xref: /titanic_50/usr/src/uts/common/inet/ip.h (revision 1d03c31e0733adea0edef54f0d5d2ea9639ecd2a)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 /* Copyright (c) 1990 Mentat Inc. */
27 
28 #ifndef	_INET_IP_H
29 #define	_INET_IP_H
30 
31 #pragma ident	"%Z%%M%	%I%	%E% SMI"
32 
33 #ifdef	__cplusplus
34 extern "C" {
35 #endif
36 
37 #include <sys/isa_defs.h>
38 #include <sys/types.h>
39 #include <inet/mib2.h>
40 #include <inet/nd.h>
41 #include <sys/atomic.h>
42 #include <net/if_dl.h>
43 #include <net/if.h>
44 #include <netinet/ip.h>
45 #include <netinet/igmp.h>
46 #include <sys/neti.h>
47 #include <sys/hook.h>
48 #include <sys/hook_event.h>
49 #include <sys/hook_impl.h>
50 #include <inet/ip_stack.h>
51 
52 #ifdef _KERNEL
53 #include <netinet/ip6.h>
54 #include <sys/avl.h>
55 #include <sys/vmem.h>
56 #include <sys/squeue.h>
57 #include <net/route.h>
58 #include <sys/systm.h>
59 #include <sys/multidata.h>
60 #include <net/radix.h>
61 #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 /* Minor numbers */
91 #define	IPV4_MINOR	0
92 #define	IPV6_MINOR	1
93 #define	TCP_MINOR	2
94 #define	TCP_MINOR6	3
95 
96 #ifndef	_IPADDR_T
97 #define	_IPADDR_T
98 typedef uint32_t ipaddr_t;
99 #endif
100 
101 /* Number of bits in an address */
102 #define	IP_ABITS		32
103 #define	IPV6_ABITS		128
104 
105 #define	IP_HOST_MASK		(ipaddr_t)0xffffffffU
106 
107 #define	IP_CSUM(mp, off, sum)		(~ip_cksum(mp, off, sum) & 0xFFFF)
108 #define	IP_CSUM_PARTIAL(mp, off, sum)	ip_cksum(mp, off, sum)
109 #define	IP_BCSUM_PARTIAL(bp, len, sum)	bcksum(bp, len, sum)
110 #define	IP_MD_CSUM(pd, off, sum)	(~ip_md_cksum(pd, off, sum) & 0xffff)
111 #define	IP_MD_CSUM_PARTIAL(pd, off, sum) ip_md_cksum(pd, off, sum)
112 
113 /*
114  * Flag to IP write side to indicate that the appln has sent in a pre-built
115  * IP header. Stored in ipha_ident (which is otherwise zero).
116  */
117 #define	IP_HDR_INCLUDED			0xFFFF
118 
119 #define	ILL_FRAG_HASH_TBL_COUNT	((unsigned int)64)
120 #define	ILL_FRAG_HASH_TBL_SIZE	(ILL_FRAG_HASH_TBL_COUNT * sizeof (ipfb_t))
121 
122 #define	IPV4_ADDR_LEN			4
123 #define	IP_ADDR_LEN			IPV4_ADDR_LEN
124 #define	IP_ARP_PROTO_TYPE		0x0800
125 
126 #define	IPV4_VERSION			4
127 #define	IP_VERSION			IPV4_VERSION
128 #define	IP_SIMPLE_HDR_LENGTH_IN_WORDS	5
129 #define	IP_SIMPLE_HDR_LENGTH		20
130 #define	IP_MAX_HDR_LENGTH		60
131 
132 #define	IP_MAX_OPT_LENGTH (IP_MAX_HDR_LENGTH-IP_SIMPLE_HDR_LENGTH)
133 
134 #define	IP_MIN_MTU			(IP_MAX_HDR_LENGTH + 8)	/* 68 bytes */
135 
136 /*
137  * XXX IP_MAXPACKET is defined in <netinet/ip.h> as well. At some point the
138  * 2 files should be cleaned up to remove all redundant definitions.
139  */
140 #define	IP_MAXPACKET			65535
141 #define	IP_SIMPLE_HDR_VERSION \
142 	((IP_VERSION << 4) | IP_SIMPLE_HDR_LENGTH_IN_WORDS)
143 
144 #define	UDPH_SIZE			8
145 
146 /* Leave room for ip_newroute to tack on the src and target addresses */
147 #define	OK_RESOLVER_MP(mp)						\
148 	((mp) && ((mp)->b_wptr - (mp)->b_rptr) >= (2 * IP_ADDR_LEN))
149 
150 /*
151  * Constants and type definitions to support IP IOCTL commands
152  */
153 #define	IP_IOCTL			(('i'<<8)|'p')
154 #define	IP_IOC_IRE_DELETE		4
155 #define	IP_IOC_IRE_DELETE_NO_REPLY	5
156 #define	IP_IOC_IRE_ADVISE_NO_REPLY	6
157 #define	IP_IOC_RTS_REQUEST		7
158 
159 /* Common definitions used by IP IOCTL data structures */
160 typedef struct ipllcmd_s {
161 	uint_t	ipllc_cmd;
162 	uint_t	ipllc_name_offset;
163 	uint_t	ipllc_name_length;
164 } ipllc_t;
165 
166 /* IP IRE Change Command Structure. */
167 typedef struct ipic_s {
168 	ipllc_t	ipic_ipllc;
169 	uint_t	ipic_ire_type;
170 	uint_t	ipic_max_frag;
171 	uint_t	ipic_addr_offset;
172 	uint_t	ipic_addr_length;
173 	uint_t	ipic_mask_offset;
174 	uint_t	ipic_mask_length;
175 	uint_t	ipic_src_addr_offset;
176 	uint_t	ipic_src_addr_length;
177 	uint_t	ipic_ll_hdr_offset;
178 	uint_t	ipic_ll_hdr_length;
179 	uint_t	ipic_gateway_addr_offset;
180 	uint_t	ipic_gateway_addr_length;
181 	clock_t	ipic_rtt;
182 	uint32_t ipic_ssthresh;
183 	clock_t	ipic_rtt_sd;
184 	uchar_t ipic_ire_marks;
185 } ipic_t;
186 
187 #define	ipic_cmd		ipic_ipllc.ipllc_cmd
188 #define	ipic_ll_name_length	ipic_ipllc.ipllc_name_length
189 #define	ipic_ll_name_offset	ipic_ipllc.ipllc_name_offset
190 
191 /* IP IRE Delete Command Structure. */
192 typedef struct ipid_s {
193 	ipllc_t	ipid_ipllc;
194 	uint_t	ipid_ire_type;
195 	uint_t	ipid_addr_offset;
196 	uint_t	ipid_addr_length;
197 	uint_t	ipid_mask_offset;
198 	uint_t	ipid_mask_length;
199 } ipid_t;
200 
201 #define	ipid_cmd		ipid_ipllc.ipllc_cmd
202 
203 #ifdef _KERNEL
204 /*
205  * Temporary state for ip options parser.
206  */
207 typedef struct ipoptp_s
208 {
209 	uint8_t		*ipoptp_next;	/* next option to look at */
210 	uint8_t		*ipoptp_end;	/* end of options */
211 	uint8_t		*ipoptp_cur;	/* start of current option */
212 	uint8_t		ipoptp_len;	/* length of current option */
213 	uint32_t	ipoptp_flags;
214 } ipoptp_t;
215 
216 /*
217  * Flag(s) for ipoptp_flags
218  */
219 #define	IPOPTP_ERROR	0x00000001
220 #endif	/* _KERNEL */
221 
222 
223 /* Controls forwarding of IP packets, set via ndd */
224 #define	IP_FORWARD_NEVER	0
225 #define	IP_FORWARD_ALWAYS	1
226 
227 #define	WE_ARE_FORWARDING(ipst)	((ipst)->ips_ip_g_forward == IP_FORWARD_ALWAYS)
228 
229 #define	IPH_HDR_LENGTH(ipha)						\
230 	((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length & 0xF) << 2)
231 
232 #define	IPH_HDR_VERSION(ipha)						\
233 	((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length) >> 4)
234 
235 #ifdef _KERNEL
236 /*
237  * IP reassembly macros.  We hide starting and ending offsets in b_next and
238  * b_prev of messages on the reassembly queue.	The messages are chained using
239  * b_cont.  These macros are used in ip_reassemble() so we don't have to see
240  * the ugly casts and assignments.
241  * Note that the offsets are <= 64k i.e. a uint_t is sufficient to represent
242  * them.
243  */
244 #define	IP_REASS_START(mp)		((uint_t)(uintptr_t)((mp)->b_next))
245 #define	IP_REASS_SET_START(mp, u)	\
246 	((mp)->b_next = (mblk_t *)(uintptr_t)(u))
247 #define	IP_REASS_END(mp)		((uint_t)(uintptr_t)((mp)->b_prev))
248 #define	IP_REASS_SET_END(mp, u)		\
249 	((mp)->b_prev = (mblk_t *)(uintptr_t)(u))
250 
251 #define	IP_REASS_COMPLETE	0x1
252 #define	IP_REASS_PARTIAL	0x2
253 #define	IP_REASS_FAILED		0x4
254 
255 /*
256  * Test to determine whether this is a module instance of IP or a
257  * driver instance of IP.
258  */
259 #define	CONN_Q(q)	(WR(q)->q_next == NULL)
260 
261 #define	Q_TO_CONN(q)	((conn_t *)(q)->q_ptr)
262 #define	Q_TO_TCP(q)	(Q_TO_CONN((q))->conn_tcp)
263 #define	Q_TO_UDP(q)	(Q_TO_CONN((q))->conn_udp)
264 
265 /*
266  * The following two macros are used by IP to get the appropriate
267  * wq and rq for a conn. If it is a TCP conn, then we need
268  * tcp_wq/tcp_rq else, conn_wq/conn_rq. IP can use conn_wq and conn_rq
269  * from a conn directly if it knows that the conn is not TCP.
270  */
271 #define	CONNP_TO_WQ(connp)	\
272 	(IPCL_IS_TCP(connp) ? (connp)->conn_tcp->tcp_wq : (connp)->conn_wq)
273 
274 #define	CONNP_TO_RQ(connp)	RD(CONNP_TO_WQ(connp))
275 
276 #define	GRAB_CONN_LOCK(q)	{				\
277 	if (q != NULL && CONN_Q(q))				\
278 		mutex_enter(&(Q_TO_CONN(q))->conn_lock);	\
279 }
280 
281 #define	RELEASE_CONN_LOCK(q)	{				\
282 	if (q != NULL && CONN_Q(q))				\
283 		mutex_exit(&(Q_TO_CONN(q))->conn_lock);		\
284 }
285 
286 /* "Congestion controlled" protocol */
287 #define	IP_FLOW_CONTROLLED_ULP(p)   ((p) == IPPROTO_TCP || (p) == IPPROTO_SCTP)
288 
289 /*
290  * Complete the pending operation. Usually an ioctl. Can also
291  * be a bind or option management request that got enqueued
292  * in an ipsq_t. Called on completion of the operation.
293  */
294 #define	CONN_OPER_PENDING_DONE(connp)	{			\
295 	mutex_enter(&(connp)->conn_lock);			\
296 	(connp)->conn_oper_pending_ill = NULL;			\
297 	cv_broadcast(&(connp)->conn_refcv);			\
298 	mutex_exit(&(connp)->conn_lock);			\
299 	CONN_DEC_REF(connp);					\
300 }
301 
302 /* Get the credential of an IP queue of unknown type */
303 #define	GET_QUEUE_CRED(wq)						\
304 	((wq)->q_next ? (((ill_t *)(wq)->q_ptr)->ill_credp) \
305 	    : ((Q_TO_CONN((wq)))->conn_cred))
306 
307 /*
308  * Flags for the various ip_fanout_* routines.
309  */
310 #define	IP_FF_SEND_ICMP		0x01	/* Send an ICMP error */
311 #define	IP_FF_HDR_COMPLETE	0x02	/* Call ip_hdr_complete if error */
312 #define	IP_FF_CKSUM		0x04	/* Recompute ipha_cksum if error */
313 #define	IP_FF_RAWIP		0x08	/* Use rawip mib variable */
314 #define	IP_FF_SRC_QUENCH	0x10	/* OK to send ICMP_SOURCE_QUENCH */
315 #define	IP_FF_SYN_ADDIRE	0x20	/* Add IRE if TCP syn packet */
316 #define	IP_FF_IPINFO		0x80	/* Used for both V4 and V6 */
317 #define	IP_FF_SEND_SLLA		0x100	/* Send source link layer info ? */
318 #define	IPV6_REACHABILITY_CONFIRMATION	0x200	/* Flags for ip_xmit_v6 */
319 #define	IP_FF_NO_MCAST_LOOP	0x400	/* No multicasts for sending zone */
320 
321 /*
322  * Following flags are used by IPQoS to determine if policy processing is
323  * required.
324  */
325 #define	IP6_NO_IPPOLICY		0x800	/* Don't do IPQoS processing */
326 #define	IP6_IN_LLMCAST		0x1000	/* Multicast */
327 
328 #define	IP_FF_LOOPBACK		0x2000	/* Loopback fanout */
329 
330 #ifndef	IRE_DB_TYPE
331 #define	IRE_DB_TYPE	M_SIG
332 #endif
333 
334 #ifndef	IRE_DB_REQ_TYPE
335 #define	IRE_DB_REQ_TYPE	M_PCSIG
336 #endif
337 
338 #ifndef	IRE_ARPRESOLVE_TYPE
339 #define	IRE_ARPRESOLVE_TYPE	M_EVENT
340 #endif
341 
342 /*
343  * Values for squeue switch:
344  */
345 
346 #define	IP_SQUEUE_ENTER_NODRAIN	1
347 #define	IP_SQUEUE_ENTER	2
348 /*
349  * This is part of the interface between Transport provider and
350  * IP which can be used to set policy information. This is usually
351  * accompanied with O_T_BIND_REQ/T_BIND_REQ.ip_bind assumes that
352  * only IPSEC_POLICY_SET is there when it is found in the chain.
353  * The information contained is an struct ipsec_req_t. On success
354  * or failure, either the T_BIND_ACK or the T_ERROR_ACK is returned.
355  * IPSEC_POLICY_SET is never returned.
356  */
357 #define	IPSEC_POLICY_SET	M_SETOPTS
358 
359 #define	IRE_IS_LOCAL(ire)	((ire != NULL) && \
360 				((ire)->ire_type & (IRE_LOCAL | IRE_LOOPBACK)))
361 
362 #define	IRE_IS_TARGET(ire)	((ire != NULL) && \
363 				((ire)->ire_type != IRE_BROADCAST))
364 
365 /* IP Fragmentation Reassembly Header */
366 typedef struct ipf_s {
367 	struct ipf_s	*ipf_hash_next;
368 	struct ipf_s	**ipf_ptphn;	/* Pointer to previous hash next. */
369 	uint32_t	ipf_ident;	/* Ident to match. */
370 	uint8_t		ipf_protocol;	/* Protocol to match. */
371 	uchar_t		ipf_last_frag_seen : 1;	/* Last fragment seen ? */
372 	time_t		ipf_timestamp;	/* Reassembly start time. */
373 	mblk_t		*ipf_mp;	/* mblk we live in. */
374 	mblk_t		*ipf_tail_mp;	/* Frag queue tail pointer. */
375 	int		ipf_hole_cnt;	/* Number of holes (hard-case). */
376 	int		ipf_end;	/* Tail end offset (0 -> hard-case). */
377 	uint_t		ipf_gen;	/* Frag queue generation */
378 	size_t		ipf_count;	/* Count of bytes used by frag */
379 	uint_t		ipf_nf_hdr_len; /* Length of nonfragmented header */
380 	in6_addr_t	ipf_v6src;	/* IPv6 source address */
381 	in6_addr_t	ipf_v6dst;	/* IPv6 dest address */
382 	uint_t		ipf_prev_nexthdr_offset; /* Offset for nexthdr value */
383 	uint8_t		ipf_ecn;	/* ECN info for the fragments */
384 	uint8_t		ipf_num_dups;	/* Number of times dup frags recvd */
385 	uint16_t	ipf_checksum_flags; /* Hardware checksum flags */
386 	uint32_t	ipf_checksum;	/* Partial checksum of fragment data */
387 } ipf_t;
388 
389 #define	ipf_src	V4_PART_OF_V6(ipf_v6src)
390 #define	ipf_dst	V4_PART_OF_V6(ipf_v6dst)
391 
392 typedef enum {
393 	IB_PKT =  0x01,
394 	OB_PKT = 0x02
395 } ip_pkt_t;
396 
397 #define	UPDATE_IB_PKT_COUNT(ire)\
398 	{ \
399 	(ire)->ire_ib_pkt_count++; \
400 	if ((ire)->ire_ipif != NULL) { \
401 		/* \
402 		 * forwarding packet \
403 		 */ \
404 		if ((ire)->ire_type & (IRE_LOCAL|IRE_BROADCAST)) \
405 			atomic_add_32(&(ire)->ire_ipif->ipif_ib_pkt_count, 1);\
406 		else \
407 			atomic_add_32(&(ire)->ire_ipif->ipif_fo_pkt_count, 1);\
408 	} \
409 	}
410 #define	UPDATE_OB_PKT_COUNT(ire)\
411 	{ \
412 	(ire)->ire_ob_pkt_count++;\
413 	if ((ire)->ire_ipif != NULL) { \
414 		atomic_add_32(&(ire)->ire_ipif->ipif_ob_pkt_count, 1); \
415 	} \
416 	}
417 
418 
419 #define	IP_RPUT_LOCAL(q, mp, ipha, ire, recv_ill) \
420 { \
421 	switch (ipha->ipha_protocol) { \
422 		case IPPROTO_UDP: \
423 			ip_udp_input(q, mp, ipha, ire, recv_ill); \
424 			break; \
425 		default: \
426 			ip_proto_input(q, mp, ipha, ire, recv_ill, B_FALSE); \
427 			break; \
428 	} \
429 }
430 
431 /*
432  * NCE_EXPIRED is TRUE when we have a non-permanent nce that was
433  * found to be REACHABLE more than ip_ire_arp_interval ms ago.
434  * This macro is used to age existing nce_t entries. The
435  * nce's will get cleaned up in the following circumstances:
436  * - ip_ire_trash_reclaim will free nce's using ndp_cache_reclaim
437  *    when memory is low,
438  * - ip_arp_news, when updates are received.
439  * - if the nce is NCE_EXPIRED(), it will deleted, so that a new
440  *   arp request will need to be triggered from an ND_INITIAL nce.
441  *
442  * Note that the nce state transition follows the pattern:
443  *	ND_INITIAL -> ND_INCOMPLETE -> ND_REACHABLE
444  * after which the nce is deleted when it has expired.
445  *
446  * nce_last is the timestamp that indicates when the nce_res_mp in the
447  * nce_t was last updated to a valid link-layer address.  nce_last gets
448  * modified/updated :
449  *  - when the nce is created
450  *  - every time we get a sane arp response for the nce.
451  */
452 #define	NCE_EXPIRED(nce, ipst)	(nce->nce_last > 0 &&	\
453 	    ((nce->nce_flags & NCE_F_PERMANENT) == 0) &&	\
454 	    ((TICK_TO_MSEC(lbolt64) - nce->nce_last) > 		\
455 		(ipst)->ips_ip_ire_arp_interval))
456 
457 #endif /* _KERNEL */
458 
459 /* ICMP types */
460 #define	ICMP_ECHO_REPLY			0
461 #define	ICMP_DEST_UNREACHABLE		3
462 #define	ICMP_SOURCE_QUENCH		4
463 #define	ICMP_REDIRECT			5
464 #define	ICMP_ECHO_REQUEST		8
465 #define	ICMP_ROUTER_ADVERTISEMENT	9
466 #define	ICMP_ROUTER_SOLICITATION	10
467 #define	ICMP_TIME_EXCEEDED		11
468 #define	ICMP_PARAM_PROBLEM		12
469 #define	ICMP_TIME_STAMP_REQUEST		13
470 #define	ICMP_TIME_STAMP_REPLY		14
471 #define	ICMP_INFO_REQUEST		15
472 #define	ICMP_INFO_REPLY			16
473 #define	ICMP_ADDRESS_MASK_REQUEST	17
474 #define	ICMP_ADDRESS_MASK_REPLY		18
475 
476 /* ICMP_TIME_EXCEEDED codes */
477 #define	ICMP_TTL_EXCEEDED		0
478 #define	ICMP_REASSEMBLY_TIME_EXCEEDED	1
479 
480 /* ICMP_DEST_UNREACHABLE codes */
481 #define	ICMP_NET_UNREACHABLE		0
482 #define	ICMP_HOST_UNREACHABLE		1
483 #define	ICMP_PROTOCOL_UNREACHABLE	2
484 #define	ICMP_PORT_UNREACHABLE		3
485 #define	ICMP_FRAGMENTATION_NEEDED	4
486 #define	ICMP_SOURCE_ROUTE_FAILED	5
487 #define	ICMP_DEST_NET_UNKNOWN		6
488 #define	ICMP_DEST_HOST_UNKNOWN		7
489 #define	ICMP_SRC_HOST_ISOLATED		8
490 #define	ICMP_DEST_NET_UNREACH_ADMIN	9
491 #define	ICMP_DEST_HOST_UNREACH_ADMIN	10
492 #define	ICMP_DEST_NET_UNREACH_TOS	11
493 #define	ICMP_DEST_HOST_UNREACH_TOS	12
494 
495 /* ICMP Header Structure */
496 typedef struct icmph_s {
497 	uint8_t		icmph_type;
498 	uint8_t		icmph_code;
499 	uint16_t	icmph_checksum;
500 	union {
501 		struct { /* ECHO request/response structure */
502 			uint16_t	u_echo_ident;
503 			uint16_t	u_echo_seqnum;
504 		} u_echo;
505 		struct { /* Destination unreachable structure */
506 			uint16_t	u_du_zero;
507 			uint16_t	u_du_mtu;
508 		} u_du;
509 		struct { /* Parameter problem structure */
510 			uint8_t		u_pp_ptr;
511 			uint8_t		u_pp_rsvd[3];
512 		} u_pp;
513 		struct { /* Redirect structure */
514 			ipaddr_t	u_rd_gateway;
515 		} u_rd;
516 	} icmph_u;
517 } icmph_t;
518 
519 #define	icmph_echo_ident	icmph_u.u_echo.u_echo_ident
520 #define	icmph_echo_seqnum	icmph_u.u_echo.u_echo_seqnum
521 #define	icmph_du_zero		icmph_u.u_du.u_du_zero
522 #define	icmph_du_mtu		icmph_u.u_du.u_du_mtu
523 #define	icmph_pp_ptr		icmph_u.u_pp.u_pp_ptr
524 #define	icmph_rd_gateway	icmph_u.u_rd.u_rd_gateway
525 
526 #define	ICMPH_SIZE	8
527 
528 /*
529  * Minimum length of transport layer header included in an ICMP error
530  * message for it to be considered valid.
531  */
532 #define	ICMP_MIN_TP_HDR_LEN	8
533 
534 /* Aligned IP header */
535 typedef struct ipha_s {
536 	uint8_t		ipha_version_and_hdr_length;
537 	uint8_t		ipha_type_of_service;
538 	uint16_t	ipha_length;
539 	uint16_t	ipha_ident;
540 	uint16_t	ipha_fragment_offset_and_flags;
541 	uint8_t		ipha_ttl;
542 	uint8_t		ipha_protocol;
543 	uint16_t	ipha_hdr_checksum;
544 	ipaddr_t	ipha_src;
545 	ipaddr_t	ipha_dst;
546 } ipha_t;
547 
548 #define	IPH_DF		0x4000	/* Don't fragment */
549 #define	IPH_MF		0x2000	/* More fragments to come */
550 #define	IPH_OFFSET	0x1FFF	/* Where the offset lives */
551 #define	IPH_FRAG_HDR	0x8000	/* IPv6 don't fragment bit */
552 
553 /* ECN code points for IPv4 TOS byte and IPv6 traffic class octet. */
554 #define	IPH_ECN_NECT	0x0	/* Not ECN-Capabable Transport */
555 #define	IPH_ECN_ECT1	0x1	/* ECN-Capable Transport, ECT(1) */
556 #define	IPH_ECN_ECT0	0x2	/* ECN-Capable Transport, ECT(0) */
557 #define	IPH_ECN_CE	0x3	/* ECN-Congestion Experienced (CE) */
558 
559 /* IP Mac info structure */
560 typedef struct ip_m_s {
561 	t_uscalar_t	ip_m_mac_type;	/* From <sys/dlpi.h> */
562 	int		ip_m_type;	/* From <net/if_types.h> */
563 	boolean_t	(*ip_m_v4mapinfo)(uint_t, uint8_t *, uint8_t *,
564 			    uint32_t *, ipaddr_t *);
565 	boolean_t	(*ip_m_v6mapinfo)(uint_t, uint8_t *, uint8_t *,
566 			    uint32_t *, in6_addr_t *);
567 	boolean_t	(*ip_m_v6intfid)(uint_t, uint8_t *, in6_addr_t *);
568 } ip_m_t;
569 
570 /*
571  * The following functions attempt to reduce the link layer dependency
572  * of the IP stack. The current set of link specific operations are:
573  * a. map from IPv4 class D (224.0/4) multicast address range to the link
574  * layer multicast address range.
575  * b. map from IPv6 multicast address range (ff00::/8) to the link
576  * layer multicast address range.
577  * c. derive the default IPv6 interface identifier from the link layer
578  * address.
579  */
580 #define	MEDIA_V4MINFO(ip_m, plen, bphys, maddr, hwxp, v4ptr) \
581 	(((ip_m)->ip_m_v4mapinfo != NULL) && \
582 	(*(ip_m)->ip_m_v4mapinfo)(plen, bphys, maddr, hwxp, v4ptr))
583 #define	MEDIA_V6INTFID(ip_m, plen, phys, v6ptr) \
584 	(((ip_m)->ip_m_v6intfid != NULL) && \
585 	(*(ip_m)->ip_m_v6intfid)(plen, phys, v6ptr))
586 #define	MEDIA_V6MINFO(ip_m, plen, bphys, maddr, hwxp, v6ptr) \
587 	(((ip_m)->ip_m_v6mapinfo != NULL) && \
588 	(*(ip_m)->ip_m_v6mapinfo)(plen, bphys, maddr, hwxp, v6ptr))
589 
590 /* Router entry types */
591 #define	IRE_BROADCAST		0x0001	/* Route entry for broadcast address */
592 #define	IRE_DEFAULT		0x0002	/* Route entry for default gateway */
593 #define	IRE_LOCAL		0x0004	/* Route entry for local address */
594 #define	IRE_LOOPBACK		0x0008	/* Route entry for loopback address */
595 #define	IRE_PREFIX		0x0010	/* Route entry for prefix routes */
596 #define	IRE_CACHE		0x0020	/* Cached Route entry */
597 #define	IRE_IF_NORESOLVER	0x0040	/* Route entry for local interface */
598 					/* net without any address mapping. */
599 #define	IRE_IF_RESOLVER		0x0080	/* Route entry for local interface */
600 					/* net with resolver. */
601 #define	IRE_HOST		0x0100	/* Host route entry */
602 #define	IRE_HOST_REDIRECT	0x0200	/* only used for T_SVR4_OPTMGMT_REQ */
603 
604 #define	IRE_INTERFACE		(IRE_IF_NORESOLVER | IRE_IF_RESOLVER)
605 #define	IRE_OFFSUBNET		(IRE_DEFAULT | IRE_PREFIX | IRE_HOST)
606 #define	IRE_CACHETABLE		(IRE_CACHE | IRE_BROADCAST | IRE_LOCAL | \
607 				IRE_LOOPBACK)
608 #define	IRE_FORWARDTABLE	(IRE_INTERFACE | IRE_OFFSUBNET)
609 
610 /*
611  * If an IRE is marked with IRE_MARK_CONDEMNED, the last walker of
612  * the bucket should delete this IRE from this bucket.
613  */
614 #define	IRE_MARK_CONDEMNED	0x0001
615 /*
616  * If a broadcast IRE is marked with IRE_MARK_NORECV, ip_rput will drop the
617  * broadcast packets received on that interface. This is marked only
618  * on broadcast ires. Employed by IPMP, where we have multiple NICs on the
619  * same subnet receiving the same broadcast packet.
620  */
621 #define	IRE_MARK_NORECV		0x0002
622 /*
623  * IRE_CACHE marked this way won't be returned by ire_cache_lookup. Need
624  * to look specifically using MATCH_IRE_MARK_HIDDEN. Used by IPMP.
625  */
626 #define	IRE_MARK_HIDDEN		0x0004	/* Typically Used by in.mpathd */
627 
628 /*
629  * ire with IRE_MARK_NOADD is created in ip_newroute_ipif, when outgoing
630  * interface is specified by IP_XMIT_IF socket option. This ire is not
631  * added in IRE_CACHE.
632  */
633 #define	IRE_MARK_NOADD		0x0008	/* Mark not to add ire in cache */
634 
635 /*
636  * IRE marked with IRE_MARK_TEMPORARY means that this IRE has been used
637  * either for forwarding a packet or has not been used for sending
638  * traffic on TCP connections terminated on this system.  In both
639  * cases, this IRE is the first to go when IRE is being cleaned up.
640  */
641 #define	IRE_MARK_TEMPORARY	0x0010
642 
643 /*
644  * IRE marked with IRE_MARK_USESRC_CHECK means that while adding an IRE with
645  * this mark, additional atomic checks need to be performed. For eg: by the
646  * time an IRE_CACHE is created, sent up to ARP and then comes back to IP; the
647  * usesrc grouping could have changed in which case we want to fail adding
648  * the IRE_CACHE entry
649  */
650 #define	IRE_MARK_USESRC_CHECK	0x0020
651 
652 /*
653  * IRE_MARK_PRIVATE_ADDR is used for IP_NEXTHOP. When IP_NEXTHOP is set, the
654  * routing table lookup for the destination is bypassed and the packet is
655  * sent directly to the specified nexthop. The associated IRE_CACHE entries
656  * should be marked with IRE_MARK_PRIVATE_ADDR flag so that they don't show up
657  * in regular ire cache lookups.
658  */
659 #define	IRE_MARK_PRIVATE_ADDR	0x0040
660 
661 /*
662  * When we send an ARP resolution query for the nexthop gateway's ire,
663  * we use esballoc to create the ire_t in the AR_ENTRY_QUERY mblk
664  * chain, and mark its ire_marks with IRE_MARK_UNCACHED. This flag
665  * indicates that information from ARP has not been transferred to a
666  * permanent IRE_CACHE entry. The flag is reset only when the
667  * information is successfully transferred to an ire_cache entry (in
668  * ire_add()). Attempting to free the AR_ENTRY_QUERY mblk chain prior
669  * to ire_add (e.g., from arp, or from ip`ip_wput_nondata) will
670  * require that the resources (incomplete ire_cache and/or nce) must
671  * be cleaned up. The free callback routine (ire_freemblk()) checks
672  * for IRE_MARK_UNCACHED to see if any resources that are pinned down
673  * will need to be cleaned up or not.
674  */
675 
676 #define	IRE_MARK_UNCACHED	0x0080
677 
678 /*
679  * The comment below (and for other netstack_t references) refers
680  * to the fact that we only do netstack_hold in particular cases,
681  * such as the references from open streams (ill_t and conn_t's
682  * pointers). Internally within IP we rely on IP's ability to cleanup e.g.
683  * ire_t's when an ill goes away.
684  */
685 typedef struct ire_expire_arg_s {
686 	int		iea_flush_flag;
687 	ip_stack_t	*iea_ipst;	/* Does not have a netstack_hold */
688 } ire_expire_arg_t;
689 
690 /* Flags with ire_expire routine */
691 #define	FLUSH_ARP_TIME		0x0001	/* ARP info potentially stale timer */
692 #define	FLUSH_REDIRECT_TIME	0x0002	/* Redirects potentially stale */
693 #define	FLUSH_MTU_TIME		0x0004	/* Include path MTU per RFC 1191 */
694 
695 /* Arguments to ire_flush_cache() */
696 #define	IRE_FLUSH_DELETE	0
697 #define	IRE_FLUSH_ADD		1
698 
699 /*
700  * Open/close synchronization flags.
701  * These are kept in a separate field in the conn and the synchronization
702  * depends on the atomic 32 bit access to that field.
703  */
704 #define	CONN_CLOSING		0x01	/* ip_close waiting for ip_wsrv */
705 #define	CONN_IPSEC_LOAD_WAIT	0x02	/* waiting for load */
706 #define	CONN_CONDEMNED		0x04	/* conn is closing, no more refs */
707 #define	CONN_INCIPIENT		0x08	/* conn not yet visible, no refs */
708 #define	CONN_QUIESCED		0x10	/* conn is now quiescent */
709 
710 /* Used to check connection state flags before caching the IRE */
711 #define	CONN_CACHE_IRE(connp)	\
712 	(!((connp)->conn_state_flags & (CONN_CLOSING|CONN_CONDEMNED)))
713 
714 /*
715  * Parameter to ip_output giving the identity of the caller.
716  * IP_WSRV means the packet was enqueued in the STREAMS queue
717  * due to flow control and is now being reprocessed in the context of
718  * the STREAMS service procedure, consequent to flow control relief.
719  * IRE_SEND means the packet is being reprocessed consequent to an
720  * ire cache creation and addition and this may or may not be happening
721  * in the service procedure context. Anything other than the above 2
722  * cases is identified as IP_WPUT. Most commonly this is the case of
723  * packets coming down from the application.
724  */
725 #ifdef _KERNEL
726 #define	IP_WSRV			1	/* Called from ip_wsrv */
727 #define	IP_WPUT			2	/* Called from ip_wput */
728 #define	IRE_SEND		3	/* Called from ire_send */
729 
730 /*
731  * Extra structures need for per-src-addr filtering (IGMPv3/MLDv2)
732  */
733 #define	MAX_FILTER_SIZE	64
734 
735 typedef struct slist_s {
736 	int		sl_numsrc;
737 	in6_addr_t	sl_addr[MAX_FILTER_SIZE];
738 } slist_t;
739 
740 /*
741  * Following struct is used to maintain retransmission state for
742  * a multicast group.  One rtx_state_t struct is an in-line field
743  * of the ilm_t struct; the slist_ts in the rtx_state_t struct are
744  * alloc'd as needed.
745  */
746 typedef struct rtx_state_s {
747 	uint_t		rtx_timer;	/* retrans timer */
748 	int		rtx_cnt;	/* retrans count */
749 	int		rtx_fmode_cnt;	/* retrans count for fmode change */
750 	slist_t		*rtx_allow;
751 	slist_t		*rtx_block;
752 } rtx_state_t;
753 
754 /*
755  * Used to construct list of multicast address records that will be
756  * sent in a single listener report.
757  */
758 typedef struct mrec_s {
759 	struct mrec_s	*mrec_next;
760 	uint8_t		mrec_type;
761 	uint8_t		mrec_auxlen;	/* currently unused */
762 	in6_addr_t	mrec_group;
763 	slist_t		mrec_srcs;
764 } mrec_t;
765 
766 /* Group membership list per upper conn */
767 /*
768  * XXX add ilg info for ifaddr/ifindex.
769  * XXX can we make ilg survive an ifconfig unplumb + plumb
770  * by setting the ipif/ill to NULL and recover that later?
771  *
772  * ilg_ipif is used by IPv4 as multicast groups are joined using an interface
773  * address (ipif).
774  * ilg_ill is used by IPv6 as multicast groups are joined using an interface
775  * index (phyint->phyint_ifindex).
776  * ilg_ill is NULL for IPv4 and ilg_ipif is NULL for IPv6.
777  *
778  * ilg records the state of multicast memberships of a socket end point.
779  * ilm records the state of multicast memberships with the driver and is
780  * maintained per interface.
781  *
782  * Notes :
783  *
784  * 1) There is no direct link between a given ilg and ilm. If the
785  *    application has joined a group G with ifindex I, we will have
786  *    an ilg with ilg_v6group and ilg_ill. There will be a corresponding
787  *    ilm with ilm_ill/ilm_v6addr recording the multicast membership.
788  *    To delete the membership,
789  *
790  *		a) Search for ilg matching on G and I with ilg_v6group
791  *		   and ilg_ill. Delete ilg_ill.
792  *		b) Search the corresponding ilm matching on G and I with
793  *		   ilm_v6addr and ilm_ill. Delete ilm.
794  *
795  *    In IPv4, the only difference is, we look using ipifs instead of
796  *    ills.
797  *
798  * 2) With IP multipathing, we want to keep receiving even after the
799  *    interface has failed. We do this by moving multicast memberships
800  *    to a new_ill within the group. This is acheived by sending
801  *    DL_DISABMULTI_REQS on ilg_ill/ilm_ill and sending DL_ENABMULTIREQS
802  *    on the new_ill and changing ilg_ill/ilm_ill to new_ill. But, we
803  *    need to be able to delete memberships which will still come down
804  *    with the ifindex of the old ill which is what the application
805  *    knows of. Thus we store the ilm_/ilg_orig_ifindex to keep track
806  *    of where we joined initially so that we can lookup even after we
807  *    moved the membership. It is also used for moving back the membership
808  *    when the old ill has been repaired. This is done by looking up for
809  *    ilms with ilm_orig_ifindex matching on the old ill's ifindex. Only
810  *    ilms actually move from old ill to new ill. ilgs don't move (just
811  *    the ilg_ill is changed when it moves) as it just records the state
812  *    of the application that has joined a group G where as ilm records
813  *    the state joined with the driver. Thus when we send DL_XXXMULTI_REQs
814  *    we also need to keep the ilm in the right ill.
815  *
816  *    In IPv4, as ipifs move from old ill to new_ill, ilgs and ilms move
817  *    implicitly as we use only ipifs in IPv4. Thus, one can always lookup
818  *    a given ilm/ilg even after it fails without the support of
819  *    orig_ifindex. We move ilms still to record the driver state as
820  *    mentioned above.
821  */
822 
823 /*
824  * The ilg_t and ilm_t members are protected by ipsq. They can be changed only
825  * by a thread executing in the ipsq. In other words add/delete of a
826  * multicast group has to execute in the ipsq.
827  */
828 #define	ILG_DELETED	0x1		/* ilg_flags */
829 typedef struct ilg_s {
830 	in6_addr_t	ilg_v6group;
831 	struct ipif_s	*ilg_ipif;	/* Logical interface we are member on */
832 	struct ill_s	*ilg_ill;	/* Used by IPv6 */
833 	int		ilg_orig_ifindex; /* Interface originally joined on */
834 	uint_t		ilg_flags;
835 	mcast_record_t	ilg_fmode;	/* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */
836 	slist_t		*ilg_filter;
837 } ilg_t;
838 
839 
840 /*
841  * Multicast address list entry for lower ill.
842  * ilm_ipif is used by IPv4 as multicast groups are joined using ipif.
843  * ilm_ill is used by IPv6 as multicast groups are joined using ill.
844  * ilm_ill is NULL for IPv4 and ilm_ipif is NULL for IPv6.
845  *
846  * The comment below (and for other netstack_t references) refers
847  * to the fact that we only do netstack_hold in particular cases,
848  * such as the references from open streams (ill_t and conn_t's
849  * pointers). Internally within IP we rely on IP's ability to cleanup e.g.
850  * ire_t's when an ill goes away.
851  */
852 #define	ILM_DELETED	0x1		/* ilm_flags */
853 typedef struct ilm_s {
854 	in6_addr_t	ilm_v6addr;
855 	int		ilm_refcnt;
856 	uint_t		ilm_timer;	/* IGMP/MLD query resp timer, in msec */
857 	struct ipif_s	*ilm_ipif;	/* Back pointer to ipif for IPv4 */
858 	struct ilm_s	*ilm_next;	/* Linked list for each ill */
859 	uint_t		ilm_state;	/* state of the membership */
860 	struct ill_s	*ilm_ill;	/* Back pointer to ill for IPv6 */
861 	int		ilm_orig_ifindex;  /* V6_MULTICAST_IF/ilm_ipif index */
862 	uint_t		ilm_flags;
863 	boolean_t	ilm_is_new;	/* new ilm */
864 	boolean_t	ilm_notify_driver; /* Need to notify the driver */
865 	zoneid_t	ilm_zoneid;
866 	int		ilm_no_ilg_cnt;	/* number of joins w/ no ilg */
867 	mcast_record_t	ilm_fmode;	/* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */
868 	slist_t		*ilm_filter;	/* source filter list */
869 	slist_t		*ilm_pendsrcs;	/* relevant src addrs for pending req */
870 	rtx_state_t	ilm_rtx;	/* SCR retransmission state */
871 	ip_stack_t	*ilm_ipst;	/* Does not have a netstack_hold */
872 } ilm_t;
873 
874 #define	ilm_addr	V4_PART_OF_V6(ilm_v6addr)
875 
876 /*
877  * ilm_walker_cleanup needs to execute when the ilm_walker_cnt goes down to
878  * zero. In addition it needs to block new walkers while it is unlinking ilm's
879  * from the list. Thus simple atomics for the ill_ilm_walker_cnt don't suffice.
880  */
881 #define	ILM_WALKER_HOLD(ill)    {               \
882 	mutex_enter(&(ill)->ill_lock);          \
883 	ill->ill_ilm_walker_cnt++;              \
884 	mutex_exit(&(ill)->ill_lock);           \
885 }
886 
887 #define	ILM_WALKER_RELE(ill)	{ 		\
888 	mutex_enter(&(ill)->ill_lock);		\
889 	(ill)->ill_ilm_walker_cnt--;		\
890 	if ((ill)->ill_ilm_walker_cnt == 0 && (ill)->ill_ilm_cleanup_reqd) \
891 		ilm_walker_cleanup(ill);	\
892 	mutex_exit(&(ill)->ill_lock);		\
893 }
894 
895 /*
896  * Soft reference to an IPsec SA.
897  *
898  * On relative terms, conn's can be persistant (living as long as the
899  * processes which create them), while SA's are ephemeral (dying when
900  * they hit their time-based or byte-based lifetimes).
901  *
902  * We could hold a hard reference to an SA from an ipsec_latch_t,
903  * but this would cause expired SA's to linger for a potentially
904  * unbounded time.
905  *
906  * Instead, we remember the hash bucket number and bucket generation
907  * in addition to the pointer.  The bucket generation is incremented on
908  * each deletion.
909  */
910 typedef struct ipsa_ref_s
911 {
912 	struct ipsa_s	*ipsr_sa;
913 	struct isaf_s	*ipsr_bucket;
914 	uint64_t	ipsr_gen;
915 } ipsa_ref_t;
916 
917 /*
918  * IPsec "latching" state.
919  *
920  * In the presence of IPsec policy, fully-bound conn's bind a connection
921  * to more than just the 5-tuple, but also a specific IPsec action and
922  * identity-pair.
923  *
924  * As an optimization, we also cache soft references to IPsec SA's
925  * here so that we can fast-path around most of the work needed for
926  * outbound IPsec SA selection.
927  *
928  * Were it not for TCP's detached connections, this state would be
929  * in-line in conn_t; instead, this is in a separate structure so it
930  * can be handed off to TCP when a connection is detached.
931  */
932 typedef struct ipsec_latch_s
933 {
934 	kmutex_t	ipl_lock;
935 	uint32_t	ipl_refcnt;
936 
937 	uint64_t	ipl_unique;
938 	struct ipsec_policy_s	*ipl_in_policy; /* latched policy (in) */
939 	struct ipsec_policy_s	*ipl_out_policy; /* latched policy (out) */
940 	struct ipsec_action_s	*ipl_in_action;	/* latched action (in) */
941 	struct ipsec_action_s	*ipl_out_action; /* latched action (out) */
942 	cred_t		*ipl_local_id;
943 	struct ipsid_s	*ipl_local_cid;
944 	struct ipsid_s	*ipl_remote_cid;
945 	unsigned int
946 			ipl_out_action_latched : 1,
947 			ipl_in_action_latched : 1,
948 			ipl_out_policy_latched : 1,
949 			ipl_in_policy_latched : 1,
950 
951 			ipl_ids_latched : 1,
952 
953 			ipl_pad_to_bit_31 : 27;
954 
955 	ipsa_ref_t	ipl_ref[2]; /* 0: ESP, 1: AH */
956 
957 } ipsec_latch_t;
958 
959 #define	IPLATCH_REFHOLD(ipl) { \
960 	atomic_add_32(&(ipl)->ipl_refcnt, 1);		\
961 	ASSERT((ipl)->ipl_refcnt != 0);			\
962 }
963 
964 #define	IPLATCH_REFRELE(ipl, ns) {				\
965 	ASSERT((ipl)->ipl_refcnt != 0);				\
966 	membar_exit();						\
967 	if (atomic_add_32_nv(&(ipl)->ipl_refcnt, -1) == 0)	\
968 		iplatch_free(ipl, ns);			\
969 }
970 
971 /*
972  * peer identity structure.
973  */
974 
975 typedef struct conn_s conn_t;
976 
977 /*
978  * The old IP client structure "ipc_t" is gone. All the data is stored in the
979  * connection structure "conn_t" now. The mapping of old and new fields looks
980  * like this:
981  *
982  * ipc_ulp			conn_ulp
983  * ipc_rq			conn_rq
984  * ipc_wq			conn_wq
985  *
986  * ipc_laddr			conn_src
987  * ipc_faddr			conn_rem
988  * ipc_v6laddr			conn_srcv6
989  * ipc_v6faddr			conn_remv6
990  *
991  * ipc_lport			conn_lport
992  * ipc_fport			conn_fport
993  * ipc_ports			conn_ports
994  *
995  * ipc_policy			conn_policy
996  * ipc_latch			conn_latch
997  *
998  * ipc_irc_lock			conn_lock
999  * ipc_ire_cache		conn_ire_cache
1000  *
1001  * ipc_state_flags		conn_state_flags
1002  * ipc_outgoing_ill		conn_outgoing_ill
1003  *
1004  * ipc_dontroute 		conn_dontroute
1005  * ipc_loopback 		conn_loopback
1006  * ipc_broadcast		conn_broadcast
1007  * ipc_reuseaddr		conn_reuseaddr
1008  *
1009  * ipc_multicast_loop		conn_multicast_loop
1010  * ipc_multi_router		conn_multi_router
1011  * ipc_priv_stream 		conn_priv_stream
1012  * ipc_draining 		conn_draining
1013  *
1014  * ipc_did_putbq		conn_did_putbq
1015  * ipc_unspec_src		conn_unspec_src
1016  * ipc_policy_cached		conn_policy_cached
1017  *
1018  * ipc_in_enforce_policy 	conn_in_enforce_policy
1019  * ipc_out_enforce_policy 	conn_out_enforce_policy
1020  * ipc_af_isv6			conn_af_isv6
1021  * ipc_pkt_isv6			conn_pkt_isv6
1022  *
1023  * ipc_ipv6_recvpktinfo		conn_ipv6_recvpktinfo
1024  *
1025  * ipc_ipv6_recvhoplimit	conn_ipv6_recvhoplimit
1026  * ipc_ipv6_recvhopopts		conn_ipv6_recvhopopts
1027  * ipc_ipv6_recvdstopts		conn_ipv6_recvdstopts
1028  *
1029  * ipc_ipv6_recvrthdr 		conn_ipv6_recvrthdr
1030  * ipc_ipv6_recvrtdstopts	conn_ipv6_recvrtdstopts
1031  * ipc_fully_bound		conn_fully_bound
1032  *
1033  * ipc_recvif			conn_recvif
1034  *
1035  * ipc_recvslla 		conn_recvslla
1036  * ipc_acking_unbind 		conn_acking_unbind
1037  * ipc_pad_to_bit_31 		conn_pad_to_bit_31
1038  *
1039  * ipc_xmit_if_ill		conn_xmit_if_ill
1040  * ipc_nofailover_ill		conn_nofailover_ill
1041  *
1042  * ipc_proto			conn_proto
1043  * ipc_incoming_ill		conn_incoming_ill
1044  * ipc_outgoing_pill		conn_outgoing_pill
1045  * ipc_pending_ill		conn_pending_ill
1046  * ipc_unbind_mp		conn_unbind_mp
1047  * ipc_ilg			conn_ilg
1048  * ipc_ilg_allocated		conn_ilg_allocated
1049  * ipc_ilg_inuse		conn_ilg_inuse
1050  * ipc_ilg_walker_cnt		conn_ilg_walker_cnt
1051  * ipc_refcv			conn_refcv
1052  * ipc_multicast_ipif		conn_multicast_ipif
1053  * ipc_multicast_ill		conn_multicast_ill
1054  * ipc_orig_bound_ifindex	conn_orig_bound_ifindex
1055  * ipc_orig_multicast_ifindex	conn_orig_multicast_ifindex
1056  * ipc_orig_xmit_ifindex	conn_orig_xmit_ifindex
1057  * ipc_drain_next		conn_drain_next
1058  * ipc_drain_prev		conn_drain_prev
1059  * ipc_idl			conn_idl
1060  */
1061 
1062 /*
1063  * This is used to match an inbound/outbound datagram with
1064  * policy.
1065  */
1066 
1067 typedef	struct ipsec_selector {
1068 	in6_addr_t	ips_local_addr_v6;
1069 	in6_addr_t	ips_remote_addr_v6;
1070 	uint16_t	ips_local_port;
1071 	uint16_t	ips_remote_port;
1072 	uint8_t		ips_icmp_type;
1073 	uint8_t		ips_icmp_code;
1074 	uint8_t		ips_protocol;
1075 	uint8_t		ips_isv4 : 1,
1076 			ips_is_icmp_inv_acq: 1;
1077 } ipsec_selector_t;
1078 
1079 /*
1080  * Note that we put v4 addresses in the *first* 32-bit word of the
1081  * selector rather than the last to simplify the prefix match/mask code
1082  * in spd.c
1083  */
1084 #define	ips_local_addr_v4 ips_local_addr_v6.s6_addr32[0]
1085 #define	ips_remote_addr_v4 ips_remote_addr_v6.s6_addr32[0]
1086 
1087 /* Values used in IP by IPSEC Code */
1088 #define		IPSEC_OUTBOUND		B_TRUE
1089 #define		IPSEC_INBOUND		B_FALSE
1090 
1091 /*
1092  * There are two variants in policy failures. The packet may come in
1093  * secure when not needed (IPSEC_POLICY_???_NOT_NEEDED) or it may not
1094  * have the desired level of protection (IPSEC_POLICY_MISMATCH).
1095  */
1096 #define	IPSEC_POLICY_NOT_NEEDED		0
1097 #define	IPSEC_POLICY_MISMATCH		1
1098 #define	IPSEC_POLICY_AUTH_NOT_NEEDED	2
1099 #define	IPSEC_POLICY_ENCR_NOT_NEEDED	3
1100 #define	IPSEC_POLICY_SE_NOT_NEEDED	4
1101 #define	IPSEC_POLICY_MAX		5	/* Always max + 1. */
1102 
1103 /*
1104  * Folowing macro is used whenever the code does not know whether there
1105  * is a M_CTL present in the front and it needs to examine the actual mp
1106  * i.e the IP header. As a M_CTL message could be in the front, this
1107  * extracts the packet into mp and the M_CTL mp into first_mp. If M_CTL
1108  * mp is not present, both first_mp and mp point to the same message.
1109  */
1110 #define	EXTRACT_PKT_MP(mp, first_mp, mctl_present)	\
1111 	(first_mp) = (mp);				\
1112 	if ((mp)->b_datap->db_type == M_CTL) {		\
1113 		(mp) = (mp)->b_cont;			\
1114 		(mctl_present) = B_TRUE;		\
1115 	} else {					\
1116 		(mctl_present) = B_FALSE;		\
1117 	}
1118 
1119 /*
1120  * Check with IPSEC inbound policy if
1121  *
1122  * 1) per-socket policy is present - indicated by conn_in_enforce_policy.
1123  * 2) Or if we have not cached policy on the conn and the global policy is
1124  *    non-empty.
1125  */
1126 #define	CONN_INBOUND_POLICY_PRESENT(connp, ipss)	\
1127 	((connp)->conn_in_enforce_policy ||		\
1128 	(!((connp)->conn_policy_cached) && 		\
1129 	(ipss)->ipsec_inbound_v4_policy_present))
1130 
1131 #define	CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)	\
1132 	((connp)->conn_in_enforce_policy ||		\
1133 	(!(connp)->conn_policy_cached &&		\
1134 	(ipss)->ipsec_inbound_v6_policy_present))
1135 
1136 #define	CONN_OUTBOUND_POLICY_PRESENT(connp, ipss)	\
1137 	((connp)->conn_out_enforce_policy ||		\
1138 	(!((connp)->conn_policy_cached) &&		\
1139 	(ipss)->ipsec_outbound_v4_policy_present))
1140 
1141 #define	CONN_OUTBOUND_POLICY_PRESENT_V6(connp, ipss)	\
1142 	((connp)->conn_out_enforce_policy ||		\
1143 	(!(connp)->conn_policy_cached &&		\
1144 	(ipss)->ipsec_outbound_v6_policy_present))
1145 
1146 /*
1147  * Information cached in IRE for upper layer protocol (ULP).
1148  *
1149  * Notice that ire_max_frag is not included in the iulp_t structure, which
1150  * it may seem that it should.  But ire_max_frag cannot really be cached.  It
1151  * is fixed for each interface.  For MTU found by PMTUd, we may want to cache
1152  * it.  But currently, we do not do that.
1153  */
1154 typedef struct iulp_s {
1155 	boolean_t	iulp_set;	/* Is any metric set? */
1156 	uint32_t	iulp_ssthresh;	/* Slow start threshold (TCP). */
1157 	clock_t		iulp_rtt;	/* Guestimate in millisecs. */
1158 	clock_t		iulp_rtt_sd;	/* Cached value of RTT variance. */
1159 	uint32_t	iulp_spipe;	/* Send pipe size. */
1160 	uint32_t	iulp_rpipe;	/* Receive pipe size. */
1161 	uint32_t	iulp_rtomax;	/* Max round trip timeout. */
1162 	uint32_t	iulp_sack;	/* Use SACK option (TCP)? */
1163 	uint32_t
1164 		iulp_tstamp_ok : 1,	/* Use timestamp option (TCP)? */
1165 		iulp_wscale_ok : 1,	/* Use window scale option (TCP)? */
1166 		iulp_ecn_ok : 1,	/* Enable ECN (for TCP)? */
1167 		iulp_pmtud_ok : 1,	/* Enable PMTUd? */
1168 
1169 		iulp_not_used : 28;
1170 } iulp_t;
1171 
1172 /* Zero iulp_t. */
1173 extern const iulp_t ire_uinfo_null;
1174 
1175 /*
1176  * The conn drain list structure.
1177  * The list is protected by idl_lock. Each conn_t inserted in the list
1178  * points back at this idl_t using conn_idl. IP primes the draining of the
1179  * conns queued in these lists, by qenabling the 1st conn of each list. This
1180  * occurs when STREAMS backenables ip_wsrv on the IP module. Each conn instance
1181  * of ip_wsrv successively qenables the next conn in the list.
1182  * idl_lock protects all other members of idl_t and conn_drain_next
1183  * and conn_drain_prev of conn_t. The conn_lock protects IPCF_DRAIN_DISABLED
1184  * flag of the conn_t and conn_idl.
1185  */
1186 typedef struct idl_s {
1187 	conn_t		*idl_conn;		/* Head of drain list */
1188 	kmutex_t	idl_lock;		/* Lock for this list */
1189 	conn_t		*idl_conn_draining;	/* conn that is draining */
1190 	uint32_t
1191 		idl_repeat : 1,			/* Last conn must re-enable */
1192 						/* drain list again */
1193 		idl_unused : 31;
1194 } idl_t;
1195 
1196 #define	CONN_DRAIN_LIST_LOCK(connp)	(&((connp)->conn_idl->idl_lock))
1197 /*
1198  * Interface route structure which holds the necessary information to recreate
1199  * routes that are tied to an interface (namely where ire_ipif != NULL).
1200  * These routes which were initially created via a routing socket or via the
1201  * SIOCADDRT ioctl may be gateway routes (RTF_GATEWAY being set) or may be
1202  * traditional interface routes.  When an interface comes back up after being
1203  * marked down, this information will be used to recreate the routes.  These
1204  * are part of an mblk_t chain that hangs off of the IPIF (ipif_saved_ire_mp).
1205  */
1206 typedef struct ifrt_s {
1207 	ushort_t	ifrt_type;		/* Type of IRE */
1208 	in6_addr_t	ifrt_v6addr;		/* Address IRE represents. */
1209 	in6_addr_t	ifrt_v6gateway_addr;	/* Gateway if IRE_OFFSUBNET */
1210 	in6_addr_t	ifrt_v6src_addr;	/* Src addr if RTF_SETSRC */
1211 	in6_addr_t	ifrt_v6mask;		/* Mask for matching IRE. */
1212 	uint32_t	ifrt_flags;		/* flags related to route */
1213 	uint_t		ifrt_max_frag;		/* MTU (next hop or path). */
1214 	iulp_t		ifrt_iulp_info;		/* Cached IRE ULP info. */
1215 } ifrt_t;
1216 
1217 #define	ifrt_addr		V4_PART_OF_V6(ifrt_v6addr)
1218 #define	ifrt_gateway_addr	V4_PART_OF_V6(ifrt_v6gateway_addr)
1219 #define	ifrt_src_addr		V4_PART_OF_V6(ifrt_v6src_addr)
1220 #define	ifrt_mask		V4_PART_OF_V6(ifrt_v6mask)
1221 
1222 /* Number of IP addresses that can be hosted on a physical interface */
1223 #define	MAX_ADDRS_PER_IF	8192
1224 /*
1225  * Number of Source addresses to be considered for source address
1226  * selection. Used by ipif_select_source[_v6].
1227  */
1228 #define	MAX_IPIF_SELECT_SOURCE	50
1229 
1230 #ifdef IP_DEBUG
1231 /*
1232  * Tracing refholds and refreleases for debugging. Existing tracing mechanisms
1233  * do not allow the granularity need to trace refrences to ipif/ill/ire's. This
1234  * mechanism should be revisited once dtrace is available.
1235  */
1236 #define	TR_STACK_DEPTH	14
1237 typedef struct tr_buf_s {
1238 	int	tr_depth;
1239 	clock_t	tr_time;
1240 	pc_t	tr_stack[TR_STACK_DEPTH];
1241 } tr_buf_t;
1242 
1243 typedef struct th_trace_s {
1244 	int		th_refcnt;
1245 	uint_t		th_trace_lastref;
1246 	kthread_t	*th_id;
1247 #define	TR_BUF_MAX	38
1248 	tr_buf_t	th_trbuf[TR_BUF_MAX];
1249 } th_trace_t;
1250 
1251 typedef struct th_hash_s {
1252 	list_node_t	thh_link;
1253 	mod_hash_t	*thh_hash;
1254 	ip_stack_t	*thh_ipst;
1255 } th_hash_t;
1256 #endif
1257 
1258 /* The following are ipif_state_flags */
1259 #define	IPIF_CONDEMNED		0x1	/* The ipif is being removed */
1260 #define	IPIF_CHANGING		0x2	/* A critcal ipif field is changing */
1261 #define	IPIF_MOVING		0x8	/* The ipif is being moved */
1262 #define	IPIF_SET_LINKLOCAL	0x10	/* transient flag during bringup */
1263 #define	IPIF_ZERO_SOURCE	0x20	/* transient flag during bringup */
1264 
1265 /* IP interface structure, one per local address */
1266 typedef struct ipif_s {
1267 	struct	ipif_s	*ipif_next;
1268 	struct	ill_s	*ipif_ill;	/* Back pointer to our ill */
1269 	int	ipif_id;		/* Logical unit number */
1270 	uint_t	ipif_mtu;		/* Starts at ipif_ill->ill_max_frag */
1271 	uint_t	ipif_saved_mtu;		/* Save of mtu during ipif_move() */
1272 	in6_addr_t ipif_v6lcl_addr;	/* Local IP address for this if. */
1273 	in6_addr_t ipif_v6src_addr;	/* Source IP address for this if. */
1274 	in6_addr_t ipif_v6subnet;	/* Subnet prefix for this if. */
1275 	in6_addr_t ipif_v6net_mask;	/* Net mask for this interface. */
1276 	in6_addr_t ipif_v6brd_addr;	/* Broadcast addr for this interface. */
1277 	in6_addr_t ipif_v6pp_dst_addr;	/* Point-to-point dest address. */
1278 	uint64_t ipif_flags;		/* Interface flags. */
1279 	uint_t	ipif_metric;		/* BSD if metric, for compatibility. */
1280 	uint_t	ipif_ire_type;		/* IRE_LOCAL or IRE_LOOPBACK */
1281 	mblk_t	*ipif_arp_del_mp;	/* Allocated at time arp comes up, to */
1282 					/* prevent awkward out of mem */
1283 					/* condition later */
1284 	mblk_t	*ipif_saved_ire_mp;	/* Allocated for each extra */
1285 					/* IRE_IF_NORESOLVER/IRE_IF_RESOLVER */
1286 					/* on this interface so that they */
1287 					/* can survive ifconfig down. */
1288 	kmutex_t ipif_saved_ire_lock;	/* Protects ipif_saved_ire_mp */
1289 
1290 	mrec_t	*ipif_igmp_rpt;		/* List of group memberships which */
1291 					/* will be reported on.  Used when */
1292 					/* handling an igmp timeout.	   */
1293 
1294 	/*
1295 	 * The packet counts in the ipif contain the sum of the
1296 	 * packet counts in dead IREs that were affiliated with
1297 	 * this ipif.
1298 	 */
1299 	uint_t	ipif_fo_pkt_count;	/* Forwarded thru our dead IREs */
1300 	uint_t	ipif_ib_pkt_count;	/* Inbound packets for our dead IREs */
1301 	uint_t	ipif_ob_pkt_count;	/* Outbound packets to our dead IREs */
1302 	/* Exclusive bit fields, protected by ipsq_t */
1303 	unsigned int
1304 		ipif_multicast_up : 1,	/* We have joined the allhosts group */
1305 		ipif_replace_zero : 1,	/* Replacement for zero */
1306 		ipif_was_up : 1,	/* ipif was up before */
1307 		ipif_addr_ready : 1,	/* DAD is done */
1308 
1309 		ipif_was_dup : 1,	/* DAD had failed */
1310 		ipif_pad_to_31 : 27;
1311 
1312 	int	ipif_orig_ifindex;	/* ifindex before SLIFFAILOVER */
1313 	uint_t	ipif_seqid;		/* unique index across all ills */
1314 	uint_t	ipif_orig_ipifid;	/* ipif_id before SLIFFAILOVER */
1315 	uint_t	ipif_state_flags;	/* See IPIF_* flag defs above */
1316 	uint_t	ipif_refcnt;		/* active consistent reader cnt */
1317 	uint_t	ipif_ire_cnt;		/* Number of ire's referencing ipif */
1318 	uint_t	ipif_saved_ire_cnt;
1319 	zoneid_t
1320 		ipif_zoneid;		/* zone ID number */
1321 	timeout_id_t ipif_recovery_id;	/* Timer for DAD recovery */
1322 	boolean_t ipif_trace_disable;	/* True when alloc fails */
1323 } ipif_t;
1324 
1325 /*
1326  * The following table lists the protection levels of the various members
1327  * of the ipif_t. The following notation is used.
1328  *
1329  * Write once - Written to only once at the time of bringing up
1330  * the interface and can be safely read after the bringup without any lock.
1331  *
1332  * ipsq - Need to execute in the ipsq to perform the indicated access.
1333  *
1334  * ill_lock - Need to hold this mutex to perform the indicated access.
1335  *
1336  * ill_g_lock - Need to hold this rw lock as reader/writer for read access or
1337  * write access respectively.
1338  *
1339  * down ill - Written to only when the ill is down (i.e all ipifs are down)
1340  * up ill - Read only when the ill is up (i.e. at least 1 ipif is up)
1341  *
1342  *		 Table of ipif_t members and their protection
1343  *
1344  * ipif_next		ipsq + ill_lock +	ipsq OR ill_lock OR
1345  *			ill_g_lock		ill_g_lock
1346  * ipif_ill		ipsq + down ipif	write once
1347  * ipif_id		ipsq + down ipif	write once
1348  * ipif_mtu		ipsq
1349  * ipif_v6lcl_addr	ipsq + down ipif	up ipif
1350  * ipif_v6src_addr	ipsq + down ipif	up ipif
1351  * ipif_v6subnet	ipsq + down ipif	up ipif
1352  * ipif_v6net_mask	ipsq + down ipif	up ipif
1353  *
1354  * ipif_v6brd_addr
1355  * ipif_v6pp_dst_addr
1356  * ipif_flags		ill_lock		ill_lock
1357  * ipif_metric
1358  * ipif_ire_type	ipsq + down ill		up ill
1359  *
1360  * ipif_arp_del_mp	ipsq			ipsq
1361  * ipif_saved_ire_mp	ipif_saved_ire_lock	ipif_saved_ire_lock
1362  * ipif_igmp_rpt	ipsq			ipsq
1363  *
1364  * ipif_fo_pkt_count	Approx
1365  * ipif_ib_pkt_count	Approx
1366  * ipif_ob_pkt_count	Approx
1367  *
1368  * bit fields		ill_lock		ill_lock
1369  *
1370  * ipif_orig_ifindex	ipsq			None
1371  * ipif_orig_ipifid	ipsq			None
1372  * ipif_seqid		ipsq			Write once
1373  *
1374  * ipif_state_flags	ill_lock		ill_lock
1375  * ipif_refcnt		ill_lock		ill_lock
1376  * ipif_ire_cnt		ill_lock		ill_lock
1377  * ipif_saved_ire_cnt
1378  */
1379 
1380 #define	IP_TR_HASH(tid)	((((uintptr_t)tid) >> 6) & (IP_TR_HASH_MAX - 1))
1381 
1382 #ifdef DEBUG
1383 #define	IPIF_TRACE_REF(ipif)	ipif_trace_ref(ipif)
1384 #define	ILL_TRACE_REF(ill)	ill_trace_ref(ill)
1385 #define	IPIF_UNTRACE_REF(ipif)	ipif_untrace_ref(ipif)
1386 #define	ILL_UNTRACE_REF(ill)	ill_untrace_ref(ill)
1387 #else
1388 #define	IPIF_TRACE_REF(ipif)
1389 #define	ILL_TRACE_REF(ill)
1390 #define	IPIF_UNTRACE_REF(ipif)
1391 #define	ILL_UNTRACE_REF(ill)
1392 #endif
1393 
1394 /* IPv4 compatability macros */
1395 #define	ipif_lcl_addr		V4_PART_OF_V6(ipif_v6lcl_addr)
1396 #define	ipif_src_addr		V4_PART_OF_V6(ipif_v6src_addr)
1397 #define	ipif_subnet		V4_PART_OF_V6(ipif_v6subnet)
1398 #define	ipif_net_mask		V4_PART_OF_V6(ipif_v6net_mask)
1399 #define	ipif_brd_addr		V4_PART_OF_V6(ipif_v6brd_addr)
1400 #define	ipif_pp_dst_addr	V4_PART_OF_V6(ipif_v6pp_dst_addr)
1401 
1402 /* Macros for easy backreferences to the ill. */
1403 #define	ipif_wq			ipif_ill->ill_wq
1404 #define	ipif_rq			ipif_ill->ill_rq
1405 #define	ipif_net_type		ipif_ill->ill_net_type
1406 #define	ipif_ipif_up_count	ipif_ill->ill_ipif_up_count
1407 #define	ipif_type		ipif_ill->ill_type
1408 #define	ipif_isv6		ipif_ill->ill_isv6
1409 
1410 #define	SIOCLIFADDR_NDX 112	/* ndx of SIOCLIFADDR in the ndx ioctl table */
1411 
1412 /*
1413  * mode value for ip_ioctl_finish for finishing an ioctl
1414  */
1415 #define	CONN_CLOSE	1		/* No mi_copy */
1416 #define	COPYOUT		2		/* do an mi_copyout if needed */
1417 #define	NO_COPYOUT	3		/* do an mi_copy_done */
1418 #define	IPI2MODE(ipi)	((ipi)->ipi_flags & IPI_GET_CMD ? COPYOUT : NO_COPYOUT)
1419 
1420 /*
1421  * The IP-MT design revolves around the serialization object ipsq_t.
1422  * It is associated with an IPMP group. If IPMP is not enabled, there is
1423  * 1 ipsq_t per phyint. Eg. an ipsq_t would cover both hme0's IPv4 stream
1424  *
1425  * ipsq_lock protects
1426  *	ipsq_reentry_cnt, ipsq_writer, ipsq_xopq_mphead, ipsq_xopq_mptail,
1427  *	ipsq_mphead, ipsq_mptail, ipsq_split
1428  *
1429  *	ipsq_pending_ipif, ipsq_current_ipif, ipsq_pending_mp, ipsq_flags,
1430  *	ipsq_waitfor
1431  *
1432  * The fields in the last line above below are set mostly by a writer thread
1433  * But there is an exception in the last call to ipif_ill_refrele_tail which
1434  * could also race with a conn close which could be cleaning up the
1435  * fields. So we choose to protect using ipsq_lock instead of depending on
1436  * the property of the writer.
1437  * ill_g_lock protects
1438  *	ipsq_refs, ipsq_phyint_list
1439  */
1440 typedef struct ipsq_s {
1441 	kmutex_t ipsq_lock;
1442 	int	ipsq_reentry_cnt;
1443 	kthread_t *ipsq_writer;		/* current owner (thread id) */
1444 	int	ipsq_flags;
1445 	mblk_t	*ipsq_xopq_mphead;	/* list of excl ops mostly ioctls */
1446 	mblk_t	*ipsq_xopq_mptail;
1447 	mblk_t	*ipsq_mphead;		/* msgs on ipsq linked thru b_next */
1448 	mblk_t	*ipsq_mptail;		/* msgs on ipsq linked thru b_next */
1449 	int	ipsq_current_ioctl;	/* current ioctl, or 0 if no ioctl */
1450 	ipif_t	*ipsq_current_ipif;	/* ipif associated with current op */
1451 	ipif_t	*ipsq_pending_ipif;	/* ipif associated w. ipsq_pending_mp */
1452 	mblk_t	*ipsq_pending_mp;	/* current ioctl mp while waiting for */
1453 					/* response from another module */
1454 	struct	ipsq_s	*ipsq_next;	/* list of all syncq's (ipsq_g_list) */
1455 	uint_t		ipsq_refs;	/* Number of phyints on this ipsq */
1456 	struct phyint	*ipsq_phyint_list; /* List of phyints on this ipsq */
1457 	boolean_t	ipsq_split;	/* ipsq may need to be split */
1458 	int		ipsq_waitfor;	/* Values encoded below */
1459 	char		ipsq_name[LIFNAMSIZ+1];	/* same as phyint_groupname */
1460 	ip_stack_t	*ipsq_ipst;	/* Does not have a netstack_hold */
1461 
1462 #ifdef DEBUG
1463 	int		ipsq_depth;	/* debugging aid */
1464 #define	IPSQ_STACK_DEPTH	15
1465 	pc_t		ipsq_stack[IPSQ_STACK_DEPTH];	/* debugging aid */
1466 #endif
1467 } ipsq_t;
1468 
1469 /* ipsq_flags */
1470 #define	IPSQ_GROUP	0x1	/* This ipsq belongs to an IPMP group */
1471 
1472 /*
1473  * ipsq_waitfor:
1474  *
1475  * IPIF_DOWN	1	ipif_down waiting for refcnts to drop
1476  * ILL_DOWN	2	ill_down waiting for refcnts to drop
1477  * IPIF_FREE	3	ipif_free waiting for refcnts to drop
1478  * ILL_FREE	4	ill unplumb waiting for refcnts to drop
1479  * ILL_MOVE_OK	5	failover waiting for refcnts to drop
1480  */
1481 
1482 enum { IPIF_DOWN = 1, ILL_DOWN, IPIF_FREE, ILL_FREE, ILL_MOVE_OK };
1483 
1484 /* Flags passed to ipsq_try_enter */
1485 #define	CUR_OP 0		/* Current ioctl continuing again */
1486 #define	NEW_OP 1		/* New ioctl starting afresh */
1487 
1488 /*
1489  * phyint represents state that is common to both IPv4 and IPv6 interfaces.
1490  * There is a separate ill_t representing IPv4 and IPv6 which has a
1491  * backpointer to the phyint structure for acessing common state.
1492  *
1493  * NOTE : It just stores the group name as there is only one name for
1494  *	  IPv4 and IPv6 i.e it is a underlying link property. Actually
1495  *        IPv4 and IPv6 ill are grouped together when their phyints have
1496  *        the same name.
1497  */
1498 typedef struct phyint {
1499 	struct ill_s	*phyint_illv4;
1500 	struct ill_s	*phyint_illv6;
1501 	uint_t		phyint_ifindex;		/* SIOCLSLIFINDEX */
1502 	char		*phyint_groupname;	/* SIOCSLIFGROUPNAME */
1503 	uint_t		phyint_groupname_len;
1504 	uint64_t	phyint_flags;
1505 	avl_node_t	phyint_avl_by_index;	/* avl tree by index */
1506 	avl_node_t	phyint_avl_by_name;	/* avl tree by name */
1507 	kmutex_t	phyint_lock;
1508 	struct ipsq_s	*phyint_ipsq;		/* back pointer to ipsq */
1509 	struct phyint	*phyint_ipsq_next;	/* phyint list on this ipsq */
1510 	/* Once Clearview IPMP is added the follow two fields can be removed */
1511 	uint_t		phyint_group_ifindex;	/* index assigned to group */
1512 	uint_t		phyint_hook_ifindex;	/* index used with neti/hook */
1513 } phyint_t;
1514 
1515 #define	CACHE_ALIGN_SIZE 64
1516 
1517 #define	CACHE_ALIGN(align_struct)	P2ROUNDUP(sizeof (struct align_struct),\
1518 							CACHE_ALIGN_SIZE)
1519 struct _phyint_list_s_ {
1520 	avl_tree_t	phyint_list_avl_by_index;	/* avl tree by index */
1521 	avl_tree_t	phyint_list_avl_by_name;	/* avl tree by name */
1522 };
1523 
1524 typedef union phyint_list_u {
1525 	struct	_phyint_list_s_ phyint_list_s;
1526 	char	phyint_list_filler[CACHE_ALIGN(_phyint_list_s_)];
1527 } phyint_list_t;
1528 
1529 #define	phyint_list_avl_by_index	phyint_list_s.phyint_list_avl_by_index
1530 #define	phyint_list_avl_by_name		phyint_list_s.phyint_list_avl_by_name
1531 /*
1532  * ILL groups. We group ills,
1533  *
1534  * - if the ills have the same group name. (New way)
1535  *
1536  * ill_group locking notes:
1537  *
1538  * illgrp_lock protects ill_grp_ill_schednext.
1539  *
1540  * ill_g_lock protects ill_grp_next, illgrp_ill, illgrp_ill_count.
1541  * Holding ill_g_lock freezes the memberships of ills in IPMP groups.
1542  * It also freezes the global list of ills and all ipifs in all ills.
1543  *
1544  * To remove an ipif from the linked list of ipifs of that ill ipif_free_tail
1545  * holds both ill_g_lock, and ill_lock. Similarly to remove an ill from the
1546  * global list of ills, ill_glist_delete() holds ill_g_lock as writer.
1547  * This simplifies things for ipif_select_source, illgrp_scheduler etc.
1548  * that need to walk the members of an illgrp. They just hold ill_g_lock
1549  * as reader to do the walk.
1550  *
1551  */
1552 typedef	struct ill_group {
1553 	kmutex_t	illgrp_lock;
1554 	struct ill_group *illgrp_next;		/* Next ill_group */
1555 	struct ill_s	*illgrp_ill_schednext;	/* Next ill to be scheduled */
1556 	struct ill_s	*illgrp_ill;		/* First ill in the group */
1557 	int		illgrp_ill_count;
1558 } ill_group_t;
1559 
1560 /*
1561  * Fragmentation hash bucket
1562  */
1563 typedef struct ipfb_s {
1564 	struct ipf_s	*ipfb_ipf;	/* List of ... */
1565 	size_t		ipfb_count;	/* Count of bytes used by frag(s) */
1566 	kmutex_t	ipfb_lock;	/* Protect all ipf in list */
1567 	uint_t		ipfb_frag_pkts; /* num of distinct fragmented pkts */
1568 } ipfb_t;
1569 
1570 /*
1571  * IRE bucket structure. Usually there is an array of such structures,
1572  * each pointing to a linked list of ires. irb_refcnt counts the number
1573  * of walkers of a given hash bucket. Usually the reference count is
1574  * bumped up if the walker wants no IRES to be DELETED while walking the
1575  * list. Bumping up does not PREVENT ADDITION. This allows walking a given
1576  * hash bucket without stumbling up on a free pointer.
1577  *
1578  * irb_t structures in ip_ftable are dynamically allocated and freed.
1579  * In order to identify the irb_t structures that can be safely kmem_free'd
1580  * we need to ensure that
1581  *  - the irb_refcnt is quiescent, indicating no other walkers,
1582  *  - no other threads or ire's are holding references to the irb,
1583  *	i.e., irb_nire == 0,
1584  *  - there are no active ire's in the bucket, i.e., irb_ire_cnt == 0
1585  */
1586 typedef struct irb {
1587 	struct ire_s	*irb_ire;	/* First ire in this bucket */
1588 					/* Should be first in this struct */
1589 	krwlock_t	irb_lock;	/* Protect this bucket */
1590 	uint_t		irb_refcnt;	/* Protected by irb_lock */
1591 	uchar_t		irb_marks;	/* CONDEMNED ires in this bucket ? */
1592 #define	IRB_MARK_CONDEMNED	0x0001
1593 #define	IRB_MARK_FTABLE		0x0002
1594 	uint_t		irb_ire_cnt;	/* Num of active IRE in this bucket */
1595 	uint_t		irb_tmp_ire_cnt; /* Num of temporary IRE */
1596 	struct ire_s	*irb_rr_origin;	/* origin for round-robin */
1597 	int		irb_nire;	/* Num of ftable ire's that ref irb */
1598 	ip_stack_t	*irb_ipst;	/* Does not have a netstack_hold */
1599 } irb_t;
1600 
1601 #define	IRB2RT(irb)	(rt_t *)((caddr_t)(irb) - offsetof(rt_t, rt_irb))
1602 
1603 /* The following are return values of ip_xmit_v4() */
1604 typedef enum {
1605 	SEND_PASSED = 0,	 /* sent packet out on wire */
1606 	SEND_FAILED,	 /* sending of packet failed */
1607 	LOOKUP_IN_PROGRESS, /* ire cache found, ARP resolution in progress */
1608 	LLHDR_RESLV_FAILED  /* macaddr resl of onlink dst or nexthop failed */
1609 } ipxmit_state_t;
1610 
1611 #define	IP_V4_G_HEAD	0
1612 #define	IP_V6_G_HEAD	1
1613 
1614 #define	MAX_G_HEADS	2
1615 
1616 /*
1617  * unpadded ill_if structure
1618  */
1619 struct 	_ill_if_s_ {
1620 	union ill_if_u	*illif_next;
1621 	union ill_if_u	*illif_prev;
1622 	avl_tree_t	illif_avl_by_ppa;	/* AVL tree sorted on ppa */
1623 	vmem_t		*illif_ppa_arena;	/* ppa index space */
1624 	uint16_t	illif_mcast_v1;		/* hints for		  */
1625 	uint16_t	illif_mcast_v2;		/* [igmp|mld]_slowtimo	  */
1626 	int		illif_name_len;		/* name length */
1627 	char		illif_name[LIFNAMSIZ];	/* name of interface type */
1628 };
1629 
1630 /* cache aligned ill_if structure */
1631 typedef union 	ill_if_u {
1632 	struct  _ill_if_s_ ill_if_s;
1633 	char 	illif_filler[CACHE_ALIGN(_ill_if_s_)];
1634 } ill_if_t;
1635 
1636 #define	illif_next		ill_if_s.illif_next
1637 #define	illif_prev		ill_if_s.illif_prev
1638 #define	illif_avl_by_ppa	ill_if_s.illif_avl_by_ppa
1639 #define	illif_ppa_arena		ill_if_s.illif_ppa_arena
1640 #define	illif_mcast_v1		ill_if_s.illif_mcast_v1
1641 #define	illif_mcast_v2		ill_if_s.illif_mcast_v2
1642 #define	illif_name		ill_if_s.illif_name
1643 #define	illif_name_len		ill_if_s.illif_name_len
1644 
1645 typedef struct ill_walk_context_s {
1646 	int	ctx_current_list; /* current list being searched */
1647 	int	ctx_last_list;	 /* last list to search */
1648 } ill_walk_context_t;
1649 
1650 /*
1651  * ill_g_heads structure, one for IPV4 and one for IPV6
1652  */
1653 struct _ill_g_head_s_ {
1654 	ill_if_t	*ill_g_list_head;
1655 	ill_if_t	*ill_g_list_tail;
1656 };
1657 
1658 typedef union ill_g_head_u {
1659 	struct _ill_g_head_s_ ill_g_head_s;
1660 	char	ill_g_head_filler[CACHE_ALIGN(_ill_g_head_s_)];
1661 } ill_g_head_t;
1662 
1663 #define	ill_g_list_head	ill_g_head_s.ill_g_list_head
1664 #define	ill_g_list_tail	ill_g_head_s.ill_g_list_tail
1665 
1666 #define	IP_V4_ILL_G_LIST(ipst)	\
1667 	(ipst)->ips_ill_g_heads[IP_V4_G_HEAD].ill_g_list_head
1668 #define	IP_V6_ILL_G_LIST(ipst)	\
1669 	(ipst)->ips_ill_g_heads[IP_V6_G_HEAD].ill_g_list_head
1670 #define	IP_VX_ILL_G_LIST(i, ipst)	\
1671 	(ipst)->ips_ill_g_heads[i].ill_g_list_head
1672 
1673 #define	ILL_START_WALK_V4(ctx_ptr, ipst)	\
1674 	ill_first(IP_V4_G_HEAD, IP_V4_G_HEAD, ctx_ptr, ipst)
1675 #define	ILL_START_WALK_V6(ctx_ptr, ipst)	\
1676 	ill_first(IP_V6_G_HEAD, IP_V6_G_HEAD, ctx_ptr, ipst)
1677 #define	ILL_START_WALK_ALL(ctx_ptr, ipst)	\
1678 	ill_first(MAX_G_HEADS, MAX_G_HEADS, ctx_ptr, ipst)
1679 
1680 /*
1681  * Capabilities, possible flags for ill_capabilities.
1682  */
1683 
1684 #define	ILL_CAPAB_AH		0x01		/* IPsec AH acceleration */
1685 #define	ILL_CAPAB_ESP		0x02		/* IPsec ESP acceleration */
1686 #define	ILL_CAPAB_MDT		0x04		/* Multidata Transmit */
1687 #define	ILL_CAPAB_HCKSUM	0x08		/* Hardware checksumming */
1688 #define	ILL_CAPAB_ZEROCOPY	0x10		/* Zero-copy */
1689 #define	ILL_CAPAB_POLL		0x20		/* Polling Toggle */
1690 #define	ILL_CAPAB_SOFT_RING	0x40		/* Soft_Ring capability */
1691 #define	ILL_CAPAB_LSO		0x80		/* Large Segment Offload */
1692 
1693 /*
1694  * Per-ill Multidata Transmit capabilities.
1695  */
1696 typedef struct ill_mdt_capab_s ill_mdt_capab_t;
1697 
1698 /*
1699  * Per-ill IPsec capabilities.
1700  */
1701 typedef struct ill_ipsec_capab_s ill_ipsec_capab_t;
1702 
1703 /*
1704  * Per-ill Hardware Checksumming capbilities.
1705  */
1706 typedef struct ill_hcksum_capab_s ill_hcksum_capab_t;
1707 
1708 /*
1709  * Per-ill Zero-copy capabilities.
1710  */
1711 typedef struct ill_zerocopy_capab_s ill_zerocopy_capab_t;
1712 
1713 /*
1714  * Per-ill Polling/soft ring capbilities.
1715  */
1716 typedef struct ill_dls_capab_s ill_dls_capab_t;
1717 
1718 /*
1719  * Per-ill polling resource map.
1720  */
1721 typedef struct ill_rx_ring ill_rx_ring_t;
1722 
1723 /*
1724  * Per-ill Large Segment Offload capabilities.
1725  */
1726 typedef struct ill_lso_capab_s ill_lso_capab_t;
1727 
1728 /* The following are ill_state_flags */
1729 #define	ILL_LL_SUBNET_PENDING	0x01	/* Waiting for DL_INFO_ACK from drv */
1730 #define	ILL_CONDEMNED		0x02	/* No more new ref's to the ILL */
1731 #define	ILL_CHANGING		0x04	/* ILL not globally visible */
1732 #define	ILL_DL_UNBIND_IN_PROGRESS	0x08	/* UNBIND_REQ is sent */
1733 #define	ILL_SOFT_RING_ASSIGN	0x10	/* Makeing soft ring assigment */
1734 
1735 /* Is this an ILL whose source address is used by other ILL's ? */
1736 #define	IS_USESRC_ILL(ill)			\
1737 	(((ill)->ill_usesrc_ifindex == 0) &&	\
1738 	((ill)->ill_usesrc_grp_next != NULL))
1739 
1740 /* Is this a client/consumer of the usesrc ILL ? */
1741 #define	IS_USESRC_CLI_ILL(ill)			\
1742 	(((ill)->ill_usesrc_ifindex != 0) &&	\
1743 	((ill)->ill_usesrc_grp_next != NULL))
1744 
1745 /* Is this an virtual network interface (vni) ILL ? */
1746 #define	IS_VNI(ill)							     \
1747 	(((ill) != NULL) &&						     \
1748 	(((ill)->ill_phyint->phyint_flags & (PHYI_LOOPBACK|PHYI_VIRTUAL)) == \
1749 	PHYI_VIRTUAL))
1750 
1751 /* Is this a loopback ILL? */
1752 #define	IS_LOOPBACK(ill) \
1753 	((ill)->ill_phyint->phyint_flags & PHYI_LOOPBACK)
1754 
1755 /*
1756  * IP Lower level Structure.
1757  * Instance data structure in ip_open when there is a device below us.
1758  */
1759 typedef struct ill_s {
1760 	ill_if_t *ill_ifptr;		/* pointer to interface type */
1761 	queue_t	*ill_rq;		/* Read queue. */
1762 	queue_t	*ill_wq;		/* Write queue. */
1763 
1764 	int	ill_error;		/* Error value sent up by device. */
1765 
1766 	ipif_t	*ill_ipif;		/* Interface chain for this ILL. */
1767 
1768 	uint_t	ill_ipif_up_count;	/* Number of IPIFs currently up. */
1769 	uint_t	ill_max_frag;		/* Max IDU from DLPI. */
1770 	char	*ill_name;		/* Our name. */
1771 	uint_t	ill_ipif_dup_count;	/* Number of duplicate addresses. */
1772 	uint_t	ill_name_length;	/* Name length, incl. terminator. */
1773 	char	*ill_ndd_name;		/* Name + ":ip?_forwarding" for NDD. */
1774 	uint_t	ill_net_type;		/* IRE_IF_RESOLVER/IRE_IF_NORESOLVER. */
1775 	/*
1776 	 * Physical Point of Attachment num.  If DLPI style 1 provider
1777 	 * then this is derived from the devname.
1778 	 */
1779 	uint_t	ill_ppa;
1780 	t_uscalar_t	ill_sap;
1781 	t_scalar_t	ill_sap_length;	/* Including sign (for position) */
1782 	uint_t	ill_phys_addr_length;	/* Excluding the sap. */
1783 	uint_t	ill_bcast_addr_length;	/* Only set when the DL provider */
1784 					/* supports broadcast. */
1785 	t_uscalar_t	ill_mactype;
1786 	uint8_t	*ill_frag_ptr;		/* Reassembly state. */
1787 	timeout_id_t ill_frag_timer_id; /* timeout id for the frag timer */
1788 	ipfb_t	*ill_frag_hash_tbl;	/* Fragment hash list head. */
1789 	ipif_t	*ill_pending_ipif;	/* IPIF waiting for DL operation. */
1790 
1791 	ilm_t	*ill_ilm;		/* Multicast mebership for lower ill */
1792 	uint_t	ill_global_timer;	/* for IGMPv3/MLDv2 general queries */
1793 	int	ill_mcast_type;		/* type of router which is querier */
1794 					/* on this interface */
1795 	uint16_t ill_mcast_v1_time;	/* # slow timeouts since last v1 qry */
1796 	uint16_t ill_mcast_v2_time;	/* # slow timeouts since last v2 qry */
1797 	uint8_t	ill_mcast_v1_tset;	/* 1 => timer is set; 0 => not set */
1798 	uint8_t	ill_mcast_v2_tset;	/* 1 => timer is set; 0 => not set */
1799 
1800 	uint8_t	ill_mcast_rv;		/* IGMPv3/MLDv2 robustness variable */
1801 	int	ill_mcast_qi;		/* IGMPv3/MLDv2 query interval var */
1802 
1803 	mblk_t	*ill_pending_mp;	/* IOCTL/DLPI awaiting completion. */
1804 	/*
1805 	 * All non-NULL cells between 'ill_first_mp_to_free' and
1806 	 * 'ill_last_mp_to_free' are freed in ill_delete.
1807 	 */
1808 #define	ill_first_mp_to_free	ill_bcast_mp
1809 	mblk_t	*ill_bcast_mp;		/* DLPI header for broadcasts. */
1810 	mblk_t	*ill_resolver_mp;	/* Resolver template. */
1811 	mblk_t	*ill_unbind_mp;		/* unbind mp from ill_dl_up() */
1812 	mblk_t	*ill_dlpi_deferred;	/* b_next chain of control messages */
1813 	mblk_t	*ill_phys_addr_mp;	/* mblk which holds ill_phys_addr */
1814 #define	ill_last_mp_to_free	ill_phys_addr_mp
1815 
1816 	cred_t	*ill_credp;		/* opener's credentials */
1817 	uint8_t	*ill_phys_addr;		/* ill_phys_addr_mp->b_rptr + off */
1818 
1819 	uint_t	ill_state_flags;	/* see ILL_* flags above */
1820 
1821 	/* Following bit fields protected by ipsq_t */
1822 	uint_t
1823 		ill_needs_attach : 1,
1824 		ill_reserved : 1,
1825 		ill_isv6 : 1,
1826 		ill_dlpi_style_set : 1,
1827 
1828 		ill_ifname_pending : 1,
1829 		ill_move_in_progress : 1, /* FAILOVER/FAILBACK in progress */
1830 		ill_join_allmulti : 1,
1831 		ill_logical_down : 1,
1832 
1833 		ill_is_6to4tun : 1,	/* Interface is a 6to4 tunnel */
1834 		ill_promisc_on_phys : 1, /* phys interface in promisc mode */
1835 		ill_dl_up : 1,
1836 		ill_up_ipifs : 1,
1837 
1838 		ill_note_link : 1,	/* supports link-up notification */
1839 		ill_capab_reneg : 1, /* capability renegotiation to be done */
1840 		ill_pad_to_bit_31 : 18;
1841 
1842 	/* Following bit fields protected by ill_lock */
1843 	uint_t
1844 		ill_fragtimer_executing : 1,
1845 		ill_fragtimer_needrestart : 1,
1846 		ill_ilm_cleanup_reqd : 1,
1847 		ill_arp_closing : 1,
1848 
1849 		ill_arp_bringup_pending : 1,
1850 		ill_mtu_userspecified : 1, /* SIOCSLNKINFO has set the mtu */
1851 		ill_arp_extend : 1,	/* ARP has DAD extensions */
1852 		ill_pad_bit_31 : 25;
1853 
1854 	/*
1855 	 * Used in SIOCSIFMUXID and SIOCGIFMUXID for 'ifconfig unplumb'.
1856 	 */
1857 	int	ill_arp_muxid;		/* muxid returned from plink for arp */
1858 	int	ill_ip_muxid;		/* muxid returned from plink for ip */
1859 
1860 	/*
1861 	 * NIC event information attached, to be used by nic event hooks.
1862 	 */
1863 	hook_nic_event_t	*ill_nic_event_info;
1864 
1865 	/*
1866 	 * Used for IP frag reassembly throttling on a per ILL basis.
1867 	 *
1868 	 * Note: frag_count is approximate, its added to and subtracted from
1869 	 *	 without any locking, so simultaneous load/modify/stores can
1870 	 *	 collide, also ill_frag_purge() recalculates its value by
1871 	 *	 summing all the ipfb_count's without locking out updates
1872 	 *	 to the ipfb's.
1873 	 */
1874 	uint_t	ill_ipf_gen;		/* Generation of next fragment queue */
1875 	uint_t	ill_frag_count;		/* Approx count of all mblk bytes */
1876 	uint_t	ill_frag_free_num_pkts;	 /* num of fragmented packets to free */
1877 	clock_t	ill_last_frag_clean_time; /* time when frag's were pruned */
1878 	int	ill_type;		/* From <net/if_types.h> */
1879 	uint_t	ill_dlpi_multicast_state;	/* See below IDS_* */
1880 	uint_t	ill_dlpi_fastpath_state;	/* See below IDS_* */
1881 
1882 	/*
1883 	 * Capabilities related fields.
1884 	 */
1885 	uint_t  ill_dlpi_capab_state;	/* State of capability query, IDS_* */
1886 	uint64_t ill_capabilities;	/* Enabled capabilities, ILL_CAPAB_* */
1887 	ill_mdt_capab_t	*ill_mdt_capab;	/* Multidata Transmit capabilities */
1888 	ill_ipsec_capab_t *ill_ipsec_capab_ah;	/* IPsec AH capabilities */
1889 	ill_ipsec_capab_t *ill_ipsec_capab_esp;	/* IPsec ESP capabilities */
1890 	ill_hcksum_capab_t *ill_hcksum_capab; /* H/W cksumming capabilities */
1891 	ill_zerocopy_capab_t *ill_zerocopy_capab; /* Zero-copy capabilities */
1892 	ill_dls_capab_t *ill_dls_capab; /* Polling, soft ring capabilities */
1893 	ill_lso_capab_t *ill_lso_capab; /* Large Segment Offload capabilities */
1894 
1895 	/*
1896 	 * New fields for IPv6
1897 	 */
1898 	uint8_t	ill_max_hops;	/* Maximum hops for any logical interface */
1899 	uint_t	ill_max_mtu;	/* Maximum MTU for any logical interface */
1900 	uint32_t ill_reachable_time;	/* Value for ND algorithm in msec */
1901 	uint32_t ill_reachable_retrans_time; /* Value for ND algorithm msec */
1902 	uint_t	ill_max_buf;		/* Max # of req to buffer for ND */
1903 	in6_addr_t	ill_token;
1904 	uint_t		ill_token_length;
1905 	uint32_t	ill_xmit_count;		/* ndp max multicast xmits */
1906 	mib2_ipIfStatsEntry_t	*ill_ip_mib;	/* ver indep. interface mib */
1907 	mib2_ipv6IfIcmpEntry_t	*ill_icmp6_mib;	/* Per interface mib */
1908 	/*
1909 	 * Following two mblks are allocated common to all
1910 	 * the ipifs when the first interface is coming up.
1911 	 * It is sent up to arp when the last ipif is coming
1912 	 * down.
1913 	 */
1914 	mblk_t			*ill_arp_down_mp;
1915 	mblk_t			*ill_arp_del_mapping_mp;
1916 	/*
1917 	 * Used for implementing IFF_NOARP. As IFF_NOARP is used
1918 	 * to turn off for all the logicals, it is here instead
1919 	 * of the ipif.
1920 	 */
1921 	mblk_t			*ill_arp_on_mp;
1922 	/* Peer ill of an IPMP move operation */
1923 	struct ill_s		*ill_move_peer;
1924 
1925 	phyint_t		*ill_phyint;
1926 	uint64_t		ill_flags;
1927 	ill_group_t		*ill_group;
1928 	struct ill_s		*ill_group_next;
1929 
1930 	kmutex_t	ill_lock;	/* Please see table below */
1931 	/*
1932 	 * The ill_nd_lla* fields handle the link layer address option
1933 	 * from neighbor discovery. This is used for external IPv6
1934 	 * address resolution.
1935 	 */
1936 	mblk_t		*ill_nd_lla_mp;	/* mblk which holds ill_nd_lla */
1937 	uint8_t		*ill_nd_lla;	/* Link Layer Address */
1938 	uint_t		ill_nd_lla_len;	/* Link Layer Address length */
1939 	/*
1940 	 * We now have 3 phys_addr_req's sent down. This field keeps track
1941 	 * of which one is pending.
1942 	 */
1943 	t_uscalar_t	ill_phys_addr_pend; /* which dl_phys_addr_req pending */
1944 	/*
1945 	 * Used to save errors that occur during plumbing
1946 	 */
1947 	uint_t		ill_ifname_pending_err;
1948 	avl_node_t	ill_avl_byppa; /* avl node based on ppa */
1949 	void		*ill_fastpath_list; /* both ire and nce hang off this */
1950 	uint_t		ill_refcnt;	/* active refcnt by threads */
1951 	uint_t		ill_ire_cnt;	/* ires associated with this ill */
1952 	kcondvar_t	ill_cv;
1953 	uint_t		ill_ilm_walker_cnt;	/* snmp ilm walkers */
1954 	uint_t		ill_nce_cnt;	/* nces associated with this ill */
1955 	uint_t		ill_waiters;	/* threads waiting in ipsq_enter */
1956 	/*
1957 	 * Contains the upper read queue pointer of the module immediately
1958 	 * beneath IP.  This field allows IP to validate sub-capability
1959 	 * acknowledgments coming up from downstream.
1960 	 */
1961 	queue_t		*ill_lmod_rq;	/* read queue pointer of module below */
1962 	uint_t		ill_lmod_cnt;	/* number of modules beneath IP */
1963 	ip_m_t		*ill_media;	/* media specific params/functions */
1964 	t_uscalar_t	ill_dlpi_pending; /* Last DLPI primitive issued */
1965 	uint_t		ill_usesrc_ifindex; /* use src addr from this ILL */
1966 	struct ill_s	*ill_usesrc_grp_next; /* Next ILL in the usesrc group */
1967 	boolean_t	ill_trace_disable;	/* True when alloc fails */
1968 	zoneid_t	ill_zoneid;
1969 	ip_stack_t	*ill_ipst;	/* Corresponds to a netstack_hold */
1970 } ill_t;
1971 
1972 /*
1973  * The following table lists the protection levels of the various members
1974  * of the ill_t. Same notation as that used for ipif_t above is used.
1975  *
1976  *				Write			Read
1977  *
1978  * ill_ifptr			ill_g_lock + s		Write once
1979  * ill_rq			ipsq			Write once
1980  * ill_wq			ipsq			Write once
1981  *
1982  * ill_error			ipsq			None
1983  * ill_ipif			ill_g_lock + ipsq	ill_g_lock OR ipsq
1984  * ill_ipif_up_count		ill_lock + ipsq		ill_lock OR ipsq
1985  * ill_max_frag			ipsq			Write once
1986  *
1987  * ill_name			ill_g_lock + ipsq	Write once
1988  * ill_name_length		ill_g_lock + ipsq	Write once
1989  * ill_ndd_name			ipsq			Write once
1990  * ill_net_type			ipsq			Write once
1991  * ill_ppa			ill_g_lock + ipsq	Write once
1992  * ill_sap			ipsq + down ill		Write once
1993  * ill_sap_length		ipsq + down ill		Write once
1994  * ill_phys_addr_length		ipsq + down ill		Write once
1995  *
1996  * ill_bcast_addr_length	ipsq			ipsq
1997  * ill_mactype			ipsq			ipsq
1998  * ill_frag_ptr			ipsq			ipsq
1999  *
2000  * ill_frag_timer_id		ill_lock		ill_lock
2001  * ill_frag_hash_tbl		ipsq			up ill
2002  * ill_ilm			ipsq + ill_lock		ill_lock
2003  * ill_mcast_type		ill_lock		ill_lock
2004  * ill_mcast_v1_time		ill_lock		ill_lock
2005  * ill_mcast_v2_time		ill_lock		ill_lock
2006  * ill_mcast_v1_tset		ill_lock		ill_lock
2007  * ill_mcast_v2_tset		ill_lock		ill_lock
2008  * ill_mcast_rv			ill_lock		ill_lock
2009  * ill_mcast_qi			ill_lock		ill_lock
2010  * ill_pending_mp		ill_lock		ill_lock
2011  *
2012  * ill_bcast_mp			ipsq			ipsq
2013  * ill_resolver_mp		ipsq			only when ill is up
2014  * ill_down_mp			ipsq			ipsq
2015  * ill_dlpi_deferred		ill_lock		ill_lock
2016  * ill_dlpi_pending		ill_lock		ill_lock
2017  * ill_phys_addr_mp		ipsq + down ill		only when ill is up
2018  * ill_phys_addr		ipsq + down ill		only when ill is up
2019  *
2020  * ill_state_flags		ill_lock		ill_lock
2021  * exclusive bit flags		ipsq_t			ipsq_t
2022  * shared bit flags		ill_lock		ill_lock
2023  *
2024  * ill_arp_muxid		ipsq			Not atomic
2025  * ill_ip_muxid			ipsq			Not atomic
2026  *
2027  * ill_ipf_gen			Not atomic
2028  * ill_frag_count		Approx. not protected
2029  * ill_type			ipsq + down ill		only when ill is up
2030  * ill_dlpi_multicast_state	ill_lock		ill_lock
2031  * ill_dlpi_fastpath_state	ill_lock		ill_lock
2032  * ill_max_hops			ipsq			Not atomic
2033  *
2034  * ill_max_mtu
2035  *
2036  * ill_reachable_time		ipsq + ill_lock		ill_lock
2037  * ill_reachable_retrans_time	ipsq + ill_lock		ill_lock
2038  * ill_max_buf			ipsq + ill_lock		ill_lock
2039  *
2040  * Next 2 fields need ill_lock because of the get ioctls. They should not
2041  * report partially updated results without executing in the ipsq.
2042  * ill_token			ipsq + ill_lock		ill_lock
2043  * ill_token_length		ipsq + ill_lock		ill_lock
2044  * ill_xmit_count		ipsq + down ill		write once
2045  * ill_ip6_mib			ipsq + down ill		only when ill is up
2046  * ill_icmp6_mib		ipsq + down ill		only when ill is up
2047  * ill_arp_down_mp		ipsq			ipsq
2048  * ill_arp_del_mapping_mp	ipsq			ipsq
2049  * ill_arp_on_mp		ipsq			ipsq
2050  * ill_move_peer		ipsq			ipsq
2051  *
2052  * ill_phyint			ipsq, ill_g_lock, ill_lock	Any of them
2053  * ill_flags			ill_lock		ill_lock
2054  * ill_group			ipsq, ill_g_lock, ill_lock	Any of them
2055  * ill_group_next		ipsq, ill_g_lock, ill_lock	Any of them
2056  * ill_nd_lla_mp		ipsq + down ill		only when ill is up
2057  * ill_nd_lla			ipsq + down ill		only when ill is up
2058  * ill_nd_lla_len		ipsq + down ill		only when ill is up
2059  * ill_phys_addr_pend		ipsq + down ill		only when ill is up
2060  * ill_ifname_pending_err	ipsq			ipsq
2061  * ill_avl_byppa		ipsq, ill_g_lock	write once
2062  *
2063  * ill_fastpath_list		ill_lock		ill_lock
2064  * ill_refcnt			ill_lock		ill_lock
2065  * ill_ire_cnt			ill_lock		ill_lock
2066  * ill_cv			ill_lock		ill_lock
2067  * ill_ilm_walker_cnt		ill_lock		ill_lock
2068  * ill_nce_cnt			ill_lock		ill_lock
2069  * ill_trace			ill_lock		ill_lock
2070  * ill_usesrc_grp_next		ill_g_usesrc_lock	ill_g_usesrc_lock
2071  */
2072 
2073 /*
2074  * For ioctl restart mechanism see ip_reprocess_ioctl()
2075  */
2076 struct ip_ioctl_cmd_s;
2077 
2078 typedef	int (*ifunc_t)(ipif_t *, struct sockaddr_in *, queue_t *, mblk_t *,
2079     struct ip_ioctl_cmd_s *, void *);
2080 
2081 typedef struct ip_ioctl_cmd_s {
2082 	int	ipi_cmd;
2083 	size_t	ipi_copyin_size;
2084 	uint_t	ipi_flags;
2085 	uint_t	ipi_cmd_type;
2086 	ifunc_t	ipi_func;
2087 	ifunc_t	ipi_func_restart;
2088 } ip_ioctl_cmd_t;
2089 
2090 /*
2091  * ipi_cmd_type:
2092  *
2093  * IF_CMD		1	old style ifreq cmd
2094  * LIF_CMD		2	new style lifreq cmd
2095  * TUN_CMD		3	tunnel related
2096  * ARP_CMD		4	arpreq cmd
2097  * XARP_CMD		5	xarpreq cmd
2098  * MSFILT_CMD		6	multicast source filter cmd
2099  * MISC_CMD		7	misc cmd (not a more specific one above)
2100  */
2101 
2102 enum { IF_CMD = 1, LIF_CMD, TUN_CMD, ARP_CMD, XARP_CMD, MSFILT_CMD, MISC_CMD };
2103 
2104 #define	IPI_DONTCARE	0	/* For ioctl encoded values that don't matter */
2105 
2106 /* Flag values in ipi_flags */
2107 #define	IPI_PRIV	0x1	/* Root only command */
2108 #define	IPI_MODOK	0x2	/* Permitted on mod instance of IP */
2109 #define	IPI_WR		0x4	/* Need to grab writer access */
2110 #define	IPI_GET_CMD	0x8	/* branch to mi_copyout on success */
2111 #define	IPI_REPL	0x10	/* valid for replacement ipif created in MOVE */
2112 #define	IPI_NULL_BCONT	0x20	/* ioctl has not data and hence no b_cont */
2113 #define	IPI_PASS_DOWN	0x40	/* pass this ioctl down when a module only */
2114 
2115 extern ip_ioctl_cmd_t	ip_ndx_ioctl_table[];
2116 extern ip_ioctl_cmd_t	ip_misc_ioctl_table[];
2117 extern int ip_ndx_ioctl_count;
2118 extern int ip_misc_ioctl_count;
2119 
2120 #define	ILL_CLEAR_MOVE(ill) {				\
2121 	ill_t *peer_ill;				\
2122 							\
2123 	peer_ill = (ill)->ill_move_peer;		\
2124 	ASSERT(peer_ill != NULL);			\
2125 	(ill)->ill_move_in_progress = B_FALSE;		\
2126 	peer_ill->ill_move_in_progress = B_FALSE;	\
2127 	(ill)->ill_move_peer = NULL;			\
2128 	peer_ill->ill_move_peer = NULL;			\
2129 }
2130 
2131 /* Passed down by ARP to IP during I_PLINK/I_PUNLINK */
2132 typedef struct ipmx_s {
2133 	char	ipmx_name[LIFNAMSIZ];		/* if name */
2134 	uint_t
2135 		ipmx_arpdev_stream : 1,		/* This is the arp stream */
2136 		ipmx_notused : 31;
2137 } ipmx_t;
2138 
2139 /*
2140  * State for detecting if a driver supports certain features.
2141  * Support for DL_ENABMULTI_REQ uses ill_dlpi_multicast_state.
2142  * Support for DLPI M_DATA fastpath uses ill_dlpi_fastpath_state.
2143  * Support for DL_CAPABILITY_REQ uses ill_dlpi_capab_state.
2144  */
2145 #define	IDS_UNKNOWN	0	/* No DLPI request sent */
2146 #define	IDS_INPROGRESS	1	/* DLPI request sent */
2147 #define	IDS_OK		2	/* DLPI request completed successfully */
2148 #define	IDS_FAILED	3	/* DLPI request failed */
2149 
2150 /* Named Dispatch Parameter Management Structure */
2151 typedef struct ipparam_s {
2152 	uint_t	ip_param_min;
2153 	uint_t	ip_param_max;
2154 	uint_t	ip_param_value;
2155 	char	*ip_param_name;
2156 } ipparam_t;
2157 
2158 /* Extended NDP Management Structure */
2159 typedef struct ipndp_s {
2160 	ndgetf_t	ip_ndp_getf;
2161 	ndsetf_t	ip_ndp_setf;
2162 	caddr_t		ip_ndp_data;
2163 	char		*ip_ndp_name;
2164 } ipndp_t;
2165 
2166 /*
2167  * The kernel stores security attributes of all gateways in a database made
2168  * up of one or more tsol_gcdb_t elements.  Each tsol_gcdb_t contains the
2169  * security-related credentials of the gateway.  More than one gateways may
2170  * share entries in the database.
2171  *
2172  * The tsol_gc_t structure represents the gateway to credential association,
2173  * and refers to an entry in the database.  One or more tsol_gc_t entities are
2174  * grouped together to form one or more tsol_gcgrp_t, each representing the
2175  * list of security attributes specific to the gateway.  A gateway may be
2176  * associated with at most one credentials group.
2177  */
2178 struct tsol_gcgrp_s;
2179 
2180 extern uchar_t	ip6opt_ls;	/* TX IPv6 enabler */
2181 
2182 /*
2183  * Gateway security credential record.
2184  */
2185 typedef struct tsol_gcdb_s {
2186 	uint_t		gcdb_refcnt;	/* reference count */
2187 	struct rtsa_s	gcdb_attr;	/* security attributes */
2188 #define	gcdb_mask	gcdb_attr.rtsa_mask
2189 #define	gcdb_doi	gcdb_attr.rtsa_doi
2190 #define	gcdb_slrange	gcdb_attr.rtsa_slrange
2191 } tsol_gcdb_t;
2192 
2193 /*
2194  * Gateway to credential association.
2195  */
2196 typedef struct tsol_gc_s {
2197 	uint_t		gc_refcnt;	/* reference count */
2198 	struct tsol_gcgrp_s *gc_grp;	/* pointer to group */
2199 	struct tsol_gc_s *gc_prev;	/* previous in list */
2200 	struct tsol_gc_s *gc_next;	/* next in list */
2201 	tsol_gcdb_t	*gc_db;		/* pointer to actual credentials */
2202 } tsol_gc_t;
2203 
2204 /*
2205  * Gateway credentials group address.
2206  */
2207 typedef struct tsol_gcgrp_addr_s {
2208 	int		ga_af;		/* address family */
2209 	in6_addr_t	ga_addr;	/* IPv4 mapped or IPv6 address */
2210 } tsol_gcgrp_addr_t;
2211 
2212 /*
2213  * Gateway credentials group.
2214  */
2215 typedef struct tsol_gcgrp_s {
2216 	uint_t		gcgrp_refcnt;	/* reference count */
2217 	krwlock_t	gcgrp_rwlock;	/* lock to protect following */
2218 	uint_t		gcgrp_count;	/* number of credentials */
2219 	tsol_gc_t	*gcgrp_head;	/* first credential in list */
2220 	tsol_gc_t	*gcgrp_tail;	/* last credential in list */
2221 	tsol_gcgrp_addr_t gcgrp_addr;	/* next-hop gateway address */
2222 } tsol_gcgrp_t;
2223 
2224 extern kmutex_t gcgrp_lock;
2225 
2226 #define	GC_REFRELE(p) {				\
2227 	ASSERT((p)->gc_grp != NULL);		\
2228 	rw_enter(&(p)->gc_grp->gcgrp_rwlock, RW_WRITER); \
2229 	ASSERT((p)->gc_refcnt > 0);		\
2230 	if (--((p)->gc_refcnt) == 0)		\
2231 		gc_inactive(p);			\
2232 	else					\
2233 		rw_exit(&(p)->gc_grp->gcgrp_rwlock); \
2234 }
2235 
2236 #define	GCGRP_REFHOLD(p) {			\
2237 	mutex_enter(&gcgrp_lock);		\
2238 	++((p)->gcgrp_refcnt);			\
2239 	ASSERT((p)->gcgrp_refcnt != 0);		\
2240 	mutex_exit(&gcgrp_lock);		\
2241 }
2242 
2243 #define	GCGRP_REFRELE(p) {			\
2244 	mutex_enter(&gcgrp_lock);		\
2245 	ASSERT((p)->gcgrp_refcnt > 0);		\
2246 	if (--((p)->gcgrp_refcnt) == 0)		\
2247 		gcgrp_inactive(p);		\
2248 	ASSERT(MUTEX_HELD(&gcgrp_lock));	\
2249 	mutex_exit(&gcgrp_lock);		\
2250 }
2251 
2252 /*
2253  * IRE gateway security attributes structure, pointed to by tsol_ire_gw_secattr
2254  */
2255 struct tsol_tnrhc;
2256 
2257 typedef struct tsol_ire_gw_secattr_s {
2258 	kmutex_t	igsa_lock;	/* lock to protect following */
2259 	struct tsol_tnrhc *igsa_rhc;	/* host entry for gateway */
2260 	tsol_gc_t	*igsa_gc;	/* for prefix IREs */
2261 	tsol_gcgrp_t	*igsa_gcgrp;	/* for cache IREs */
2262 } tsol_ire_gw_secattr_t;
2263 
2264 /*
2265  * Following are the macros to increment/decrement the reference
2266  * count of the IREs and IRBs (ire bucket).
2267  *
2268  * 1) We bump up the reference count of an IRE to make sure that
2269  *    it does not get deleted and freed while we are using it.
2270  *    Typically all the lookup functions hold the bucket lock,
2271  *    and look for the IRE. If it finds an IRE, it bumps up the
2272  *    reference count before dropping the lock. Sometimes we *may* want
2273  *    to bump up the reference count after we *looked* up i.e without
2274  *    holding the bucket lock. So, the IRE_REFHOLD macro does not assert
2275  *    on the bucket lock being held. Any thread trying to delete from
2276  *    the hash bucket can still do so but cannot free the IRE if
2277  *    ire_refcnt is not 0.
2278  *
2279  * 2) We bump up the reference count on the bucket where the IRE resides
2280  *    (IRB), when we want to prevent the IREs getting deleted from a given
2281  *    hash bucket. This makes life easier for ire_walk type functions which
2282  *    wants to walk the IRE list, call a function, but needs to drop
2283  *    the bucket lock to prevent recursive rw_enters. While the
2284  *    lock is dropped, the list could be changed by other threads or
2285  *    the same thread could end up deleting the ire or the ire pointed by
2286  *    ire_next. IRE_REFHOLDing the ire or ire_next is not sufficient as
2287  *    a delete will still remove the ire from the bucket while we have
2288  *    dropped the lock and hence the ire_next would be NULL. Thus, we
2289  *    need a mechanism to prevent deletions from a given bucket.
2290  *
2291  *    To prevent deletions, we bump up the reference count on the
2292  *    bucket. If the bucket is held, ire_delete just marks IRE_MARK_CONDEMNED
2293  *    both on the ire's ire_marks and the bucket's irb_marks. When the
2294  *    reference count on the bucket drops to zero, all the CONDEMNED ires
2295  *    are deleted. We don't have to bump up the reference count on the
2296  *    bucket if we are walking the bucket and never have to drop the bucket
2297  *    lock. Note that IRB_REFHOLD does not prevent addition of new ires
2298  *    in the list. It is okay because addition of new ires will not cause
2299  *    ire_next to point to freed memory. We do IRB_REFHOLD only when
2300  *    all of the 3 conditions are true :
2301  *
2302  *    1) The code needs to walk the IRE bucket from start to end.
2303  *    2) It may have to drop the bucket lock sometimes while doing (1)
2304  *    3) It does not want any ires to be deleted meanwhile.
2305  */
2306 
2307 /*
2308  * Bump up the reference count on the IRE. We cannot assert that the
2309  * bucket lock is being held as it is legal to bump up the reference
2310  * count after the first lookup has returned the IRE without
2311  * holding the lock. Currently ip_wput does this for caching IRE_CACHEs.
2312  */
2313 
2314 #ifdef DEBUG
2315 #define	IRE_UNTRACE_REF(ire)	ire_untrace_ref(ire);
2316 #define	IRE_TRACE_REF(ire)	ire_trace_ref(ire);
2317 #else
2318 #define	IRE_UNTRACE_REF(ire)
2319 #define	IRE_TRACE_REF(ire)
2320 #endif
2321 
2322 #define	IRE_REFHOLD_NOTR(ire) {				\
2323 	atomic_add_32(&(ire)->ire_refcnt, 1);		\
2324 	ASSERT((ire)->ire_refcnt != 0);			\
2325 }
2326 
2327 #define	IRE_REFHOLD(ire) {				\
2328 	IRE_REFHOLD_NOTR(ire);				\
2329 	IRE_TRACE_REF(ire);				\
2330 }
2331 
2332 #define	IRE_REFHOLD_LOCKED(ire)	{			\
2333 	IRE_TRACE_REF(ire);				\
2334 	(ire)->ire_refcnt++;				\
2335 }
2336 
2337 /*
2338  * Decrement the reference count on the IRE.
2339  * In architectures e.g sun4u, where atomic_add_32_nv is just
2340  * a cas, we need to maintain the right memory barrier semantics
2341  * as that of mutex_exit i.e all the loads and stores should complete
2342  * before the cas is executed. membar_exit() does that here.
2343  *
2344  * NOTE : This macro is used only in places where we want performance.
2345  *	  To avoid bloating the code, we use the function "ire_refrele"
2346  *	  which essentially calls the macro.
2347  */
2348 #define	IRE_REFRELE_NOTR(ire) {					\
2349 	ASSERT((ire)->ire_refcnt != 0);				\
2350 	membar_exit();						\
2351 	if (atomic_add_32_nv(&(ire)->ire_refcnt, -1) == 0)	\
2352 		ire_inactive(ire);				\
2353 }
2354 
2355 #define	IRE_REFRELE(ire) {					\
2356 	if (ire->ire_bucket != NULL) {				\
2357 		IRE_UNTRACE_REF(ire);				\
2358 	}							\
2359 	IRE_REFRELE_NOTR(ire);					\
2360 }
2361 
2362 /*
2363  * Bump up the reference count on the hash bucket - IRB to
2364  * prevent ires from being deleted in this bucket.
2365  */
2366 #define	IRB_REFHOLD(irb) {				\
2367 	rw_enter(&(irb)->irb_lock, RW_WRITER);		\
2368 	(irb)->irb_refcnt++;				\
2369 	ASSERT((irb)->irb_refcnt != 0);			\
2370 	rw_exit(&(irb)->irb_lock);			\
2371 }
2372 #define	IRB_REFHOLD_LOCKED(irb) {			\
2373 	ASSERT(RW_WRITE_HELD(&(irb)->irb_lock));	\
2374 	(irb)->irb_refcnt++;				\
2375 	ASSERT((irb)->irb_refcnt != 0);			\
2376 }
2377 
2378 void irb_refrele_ftable(irb_t *);
2379 /*
2380  * Note: when IRB_MARK_FTABLE (i.e., IRE_CACHETABLE entry), the irb_t
2381  * is statically allocated, so that when the irb_refcnt goes to 0,
2382  * we simply clean up the ire list and continue.
2383  */
2384 #define	IRB_REFRELE(irb) {				\
2385 	if ((irb)->irb_marks & IRB_MARK_FTABLE) {	\
2386 		irb_refrele_ftable((irb));		\
2387 	} else {					\
2388 		rw_enter(&(irb)->irb_lock, RW_WRITER);		\
2389 		ASSERT((irb)->irb_refcnt != 0);			\
2390 		if (--(irb)->irb_refcnt	== 0 &&			\
2391 		    ((irb)->irb_marks & IRE_MARK_CONDEMNED)) {	\
2392 			ire_t *ire_list;			\
2393 								\
2394 			ire_list = ire_unlink(irb);		\
2395 			rw_exit(&(irb)->irb_lock);		\
2396 			ASSERT(ire_list != NULL);		\
2397 			ire_cleanup(ire_list);			\
2398 		} else {					\
2399 			rw_exit(&(irb)->irb_lock);		\
2400 		}						\
2401 	}							\
2402 }
2403 
2404 extern struct kmem_cache *rt_entry_cache;
2405 
2406 /*
2407  * Lock the fast path mp for access, since the fp_mp can be deleted
2408  * due a DL_NOTE_FASTPATH_FLUSH in the case of IRE_BROADCAST
2409  */
2410 
2411 #define	LOCK_IRE_FP_MP(ire) {				\
2412 		if ((ire)->ire_type == IRE_BROADCAST)	\
2413 			mutex_enter(&ire->ire_nce->nce_lock);	\
2414 	}
2415 #define	UNLOCK_IRE_FP_MP(ire) {				\
2416 		if ((ire)->ire_type == IRE_BROADCAST)	\
2417 			mutex_exit(&ire->ire_nce->nce_lock);	\
2418 	}
2419 
2420 typedef struct ire4 {
2421 	ipaddr_t ire4_src_addr;		/* Source address to use. */
2422 	ipaddr_t ire4_mask;		/* Mask for matching this IRE. */
2423 	ipaddr_t ire4_addr;		/* Address this IRE represents. */
2424 	ipaddr_t ire4_gateway_addr;	/* Gateway if IRE_CACHE/IRE_OFFSUBNET */
2425 	ipaddr_t ire4_cmask;		/* Mask from parent prefix route */
2426 } ire4_t;
2427 
2428 typedef struct ire6 {
2429 	in6_addr_t ire6_src_addr;	/* Source address to use. */
2430 	in6_addr_t ire6_mask;		/* Mask for matching this IRE. */
2431 	in6_addr_t ire6_addr;		/* Address this IRE represents. */
2432 	in6_addr_t ire6_gateway_addr;	/* Gateway if IRE_CACHE/IRE_OFFSUBNET */
2433 	in6_addr_t ire6_cmask;		/* Mask from parent prefix route */
2434 } ire6_t;
2435 
2436 typedef union ire_addr {
2437 	ire6_t	ire6_u;
2438 	ire4_t	ire4_u;
2439 } ire_addr_u_t;
2440 
2441 /* Internet Routing Entry */
2442 typedef struct ire_s {
2443 	struct	ire_s	*ire_next;	/* The hash chain must be first. */
2444 	struct	ire_s	**ire_ptpn;	/* Pointer to previous next. */
2445 	uint32_t	ire_refcnt;	/* Number of references */
2446 	mblk_t		*ire_mp;	/* Non-null if allocated as mblk */
2447 	queue_t		*ire_rfq;	/* recv from this queue */
2448 	queue_t		*ire_stq;	/* send to this queue */
2449 	union {
2450 		uint_t	*max_fragp;	/* Used only during ire creation */
2451 		uint_t	max_frag;	/* MTU (next hop or path). */
2452 	} imf_u;
2453 #define	ire_max_frag	imf_u.max_frag
2454 #define	ire_max_fragp	imf_u.max_fragp
2455 	uint32_t	ire_frag_flag;	/* IPH_DF or zero. */
2456 	uint32_t	ire_ident;	/* Per IRE IP ident. */
2457 	uint32_t	ire_tire_mark;	/* Used for reclaim of unused. */
2458 	uchar_t		ire_ipversion;	/* IPv4/IPv6 version */
2459 	uchar_t		ire_marks;	/* IRE_MARK_CONDEMNED etc. */
2460 	ushort_t	ire_type;	/* Type of IRE */
2461 	uint_t	ire_ib_pkt_count;	/* Inbound packets for ire_addr */
2462 	uint_t	ire_ob_pkt_count;	/* Outbound packets to ire_addr */
2463 	uint_t	ire_ll_hdr_length;	/* Non-zero if we do M_DATA prepends */
2464 	time_t	ire_create_time;	/* Time (in secs) IRE was created. */
2465 	uint32_t	ire_phandle;	/* Associate prefix IREs to cache */
2466 	uint32_t	ire_ihandle;	/* Associate interface IREs to cache */
2467 	ipif_t		*ire_ipif;	/* the interface that this ire uses */
2468 	uint32_t	ire_flags;	/* flags related to route (RTF_*) */
2469 	uint_t	ire_ipsec_overhead;	/* IPSEC overhead */
2470 	/*
2471 	 * Neighbor Cache Entry for IPv6; arp info for IPv4
2472 	 */
2473 	struct	nce_s	*ire_nce;
2474 	uint_t		ire_masklen;	/* # bits in ire_mask{,_v6} */
2475 	ire_addr_u_t	ire_u;		/* IPv4/IPv6 address info. */
2476 
2477 	irb_t		*ire_bucket;	/* Hash bucket when ire_ptphn is set */
2478 	iulp_t		ire_uinfo;	/* Upper layer protocol info. */
2479 	/*
2480 	 * Protects ire_uinfo, ire_max_frag, and ire_frag_flag.
2481 	 */
2482 	kmutex_t	ire_lock;
2483 	uint_t		ire_ipif_seqid; /* ipif_seqid of ire_ipif */
2484 	clock_t		ire_last_used_time;	/* Last used time */
2485 	tsol_ire_gw_secattr_t *ire_gw_secattr; /* gateway security attributes */
2486 	zoneid_t	ire_zoneid;	/* for local address discrimination */
2487 	/*
2488 	 * ire's that are embedded inside mblk_t and sent to the external
2489 	 * resolver use the ire_stq_ifindex to track the ifindex of the
2490 	 * ire_stq, so that the ill (if it exists) can be correctly recovered
2491 	 * for cleanup in the esbfree routine when arp failure occurs
2492 	 */
2493 	uint_t	ire_stq_ifindex;
2494 	uint_t		ire_defense_count;	/* number of ARP conflicts */
2495 	uint_t		ire_defense_time;	/* last time defended (secs) */
2496 	boolean_t	ire_trace_disable;	/* True when alloc fails */
2497 	ip_stack_t	*ire_ipst;	/* Does not have a netstack_hold */
2498 } ire_t;
2499 
2500 /* IPv4 compatiblity macros */
2501 #define	ire_src_addr		ire_u.ire4_u.ire4_src_addr
2502 #define	ire_mask		ire_u.ire4_u.ire4_mask
2503 #define	ire_addr		ire_u.ire4_u.ire4_addr
2504 #define	ire_gateway_addr	ire_u.ire4_u.ire4_gateway_addr
2505 #define	ire_cmask		ire_u.ire4_u.ire4_cmask
2506 
2507 #define	ire_src_addr_v6		ire_u.ire6_u.ire6_src_addr
2508 #define	ire_mask_v6		ire_u.ire6_u.ire6_mask
2509 #define	ire_addr_v6		ire_u.ire6_u.ire6_addr
2510 #define	ire_gateway_addr_v6	ire_u.ire6_u.ire6_gateway_addr
2511 #define	ire_cmask_v6		ire_u.ire6_u.ire6_cmask
2512 
2513 /* Convenient typedefs for sockaddrs */
2514 typedef	struct sockaddr_in	sin_t;
2515 typedef	struct sockaddr_in6	sin6_t;
2516 
2517 /* Address structure used for internal bind with IP */
2518 typedef struct ipa_conn_s {
2519 	ipaddr_t	ac_laddr;
2520 	ipaddr_t	ac_faddr;
2521 	uint16_t	ac_fport;
2522 	uint16_t	ac_lport;
2523 } ipa_conn_t;
2524 
2525 typedef struct ipa6_conn_s {
2526 	in6_addr_t	ac6_laddr;
2527 	in6_addr_t	ac6_faddr;
2528 	uint16_t	ac6_fport;
2529 	uint16_t	ac6_lport;
2530 } ipa6_conn_t;
2531 
2532 /*
2533  * Using ipa_conn_x_t or ipa6_conn_x_t allows us to modify the behavior of IP's
2534  * bind handler.
2535  */
2536 typedef struct ipa_conn_extended_s {
2537 	uint64_t	acx_flags;
2538 	ipa_conn_t	acx_conn;
2539 } ipa_conn_x_t;
2540 
2541 typedef struct ipa6_conn_extended_s {
2542 	uint64_t	ac6x_flags;
2543 	ipa6_conn_t	ac6x_conn;
2544 } ipa6_conn_x_t;
2545 
2546 /* flag values for ipa_conn_x_t and ipa6_conn_x_t. */
2547 #define	ACX_VERIFY_DST	0x1ULL	/* verify destination address is reachable */
2548 
2549 /* Name/Value Descriptor. */
2550 typedef struct nv_s {
2551 	uint64_t nv_value;
2552 	char	*nv_name;
2553 } nv_t;
2554 
2555 #define	ILL_FRAG_HASH(s, i) \
2556 	((ntohl(s) ^ ((i) ^ ((i) >> 8))) % ILL_FRAG_HASH_TBL_COUNT)
2557 
2558 /*
2559  * The MAX number of allowed fragmented packets per hash bucket
2560  * calculation is based on the most common mtu size of 1500. This limit
2561  * will work well for other mtu sizes as well.
2562  */
2563 #define	COMMON_IP_MTU 1500
2564 #define	MAX_FRAG_MIN 10
2565 #define	MAX_FRAG_PKTS(ipst)	\
2566 	MAX(MAX_FRAG_MIN, (2 * (ipst->ips_ip_reass_queue_bytes / \
2567 	    (COMMON_IP_MTU * ILL_FRAG_HASH_TBL_COUNT))))
2568 
2569 /*
2570  * Maximum dups allowed per packet.
2571  */
2572 extern uint_t ip_max_frag_dups;
2573 
2574 /*
2575  * Per-packet information for received packets and transmitted.
2576  * Used by the transport protocols when converting between the packet
2577  * and ancillary data and socket options.
2578  *
2579  * Note: This private data structure and related IPPF_* constant
2580  * definitions are exposed to enable compilation of some debugging tools
2581  * like lsof which use struct tcp_t in <inet/tcp.h>. This is intended to be
2582  * a temporary hack and long term alternate interfaces should be defined
2583  * to support the needs of such tools and private definitions moved to
2584  * private headers.
2585  */
2586 struct ip6_pkt_s {
2587 	uint_t		ipp_fields;		/* Which fields are valid */
2588 	uint_t		ipp_sticky_ignored;	/* sticky fields to ignore */
2589 	uint_t		ipp_ifindex;		/* pktinfo ifindex */
2590 	in6_addr_t	ipp_addr;		/* pktinfo src/dst addr */
2591 	uint_t		ipp_unicast_hops;	/* IPV6_UNICAST_HOPS */
2592 	uint_t		ipp_multicast_hops;	/* IPV6_MULTICAST_HOPS */
2593 	uint_t		ipp_hoplimit;		/* IPV6_HOPLIMIT */
2594 	uint_t		ipp_hopoptslen;
2595 	uint_t		ipp_rtdstoptslen;
2596 	uint_t		ipp_rthdrlen;
2597 	uint_t		ipp_dstoptslen;
2598 	uint_t		ipp_pathmtulen;
2599 	uint_t		ipp_fraghdrlen;
2600 	ip6_hbh_t	*ipp_hopopts;
2601 	ip6_dest_t	*ipp_rtdstopts;
2602 	ip6_rthdr_t	*ipp_rthdr;
2603 	ip6_dest_t	*ipp_dstopts;
2604 	ip6_frag_t	*ipp_fraghdr;
2605 	struct ip6_mtuinfo *ipp_pathmtu;
2606 	in6_addr_t	ipp_nexthop;		/* Transmit only */
2607 	uint8_t		ipp_tclass;
2608 	int8_t		ipp_use_min_mtu;
2609 };
2610 typedef struct ip6_pkt_s ip6_pkt_t;
2611 
2612 extern void ip6_pkt_free(ip6_pkt_t *);	/* free storage inside ip6_pkt_t */
2613 
2614 /*
2615  * This struct is used by ULP_opt_set() functions to return value of IPv4
2616  * ancillary options. Currently this is only used by udp and icmp and only
2617  * IP_PKTINFO option is supported.
2618  */
2619 typedef struct ip4_pkt_s {
2620 	uint_t		ip4_ill_index;	/* interface index */
2621 	ipaddr_t	ip4_addr;	/* source address */
2622 } ip4_pkt_t;
2623 
2624 /*
2625  * Used by ULP's to pass options info to ip_output
2626  * currently only IP_PKTINFO is supported.
2627  */
2628 typedef struct ip_opt_info_s {
2629 	uint_t ip_opt_ill_index;
2630 	uint_t ip_opt_flags;
2631 } ip_opt_info_t;
2632 
2633 /*
2634  * value for ip_opt_flags
2635  */
2636 #define	IP_VERIFY_SRC	0x1
2637 
2638 /*
2639  * This structure is used to convey information from IP and the ULP.
2640  * Currently used for the IP_RECVSLLA, IP_RECVIF and IP_RECVPKTINFO options.
2641  * The type of information field is set to IN_PKTINFO (i.e inbound pkt info)
2642  */
2643 typedef struct ip_pktinfo {
2644 	uint32_t		ip_pkt_ulp_type;	/* type of info sent */
2645 	uint32_t		ip_pkt_flags;	/* what is sent up by IP */
2646 	uint32_t		ip_pkt_ifindex;	/* inbound interface index */
2647 	struct sockaddr_dl	ip_pkt_slla;	/* has source link layer addr */
2648 	struct in_addr		ip_pkt_match_addr; /* matched address */
2649 } ip_pktinfo_t;
2650 
2651 /*
2652  * flags to tell UDP what IP is sending; in_pkt_flags
2653  */
2654 #define	IPF_RECVIF	0x01	/* inbound interface index */
2655 #define	IPF_RECVSLLA	0x02	/* source link layer address */
2656 /*
2657  * Inbound interface index + matched address.
2658  * Used only by IPV4.
2659  */
2660 #define	IPF_RECVADDR	0x04
2661 
2662 /* ipp_fields values */
2663 #define	IPPF_IFINDEX	0x0001	/* Part of in6_pktinfo: ifindex */
2664 #define	IPPF_ADDR	0x0002	/* Part of in6_pktinfo: src/dst addr */
2665 #define	IPPF_SCOPE_ID	0x0004	/* Add xmit ip6i_t for sin6_scope_id */
2666 #define	IPPF_NO_CKSUM	0x0008	/* Add xmit ip6i_t for IP6I_NO_*_CKSUM */
2667 
2668 #define	IPPF_RAW_CKSUM	0x0010	/* Add xmit ip6i_t for IP6I_RAW_CHECKSUM */
2669 #define	IPPF_HOPLIMIT	0x0020
2670 #define	IPPF_HOPOPTS	0x0040
2671 #define	IPPF_RTHDR	0x0080
2672 
2673 #define	IPPF_RTDSTOPTS	0x0100
2674 #define	IPPF_DSTOPTS	0x0200
2675 #define	IPPF_NEXTHOP	0x0400
2676 #define	IPPF_PATHMTU	0x0800
2677 
2678 #define	IPPF_TCLASS	0x1000
2679 #define	IPPF_DONTFRAG	0x2000
2680 #define	IPPF_USE_MIN_MTU	0x04000
2681 #define	IPPF_MULTICAST_HOPS	0x08000
2682 
2683 #define	IPPF_UNICAST_HOPS	0x10000
2684 #define	IPPF_FRAGHDR		0x20000
2685 
2686 #define	IPPF_HAS_IP6I \
2687 	(IPPF_IFINDEX|IPPF_ADDR|IPPF_NEXTHOP|IPPF_SCOPE_ID| \
2688 	IPPF_NO_CKSUM|IPPF_RAW_CKSUM|IPPF_HOPLIMIT|IPPF_DONTFRAG| \
2689 	IPPF_USE_MIN_MTU|IPPF_MULTICAST_HOPS|IPPF_UNICAST_HOPS)
2690 
2691 #define	TCP_PORTS_OFFSET	0
2692 #define	UDP_PORTS_OFFSET	0
2693 
2694 /*
2695  * lookups return the ill/ipif only if the flags are clear OR Iam writer.
2696  * ill / ipif lookup functions increment the refcnt on the ill / ipif only
2697  * after calling these macros. This ensures that the refcnt on the ipif or
2698  * ill will eventually drop down to zero.
2699  */
2700 #define	ILL_LOOKUP_FAILED	1	/* Used as error code */
2701 #define	IPIF_LOOKUP_FAILED	2	/* Used as error code */
2702 
2703 #define	ILL_CAN_LOOKUP(ill) 						\
2704 	(!((ill)->ill_state_flags & (ILL_CONDEMNED | ILL_CHANGING)) ||	\
2705 	IAM_WRITER_ILL(ill))
2706 
2707 #define	ILL_CAN_WAIT(ill, q)	\
2708 	(((q) != NULL) && !((ill)->ill_state_flags & (ILL_CONDEMNED)))
2709 
2710 #define	ILL_CAN_LOOKUP_WALKER(ill)	\
2711 	(!((ill)->ill_state_flags & ILL_CONDEMNED))
2712 
2713 #define	IPIF_CAN_LOOKUP(ipif)						\
2714 	(!((ipif)->ipif_state_flags & (IPIF_CONDEMNED | IPIF_CHANGING)) || \
2715 	IAM_WRITER_IPIF(ipif))
2716 
2717 /*
2718  * If the parameter 'q' is NULL, the caller is not interested in wait and
2719  * restart of the operation if the ILL or IPIF cannot be looked up when it is
2720  * marked as 'CHANGING'. Typically a thread that tries to send out data  will
2721  * end up passing NULLs as the last 4 parameters to ill_lookup_on_ifindex and
2722  * in this case 'q' is NULL
2723  */
2724 #define	IPIF_CAN_WAIT(ipif, q)	\
2725 	(((q) != NULL) && !((ipif)->ipif_state_flags & (IPIF_CONDEMNED)))
2726 
2727 #define	IPIF_CAN_LOOKUP_WALKER(ipif)					\
2728 	(!((ipif)->ipif_state_flags & (IPIF_CONDEMNED)) ||		\
2729 	IAM_WRITER_IPIF(ipif))
2730 
2731 /*
2732  * These macros are used by critical set ioctls and failover ioctls to
2733  * mark the ipif appropriately before starting the operation and to clear the
2734  * marks after completing the operation.
2735  */
2736 #define	IPIF_UNMARK_MOVING(ipif)                                \
2737 	(ipif)->ipif_state_flags &= ~IPIF_MOVING & ~IPIF_CHANGING;
2738 
2739 #define	ILL_UNMARK_CHANGING(ill)                                \
2740 	(ill)->ill_state_flags &= ~ILL_CHANGING;
2741 
2742 /* Macros used to assert that this thread is a writer  */
2743 #define	IAM_WRITER_IPSQ(ipsq)	((ipsq)->ipsq_writer == curthread)
2744 #define	IAM_WRITER_ILL(ill)					\
2745 	((ill)->ill_phyint->phyint_ipsq->ipsq_writer == curthread)
2746 #define	IAM_WRITER_IPIF(ipif)					\
2747 	((ipif)->ipif_ill->ill_phyint->phyint_ipsq->ipsq_writer == curthread)
2748 
2749 /*
2750  * Grab ill locks in the proper order. The order is highest addressed
2751  * ill is locked first.
2752  */
2753 #define	GRAB_ILL_LOCKS(ill_1, ill_2)				\
2754 {								\
2755 	if ((ill_1) > (ill_2)) {				\
2756 		if (ill_1 != NULL)				\
2757 			mutex_enter(&(ill_1)->ill_lock);	\
2758 		if (ill_2 != NULL)				\
2759 			mutex_enter(&(ill_2)->ill_lock);	\
2760 	} else {						\
2761 		if (ill_2 != NULL)				\
2762 			mutex_enter(&(ill_2)->ill_lock);	\
2763 		if (ill_1 != NULL && ill_1 != ill_2)		\
2764 			mutex_enter(&(ill_1)->ill_lock);	\
2765 	}							\
2766 }
2767 
2768 #define	RELEASE_ILL_LOCKS(ill_1, ill_2)		\
2769 {						\
2770 	if (ill_1 != NULL)			\
2771 		mutex_exit(&(ill_1)->ill_lock);	\
2772 	if (ill_2 != NULL && ill_2 != ill_1)	\
2773 		mutex_exit(&(ill_2)->ill_lock);	\
2774 }
2775 
2776 /* Get the other protocol instance ill */
2777 #define	ILL_OTHER(ill)						\
2778 	((ill)->ill_isv6 ? (ill)->ill_phyint->phyint_illv4 :	\
2779 	    (ill)->ill_phyint->phyint_illv6)
2780 
2781 #define	MATCH_V4_ONLY	0x1
2782 #define	MATCH_V6_ONLY	0x2
2783 #define	MATCH_ILL_ONLY	0x4
2784 
2785 /* ioctl command info: Ioctl properties extracted and stored in here */
2786 typedef struct cmd_info_s
2787 {
2788 	char    ci_groupname[LIFNAMSIZ + 1];	/* SIOCSLIFGROUPNAME */
2789 	ipif_t  *ci_ipif;	/* ipif associated with [l]ifreq ioctl's */
2790 	sin_t	*ci_sin;	/* the sin struct passed down */
2791 	sin6_t	*ci_sin6;	/* the sin6_t struct passed down */
2792 	struct lifreq *ci_lifr;	/* the lifreq struct passed down */
2793 } cmd_info_t;
2794 
2795 /*
2796  * List of AH and ESP IPsec acceleration capable ills
2797  */
2798 typedef struct ipsec_capab_ill_s {
2799 	uint_t ill_index;
2800 	boolean_t ill_isv6;
2801 	struct ipsec_capab_ill_s *next;
2802 } ipsec_capab_ill_t;
2803 
2804 extern struct kmem_cache *ire_cache;
2805 
2806 extern ipaddr_t	ip_g_all_ones;
2807 
2808 extern	uint_t	ip_loopback_mtu;	/* /etc/system */
2809 
2810 extern vmem_t *ip_minor_arena;
2811 
2812 /*
2813  * ip_g_forward controls IP forwarding.  It takes two values:
2814  *	0: IP_FORWARD_NEVER	Don't forward packets ever.
2815  *	1: IP_FORWARD_ALWAYS	Forward packets for elsewhere.
2816  *
2817  * RFC1122 says there must be a configuration switch to control forwarding,
2818  * but that the default MUST be to not forward packets ever.  Implicit
2819  * control based on configuration of multiple interfaces MUST NOT be
2820  * implemented (Section 3.1).  SunOS 4.1 did provide the "automatic" capability
2821  * and, in fact, it was the default.  That capability is now provided in the
2822  * /etc/rc2.d/S69inet script.
2823  */
2824 
2825 #define	ips_ip_respond_to_address_mask_broadcast ips_param_arr[0].ip_param_value
2826 #define	ips_ip_g_resp_to_echo_bcast	ips_param_arr[1].ip_param_value
2827 #define	ips_ip_g_resp_to_echo_mcast	ips_param_arr[2].ip_param_value
2828 #define	ips_ip_g_resp_to_timestamp	ips_param_arr[3].ip_param_value
2829 #define	ips_ip_g_resp_to_timestamp_bcast ips_param_arr[4].ip_param_value
2830 #define	ips_ip_g_send_redirects		ips_param_arr[5].ip_param_value
2831 #define	ips_ip_g_forward_directed_bcast	ips_param_arr[6].ip_param_value
2832 #define	ips_ip_debug			ips_param_arr[7].ip_param_value
2833 #define	ips_ip_mrtdebug			ips_param_arr[8].ip_param_value
2834 #define	ips_ip_timer_interval		ips_param_arr[9].ip_param_value
2835 #define	ips_ip_ire_arp_interval		ips_param_arr[10].ip_param_value
2836 #define	ips_ip_ire_redir_interval	ips_param_arr[11].ip_param_value
2837 #define	ips_ip_def_ttl			ips_param_arr[12].ip_param_value
2838 #define	ips_ip_forward_src_routed	ips_param_arr[13].ip_param_value
2839 #define	ips_ip_wroff_extra		ips_param_arr[14].ip_param_value
2840 #define	ips_ip_ire_pathmtu_interval	ips_param_arr[15].ip_param_value
2841 #define	ips_ip_icmp_return		ips_param_arr[16].ip_param_value
2842 #define	ips_ip_path_mtu_discovery	ips_param_arr[17].ip_param_value
2843 #define	ips_ip_ignore_delete_time	ips_param_arr[18].ip_param_value
2844 #define	ips_ip_ignore_redirect		ips_param_arr[19].ip_param_value
2845 #define	ips_ip_output_queue		ips_param_arr[20].ip_param_value
2846 #define	ips_ip_broadcast_ttl		ips_param_arr[21].ip_param_value
2847 #define	ips_ip_icmp_err_interval	ips_param_arr[22].ip_param_value
2848 #define	ips_ip_icmp_err_burst		ips_param_arr[23].ip_param_value
2849 #define	ips_ip_reass_queue_bytes	ips_param_arr[24].ip_param_value
2850 #define	ips_ip_strict_dst_multihoming	ips_param_arr[25].ip_param_value
2851 #define	ips_ip_addrs_per_if		ips_param_arr[26].ip_param_value
2852 #define	ips_ipsec_override_persocket_policy ips_param_arr[27].ip_param_value
2853 #define	ips_icmp_accept_clear_messages	ips_param_arr[28].ip_param_value
2854 #define	ips_igmp_accept_clear_messages	ips_param_arr[29].ip_param_value
2855 
2856 /* IPv6 configuration knobs */
2857 #define	ips_delay_first_probe_time	ips_param_arr[30].ip_param_value
2858 #define	ips_max_unicast_solicit		ips_param_arr[31].ip_param_value
2859 #define	ips_ipv6_def_hops		ips_param_arr[32].ip_param_value
2860 #define	ips_ipv6_icmp_return		ips_param_arr[33].ip_param_value
2861 #define	ips_ipv6_forward_src_routed	ips_param_arr[34].ip_param_value
2862 #define	ips_ipv6_resp_echo_mcast	ips_param_arr[35].ip_param_value
2863 #define	ips_ipv6_send_redirects		ips_param_arr[36].ip_param_value
2864 #define	ips_ipv6_ignore_redirect	ips_param_arr[37].ip_param_value
2865 #define	ips_ipv6_strict_dst_multihoming	ips_param_arr[38].ip_param_value
2866 #define	ips_ip_ire_reclaim_fraction	ips_param_arr[39].ip_param_value
2867 #define	ips_ipsec_policy_log_interval	ips_param_arr[40].ip_param_value
2868 #define	ips_pim_accept_clear_messages	ips_param_arr[41].ip_param_value
2869 #define	ips_ip_ndp_unsolicit_interval	ips_param_arr[42].ip_param_value
2870 #define	ips_ip_ndp_unsolicit_count	ips_param_arr[43].ip_param_value
2871 #define	ips_ipv6_ignore_home_address_opt ips_param_arr[44].ip_param_value
2872 #define	ips_ip_policy_mask		ips_param_arr[45].ip_param_value
2873 #define	ips_ip_multirt_resolution_interval ips_param_arr[46].ip_param_value
2874 #define	ips_ip_multirt_ttl  		ips_param_arr[47].ip_param_value
2875 #define	ips_ip_multidata_outbound	ips_param_arr[48].ip_param_value
2876 #define	ips_ip_ndp_defense_interval	ips_param_arr[49].ip_param_value
2877 #define	ips_ip_max_temp_idle		ips_param_arr[50].ip_param_value
2878 #define	ips_ip_max_temp_defend		ips_param_arr[51].ip_param_value
2879 #define	ips_ip_max_defend		ips_param_arr[52].ip_param_value
2880 #define	ips_ip_defend_interval		ips_param_arr[53].ip_param_value
2881 #define	ips_ip_dup_recovery		ips_param_arr[54].ip_param_value
2882 #define	ips_ip_restrict_interzone_loopback ips_param_arr[55].ip_param_value
2883 #define	ips_ip_lso_outbound		ips_param_arr[56].ip_param_value
2884 #define	ips_igmp_max_version		ips_param_arr[57].ip_param_value
2885 #define	ips_mld_max_version		ips_param_arr[58].ip_param_value
2886 #define	ips_ipv6_drop_inbound_icmpv6	ips_param_arr[59].ip_param_value
2887 
2888 extern int	dohwcksum;	/* use h/w cksum if supported by the h/w */
2889 #ifdef ZC_TEST
2890 extern int	noswcksum;
2891 #endif
2892 
2893 extern char	ipif_loopback_name[];
2894 
2895 extern nv_t	*ire_nv_tbl;
2896 
2897 extern struct module_info ip_mod_info;
2898 
2899 #define	HOOKS4_INTERESTED_PHYSICAL_IN(ipst)	\
2900 	((ipst)->ips_ip4_physical_in_event.he_interested)
2901 #define	HOOKS6_INTERESTED_PHYSICAL_IN(ipst)	\
2902 	((ipst)->ips_ip6_physical_in_event.he_interested)
2903 #define	HOOKS4_INTERESTED_PHYSICAL_OUT(ipst)	\
2904 	((ipst)->ips_ip4_physical_out_event.he_interested)
2905 #define	HOOKS6_INTERESTED_PHYSICAL_OUT(ipst)	\
2906 	((ipst)->ips_ip6_physical_out_event.he_interested)
2907 #define	HOOKS4_INTERESTED_FORWARDING(ipst)	\
2908 	((ipst)->ips_ip4_forwarding_event.he_interested)
2909 #define	HOOKS6_INTERESTED_FORWARDING(ipst)	\
2910 	((ipst)->ips_ip6_forwarding_event.he_interested)
2911 #define	HOOKS4_INTERESTED_LOOPBACK_IN(ipst)	\
2912 	((ipst)->ips_ip4_loopback_in_event.he_interested)
2913 #define	HOOKS6_INTERESTED_LOOPBACK_IN(ipst)	\
2914 	((ipst)->ips_ip6_loopback_in_event.he_interested)
2915 #define	HOOKS4_INTERESTED_LOOPBACK_OUT(ipst)	\
2916 	((ipst)->ips_ip4_loopback_out_event.he_interested)
2917 #define	HOOKS6_INTERESTED_LOOPBACK_OUT(ipst)	\
2918 	((ipst)->ips_ip6_loopback_out_event.he_interested)
2919 
2920 /*
2921  * Hooks marcos used inside of ip
2922  */
2923 #define	IPHA_VHL	ipha_version_and_hdr_length
2924 
2925 #define	FW_HOOKS(_hook, _event, _ilp, _olp, _iph, _fm, _m, ipst)	\
2926 									\
2927 	if ((_hook).he_interested) {	\
2928 		hook_pkt_event_t info;					\
2929 									\
2930 		_NOTE(CONSTCOND)					\
2931 		ASSERT((_ilp != NULL) || (_olp != NULL));		\
2932 									\
2933 		_NOTE(CONSTCOND)					\
2934 		if ((_ilp != NULL) &&					\
2935 		    (((ill_t *)(_ilp))->ill_phyint != NULL))		\
2936 			info.hpe_ifp = (phy_if_t)((ill_t *)		\
2937 			    (_ilp))->ill_phyint->phyint_hook_ifindex;	\
2938 		else							\
2939 			info.hpe_ifp = 0;				\
2940 									\
2941 		_NOTE(CONSTCOND)					\
2942 		if ((_olp != NULL) &&					\
2943 		    (((ill_t *)(_olp))->ill_phyint != NULL))		\
2944 			info.hpe_ofp = (phy_if_t)((ill_t *)		\
2945 			    (_olp))->ill_phyint->phyint_hook_ifindex;	\
2946 		else							\
2947 			info.hpe_ofp = 0;				\
2948 		info.hpe_hdr = _iph;					\
2949 		info.hpe_mp = &(_fm);					\
2950 		info.hpe_mb = _m;					\
2951 		if (hook_run(_event, (hook_data_t)&info,		\
2952 		    ipst->ips_netstack) != 0) {				\
2953 			ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
2954 			    (_hook).he_name, (void *)_fm, (void *)_m));	\
2955 			if (_fm != NULL) {				\
2956 				freemsg(_fm);				\
2957 				_fm = NULL;				\
2958 			}						\
2959 			_iph = NULL;					\
2960 			_m = NULL;					\
2961 		} else {						\
2962 			_iph = info.hpe_hdr;				\
2963 			_m = info.hpe_mb;				\
2964 		}							\
2965 	}
2966 
2967 #define	FW_HOOKS6(_hook, _event, _ilp, _olp, _iph, _fm, _m, ipst)	\
2968 									\
2969 	if ((_hook).he_interested) {	\
2970 		hook_pkt_event_t info;					\
2971 									\
2972 		_NOTE(CONSTCOND)					\
2973 		ASSERT((_ilp != NULL) || (_olp != NULL));		\
2974 									\
2975 		_NOTE(CONSTCOND)					\
2976 		if ((_ilp != NULL) &&					\
2977 		    (((ill_t *)(_ilp))->ill_phyint != NULL))		\
2978 			info.hpe_ifp = (phy_if_t)((ill_t *)		\
2979 			    (_ilp))->ill_phyint->phyint_hook_ifindex;	\
2980 		else							\
2981 			info.hpe_ifp = 0;				\
2982 									\
2983 		_NOTE(CONSTCOND)					\
2984 		if ((_olp != NULL) &&					\
2985 		    (((ill_t *)(_olp))->ill_phyint != NULL))		\
2986 			info.hpe_ofp = (phy_if_t)((ill_t *)		\
2987 			    (_olp))->ill_phyint->phyint_hook_ifindex;	\
2988 		else							\
2989 			info.hpe_ofp = 0;				\
2990 		info.hpe_hdr = _iph;					\
2991 		info.hpe_mp = &(_fm);					\
2992 		info.hpe_mb = _m;					\
2993 		if (hook_run(_event, (hook_data_t)&info,		\
2994 		    ipst->ips_netstack) != 0) {				\
2995 			ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
2996 			    (_hook).he_name, (void *)_fm, (void *)_m));	\
2997 			if (_fm != NULL) {				\
2998 				freemsg(_fm);				\
2999 				_fm = NULL;				\
3000 			}						\
3001 			_iph = NULL;					\
3002 			_m = NULL;					\
3003 		} else {						\
3004 			_iph = info.hpe_hdr;				\
3005 			_m = info.hpe_mb;				\
3006 		}							\
3007 	}
3008 
3009 /*
3010  * Network byte order macros
3011  */
3012 #ifdef	_BIG_ENDIAN
3013 #define	N_IN_CLASSD_NET		IN_CLASSD_NET
3014 #define	N_INADDR_UNSPEC_GROUP	INADDR_UNSPEC_GROUP
3015 #else /* _BIG_ENDIAN */
3016 #define	N_IN_CLASSD_NET		(ipaddr_t)0x000000f0U
3017 #define	N_INADDR_UNSPEC_GROUP	(ipaddr_t)0x000000e0U
3018 #endif /* _BIG_ENDIAN */
3019 #define	CLASSD(addr)	(((addr) & N_IN_CLASSD_NET) == N_INADDR_UNSPEC_GROUP)
3020 
3021 #define	IP_LOOPBACK_ADDR(addr)			\
3022 	((ntohl(addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET)
3023 
3024 #ifdef DEBUG
3025 /* IPsec HW acceleration debugging support */
3026 
3027 #define	IPSECHW_CAPAB		0x0001	/* capability negotiation */
3028 #define	IPSECHW_SADB		0x0002	/* SADB exchange */
3029 #define	IPSECHW_PKT		0x0004	/* general packet flow */
3030 #define	IPSECHW_PKTIN		0x0008	/* driver in pkt processing details */
3031 #define	IPSECHW_PKTOUT		0x0010	/* driver out pkt processing details */
3032 
3033 #define	IPSECHW_DEBUG(f, x)	if (ipsechw_debug & (f)) { (void) printf x; }
3034 #define	IPSECHW_CALL(f, r, x)	if (ipsechw_debug & (f)) { (void) r x; }
3035 
3036 extern uint32_t ipsechw_debug;
3037 #else
3038 #define	IPSECHW_DEBUG(f, x)	{}
3039 #define	IPSECHW_CALL(f, r, x)	{}
3040 #endif
3041 
3042 extern int	ip_debug;
3043 extern uint_t	ip_thread_data;
3044 extern krwlock_t ip_thread_rwlock;
3045 extern list_t	ip_thread_list;
3046 
3047 #ifdef IP_DEBUG
3048 #include <sys/debug.h>
3049 #include <sys/promif.h>
3050 
3051 #define	ip0dbg(a)	printf a
3052 #define	ip1dbg(a)	if (ip_debug > 2) printf a
3053 #define	ip2dbg(a)	if (ip_debug > 3) printf a
3054 #define	ip3dbg(a)	if (ip_debug > 4) printf a
3055 #else
3056 #define	ip0dbg(a)	/* */
3057 #define	ip1dbg(a)	/* */
3058 #define	ip2dbg(a)	/* */
3059 #define	ip3dbg(a)	/* */
3060 #endif	/* IP_DEBUG */
3061 
3062 /* Default MAC-layer address string length for mac_colon_addr */
3063 #define	MAC_STR_LEN	128
3064 
3065 struct	ipsec_out_s;
3066 
3067 struct	mac_header_info_s;
3068 
3069 extern boolean_t ip_assign_ifindex(uint_t *, ip_stack_t *);
3070 extern const char *dlpi_prim_str(int);
3071 extern const char *dlpi_err_str(int);
3072 extern void	ill_frag_timer(void *);
3073 extern ill_t	*ill_first(int, int, ill_walk_context_t *, ip_stack_t *);
3074 extern ill_t	*ill_next(ill_walk_context_t *, ill_t *);
3075 extern void	ill_frag_timer_start(ill_t *);
3076 extern void	ill_nic_info_dispatch(ill_t *);
3077 extern void	ill_nic_info_plumb(ill_t *, boolean_t);
3078 extern mblk_t	*ip_carve_mp(mblk_t **, ssize_t);
3079 extern mblk_t	*ip_dlpi_alloc(size_t, t_uscalar_t);
3080 extern char	*ip_dot_addr(ipaddr_t, char *);
3081 extern const char *mac_colon_addr(const uint8_t *, size_t, char *, size_t);
3082 extern void	ip_lwput(queue_t *, mblk_t *);
3083 extern boolean_t icmp_err_rate_limit(ip_stack_t *);
3084 extern void	icmp_time_exceeded(queue_t *, mblk_t *, uint8_t, zoneid_t,
3085     ip_stack_t *);
3086 extern void	icmp_unreachable(queue_t *, mblk_t *, uint8_t, zoneid_t,
3087     ip_stack_t *);
3088 extern mblk_t	*ip_add_info(mblk_t *, ill_t *, uint_t, zoneid_t, ip_stack_t *);
3089 extern mblk_t	*ip_bind_v4(queue_t *, mblk_t *, conn_t *);
3090 extern int	ip_bind_connected(conn_t *, mblk_t *, ipaddr_t *, uint16_t,
3091     ipaddr_t, uint16_t, boolean_t, boolean_t, boolean_t,
3092     boolean_t);
3093 extern boolean_t ip_bind_ipsec_policy_set(conn_t *, mblk_t *);
3094 extern int	ip_bind_laddr(conn_t *, mblk_t *, ipaddr_t, uint16_t,
3095     boolean_t, boolean_t, boolean_t);
3096 extern uint_t	ip_cksum(mblk_t *, int, uint32_t);
3097 extern int	ip_close(queue_t *, int);
3098 extern uint16_t	ip_csum_hdr(ipha_t *);
3099 extern void	ip_proto_not_sup(queue_t *, mblk_t *, uint_t, zoneid_t,
3100     ip_stack_t *);
3101 extern void	ip_ire_g_fini(void);
3102 extern void	ip_ire_g_init(void);
3103 extern void	ip_ire_fini(ip_stack_t *);
3104 extern void	ip_ire_init(ip_stack_t *);
3105 extern int	ip_open(queue_t *, dev_t *, int, int, cred_t *);
3106 extern int	ip_reassemble(mblk_t *, ipf_t *, uint_t, boolean_t, ill_t *,
3107     size_t);
3108 extern int	ip_opt_set_ill(conn_t *, int, boolean_t, boolean_t,
3109     int, int, mblk_t *);
3110 extern void	ip_rput(queue_t *, mblk_t *);
3111 extern void	ip_input(ill_t *, ill_rx_ring_t *, mblk_t *,
3112     struct mac_header_info_s *);
3113 extern void	ip_rput_dlpi(queue_t *, mblk_t *);
3114 extern void	ip_rput_forward(ire_t *, ipha_t *, mblk_t *, ill_t *);
3115 extern void	ip_rput_forward_multicast(ipaddr_t, mblk_t *, ipif_t *);
3116 
3117 extern int	ip_snmpmod_close(queue_t *);
3118 extern void	ip_snmpmod_wput(queue_t *, mblk_t *);
3119 extern void	ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *,
3120 		    mib2_ipIfStatsEntry_t *);
3121 extern void	ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *,
3122 		    mib2_ipv6IfIcmpEntry_t *);
3123 extern void	ip_udp_input(queue_t *, mblk_t *, ipha_t *, ire_t *, ill_t *);
3124 extern void	ip_proto_input(queue_t *, mblk_t *, ipha_t *, ire_t *, ill_t *,
3125     boolean_t);
3126 extern void	ip_rput_other(ipsq_t *, queue_t *, mblk_t *, void *);
3127 extern ire_t	*ip_check_multihome(void *, ire_t *, ill_t *);
3128 extern void	ip_setqinfo(queue_t *, minor_t, boolean_t, ip_stack_t *);
3129 extern void	ip_trash_ire_reclaim(void *);
3130 extern void	ip_trash_timer_expire(void *);
3131 extern void	ip_wput(queue_t *, mblk_t *);
3132 extern void	ip_output(void *, mblk_t *, void *, int);
3133 extern void	ip_output_options(void *, mblk_t *, void *, int,
3134     ip_opt_info_t *);
3135 
3136 extern void	ip_wput_ire(queue_t *, mblk_t *, ire_t *, conn_t *, int,
3137 		    zoneid_t);
3138 extern void	ip_wput_local(queue_t *, ill_t *, ipha_t *, mblk_t *, ire_t *,
3139 		    int, zoneid_t);
3140 extern void	ip_wput_multicast(queue_t *, mblk_t *, ipif_t *, zoneid_t);
3141 extern void	ip_wput_nondata(ipsq_t *, queue_t *, mblk_t *, void *);
3142 extern void	ip_wsrv(queue_t *);
3143 extern char	*ip_nv_lookup(nv_t *, int);
3144 extern boolean_t ip_local_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3145 extern boolean_t ip_remote_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3146 extern ipaddr_t ip_massage_options(ipha_t *, netstack_t *);
3147 extern ipaddr_t ip_net_mask(ipaddr_t);
3148 extern void	ip_newroute(queue_t *, mblk_t *, ipaddr_t, conn_t *, zoneid_t,
3149 		    ip_stack_t *);
3150 extern ipxmit_state_t	ip_xmit_v4(mblk_t *, ire_t *, struct ipsec_out_s *,
3151     boolean_t);
3152 extern int	ip_hdr_complete(ipha_t *, zoneid_t, ip_stack_t *);
3153 
3154 extern struct qinit rinit_ipv6;
3155 extern struct qinit winit_ipv6;
3156 
3157 extern void	conn_drain_insert(conn_t *connp);
3158 extern	int	conn_ipsec_length(conn_t *connp);
3159 extern void	ip_wput_ipsec_out(queue_t *, mblk_t *, ipha_t *, ill_t *,
3160     ire_t *);
3161 extern ipaddr_t	ip_get_dst(ipha_t *);
3162 extern int	ipsec_out_extra_length(mblk_t *);
3163 extern int	ipsec_in_extra_length(mblk_t *);
3164 extern mblk_t	*ipsec_in_alloc(boolean_t, netstack_t *);
3165 extern boolean_t ipsec_in_is_secure(mblk_t *);
3166 extern void	ipsec_out_process(queue_t *, mblk_t *, ire_t *, uint_t);
3167 extern void	ipsec_out_to_in(mblk_t *);
3168 extern void	ip_fanout_proto_again(mblk_t *, ill_t *, ill_t *, ire_t *);
3169 
3170 extern void	ire_cleanup(ire_t *);
3171 extern void	ire_inactive(ire_t *);
3172 extern boolean_t irb_inactive(irb_t *);
3173 extern ire_t	*ire_unlink(irb_t *);
3174 
3175 #ifdef DEBUG
3176 extern	boolean_t th_trace_ref(const void *, ip_stack_t *);
3177 extern	void	th_trace_unref(const void *);
3178 extern	void	th_trace_cleanup(const void *, boolean_t);
3179 extern	void	ire_trace_ref(ire_t *);
3180 extern	void	ire_untrace_ref(ire_t *);
3181 #endif
3182 
3183 extern int	ip_srcid_insert(const in6_addr_t *, zoneid_t, ip_stack_t *);
3184 extern int	ip_srcid_remove(const in6_addr_t *, zoneid_t, ip_stack_t *);
3185 extern void	ip_srcid_find_id(uint_t, in6_addr_t *, zoneid_t, netstack_t *);
3186 extern uint_t	ip_srcid_find_addr(const in6_addr_t *, zoneid_t, netstack_t *);
3187 extern int	ip_srcid_report(queue_t *, mblk_t *, caddr_t, cred_t *);
3188 
3189 extern uint8_t	ipoptp_next(ipoptp_t *);
3190 extern uint8_t	ipoptp_first(ipoptp_t *, ipha_t *);
3191 extern int	ip_opt_get_user(const ipha_t *, uchar_t *);
3192 extern ill_t	*ip_grab_attach_ill(ill_t *, mblk_t *, int, boolean_t,
3193     ip_stack_t *);
3194 extern ire_t	*conn_set_outgoing_ill(conn_t *, ire_t *, ill_t **);
3195 extern int	ipsec_req_from_conn(conn_t *, ipsec_req_t *, int);
3196 extern int	ip_snmp_get(queue_t *q, mblk_t *mctl);
3197 extern int	ip_snmp_set(queue_t *q, int, int, uchar_t *, int);
3198 extern void	ip_process_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3199 extern void	ip_quiesce_conn(conn_t *);
3200 extern  void    ip_reprocess_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3201 extern void	ip_restart_optmgmt(ipsq_t *, queue_t *, mblk_t *, void *);
3202 extern void	ip_ioctl_finish(queue_t *, mblk_t *, int, int, ipsq_t *);
3203 
3204 extern boolean_t ip_cmpbuf(const void *, uint_t, boolean_t, const void *,
3205     uint_t);
3206 extern boolean_t ip_allocbuf(void **, uint_t *, boolean_t, const void *,
3207     uint_t);
3208 extern void	ip_savebuf(void **, uint_t *, boolean_t, const void *, uint_t);
3209 
3210 extern boolean_t	ipsq_pending_mp_cleanup(ill_t *, conn_t *);
3211 extern void	conn_ioctl_cleanup(conn_t *);
3212 extern ill_t	*conn_get_held_ill(conn_t *, ill_t **, int *);
3213 extern ill_t	*ip_newroute_get_dst_ill(ill_t *);
3214 
3215 struct multidata_s;
3216 struct pdesc_s;
3217 
3218 extern mblk_t	*ip_mdinfo_alloc(ill_mdt_capab_t *);
3219 extern mblk_t	*ip_mdinfo_return(ire_t *, conn_t *, char *, ill_mdt_capab_t *);
3220 extern mblk_t	*ip_lsoinfo_alloc(ill_lso_capab_t *);
3221 extern mblk_t	*ip_lsoinfo_return(ire_t *, conn_t *, char *,
3222     ill_lso_capab_t *);
3223 extern uint_t	ip_md_cksum(struct pdesc_s *, int, uint_t);
3224 extern boolean_t ip_md_addr_attr(struct multidata_s *, struct pdesc_s *,
3225 			const mblk_t *);
3226 extern boolean_t ip_md_hcksum_attr(struct multidata_s *, struct pdesc_s *,
3227 			uint32_t, uint32_t, uint32_t, uint32_t);
3228 extern boolean_t ip_md_zcopy_attr(struct multidata_s *, struct pdesc_s *,
3229 			uint_t);
3230 extern mblk_t	*ip_unbind(queue_t *, mblk_t *);
3231 
3232 extern phyint_t *phyint_lookup_group(char *, boolean_t, ip_stack_t *);
3233 extern phyint_t *phyint_lookup_group_ifindex(uint_t, ip_stack_t *);
3234 
3235 extern void tnet_init(void);
3236 extern void tnet_fini(void);
3237 
3238 /* Hooks for CGTP (multirt routes) filtering module */
3239 #define	CGTP_FILTER_REV_1	1
3240 #define	CGTP_FILTER_REV_2	2
3241 #define	CGTP_FILTER_REV_3	3
3242 #define	CGTP_FILTER_REV		CGTP_FILTER_REV_3
3243 
3244 /* cfo_filter and cfo_filter_v6 hooks return values */
3245 #define	CGTP_IP_PKT_NOT_CGTP	0
3246 #define	CGTP_IP_PKT_PREMIUM	1
3247 #define	CGTP_IP_PKT_DUPLICATE	2
3248 
3249 /* Version 3 of the filter interface */
3250 typedef struct cgtp_filter_ops {
3251 	int	cfo_filter_rev;			/* CGTP_FILTER_REV_3 */
3252 	int	(*cfo_change_state)(netstackid_t, int);
3253 	int	(*cfo_add_dest_v4)(netstackid_t, ipaddr_t, ipaddr_t,
3254 		    ipaddr_t, ipaddr_t);
3255 	int	(*cfo_del_dest_v4)(netstackid_t, ipaddr_t, ipaddr_t);
3256 	int	(*cfo_add_dest_v6)(netstackid_t, in6_addr_t *, in6_addr_t *,
3257 		    in6_addr_t *, in6_addr_t *);
3258 	int	(*cfo_del_dest_v6)(netstackid_t, in6_addr_t *, in6_addr_t *);
3259 	int	(*cfo_filter)(netstackid_t, uint_t, mblk_t *);
3260 	int	(*cfo_filter_v6)(netstackid_t, uint_t, ip6_t *,
3261 		    ip6_frag_t *);
3262 } cgtp_filter_ops_t;
3263 
3264 #define	CGTP_MCAST_SUCCESS	1
3265 
3266 /*
3267  * The separate CGTP module needs this global symbol so that it
3268  * can check the version and determine whether to use the old or the new
3269  * version of the filtering interface.
3270  */
3271 extern int	ip_cgtp_filter_rev;
3272 
3273 extern int	ip_cgtp_filter_supported(void);
3274 extern int	ip_cgtp_filter_register(netstackid_t, cgtp_filter_ops_t *);
3275 extern int	ip_cgtp_filter_unregister(netstackid_t);
3276 extern int	ip_cgtp_filter_is_registered(netstackid_t);
3277 
3278 /* Flags for ire_multirt_lookup() */
3279 
3280 #define	MULTIRT_USESTAMP	0x0001
3281 #define	MULTIRT_SETSTAMP	0x0002
3282 #define	MULTIRT_CACHEGW		0x0004
3283 
3284 /* Debug stuff for multirt route resolution. */
3285 #if defined(DEBUG) && !defined(__lint)
3286 /* Our "don't send, rather drop" flag. */
3287 #define	MULTIRT_DEBUG_FLAG	0x8000
3288 
3289 #define	MULTIRT_TRACE(x)	ip2dbg(x)
3290 
3291 #define	MULTIRT_DEBUG_TAG(mblk)	\
3292 	do { \
3293 		ASSERT(mblk != NULL); \
3294 		MULTIRT_TRACE(("%s[%d]: tagging mblk %p, tag was %d\n", \
3295 		__FILE__, __LINE__, \
3296 		(void *)(mblk), (mblk)->b_flag & MULTIRT_DEBUG_FLAG)); \
3297 		(mblk)->b_flag |= MULTIRT_DEBUG_FLAG; \
3298 	} while (0)
3299 
3300 #define	MULTIRT_DEBUG_UNTAG(mblk) \
3301 	do { \
3302 		ASSERT(mblk != NULL); \
3303 		MULTIRT_TRACE(("%s[%d]: untagging mblk %p, tag was %d\n", \
3304 		__FILE__, __LINE__, \
3305 		(void *)(mblk), (mblk)->b_flag & MULTIRT_DEBUG_FLAG)); \
3306 		(mblk)->b_flag &= ~MULTIRT_DEBUG_FLAG; \
3307 	} while (0)
3308 
3309 #define	MULTIRT_DEBUG_TAGGED(mblk) \
3310 	(((mblk)->b_flag & MULTIRT_DEBUG_FLAG) ? B_TRUE : B_FALSE)
3311 #else
3312 #define	MULTIRT_DEBUG_TAG(mblk)		ASSERT(mblk != NULL)
3313 #define	MULTIRT_DEBUG_UNTAG(mblk)	ASSERT(mblk != NULL)
3314 #define	MULTIRT_DEBUG_TAGGED(mblk)	B_FALSE
3315 #endif
3316 
3317 /*
3318  * Per-ILL Multidata Transmit capabilities.
3319  */
3320 struct ill_mdt_capab_s {
3321 	uint_t ill_mdt_version;  /* interface version */
3322 	uint_t ill_mdt_on;	 /* on/off switch for MDT on this ILL */
3323 	uint_t ill_mdt_hdr_head; /* leading header fragment extra space */
3324 	uint_t ill_mdt_hdr_tail; /* trailing header fragment extra space */
3325 	uint_t ill_mdt_max_pld;	 /* maximum payload buffers per Multidata */
3326 	uint_t ill_mdt_span_limit; /* maximum payload span per packet */
3327 };
3328 
3329 struct ill_hcksum_capab_s {
3330 	uint_t	ill_hcksum_version;	/* interface version */
3331 	uint_t	ill_hcksum_txflags;	/* capabilities on transmit */
3332 };
3333 
3334 struct ill_zerocopy_capab_s {
3335 	uint_t	ill_zerocopy_version;	/* interface version */
3336 	uint_t	ill_zerocopy_flags;	/* capabilities */
3337 };
3338 
3339 struct ill_lso_capab_s {
3340 	uint_t	ill_lso_version;	/* interface version */
3341 	uint_t	ill_lso_on;		/* on/off switch for LSO on this ILL */
3342 	uint_t	ill_lso_flags;		/* capabilities */
3343 	uint_t	ill_lso_max;		/* maximum size of payload */
3344 };
3345 
3346 /* Possible ill_states */
3347 #define	ILL_RING_INPROC		3	/* Being assigned to squeue */
3348 #define	ILL_RING_INUSE		2	/* Already Assigned to Rx Ring */
3349 #define	ILL_RING_BEING_FREED	1	/* Being Unassigned */
3350 #define	ILL_RING_FREE		0	/* Available to be assigned to Ring */
3351 
3352 #define	ILL_MAX_RINGS		256	/* Max num of rx rings we can manage */
3353 #define	ILL_POLLING		0x01	/* Polling in use */
3354 
3355 /*
3356  * These functions pointer types are exported by the mac/dls layer.
3357  * we need to duplicate the definitions here because we cannot
3358  * include mac/dls header files here.
3359  */
3360 typedef void	(*ip_mac_blank_t)(void *, time_t, uint_t);
3361 typedef void	(*ip_dld_tx_t)(void *, mblk_t *);
3362 
3363 typedef void	(*ip_dls_chg_soft_ring_t)(void *, int);
3364 typedef void	(*ip_dls_bind_t)(void *, processorid_t);
3365 typedef void	(*ip_dls_unbind_t)(void *);
3366 
3367 struct ill_rx_ring {
3368 	ip_mac_blank_t		rr_blank; /* Driver interrupt blanking func */
3369 	void			*rr_handle; /* Handle for Rx ring */
3370 	squeue_t		*rr_sqp; /* Squeue the ring is bound to */
3371 	ill_t			*rr_ill; /* back pointer to ill */
3372 	clock_t			rr_poll_time; /* Last lbolt polling was used */
3373 	uint32_t		rr_poll_state; /* polling state flags */
3374 	uint32_t		rr_max_blank_time; /* Max interrupt blank */
3375 	uint32_t		rr_min_blank_time; /* Min interrupt blank */
3376 	uint32_t		rr_max_pkt_cnt; /* Max pkts before interrupt */
3377 	uint32_t		rr_min_pkt_cnt; /* Mix pkts before interrupt */
3378 	uint32_t		rr_normal_blank_time; /* Normal intr freq */
3379 	uint32_t		rr_normal_pkt_cnt; /* Normal intr pkt cnt */
3380 	uint32_t		rr_ring_state; /* State of this ring */
3381 };
3382 
3383 struct ill_dls_capab_s {
3384 	ip_dld_tx_t		ill_tx;		/* Driver Tx routine */
3385 	void			*ill_tx_handle;	/* Driver Tx handle */
3386 	ip_dls_chg_soft_ring_t	ill_dls_change_status;
3387 						/* change soft ring fanout */
3388 	ip_dls_bind_t		ill_dls_bind;	/* to add CPU affinity */
3389 	ip_dls_unbind_t		ill_dls_unbind;	/* remove CPU affinity */
3390 	ill_rx_ring_t		*ill_ring_tbl; /* Ring to Sqp mapping table */
3391 	uint_t			ill_dls_soft_ring_cnt; /* Number of soft ring */
3392 	conn_t			*ill_unbind_conn; /* Conn used during unplumb */
3393 };
3394 
3395 /*
3396  * This message is sent by an upper-layer protocol to tell IP that it knows all
3397  * about labels and will construct them itself.  IP takes the slow path and
3398  * recomputes the label on every packet when this isn't true.
3399  */
3400 #define	IP_ULP_OUT_LABELED		(('O' << 8) + 'L')
3401 typedef struct out_labeled_s {
3402 	uint32_t	out_labeled_type;	/* OUT_LABELED */
3403 	queue_t		*out_qnext;		/* intermediate detection */
3404 } out_labeled_t;
3405 
3406 /*
3407  * IP squeues exports
3408  */
3409 extern int 		ip_squeue_profile;
3410 extern int 		ip_squeue_bind;
3411 extern boolean_t 	ip_squeue_fanout;
3412 extern boolean_t	ip_squeue_soft_ring;
3413 extern uint_t		ip_threads_per_cpu;
3414 extern uint_t		ip_squeues_per_cpu;
3415 extern uint_t		ip_soft_rings_cnt;
3416 
3417 typedef struct squeue_set_s {
3418 	kmutex_t	sqs_lock;
3419 	struct squeue_s	**sqs_list;
3420 	int		sqs_size;
3421 	int		sqs_max_size;
3422 	processorid_t	sqs_bind;
3423 } squeue_set_t;
3424 
3425 #define	IP_SQUEUE_GET(hint) 						\
3426 	((!ip_squeue_fanout) ?	(CPU->cpu_squeue_set->sqs_list[0]) :	\
3427 		ip_squeue_random(hint))
3428 
3429 typedef void (*squeue_func_t)(squeue_t *, mblk_t *, sqproc_t, void *, uint8_t);
3430 
3431 extern void ip_squeue_init(void (*)(squeue_t *));
3432 extern squeue_t	*ip_squeue_random(uint_t);
3433 extern squeue_t *ip_squeue_get(ill_rx_ring_t *);
3434 extern int ip_squeue_bind_set(queue_t *, mblk_t *, char *, caddr_t, cred_t *);
3435 extern void ip_squeue_clean_all(ill_t *);
3436 extern void ip_soft_ring_assignment(ill_t *, ill_rx_ring_t *,
3437     mblk_t *, struct mac_header_info_s *);
3438 
3439 extern void ip_resume_tcp_bind(void *, mblk_t *, void *);
3440 extern void tcp_wput(queue_t *, mblk_t *);
3441 
3442 extern int	ip_fill_mtuinfo(struct in6_addr *, in_port_t,
3443 	struct ip6_mtuinfo *, netstack_t *);
3444 extern	ipif_t *conn_get_held_ipif(conn_t *, ipif_t **, int *);
3445 
3446 typedef void    (*ipsq_func_t)(ipsq_t *, queue_t *, mblk_t *, void *);
3447 
3448 /*
3449  * Squeue tags. Tags only need to be unique when the callback function is the
3450  * same to distinguish between different calls, but we use unique tags for
3451  * convenience anyway.
3452  */
3453 #define	SQTAG_IP_INPUT			1
3454 #define	SQTAG_TCP_INPUT_ICMP_ERR	2
3455 #define	SQTAG_TCP6_INPUT_ICMP_ERR	3
3456 #define	SQTAG_IP_TCP_INPUT		4
3457 #define	SQTAG_IP6_TCP_INPUT		5
3458 #define	SQTAG_IP_TCP_CLOSE		6
3459 #define	SQTAG_TCP_OUTPUT		7
3460 #define	SQTAG_TCP_TIMER			8
3461 #define	SQTAG_TCP_TIMEWAIT		9
3462 #define	SQTAG_TCP_ACCEPT_FINISH		10
3463 #define	SQTAG_TCP_ACCEPT_FINISH_Q0	11
3464 #define	SQTAG_TCP_ACCEPT_PENDING	12
3465 #define	SQTAG_TCP_LISTEN_DISCON		13
3466 #define	SQTAG_TCP_CONN_REQ_1		14
3467 #define	SQTAG_TCP_EAGER_BLOWOFF		15
3468 #define	SQTAG_TCP_EAGER_CLEANUP		16
3469 #define	SQTAG_TCP_EAGER_CLEANUP_Q0	17
3470 #define	SQTAG_TCP_CONN_IND		18
3471 #define	SQTAG_TCP_RSRV			19
3472 #define	SQTAG_TCP_ABORT_BUCKET		20
3473 #define	SQTAG_TCP_REINPUT		21
3474 #define	SQTAG_TCP_REINPUT_EAGER		22
3475 #define	SQTAG_TCP_INPUT_MCTL		23
3476 #define	SQTAG_TCP_RPUTOTHER		24
3477 #define	SQTAG_IP_PROTO_AGAIN		25
3478 #define	SQTAG_IP_FANOUT_TCP		26
3479 #define	SQTAG_IPSQ_CLEAN_RING		27
3480 #define	SQTAG_TCP_WPUT_OTHER		28
3481 #define	SQTAG_TCP_CONN_REQ_UNBOUND	29
3482 #define	SQTAG_TCP_SEND_PENDING		30
3483 #define	SQTAG_BIND_RETRY		31
3484 #define	SQTAG_UDP_FANOUT		32
3485 #define	SQTAG_UDP_INPUT			33
3486 #define	SQTAG_UDP_WPUT			34
3487 #define	SQTAG_UDP_OUTPUT		35
3488 #define	SQTAG_TCP_KSSL_INPUT		36
3489 #define	SQTAG_TCP_DROP_Q0		37
3490 #define	SQTAG_TCP_CONN_REQ_2		38
3491 
3492 #define	NOT_OVER_IP(ip_wq)	\
3493 	(ip_wq->q_next != NULL ||	\
3494 	    (ip_wq->q_qinfo->qi_minfo->mi_idname) == NULL ||	\
3495 	    strcmp(ip_wq->q_qinfo->qi_minfo->mi_idname,	\
3496 	    IP_MOD_NAME) != 0 ||	\
3497 	    ip_wq->q_qinfo->qi_minfo->mi_idnum != IP_MOD_ID)
3498 
3499 #endif	/* _KERNEL */
3500 
3501 #ifdef	__cplusplus
3502 }
3503 #endif
3504 
3505 #endif	/* _INET_IP_H */
3506