_other(ipsq_t *, queue_t *, mblk_t *, void *);
3175 extern ire_t	*ip_check_multihome(void *, ire_t *, ill_t *);
3176 extern void	ip_setpktversion(conn_t *, boolean_t, boolean_t, ip_stack_t *);
3177 extern void	ip_trash_ire_reclaim(void *);
3178 extern void	ip_trash_timer_expire(void *);
3179 extern void	ip_wput(queue_t *, mblk_t *);
3180 extern void	ip_output(void *, mblk_t *, void *, int);
3181 extern void	ip_output_options(void *, mblk_t *, void *, int,
3182     ip_opt_info_t *);
3183 
3184 extern void	ip_wput_ire(queue_t *, mblk_t *, ire_t *, conn_t *, int,
3185 		    zoneid_t);
3186 extern void	ip_wput_local(queue_t *, ill_t *, ipha_t *, mblk_t *, ire_t *,
3187 		    int, zoneid_t);
3188 extern void	ip_wput_multicast(queue_t *, mblk_t *, ipif_t *, zoneid_t);
3189 extern void	ip_wput_nondata(ipsq_t *, queue_t *, mblk_t *, void *);
3190 extern void	ip_wsrv(queue_t *);
3191 extern char	*ip_nv_lookup(nv_t *, int);
3192 extern boolean_t ip_local_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3193 extern boolean_t ip_remote_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3194 extern ipaddr_t ip_massage_options(ipha_t *, netstack_t *);
3195 extern ipaddr_t ip_net_mask(ipaddr_t);
3196 extern void	ip_newroute(queue_t *, mblk_t *, ipaddr_t, conn_t *, zoneid_t,
3197 		    ip_stack_t *);
3198 extern ipxmit_state_t	ip_xmit_v4(mblk_t *, ire_t *, struct ipsec_out_s *,
3199     boolean_t);
3200 extern int	ip_hdr_complete(ipha_t *, zoneid_t, ip_stack_t *);
3201 
3202 extern struct qinit iprinitv6;
3203 extern struct qinit ipwinitv6;
3204 
3205 extern void	conn_drain_insert(conn_t *connp);
3206 extern	int	conn_ipsec_length(conn_t *connp);
3207 extern void	ip_wput_ipsec_out(queue_t *, mblk_t *, ipha_t *, ill_t *,
3208     ire_t *);
3209 extern ipaddr_t	ip_get_dst(ipha_t *);
3210 extern int	ipsec_out_extra_length(mblk_t *);
3211 extern int	ipsec_in_extra_length(mblk_t *);
3212 extern mblk_t	*ipsec_in_alloc(boolean_t, netstack_t *);
3213 extern boolean_t ipsec_in_is_secure(mblk_t *);
3214 extern void	ipsec_out_process(queue_t *, mblk_t *, ire_t *, uint_t);
3215 extern void	ipsec_out_to_in(mblk_t *);
3216 extern void	ip_fanout_proto_again(mblk_t *, ill_t *, ill_t *, ire_t *);
3217 
3218 extern void	ire_cleanup(ire_t *);
3219 extern void	ire_inactive(ire_t *);
3220 extern boolean_t irb_inactive(irb_t *);
3221 extern ire_t	*ire_unlink(irb_t *);
3222 
3223 #ifdef DEBUG
3224 extern	boolean_t th_trace_ref(const void *, ip_stack_t *);
3225 extern	void	th_trace_unref(const void *);
3226 extern	void	th_trace_cleanup(const void *, boolean_t);
3227 extern	void	ire_trace_ref(ire_t *);
3228 extern	void	ire_untrace_ref(ire_t *);
3229 #endif
3230 
3231 extern int	ip_srcid_insert(const in6_addr_t *, zoneid_t, ip_stack_t *);
3232 extern int	ip_srcid_remove(const in6_addr_t *, zoneid_t, ip_stack_t *);
3233 extern void	ip_srcid_find_id(uint_t, in6_addr_t *, zoneid_t, netstack_t *);
3234 extern uint_t	ip_srcid_find_addr(const in6_addr_t *, zoneid_t, netstack_t *);
3235 extern int	ip_srcid_report(queue_t *, mblk_t *, caddr_t, cred_t *);
3236 
3237 extern uint8_t	ipoptp_next(ipoptp_t *);
3238 extern uint8_t	ipoptp_first(ipoptp_t *, ipha_t *);
3239 extern int	ip_opt_get_user(const ipha_t *, uchar_t *);
3240 extern ill_t	*ip_grab_attach_ill(ill_t *, mblk_t *, int, boolean_t,
3241     ip_stack_t *);
3242 extern ire_t	*conn_set_outgoing_ill(conn_t *, ire_t *, ill_t **);
3243 extern int	ipsec_req_from_conn(conn_t *, ipsec_req_t *, int);
3244 extern int	ip_snmp_get(queue_t *q, mblk_t *mctl, int level);
3245 extern int	ip_snmp_set(queue_t *q, int, int, uchar_t *, int);
3246 extern void	ip_process_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3247 extern void	ip_quiesce_conn(conn_t *);
3248 extern  void    ip_reprocess_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3249 extern void	ip_restart_optmgmt(ipsq_t *, queue_t *, mblk_t *, void *);
3250 extern void	ip_ioctl_finish(queue_t *, mblk_t *, int, int, ipsq_t *);
3251 
3252 extern boolean_t ip_cmpbuf(const void *, uint_t, boolean_t, const void *,
3253     uint_t);
3254 extern boolean_t ip_allocbuf(void **, uint_t *, boolean_t, const void *,
3255     uint_t);
3256 extern void	ip_savebuf(void **, uint_t *, boolean_t, const void *, uint_t);
3257 
3258 extern boolean_t	ipsq_pending_mp_cleanup(ill_t *, conn_t *);
3259 extern void	conn_ioctl_cleanup(conn_t *);
3260 extern ill_t	*conn_get_held_ill(conn_t *, ill_t **, int *);
3261 extern ill_t	*ip_newroute_get_dst_ill(ill_t *);
3262 
3263 struct multidata_s;
3264 struct pdesc_s;
3265 
3266 extern mblk_t	*ip_mdinfo_alloc(ill_mdt_capab_t *);
3267 extern mblk_t	*ip_mdinfo_return(ire_t *, conn_t *, char *, ill_mdt_capab_t *);
3268 extern mblk_t	*ip_lsoinfo_alloc(ill_lso_capab_t *);
3269 extern mblk_t	*ip_lsoinfo_return(ire_t *, conn_t *, char *,
3270     ill_lso_capab_t *);
3271 extern uint_t	ip_md_cksum(struct pdesc_s *, int, uint_t);
3272 extern boolean_t ip_md_addr_attr(struct multidata_s *, struct pdesc_s *,
3273 			const mblk_t *);
3274 extern boolean_t ip_md_hcksum_attr(struct multidata_s *, struct pdesc_s *,
3275 			uint32_t, uint32_t, uint32_t, uint32_t);
3276 extern boolean_t ip_md_zcopy_attr(struct multidata_s *, struct pdesc_s *,
3277 			uint_t);
3278 extern mblk_t	*ip_unbind(queue_t *, mblk_t *);
3279 
3280 extern phyint_t *phyint_lookup_group(char *, boolean_t, ip_stack_t *);
3281 extern phyint_t *phyint_lookup_group_ifindex(uint_t, ip_stack_t *);
3282 
3283 extern void tnet_init(void);
3284 extern void tnet_fini(void);
3285 
3286 /* Hooks for CGTP (multirt routes) filtering module */
3287 #define	CGTP_FILTER_REV_1	1
3288 #define	CGTP_FILTER_REV_2	2
3289 #define	CGTP_FILTER_REV_3	3
3290 #define	CGTP_FILTER_REV		CGTP_FILTER_REV_3
3291 
3292 /* cfo_filter and cfo_filter_v6 hooks return values */
3293 #define	CGTP_IP_PKT_NOT_CGTP	0
3294 #define	CGTP_IP_PKT_PREMIUM	1
3295 #define	CGTP_IP_PKT_DUPLICATE	2
3296 
3297 /* Version 3 of the filter interface */
3298 typedef struct cgtp_filter_ops {
3299 	int	cfo_filter_rev;			/* CGTP_FILTER_REV_3 */
3300 	int	(*cfo_change_state)(netstackid_t, int);
3301 	int	(*cfo_add_dest_v4)(netstackid_t, ipaddr_t, ipaddr_t,
3302 		    ipaddr_t, ipaddr_t);
3303 	int	(*cfo_del_dest_v4)(netstackid_t, ipaddr_t, ipaddr_t);
3304 	int	(*cfo_add_dest_v6)(netstackid_t, in6_addr_t *, in6_addr_t *,
3305 		    in6_addr_t *, in6_addr_t *);
3306 	int	(*cfo_del_dest_v6)(netstackid_t, in6_addr_t *, in6_addr_t *);
3307 	int	(*cfo_filter)(netstackid_t, uint_t, mblk_t *);
3308 	int	(*cfo_filter_v6)(netstackid_t, uint_t, ip6_t *,
3309 		    ip6_frag_t *);
3310 } cgtp_filter_ops_t;
3311 
3312 #define	CGTP_MCAST_SUCCESS	1
3313 
3314 /*
3315  * The separate CGTP module needs this global symbol so that it
3316  * can check the version and determine whether to use the old or the new
3317  * version of the filtering interface.
3318  */
3319 extern int	ip_cgtp_filter_rev;
3320 
3321 extern int	ip_cgtp_filter_supported(void);
3322 extern int	ip_cgtp_filter_register(netstackid_t, cgtp_filter_ops_t *);
3323 extern int	ip_cgtp_filter_unregister(netstackid_t);
3324 extern int	ip_cgtp_filter_is_registered(netstackid_t);
3325 
3326 /* Flags for ire_multirt_lookup() */
3327 
3328 #define	MULTIRT_USESTAMP	0x0001
3329 #define	MULTIRT_SETSTAMP	0x0002
3330 #define	MULTIRT_CACHEGW		0x0004
3331 
3332 /* Debug stuff for multirt route resolution. */
3333 #if defined(DEBUG) && !defined(__lint)
3334 /* Our "don't send, rather drop" flag. */
3335 #define	MULTIRT_DEBUG_FLAG	0x8000
3336 
3337 #define	MULTIRT_TRACE(x)	ip2dbg(x)
3338 
3339 #define	MULTIRT_DEBUG_TAG(mblk)	\
3340 	do { \
3341 		ASSERT(mblk != NULL); \
3342 		MULTIRT_TRACE(("%s[%d]: tagging mblk %p, tag was %d\n", \
3343 		__FILE__, __LINE__, \
3344 		(void *)(mblk), (mblk)->b_flag & MULTIRT_DEBUG_FLAG)); \
3345 		(mblk)->b_flag |= MULTIRT_DEBUG_FLAG; \
3346 	} while (0)
3347 
3348 #define	MULTIRT_DEBUG_UNTAG(mblk) \
3349 	do { \
3350 		ASSERT(mblk != NULL); \
3351 		MULTIRT_TRACE(("%s[%d]: untagging mblk %p, tag was %d\n", \
3352 		__FILE__, __LINE__, \
3353 		(void *)(mblk), (mblk)->b_flag & MULTIRT_DEBUG_FLAG)); \
3354 		(mblk)->b_flag &= ~MULTIRT_DEBUG_FLAG; \
3355 	} while (0)
3356 
3357 #define	MULTIRT_DEBUG_TAGGED(mblk) \
3358 	(((mblk)->b_flag & MULTIRT_DEBUG_FLAG) ? B_TRUE : B_FALSE)
3359 #else
3360 #define	MULTIRT_DEBUG_TAG(mblk)		ASSERT(mblk != NULL)
3361 #define	MULTIRT_DEBUG_UNTAG(mblk)	ASSERT(mblk != NULL)
3362 #define	MULTIRT_DEBUG_TAGGED(mblk)	B_FALSE
3363 #endif
3364 
3365 /*
3366  * Per-ILL Multidata Transmit capabilities.
3367  */
3368 struct ill_mdt_capab_s {
3369 	uint_t ill_mdt_version;  /* interface version */
3370 	uint_t ill_mdt_on;	 /* on/off switch for MDT on this ILL */
3371 	uint_t ill_mdt_hdr_head; /* leading header fragment extra space */
3372 	uint_t ill_mdt_hdr_tail; /* trailing header fragment extra space */
3373 	uint_t ill_mdt_max_pld;	 /* maximum payload buffers per Multidata */
3374 	uint_t ill_mdt_span_limit; /* maximum payload span per packet */
3375 };
3376 
3377 struct ill_hcksum_capab_s {
3378 	uint_t	ill_hcksum_version;	/* interface version */
3379 	uint_t	ill_hcksum_txflags;	/* capabilities on transmit */
3380 };
3381 
3382 struct ill_zerocopy_capab_s {
3383 	uint_t	ill_zerocopy_version;	/* interface version */
3384 	uint_t	ill_zerocopy_flags;	/* capabilities */
3385 };
3386 
3387 struct ill_lso_capab_s {
3388 	uint_t	ill_lso_version;	/* interface version */
3389 	uint_t	ill_lso_on;		/* on/off switch for LSO on this ILL */
3390 	uint_t	ill_lso_flags;		/* capabilities */
3391 	uint_t	ill_lso_max;		/* maximum size of payload */
3392 };
3393 
3394 /* Possible ill_states */
3395 #define	ILL_RING_INPROC		3	/* Being assigned to squeue */
3396 #define	ILL_RING_INUSE		2	/* Already Assigned to Rx Ring */
3397 #define	ILL_RING_BEING_FREED	1	/* Being Unassigned */
3398 #define	ILL_RING_FREE		0	/* Available to be assigned to Ring */
3399 
3400 #define	ILL_MAX_RINGS		256	/* Max num of rx rings we can manage */
3401 #define	ILL_POLLING		0x01	/* Polling in use */
3402 
3403 /*
3404  * These functions pointer types are exported by the mac/dls layer.
3405  * we need to duplicate the definitions here because we cannot
3406  * include mac/dls header files here.
3407  */
3408 typedef void	(*ip_mac_blank_t)(void *, time_t, uint_t);
3409 typedef void	(*ip_dld_tx_t)(void *, mblk_t *);
3410 
3411 typedef void	(*ip_dls_chg_soft_ring_t)(void *, int);
3412 typedef void	(*ip_dls_bind_t)(void *, processorid_t);
3413 typedef void	(*ip_dls_unbind_t)(void *);
3414 
3415 struct ill_rx_ring {
3416 	ip_mac_blank_t		rr_blank; /* Driver interrupt blanking func */
3417 	void			*rr_handle; /* Handle for Rx ring */
3418 	squeue_t		*rr_sqp; /* Squeue the ring is bound to */
3419 	ill_t			*rr_ill; /* back pointer to ill */
3420 	clock_t			rr_poll_time; /* Last lbolt polling was used */
3421 	uint32_t		rr_poll_state; /* polling state flags */
3422 	uint32_t		rr_max_blank_time; /* Max interrupt blank */
3423 	uint32_t		rr_min_blank_time; /* Min interrupt blank */
3424 	uint32_t		rr_max_pkt_cnt; /* Max pkts before interrupt */
3425 	uint32_t		rr_min_pkt_cnt; /* Mix pkts before interrupt */
3426 	uint32_t		rr_normal_blank_time; /* Normal intr freq */
3427 	uint32_t		rr_normal_pkt_cnt; /* Normal intr pkt cnt */
3428 	uint32_t		rr_ring_state; /* State of this ring */
3429 };
3430 
3431 struct ill_dls_capab_s {
3432 	ip_dld_tx_t		ill_tx;		/* Driver Tx routine */
3433 	void			*ill_tx_handle;	/* Driver Tx handle */
3434 	ip_dls_chg_soft_ring_t	ill_dls_change_status;
3435 						/* change soft ring fanout */
3436 	ip_dls_bind_t		ill_dls_bind;	/* to add CPU affinity */
3437 	ip_dls_unbind_t		ill_dls_unbind;	/* remove CPU affinity */
3438 	ill_rx_ring_t		*ill_ring_tbl; /* Ring to Sqp mapping table */
3439 	uint_t			ill_dls_soft_ring_cnt; /* Number of soft ring */
3440 	conn_t			*ill_unbind_conn; /* Conn used during unplumb */
3441 };
3442 
3443 /*
3444  * IP squeues exports
3445  */
3446 extern int 		ip_squeue_profile;
3447 extern int 		ip_squeue_bind;
3448 extern boolean_t 	ip_squeue_fanout;
3449 extern boolean_t	ip_squeue_soft_ring;
3450 extern uint_t		ip_threads_per_cpu;
3451 extern uint_t		ip_squeues_per_cpu;
3452 extern uint_t		ip_soft_rings_cnt;
3453 
3454 typedef struct squeue_set_s {
3455 	kmutex_t	sqs_lock;
3456 	struct squeue_s	**sqs_list;
3457 	int		sqs_size;
3458 	int		sqs_max_size;
3459 	processorid_t	sqs_bind;
3460 } squeue_set_t;
3461 
3462 #define	IP_SQUEUE_GET(hint) 						\
3463 	((!ip_squeue_fanout) ?	(CPU->cpu_squeue_set->sqs_list[0]) :	\
3464 		ip_squeue_random(hint))
3465 
3466 typedef void (*squeue_func_t)(squeue_t *, mblk_t *, sqproc_t, void *, uint8_t);
3467 
3468 extern void ip_squeue_init(void (*)(squeue_t *));
3469 extern squeue_t	*ip_squeue_random(uint_t);
3470 extern squeue_t *ip_squeue_get(ill_rx_ring_t *);
3471 extern int ip_squeue_bind_set(queue_t *, mblk_t *, char *, caddr_t, cred_t *);
3472 extern void ip_squeue_clean_all(ill_t *);
3473 extern void ip_soft_ring_assignment(ill_t *, ill_rx_ring_t *,
3474     mblk_t *, struct mac_header_info_s *);
3475 
3476 extern void ip_resume_tcp_bind(void *, mblk_t *, void *);
3477 extern void tcp_wput(queue_t *, mblk_t *);
3478 
3479 extern int	ip_fill_mtuinfo(struct in6_addr *, in_port_t,
3480 	struct ip6_mtuinfo *, netstack_t *);
3481 extern	ipif_t *conn_get_held_ipif(conn_t *, ipif_t **, int *);
3482 
3483 typedef void    (*ipsq_func_t)(ipsq_t *, queue_t *, mblk_t *, void *);
3484 
3485 /*
3486  * Squeue tags. Tags only need to be unique when the callback function is the
3487  * same to distinguish between different calls, but we use unique tags for
3488  * convenience anyway.
3489  */
3490 #define	SQTAG_IP_INPUT			1
3491 #define	SQTAG_TCP_INPUT_ICMP_ERR	2
3492 #define	SQTAG_TCP6_INPUT_ICMP_ERR	3
3493 #define	SQTAG_IP_TCP_INPUT		4
3494 #define	SQTAG_IP6_TCP_INPUT		5
3495 #define	SQTAG_IP_TCP_CLOSE		6
3496 #define	SQTAG_TCP_OUTPUT		7
3497 #define	SQTAG_TCP_TIMER			8
3498 #define	SQTAG_TCP_TIMEWAIT		9
3499 #define	SQTAG_TCP_ACCEPT_FINISH		10
3500 #define	SQTAG_TCP_ACCEPT_FINISH_Q0	11
3501 #define	SQTAG_TCP_ACCEPT_PENDING	12
3502 #define	SQTAG_TCP_LISTEN_DISCON		13
3503 #define	SQTAG_TCP_CONN_REQ_1		14
3504 #define	SQTAG_TCP_EAGER_BLOWOFF		15
3505 #define	SQTAG_TCP_EAGER_CLEANUP		16
3506 #define	SQTAG_TCP_EAGER_CLEANUP_Q0	17
3507 #define	SQTAG_TCP_CONN_IND		18
3508 #define	SQTAG_TCP_RSRV			19
3509 #define	SQTAG_TCP_ABORT_BUCKET		20
3510 #define	SQTAG_TCP_REINPUT		21
3511 #define	SQTAG_TCP_REINPUT_EAGER		22
3512 #define	SQTAG_TCP_INPUT_MCTL		23
3513 #define	SQTAG_TCP_RPUTOTHER		24
3514 #define	SQTAG_IP_PROTO_AGAIN		25
3515 #define	SQTAG_IP_FANOUT_TCP		26
3516 #define	SQTAG_IPSQ_CLEAN_RING		27
3517 #define	SQTAG_TCP_WPUT_OTHER		28
3518 #define	SQTAG_TCP_CONN_REQ_UNBOUND	29
3519 #define	SQTAG_TCP_SEND_PENDING		30
3520 #define	SQTAG_BIND_RETRY		31
3521 #define	SQTAG_UDP_FANOUT		32
3522 #define	SQTAG_UDP_INPUT			33
3523 #define	SQTAG_UDP_WPUT			34
3524 #define	SQTAG_UDP_OUTPUT		35
3525 #define	SQTAG_TCP_KSSL_INPUT		36
3526 #define	SQTAG_TCP_DROP_Q0		37
3527 #define	SQTAG_TCP_CONN_REQ_2		38
3528 
3529 #define	NOT_OVER_IP(ip_wq)	\
3530 	(ip_wq->q_next != NULL ||	\
3531 	    (ip_wq->q_qinfo->qi_minfo->mi_idname) == NULL ||	\
3532 	    strcmp(ip_wq->q_qinfo->qi_minfo->mi_idname,	\
3533 	    IP_MOD_NAME) != 0 ||	\
3534 	    ip_wq->q_qinfo->qi_minfo->mi_idnum != IP_MOD_ID)
3535 
3536 #endif	/* _KERNEL */
3537 
3538 #ifdef	__cplusplus
3539 }
3540 #endif
3541 
3542 #endif	/* _INET_IP_H */
3543