xref: /illumos-gate/usr/src/uts/common/inet/ip.h (revision 7f125a531d922638f1af62a3d69f9ffc00bc2ee7)
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 (c) 1991, 2010, Oracle and/or its affiliates. All rights reserved.
24  * Copyright (c) 1990 Mentat Inc.
25  */
26 
27 #ifndef	_INET_IP_H
28 #define	_INET_IP_H
29 
30 #ifdef	__cplusplus
31 extern "C" {
32 #endif
33 
34 #include <sys/isa_defs.h>
35 #include <sys/types.h>
36 #include <inet/mib2.h>
37 #include <inet/nd.h>
38 #include <sys/atomic.h>
39 #include <net/if_dl.h>
40 #include <net/if.h>
41 #include <netinet/ip.h>
42 #include <netinet/igmp.h>
43 #include <sys/neti.h>
44 #include <sys/hook.h>
45 #include <sys/hook_event.h>
46 #include <sys/hook_impl.h>
47 #include <inet/ip_stack.h>
48 
49 #ifdef _KERNEL
50 #include <netinet/ip6.h>
51 #include <sys/avl.h>
52 #include <sys/list.h>
53 #include <sys/vmem.h>
54 #include <sys/squeue.h>
55 #include <net/route.h>
56 #include <sys/systm.h>
57 #include <net/radix.h>
58 #include <sys/modhash.h>
59 
60 #ifdef DEBUG
61 #define	CONN_DEBUG
62 #endif
63 
64 #define	IP_DEBUG
65 /*
66  * The mt-streams(9F) flags for the IP module; put here so that other
67  * "drivers" that are actually IP (e.g., ICMP, UDP) can use the same set
68  * of flags.
69  */
70 #define	IP_DEVMTFLAGS D_MP
71 #endif	/* _KERNEL */
72 
73 #define	IP_MOD_NAME	"ip"
74 #define	IP_DEV_NAME	"/dev/ip"
75 #define	IP6_DEV_NAME	"/dev/ip6"
76 
77 #define	UDP_MOD_NAME	"udp"
78 #define	UDP_DEV_NAME	"/dev/udp"
79 #define	UDP6_DEV_NAME	"/dev/udp6"
80 
81 #define	TCP_MOD_NAME	"tcp"
82 #define	TCP_DEV_NAME	"/dev/tcp"
83 #define	TCP6_DEV_NAME	"/dev/tcp6"
84 
85 #define	SCTP_MOD_NAME	"sctp"
86 
87 #ifndef	_IPADDR_T
88 #define	_IPADDR_T
89 typedef uint32_t ipaddr_t;
90 #endif
91 
92 /* Number of bits in an address */
93 #define	IP_ABITS		32
94 #define	IPV4_ABITS		IP_ABITS
95 #define	IPV6_ABITS		128
96 #define	IP_MAX_HW_LEN	40
97 
98 #define	IP_HOST_MASK		(ipaddr_t)0xffffffffU
99 
100 #define	IP_CSUM(mp, off, sum)		(~ip_cksum(mp, off, sum) & 0xFFFF)
101 #define	IP_CSUM_PARTIAL(mp, off, sum)	ip_cksum(mp, off, sum)
102 #define	IP_BCSUM_PARTIAL(bp, len, sum)	bcksum(bp, len, sum)
103 
104 #define	ILL_FRAG_HASH_TBL_COUNT	((unsigned int)64)
105 #define	ILL_FRAG_HASH_TBL_SIZE	(ILL_FRAG_HASH_TBL_COUNT * sizeof (ipfb_t))
106 
107 #define	IPV4_ADDR_LEN			4
108 #define	IP_ADDR_LEN			IPV4_ADDR_LEN
109 #define	IP_ARP_PROTO_TYPE		0x0800
110 
111 #define	IPV4_VERSION			4
112 #define	IP_VERSION			IPV4_VERSION
113 #define	IP_SIMPLE_HDR_LENGTH_IN_WORDS	5
114 #define	IP_SIMPLE_HDR_LENGTH		20
115 #define	IP_MAX_HDR_LENGTH		60
116 
117 #define	IP_MAX_OPT_LENGTH (IP_MAX_HDR_LENGTH-IP_SIMPLE_HDR_LENGTH)
118 
119 #define	IP_MIN_MTU			(IP_MAX_HDR_LENGTH + 8)	/* 68 bytes */
120 
121 /*
122  * XXX IP_MAXPACKET is defined in <netinet/ip.h> as well. At some point the
123  * 2 files should be cleaned up to remove all redundant definitions.
124  */
125 #define	IP_MAXPACKET			65535
126 #define	IP_SIMPLE_HDR_VERSION \
127 	((IP_VERSION << 4) | IP_SIMPLE_HDR_LENGTH_IN_WORDS)
128 
129 #define	UDPH_SIZE			8
130 
131 /*
132  * Constants and type definitions to support IP IOCTL commands
133  */
134 #define	IP_IOCTL			(('i'<<8)|'p')
135 #define	IP_IOC_IRE_DELETE		4
136 #define	IP_IOC_IRE_DELETE_NO_REPLY	5
137 #define	IP_IOC_RTS_REQUEST		7
138 
139 /* Common definitions used by IP IOCTL data structures */
140 typedef struct ipllcmd_s {
141 	uint_t	ipllc_cmd;
142 	uint_t	ipllc_name_offset;
143 	uint_t	ipllc_name_length;
144 } ipllc_t;
145 
146 /* IP IRE Delete Command Structure. */
147 typedef struct ipid_s {
148 	ipllc_t	ipid_ipllc;
149 	uint_t	ipid_ire_type;
150 	uint_t	ipid_addr_offset;
151 	uint_t	ipid_addr_length;
152 	uint_t	ipid_mask_offset;
153 	uint_t	ipid_mask_length;
154 } ipid_t;
155 
156 #define	ipid_cmd		ipid_ipllc.ipllc_cmd
157 
158 #ifdef _KERNEL
159 /*
160  * Temporary state for ip options parser.
161  */
162 typedef struct ipoptp_s
163 {
164 	uint8_t		*ipoptp_next;	/* next option to look at */
165 	uint8_t		*ipoptp_end;	/* end of options */
166 	uint8_t		*ipoptp_cur;	/* start of current option */
167 	uint8_t		ipoptp_len;	/* length of current option */
168 	uint32_t	ipoptp_flags;
169 } ipoptp_t;
170 
171 /*
172  * Flag(s) for ipoptp_flags
173  */
174 #define	IPOPTP_ERROR	0x00000001
175 #endif	/* _KERNEL */
176 
177 /* Controls forwarding of IP packets, set via ipadm(1M)/ndd(1M) */
178 #define	IP_FORWARD_NEVER	0
179 #define	IP_FORWARD_ALWAYS	1
180 
181 #define	WE_ARE_FORWARDING(ipst)	((ipst)->ips_ip_forwarding == IP_FORWARD_ALWAYS)
182 
183 #define	IPH_HDR_LENGTH(ipha)						\
184 	((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length & 0xF) << 2)
185 
186 #define	IPH_HDR_VERSION(ipha)						\
187 	((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length) >> 4)
188 
189 #ifdef _KERNEL
190 /*
191  * IP reassembly macros.  We hide starting and ending offsets in b_next and
192  * b_prev of messages on the reassembly queue.	The messages are chained using
193  * b_cont.  These macros are used in ip_reassemble() so we don't have to see
194  * the ugly casts and assignments.
195  * Note that the offsets are <= 64k i.e. a uint_t is sufficient to represent
196  * them.
197  */
198 #define	IP_REASS_START(mp)		((uint_t)(uintptr_t)((mp)->b_next))
199 #define	IP_REASS_SET_START(mp, u)	\
200 	((mp)->b_next = (mblk_t *)(uintptr_t)(u))
201 #define	IP_REASS_END(mp)		((uint_t)(uintptr_t)((mp)->b_prev))
202 #define	IP_REASS_SET_END(mp, u)		\
203 	((mp)->b_prev = (mblk_t *)(uintptr_t)(u))
204 
205 #define	IP_REASS_COMPLETE	0x1
206 #define	IP_REASS_PARTIAL	0x2
207 #define	IP_REASS_FAILED		0x4
208 
209 /*
210  * Test to determine whether this is a module instance of IP or a
211  * driver instance of IP.
212  */
213 #define	CONN_Q(q)	(WR(q)->q_next == NULL)
214 
215 #define	Q_TO_CONN(q)	((conn_t *)(q)->q_ptr)
216 #define	Q_TO_TCP(q)	(Q_TO_CONN((q))->conn_tcp)
217 #define	Q_TO_UDP(q)	(Q_TO_CONN((q))->conn_udp)
218 #define	Q_TO_ICMP(q)	(Q_TO_CONN((q))->conn_icmp)
219 #define	Q_TO_RTS(q)	(Q_TO_CONN((q))->conn_rts)
220 
221 #define	CONNP_TO_WQ(connp)	((connp)->conn_wq)
222 #define	CONNP_TO_RQ(connp)	((connp)->conn_rq)
223 
224 #define	GRAB_CONN_LOCK(q)	{				\
225 	if (q != NULL && CONN_Q(q))				\
226 		mutex_enter(&(Q_TO_CONN(q))->conn_lock);	\
227 }
228 
229 #define	RELEASE_CONN_LOCK(q)	{				\
230 	if (q != NULL && CONN_Q(q))				\
231 		mutex_exit(&(Q_TO_CONN(q))->conn_lock);		\
232 }
233 
234 /*
235  * Ref counter macros for ioctls. This provides a guard for TCP to stop
236  * tcp_close from removing the rq/wq whilst an ioctl is still in flight on the
237  * stream. The ioctl could have been queued on e.g. an ipsq. tcp_close will wait
238  * until the ioctlref count is zero before proceeding.
239  * Ideally conn_oper_pending_ill would be used for this purpose. However, in the
240  * case where an ioctl is aborted or interrupted, it can be cleared prematurely.
241  * There are also some race possibilities between ip and the stream head which
242  * can also end up with conn_oper_pending_ill being cleared prematurely. So, to
243  * avoid these situations, we use a dedicated ref counter for ioctls which is
244  * used in addition to and in parallel with the normal conn_ref count.
245  */
246 #define	CONN_INC_IOCTLREF_LOCKED(connp)	{			\
247 	ASSERT(MUTEX_HELD(&(connp)->conn_lock));		\
248 	DTRACE_PROBE1(conn__inc__ioctlref, conn_t *, (connp));	\
249 	(connp)->conn_ioctlref++;				\
250 	mutex_exit(&(connp)->conn_lock);			\
251 }
252 
253 #define	CONN_INC_IOCTLREF(connp)	{			\
254 	mutex_enter(&(connp)->conn_lock);			\
255 	CONN_INC_IOCTLREF_LOCKED(connp);			\
256 }
257 
258 #define	CONN_DEC_IOCTLREF(connp)	{			\
259 	mutex_enter(&(connp)->conn_lock);			\
260 	DTRACE_PROBE1(conn__dec__ioctlref, conn_t *, (connp));	\
261 	/* Make sure conn_ioctlref will not underflow. */	\
262 	ASSERT((connp)->conn_ioctlref != 0);			\
263 	if ((--(connp)->conn_ioctlref == 0) &&			\
264 	    ((connp)->conn_state_flags & CONN_CLOSING)) {	\
265 		cv_broadcast(&(connp)->conn_cv);		\
266 	}							\
267 	mutex_exit(&(connp)->conn_lock);			\
268 }
269 
270 
271 /*
272  * Complete the pending operation. Usually an ioctl. Can also
273  * be a bind or option management request that got enqueued
274  * in an ipsq_t. Called on completion of the operation.
275  */
276 #define	CONN_OPER_PENDING_DONE(connp)	{			\
277 	mutex_enter(&(connp)->conn_lock);			\
278 	(connp)->conn_oper_pending_ill = NULL;			\
279 	cv_broadcast(&(connp)->conn_refcv);			\
280 	mutex_exit(&(connp)->conn_lock);			\
281 	CONN_DEC_REF(connp);					\
282 }
283 
284 /*
285  * Values for squeue switch:
286  */
287 #define	IP_SQUEUE_ENTER_NODRAIN	1
288 #define	IP_SQUEUE_ENTER	2
289 #define	IP_SQUEUE_FILL 3
290 
291 extern int ip_squeue_flag;
292 
293 /* IP Fragmentation Reassembly Header */
294 typedef struct ipf_s {
295 	struct ipf_s	*ipf_hash_next;
296 	struct ipf_s	**ipf_ptphn;	/* Pointer to previous hash next. */
297 	uint32_t	ipf_ident;	/* Ident to match. */
298 	uint8_t		ipf_protocol;	/* Protocol to match. */
299 	uchar_t		ipf_last_frag_seen : 1;	/* Last fragment seen ? */
300 	time_t		ipf_timestamp;	/* Reassembly start time. */
301 	mblk_t		*ipf_mp;	/* mblk we live in. */
302 	mblk_t		*ipf_tail_mp;	/* Frag queue tail pointer. */
303 	int		ipf_hole_cnt;	/* Number of holes (hard-case). */
304 	int		ipf_end;	/* Tail end offset (0 -> hard-case). */
305 	uint_t		ipf_gen;	/* Frag queue generation */
306 	size_t		ipf_count;	/* Count of bytes used by frag */
307 	uint_t		ipf_nf_hdr_len; /* Length of nonfragmented header */
308 	in6_addr_t	ipf_v6src;	/* IPv6 source address */
309 	in6_addr_t	ipf_v6dst;	/* IPv6 dest address */
310 	uint_t		ipf_prev_nexthdr_offset; /* Offset for nexthdr value */
311 	uint8_t		ipf_ecn;	/* ECN info for the fragments */
312 	uint8_t		ipf_num_dups;	/* Number of times dup frags recvd */
313 	uint16_t	ipf_checksum_flags; /* Hardware checksum flags */
314 	uint32_t	ipf_checksum;	/* Partial checksum of fragment data */
315 } ipf_t;
316 
317 /*
318  * IPv4 Fragments
319  */
320 #define	IS_V4_FRAGMENT(ipha_fragment_offset_and_flags)			\
321 	(((ntohs(ipha_fragment_offset_and_flags) & IPH_OFFSET) != 0) ||	\
322 	((ntohs(ipha_fragment_offset_and_flags) & IPH_MF) != 0))
323 
324 #define	ipf_src	V4_PART_OF_V6(ipf_v6src)
325 #define	ipf_dst	V4_PART_OF_V6(ipf_v6dst)
326 
327 #endif /* _KERNEL */
328 
329 /* ICMP types */
330 #define	ICMP_ECHO_REPLY			0
331 #define	ICMP_DEST_UNREACHABLE		3
332 #define	ICMP_SOURCE_QUENCH		4
333 #define	ICMP_REDIRECT			5
334 #define	ICMP_ECHO_REQUEST		8
335 #define	ICMP_ROUTER_ADVERTISEMENT	9
336 #define	ICMP_ROUTER_SOLICITATION	10
337 #define	ICMP_TIME_EXCEEDED		11
338 #define	ICMP_PARAM_PROBLEM		12
339 #define	ICMP_TIME_STAMP_REQUEST		13
340 #define	ICMP_TIME_STAMP_REPLY		14
341 #define	ICMP_INFO_REQUEST		15
342 #define	ICMP_INFO_REPLY			16
343 #define	ICMP_ADDRESS_MASK_REQUEST	17
344 #define	ICMP_ADDRESS_MASK_REPLY		18
345 
346 /* Evaluates to true if the ICMP type is an ICMP error */
347 #define	ICMP_IS_ERROR(type)	(		\
348 	(type) == ICMP_DEST_UNREACHABLE ||	\
349 	(type) == ICMP_SOURCE_QUENCH ||		\
350 	(type) == ICMP_TIME_EXCEEDED ||		\
351 	(type) == ICMP_PARAM_PROBLEM)
352 
353 /* ICMP_TIME_EXCEEDED codes */
354 #define	ICMP_TTL_EXCEEDED		0
355 #define	ICMP_REASSEMBLY_TIME_EXCEEDED	1
356 
357 /* ICMP_DEST_UNREACHABLE codes */
358 #define	ICMP_NET_UNREACHABLE		0
359 #define	ICMP_HOST_UNREACHABLE		1
360 #define	ICMP_PROTOCOL_UNREACHABLE	2
361 #define	ICMP_PORT_UNREACHABLE		3
362 #define	ICMP_FRAGMENTATION_NEEDED	4
363 #define	ICMP_SOURCE_ROUTE_FAILED	5
364 #define	ICMP_DEST_NET_UNKNOWN		6
365 #define	ICMP_DEST_HOST_UNKNOWN		7
366 #define	ICMP_SRC_HOST_ISOLATED		8
367 #define	ICMP_DEST_NET_UNREACH_ADMIN	9
368 #define	ICMP_DEST_HOST_UNREACH_ADMIN	10
369 #define	ICMP_DEST_NET_UNREACH_TOS	11
370 #define	ICMP_DEST_HOST_UNREACH_TOS	12
371 
372 /* ICMP Header Structure */
373 typedef struct icmph_s {
374 	uint8_t		icmph_type;
375 	uint8_t		icmph_code;
376 	uint16_t	icmph_checksum;
377 	union {
378 		struct { /* ECHO request/response structure */
379 			uint16_t	u_echo_ident;
380 			uint16_t	u_echo_seqnum;
381 		} u_echo;
382 		struct { /* Destination unreachable structure */
383 			uint16_t	u_du_zero;
384 			uint16_t	u_du_mtu;
385 		} u_du;
386 		struct { /* Parameter problem structure */
387 			uint8_t		u_pp_ptr;
388 			uint8_t		u_pp_rsvd[3];
389 		} u_pp;
390 		struct { /* Redirect structure */
391 			ipaddr_t	u_rd_gateway;
392 		} u_rd;
393 	} icmph_u;
394 } icmph_t;
395 
396 #define	icmph_echo_ident	icmph_u.u_echo.u_echo_ident
397 #define	icmph_echo_seqnum	icmph_u.u_echo.u_echo_seqnum
398 #define	icmph_du_zero		icmph_u.u_du.u_du_zero
399 #define	icmph_du_mtu		icmph_u.u_du.u_du_mtu
400 #define	icmph_pp_ptr		icmph_u.u_pp.u_pp_ptr
401 #define	icmph_rd_gateway	icmph_u.u_rd.u_rd_gateway
402 
403 #define	ICMPH_SIZE	8
404 
405 /*
406  * Minimum length of transport layer header included in an ICMP error
407  * message for it to be considered valid.
408  */
409 #define	ICMP_MIN_TP_HDR_LEN	8
410 
411 /* Aligned IP header */
412 typedef struct ipha_s {
413 	uint8_t		ipha_version_and_hdr_length;
414 	uint8_t		ipha_type_of_service;
415 	uint16_t	ipha_length;
416 	uint16_t	ipha_ident;
417 	uint16_t	ipha_fragment_offset_and_flags;
418 	uint8_t		ipha_ttl;
419 	uint8_t		ipha_protocol;
420 	uint16_t	ipha_hdr_checksum;
421 	ipaddr_t	ipha_src;
422 	ipaddr_t	ipha_dst;
423 } ipha_t;
424 
425 /*
426  * IP Flags
427  *
428  * Some of these constant names are copied for the DTrace IP provider in
429  * usr/src/lib/libdtrace/common/{ip.d.in, ip.sed.in}, which should be kept
430  * in sync.
431  */
432 #define	IPH_DF		0x4000	/* Don't fragment */
433 #define	IPH_MF		0x2000	/* More fragments to come */
434 #define	IPH_OFFSET	0x1FFF	/* Where the offset lives */
435 
436 /* Byte-order specific values */
437 #ifdef	_BIG_ENDIAN
438 #define	IPH_DF_HTONS	0x4000	/* Don't fragment */
439 #define	IPH_MF_HTONS	0x2000	/* More fragments to come */
440 #define	IPH_OFFSET_HTONS 0x1FFF	/* Where the offset lives */
441 #else
442 #define	IPH_DF_HTONS	0x0040	/* Don't fragment */
443 #define	IPH_MF_HTONS	0x0020	/* More fragments to come */
444 #define	IPH_OFFSET_HTONS 0xFF1F	/* Where the offset lives */
445 #endif
446 
447 /* ECN code points for IPv4 TOS byte and IPv6 traffic class octet. */
448 #define	IPH_ECN_NECT	0x0	/* Not ECN-Capable Transport */
449 #define	IPH_ECN_ECT1	0x1	/* ECN-Capable Transport, ECT(1) */
450 #define	IPH_ECN_ECT0	0x2	/* ECN-Capable Transport, ECT(0) */
451 #define	IPH_ECN_CE	0x3	/* ECN-Congestion Experienced (CE) */
452 
453 struct ill_s;
454 
455 typedef	void ip_v6intfid_func_t(struct ill_s *, in6_addr_t *);
456 typedef void ip_v6mapinfo_func_t(struct ill_s *, uchar_t *, uchar_t *);
457 typedef void ip_v4mapinfo_func_t(struct ill_s *, uchar_t *, uchar_t *);
458 
459 /* IP Mac info structure */
460 typedef struct ip_m_s {
461 	t_uscalar_t		ip_m_mac_type;	/* From <sys/dlpi.h> */
462 	int			ip_m_type;	/* From <net/if_types.h> */
463 	t_uscalar_t		ip_m_ipv4sap;
464 	t_uscalar_t		ip_m_ipv6sap;
465 	ip_v4mapinfo_func_t	*ip_m_v4mapping;
466 	ip_v6mapinfo_func_t	*ip_m_v6mapping;
467 	ip_v6intfid_func_t	*ip_m_v6intfid;
468 	ip_v6intfid_func_t	*ip_m_v6destintfid;
469 } ip_m_t;
470 
471 /*
472  * The following functions attempt to reduce the link layer dependency
473  * of the IP stack. The current set of link specific operations are:
474  * a. map from IPv4 class D (224.0/4) multicast address range or the
475  * IPv6 multicast address range (ff00::/8) to the link layer multicast
476  * address.
477  * b. derive the default IPv6 interface identifier from the interface.
478  * c. derive the default IPv6 destination interface identifier from
479  * the interface (point-to-point only).
480  */
481 extern	void ip_mcast_mapping(struct ill_s *, uchar_t *, uchar_t *);
482 /* ip_m_v6*intfid return void and are never NULL */
483 #define	MEDIA_V6INTFID(ip_m, ill, v6ptr) (ip_m)->ip_m_v6intfid(ill, v6ptr)
484 #define	MEDIA_V6DESTINTFID(ip_m, ill, v6ptr) \
485 	(ip_m)->ip_m_v6destintfid(ill, v6ptr)
486 
487 /* Router entry types */
488 #define	IRE_BROADCAST		0x0001	/* Route entry for broadcast address */
489 #define	IRE_DEFAULT		0x0002	/* Route entry for default gateway */
490 #define	IRE_LOCAL		0x0004	/* Route entry for local address */
491 #define	IRE_LOOPBACK		0x0008	/* Route entry for loopback address */
492 #define	IRE_PREFIX		0x0010	/* Route entry for prefix routes */
493 #ifndef _KERNEL
494 /* Keep so user-level still compiles */
495 #define	IRE_CACHE		0x0020	/* Cached Route entry */
496 #endif
497 #define	IRE_IF_NORESOLVER	0x0040	/* Route entry for local interface */
498 					/* net without any address mapping. */
499 #define	IRE_IF_RESOLVER		0x0080	/* Route entry for local interface */
500 					/* net with resolver. */
501 #define	IRE_HOST		0x0100	/* Host route entry */
502 /* Keep so user-level still compiles */
503 #define	IRE_HOST_REDIRECT	0x0200	/* only used for T_SVR4_OPTMGMT_REQ */
504 #define	IRE_IF_CLONE		0x0400	/* Per host clone of IRE_IF */
505 #define	IRE_MULTICAST		0x0800	/* Special - not in table */
506 #define	IRE_NOROUTE		0x1000	/* Special - not in table */
507 
508 #define	IRE_INTERFACE		(IRE_IF_NORESOLVER | IRE_IF_RESOLVER)
509 
510 #define	IRE_IF_ALL		(IRE_IF_NORESOLVER | IRE_IF_RESOLVER | \
511 				    IRE_IF_CLONE)
512 #define	IRE_OFFSUBNET		(IRE_DEFAULT | IRE_PREFIX | IRE_HOST)
513 #define	IRE_OFFLINK		IRE_OFFSUBNET
514 /*
515  * Note that we view IRE_NOROUTE as ONLINK since we can "send" to them without
516  * going through a router; the result of sending will be an error/icmp error.
517  */
518 #define	IRE_ONLINK		(IRE_IF_ALL|IRE_LOCAL|IRE_LOOPBACK| \
519 				    IRE_BROADCAST|IRE_MULTICAST|IRE_NOROUTE)
520 
521 /* Arguments to ire_flush_cache() */
522 #define	IRE_FLUSH_DELETE	0
523 #define	IRE_FLUSH_ADD		1
524 #define	IRE_FLUSH_GWCHANGE	2
525 
526 /*
527  * Flags to ire_route_recursive
528  */
529 #define	IRR_NONE		0
530 #define	IRR_ALLOCATE		1	/* OK to allocate IRE_IF_CLONE */
531 #define	IRR_INCOMPLETE		2	/* OK to return incomplete chain */
532 
533 /*
534  * Open/close synchronization flags.
535  * These are kept in a separate field in the conn and the synchronization
536  * depends on the atomic 32 bit access to that field.
537  */
538 #define	CONN_CLOSING		0x01	/* ip_close waiting for ip_wsrv */
539 #define	CONN_CONDEMNED		0x02	/* conn is closing, no more refs */
540 #define	CONN_INCIPIENT		0x04	/* conn not yet visible, no refs */
541 #define	CONN_QUIESCED		0x08	/* conn is now quiescent */
542 #define	CONN_UPDATE_ILL		0x10	/* conn_update_ill in progress */
543 
544 /*
545  * Flags for dce_flags field. Specifies which information has been set.
546  * dce_ident is always present, but the other ones are identified by the flags.
547  */
548 #define	DCEF_DEFAULT		0x0001	/* Default DCE - no pmtu or uinfo */
549 #define	DCEF_PMTU		0x0002	/* Different than interface MTU */
550 #define	DCEF_UINFO		0x0004	/* dce_uinfo set */
551 #define	DCEF_TOO_SMALL_PMTU	0x0008	/* Smaller than IPv4/IPv6 MIN */
552 
553 #ifdef _KERNEL
554 /*
555  * Extra structures need for per-src-addr filtering (IGMPv3/MLDv2)
556  */
557 #define	MAX_FILTER_SIZE	64
558 
559 typedef struct slist_s {
560 	int		sl_numsrc;
561 	in6_addr_t	sl_addr[MAX_FILTER_SIZE];
562 } slist_t;
563 
564 /*
565  * Following struct is used to maintain retransmission state for
566  * a multicast group.  One rtx_state_t struct is an in-line field
567  * of the ilm_t struct; the slist_ts in the rtx_state_t struct are
568  * alloc'd as needed.
569  */
570 typedef struct rtx_state_s {
571 	uint_t		rtx_timer;	/* retrans timer */
572 	int		rtx_cnt;	/* retrans count */
573 	int		rtx_fmode_cnt;	/* retrans count for fmode change */
574 	slist_t		*rtx_allow;
575 	slist_t		*rtx_block;
576 } rtx_state_t;
577 
578 /*
579  * Used to construct list of multicast address records that will be
580  * sent in a single listener report.
581  */
582 typedef struct mrec_s {
583 	struct mrec_s	*mrec_next;
584 	uint8_t		mrec_type;
585 	uint8_t		mrec_auxlen;	/* currently unused */
586 	in6_addr_t	mrec_group;
587 	slist_t		mrec_srcs;
588 } mrec_t;
589 
590 /* Group membership list per upper conn */
591 
592 /*
593  * We record the multicast information from the socket option in
594  * ilg_ifaddr/ilg_ifindex. This allows rejoining the group in the case when
595  * the ifaddr (or ifindex) disappears and later reappears, potentially on
596  * a different ill. The IPv6 multicast socket options and ioctls all specify
597  * the interface using an ifindex. For IPv4 some socket options/ioctls use
598  * the interface address and others use the index. We record here the method
599  * that was actually used (and leave the other of ilg_ifaddr or ilg_ifindex)
600  * at zero so that we can rejoin the way the application intended.
601  *
602  * We track the ill on which we will or already have joined an ilm using
603  * ilg_ill. When we have succeeded joining the ilm and have a refhold on it
604  * then we set ilg_ilm. Thus intentionally there is a window where ilg_ill is
605  * set and ilg_ilm is not set. This allows clearing ilg_ill as a signal that
606  * the ill is being unplumbed and the ilm should be discarded.
607  *
608  * ilg records the state of multicast memberships of a socket end point.
609  * ilm records the state of multicast memberships with the driver and is
610  * maintained per interface.
611  *
612  * The ilg state is protected by conn_ilg_lock.
613  * The ilg will not be freed until ilg_refcnt drops to zero.
614  */
615 typedef struct ilg_s {
616 	struct ilg_s	*ilg_next;
617 	struct ilg_s	**ilg_ptpn;
618 	struct conn_s	*ilg_connp;	/* Back pointer to get lock */
619 	in6_addr_t	ilg_v6group;
620 	ipaddr_t	ilg_ifaddr;	/* For some IPv4 cases */
621 	uint_t		ilg_ifindex;	/* IPv6 and some other IPv4 cases */
622 	struct ill_s	*ilg_ill;	/* Where ilm is joined. No refhold */
623 	struct ilm_s	*ilg_ilm;	/* With ilm_refhold */
624 	uint_t		ilg_refcnt;
625 	mcast_record_t	ilg_fmode;	/* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */
626 	slist_t		*ilg_filter;
627 	boolean_t	ilg_condemned;	/* Conceptually deleted */
628 } ilg_t;
629 
630 /*
631  * Multicast address list entry for ill.
632  * ilm_ill is used by IPv4 and IPv6
633  *
634  * The ilm state (and other multicast state on the ill) is protected by
635  * ill_mcast_lock. Operations that change state on both an ilg and ilm
636  * in addition use ill_mcast_serializer to ensure that we can't have
637  * interleaving between e.g., add and delete operations for the same conn_t,
638  * group, and ill. The ill_mcast_serializer is also used to ensure that
639  * multicast group joins do not occur on an interface that is in the process
640  * of joining an IPMP group.
641  *
642  * The comment below (and for other netstack_t references) refers
643  * to the fact that we only do netstack_hold in particular cases,
644  * such as the references from open endpoints (ill_t and conn_t's
645  * pointers). Internally within IP we rely on IP's ability to cleanup e.g.
646  * ire_t's when an ill goes away.
647  */
648 typedef struct ilm_s {
649 	in6_addr_t	ilm_v6addr;
650 	int		ilm_refcnt;
651 	uint_t		ilm_timer;	/* IGMP/MLD query resp timer, in msec */
652 	struct ilm_s	*ilm_next;	/* Linked list for each ill */
653 	uint_t		ilm_state;	/* state of the membership */
654 	struct ill_s	*ilm_ill;	/* Back pointer to ill - ill_ilm_cnt */
655 	zoneid_t	ilm_zoneid;
656 	int		ilm_no_ilg_cnt;	/* number of joins w/ no ilg */
657 	mcast_record_t	ilm_fmode;	/* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */
658 	slist_t		*ilm_filter;	/* source filter list */
659 	slist_t		*ilm_pendsrcs;	/* relevant src addrs for pending req */
660 	rtx_state_t	ilm_rtx;	/* SCR retransmission state */
661 	ipaddr_t	ilm_ifaddr;	/* For IPv4 netstat */
662 	ip_stack_t	*ilm_ipst;	/* Does not have a netstack_hold */
663 } ilm_t;
664 
665 #define	ilm_addr	V4_PART_OF_V6(ilm_v6addr)
666 
667 /*
668  * Soft reference to an IPsec SA.
669  *
670  * On relative terms, conn's can be persistent (living as long as the
671  * processes which create them), while SA's are ephemeral (dying when
672  * they hit their time-based or byte-based lifetimes).
673  *
674  * We could hold a hard reference to an SA from an ipsec_latch_t,
675  * but this would cause expired SA's to linger for a potentially
676  * unbounded time.
677  *
678  * Instead, we remember the hash bucket number and bucket generation
679  * in addition to the pointer.  The bucket generation is incremented on
680  * each deletion.
681  */
682 typedef struct ipsa_ref_s
683 {
684 	struct ipsa_s	*ipsr_sa;
685 	struct isaf_s	*ipsr_bucket;
686 	uint64_t	ipsr_gen;
687 } ipsa_ref_t;
688 
689 /*
690  * IPsec "latching" state.
691  *
692  * In the presence of IPsec policy, fully-bound conn's bind a connection
693  * to more than just the 5-tuple, but also a specific IPsec action and
694  * identity-pair.
695  * The identity pair is accessed from both the receive and transmit side
696  * hence it is maintained in the ipsec_latch_t structure. conn_latch and
697  * ixa_ipsec_latch points to it.
698  * The policy and actions are stored in conn_latch_in_policy and
699  * conn_latch_in_action for the inbound side, and in ixa_ipsec_policy and
700  * ixa_ipsec_action for the transmit side.
701  *
702  * As an optimization, we also cache soft references to IPsec SA's in
703  * ip_xmit_attr_t so that we can fast-path around most of the work needed for
704  * outbound IPsec SA selection.
705  */
706 typedef struct ipsec_latch_s
707 {
708 	kmutex_t	ipl_lock;
709 	uint32_t	ipl_refcnt;
710 
711 	struct ipsid_s	*ipl_local_cid;
712 	struct ipsid_s	*ipl_remote_cid;
713 	unsigned int
714 			ipl_ids_latched : 1,
715 
716 			ipl_pad_to_bit_31 : 31;
717 } ipsec_latch_t;
718 
719 #define	IPLATCH_REFHOLD(ipl) { \
720 	atomic_add_32(&(ipl)->ipl_refcnt, 1);		\
721 	ASSERT((ipl)->ipl_refcnt != 0);			\
722 }
723 
724 #define	IPLATCH_REFRELE(ipl) {				\
725 	ASSERT((ipl)->ipl_refcnt != 0);				\
726 	membar_exit();						\
727 	if (atomic_add_32_nv(&(ipl)->ipl_refcnt, -1) == 0)	\
728 		iplatch_free(ipl);				\
729 }
730 
731 /*
732  * peer identity structure.
733  */
734 typedef struct conn_s conn_t;
735 
736 /*
737  * This is used to match an inbound/outbound datagram with policy.
738  */
739 typedef	struct ipsec_selector {
740 	in6_addr_t	ips_local_addr_v6;
741 	in6_addr_t	ips_remote_addr_v6;
742 	uint16_t	ips_local_port;
743 	uint16_t	ips_remote_port;
744 	uint8_t		ips_icmp_type;
745 	uint8_t		ips_icmp_code;
746 	uint8_t		ips_protocol;
747 	uint8_t		ips_isv4 : 1,
748 			ips_is_icmp_inv_acq: 1;
749 } ipsec_selector_t;
750 
751 /*
752  * Note that we put v4 addresses in the *first* 32-bit word of the
753  * selector rather than the last to simplify the prefix match/mask code
754  * in spd.c
755  */
756 #define	ips_local_addr_v4 ips_local_addr_v6.s6_addr32[0]
757 #define	ips_remote_addr_v4 ips_remote_addr_v6.s6_addr32[0]
758 
759 /* Values used in IP by IPSEC Code */
760 #define		IPSEC_OUTBOUND		B_TRUE
761 #define		IPSEC_INBOUND		B_FALSE
762 
763 /*
764  * There are two variants in policy failures. The packet may come in
765  * secure when not needed (IPSEC_POLICY_???_NOT_NEEDED) or it may not
766  * have the desired level of protection (IPSEC_POLICY_MISMATCH).
767  */
768 #define	IPSEC_POLICY_NOT_NEEDED		0
769 #define	IPSEC_POLICY_MISMATCH		1
770 #define	IPSEC_POLICY_AUTH_NOT_NEEDED	2
771 #define	IPSEC_POLICY_ENCR_NOT_NEEDED	3
772 #define	IPSEC_POLICY_SE_NOT_NEEDED	4
773 #define	IPSEC_POLICY_MAX		5	/* Always max + 1. */
774 
775 /*
776  * Check with IPSEC inbound policy if
777  *
778  * 1) per-socket policy is present - indicated by conn_in_enforce_policy.
779  * 2) Or if we have not cached policy on the conn and the global policy is
780  *    non-empty.
781  */
782 #define	CONN_INBOUND_POLICY_PRESENT(connp, ipss)	\
783 	((connp)->conn_in_enforce_policy ||		\
784 	(!((connp)->conn_policy_cached) && 		\
785 	(ipss)->ipsec_inbound_v4_policy_present))
786 
787 #define	CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)	\
788 	((connp)->conn_in_enforce_policy ||		\
789 	(!(connp)->conn_policy_cached &&		\
790 	(ipss)->ipsec_inbound_v6_policy_present))
791 
792 #define	CONN_OUTBOUND_POLICY_PRESENT(connp, ipss)	\
793 	((connp)->conn_out_enforce_policy ||		\
794 	(!((connp)->conn_policy_cached) &&		\
795 	(ipss)->ipsec_outbound_v4_policy_present))
796 
797 #define	CONN_OUTBOUND_POLICY_PRESENT_V6(connp, ipss)	\
798 	((connp)->conn_out_enforce_policy ||		\
799 	(!(connp)->conn_policy_cached &&		\
800 	(ipss)->ipsec_outbound_v6_policy_present))
801 
802 /*
803  * Information cached in IRE for upper layer protocol (ULP).
804  */
805 typedef struct iulp_s {
806 	boolean_t	iulp_set;	/* Is any metric set? */
807 	uint32_t	iulp_ssthresh;	/* Slow start threshold (TCP). */
808 	clock_t		iulp_rtt;	/* Guestimate in millisecs. */
809 	clock_t		iulp_rtt_sd;	/* Cached value of RTT variance. */
810 	uint32_t	iulp_spipe;	/* Send pipe size. */
811 	uint32_t	iulp_rpipe;	/* Receive pipe size. */
812 	uint32_t	iulp_rtomax;	/* Max round trip timeout. */
813 	uint32_t	iulp_sack;	/* Use SACK option (TCP)? */
814 	uint32_t	iulp_mtu;	/* Setable with routing sockets */
815 
816 	uint32_t
817 		iulp_tstamp_ok : 1,	/* Use timestamp option (TCP)? */
818 		iulp_wscale_ok : 1,	/* Use window scale option (TCP)? */
819 		iulp_ecn_ok : 1,	/* Enable ECN (for TCP)? */
820 		iulp_pmtud_ok : 1,	/* Enable PMTUd? */
821 
822 		/* These three are passed out by ip_set_destination */
823 		iulp_localnet: 1,	/* IRE_ONLINK */
824 		iulp_loopback: 1,	/* IRE_LOOPBACK */
825 		iulp_local: 1,		/* IRE_LOCAL */
826 
827 		iulp_not_used : 25;
828 } iulp_t;
829 
830 /*
831  * The conn drain list structure (idl_t), protected by idl_lock.  Each conn_t
832  * inserted in the list points back at this idl_t using conn_idl, and is
833  * chained by conn_drain_next and conn_drain_prev, which are also protected by
834  * idl_lock.  When flow control is relieved, either ip_wsrv() (STREAMS) or
835  * ill_flow_enable() (non-STREAMS) will call conn_drain().
836  *
837  * The conn drain list, idl_t, itself is part of tx cookie list structure.
838  * A tx cookie list points to a blocked Tx ring and contains the list of
839  * all conn's that are blocked due to the flow-controlled Tx ring (via
840  * the idl drain list). Note that a link can have multiple Tx rings. The
841  * drain list will store the conn's blocked due to Tx ring being flow
842  * controlled.
843  */
844 
845 typedef uintptr_t ip_mac_tx_cookie_t;
846 typedef	struct idl_s idl_t;
847 typedef	struct idl_tx_list_s idl_tx_list_t;
848 
849 struct idl_tx_list_s {
850 	ip_mac_tx_cookie_t	txl_cookie;
851 	kmutex_t		txl_lock;	/* Lock for this list */
852 	idl_t			*txl_drain_list;
853 	int			txl_drain_index;
854 };
855 
856 struct idl_s {
857 	conn_t		*idl_conn;		/* Head of drain list */
858 	kmutex_t	idl_lock;		/* Lock for this list */
859 	idl_tx_list_t	*idl_itl;
860 };
861 
862 /*
863  * Interface route structure which holds the necessary information to recreate
864  * routes that are tied to an interface i.e. have ire_ill set.
865  *
866  * These routes which were initially created via a routing socket or via the
867  * SIOCADDRT ioctl may be gateway routes (RTF_GATEWAY being set) or may be
868  * traditional interface routes.  When an ill comes back up after being
869  * down, this information will be used to recreate the routes.  These
870  * are part of an mblk_t chain that hangs off of the ILL (ill_saved_ire_mp).
871  */
872 typedef struct ifrt_s {
873 	ushort_t	ifrt_type;		/* Type of IRE */
874 	in6_addr_t	ifrt_v6addr;		/* Address IRE represents. */
875 	in6_addr_t	ifrt_v6gateway_addr;	/* Gateway if IRE_OFFLINK */
876 	in6_addr_t	ifrt_v6setsrc_addr;	/* Src addr if RTF_SETSRC */
877 	in6_addr_t	ifrt_v6mask;		/* Mask for matching IRE. */
878 	uint32_t	ifrt_flags;		/* flags related to route */
879 	iulp_t		ifrt_metrics;		/* Routing socket metrics */
880 	zoneid_t	ifrt_zoneid;		/* zoneid for route */
881 } ifrt_t;
882 
883 #define	ifrt_addr		V4_PART_OF_V6(ifrt_v6addr)
884 #define	ifrt_gateway_addr	V4_PART_OF_V6(ifrt_v6gateway_addr)
885 #define	ifrt_mask		V4_PART_OF_V6(ifrt_v6mask)
886 #define	ifrt_setsrc_addr	V4_PART_OF_V6(ifrt_v6setsrc_addr)
887 
888 /* Number of IP addresses that can be hosted on a physical interface */
889 #define	MAX_ADDRS_PER_IF	8192
890 /*
891  * Number of Source addresses to be considered for source address
892  * selection. Used by ipif_select_source_v4/v6.
893  */
894 #define	MAX_IPIF_SELECT_SOURCE	50
895 
896 #ifdef IP_DEBUG
897 /*
898  * Trace refholds and refreles for debugging.
899  */
900 #define	TR_STACK_DEPTH	14
901 typedef struct tr_buf_s {
902 	int	tr_depth;
903 	clock_t	tr_time;
904 	pc_t	tr_stack[TR_STACK_DEPTH];
905 } tr_buf_t;
906 
907 typedef struct th_trace_s {
908 	int		th_refcnt;
909 	uint_t		th_trace_lastref;
910 	kthread_t	*th_id;
911 #define	TR_BUF_MAX	38
912 	tr_buf_t	th_trbuf[TR_BUF_MAX];
913 } th_trace_t;
914 
915 typedef struct th_hash_s {
916 	list_node_t	thh_link;
917 	mod_hash_t	*thh_hash;
918 	ip_stack_t	*thh_ipst;
919 } th_hash_t;
920 #endif
921 
922 /* The following are ipif_state_flags */
923 #define	IPIF_CONDEMNED		0x1	/* The ipif is being removed */
924 #define	IPIF_CHANGING		0x2	/* A critcal ipif field is changing */
925 #define	IPIF_SET_LINKLOCAL	0x10	/* transient flag during bringup */
926 
927 /* IP interface structure, one per local address */
928 typedef struct ipif_s {
929 	struct	ipif_s	*ipif_next;
930 	struct	ill_s	*ipif_ill;	/* Back pointer to our ill */
931 	int	ipif_id;		/* Logical unit number */
932 	in6_addr_t ipif_v6lcl_addr;	/* Local IP address for this if. */
933 	in6_addr_t ipif_v6subnet;	/* Subnet prefix for this if. */
934 	in6_addr_t ipif_v6net_mask;	/* Net mask for this interface. */
935 	in6_addr_t ipif_v6brd_addr;	/* Broadcast addr for this interface. */
936 	in6_addr_t ipif_v6pp_dst_addr;	/* Point-to-point dest address. */
937 	uint64_t ipif_flags;		/* Interface flags. */
938 	uint_t	ipif_ire_type;		/* IRE_LOCAL or IRE_LOOPBACK */
939 
940 	/*
941 	 * The packet count in the ipif contain the sum of the
942 	 * packet counts in dead IRE_LOCAL/LOOPBACK for this ipif.
943 	 */
944 	uint_t	ipif_ib_pkt_count;	/* Inbound packets for our dead IREs */
945 
946 	/* Exclusive bit fields, protected by ipsq_t */
947 	unsigned int
948 		ipif_was_up : 1,	/* ipif was up before */
949 		ipif_addr_ready : 1,	/* DAD is done */
950 		ipif_was_dup : 1,	/* DAD had failed */
951 		ipif_added_nce : 1,	/* nce added for local address */
952 
953 		ipif_pad_to_31 : 28;
954 
955 	ilm_t	*ipif_allhosts_ilm;	/* For all-nodes join */
956 	ilm_t	*ipif_solmulti_ilm;	/* For IPv6 solicited multicast join */
957 
958 	uint_t	ipif_seqid;		/* unique index across all ills */
959 	uint_t	ipif_state_flags;	/* See IPIF_* flag defs above */
960 	uint_t	ipif_refcnt;		/* active consistent reader cnt */
961 
962 	zoneid_t ipif_zoneid;		/* zone ID number */
963 	timeout_id_t ipif_recovery_id;	/* Timer for DAD recovery */
964 	boolean_t ipif_trace_disable;	/* True when alloc fails */
965 	/*
966 	 * For an IPMP interface, ipif_bound_ill tracks the ill whose hardware
967 	 * information this ipif is associated with via ARP/NDP.  We can use
968 	 * an ill pointer (rather than an index) because only ills that are
969 	 * part of a group will be pointed to, and an ill cannot disappear
970 	 * while it's in a group.
971 	 */
972 	struct ill_s    *ipif_bound_ill;
973 	struct ipif_s   *ipif_bound_next; /* bound ipif chain */
974 	boolean_t	ipif_bound;	/* B_TRUE if we successfully bound */
975 
976 	struct ire_s	*ipif_ire_local; /* Our IRE_LOCAL or LOOPBACK */
977 	struct ire_s	*ipif_ire_if;	 /* Our IRE_INTERFACE */
978 } ipif_t;
979 
980 /*
981  * The following table lists the protection levels of the various members
982  * of the ipif_t. The following notation is used.
983  *
984  * Write once - Written to only once at the time of bringing up
985  * the interface and can be safely read after the bringup without any lock.
986  *
987  * ipsq - Need to execute in the ipsq to perform the indicated access.
988  *
989  * ill_lock - Need to hold this mutex to perform the indicated access.
990  *
991  * ill_g_lock - Need to hold this rw lock as reader/writer for read access or
992  * write access respectively.
993  *
994  * down ill - Written to only when the ill is down (i.e all ipifs are down)
995  * up ill - Read only when the ill is up (i.e. at least 1 ipif is up)
996  *
997  *		 Table of ipif_t members and their protection
998  *
999  * ipif_next		ipsq + ill_lock +	ipsq OR ill_lock OR
1000  *			ill_g_lock		ill_g_lock
1001  * ipif_ill		ipsq + down ipif	write once
1002  * ipif_id		ipsq + down ipif	write once
1003  * ipif_v6lcl_addr	ipsq + down ipif	up ipif
1004  * ipif_v6subnet	ipsq + down ipif	up ipif
1005  * ipif_v6net_mask	ipsq + down ipif	up ipif
1006  *
1007  * ipif_v6brd_addr
1008  * ipif_v6pp_dst_addr
1009  * ipif_flags		ill_lock		ill_lock
1010  * ipif_ire_type	ipsq + down ill		up ill
1011  *
1012  * ipif_ib_pkt_count	Approx
1013  *
1014  * bit fields		ill_lock		ill_lock
1015  *
1016  * ipif_allhosts_ilm	ipsq			ipsq
1017  * ipif_solmulti_ilm	ipsq			ipsq
1018  *
1019  * ipif_seqid		ipsq			Write once
1020  *
1021  * ipif_state_flags	ill_lock		ill_lock
1022  * ipif_refcnt		ill_lock		ill_lock
1023  * ipif_bound_ill	ipsq + ipmp_lock	ipsq OR ipmp_lock
1024  * ipif_bound_next	ipsq			ipsq
1025  * ipif_bound		ipsq			ipsq
1026  *
1027  * ipif_ire_local	ipsq + ips_ill_g_lock	ipsq OR ips_ill_g_lock
1028  * ipif_ire_if		ipsq + ips_ill_g_lock	ipsq OR ips_ill_g_lock
1029  */
1030 
1031 /*
1032  * Return values from ip_laddr_verify_{v4,v6}
1033  */
1034 typedef enum { IPVL_UNICAST_UP, IPVL_UNICAST_DOWN, IPVL_MCAST, IPVL_BCAST,
1035 	    IPVL_BAD} ip_laddr_t;
1036 
1037 
1038 #define	IP_TR_HASH(tid)	((((uintptr_t)tid) >> 6) & (IP_TR_HASH_MAX - 1))
1039 
1040 #ifdef DEBUG
1041 #define	IPIF_TRACE_REF(ipif)	ipif_trace_ref(ipif)
1042 #define	ILL_TRACE_REF(ill)	ill_trace_ref(ill)
1043 #define	IPIF_UNTRACE_REF(ipif)	ipif_untrace_ref(ipif)
1044 #define	ILL_UNTRACE_REF(ill)	ill_untrace_ref(ill)
1045 #else
1046 #define	IPIF_TRACE_REF(ipif)
1047 #define	ILL_TRACE_REF(ill)
1048 #define	IPIF_UNTRACE_REF(ipif)
1049 #define	ILL_UNTRACE_REF(ill)
1050 #endif
1051 
1052 /* IPv4 compatibility macros */
1053 #define	ipif_lcl_addr		V4_PART_OF_V6(ipif_v6lcl_addr)
1054 #define	ipif_subnet		V4_PART_OF_V6(ipif_v6subnet)
1055 #define	ipif_net_mask		V4_PART_OF_V6(ipif_v6net_mask)
1056 #define	ipif_brd_addr		V4_PART_OF_V6(ipif_v6brd_addr)
1057 #define	ipif_pp_dst_addr	V4_PART_OF_V6(ipif_v6pp_dst_addr)
1058 
1059 /* Macros for easy backreferences to the ill. */
1060 #define	ipif_isv6		ipif_ill->ill_isv6
1061 
1062 #define	SIOCLIFADDR_NDX 112	/* ndx of SIOCLIFADDR in the ndx ioctl table */
1063 
1064 /*
1065  * mode value for ip_ioctl_finish for finishing an ioctl
1066  */
1067 #define	CONN_CLOSE	1		/* No mi_copy */
1068 #define	COPYOUT		2		/* do an mi_copyout if needed */
1069 #define	NO_COPYOUT	3		/* do an mi_copy_done */
1070 #define	IPI2MODE(ipi)	((ipi)->ipi_flags & IPI_GET_CMD ? COPYOUT : NO_COPYOUT)
1071 
1072 /*
1073  * The IP-MT design revolves around the serialization objects ipsq_t (IPSQ)
1074  * and ipxop_t (exclusive operation or "xop").  Becoming "writer" on an IPSQ
1075  * ensures that no other threads can become "writer" on any IPSQs sharing that
1076  * IPSQ's xop until the writer thread is done.
1077  *
1078  * Each phyint points to one IPSQ that remains fixed over the phyint's life.
1079  * Each IPSQ points to one xop that can change over the IPSQ's life.  If a
1080  * phyint is *not* in an IPMP group, then its IPSQ will refer to the IPSQ's
1081  * "own" xop (ipsq_ownxop).  If a phyint *is* part of an IPMP group, then its
1082  * IPSQ will refer to the "group" xop, which is shorthand for the xop of the
1083  * IPSQ of the IPMP meta-interface's phyint.  Thus, all phyints that are part
1084  * of the same IPMP group will have their IPSQ's point to the group xop, and
1085  * thus becoming "writer" on any phyint in the group will prevent any other
1086  * writer on any other phyint in the group.  All IPSQs sharing the same xop
1087  * are chained together through ipsq_next (in the degenerate common case,
1088  * ipsq_next simply refers to itself).  Note that the group xop is guaranteed
1089  * to exist at least as long as there are members in the group, since the IPMP
1090  * meta-interface can only be destroyed if the group is empty.
1091  *
1092  * Incoming exclusive operation requests are enqueued on the IPSQ they arrived
1093  * on rather than the xop.  This makes switching xop's (as would happen when a
1094  * phyint leaves an IPMP group) simple, because after the phyint leaves the
1095  * group, any operations enqueued on its IPSQ can be safely processed with
1096  * respect to its new xop, and any operations enqueued on the IPSQs of its
1097  * former group can be processed with respect to their existing group xop.
1098  * Even so, switching xops is a subtle dance; see ipsq_dq() for details.
1099  *
1100  * An IPSQ's "own" xop is embedded within the IPSQ itself since they have have
1101  * identical lifetimes, and because doing so simplifies pointer management.
1102  * While each phyint and IPSQ point to each other, it is not possible to free
1103  * the IPSQ when the phyint is freed, since we may still *inside* the IPSQ
1104  * when the phyint is being freed.  Thus, ipsq_phyint is set to NULL when the
1105  * phyint is freed, and the IPSQ free is later done in ipsq_exit().
1106  *
1107  * ipsq_t synchronization:	read			write
1108  *
1109  *	ipsq_xopq_mphead	ipx_lock		ipx_lock
1110  *	ipsq_xopq_mptail	ipx_lock		ipx_lock
1111  *	ipsq_xop_switch_mp	ipsq_lock		ipsq_lock
1112  *	ipsq_phyint		write once		write once
1113  *	ipsq_next		RW_READER ill_g_lock	RW_WRITER ill_g_lock
1114  *	ipsq_xop 		ipsq_lock or ipsq	ipsq_lock + ipsq
1115  *	ipsq_swxop		ipsq			ipsq
1116  * 	ipsq_ownxop		see ipxop_t		see ipxop_t
1117  *	ipsq_ipst		write once		write once
1118  *
1119  * ipxop_t synchronization:     read			write
1120  *
1121  *	ipx_writer  		ipx_lock		ipx_lock
1122  *	ipx_xop_queued		ipx_lock 		ipx_lock
1123  *	ipx_mphead		ipx_lock		ipx_lock
1124  *	ipx_mptail		ipx_lock		ipx_lock
1125  *	ipx_ipsq		write once		write once
1126  *	ips_ipsq_queued		ipx_lock		ipx_lock
1127  *	ipx_waitfor		ipsq or ipx_lock	ipsq + ipx_lock
1128  *	ipx_reentry_cnt		ipsq or ipx_lock	ipsq + ipx_lock
1129  *	ipx_current_done	ipsq			ipsq
1130  *	ipx_current_ioctl	ipsq			ipsq
1131  *	ipx_current_ipif	ipsq or ipx_lock	ipsq + ipx_lock
1132  *	ipx_pending_ipif	ipsq or ipx_lock	ipsq + ipx_lock
1133  *	ipx_pending_mp		ipsq or ipx_lock	ipsq + ipx_lock
1134  *	ipx_forced		ipsq			ipsq
1135  *	ipx_depth		ipsq			ipsq
1136  *	ipx_stack		ipsq			ipsq
1137  */
1138 typedef struct ipxop_s {
1139 	kmutex_t	ipx_lock;	/* see above */
1140 	kthread_t	*ipx_writer;  	/* current owner */
1141 	mblk_t		*ipx_mphead;	/* messages tied to this op */
1142 	mblk_t		*ipx_mptail;
1143 	struct ipsq_s	*ipx_ipsq;	/* associated ipsq */
1144 	boolean_t	ipx_ipsq_queued; /* ipsq using xop has queued op */
1145 	int		ipx_waitfor;	/* waiting; values encoded below */
1146 	int		ipx_reentry_cnt;
1147 	boolean_t	ipx_current_done;  /* is the current operation done? */
1148 	int		ipx_current_ioctl; /* current ioctl, or 0 if no ioctl */
1149 	ipif_t		*ipx_current_ipif; /* ipif for current op */
1150 	ipif_t		*ipx_pending_ipif; /* ipif for ipx_pending_mp */
1151 	mblk_t 		*ipx_pending_mp;   /* current ioctl mp while waiting */
1152 	boolean_t	ipx_forced; 			/* debugging aid */
1153 #ifdef DEBUG
1154 	int		ipx_depth;			/* debugging aid */
1155 #define	IPX_STACK_DEPTH	15
1156 	pc_t		ipx_stack[IPX_STACK_DEPTH];	/* debugging aid */
1157 #endif
1158 } ipxop_t;
1159 
1160 typedef struct ipsq_s {
1161 	kmutex_t ipsq_lock;		/* see above */
1162 	mblk_t	*ipsq_switch_mp;	/* op to handle right after switch */
1163 	mblk_t	*ipsq_xopq_mphead;	/* list of excl ops (mostly ioctls) */
1164 	mblk_t	*ipsq_xopq_mptail;
1165 	struct phyint	*ipsq_phyint;	/* associated phyint */
1166 	struct ipsq_s	*ipsq_next;	/* next ipsq sharing ipsq_xop */
1167 	struct ipxop_s	*ipsq_xop;	/* current xop synchronization info */
1168 	struct ipxop_s	*ipsq_swxop;	/* switch xop to on ipsq_exit() */
1169 	struct ipxop_s	ipsq_ownxop;	/* our own xop (may not be in-use) */
1170 	ip_stack_t	*ipsq_ipst;	/* does not have a netstack_hold */
1171 } ipsq_t;
1172 
1173 /*
1174  * ipx_waitfor values:
1175  */
1176 enum {
1177 	IPIF_DOWN = 1,	/* ipif_down() waiting for refcnts to drop */
1178 	ILL_DOWN,	/* ill_down() waiting for refcnts to drop */
1179 	IPIF_FREE,	/* ipif_free() waiting for refcnts to drop */
1180 	ILL_FREE	/* ill unplumb waiting for refcnts to drop */
1181 };
1182 
1183 /* Operation types for ipsq_try_enter() */
1184 #define	CUR_OP 0	/* request writer within current operation */
1185 #define	NEW_OP 1	/* request writer for a new operation */
1186 #define	SWITCH_OP 2	/* request writer once IPSQ XOP switches */
1187 
1188 /*
1189  * Kstats tracked on each IPMP meta-interface.  Order here must match
1190  * ipmp_kstats[] in ip/ipmp.c.
1191  */
1192 enum {
1193 	IPMP_KSTAT_OBYTES,	IPMP_KSTAT_OBYTES64,	IPMP_KSTAT_RBYTES,
1194 	IPMP_KSTAT_RBYTES64,	IPMP_KSTAT_OPACKETS,	IPMP_KSTAT_OPACKETS64,
1195 	IPMP_KSTAT_OERRORS,	IPMP_KSTAT_IPACKETS,	IPMP_KSTAT_IPACKETS64,
1196 	IPMP_KSTAT_IERRORS,	IPMP_KSTAT_MULTIRCV,	IPMP_KSTAT_MULTIXMT,
1197 	IPMP_KSTAT_BRDCSTRCV,	IPMP_KSTAT_BRDCSTXMT,	IPMP_KSTAT_LINK_UP,
1198 	IPMP_KSTAT_MAX		/* keep last */
1199 };
1200 
1201 /*
1202  * phyint represents state that is common to both IPv4 and IPv6 interfaces.
1203  * There is a separate ill_t representing IPv4 and IPv6 which has a
1204  * backpointer to the phyint structure for accessing common state.
1205  */
1206 typedef struct phyint {
1207 	struct ill_s	*phyint_illv4;
1208 	struct ill_s	*phyint_illv6;
1209 	uint_t		phyint_ifindex;		/* SIOCSLIFINDEX */
1210 	uint64_t	phyint_flags;
1211 	avl_node_t	phyint_avl_by_index;	/* avl tree by index */
1212 	avl_node_t	phyint_avl_by_name;	/* avl tree by name */
1213 	kmutex_t	phyint_lock;
1214 	struct ipsq_s	*phyint_ipsq;		/* back pointer to ipsq */
1215 	struct ipmp_grp_s *phyint_grp;		/* associated IPMP group */
1216 	char		phyint_name[LIFNAMSIZ];	/* physical interface name */
1217 	uint64_t	phyint_kstats0[IPMP_KSTAT_MAX];	/* baseline kstats */
1218 } phyint_t;
1219 
1220 #define	CACHE_ALIGN_SIZE 64
1221 #define	CACHE_ALIGN(align_struct)	P2ROUNDUP(sizeof (struct align_struct),\
1222 							CACHE_ALIGN_SIZE)
1223 struct _phyint_list_s_ {
1224 	avl_tree_t	phyint_list_avl_by_index;	/* avl tree by index */
1225 	avl_tree_t	phyint_list_avl_by_name;	/* avl tree by name */
1226 };
1227 
1228 typedef union phyint_list_u {
1229 	struct	_phyint_list_s_ phyint_list_s;
1230 	char	phyint_list_filler[CACHE_ALIGN(_phyint_list_s_)];
1231 } phyint_list_t;
1232 
1233 #define	phyint_list_avl_by_index	phyint_list_s.phyint_list_avl_by_index
1234 #define	phyint_list_avl_by_name		phyint_list_s.phyint_list_avl_by_name
1235 
1236 /*
1237  * Fragmentation hash bucket
1238  */
1239 typedef struct ipfb_s {
1240 	struct ipf_s	*ipfb_ipf;	/* List of ... */
1241 	size_t		ipfb_count;	/* Count of bytes used by frag(s) */
1242 	kmutex_t	ipfb_lock;	/* Protect all ipf in list */
1243 	uint_t		ipfb_frag_pkts; /* num of distinct fragmented pkts */
1244 } ipfb_t;
1245 
1246 /*
1247  * IRE bucket structure. Usually there is an array of such structures,
1248  * each pointing to a linked list of ires. irb_refcnt counts the number
1249  * of walkers of a given hash bucket. Usually the reference count is
1250  * bumped up if the walker wants no IRES to be DELETED while walking the
1251  * list. Bumping up does not PREVENT ADDITION. This allows walking a given
1252  * hash bucket without stumbling up on a free pointer.
1253  *
1254  * irb_t structures in ip_ftable are dynamically allocated and freed.
1255  * In order to identify the irb_t structures that can be safely kmem_free'd
1256  * we need to ensure that
1257  *  - the irb_refcnt is quiescent, indicating no other walkers,
1258  *  - no other threads or ire's are holding references to the irb,
1259  *	i.e., irb_nire == 0,
1260  *  - there are no active ire's in the bucket, i.e., irb_ire_cnt == 0
1261  */
1262 typedef struct irb {
1263 	struct ire_s	*irb_ire;	/* First ire in this bucket */
1264 					/* Should be first in this struct */
1265 	krwlock_t	irb_lock;	/* Protect this bucket */
1266 	uint_t		irb_refcnt;	/* Protected by irb_lock */
1267 	uchar_t		irb_marks;	/* CONDEMNED ires in this bucket ? */
1268 #define	IRB_MARK_CONDEMNED	0x0001	/* Contains some IRE_IS_CONDEMNED */
1269 #define	IRB_MARK_DYNAMIC	0x0002	/* Dynamically allocated */
1270 	/* Once IPv6 uses radix then IRB_MARK_DYNAMIC will be always be set */
1271 	uint_t		irb_ire_cnt;	/* Num of active IRE in this bucket */
1272 	int		irb_nire;	/* Num of ftable ire's that ref irb */
1273 	ip_stack_t	*irb_ipst;	/* Does not have a netstack_hold */
1274 } irb_t;
1275 
1276 /*
1277  * This is the structure used to store the multicast physical addresses
1278  * that an interface has joined.
1279  * The refcnt keeps track of the number of multicast IP addresses mapping
1280  * to a physical multicast address.
1281  */
1282 typedef struct multiphysaddr_s {
1283 	struct	multiphysaddr_s  *mpa_next;
1284 	char	mpa_addr[IP_MAX_HW_LEN];
1285 	int	mpa_refcnt;
1286 } multiphysaddr_t;
1287 
1288 #define	IRB2RT(irb)	(rt_t *)((caddr_t)(irb) - offsetof(rt_t, rt_irb))
1289 
1290 /* Forward declarations */
1291 struct dce_s;
1292 typedef struct dce_s dce_t;
1293 struct ire_s;
1294 typedef struct ire_s ire_t;
1295 struct ncec_s;
1296 typedef struct ncec_s ncec_t;
1297 struct nce_s;
1298 typedef struct nce_s nce_t;
1299 struct ip_recv_attr_s;
1300 typedef struct ip_recv_attr_s ip_recv_attr_t;
1301 struct ip_xmit_attr_s;
1302 typedef struct ip_xmit_attr_s ip_xmit_attr_t;
1303 
1304 struct tsol_ire_gw_secattr_s;
1305 typedef struct tsol_ire_gw_secattr_s tsol_ire_gw_secattr_t;
1306 
1307 /*
1308  * This is a structure for a one-element route cache that is passed
1309  * by reference between ip_input and ill_inputfn.
1310  */
1311 typedef struct {
1312 	ire_t		*rtc_ire;
1313 	ipaddr_t	rtc_ipaddr;
1314 	in6_addr_t	rtc_ip6addr;
1315 } rtc_t;
1316 
1317 /*
1318  * Note: Temporarily use 64 bits, and will probably go back to 32 bits after
1319  * more cleanup work is done.
1320  */
1321 typedef uint64_t iaflags_t;
1322 
1323 /* The ill input function pointer type */
1324 typedef void (*pfillinput_t)(mblk_t *, void *, void *, ip_recv_attr_t *,
1325     rtc_t *);
1326 
1327 /* The ire receive function pointer type */
1328 typedef void (*pfirerecv_t)(ire_t *, mblk_t *, void *, ip_recv_attr_t *);
1329 
1330 /* The ire send and postfrag function pointer types */
1331 typedef int (*pfiresend_t)(ire_t *, mblk_t *, void *,
1332     ip_xmit_attr_t *, uint32_t *);
1333 typedef int (*pfirepostfrag_t)(mblk_t *, nce_t *, iaflags_t, uint_t, uint32_t,
1334     zoneid_t, zoneid_t, uintptr_t *);
1335 
1336 
1337 #define	IP_V4_G_HEAD	0
1338 #define	IP_V6_G_HEAD	1
1339 
1340 #define	MAX_G_HEADS	2
1341 
1342 /*
1343  * unpadded ill_if structure
1344  */
1345 struct 	_ill_if_s_ {
1346 	union ill_if_u	*illif_next;
1347 	union ill_if_u	*illif_prev;
1348 	avl_tree_t	illif_avl_by_ppa;	/* AVL tree sorted on ppa */
1349 	vmem_t		*illif_ppa_arena;	/* ppa index space */
1350 	uint16_t	illif_mcast_v1;		/* hints for		  */
1351 	uint16_t	illif_mcast_v2;		/* [igmp|mld]_slowtimo	  */
1352 	int		illif_name_len;		/* name length */
1353 	char		illif_name[LIFNAMSIZ];	/* name of interface type */
1354 };
1355 
1356 /* cache aligned ill_if structure */
1357 typedef union 	ill_if_u {
1358 	struct  _ill_if_s_ ill_if_s;
1359 	char 	illif_filler[CACHE_ALIGN(_ill_if_s_)];
1360 } ill_if_t;
1361 
1362 #define	illif_next		ill_if_s.illif_next
1363 #define	illif_prev		ill_if_s.illif_prev
1364 #define	illif_avl_by_ppa	ill_if_s.illif_avl_by_ppa
1365 #define	illif_ppa_arena		ill_if_s.illif_ppa_arena
1366 #define	illif_mcast_v1		ill_if_s.illif_mcast_v1
1367 #define	illif_mcast_v2		ill_if_s.illif_mcast_v2
1368 #define	illif_name		ill_if_s.illif_name
1369 #define	illif_name_len		ill_if_s.illif_name_len
1370 
1371 typedef struct ill_walk_context_s {
1372 	int	ctx_current_list; /* current list being searched */
1373 	int	ctx_last_list;	 /* last list to search */
1374 } ill_walk_context_t;
1375 
1376 /*
1377  * ill_g_heads structure, one for IPV4 and one for IPV6
1378  */
1379 struct _ill_g_head_s_ {
1380 	ill_if_t	*ill_g_list_head;
1381 	ill_if_t	*ill_g_list_tail;
1382 };
1383 
1384 typedef union ill_g_head_u {
1385 	struct _ill_g_head_s_ ill_g_head_s;
1386 	char	ill_g_head_filler[CACHE_ALIGN(_ill_g_head_s_)];
1387 } ill_g_head_t;
1388 
1389 #define	ill_g_list_head	ill_g_head_s.ill_g_list_head
1390 #define	ill_g_list_tail	ill_g_head_s.ill_g_list_tail
1391 
1392 #define	IP_V4_ILL_G_LIST(ipst)	\
1393 	(ipst)->ips_ill_g_heads[IP_V4_G_HEAD].ill_g_list_head
1394 #define	IP_V6_ILL_G_LIST(ipst)	\
1395 	(ipst)->ips_ill_g_heads[IP_V6_G_HEAD].ill_g_list_head
1396 #define	IP_VX_ILL_G_LIST(i, ipst)	\
1397 	(ipst)->ips_ill_g_heads[i].ill_g_list_head
1398 
1399 #define	ILL_START_WALK_V4(ctx_ptr, ipst)	\
1400 	ill_first(IP_V4_G_HEAD, IP_V4_G_HEAD, ctx_ptr, ipst)
1401 #define	ILL_START_WALK_V6(ctx_ptr, ipst)	\
1402 	ill_first(IP_V6_G_HEAD, IP_V6_G_HEAD, ctx_ptr, ipst)
1403 #define	ILL_START_WALK_ALL(ctx_ptr, ipst)	\
1404 	ill_first(MAX_G_HEADS, MAX_G_HEADS, ctx_ptr, ipst)
1405 
1406 /*
1407  * Capabilities, possible flags for ill_capabilities.
1408  */
1409 #define	ILL_CAPAB_LSO		0x04		/* Large Send Offload */
1410 #define	ILL_CAPAB_HCKSUM	0x08		/* Hardware checksumming */
1411 #define	ILL_CAPAB_ZEROCOPY	0x10		/* Zero-copy */
1412 #define	ILL_CAPAB_DLD		0x20		/* DLD capabilities */
1413 #define	ILL_CAPAB_DLD_POLL	0x40		/* Polling */
1414 #define	ILL_CAPAB_DLD_DIRECT	0x80		/* Direct function call */
1415 
1416 /*
1417  * Per-ill Hardware Checksumming capbilities.
1418  */
1419 typedef struct ill_hcksum_capab_s ill_hcksum_capab_t;
1420 
1421 /*
1422  * Per-ill Zero-copy capabilities.
1423  */
1424 typedef struct ill_zerocopy_capab_s ill_zerocopy_capab_t;
1425 
1426 /*
1427  * DLD capbilities.
1428  */
1429 typedef struct ill_dld_capab_s ill_dld_capab_t;
1430 
1431 /*
1432  * Per-ill polling resource map.
1433  */
1434 typedef struct ill_rx_ring ill_rx_ring_t;
1435 
1436 /*
1437  * Per-ill Large Send Offload capabilities.
1438  */
1439 typedef struct ill_lso_capab_s ill_lso_capab_t;
1440 
1441 /* The following are ill_state_flags */
1442 #define	ILL_LL_SUBNET_PENDING	0x01	/* Waiting for DL_INFO_ACK from drv */
1443 #define	ILL_CONDEMNED		0x02	/* No more new ref's to the ILL */
1444 #define	ILL_DL_UNBIND_IN_PROGRESS	0x04	/* UNBIND_REQ is sent */
1445 /*
1446  * ILL_DOWN_IN_PROGRESS is set to ensure the following:
1447  * - no packets are sent to the driver after the DL_UNBIND_REQ is sent,
1448  * - no longstanding references will be acquired on objects that are being
1449  *   brought down.
1450  */
1451 #define	ILL_DOWN_IN_PROGRESS	0x08
1452 
1453 /* Is this an ILL whose source address is used by other ILL's ? */
1454 #define	IS_USESRC_ILL(ill)			\
1455 	(((ill)->ill_usesrc_ifindex == 0) &&	\
1456 	((ill)->ill_usesrc_grp_next != NULL))
1457 
1458 /* Is this a client/consumer of the usesrc ILL ? */
1459 #define	IS_USESRC_CLI_ILL(ill)			\
1460 	(((ill)->ill_usesrc_ifindex != 0) &&	\
1461 	((ill)->ill_usesrc_grp_next != NULL))
1462 
1463 /* Is this an virtual network interface (vni) ILL ? */
1464 #define	IS_VNI(ill)							\
1465 	(((ill)->ill_phyint->phyint_flags & (PHYI_LOOPBACK|PHYI_VIRTUAL)) == \
1466 	PHYI_VIRTUAL)
1467 
1468 /* Is this a loopback ILL? */
1469 #define	IS_LOOPBACK(ill) \
1470 	((ill)->ill_phyint->phyint_flags & PHYI_LOOPBACK)
1471 
1472 /* Is this an IPMP meta-interface ILL? */
1473 #define	IS_IPMP(ill)							\
1474 	((ill)->ill_phyint->phyint_flags & PHYI_IPMP)
1475 
1476 /* Is this ILL under an IPMP meta-interface? (aka "in a group?") */
1477 #define	IS_UNDER_IPMP(ill)						\
1478 	((ill)->ill_grp != NULL && !IS_IPMP(ill))
1479 
1480 /* Is ill1 in the same illgrp as ill2? */
1481 #define	IS_IN_SAME_ILLGRP(ill1, ill2)					\
1482 	((ill1)->ill_grp != NULL && ((ill1)->ill_grp == (ill2)->ill_grp))
1483 
1484 /* Is ill1 on the same LAN as ill2? */
1485 #define	IS_ON_SAME_LAN(ill1, ill2)					\
1486 	((ill1) == (ill2) || IS_IN_SAME_ILLGRP(ill1, ill2))
1487 
1488 #define	ILL_OTHER(ill)							\
1489 	((ill)->ill_isv6 ? (ill)->ill_phyint->phyint_illv4 :		\
1490 	    (ill)->ill_phyint->phyint_illv6)
1491 
1492 /*
1493  * IPMP group ILL state structure -- up to two per IPMP group (V4 and V6).
1494  * Created when the V4 and/or V6 IPMP meta-interface is I_PLINK'd.  It is
1495  * guaranteed to persist while there are interfaces of that type in the group.
1496  * In general, most fields are accessed outside of the IPSQ (e.g., in the
1497  * datapath), and thus use locks in addition to the IPSQ for protection.
1498  *
1499  * synchronization:		read			write
1500  *
1501  *	ig_if			ipsq or ill_g_lock	ipsq and ill_g_lock
1502  *	ig_actif		ipsq or ipmp_lock	ipsq and ipmp_lock
1503  *	ig_nactif		ipsq or ipmp_lock	ipsq and ipmp_lock
1504  *	ig_next_ill		ipsq or ipmp_lock	ipsq and ipmp_lock
1505  *	ig_ipmp_ill		write once		write once
1506  *	ig_cast_ill		ipsq or ipmp_lock	ipsq and ipmp_lock
1507  *	ig_arpent		ipsq			ipsq
1508  *	ig_mtu			ipsq			ipsq
1509  */
1510 typedef struct ipmp_illgrp_s {
1511 	list_t		ig_if; 		/* list of all interfaces */
1512 	list_t		ig_actif;	/* list of active interfaces */
1513 	uint_t		ig_nactif;	/* number of active interfaces */
1514 	struct ill_s	*ig_next_ill;	/* next active interface to use */
1515 	struct ill_s	*ig_ipmp_ill;	/* backpointer to IPMP meta-interface */
1516 	struct ill_s	*ig_cast_ill;	/* nominated ill for multi/broadcast */
1517 	list_t		ig_arpent;	/* list of ARP entries */
1518 	uint_t		ig_mtu;		/* ig_ipmp_ill->ill_max_mtu */
1519 } ipmp_illgrp_t;
1520 
1521 /*
1522  * IPMP group state structure -- one per IPMP group.  Created when the
1523  * IPMP meta-interface is plumbed; it is guaranteed to persist while there
1524  * are interfaces in it.
1525  *
1526  * ipmp_grp_t synchronization:		read			write
1527  *
1528  *	gr_name				ipmp_lock		ipmp_lock
1529  *	gr_ifname			write once		write once
1530  *	gr_mactype			ipmp_lock		ipmp_lock
1531  *	gr_phyint			write once		write once
1532  *	gr_nif				ipmp_lock		ipmp_lock
1533  *	gr_nactif			ipsq			ipsq
1534  *	gr_v4				ipmp_lock		ipmp_lock
1535  *	gr_v6				ipmp_lock		ipmp_lock
1536  *	gr_nv4				ipmp_lock		ipmp_lock
1537  *	gr_nv6				ipmp_lock		ipmp_lock
1538  *	gr_pendv4			ipmp_lock		ipmp_lock
1539  *	gr_pendv6			ipmp_lock		ipmp_lock
1540  *	gr_linkdownmp			ipsq			ipsq
1541  *	gr_ksp				ipmp_lock		ipmp_lock
1542  *	gr_kstats0			atomic			atomic
1543  */
1544 typedef struct ipmp_grp_s {
1545 	char		gr_name[LIFGRNAMSIZ];	/* group name */
1546 	char		gr_ifname[LIFNAMSIZ];	/* interface name */
1547 	t_uscalar_t	gr_mactype;	/* DLPI mactype of group */
1548 	phyint_t	*gr_phyint;	/* IPMP group phyint */
1549 	uint_t		gr_nif;		/* number of interfaces in group */
1550 	uint_t		gr_nactif; 	/* number of active interfaces */
1551 	ipmp_illgrp_t	*gr_v4;		/* V4 group information */
1552 	ipmp_illgrp_t	*gr_v6;		/* V6 group information */
1553 	uint_t		gr_nv4;		/* number of ills in V4 group */
1554 	uint_t		gr_nv6;		/* number of ills in V6 group */
1555 	uint_t		gr_pendv4; 	/* number of pending ills in V4 group */
1556 	uint_t		gr_pendv6; 	/* number of pending ills in V6 group */
1557 	mblk_t		*gr_linkdownmp;	/* message used to bring link down */
1558 	kstat_t		*gr_ksp;	/* group kstat pointer */
1559 	uint64_t	gr_kstats0[IPMP_KSTAT_MAX]; /* baseline group kstats */
1560 } ipmp_grp_t;
1561 
1562 /*
1563  * IPMP ARP entry -- one per SIOCS*ARP entry tied to the group.  Used to keep
1564  * ARP up-to-date as the active set of interfaces in the group changes.
1565  */
1566 typedef struct ipmp_arpent_s {
1567 	ipaddr_t	ia_ipaddr; 	/* IP address for this entry */
1568 	boolean_t	ia_proxyarp; 	/* proxy ARP entry? */
1569 	boolean_t	ia_notified; 	/* ARP notified about this entry? */
1570 	list_node_t	ia_node; 	/* next ARP entry in list */
1571 	uint16_t	ia_flags;	/* nce_flags for the address */
1572 	size_t		ia_lladdr_len;
1573 	uchar_t		*ia_lladdr;
1574 } ipmp_arpent_t;
1575 
1576 struct arl_s;
1577 
1578 /*
1579  * Per-ill capabilities.
1580  */
1581 struct ill_hcksum_capab_s {
1582 	uint_t	ill_hcksum_version;	/* interface version */
1583 	uint_t	ill_hcksum_txflags;	/* capabilities on transmit */
1584 };
1585 
1586 struct ill_zerocopy_capab_s {
1587 	uint_t	ill_zerocopy_version;	/* interface version */
1588 	uint_t	ill_zerocopy_flags;	/* capabilities */
1589 };
1590 
1591 struct ill_lso_capab_s {
1592 	uint_t	ill_lso_flags;		/* capabilities */
1593 	uint_t	ill_lso_max;		/* maximum size of payload */
1594 };
1595 
1596 /*
1597  * IP Lower level Structure.
1598  * Instance data structure in ip_open when there is a device below us.
1599  */
1600 typedef struct ill_s {
1601 	pfillinput_t ill_inputfn;	/* Fast input function selector */
1602 	ill_if_t *ill_ifptr;		/* pointer to interface type */
1603 	queue_t	*ill_rq;		/* Read queue. */
1604 	queue_t	*ill_wq;		/* Write queue. */
1605 
1606 	int	ill_error;		/* Error value sent up by device. */
1607 
1608 	ipif_t	*ill_ipif;		/* Interface chain for this ILL. */
1609 
1610 	uint_t	ill_ipif_up_count;	/* Number of IPIFs currently up. */
1611 	uint_t	ill_max_frag;		/* Max IDU from DLPI. */
1612 	uint_t	ill_current_frag;	/* Current IDU from DLPI. */
1613 	uint_t	ill_mtu;		/* User-specified MTU; SIOCSLIFMTU */
1614 	uint_t	ill_metric;		/* BSD if metric, for compatibility. */
1615 	char	*ill_name;		/* Our name. */
1616 	uint_t	ill_ipif_dup_count;	/* Number of duplicate addresses. */
1617 	uint_t	ill_name_length;	/* Name length, incl. terminator. */
1618 	uint_t	ill_net_type;		/* IRE_IF_RESOLVER/IRE_IF_NORESOLVER. */
1619 	/*
1620 	 * Physical Point of Attachment num.  If DLPI style 1 provider
1621 	 * then this is derived from the devname.
1622 	 */
1623 	uint_t	ill_ppa;
1624 	t_uscalar_t	ill_sap;
1625 	t_scalar_t	ill_sap_length;	/* Including sign (for position) */
1626 	uint_t	ill_phys_addr_length;	/* Excluding the sap. */
1627 	uint_t	ill_bcast_addr_length;	/* Only set when the DL provider */
1628 					/* supports broadcast. */
1629 	t_uscalar_t	ill_mactype;
1630 	uint8_t	*ill_frag_ptr;		/* Reassembly state. */
1631 	timeout_id_t ill_frag_timer_id; /* timeout id for the frag timer */
1632 	ipfb_t	*ill_frag_hash_tbl;	/* Fragment hash list head. */
1633 
1634 	krwlock_t ill_mcast_lock;	/* Protects multicast state */
1635 	kmutex_t ill_mcast_serializer;	/* Serialize across ilg and ilm state */
1636 	ilm_t	*ill_ilm;		/* Multicast membership for ill */
1637 	uint_t	ill_global_timer;	/* for IGMPv3/MLDv2 general queries */
1638 	int	ill_mcast_type;		/* type of router which is querier */
1639 					/* on this interface */
1640 	uint16_t ill_mcast_v1_time;	/* # slow timeouts since last v1 qry */
1641 	uint16_t ill_mcast_v2_time;	/* # slow timeouts since last v2 qry */
1642 	uint8_t	ill_mcast_v1_tset;	/* 1 => timer is set; 0 => not set */
1643 	uint8_t	ill_mcast_v2_tset;	/* 1 => timer is set; 0 => not set */
1644 
1645 	uint8_t	ill_mcast_rv;		/* IGMPv3/MLDv2 robustness variable */
1646 	int	ill_mcast_qi;		/* IGMPv3/MLDv2 query interval var */
1647 
1648 	/*
1649 	 * All non-NULL cells between 'ill_first_mp_to_free' and
1650 	 * 'ill_last_mp_to_free' are freed in ill_delete.
1651 	 */
1652 #define	ill_first_mp_to_free	ill_bcast_mp
1653 	mblk_t	*ill_bcast_mp;		/* DLPI header for broadcasts. */
1654 	mblk_t	*ill_unbind_mp;		/* unbind mp from ill_dl_up() */
1655 	mblk_t	*ill_promiscoff_mp;	/* for ill_leave_allmulti() */
1656 	mblk_t	*ill_dlpi_deferred;	/* b_next chain of control messages */
1657 	mblk_t	*ill_dest_addr_mp;	/* mblk which holds ill_dest_addr */
1658 	mblk_t	*ill_replumb_mp;	/* replumb mp from ill_replumb() */
1659 	mblk_t	*ill_phys_addr_mp;	/* mblk which holds ill_phys_addr */
1660 	mblk_t	*ill_mcast_deferred;	/* b_next chain of IGMP/MLD packets */
1661 #define	ill_last_mp_to_free	ill_mcast_deferred
1662 
1663 	cred_t	*ill_credp;		/* opener's credentials */
1664 	uint8_t	*ill_phys_addr;		/* ill_phys_addr_mp->b_rptr + off */
1665 	uint8_t *ill_dest_addr;		/* ill_dest_addr_mp->b_rptr + off */
1666 
1667 	uint_t	ill_state_flags;	/* see ILL_* flags above */
1668 
1669 	/* Following bit fields protected by ipsq_t */
1670 	uint_t
1671 		ill_needs_attach : 1,
1672 		ill_reserved : 1,
1673 		ill_isv6 : 1,
1674 		ill_dlpi_style_set : 1,
1675 
1676 		ill_ifname_pending : 1,
1677 		ill_logical_down : 1,
1678 		ill_dl_up : 1,
1679 		ill_up_ipifs : 1,
1680 
1681 		ill_note_link : 1,	/* supports link-up notification */
1682 		ill_capab_reneg : 1, /* capability renegotiation to be done */
1683 		ill_dld_capab_inprog : 1, /* direct dld capab call in prog */
1684 		ill_need_recover_multicast : 1,
1685 
1686 		ill_replumbing : 1,
1687 		ill_arl_dlpi_pending : 1,
1688 		ill_grp_pending : 1,
1689 
1690 		ill_pad_to_bit_31 : 17;
1691 
1692 	/* Following bit fields protected by ill_lock */
1693 	uint_t
1694 		ill_fragtimer_executing : 1,
1695 		ill_fragtimer_needrestart : 1,
1696 		ill_manual_token : 1,	/* system won't override ill_token */
1697 		/*
1698 		 * ill_manual_linklocal : system will not change the
1699 		 * linklocal whenever ill_token changes.
1700 		 */
1701 		ill_manual_linklocal : 1,
1702 
1703 		ill_manual_dst_linklocal : 1, /* same for pt-pt dst linklocal */
1704 
1705 		ill_pad_bit_31 : 27;
1706 
1707 	/*
1708 	 * Used in SIOCSIFMUXID and SIOCGIFMUXID for 'ifconfig unplumb'.
1709 	 */
1710 	int	ill_muxid;		/* muxid returned from plink */
1711 
1712 	/* Used for IP frag reassembly throttling on a per ILL basis.  */
1713 	uint_t	ill_ipf_gen;		/* Generation of next fragment queue */
1714 	uint_t	ill_frag_count;		/* Count of all reassembly mblk bytes */
1715 	uint_t	ill_frag_free_num_pkts;	 /* num of fragmented packets to free */
1716 	clock_t	ill_last_frag_clean_time; /* time when frag's were pruned */
1717 	int	ill_type;		/* From <net/if_types.h> */
1718 	uint_t	ill_dlpi_multicast_state;	/* See below IDS_* */
1719 	uint_t	ill_dlpi_fastpath_state;	/* See below IDS_* */
1720 
1721 	/*
1722 	 * Capabilities related fields.
1723 	 */
1724 	uint_t  ill_dlpi_capab_state;	/* State of capability query, IDCS_* */
1725 	uint_t	ill_capab_pending_cnt;
1726 	uint64_t ill_capabilities;	/* Enabled capabilities, ILL_CAPAB_* */
1727 	ill_hcksum_capab_t *ill_hcksum_capab; /* H/W cksumming capabilities */
1728 	ill_zerocopy_capab_t *ill_zerocopy_capab; /* Zero-copy capabilities */
1729 	ill_dld_capab_t *ill_dld_capab; /* DLD capabilities */
1730 	ill_lso_capab_t	*ill_lso_capab;	/* Large Segment Offload capabilities */
1731 	mblk_t	*ill_capab_reset_mp;	/* Preallocated mblk for capab reset */
1732 
1733 	uint8_t	ill_max_hops;	/* Maximum hops for any logical interface */
1734 	uint_t	ill_user_mtu;	/* User-specified MTU via SIOCSLIFLNKINFO */
1735 	uint32_t ill_reachable_time;	/* Value for ND algorithm in msec */
1736 	uint32_t ill_reachable_retrans_time; /* Value for ND algorithm msec */
1737 	uint_t	ill_max_buf;		/* Max # of req to buffer for ND */
1738 	in6_addr_t	ill_token;	/* IPv6 interface id */
1739 	in6_addr_t	ill_dest_token;	/* Destination IPv6 interface id */
1740 	uint_t		ill_token_length;
1741 	uint32_t	ill_xmit_count;		/* ndp max multicast xmits */
1742 	mib2_ipIfStatsEntry_t	*ill_ip_mib;	/* ver indep. interface mib */
1743 	mib2_ipv6IfIcmpEntry_t	*ill_icmp6_mib;	/* Per interface mib */
1744 
1745 	phyint_t		*ill_phyint;
1746 	uint64_t		ill_flags;
1747 
1748 	kmutex_t	ill_lock;	/* Please see table below */
1749 	/*
1750 	 * The ill_nd_lla* fields handle the link layer address option
1751 	 * from neighbor discovery. This is used for external IPv6
1752 	 * address resolution.
1753 	 */
1754 	mblk_t		*ill_nd_lla_mp;	/* mblk which holds ill_nd_lla */
1755 	uint8_t		*ill_nd_lla;	/* Link Layer Address */
1756 	uint_t		ill_nd_lla_len;	/* Link Layer Address length */
1757 	/*
1758 	 * We have 4 phys_addr_req's sent down. This field keeps track
1759 	 * of which one is pending.
1760 	 */
1761 	t_uscalar_t	ill_phys_addr_pend; /* which dl_phys_addr_req pending */
1762 	/*
1763 	 * Used to save errors that occur during plumbing
1764 	 */
1765 	uint_t		ill_ifname_pending_err;
1766 	avl_node_t	ill_avl_byppa; /* avl node based on ppa */
1767 	list_t		ill_nce; /* pointer to nce_s list */
1768 	uint_t		ill_refcnt;	/* active refcnt by threads */
1769 	uint_t		ill_ire_cnt;	/* ires associated with this ill */
1770 	kcondvar_t	ill_cv;
1771 	uint_t		ill_ncec_cnt;	/* ncecs associated with this ill */
1772 	uint_t		ill_nce_cnt;	/* nces associated with this ill */
1773 	uint_t		ill_waiters;	/* threads waiting in ipsq_enter */
1774 	/*
1775 	 * Contains the upper read queue pointer of the module immediately
1776 	 * beneath IP.  This field allows IP to validate sub-capability
1777 	 * acknowledgments coming up from downstream.
1778 	 */
1779 	queue_t		*ill_lmod_rq;	/* read queue pointer of module below */
1780 	uint_t		ill_lmod_cnt;	/* number of modules beneath IP */
1781 	ip_m_t		*ill_media;	/* media specific params/functions */
1782 	t_uscalar_t	ill_dlpi_pending; /* Last DLPI primitive issued */
1783 	uint_t		ill_usesrc_ifindex; /* use src addr from this ILL */
1784 	struct ill_s	*ill_usesrc_grp_next; /* Next ILL in the usesrc group */
1785 	boolean_t	ill_trace_disable;	/* True when alloc fails */
1786 	zoneid_t	ill_zoneid;
1787 	ip_stack_t	*ill_ipst;	/* Corresponds to a netstack_hold */
1788 	uint32_t	ill_dhcpinit;	/* IP_DHCPINIT_IFs for ill */
1789 	void		*ill_flownotify_mh; /* Tx flow ctl, mac cb handle */
1790 	uint_t		ill_ilm_cnt;    /* ilms referencing this ill */
1791 	uint_t		ill_ipallmulti_cnt; /* ip_join_allmulti() calls */
1792 	ilm_t		*ill_ipallmulti_ilm;
1793 
1794 	mblk_t		*ill_saved_ire_mp; /* Allocated for each extra IRE */
1795 					/* with ire_ill set so they can */
1796 					/* survive the ill going down and up. */
1797 	kmutex_t	ill_saved_ire_lock; /* Protects ill_saved_ire_mp, cnt */
1798 	uint_t		ill_saved_ire_cnt;	/* # entries */
1799 	struct arl_ill_common_s    *ill_common;
1800 	ire_t		*ill_ire_multicast; /* IRE_MULTICAST for ill */
1801 	clock_t		ill_defend_start;   /* start of 1 hour period */
1802 	uint_t		ill_defend_count;   /* # of announce/defends per ill */
1803 	/*
1804 	 * IPMP fields.
1805 	 */
1806 	ipmp_illgrp_t	*ill_grp;	/* IPMP group information */
1807 	list_node_t	ill_actnode; 	/* next active ill in group */
1808 	list_node_t	ill_grpnode;	/* next ill in group */
1809 	ipif_t		*ill_src_ipif;	/* source address selection rotor */
1810 	ipif_t		*ill_move_ipif;	/* ipif awaiting move to new ill */
1811 	boolean_t	ill_nom_cast;	/* nominated for mcast/bcast */
1812 	uint_t		ill_bound_cnt;	/* # of data addresses bound to ill */
1813 	ipif_t		*ill_bound_ipif; /* ipif chain bound to ill */
1814 	timeout_id_t	ill_refresh_tid; /* ill refresh retry timeout id */
1815 
1816 	uint32_t	ill_mrouter_cnt; /* mrouter allmulti joins */
1817 	uint32_t	ill_allowed_ips_cnt;
1818 	in6_addr_t	*ill_allowed_ips;
1819 
1820 	/* list of multicast physical addresses joined on this ill */
1821 	multiphysaddr_t *ill_mphysaddr_list;
1822 } ill_t;
1823 
1824 /*
1825  * ILL_FREE_OK() means that there are no incoming pointer references
1826  * to the ill.
1827  */
1828 #define	ILL_FREE_OK(ill)					\
1829 	((ill)->ill_ire_cnt == 0 && (ill)->ill_ilm_cnt == 0 &&	\
1830 	(ill)->ill_ncec_cnt == 0 && (ill)->ill_nce_cnt == 0)
1831 
1832 /*
1833  * An ipif/ill can be marked down only when the ire and ncec references
1834  * to that ipif/ill goes to zero. ILL_DOWN_OK() is a necessary condition
1835  * quiescence checks. See comments above IPIF_DOWN_OK for details
1836  * on why ires and nces are selectively considered for this macro.
1837  */
1838 #define	ILL_DOWN_OK(ill)					\
1839 	(ill->ill_ire_cnt == 0 && ill->ill_ncec_cnt == 0 &&	\
1840 	ill->ill_nce_cnt == 0)
1841 
1842 /*
1843  * The following table lists the protection levels of the various members
1844  * of the ill_t. Same notation as that used for ipif_t above is used.
1845  *
1846  *				Write			Read
1847  *
1848  * ill_ifptr			ill_g_lock + s		Write once
1849  * ill_rq			ipsq			Write once
1850  * ill_wq			ipsq			Write once
1851  *
1852  * ill_error			ipsq			None
1853  * ill_ipif			ill_g_lock + ipsq	ill_g_lock OR ipsq
1854  * ill_ipif_up_count		ill_lock + ipsq		ill_lock OR ipsq
1855  * ill_max_frag			ill_lock		ill_lock
1856  * ill_current_frag		ill_lock		ill_lock
1857  *
1858  * ill_name			ill_g_lock + ipsq	Write once
1859  * ill_name_length		ill_g_lock + ipsq	Write once
1860  * ill_ndd_name			ipsq			Write once
1861  * ill_net_type			ipsq			Write once
1862  * ill_ppa			ill_g_lock + ipsq	Write once
1863  * ill_sap			ipsq + down ill		Write once
1864  * ill_sap_length		ipsq + down ill		Write once
1865  * ill_phys_addr_length		ipsq + down ill		Write once
1866  *
1867  * ill_bcast_addr_length	ipsq			ipsq
1868  * ill_mactype			ipsq			ipsq
1869  * ill_frag_ptr			ipsq			ipsq
1870  *
1871  * ill_frag_timer_id		ill_lock		ill_lock
1872  * ill_frag_hash_tbl		ipsq			up ill
1873  * ill_ilm			ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1874  * ill_global_timer		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1875  * ill_mcast_type		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1876  * ill_mcast_v1_time		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1877  * ill_mcast_v2_time		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1878  * ill_mcast_v1_tset		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1879  * ill_mcast_v2_tset		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1880  * ill_mcast_rv			ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1881  * ill_mcast_qi			ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1882  *
1883  * ill_down_mp			ipsq			ipsq
1884  * ill_dlpi_deferred		ill_lock		ill_lock
1885  * ill_dlpi_pending		ipsq + ill_lock		ipsq or ill_lock or
1886  *							absence of ipsq writer.
1887  * ill_phys_addr_mp		ipsq + down ill		only when ill is up
1888  * ill_mcast_deferred		ill_lock		ill_lock
1889  * ill_phys_addr		ipsq + down ill		only when ill is up
1890  * ill_dest_addr_mp		ipsq + down ill		only when ill is up
1891  * ill_dest_addr		ipsq + down ill		only when ill is up
1892  *
1893  * ill_state_flags		ill_lock		ill_lock
1894  * exclusive bit flags		ipsq_t			ipsq_t
1895  * shared bit flags		ill_lock		ill_lock
1896  *
1897  * ill_muxid			ipsq			Not atomic
1898  *
1899  * ill_ipf_gen			Not atomic
1900  * ill_frag_count		atomics			atomics
1901  * ill_type			ipsq + down ill		only when ill is up
1902  * ill_dlpi_multicast_state	ill_lock		ill_lock
1903  * ill_dlpi_fastpath_state	ill_lock		ill_lock
1904  * ill_dlpi_capab_state		ipsq			ipsq
1905  * ill_max_hops			ipsq			Not atomic
1906  *
1907  * ill_mtu			ill_lock		None
1908  *
1909  * ill_user_mtu			ipsq + ill_lock		ill_lock
1910  * ill_reachable_time		ipsq + ill_lock		ill_lock
1911  * ill_reachable_retrans_time	ipsq + ill_lock		ill_lock
1912  * ill_max_buf			ipsq + ill_lock		ill_lock
1913  *
1914  * Next 2 fields need ill_lock because of the get ioctls. They should not
1915  * report partially updated results without executing in the ipsq.
1916  * ill_token			ipsq + ill_lock		ill_lock
1917  * ill_token_length		ipsq + ill_lock		ill_lock
1918  * ill_dest_token		ipsq + down ill		only when ill is up
1919  * ill_xmit_count		ipsq + down ill		write once
1920  * ill_ip6_mib			ipsq + down ill		only when ill is up
1921  * ill_icmp6_mib		ipsq + down ill		only when ill is up
1922  *
1923  * ill_phyint			ipsq, ill_g_lock, ill_lock	Any of them
1924  * ill_flags			ill_lock		ill_lock
1925  * ill_nd_lla_mp		ipsq + down ill		only when ill is up
1926  * ill_nd_lla			ipsq + down ill		only when ill is up
1927  * ill_nd_lla_len		ipsq + down ill		only when ill is up
1928  * ill_phys_addr_pend		ipsq + down ill		only when ill is up
1929  * ill_ifname_pending_err	ipsq			ipsq
1930  * ill_avl_byppa		ipsq, ill_g_lock	write once
1931  *
1932  * ill_fastpath_list		ill_lock		ill_lock
1933  * ill_refcnt			ill_lock		ill_lock
1934  * ill_ire_cnt			ill_lock		ill_lock
1935  * ill_cv			ill_lock		ill_lock
1936  * ill_ncec_cnt			ill_lock		ill_lock
1937  * ill_nce_cnt			ill_lock		ill_lock
1938  * ill_ilm_cnt			ill_lock		ill_lock
1939  * ill_src_ipif			ill_g_lock		ill_g_lock
1940  * ill_trace			ill_lock		ill_lock
1941  * ill_usesrc_grp_next		ill_g_usesrc_lock	ill_g_usesrc_lock
1942  * ill_dhcpinit			atomics			atomics
1943  * ill_flownotify_mh		write once		write once
1944  * ill_capab_pending_cnt	ipsq			ipsq
1945  * ill_ipallmulti_cnt		ill_lock		ill_lock
1946  * ill_ipallmulti_ilm		ill_lock		ill_lock
1947  * ill_saved_ire_mp		ill_saved_ire_lock	ill_saved_ire_lock
1948  * ill_saved_ire_cnt		ill_saved_ire_lock	ill_saved_ire_lock
1949  * ill_arl			???			???
1950  * ill_ire_multicast		ipsq + quiescent	none
1951  * ill_bound_ipif		ipsq			ipsq
1952  * ill_actnode			ipsq + ipmp_lock	ipsq OR ipmp_lock
1953  * ill_grpnode			ipsq + ill_g_lock	ipsq OR ill_g_lock
1954  * ill_src_ipif			ill_g_lock		ill_g_lock
1955  * ill_move_ipif		ipsq			ipsq
1956  * ill_nom_cast			ipsq			ipsq OR advisory
1957  * ill_refresh_tid		ill_lock		ill_lock
1958  * ill_grp (for IPMP ill)	write once		write once
1959  * ill_grp (for underlying ill)	ipsq + ill_g_lock	ipsq OR ill_g_lock
1960  * ill_grp_pending		ill_mcast_serializer	ill_mcast_serializer
1961  * ill_mrouter_cnt		atomics			atomics
1962  * ill_mphysaddr_list	ill_lock		ill_lock
1963  *
1964  * NOTE: It's OK to make heuristic decisions on an underlying interface
1965  *	 by using IS_UNDER_IPMP() or comparing ill_grp's raw pointer value.
1966  */
1967 
1968 /*
1969  * For ioctl restart mechanism see ip_reprocess_ioctl()
1970  */
1971 struct ip_ioctl_cmd_s;
1972 
1973 typedef	int (*ifunc_t)(ipif_t *, struct sockaddr_in *, queue_t *, mblk_t *,
1974     struct ip_ioctl_cmd_s *, void *);
1975 
1976 typedef struct ip_ioctl_cmd_s {
1977 	int	ipi_cmd;
1978 	size_t	ipi_copyin_size;
1979 	uint_t	ipi_flags;
1980 	uint_t	ipi_cmd_type;
1981 	ifunc_t	ipi_func;
1982 	ifunc_t	ipi_func_restart;
1983 } ip_ioctl_cmd_t;
1984 
1985 /*
1986  * ipi_cmd_type:
1987  *
1988  * IF_CMD		1	old style ifreq cmd
1989  * LIF_CMD		2	new style lifreq cmd
1990  * ARP_CMD		3	arpreq cmd
1991  * XARP_CMD		4	xarpreq cmd
1992  * MSFILT_CMD		5	multicast source filter cmd
1993  * MISC_CMD		6	misc cmd (not a more specific one above)
1994  */
1995 
1996 enum { IF_CMD = 1, LIF_CMD, ARP_CMD, XARP_CMD, MSFILT_CMD, MISC_CMD };
1997 
1998 #define	IPI_DONTCARE	0	/* For ioctl encoded values that don't matter */
1999 
2000 /* Flag values in ipi_flags */
2001 #define	IPI_PRIV	0x1	/* Root only command */
2002 #define	IPI_MODOK	0x2	/* Permitted on mod instance of IP */
2003 #define	IPI_WR		0x4	/* Need to grab writer access */
2004 #define	IPI_GET_CMD	0x8	/* branch to mi_copyout on success */
2005 /*	unused		0x10	*/
2006 #define	IPI_NULL_BCONT	0x20	/* ioctl has not data and hence no b_cont */
2007 
2008 extern ip_ioctl_cmd_t	ip_ndx_ioctl_table[];
2009 extern ip_ioctl_cmd_t	ip_misc_ioctl_table[];
2010 extern int ip_ndx_ioctl_count;
2011 extern int ip_misc_ioctl_count;
2012 
2013 /* Passed down by ARP to IP during I_PLINK/I_PUNLINK */
2014 typedef struct ipmx_s {
2015 	char	ipmx_name[LIFNAMSIZ];		/* if name */
2016 	uint_t
2017 		ipmx_arpdev_stream : 1,		/* This is the arp stream */
2018 		ipmx_notused : 31;
2019 } ipmx_t;
2020 
2021 /*
2022  * State for detecting if a driver supports certain features.
2023  * Support for DL_ENABMULTI_REQ uses ill_dlpi_multicast_state.
2024  * Support for DLPI M_DATA fastpath uses ill_dlpi_fastpath_state.
2025  */
2026 #define	IDS_UNKNOWN	0	/* No DLPI request sent */
2027 #define	IDS_INPROGRESS	1	/* DLPI request sent */
2028 #define	IDS_OK		2	/* DLPI request completed successfully */
2029 #define	IDS_FAILED	3	/* DLPI request failed */
2030 
2031 /* Support for DL_CAPABILITY_REQ uses ill_dlpi_capab_state. */
2032 enum {
2033 	IDCS_UNKNOWN,
2034 	IDCS_PROBE_SENT,
2035 	IDCS_OK,
2036 	IDCS_RESET_SENT,
2037 	IDCS_RENEG,
2038 	IDCS_FAILED
2039 };
2040 
2041 /* Extended NDP Management Structure */
2042 typedef struct ipndp_s {
2043 	ndgetf_t	ip_ndp_getf;
2044 	ndsetf_t	ip_ndp_setf;
2045 	caddr_t		ip_ndp_data;
2046 	char		*ip_ndp_name;
2047 } ipndp_t;
2048 
2049 /* IXA Notification types */
2050 typedef enum {
2051 	IXAN_LSO,	/* LSO capability change */
2052 	IXAN_PMTU,	/* PMTU change */
2053 	IXAN_ZCOPY	/* ZEROCOPY capability change */
2054 } ixa_notify_type_t;
2055 
2056 typedef uint_t ixa_notify_arg_t;
2057 
2058 typedef	void	(*ixa_notify_t)(void *, ip_xmit_attr_t *ixa, ixa_notify_type_t,
2059     ixa_notify_arg_t);
2060 
2061 /*
2062  * Attribute flags that are common to the transmit and receive attributes
2063  */
2064 #define	IAF_IS_IPV4		0x80000000	/* ipsec_*_v4 */
2065 #define	IAF_TRUSTED_ICMP	0x40000000	/* ipsec_*_icmp_loopback */
2066 #define	IAF_NO_LOOP_ZONEID_SET	0x20000000	/* Zone that shouldn't have */
2067 						/* a copy */
2068 #define	IAF_LOOPBACK_COPY	0x10000000	/* For multi and broadcast */
2069 
2070 #define	IAF_MASK		0xf0000000	/* Flags that are common */
2071 
2072 /*
2073  * Transmit side attributes used between the transport protocols and IP as
2074  * well as inside IP. It is also used to cache information in the conn_t i.e.
2075  * replaces conn_ire and the IPsec caching in the conn_t.
2076  */
2077 struct ip_xmit_attr_s {
2078 	iaflags_t	ixa_flags;	/* IXAF_*. See below */
2079 
2080 	uint32_t	ixa_free_flags;	/* IXA_FREE_*. See below */
2081 	uint32_t	ixa_refcnt;	/* Using atomics */
2082 
2083 	/*
2084 	 * Always initialized independently of ixa_flags settings.
2085 	 * Used by ip_xmit so we keep them up front for cache locality.
2086 	 */
2087 	uint32_t	ixa_xmit_hint;	/* For ECMP and GLD TX ring fanout */
2088 	uint_t		ixa_pktlen;	/* Always set. For frag and stats */
2089 	zoneid_t	ixa_zoneid;	/* Assumed always set */
2090 
2091 	/* Always set for conn_ip_output(); might be stale */
2092 	/*
2093 	 * Since TCP keeps the conn_t around past the process going away
2094 	 * we need to use the "notr" (e.g, ire_refhold_notr) for ixa_ire,
2095 	 * ixa_nce, and ixa_dce.
2096 	 */
2097 	ire_t		*ixa_ire;	/* Forwarding table entry */
2098 	uint_t		ixa_ire_generation;
2099 	nce_t		*ixa_nce;	/* Neighbor cache entry */
2100 	dce_t		*ixa_dce;	/* Destination cache entry */
2101 	uint_t		ixa_dce_generation;
2102 	uint_t		ixa_src_generation;	/* If IXAF_VERIFY_SOURCE */
2103 
2104 	uint32_t	ixa_src_preferences;	/* prefs for src addr select */
2105 	uint32_t	ixa_pmtu;		/* IXAF_VERIFY_PMTU */
2106 
2107 	/* Set by ULP if IXAF_VERIFY_PMTU; otherwise set by IP */
2108 	uint32_t	ixa_fragsize;
2109 
2110 	int8_t		ixa_use_min_mtu;	/* IXAF_USE_MIN_MTU values */
2111 
2112 	pfirepostfrag_t	ixa_postfragfn;		/* Set internally in IP */
2113 
2114 	in6_addr_t	ixa_nexthop_v6;		/* IXAF_NEXTHOP_SET */
2115 #define	ixa_nexthop_v4	V4_PART_OF_V6(ixa_nexthop_v6)
2116 
2117 	zoneid_t	ixa_no_loop_zoneid;	/* IXAF_NO_LOOP_ZONEID_SET */
2118 
2119 	uint_t		ixa_scopeid;		/* For IPv6 link-locals */
2120 
2121 	uint_t		ixa_broadcast_ttl;	/* IXAF_BROACAST_TTL_SET */
2122 
2123 	uint_t		ixa_multicast_ttl;	/* Assumed set for multicast */
2124 	uint_t		ixa_multicast_ifindex;	/* Assumed set for multicast */
2125 	ipaddr_t	ixa_multicast_ifaddr;	/* Assumed set for multicast */
2126 
2127 	int		ixa_raw_cksum_offset;	/* If IXAF_SET_RAW_CKSUM */
2128 
2129 	uint32_t	ixa_ident;		/* For IPv6 fragment header */
2130 
2131 	uint64_t	ixa_conn_id;		/* Used by DTrace */
2132 	/*
2133 	 * Cached LSO information.
2134 	 */
2135 	ill_lso_capab_t	ixa_lso_capab;		/* Valid when IXAF_LSO_CAPAB */
2136 
2137 	uint64_t	ixa_ipsec_policy_gen;	/* Generation from iph_gen */
2138 	/*
2139 	 * The following IPsec fields are only initialized when
2140 	 * IXAF_IPSEC_SECURE is set. Otherwise they contain garbage.
2141 	 */
2142 	ipsec_latch_t	*ixa_ipsec_latch;	/* Just the ids */
2143 	struct ipsa_s 	*ixa_ipsec_ah_sa;	/* Hard reference SA for AH */
2144 	struct ipsa_s 	*ixa_ipsec_esp_sa;	/* Hard reference SA for ESP */
2145 	struct ipsec_policy_s 	*ixa_ipsec_policy; /* why are we here? */
2146 	struct ipsec_action_s	*ixa_ipsec_action; /* For reflected packets */
2147 	ipsa_ref_t	ixa_ipsec_ref[2];	/* Soft reference to SA */
2148 						/* 0: ESP, 1: AH */
2149 
2150 	/*
2151 	 * The selectors here are potentially different than the SPD rule's
2152 	 * selectors, and we need to have both available for IKEv2.
2153 	 *
2154 	 * NOTE: "Source" and "Dest" are w.r.t. outbound datagrams.  Ports can
2155 	 *	 be zero, and the protocol number is needed to make the ports
2156 	 *	 significant.
2157 	 */
2158 	uint16_t ixa_ipsec_src_port;	/* Source port number of d-gram. */
2159 	uint16_t ixa_ipsec_dst_port;	/* Destination port number of d-gram. */
2160 	uint8_t  ixa_ipsec_icmp_type;	/* ICMP type of d-gram */
2161 	uint8_t  ixa_ipsec_icmp_code;	/* ICMP code of d-gram */
2162 
2163 	sa_family_t ixa_ipsec_inaf;	/* Inner address family */
2164 #define	IXA_MAX_ADDRLEN 4	/* Max addr len. (in 32-bit words) */
2165 	uint32_t ixa_ipsec_insrc[IXA_MAX_ADDRLEN];	/* Inner src address */
2166 	uint32_t ixa_ipsec_indst[IXA_MAX_ADDRLEN];	/* Inner dest address */
2167 	uint8_t  ixa_ipsec_insrcpfx;	/* Inner source prefix */
2168 	uint8_t  ixa_ipsec_indstpfx;	/* Inner destination prefix */
2169 
2170 	uint8_t ixa_ipsec_proto;	/* IP protocol number for d-gram. */
2171 
2172 	/* Always initialized independently of ixa_flags settings */
2173 	uint_t		ixa_ifindex;	/* Assumed always set */
2174 	uint16_t	ixa_ip_hdr_length; /* Points to ULP header */
2175 	uint8_t		ixa_protocol;	/* Protocol number for ULP cksum */
2176 	ts_label_t	*ixa_tsl;	/* Always set. NULL if not TX */
2177 	ip_stack_t	*ixa_ipst;	/* Always set */
2178 	uint32_t	ixa_extra_ident; /* Set if LSO */
2179 	cred_t		*ixa_cred;	/* For getpeerucred */
2180 	pid_t		ixa_cpid;	/* For getpeerucred */
2181 
2182 #ifdef DEBUG
2183 	kthread_t	*ixa_curthread;	/* For serialization assert */
2184 #endif
2185 	squeue_t	*ixa_sqp;	/* Set from conn_sqp as a hint */
2186 	uintptr_t	ixa_cookie;	/* cookie to use for tx flow control */
2187 
2188 	/*
2189 	 * Must be set by ULP if any of IXAF_VERIFY_LSO, IXAF_VERIFY_PMTU,
2190 	 * or IXAF_VERIFY_ZCOPY is set.
2191 	 */
2192 	ixa_notify_t	ixa_notify;	/* Registered upcall notify function */
2193 	void		*ixa_notify_cookie; /* ULP cookie for ixa_notify */
2194 };
2195 
2196 /*
2197  * Flags to indicate which transmit attributes are set.
2198  * Split into "xxx_SET" ones which indicate that the "xxx" field it set, and
2199  * single flags.
2200  */
2201 #define	IXAF_REACH_CONF		0x00000001	/* Reachability confirmation */
2202 #define	IXAF_BROADCAST_TTL_SET	0x00000002	/* ixa_broadcast_ttl valid */
2203 #define	IXAF_SET_SOURCE		0x00000004	/* Replace if broadcast */
2204 #define	IXAF_USE_MIN_MTU	0x00000008	/* IPV6_USE_MIN_MTU */
2205 
2206 #define	IXAF_DONTFRAG		0x00000010	/* IP*_DONTFRAG */
2207 #define	IXAF_VERIFY_PMTU	0x00000020	/* ixa_pmtu/ixa_fragsize set */
2208 #define	IXAF_PMTU_DISCOVERY	0x00000040	/* Create/use PMTU state */
2209 #define	IXAF_MULTICAST_LOOP	0x00000080	/* IP_MULTICAST_LOOP */
2210 
2211 #define	IXAF_IPSEC_SECURE	0x00000100	/* Need IPsec processing */
2212 #define	IXAF_UCRED_TSL		0x00000200	/* ixa_tsl from SCM_UCRED */
2213 #define	IXAF_DONTROUTE		0x00000400	/* SO_DONTROUTE */
2214 #define	IXAF_NO_IPSEC		0x00000800	/* Ignore policy */
2215 
2216 #define	IXAF_PMTU_TOO_SMALL	0x00001000	/* PMTU too small */
2217 #define	IXAF_SET_ULP_CKSUM	0x00002000	/* Calculate ULP checksum */
2218 #define	IXAF_VERIFY_SOURCE	0x00004000	/* Check that source is ok */
2219 #define	IXAF_NEXTHOP_SET	0x00008000	/* ixa_nexthop set */
2220 
2221 #define	IXAF_PMTU_IPV4_DF	0x00010000	/* Set IPv4 DF */
2222 #define	IXAF_NO_DEV_FLOW_CTL	0x00020000	/* Protocol needs no flow ctl */
2223 #define	IXAF_NO_TTL_CHANGE	0x00040000	/* Internal to IP */
2224 #define	IXAF_IPV6_ADD_FRAGHDR	0x00080000	/* Add fragment header */
2225 
2226 #define	IXAF_IPSEC_TUNNEL	0x00100000	/* Tunnel mode */
2227 #define	IXAF_NO_PFHOOK		0x00200000	/* Skip xmit pfhook */
2228 #define	IXAF_NO_TRACE		0x00400000	/* When back from ARP/ND */
2229 #define	IXAF_SCOPEID_SET	0x00800000	/* ixa_scopeid set */
2230 
2231 #define	IXAF_MULTIRT_MULTICAST	0x01000000	/* MULTIRT for multicast */
2232 #define	IXAF_NO_HW_CKSUM	0x02000000	/* Force software cksum */
2233 #define	IXAF_SET_RAW_CKSUM	0x04000000	/* Use ixa_raw_cksum_offset */
2234 #define	IXAF_IPSEC_GLOBAL_POLICY 0x08000000	/* Policy came from global */
2235 
2236 /* Note the following uses bits 0x10000000 through 0x80000000 */
2237 #define	IXAF_IS_IPV4		IAF_IS_IPV4
2238 #define	IXAF_TRUSTED_ICMP	IAF_TRUSTED_ICMP
2239 #define	IXAF_NO_LOOP_ZONEID_SET	IAF_NO_LOOP_ZONEID_SET
2240 #define	IXAF_LOOPBACK_COPY	IAF_LOOPBACK_COPY
2241 
2242 /* Note: use the upper 32 bits */
2243 #define	IXAF_VERIFY_LSO		0x100000000	/* Check LSO capability */
2244 #define	IXAF_LSO_CAPAB		0x200000000	/* Capable of LSO */
2245 #define	IXAF_VERIFY_ZCOPY	0x400000000	/* Check Zero Copy capability */
2246 #define	IXAF_ZCOPY_CAPAB	0x800000000	/* Capable of ZEROCOPY */
2247 
2248 /*
2249  * The normal flags for sending packets e.g., icmp errors
2250  */
2251 #define	IXAF_BASIC_SIMPLE_V4	\
2252 	(IXAF_SET_ULP_CKSUM | IXAF_IS_IPV4 | IXAF_VERIFY_SOURCE)
2253 #define	IXAF_BASIC_SIMPLE_V6	(IXAF_SET_ULP_CKSUM | IXAF_VERIFY_SOURCE)
2254 
2255 /*
2256  * Normally these fields do not have a hold. But in some cases they do, for
2257  * instance when we've gone through ip_*_attr_to/from_mblk.
2258  * We use ixa_free_flags to indicate that they have a hold and need to be
2259  * released on cleanup.
2260  */
2261 #define	IXA_FREE_CRED		0x00000001	/* ixa_cred needs to be rele */
2262 #define	IXA_FREE_TSL		0x00000002	/* ixa_tsl needs to be rele */
2263 
2264 /*
2265  * Simplistic way to set the ixa_xmit_hint for locally generated traffic
2266  * and forwarded traffic. The shift amount are based on the size of the
2267  * structs to discard the low order bits which don't have much if any variation
2268  * (coloring in kmem_cache_alloc might provide some variation).
2269  *
2270  * Basing the locally generated hint on the address of the conn_t means that
2271  * the packets from the same socket/connection do not get reordered.
2272  * Basing the hint for forwarded traffic on the ill_ring_t means that
2273  * packets from the same NIC+ring are likely to use the same outbound ring
2274  * hence we get low contention on the ring in the transmitting driver.
2275  */
2276 #define	CONN_TO_XMIT_HINT(connp)	((uint32_t)(((uintptr_t)connp) >> 11))
2277 #define	ILL_RING_TO_XMIT_HINT(ring)	((uint32_t)(((uintptr_t)ring) >> 7))
2278 
2279 /*
2280  * IP set Destination Flags used by function ip_set_destination,
2281  * ip_attr_connect, and conn_connect.
2282  */
2283 #define	IPDF_ALLOW_MCBC		0x1	/* Allow multi/broadcast */
2284 #define	IPDF_VERIFY_DST		0x2	/* Verify destination addr */
2285 #define	IPDF_SELECT_SRC		0x4	/* Select source address */
2286 #define	IPDF_LSO		0x8	/* Try LSO */
2287 #define	IPDF_IPSEC		0x10	/* Set IPsec policy */
2288 #define	IPDF_ZONE_IS_GLOBAL	0x20	/* From conn_zone_is_global */
2289 #define	IPDF_ZCOPY		0x40	/* Try ZEROCOPY */
2290 #define	IPDF_UNIQUE_DCE		0x80	/* Get a per-destination DCE */
2291 
2292 /*
2293  * Receive side attributes used between the transport protocols and IP as
2294  * well as inside IP.
2295  */
2296 struct ip_recv_attr_s {
2297 	iaflags_t	ira_flags;	/* See below */
2298 
2299 	uint32_t	ira_free_flags;	/* IRA_FREE_*. See below */
2300 
2301 	/*
2302 	 * This is a hint for TCP SYN packets.
2303 	 * Always initialized independently of ira_flags settings
2304 	 */
2305 	squeue_t	*ira_sqp;
2306 	ill_rx_ring_t	*ira_ring;	/* Internal to IP */
2307 
2308 	/* For ip_accept_tcp when IRAF_TARGET_SQP is set */
2309 	squeue_t	*ira_target_sqp;
2310 	mblk_t		*ira_target_sqp_mp;
2311 
2312 	/* Always initialized independently of ira_flags settings */
2313 	uint32_t	ira_xmit_hint;	/* For ECMP and GLD TX ring fanout */
2314 	zoneid_t	ira_zoneid;	/* ALL_ZONES unless local delivery */
2315 	uint_t		ira_pktlen;	/* Always set. For frag and stats */
2316 	uint16_t	ira_ip_hdr_length; /* Points to ULP header */
2317 	uint8_t		ira_protocol;	/* Protocol number for ULP cksum */
2318 	uint_t		ira_rifindex;	/* Received ifindex */
2319 	uint_t		ira_ruifindex;	/* Received upper ifindex */
2320 	ts_label_t	*ira_tsl;	/* Always set. NULL if not TX */
2321 	/*
2322 	 * ira_rill and ira_ill is set inside IP, but not when conn_recv is
2323 	 * called; ULPs should use ira_ruifindex instead.
2324 	 */
2325 	ill_t		*ira_rill;	/* ill where packet came */
2326 	ill_t		*ira_ill;	/* ill where IP address hosted */
2327 	cred_t		*ira_cred;	/* For getpeerucred */
2328 	pid_t		ira_cpid;	/* For getpeerucred */
2329 
2330 	/* Used when IRAF_VERIFIED_SRC is set; this source was ok */
2331 	ipaddr_t	ira_verified_src;
2332 
2333 	/*
2334 	 * The following IPsec fields are only initialized when
2335 	 * IRAF_IPSEC_SECURE is set. Otherwise they contain garbage.
2336 	 */
2337 	struct ipsec_action_s *ira_ipsec_action; /* how we made it in.. */
2338 	struct ipsa_s 	*ira_ipsec_ah_sa;	/* SA for AH */
2339 	struct ipsa_s 	*ira_ipsec_esp_sa;	/* SA for ESP */
2340 
2341 	ipaddr_t	ira_mroute_tunnel;	/* IRAF_MROUTE_TUNNEL_SET */
2342 
2343 	zoneid_t	ira_no_loop_zoneid;	/* IRAF_NO_LOOP_ZONEID_SET */
2344 
2345 	uint32_t	ira_esp_udp_ports;	/* IRAF_ESP_UDP_PORTS */
2346 
2347 	/*
2348 	 * For IP_RECVSLLA and ip_ndp_conflict/find_solicitation.
2349 	 * Same size as max for sockaddr_dl
2350 	 */
2351 #define	IRA_L2SRC_SIZE	244
2352 	uint8_t		ira_l2src[IRA_L2SRC_SIZE];	/* If IRAF_L2SRC_SET */
2353 
2354 	/*
2355 	 * Local handle that we use to do lazy setting of ira_l2src.
2356 	 * We defer setting l2src until needed but we do before any
2357 	 * ip_input pullupmsg or copymsg.
2358 	 */
2359 	struct mac_header_info_s *ira_mhip;	/* Could be NULL */
2360 };
2361 
2362 /*
2363  * Flags to indicate which receive attributes are set.
2364  */
2365 #define	IRAF_SYSTEM_LABELED	0x00000001	/* is_system_labeled() */
2366 #define	IRAF_IPV4_OPTIONS	0x00000002	/* Performance */
2367 #define	IRAF_MULTICAST		0x00000004	/* Was multicast at L3 */
2368 #define	IRAF_BROADCAST		0x00000008	/* Was broadcast at L3 */
2369 #define	IRAF_MULTIBROADCAST	(IRAF_MULTICAST|IRAF_BROADCAST)
2370 
2371 #define	IRAF_LOOPBACK		0x00000010	/* Looped back by IP */
2372 #define	IRAF_VERIFY_IP_CKSUM	0x00000020	/* Need to verify IP */
2373 #define	IRAF_VERIFY_ULP_CKSUM	0x00000040	/* Need to verify TCP,UDP,etc */
2374 #define	IRAF_SCTP_CSUM_ERR	0x00000080	/* sctp pkt has failed chksum */
2375 
2376 #define	IRAF_IPSEC_SECURE	0x00000100	/* Passed AH and/or ESP */
2377 #define	IRAF_DHCP_UNICAST	0x00000200
2378 #define	IRAF_IPSEC_DECAPS	0x00000400	/* Was packet decapsulated */
2379 					/* from a matching inner packet? */
2380 #define	IRAF_TARGET_SQP		0x00000800	/* ira_target_sqp is set */
2381 #define	IRAF_VERIFIED_SRC	0x00001000	/* ira_verified_src set */
2382 #define	IRAF_RSVP		0x00002000	/* RSVP packet for rsvpd */
2383 #define	IRAF_MROUTE_TUNNEL_SET	0x00004000	/* From ip_mroute_decap */
2384 #define	IRAF_PIM_REGISTER	0x00008000	/* From register_mforward */
2385 
2386 #define	IRAF_TX_MAC_EXEMPTABLE	0x00010000	/* Allow MAC_EXEMPT readdown */
2387 #define	IRAF_TX_SHARED_ADDR	0x00020000	/* Arrived on ALL_ZONES addr */
2388 #define	IRAF_ESP_UDP_PORTS	0x00040000	/* NAT-traversal packet */
2389 #define	IRAF_NO_HW_CKSUM	0x00080000	/* Force software cksum */
2390 
2391 #define	IRAF_ICMP_ERROR		0x00100000	/* Send to conn_recvicmp */
2392 #define	IRAF_ROUTER_ALERT	0x00200000	/* IPv6 router alert */
2393 #define	IRAF_L2SRC_SET		0x00400000	/* ira_l2src has been set */
2394 #define	IRAF_L2SRC_LOOPBACK	0x00800000	/* Came from us */
2395 
2396 #define	IRAF_L2DST_MULTICAST	0x01000000	/* Multicast at L2 */
2397 #define	IRAF_L2DST_BROADCAST	0x02000000	/* Broadcast at L2 */
2398 /* Unused 0x04000000 */
2399 /* Unused 0x08000000 */
2400 
2401 /* Below starts with 0x10000000 */
2402 #define	IRAF_IS_IPV4		IAF_IS_IPV4
2403 #define	IRAF_TRUSTED_ICMP	IAF_TRUSTED_ICMP
2404 #define	IRAF_NO_LOOP_ZONEID_SET	IAF_NO_LOOP_ZONEID_SET
2405 #define	IRAF_LOOPBACK_COPY	IAF_LOOPBACK_COPY
2406 
2407 /*
2408  * Normally these fields do not have a hold. But in some cases they do, for
2409  * instance when we've gone through ip_*_attr_to/from_mblk.
2410  * We use ira_free_flags to indicate that they have a hold and need to be
2411  * released on cleanup.
2412  */
2413 #define	IRA_FREE_CRED		0x00000001	/* ira_cred needs to be rele */
2414 #define	IRA_FREE_TSL		0x00000002	/* ira_tsl needs to be rele */
2415 
2416 /*
2417  * Optional destination cache entry for path MTU information,
2418  * and ULP metrics.
2419  */
2420 struct dce_s {
2421 	uint_t		dce_generation;	/* Changed since cached? */
2422 	uint_t		dce_flags;	/* See below */
2423 	uint_t		dce_ipversion;	/* IPv4/IPv6 version */
2424 	uint32_t	dce_pmtu;	/* Path MTU if DCEF_PMTU */
2425 	uint32_t	dce_ident;	/* Per destination IP ident. */
2426 	iulp_t		dce_uinfo;	/* Metrics if DCEF_UINFO */
2427 
2428 	struct dce_s	*dce_next;
2429 	struct dce_s	**dce_ptpn;
2430 	struct dcb_s	*dce_bucket;
2431 
2432 	union {
2433 		in6_addr_t	dceu_v6addr;
2434 		ipaddr_t	dceu_v4addr;
2435 	} dce_u;
2436 #define	dce_v4addr	dce_u.dceu_v4addr
2437 #define	dce_v6addr	dce_u.dceu_v6addr
2438 	/* Note that for IPv6+IPMP we use the ifindex for the upper interface */
2439 	uint_t		dce_ifindex;	/* For IPv6 link-locals */
2440 
2441 	kmutex_t	dce_lock;
2442 	uint_t		dce_refcnt;
2443 	uint64_t	dce_last_change_time;	/* Path MTU. In seconds */
2444 
2445 	ip_stack_t	*dce_ipst;	/* Does not have a netstack_hold */
2446 };
2447 
2448 /*
2449  * Values for dce_generation.
2450  *
2451  * If a DCE has DCE_GENERATION_CONDEMNED, the last dce_refrele should delete
2452  * it.
2453  *
2454  * DCE_GENERATION_VERIFY is never stored in dce_generation but it is
2455  * stored in places that cache DCE (such as ixa_dce_generation).
2456  * It is used as a signal that the cache is stale and needs to be reverified.
2457  */
2458 #define	DCE_GENERATION_CONDEMNED	0
2459 #define	DCE_GENERATION_VERIFY		1
2460 #define	DCE_GENERATION_INITIAL		2
2461 #define	DCE_IS_CONDEMNED(dce) \
2462 	((dce)->dce_generation == DCE_GENERATION_CONDEMNED)
2463 
2464 
2465 /*
2466  * Values for ips_src_generation.
2467  *
2468  * SRC_GENERATION_VERIFY is never stored in ips_src_generation but it is
2469  * stored in places that cache IREs (ixa_src_generation). It is used as a
2470  * signal that the cache is stale and needs to be reverified.
2471  */
2472 #define	SRC_GENERATION_VERIFY		0
2473 #define	SRC_GENERATION_INITIAL		1
2474 
2475 /*
2476  * The kernel stores security attributes of all gateways in a database made
2477  * up of one or more tsol_gcdb_t elements.  Each tsol_gcdb_t contains the
2478  * security-related credentials of the gateway.  More than one gateways may
2479  * share entries in the database.
2480  *
2481  * The tsol_gc_t structure represents the gateway to credential association,
2482  * and refers to an entry in the database.  One or more tsol_gc_t entities are
2483  * grouped together to form one or more tsol_gcgrp_t, each representing the
2484  * list of security attributes specific to the gateway.  A gateway may be
2485  * associated with at most one credentials group.
2486  */
2487 struct tsol_gcgrp_s;
2488 
2489 extern uchar_t	ip6opt_ls;	/* TX IPv6 enabler */
2490 
2491 /*
2492  * Gateway security credential record.
2493  */
2494 typedef struct tsol_gcdb_s {
2495 	uint_t		gcdb_refcnt;	/* reference count */
2496 	struct rtsa_s	gcdb_attr;	/* security attributes */
2497 #define	gcdb_mask	gcdb_attr.rtsa_mask
2498 #define	gcdb_doi	gcdb_attr.rtsa_doi
2499 #define	gcdb_slrange	gcdb_attr.rtsa_slrange
2500 } tsol_gcdb_t;
2501 
2502 /*
2503  * Gateway to credential association.
2504  */
2505 typedef struct tsol_gc_s {
2506 	uint_t		gc_refcnt;	/* reference count */
2507 	struct tsol_gcgrp_s *gc_grp;	/* pointer to group */
2508 	struct tsol_gc_s *gc_prev;	/* previous in list */
2509 	struct tsol_gc_s *gc_next;	/* next in list */
2510 	tsol_gcdb_t	*gc_db;		/* pointer to actual credentials */
2511 } tsol_gc_t;
2512 
2513 /*
2514  * Gateway credentials group address.
2515  */
2516 typedef struct tsol_gcgrp_addr_s {
2517 	int		ga_af;		/* address family */
2518 	in6_addr_t	ga_addr;	/* IPv4 mapped or IPv6 address */
2519 } tsol_gcgrp_addr_t;
2520 
2521 /*
2522  * Gateway credentials group.
2523  */
2524 typedef struct tsol_gcgrp_s {
2525 	uint_t		gcgrp_refcnt;	/* reference count */
2526 	krwlock_t	gcgrp_rwlock;	/* lock to protect following */
2527 	uint_t		gcgrp_count;	/* number of credentials */
2528 	tsol_gc_t	*gcgrp_head;	/* first credential in list */
2529 	tsol_gc_t	*gcgrp_tail;	/* last credential in list */
2530 	tsol_gcgrp_addr_t gcgrp_addr;	/* next-hop gateway address */
2531 } tsol_gcgrp_t;
2532 
2533 extern kmutex_t gcgrp_lock;
2534 
2535 #define	GC_REFRELE(p) {				\
2536 	ASSERT((p)->gc_grp != NULL);		\
2537 	rw_enter(&(p)->gc_grp->gcgrp_rwlock, RW_WRITER); \
2538 	ASSERT((p)->gc_refcnt > 0);		\
2539 	if (--((p)->gc_refcnt) == 0)		\
2540 		gc_inactive(p);			\
2541 	else					\
2542 		rw_exit(&(p)->gc_grp->gcgrp_rwlock); \
2543 }
2544 
2545 #define	GCGRP_REFHOLD(p) {			\
2546 	mutex_enter(&gcgrp_lock);		\
2547 	++((p)->gcgrp_refcnt);			\
2548 	ASSERT((p)->gcgrp_refcnt != 0);		\
2549 	mutex_exit(&gcgrp_lock);		\
2550 }
2551 
2552 #define	GCGRP_REFRELE(p) {			\
2553 	mutex_enter(&gcgrp_lock);		\
2554 	ASSERT((p)->gcgrp_refcnt > 0);		\
2555 	if (--((p)->gcgrp_refcnt) == 0)		\
2556 		gcgrp_inactive(p);		\
2557 	ASSERT(MUTEX_HELD(&gcgrp_lock));	\
2558 	mutex_exit(&gcgrp_lock);		\
2559 }
2560 
2561 /*
2562  * IRE gateway security attributes structure, pointed to by tsol_ire_gw_secattr
2563  */
2564 struct tsol_tnrhc;
2565 
2566 struct tsol_ire_gw_secattr_s {
2567 	kmutex_t	igsa_lock;	/* lock to protect following */
2568 	struct tsol_tnrhc *igsa_rhc;	/* host entry for gateway */
2569 	tsol_gc_t	*igsa_gc;	/* for prefix IREs */
2570 };
2571 
2572 void irb_refrele_ftable(irb_t *);
2573 
2574 extern struct kmem_cache *rt_entry_cache;
2575 
2576 typedef struct ire4 {
2577 	ipaddr_t ire4_mask;		/* Mask for matching this IRE. */
2578 	ipaddr_t ire4_addr;		/* Address this IRE represents. */
2579 	ipaddr_t ire4_gateway_addr;	/* Gateway including for IRE_ONLINK */
2580 	ipaddr_t ire4_setsrc_addr;	/* RTF_SETSRC */
2581 } ire4_t;
2582 
2583 typedef struct ire6 {
2584 	in6_addr_t ire6_mask;		/* Mask for matching this IRE. */
2585 	in6_addr_t ire6_addr;		/* Address this IRE represents. */
2586 	in6_addr_t ire6_gateway_addr;	/* Gateway including for IRE_ONLINK */
2587 	in6_addr_t ire6_setsrc_addr;	/* RTF_SETSRC */
2588 } ire6_t;
2589 
2590 typedef union ire_addr {
2591 	ire6_t	ire6_u;
2592 	ire4_t	ire4_u;
2593 } ire_addr_u_t;
2594 
2595 /*
2596  * Internet Routing Entry
2597  * When we have multiple identical IREs we logically add them by manipulating
2598  * ire_identical_ref and ire_delete first decrements
2599  * that and when it reaches 1 we know it is the last IRE.
2600  * "identical" is defined as being the same for:
2601  * ire_addr, ire_netmask, ire_gateway, ire_ill, ire_zoneid, and ire_type
2602  * For instance, multiple IRE_BROADCASTs for the same subnet number are
2603  * viewed as identical, and so are the IRE_INTERFACEs when there are
2604  * multiple logical interfaces (on the same ill) with the same subnet prefix.
2605  */
2606 struct ire_s {
2607 	struct	ire_s	*ire_next;	/* The hash chain must be first. */
2608 	struct	ire_s	**ire_ptpn;	/* Pointer to previous next. */
2609 	uint32_t	ire_refcnt;	/* Number of references */
2610 	ill_t		*ire_ill;
2611 	uint32_t	ire_identical_ref; /* IRE_INTERFACE, IRE_BROADCAST */
2612 	uchar_t		ire_ipversion;	/* IPv4/IPv6 version */
2613 	ushort_t	ire_type;	/* Type of IRE */
2614 	uint_t		ire_generation;	/* Generation including CONDEMNED */
2615 	uint_t	ire_ib_pkt_count;	/* Inbound packets for ire_addr */
2616 	uint_t	ire_ob_pkt_count;	/* Outbound packets to ire_addr */
2617 	time_t	ire_create_time;	/* Time (in secs) IRE was created. */
2618 	uint32_t	ire_flags;	/* flags related to route (RTF_*) */
2619 	/*
2620 	 * ire_testhidden is TRUE for INTERFACE IREs of IS_UNDER_IPMP(ill)
2621 	 * interfaces
2622 	 */
2623 	boolean_t	ire_testhidden;
2624 	pfirerecv_t	ire_recvfn;	/* Receive side handling */
2625 	pfiresend_t	ire_sendfn;	/* Send side handling */
2626 	pfirepostfrag_t	ire_postfragfn;	/* Bottom end of send handling */
2627 
2628 	uint_t		ire_masklen;	/* # bits in ire_mask{,_v6} */
2629 	ire_addr_u_t	ire_u;		/* IPv4/IPv6 address info. */
2630 
2631 	irb_t		*ire_bucket;	/* Hash bucket when ire_ptphn is set */
2632 	kmutex_t	ire_lock;
2633 	clock_t		ire_last_used_time;	/* For IRE_LOCAL reception */
2634 	tsol_ire_gw_secattr_t *ire_gw_secattr; /* gateway security attributes */
2635 	zoneid_t	ire_zoneid;
2636 
2637 	/*
2638 	 * Cached information of where to send packets that match this route.
2639 	 * The ire_dep_* information is used to determine when ire_nce_cache
2640 	 * needs to be updated.
2641 	 * ire_nce_cache is the fastpath for the Neighbor Cache Entry
2642 	 * for IPv6; arp info for IPv4
2643 	 * Since this is a cache setup and torn down independently of
2644 	 * applications we need to use nce_ref{rele,hold}_notr for it.
2645 	 */
2646 	nce_t		*ire_nce_cache;
2647 
2648 	/*
2649 	 * Quick check whether the ire_type and ire_masklen indicates
2650 	 * that the IRE can have ire_nce_cache set i.e., whether it is
2651 	 * IRE_ONLINK and for a single destination.
2652 	 */
2653 	boolean_t	ire_nce_capable;
2654 
2655 	/*
2656 	 * Dependency tracking so we can safely cache IRE and NCE pointers
2657 	 * in offlink and onlink IREs.
2658 	 * These are locked under the ips_ire_dep_lock rwlock. Write held
2659 	 * when modifying the linkage.
2660 	 * ire_dep_parent (Also chain towards IRE for nexthop)
2661 	 * ire_dep_parent_generation: ire_generation of ire_dep_parent
2662 	 * ire_dep_children (From parent to first child)
2663 	 * ire_dep_sib_next (linked list of siblings)
2664 	 * ire_dep_sib_ptpn (linked list of siblings)
2665 	 *
2666 	 * The parent has a ire_refhold on each child, and each child has
2667 	 * an ire_refhold on its parent.
2668 	 * Since ire_dep_parent is a cache setup and torn down independently of
2669 	 * applications we need to use ire_ref{rele,hold}_notr for it.
2670 	 */
2671 	ire_t		*ire_dep_parent;
2672 	ire_t		*ire_dep_children;
2673 	ire_t		*ire_dep_sib_next;
2674 	ire_t		**ire_dep_sib_ptpn;	/* Pointer to previous next */
2675 	uint_t		ire_dep_parent_generation;
2676 
2677 	uint_t		ire_badcnt;	/* Number of times ND_UNREACHABLE */
2678 	uint64_t	ire_last_badcnt;	/* In seconds */
2679 
2680 	/* ire_defense* and ire_last_used_time are only used on IRE_LOCALs */
2681 	uint_t		ire_defense_count;	/* number of ARP conflicts */
2682 	uint_t		ire_defense_time;	/* last time defended (secs) */
2683 
2684 	boolean_t	ire_trace_disable;	/* True when alloc fails */
2685 	ip_stack_t	*ire_ipst;	/* Does not have a netstack_hold */
2686 	iulp_t		ire_metrics;
2687 	/*
2688 	 * default and prefix routes that are added without explicitly
2689 	 * specifying the interface are termed "unbound" routes, and will
2690 	 * have ire_unbound set to true.
2691 	 */
2692 	boolean_t	ire_unbound;
2693 };
2694 
2695 /* IPv4 compatibility macros */
2696 #define	ire_mask		ire_u.ire4_u.ire4_mask
2697 #define	ire_addr		ire_u.ire4_u.ire4_addr
2698 #define	ire_gateway_addr	ire_u.ire4_u.ire4_gateway_addr
2699 #define	ire_setsrc_addr		ire_u.ire4_u.ire4_setsrc_addr
2700 
2701 #define	ire_mask_v6		ire_u.ire6_u.ire6_mask
2702 #define	ire_addr_v6		ire_u.ire6_u.ire6_addr
2703 #define	ire_gateway_addr_v6	ire_u.ire6_u.ire6_gateway_addr
2704 #define	ire_setsrc_addr_v6	ire_u.ire6_u.ire6_setsrc_addr
2705 
2706 /*
2707  * Values for ire_generation.
2708  *
2709  * If an IRE is marked with IRE_IS_CONDEMNED, the last walker of
2710  * the bucket should delete this IRE from this bucket.
2711  *
2712  * IRE_GENERATION_VERIFY is never stored in ire_generation but it is
2713  * stored in places that cache IREs (such as ixa_ire_generation and
2714  * ire_dep_parent_generation). It is used as a signal that the cache is
2715  * stale and needs to be reverified.
2716  */
2717 #define	IRE_GENERATION_CONDEMNED	0
2718 #define	IRE_GENERATION_VERIFY		1
2719 #define	IRE_GENERATION_INITIAL		2
2720 #define	IRE_IS_CONDEMNED(ire) \
2721 	((ire)->ire_generation == IRE_GENERATION_CONDEMNED)
2722 
2723 /* Convenient typedefs for sockaddrs */
2724 typedef	struct sockaddr_in	sin_t;
2725 typedef	struct sockaddr_in6	sin6_t;
2726 
2727 /* Name/Value Descriptor. */
2728 typedef struct nv_s {
2729 	uint64_t nv_value;
2730 	char	*nv_name;
2731 } nv_t;
2732 
2733 #define	ILL_FRAG_HASH(s, i) \
2734 	((ntohl(s) ^ ((i) ^ ((i) >> 8))) % ILL_FRAG_HASH_TBL_COUNT)
2735 
2736 /*
2737  * The MAX number of allowed fragmented packets per hash bucket
2738  * calculation is based on the most common mtu size of 1500. This limit
2739  * will work well for other mtu sizes as well.
2740  */
2741 #define	COMMON_IP_MTU 1500
2742 #define	MAX_FRAG_MIN 10
2743 #define	MAX_FRAG_PKTS(ipst)	\
2744 	MAX(MAX_FRAG_MIN, (2 * (ipst->ips_ip_reass_queue_bytes / \
2745 	    (COMMON_IP_MTU * ILL_FRAG_HASH_TBL_COUNT))))
2746 
2747 /*
2748  * Maximum dups allowed per packet.
2749  */
2750 extern uint_t ip_max_frag_dups;
2751 
2752 /*
2753  * Per-packet information for received packets and transmitted.
2754  * Used by the transport protocols when converting between the packet
2755  * and ancillary data and socket options.
2756  *
2757  * Note: This private data structure and related IPPF_* constant
2758  * definitions are exposed to enable compilation of some debugging tools
2759  * like lsof which use struct tcp_t in <inet/tcp.h>. This is intended to be
2760  * a temporary hack and long term alternate interfaces should be defined
2761  * to support the needs of such tools and private definitions moved to
2762  * private headers.
2763  */
2764 struct ip_pkt_s {
2765 	uint_t		ipp_fields;		/* Which fields are valid */
2766 	in6_addr_t	ipp_addr;		/* pktinfo src/dst addr */
2767 #define	ipp_addr_v4	V4_PART_OF_V6(ipp_addr)
2768 	uint_t		ipp_unicast_hops;	/* IPV6_UNICAST_HOPS, IP_TTL */
2769 	uint_t		ipp_hoplimit;		/* IPV6_HOPLIMIT */
2770 	uint_t		ipp_hopoptslen;
2771 	uint_t		ipp_rthdrdstoptslen;
2772 	uint_t		ipp_rthdrlen;
2773 	uint_t		ipp_dstoptslen;
2774 	uint_t		ipp_fraghdrlen;
2775 	ip6_hbh_t	*ipp_hopopts;
2776 	ip6_dest_t	*ipp_rthdrdstopts;
2777 	ip6_rthdr_t	*ipp_rthdr;
2778 	ip6_dest_t	*ipp_dstopts;
2779 	ip6_frag_t	*ipp_fraghdr;
2780 	uint8_t		ipp_tclass;		/* IPV6_TCLASS */
2781 	uint8_t		ipp_type_of_service;	/* IP_TOS */
2782 	uint_t		ipp_ipv4_options_len;	/* Len of IPv4 options */
2783 	uint8_t		*ipp_ipv4_options;	/* Ptr to IPv4 options */
2784 	uint_t		ipp_label_len_v4;	/* Len of TX label for IPv4 */
2785 	uint8_t		*ipp_label_v4;		/* TX label for IPv4 */
2786 	uint_t		ipp_label_len_v6;	/* Len of TX label for IPv6 */
2787 	uint8_t		*ipp_label_v6;		/* TX label for IPv6 */
2788 };
2789 typedef struct ip_pkt_s ip_pkt_t;
2790 
2791 extern void ip_pkt_free(ip_pkt_t *);	/* free storage inside ip_pkt_t */
2792 extern ipaddr_t ip_pkt_source_route_v4(const ip_pkt_t *);
2793 extern in6_addr_t *ip_pkt_source_route_v6(const ip_pkt_t *);
2794 extern int ip_pkt_copy(ip_pkt_t *, ip_pkt_t *, int);
2795 extern void ip_pkt_source_route_reverse_v4(ip_pkt_t *);
2796 
2797 /* ipp_fields values */
2798 #define	IPPF_ADDR		0x0001	/* Part of in6_pktinfo: src/dst addr */
2799 #define	IPPF_HOPLIMIT		0x0002	/* Overrides unicast and multicast */
2800 #define	IPPF_TCLASS		0x0004	/* Overrides class in sin6_flowinfo */
2801 
2802 #define	IPPF_HOPOPTS		0x0010	/* ipp_hopopts set */
2803 #define	IPPF_RTHDR		0x0020	/* ipp_rthdr set */
2804 #define	IPPF_RTHDRDSTOPTS	0x0040	/* ipp_rthdrdstopts set */
2805 #define	IPPF_DSTOPTS		0x0080	/* ipp_dstopts set */
2806 
2807 #define	IPPF_IPV4_OPTIONS	0x0100	/* ipp_ipv4_options set */
2808 #define	IPPF_LABEL_V4		0x0200	/* ipp_label_v4 set */
2809 #define	IPPF_LABEL_V6		0x0400	/* ipp_label_v6 set */
2810 
2811 #define	IPPF_FRAGHDR		0x0800	/* Used for IPsec receive side */
2812 
2813 /*
2814  * Data structure which is passed to conn_opt_get/set.
2815  * The conn_t is included even though it can be inferred from queue_t.
2816  * setsockopt and getsockopt use conn_ixa and conn_xmit_ipp. However,
2817  * when handling ancillary data we use separate ixa and ipps.
2818  */
2819 typedef struct conn_opt_arg_s {
2820 	conn_t		*coa_connp;
2821 	ip_xmit_attr_t	*coa_ixa;
2822 	ip_pkt_t	*coa_ipp;
2823 	boolean_t	coa_ancillary;	/* Ancillary data and not setsockopt */
2824 	uint_t		coa_changed;	/* See below */
2825 } conn_opt_arg_t;
2826 
2827 /*
2828  * Flags for what changed.
2829  * If we want to be more efficient in the future we can have more fine
2830  * grained flags e.g., a flag for just IP_TOS changing.
2831  * For now we either call ip_set_destination (for "route changed")
2832  * and/or conn_build_hdr_template/conn_prepend_hdr (for "header changed").
2833  */
2834 #define	COA_HEADER_CHANGED	0x0001
2835 #define	COA_ROUTE_CHANGED	0x0002
2836 #define	COA_RCVBUF_CHANGED	0x0004	/* SO_RCVBUF */
2837 #define	COA_SNDBUF_CHANGED	0x0008	/* SO_SNDBUF */
2838 #define	COA_WROFF_CHANGED	0x0010	/* Header size changed */
2839 #define	COA_ICMP_BIND_NEEDED	0x0020
2840 #define	COA_OOBINLINE_CHANGED	0x0040
2841 
2842 #define	TCP_PORTS_OFFSET	0
2843 #define	UDP_PORTS_OFFSET	0
2844 
2845 /*
2846  * lookups return the ill/ipif only if the flags are clear OR Iam writer.
2847  * ill / ipif lookup functions increment the refcnt on the ill / ipif only
2848  * after calling these macros. This ensures that the refcnt on the ipif or
2849  * ill will eventually drop down to zero.
2850  */
2851 #define	ILL_LOOKUP_FAILED	1	/* Used as error code */
2852 #define	IPIF_LOOKUP_FAILED	2	/* Used as error code */
2853 
2854 #define	ILL_CAN_LOOKUP(ill)						\
2855 	(!((ill)->ill_state_flags & ILL_CONDEMNED) ||			\
2856 	IAM_WRITER_ILL(ill))
2857 
2858 #define	ILL_IS_CONDEMNED(ill)	\
2859 	((ill)->ill_state_flags & ILL_CONDEMNED)
2860 
2861 #define	IPIF_CAN_LOOKUP(ipif)	\
2862 	(!((ipif)->ipif_state_flags & IPIF_CONDEMNED) || \
2863 	IAM_WRITER_IPIF(ipif))
2864 
2865 #define	IPIF_IS_CONDEMNED(ipif)	\
2866 	((ipif)->ipif_state_flags & IPIF_CONDEMNED)
2867 
2868 #define	IPIF_IS_CHANGING(ipif)	\
2869 	((ipif)->ipif_state_flags & IPIF_CHANGING)
2870 
2871 /* Macros used to assert that this thread is a writer */
2872 #define	IAM_WRITER_IPSQ(ipsq)	((ipsq)->ipsq_xop->ipx_writer == curthread)
2873 #define	IAM_WRITER_ILL(ill)	IAM_WRITER_IPSQ((ill)->ill_phyint->phyint_ipsq)
2874 #define	IAM_WRITER_IPIF(ipif)	IAM_WRITER_ILL((ipif)->ipif_ill)
2875 
2876 /*
2877  * Grab ill locks in the proper order. The order is highest addressed
2878  * ill is locked first.
2879  */
2880 #define	GRAB_ILL_LOCKS(ill_1, ill_2)				\
2881 {								\
2882 	if ((ill_1) > (ill_2)) {				\
2883 		if (ill_1 != NULL)				\
2884 			mutex_enter(&(ill_1)->ill_lock);	\
2885 		if (ill_2 != NULL)				\
2886 			mutex_enter(&(ill_2)->ill_lock);	\
2887 	} else {						\
2888 		if (ill_2 != NULL)				\
2889 			mutex_enter(&(ill_2)->ill_lock);	\
2890 		if (ill_1 != NULL && ill_1 != ill_2)		\
2891 			mutex_enter(&(ill_1)->ill_lock);	\
2892 	}							\
2893 }
2894 
2895 #define	RELEASE_ILL_LOCKS(ill_1, ill_2)		\
2896 {						\
2897 	if (ill_1 != NULL)			\
2898 		mutex_exit(&(ill_1)->ill_lock);	\
2899 	if (ill_2 != NULL && ill_2 != ill_1)	\
2900 		mutex_exit(&(ill_2)->ill_lock);	\
2901 }
2902 
2903 /* Get the other protocol instance ill */
2904 #define	ILL_OTHER(ill)						\
2905 	((ill)->ill_isv6 ? (ill)->ill_phyint->phyint_illv4 :	\
2906 	    (ill)->ill_phyint->phyint_illv6)
2907 
2908 /* ioctl command info: Ioctl properties extracted and stored in here */
2909 typedef struct cmd_info_s
2910 {
2911 	ipif_t  *ci_ipif;	/* ipif associated with [l]ifreq ioctl's */
2912 	sin_t	*ci_sin;	/* the sin struct passed down */
2913 	sin6_t	*ci_sin6;	/* the sin6_t struct passed down */
2914 	struct lifreq *ci_lifr;	/* the lifreq struct passed down */
2915 } cmd_info_t;
2916 
2917 extern struct kmem_cache *ire_cache;
2918 
2919 extern ipaddr_t	ip_g_all_ones;
2920 
2921 extern uint_t	ip_loopback_mtu;	/* /etc/system */
2922 extern uint_t	ip_loopback_mtuplus;
2923 extern uint_t	ip_loopback_mtu_v6plus;
2924 
2925 extern vmem_t *ip_minor_arena_sa;
2926 extern vmem_t *ip_minor_arena_la;
2927 
2928 /*
2929  * ip_g_forward controls IP forwarding.  It takes two values:
2930  *	0: IP_FORWARD_NEVER	Don't forward packets ever.
2931  *	1: IP_FORWARD_ALWAYS	Forward packets for elsewhere.
2932  *
2933  * RFC1122 says there must be a configuration switch to control forwarding,
2934  * but that the default MUST be to not forward packets ever.  Implicit
2935  * control based on configuration of multiple interfaces MUST NOT be
2936  * implemented (Section 3.1).  SunOS 4.1 did provide the "automatic" capability
2937  * and, in fact, it was the default.  That capability is now provided in the
2938  * /etc/rc2.d/S69inet script.
2939  */
2940 
2941 #define	ips_ip_respond_to_address_mask_broadcast \
2942 					ips_propinfo_tbl[0].prop_cur_bval
2943 #define	ips_ip_g_resp_to_echo_bcast	ips_propinfo_tbl[1].prop_cur_bval
2944 #define	ips_ip_g_resp_to_echo_mcast	ips_propinfo_tbl[2].prop_cur_bval
2945 #define	ips_ip_g_resp_to_timestamp	ips_propinfo_tbl[3].prop_cur_bval
2946 #define	ips_ip_g_resp_to_timestamp_bcast ips_propinfo_tbl[4].prop_cur_bval
2947 #define	ips_ip_g_send_redirects		ips_propinfo_tbl[5].prop_cur_bval
2948 #define	ips_ip_g_forward_directed_bcast	ips_propinfo_tbl[6].prop_cur_bval
2949 #define	ips_ip_mrtdebug			ips_propinfo_tbl[7].prop_cur_uval
2950 #define	ips_ip_ire_reclaim_fraction	ips_propinfo_tbl[8].prop_cur_uval
2951 #define	ips_ip_nce_reclaim_fraction	ips_propinfo_tbl[9].prop_cur_uval
2952 #define	ips_ip_dce_reclaim_fraction	ips_propinfo_tbl[10].prop_cur_uval
2953 #define	ips_ip_def_ttl			ips_propinfo_tbl[11].prop_cur_uval
2954 #define	ips_ip_forward_src_routed	ips_propinfo_tbl[12].prop_cur_bval
2955 #define	ips_ip_wroff_extra		ips_propinfo_tbl[13].prop_cur_uval
2956 #define	ips_ip_pathmtu_interval		ips_propinfo_tbl[14].prop_cur_uval
2957 #define	ips_ip_icmp_return		ips_propinfo_tbl[15].prop_cur_uval
2958 #define	ips_ip_path_mtu_discovery	ips_propinfo_tbl[16].prop_cur_bval
2959 #define	ips_ip_pmtu_min			ips_propinfo_tbl[17].prop_cur_uval
2960 #define	ips_ip_ignore_redirect		ips_propinfo_tbl[18].prop_cur_bval
2961 #define	ips_ip_arp_icmp_error		ips_propinfo_tbl[19].prop_cur_bval
2962 #define	ips_ip_broadcast_ttl		ips_propinfo_tbl[20].prop_cur_uval
2963 #define	ips_ip_icmp_err_interval	ips_propinfo_tbl[21].prop_cur_uval
2964 #define	ips_ip_icmp_err_burst		ips_propinfo_tbl[22].prop_cur_uval
2965 #define	ips_ip_reass_queue_bytes	ips_propinfo_tbl[23].prop_cur_uval
2966 #define	ips_ip_strict_dst_multihoming	ips_propinfo_tbl[24].prop_cur_uval
2967 #define	ips_ip_addrs_per_if		ips_propinfo_tbl[25].prop_cur_uval
2968 #define	ips_ipsec_override_persocket_policy ips_propinfo_tbl[26].prop_cur_bval
2969 #define	ips_icmp_accept_clear_messages	ips_propinfo_tbl[27].prop_cur_bval
2970 #define	ips_igmp_accept_clear_messages	ips_propinfo_tbl[28].prop_cur_bval
2971 
2972 /* IPv6 configuration knobs */
2973 #define	ips_delay_first_probe_time	ips_propinfo_tbl[29].prop_cur_uval
2974 #define	ips_max_unicast_solicit		ips_propinfo_tbl[30].prop_cur_uval
2975 #define	ips_ipv6_def_hops		ips_propinfo_tbl[31].prop_cur_uval
2976 #define	ips_ipv6_icmp_return		ips_propinfo_tbl[32].prop_cur_uval
2977 #define	ips_ipv6_forward_src_routed	ips_propinfo_tbl[33].prop_cur_bval
2978 #define	ips_ipv6_resp_echo_mcast	ips_propinfo_tbl[34].prop_cur_bval
2979 #define	ips_ipv6_send_redirects		ips_propinfo_tbl[35].prop_cur_bval
2980 #define	ips_ipv6_ignore_redirect	ips_propinfo_tbl[36].prop_cur_bval
2981 #define	ips_ipv6_strict_dst_multihoming	ips_propinfo_tbl[37].prop_cur_uval
2982 #define	ips_src_check			ips_propinfo_tbl[38].prop_cur_uval
2983 #define	ips_ipsec_policy_log_interval	ips_propinfo_tbl[39].prop_cur_uval
2984 #define	ips_pim_accept_clear_messages	ips_propinfo_tbl[40].prop_cur_bval
2985 #define	ips_ip_ndp_unsolicit_interval	ips_propinfo_tbl[41].prop_cur_uval
2986 #define	ips_ip_ndp_unsolicit_count	ips_propinfo_tbl[42].prop_cur_uval
2987 #define	ips_ipv6_ignore_home_address_opt ips_propinfo_tbl[43].prop_cur_bval
2988 
2989 /* Misc IP configuration knobs */
2990 #define	ips_ip_policy_mask		ips_propinfo_tbl[44].prop_cur_uval
2991 #define	ips_ip_ecmp_behavior		ips_propinfo_tbl[45].prop_cur_uval
2992 #define	ips_ip_multirt_ttl  		ips_propinfo_tbl[46].prop_cur_uval
2993 #define	ips_ip_ire_badcnt_lifetime	ips_propinfo_tbl[47].prop_cur_uval
2994 #define	ips_ip_max_temp_idle		ips_propinfo_tbl[48].prop_cur_uval
2995 #define	ips_ip_max_temp_defend		ips_propinfo_tbl[49].prop_cur_uval
2996 #define	ips_ip_max_defend		ips_propinfo_tbl[50].prop_cur_uval
2997 #define	ips_ip_defend_interval		ips_propinfo_tbl[51].prop_cur_uval
2998 #define	ips_ip_dup_recovery		ips_propinfo_tbl[52].prop_cur_uval
2999 #define	ips_ip_restrict_interzone_loopback ips_propinfo_tbl[53].prop_cur_bval
3000 #define	ips_ip_lso_outbound		ips_propinfo_tbl[54].prop_cur_bval
3001 #define	ips_igmp_max_version		ips_propinfo_tbl[55].prop_cur_uval
3002 #define	ips_mld_max_version		ips_propinfo_tbl[56].prop_cur_uval
3003 #define	ips_ip_forwarding		ips_propinfo_tbl[57].prop_cur_bval
3004 #define	ips_ipv6_forwarding		ips_propinfo_tbl[58].prop_cur_bval
3005 #define	ips_ip_reassembly_timeout	ips_propinfo_tbl[59].prop_cur_uval
3006 #define	ips_ipv6_reassembly_timeout	ips_propinfo_tbl[60].prop_cur_uval
3007 #define	ips_ip_cgtp_filter		ips_propinfo_tbl[61].prop_cur_bval
3008 #define	ips_arp_probe_delay		ips_propinfo_tbl[62].prop_cur_uval
3009 #define	ips_arp_fastprobe_delay		ips_propinfo_tbl[63].prop_cur_uval
3010 #define	ips_arp_probe_interval		ips_propinfo_tbl[64].prop_cur_uval
3011 #define	ips_arp_fastprobe_interval	ips_propinfo_tbl[65].prop_cur_uval
3012 #define	ips_arp_probe_count		ips_propinfo_tbl[66].prop_cur_uval
3013 #define	ips_arp_fastprobe_count		ips_propinfo_tbl[67].prop_cur_uval
3014 #define	ips_ipv4_dad_announce_interval	ips_propinfo_tbl[68].prop_cur_uval
3015 #define	ips_ipv6_dad_announce_interval	ips_propinfo_tbl[69].prop_cur_uval
3016 #define	ips_arp_defend_interval		ips_propinfo_tbl[70].prop_cur_uval
3017 #define	ips_arp_defend_rate		ips_propinfo_tbl[71].prop_cur_uval
3018 #define	ips_ndp_defend_interval		ips_propinfo_tbl[72].prop_cur_uval
3019 #define	ips_ndp_defend_rate		ips_propinfo_tbl[73].prop_cur_uval
3020 #define	ips_arp_defend_period		ips_propinfo_tbl[74].prop_cur_uval
3021 #define	ips_ndp_defend_period		ips_propinfo_tbl[75].prop_cur_uval
3022 #define	ips_ipv4_icmp_return_pmtu	ips_propinfo_tbl[76].prop_cur_bval
3023 #define	ips_ipv6_icmp_return_pmtu	ips_propinfo_tbl[77].prop_cur_bval
3024 #define	ips_ip_arp_publish_count	ips_propinfo_tbl[78].prop_cur_uval
3025 #define	ips_ip_arp_publish_interval	ips_propinfo_tbl[79].prop_cur_uval
3026 #define	ips_ip_strict_src_multihoming	ips_propinfo_tbl[80].prop_cur_uval
3027 #define	ips_ipv6_strict_src_multihoming	ips_propinfo_tbl[81].prop_cur_uval
3028 #define	ips_ipv6_drop_inbound_icmpv6	ips_propinfo_tbl[82].prop_cur_bval
3029 
3030 extern int	dohwcksum;	/* use h/w cksum if supported by the h/w */
3031 #ifdef ZC_TEST
3032 extern int	noswcksum;
3033 #endif
3034 
3035 extern char	ipif_loopback_name[];
3036 
3037 extern nv_t	*ire_nv_tbl;
3038 
3039 extern struct module_info ip_mod_info;
3040 
3041 #define	HOOKS4_INTERESTED_PHYSICAL_IN(ipst)	\
3042 	((ipst)->ips_ip4_physical_in_event.he_interested)
3043 #define	HOOKS6_INTERESTED_PHYSICAL_IN(ipst)	\
3044 	((ipst)->ips_ip6_physical_in_event.he_interested)
3045 #define	HOOKS4_INTERESTED_PHYSICAL_OUT(ipst)	\
3046 	((ipst)->ips_ip4_physical_out_event.he_interested)
3047 #define	HOOKS6_INTERESTED_PHYSICAL_OUT(ipst)	\
3048 	((ipst)->ips_ip6_physical_out_event.he_interested)
3049 #define	HOOKS4_INTERESTED_FORWARDING(ipst)	\
3050 	((ipst)->ips_ip4_forwarding_event.he_interested)
3051 #define	HOOKS6_INTERESTED_FORWARDING(ipst)	\
3052 	((ipst)->ips_ip6_forwarding_event.he_interested)
3053 #define	HOOKS4_INTERESTED_LOOPBACK_IN(ipst)	\
3054 	((ipst)->ips_ip4_loopback_in_event.he_interested)
3055 #define	HOOKS6_INTERESTED_LOOPBACK_IN(ipst)	\
3056 	((ipst)->ips_ip6_loopback_in_event.he_interested)
3057 #define	HOOKS4_INTERESTED_LOOPBACK_OUT(ipst)	\
3058 	((ipst)->ips_ip4_loopback_out_event.he_interested)
3059 #define	HOOKS6_INTERESTED_LOOPBACK_OUT(ipst)	\
3060 	((ipst)->ips_ip6_loopback_out_event.he_interested)
3061 /*
3062  * Hooks marcos used inside of ip
3063  * The callers use the above INTERESTED macros first, hence
3064  * the he_interested check is superflous.
3065  */
3066 #define	FW_HOOKS(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst, _err) \
3067 	if ((_hook).he_interested) {					\
3068 		hook_pkt_event_t info;					\
3069 									\
3070 		_NOTE(CONSTCOND)					\
3071 		ASSERT((_ilp != NULL) || (_olp != NULL));		\
3072 									\
3073 		FW_SET_ILL_INDEX(info.hpe_ifp, (ill_t *)_ilp);		\
3074 		FW_SET_ILL_INDEX(info.hpe_ofp, (ill_t *)_olp);		\
3075 		info.hpe_protocol = ipst->ips_ipv4_net_data;		\
3076 		info.hpe_hdr = _iph;					\
3077 		info.hpe_mp = &(_fm);					\
3078 		info.hpe_mb = _m;					\
3079 		info.hpe_flags = _llm;					\
3080 		_err = hook_run(ipst->ips_ipv4_net_data->netd_hooks,	\
3081 		    _event, (hook_data_t)&info);			\
3082 		if (_err != 0) {					\
3083 			ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
3084 			    (_hook).he_name, (void *)_fm, (void *)_m));	\
3085 			if (_fm != NULL) {				\
3086 				freemsg(_fm);				\
3087 				_fm = NULL;				\
3088 			}						\
3089 			_iph = NULL;					\
3090 			_m = NULL;					\
3091 		} else {						\
3092 			_iph = info.hpe_hdr;				\
3093 			_m = info.hpe_mb;				\
3094 		}							\
3095 	}
3096 
3097 #define	FW_HOOKS6(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst, _err) \
3098 	if ((_hook).he_interested) {					\
3099 		hook_pkt_event_t info;					\
3100 									\
3101 		_NOTE(CONSTCOND)					\
3102 		ASSERT((_ilp != NULL) || (_olp != NULL));		\
3103 									\
3104 		FW_SET_ILL_INDEX(info.hpe_ifp, (ill_t *)_ilp);		\
3105 		FW_SET_ILL_INDEX(info.hpe_ofp, (ill_t *)_olp);		\
3106 		info.hpe_protocol = ipst->ips_ipv6_net_data;		\
3107 		info.hpe_hdr = _iph;					\
3108 		info.hpe_mp = &(_fm);					\
3109 		info.hpe_mb = _m;					\
3110 		info.hpe_flags = _llm;					\
3111 		_err = hook_run(ipst->ips_ipv6_net_data->netd_hooks,	\
3112 		    _event, (hook_data_t)&info);			\
3113 		if (_err != 0) {					\
3114 			ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
3115 			    (_hook).he_name, (void *)_fm, (void *)_m));	\
3116 			if (_fm != NULL) {				\
3117 				freemsg(_fm);				\
3118 				_fm = NULL;				\
3119 			}						\
3120 			_iph = NULL;					\
3121 			_m = NULL;					\
3122 		} else {						\
3123 			_iph = info.hpe_hdr;				\
3124 			_m = info.hpe_mb;				\
3125 		}							\
3126 	}
3127 
3128 #define	FW_SET_ILL_INDEX(fp, ill)					\
3129 	_NOTE(CONSTCOND)						\
3130 	if ((ill) == NULL || (ill)->ill_phyint == NULL) {		\
3131 		(fp) = 0;						\
3132 		_NOTE(CONSTCOND)					\
3133 	} else if (IS_UNDER_IPMP(ill)) {				\
3134 		(fp) = ipmp_ill_get_ipmp_ifindex(ill);			\
3135 	} else {							\
3136 		(fp) = (ill)->ill_phyint->phyint_ifindex;		\
3137 	}
3138 
3139 /*
3140  * Network byte order macros
3141  */
3142 #ifdef	_BIG_ENDIAN
3143 #define	N_IN_CLASSA_NET		IN_CLASSA_NET
3144 #define	N_IN_CLASSD_NET		IN_CLASSD_NET
3145 #define	N_INADDR_UNSPEC_GROUP	INADDR_UNSPEC_GROUP
3146 #define	N_IN_LOOPBACK_NET	(ipaddr_t)0x7f000000U
3147 #else /* _BIG_ENDIAN */
3148 #define	N_IN_CLASSA_NET		(ipaddr_t)0x000000ffU
3149 #define	N_IN_CLASSD_NET		(ipaddr_t)0x000000f0U
3150 #define	N_INADDR_UNSPEC_GROUP	(ipaddr_t)0x000000e0U
3151 #define	N_IN_LOOPBACK_NET	(ipaddr_t)0x0000007fU
3152 #endif /* _BIG_ENDIAN */
3153 #define	CLASSD(addr)	(((addr) & N_IN_CLASSD_NET) == N_INADDR_UNSPEC_GROUP)
3154 #define	CLASSE(addr)	(((addr) & N_IN_CLASSD_NET) == N_IN_CLASSD_NET)
3155 #define	IP_LOOPBACK_ADDR(addr)			\
3156 	(((addr) & N_IN_CLASSA_NET == N_IN_LOOPBACK_NET))
3157 
3158 extern int	ip_debug;
3159 extern uint_t	ip_thread_data;
3160 extern krwlock_t ip_thread_rwlock;
3161 extern list_t	ip_thread_list;
3162 
3163 #ifdef IP_DEBUG
3164 #include <sys/debug.h>
3165 #include <sys/promif.h>
3166 
3167 #define	ip0dbg(a)	printf a
3168 #define	ip1dbg(a)	if (ip_debug > 2) printf a
3169 #define	ip2dbg(a)	if (ip_debug > 3) printf a
3170 #define	ip3dbg(a)	if (ip_debug > 4) printf a
3171 #else
3172 #define	ip0dbg(a)	/* */
3173 #define	ip1dbg(a)	/* */
3174 #define	ip2dbg(a)	/* */
3175 #define	ip3dbg(a)	/* */
3176 #endif	/* IP_DEBUG */
3177 
3178 /* Default MAC-layer address string length for mac_colon_addr */
3179 #define	MAC_STR_LEN	128
3180 
3181 struct	mac_header_info_s;
3182 
3183 extern void	ill_frag_timer(void *);
3184 extern ill_t	*ill_first(int, int, ill_walk_context_t *, ip_stack_t *);
3185 extern ill_t	*ill_next(ill_walk_context_t *, ill_t *);
3186 extern void	ill_frag_timer_start(ill_t *);
3187 extern void	ill_nic_event_dispatch(ill_t *, lif_if_t, nic_event_t,
3188     nic_event_data_t, size_t);
3189 extern mblk_t	*ip_carve_mp(mblk_t **, ssize_t);
3190 extern mblk_t	*ip_dlpi_alloc(size_t, t_uscalar_t);
3191 extern mblk_t	*ip_dlnotify_alloc(uint_t, uint_t);
3192 extern char	*ip_dot_addr(ipaddr_t, char *);
3193 extern const char *mac_colon_addr(const uint8_t *, size_t, char *, size_t);
3194 extern void	ip_lwput(queue_t *, mblk_t *);
3195 extern boolean_t icmp_err_rate_limit(ip_stack_t *);
3196 extern void	icmp_frag_needed(mblk_t *, int, ip_recv_attr_t *);
3197 extern mblk_t	*icmp_inbound_v4(mblk_t *, ip_recv_attr_t *);
3198 extern void	icmp_time_exceeded(mblk_t *, uint8_t, ip_recv_attr_t *);
3199 extern void	icmp_unreachable(mblk_t *, uint8_t, ip_recv_attr_t *);
3200 extern boolean_t ip_ipsec_policy_inherit(conn_t *, conn_t *, ip_recv_attr_t *);
3201 extern void	*ip_pullup(mblk_t *, ssize_t, ip_recv_attr_t *);
3202 extern void	ip_setl2src(mblk_t *, ip_recv_attr_t *, ill_t *);
3203 extern mblk_t	*ip_check_and_align_header(mblk_t *, uint_t, ip_recv_attr_t *);
3204 extern mblk_t	*ip_check_length(mblk_t *, uchar_t *, ssize_t, uint_t, uint_t,
3205     ip_recv_attr_t *);
3206 extern mblk_t	*ip_check_optlen(mblk_t *, ipha_t *, uint_t, uint_t,
3207     ip_recv_attr_t *);
3208 extern mblk_t	*ip_fix_dbref(mblk_t *, ip_recv_attr_t *);
3209 extern uint_t	ip_cksum(mblk_t *, int, uint32_t);
3210 extern int	ip_close(queue_t *, int);
3211 extern uint16_t	ip_csum_hdr(ipha_t *);
3212 extern void	ip_forward_xmit_v4(nce_t *, ill_t *, mblk_t *, ipha_t *,
3213     ip_recv_attr_t *, uint32_t, uint32_t);
3214 extern boolean_t ip_forward_options(mblk_t *, ipha_t *, ill_t *,
3215     ip_recv_attr_t *);
3216 extern int	ip_fragment_v4(mblk_t *, nce_t *, iaflags_t, uint_t, uint32_t,
3217     uint32_t, zoneid_t, zoneid_t, pfirepostfrag_t postfragfn,
3218     uintptr_t *cookie);
3219 extern void	ip_proto_not_sup(mblk_t *, ip_recv_attr_t *);
3220 extern void	ip_ire_g_fini(void);
3221 extern void	ip_ire_g_init(void);
3222 extern void	ip_ire_fini(ip_stack_t *);
3223 extern void	ip_ire_init(ip_stack_t *);
3224 extern void	ip_mdata_to_mhi(ill_t *, mblk_t *, struct mac_header_info_s *);
3225 extern int	ip_openv4(queue_t *q, dev_t *devp, int flag, int sflag,
3226 		    cred_t *credp);
3227 extern int	ip_openv6(queue_t *q, dev_t *devp, int flag, int sflag,
3228 		    cred_t *credp);
3229 extern int	ip_reassemble(mblk_t *, ipf_t *, uint_t, boolean_t, ill_t *,
3230     size_t);
3231 extern void	ip_rput(queue_t *, mblk_t *);
3232 extern void	ip_input(ill_t *, ill_rx_ring_t *, mblk_t *,
3233     struct mac_header_info_s *);
3234 extern void	ip_input_v6(ill_t *, ill_rx_ring_t *, mblk_t *,
3235     struct mac_header_info_s *);
3236 extern mblk_t	*ip_input_common_v4(ill_t *, ill_rx_ring_t *, mblk_t *,
3237     struct mac_header_info_s *, squeue_t *, mblk_t **, uint_t *);
3238 extern mblk_t	*ip_input_common_v6(ill_t *, ill_rx_ring_t *, mblk_t *,
3239     struct mac_header_info_s *, squeue_t *, mblk_t **, uint_t *);
3240 extern void	ill_input_full_v4(mblk_t *, void *, void *,
3241     ip_recv_attr_t *, rtc_t *);
3242 extern void	ill_input_short_v4(mblk_t *, void *, void *,
3243     ip_recv_attr_t *, rtc_t *);
3244 extern void	ill_input_full_v6(mblk_t *, void *, void *,
3245     ip_recv_attr_t *, rtc_t *);
3246 extern void	ill_input_short_v6(mblk_t *, void *, void *,
3247     ip_recv_attr_t *, rtc_t *);
3248 extern ipaddr_t	ip_input_options(ipha_t *, ipaddr_t, mblk_t *,
3249     ip_recv_attr_t *, int *);
3250 extern boolean_t ip_input_local_options(mblk_t *, ipha_t *, ip_recv_attr_t *);
3251 extern mblk_t	*ip_input_fragment(mblk_t *, ipha_t *, ip_recv_attr_t *);
3252 extern mblk_t	*ip_input_fragment_v6(mblk_t *, ip6_t *, ip6_frag_t *, uint_t,
3253     ip_recv_attr_t *);
3254 extern void	ip_input_post_ipsec(mblk_t *, ip_recv_attr_t *);
3255 extern void	ip_fanout_v4(mblk_t *, ipha_t *, ip_recv_attr_t *);
3256 extern void	ip_fanout_v6(mblk_t *, ip6_t *, ip_recv_attr_t *);
3257 extern void	ip_fanout_proto_conn(conn_t *, mblk_t *, ipha_t *, ip6_t *,
3258     ip_recv_attr_t *);
3259 extern void	ip_fanout_proto_v4(mblk_t *, ipha_t *, ip_recv_attr_t *);
3260 extern void	ip_fanout_send_icmp_v4(mblk_t *, uint_t, uint_t,
3261     ip_recv_attr_t *);
3262 extern void	ip_fanout_udp_conn(conn_t *, mblk_t *, ipha_t *, ip6_t *,
3263     ip_recv_attr_t *);
3264 extern void	ip_fanout_udp_multi_v4(mblk_t *, ipha_t *, uint16_t, uint16_t,
3265     ip_recv_attr_t *);
3266 extern mblk_t	*zero_spi_check(mblk_t *, ip_recv_attr_t *);
3267 extern void	ip_build_hdrs_v4(uchar_t *, uint_t, const ip_pkt_t *, uint8_t);
3268 extern int	ip_find_hdr_v4(ipha_t *, ip_pkt_t *, boolean_t);
3269 extern int	ip_total_hdrs_len_v4(const ip_pkt_t *);
3270 
3271 extern mblk_t	*ip_accept_tcp(ill_t *, ill_rx_ring_t *, squeue_t *,
3272     mblk_t *, mblk_t **, uint_t *cnt);
3273 extern void	ip_rput_dlpi(ill_t *, mblk_t *);
3274 extern void	ip_rput_notdata(ill_t *, mblk_t *);
3275 
3276 extern void	ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *,
3277 		    mib2_ipIfStatsEntry_t *);
3278 extern void	ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *,
3279 		    mib2_ipv6IfIcmpEntry_t *);
3280 extern void	ip_rput_other(ipsq_t *, queue_t *, mblk_t *, void *);
3281 extern ire_t	*ip_check_multihome(void *, ire_t *, ill_t *);
3282 extern void	ip_send_potential_redirect_v4(mblk_t *, ipha_t *, ire_t *,
3283     ip_recv_attr_t *);
3284 extern int	ip_set_destination_v4(ipaddr_t *, ipaddr_t, ipaddr_t,
3285     ip_xmit_attr_t *, iulp_t *, uint32_t, uint_t);
3286 extern int	ip_set_destination_v6(in6_addr_t *, const in6_addr_t *,
3287     const in6_addr_t *, ip_xmit_attr_t *, iulp_t *, uint32_t, uint_t);
3288 
3289 extern int	ip_output_simple(mblk_t *, ip_xmit_attr_t *);
3290 extern int	ip_output_simple_v4(mblk_t *, ip_xmit_attr_t *);
3291 extern int	ip_output_simple_v6(mblk_t *, ip_xmit_attr_t *);
3292 extern int	ip_output_options(mblk_t *, ipha_t *, ip_xmit_attr_t *,
3293     ill_t *);
3294 extern void	ip_output_local_options(ipha_t *, ip_stack_t *);
3295 
3296 extern ip_xmit_attr_t *conn_get_ixa(conn_t *, boolean_t);
3297 extern ip_xmit_attr_t *conn_get_ixa_tryhard(conn_t *, boolean_t);
3298 extern ip_xmit_attr_t *conn_replace_ixa(conn_t *, ip_xmit_attr_t *);
3299 extern ip_xmit_attr_t *conn_get_ixa_exclusive(conn_t *);
3300 extern ip_xmit_attr_t *ip_xmit_attr_duplicate(ip_xmit_attr_t *);
3301 extern void	ip_xmit_attr_replace_tsl(ip_xmit_attr_t *, ts_label_t *);
3302 extern void	ip_xmit_attr_restore_tsl(ip_xmit_attr_t *, cred_t *);
3303 boolean_t	ip_recv_attr_replace_label(ip_recv_attr_t *, ts_label_t *);
3304 extern void	ixa_inactive(ip_xmit_attr_t *);
3305 extern void	ixa_refrele(ip_xmit_attr_t *);
3306 extern boolean_t ixa_check_drain_insert(conn_t *, ip_xmit_attr_t *);
3307 extern void	ixa_cleanup(ip_xmit_attr_t *);
3308 extern void	ira_cleanup(ip_recv_attr_t *, boolean_t);
3309 extern void	ixa_safe_copy(ip_xmit_attr_t *, ip_xmit_attr_t *);
3310 
3311 extern int	conn_ip_output(mblk_t *, ip_xmit_attr_t *);
3312 extern boolean_t ip_output_verify_local(ip_xmit_attr_t *);
3313 extern mblk_t	*ip_output_process_local(mblk_t *, ip_xmit_attr_t *, boolean_t,
3314     boolean_t, conn_t *);
3315 
3316 extern int	conn_opt_get(conn_opt_arg_t *, t_scalar_t, t_scalar_t,
3317     uchar_t *);
3318 extern int	conn_opt_set(conn_opt_arg_t *, t_scalar_t, t_scalar_t, uint_t,
3319     uchar_t *, boolean_t, cred_t *);
3320 extern boolean_t	conn_same_as_last_v4(conn_t *, sin_t *);
3321 extern boolean_t	conn_same_as_last_v6(conn_t *, sin6_t *);
3322 extern int	conn_update_label(const conn_t *, const ip_xmit_attr_t *,
3323     const in6_addr_t *, ip_pkt_t *);
3324 
3325 extern int	ip_opt_set_multicast_group(conn_t *, t_scalar_t,
3326     uchar_t *, boolean_t, boolean_t);
3327 extern int	ip_opt_set_multicast_sources(conn_t *, t_scalar_t,
3328     uchar_t *, boolean_t, boolean_t);
3329 extern int	conn_getsockname(conn_t *, struct sockaddr *, uint_t *);
3330 extern int	conn_getpeername(conn_t *, struct sockaddr *, uint_t *);
3331 
3332 extern int	conn_build_hdr_template(conn_t *, uint_t, uint_t,
3333     const in6_addr_t *, const in6_addr_t *, uint32_t);
3334 extern mblk_t	*conn_prepend_hdr(ip_xmit_attr_t *, const ip_pkt_t *,
3335     const in6_addr_t *, const in6_addr_t *, uint8_t, uint32_t, uint_t,
3336     mblk_t *, uint_t, uint_t, uint32_t *, int *);
3337 extern void	ip_attr_newdst(ip_xmit_attr_t *);
3338 extern void	ip_attr_nexthop(const ip_pkt_t *, const ip_xmit_attr_t *,
3339     const in6_addr_t *, in6_addr_t *);
3340 extern int	conn_connect(conn_t *, iulp_t *, uint32_t);
3341 extern int	ip_attr_connect(const conn_t *, ip_xmit_attr_t *,
3342     const in6_addr_t *, const in6_addr_t *, const in6_addr_t *, in_port_t,
3343     in6_addr_t *, iulp_t *, uint32_t);
3344 extern int	conn_inherit_parent(conn_t *, conn_t *);
3345 
3346 extern void	conn_ixa_cleanup(conn_t *connp, void *arg);
3347 
3348 extern boolean_t conn_wantpacket(conn_t *, ip_recv_attr_t *, ipha_t *);
3349 extern uint_t	ip_type_v4(ipaddr_t, ip_stack_t *);
3350 extern uint_t	ip_type_v6(const in6_addr_t *, ip_stack_t *);
3351 
3352 extern void	ip_wput_nondata(queue_t *, mblk_t *);
3353 extern void	ip_wsrv(queue_t *);
3354 extern char	*ip_nv_lookup(nv_t *, int);
3355 extern boolean_t ip_local_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3356 extern boolean_t ip_remote_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3357 extern ipaddr_t ip_massage_options(ipha_t *, netstack_t *);
3358 extern ipaddr_t ip_net_mask(ipaddr_t);
3359 extern void	arp_bringup_done(ill_t *, int);
3360 extern void	arp_replumb_done(ill_t *, int);
3361 
3362 extern struct qinit iprinitv6;
3363 
3364 extern void	ipmp_init(ip_stack_t *);
3365 extern void	ipmp_destroy(ip_stack_t *);
3366 extern ipmp_grp_t *ipmp_grp_create(const char *, phyint_t *);
3367 extern void	ipmp_grp_destroy(ipmp_grp_t *);
3368 extern void	ipmp_grp_info(const ipmp_grp_t *, lifgroupinfo_t *);
3369 extern int	ipmp_grp_rename(ipmp_grp_t *, const char *);
3370 extern ipmp_grp_t *ipmp_grp_lookup(const char *, ip_stack_t *);
3371 extern int	ipmp_grp_vet_phyint(ipmp_grp_t *, phyint_t *);
3372 extern ipmp_illgrp_t *ipmp_illgrp_create(ill_t *);
3373 extern void	ipmp_illgrp_destroy(ipmp_illgrp_t *);
3374 extern ill_t	*ipmp_illgrp_add_ipif(ipmp_illgrp_t *, ipif_t *);
3375 extern void	ipmp_illgrp_del_ipif(ipmp_illgrp_t *, ipif_t *);
3376 extern ill_t	*ipmp_illgrp_next_ill(ipmp_illgrp_t *);
3377 extern ill_t	*ipmp_illgrp_hold_next_ill(ipmp_illgrp_t *);
3378 extern ill_t	*ipmp_illgrp_hold_cast_ill(ipmp_illgrp_t *);
3379 extern ill_t	*ipmp_illgrp_ipmp_ill(ipmp_illgrp_t *);
3380 extern void	ipmp_illgrp_refresh_mtu(ipmp_illgrp_t *);
3381 extern ipmp_arpent_t *ipmp_illgrp_create_arpent(ipmp_illgrp_t *,
3382     boolean_t, ipaddr_t, uchar_t *, size_t, uint16_t);
3383 extern void	ipmp_illgrp_destroy_arpent(ipmp_illgrp_t *, ipmp_arpent_t *);
3384 extern ipmp_arpent_t *ipmp_illgrp_lookup_arpent(ipmp_illgrp_t *, ipaddr_t *);
3385 extern void	ipmp_illgrp_refresh_arpent(ipmp_illgrp_t *);
3386 extern void	ipmp_illgrp_mark_arpent(ipmp_illgrp_t *, ipmp_arpent_t *);
3387 extern ill_t	*ipmp_illgrp_find_ill(ipmp_illgrp_t *, uchar_t *, uint_t);
3388 extern void	ipmp_illgrp_link_grp(ipmp_illgrp_t *, ipmp_grp_t *);
3389 extern int	ipmp_illgrp_unlink_grp(ipmp_illgrp_t *);
3390 extern uint_t	ipmp_ill_get_ipmp_ifindex(const ill_t *);
3391 extern void	ipmp_ill_join_illgrp(ill_t *, ipmp_illgrp_t *);
3392 extern void	ipmp_ill_leave_illgrp(ill_t *);
3393 extern ill_t	*ipmp_ill_hold_ipmp_ill(ill_t *);
3394 extern ill_t	*ipmp_ill_hold_xmit_ill(ill_t *, boolean_t);
3395 extern boolean_t ipmp_ill_is_active(ill_t *);
3396 extern void	ipmp_ill_refresh_active(ill_t *);
3397 extern void	ipmp_phyint_join_grp(phyint_t *, ipmp_grp_t *);
3398 extern void	ipmp_phyint_leave_grp(phyint_t *);
3399 extern void	ipmp_phyint_refresh_active(phyint_t *);
3400 extern ill_t	*ipmp_ipif_bound_ill(const ipif_t *);
3401 extern ill_t	*ipmp_ipif_hold_bound_ill(const ipif_t *);
3402 extern boolean_t ipmp_ipif_is_dataaddr(const ipif_t *);
3403 extern boolean_t ipmp_ipif_is_stubaddr(const ipif_t *);
3404 extern boolean_t ipmp_packet_is_probe(mblk_t *, ill_t *);
3405 extern void	ipmp_ncec_delete_nce(ncec_t *);
3406 extern void	ipmp_ncec_refresh_nce(ncec_t *);
3407 
3408 extern void	conn_drain_insert(conn_t *, idl_tx_list_t *);
3409 extern void	conn_setqfull(conn_t *, boolean_t *);
3410 extern void	conn_clrqfull(conn_t *, boolean_t *);
3411 extern int	conn_ipsec_length(conn_t *);
3412 extern ipaddr_t	ip_get_dst(ipha_t *);
3413 extern uint_t	ip_get_pmtu(ip_xmit_attr_t *);
3414 extern uint_t	ip_get_base_mtu(ill_t *, ire_t *);
3415 extern mblk_t *ip_output_attach_policy(mblk_t *, ipha_t *, ip6_t *,
3416     const conn_t *, ip_xmit_attr_t *);
3417 extern int	ipsec_out_extra_length(ip_xmit_attr_t *);
3418 extern int	ipsec_out_process(mblk_t *, ip_xmit_attr_t *);
3419 extern int	ip_output_post_ipsec(mblk_t *, ip_xmit_attr_t *);
3420 extern void	ipsec_out_to_in(ip_xmit_attr_t *, ill_t *ill,
3421     ip_recv_attr_t *);
3422 
3423 extern void	ire_cleanup(ire_t *);
3424 extern void	ire_inactive(ire_t *);
3425 extern boolean_t irb_inactive(irb_t *);
3426 extern ire_t	*ire_unlink(irb_t *);
3427 
3428 #ifdef DEBUG
3429 extern	boolean_t th_trace_ref(const void *, ip_stack_t *);
3430 extern	void	th_trace_unref(const void *);
3431 extern	void	th_trace_cleanup(const void *, boolean_t);
3432 extern	void	ire_trace_ref(ire_t *);
3433 extern	void	ire_untrace_ref(ire_t *);
3434 #endif
3435 
3436 extern int	ip_srcid_insert(const in6_addr_t *, zoneid_t, ip_stack_t *);
3437 extern int	ip_srcid_remove(const in6_addr_t *, zoneid_t, ip_stack_t *);
3438 extern void	ip_srcid_find_id(uint_t, in6_addr_t *, zoneid_t, netstack_t *);
3439 extern uint_t	ip_srcid_find_addr(const in6_addr_t *, zoneid_t, netstack_t *);
3440 
3441 extern uint8_t	ipoptp_next(ipoptp_t *);
3442 extern uint8_t	ipoptp_first(ipoptp_t *, ipha_t *);
3443 extern int	ip_opt_get_user(conn_t *, uchar_t *);
3444 extern int	ipsec_req_from_conn(conn_t *, ipsec_req_t *, int);
3445 extern int	ip_snmp_get(queue_t *q, mblk_t *mctl, int level);
3446 extern int	ip_snmp_set(queue_t *q, int, int, uchar_t *, int);
3447 extern void	ip_process_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3448 extern void	ip_quiesce_conn(conn_t *);
3449 extern  void    ip_reprocess_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3450 extern void	ip_ioctl_finish(queue_t *, mblk_t *, int, int, ipsq_t *);
3451 
3452 extern boolean_t ip_cmpbuf(const void *, uint_t, boolean_t, const void *,
3453     uint_t);
3454 extern boolean_t ip_allocbuf(void **, uint_t *, boolean_t, const void *,
3455     uint_t);
3456 extern void	ip_savebuf(void **, uint_t *, boolean_t, const void *, uint_t);
3457 
3458 extern boolean_t	ipsq_pending_mp_cleanup(ill_t *, conn_t *);
3459 extern void	conn_ioctl_cleanup(conn_t *);
3460 
3461 extern void	ip_unbind(conn_t *);
3462 
3463 extern void tnet_init(void);
3464 extern void tnet_fini(void);
3465 
3466 /*
3467  * Hook functions to enable cluster networking
3468  * On non-clustered systems these vectors must always be NULL.
3469  */
3470 extern int (*cl_inet_isclusterwide)(netstackid_t stack_id, uint8_t protocol,
3471     sa_family_t addr_family, uint8_t *laddrp, void *args);
3472 extern uint32_t (*cl_inet_ipident)(netstackid_t stack_id, uint8_t protocol,
3473     sa_family_t addr_family, uint8_t *laddrp, uint8_t *faddrp,
3474     void *args);
3475 extern int (*cl_inet_connect2)(netstackid_t stack_id, uint8_t protocol,
3476     boolean_t is_outgoing, sa_family_t addr_family, uint8_t *laddrp,
3477     in_port_t lport, uint8_t *faddrp, in_port_t fport, void *args);
3478 extern void (*cl_inet_getspi)(netstackid_t, uint8_t, uint8_t *, size_t,
3479     void *);
3480 extern void (*cl_inet_getspi)(netstackid_t stack_id, uint8_t protocol,
3481     uint8_t *ptr, size_t len, void *args);
3482 extern int (*cl_inet_checkspi)(netstackid_t stack_id, uint8_t protocol,
3483     uint32_t spi, void *args);
3484 extern void (*cl_inet_deletespi)(netstackid_t stack_id, uint8_t protocol,
3485     uint32_t spi, void *args);
3486 extern void (*cl_inet_idlesa)(netstackid_t, uint8_t, uint32_t,
3487     sa_family_t, in6_addr_t, in6_addr_t, void *);
3488 
3489 
3490 /* Hooks for CGTP (multirt routes) filtering module */
3491 #define	CGTP_FILTER_REV_1	1
3492 #define	CGTP_FILTER_REV_2	2
3493 #define	CGTP_FILTER_REV_3	3
3494 #define	CGTP_FILTER_REV		CGTP_FILTER_REV_3
3495 
3496 /* cfo_filter and cfo_filter_v6 hooks return values */
3497 #define	CGTP_IP_PKT_NOT_CGTP	0
3498 #define	CGTP_IP_PKT_PREMIUM	1
3499 #define	CGTP_IP_PKT_DUPLICATE	2
3500 
3501 /* Version 3 of the filter interface */
3502 typedef struct cgtp_filter_ops {
3503 	int	cfo_filter_rev;			/* CGTP_FILTER_REV_3 */
3504 	int	(*cfo_change_state)(netstackid_t, int);
3505 	int	(*cfo_add_dest_v4)(netstackid_t, ipaddr_t, ipaddr_t,
3506 		    ipaddr_t, ipaddr_t);
3507 	int	(*cfo_del_dest_v4)(netstackid_t, ipaddr_t, ipaddr_t);
3508 	int	(*cfo_add_dest_v6)(netstackid_t, in6_addr_t *, in6_addr_t *,
3509 		    in6_addr_t *, in6_addr_t *);
3510 	int	(*cfo_del_dest_v6)(netstackid_t, in6_addr_t *, in6_addr_t *);
3511 	int	(*cfo_filter)(netstackid_t, uint_t, mblk_t *);
3512 	int	(*cfo_filter_v6)(netstackid_t, uint_t, ip6_t *,
3513 		    ip6_frag_t *);
3514 } cgtp_filter_ops_t;
3515 
3516 #define	CGTP_MCAST_SUCCESS	1
3517 
3518 /*
3519  * The separate CGTP module needs this global symbol so that it
3520  * can check the version and determine whether to use the old or the new
3521  * version of the filtering interface.
3522  */
3523 extern int	ip_cgtp_filter_rev;
3524 
3525 extern int	ip_cgtp_filter_supported(void);
3526 extern int	ip_cgtp_filter_register(netstackid_t, cgtp_filter_ops_t *);
3527 extern int	ip_cgtp_filter_unregister(netstackid_t);
3528 extern int	ip_cgtp_filter_is_registered(netstackid_t);
3529 
3530 /*
3531  * rr_ring_state cycles in the order shown below from RR_FREE through
3532  * RR_FREE_IN_PROG and  back to RR_FREE.
3533  */
3534 typedef enum {
3535 	RR_FREE,			/* Free slot */
3536 	RR_SQUEUE_UNBOUND,		/* Ring's squeue is unbound */
3537 	RR_SQUEUE_BIND_INPROG,		/* Ring's squeue bind in progress */
3538 	RR_SQUEUE_BOUND,		/* Ring's squeue bound to cpu */
3539 	RR_FREE_INPROG			/* Ring is being freed */
3540 } ip_ring_state_t;
3541 
3542 #define	ILL_MAX_RINGS		256	/* Max num of rx rings we can manage */
3543 #define	ILL_POLLING		0x01	/* Polling in use */
3544 
3545 /*
3546  * These functions pointer types are exported by the mac/dls layer.
3547  * we need to duplicate the definitions here because we cannot
3548  * include mac/dls header files here.
3549  */
3550 typedef boolean_t		(*ip_mac_intr_disable_t)(void *);
3551 typedef void			(*ip_mac_intr_enable_t)(void *);
3552 typedef ip_mac_tx_cookie_t	(*ip_dld_tx_t)(void *, mblk_t *,
3553     uint64_t, uint16_t);
3554 typedef	void			(*ip_flow_enable_t)(void *, ip_mac_tx_cookie_t);
3555 typedef void			*(*ip_dld_callb_t)(void *,
3556     ip_flow_enable_t, void *);
3557 typedef boolean_t		(*ip_dld_fctl_t)(void *, ip_mac_tx_cookie_t);
3558 typedef int			(*ip_capab_func_t)(void *, uint_t,
3559     void *, uint_t);
3560 
3561 /*
3562  * POLLING README
3563  * sq_get_pkts() is called to pick packets from softring in poll mode. It
3564  * calls rr_rx to get the chain and process it with rr_ip_accept.
3565  * rr_rx = mac_soft_ring_poll() to pick packets
3566  * rr_ip_accept = ip_accept_tcp() to process packets
3567  */
3568 
3569 /*
3570  * XXX: With protocol, service specific squeues, they will have
3571  * specific acceptor functions.
3572  */
3573 typedef	mblk_t *(*ip_mac_rx_t)(void *, size_t);
3574 typedef mblk_t *(*ip_accept_t)(ill_t *, ill_rx_ring_t *,
3575     squeue_t *, mblk_t *, mblk_t **, uint_t *);
3576 
3577 /*
3578  * rr_intr_enable, rr_intr_disable, rr_rx_handle, rr_rx:
3579  * May be accessed while in the squeue AND after checking that SQS_POLL_CAPAB
3580  * is set.
3581  *
3582  * rr_ring_state: Protected by ill_lock.
3583  */
3584 struct ill_rx_ring {
3585 	ip_mac_intr_disable_t	rr_intr_disable; /* Interrupt disabling func */
3586 	ip_mac_intr_enable_t	rr_intr_enable;	/* Interrupt enabling func */
3587 	void			*rr_intr_handle; /* Handle interrupt funcs */
3588 	ip_mac_rx_t		rr_rx;		/* Driver receive function */
3589 	ip_accept_t		rr_ip_accept;	/* IP accept function */
3590 	void			*rr_rx_handle;	/* Handle for Rx ring */
3591 	squeue_t		*rr_sqp; /* Squeue the ring is bound to */
3592 	ill_t			*rr_ill;	/* back pointer to ill */
3593 	ip_ring_state_t		rr_ring_state;	/* State of this ring */
3594 };
3595 
3596 /*
3597  * IP - DLD direct function call capability
3598  * Suffixes, df - dld function, dh - dld handle,
3599  * cf - client (IP) function, ch - client handle
3600  */
3601 typedef struct ill_dld_direct_s {		/* DLD provided driver Tx */
3602 	ip_dld_tx_t		idd_tx_df;	/* str_mdata_fastpath_put */
3603 	void			*idd_tx_dh;	/* dld_str_t *dsp */
3604 	ip_dld_callb_t		idd_tx_cb_df;	/* mac_tx_srs_notify */
3605 	void			*idd_tx_cb_dh;	/* mac_client_handle_t *mch */
3606 	ip_dld_fctl_t		idd_tx_fctl_df;	/* mac_tx_is_flow_blocked */
3607 	void			*idd_tx_fctl_dh;	/* mac_client_handle */
3608 } ill_dld_direct_t;
3609 
3610 /* IP - DLD polling capability */
3611 typedef struct ill_dld_poll_s {
3612 	ill_rx_ring_t		idp_ring_tbl[ILL_MAX_RINGS];
3613 } ill_dld_poll_t;
3614 
3615 /* Describes ill->ill_dld_capab */
3616 struct ill_dld_capab_s {
3617 	ip_capab_func_t		idc_capab_df;	/* dld_capab_func */
3618 	void			*idc_capab_dh;	/* dld_str_t *dsp */
3619 	ill_dld_direct_t	idc_direct;
3620 	ill_dld_poll_t		idc_poll;
3621 };
3622 
3623 /*
3624  * IP squeues exports
3625  */
3626 extern boolean_t 	ip_squeue_fanout;
3627 
3628 #define	IP_SQUEUE_GET(hint) ip_squeue_random(hint)
3629 
3630 extern void ip_squeue_init(void (*)(squeue_t *));
3631 extern squeue_t	*ip_squeue_random(uint_t);
3632 extern squeue_t *ip_squeue_get(ill_rx_ring_t *);
3633 extern squeue_t *ip_squeue_getfree(pri_t);
3634 extern int ip_squeue_cpu_move(squeue_t *, processorid_t);
3635 extern void *ip_squeue_add_ring(ill_t *, void *);
3636 extern void ip_squeue_bind_ring(ill_t *, ill_rx_ring_t *, processorid_t);
3637 extern void ip_squeue_clean_ring(ill_t *, ill_rx_ring_t *);
3638 extern void ip_squeue_quiesce_ring(ill_t *, ill_rx_ring_t *);
3639 extern void ip_squeue_restart_ring(ill_t *, ill_rx_ring_t *);
3640 extern void ip_squeue_clean_all(ill_t *);
3641 extern boolean_t	ip_source_routed(ipha_t *, ip_stack_t *);
3642 
3643 extern void tcp_wput(queue_t *, mblk_t *);
3644 
3645 extern int	ip_fill_mtuinfo(conn_t *, ip_xmit_attr_t *,
3646     struct ip6_mtuinfo *);
3647 extern hook_t *ipobs_register_hook(netstack_t *, pfv_t);
3648 extern void ipobs_unregister_hook(netstack_t *, hook_t *);
3649 extern void ipobs_hook(mblk_t *, int, zoneid_t, zoneid_t, const ill_t *,
3650     ip_stack_t *);
3651 typedef void    (*ipsq_func_t)(ipsq_t *, queue_t *, mblk_t *, void *);
3652 
3653 extern void	dce_g_init(void);
3654 extern void	dce_g_destroy(void);
3655 extern void	dce_stack_init(ip_stack_t *);
3656 extern void	dce_stack_destroy(ip_stack_t *);
3657 extern void	dce_cleanup(uint_t, ip_stack_t *);
3658 extern dce_t	*dce_get_default(ip_stack_t *);
3659 extern dce_t	*dce_lookup_pkt(mblk_t *, ip_xmit_attr_t *, uint_t *);
3660 extern dce_t	*dce_lookup_v4(ipaddr_t, ip_stack_t *, uint_t *);
3661 extern dce_t	*dce_lookup_v6(const in6_addr_t *, uint_t, ip_stack_t *,
3662     uint_t *);
3663 extern dce_t	*dce_lookup_and_add_v4(ipaddr_t, ip_stack_t *);
3664 extern dce_t	*dce_lookup_and_add_v6(const in6_addr_t *, uint_t,
3665     ip_stack_t *);
3666 extern int	dce_update_uinfo_v4(ipaddr_t, iulp_t *, ip_stack_t *);
3667 extern int	dce_update_uinfo_v6(const in6_addr_t *, uint_t, iulp_t *,
3668     ip_stack_t *);
3669 extern int	dce_update_uinfo(const in6_addr_t *, uint_t, iulp_t *,
3670     ip_stack_t *);
3671 extern void	dce_increment_generation(dce_t *);
3672 extern void	dce_increment_all_generations(boolean_t, ip_stack_t *);
3673 extern void	dce_refrele(dce_t *);
3674 extern void	dce_refhold(dce_t *);
3675 extern void	dce_refrele_notr(dce_t *);
3676 extern void	dce_refhold_notr(dce_t *);
3677 mblk_t		*ip_snmp_get_mib2_ip_dce(queue_t *, mblk_t *, ip_stack_t *ipst);
3678 
3679 extern ip_laddr_t ip_laddr_verify_v4(ipaddr_t, zoneid_t,
3680     ip_stack_t *, boolean_t);
3681 extern ip_laddr_t ip_laddr_verify_v6(const in6_addr_t *, zoneid_t,
3682     ip_stack_t *, boolean_t, uint_t);
3683 extern int	ip_laddr_fanout_insert(conn_t *);
3684 
3685 extern boolean_t ip_verify_src(mblk_t *, ip_xmit_attr_t *, uint_t *);
3686 extern int	ip_verify_ire(mblk_t *, ip_xmit_attr_t *);
3687 
3688 extern mblk_t	*ip_xmit_attr_to_mblk(ip_xmit_attr_t *);
3689 extern boolean_t ip_xmit_attr_from_mblk(mblk_t *, ip_xmit_attr_t *);
3690 extern mblk_t	*ip_xmit_attr_free_mblk(mblk_t *);
3691 extern mblk_t	*ip_recv_attr_to_mblk(ip_recv_attr_t *);
3692 extern boolean_t ip_recv_attr_from_mblk(mblk_t *, ip_recv_attr_t *);
3693 extern mblk_t	*ip_recv_attr_free_mblk(mblk_t *);
3694 extern boolean_t ip_recv_attr_is_mblk(mblk_t *);
3695 
3696 /*
3697  * Squeue tags. Tags only need to be unique when the callback function is the
3698  * same to distinguish between different calls, but we use unique tags for
3699  * convenience anyway.
3700  */
3701 #define	SQTAG_IP_INPUT			1
3702 #define	SQTAG_TCP_INPUT_ICMP_ERR	2
3703 #define	SQTAG_TCP6_INPUT_ICMP_ERR	3
3704 #define	SQTAG_IP_TCP_INPUT		4
3705 #define	SQTAG_IP6_TCP_INPUT		5
3706 #define	SQTAG_IP_TCP_CLOSE		6
3707 #define	SQTAG_TCP_OUTPUT		7
3708 #define	SQTAG_TCP_TIMER			8
3709 #define	SQTAG_TCP_TIMEWAIT		9
3710 #define	SQTAG_TCP_ACCEPT_FINISH		10
3711 #define	SQTAG_TCP_ACCEPT_FINISH_Q0	11
3712 #define	SQTAG_TCP_ACCEPT_PENDING	12
3713 #define	SQTAG_TCP_LISTEN_DISCON		13
3714 #define	SQTAG_TCP_CONN_REQ_1		14
3715 #define	SQTAG_TCP_EAGER_BLOWOFF		15
3716 #define	SQTAG_TCP_EAGER_CLEANUP		16
3717 #define	SQTAG_TCP_EAGER_CLEANUP_Q0	17
3718 #define	SQTAG_TCP_CONN_IND		18
3719 #define	SQTAG_TCP_RSRV			19
3720 #define	SQTAG_TCP_ABORT_BUCKET		20
3721 #define	SQTAG_TCP_REINPUT		21
3722 #define	SQTAG_TCP_REINPUT_EAGER		22
3723 #define	SQTAG_TCP_INPUT_MCTL		23
3724 #define	SQTAG_TCP_RPUTOTHER		24
3725 #define	SQTAG_IP_PROTO_AGAIN		25
3726 #define	SQTAG_IP_FANOUT_TCP		26
3727 #define	SQTAG_IPSQ_CLEAN_RING		27
3728 #define	SQTAG_TCP_WPUT_OTHER		28
3729 #define	SQTAG_TCP_CONN_REQ_UNBOUND	29
3730 #define	SQTAG_TCP_SEND_PENDING		30
3731 #define	SQTAG_BIND_RETRY		31
3732 #define	SQTAG_UDP_FANOUT		32
3733 #define	SQTAG_UDP_INPUT			33
3734 #define	SQTAG_UDP_WPUT			34
3735 #define	SQTAG_UDP_OUTPUT		35
3736 #define	SQTAG_TCP_KSSL_INPUT		36
3737 #define	SQTAG_TCP_DROP_Q0		37
3738 #define	SQTAG_TCP_CONN_REQ_2		38
3739 #define	SQTAG_IP_INPUT_RX_RING		39
3740 #define	SQTAG_SQUEUE_CHANGE		40
3741 #define	SQTAG_CONNECT_FINISH		41
3742 #define	SQTAG_SYNCHRONOUS_OP		42
3743 #define	SQTAG_TCP_SHUTDOWN_OUTPUT	43
3744 #define	SQTAG_TCP_IXA_CLEANUP		44
3745 #define	SQTAG_TCP_SEND_SYNACK		45
3746 
3747 extern sin_t	sin_null;	/* Zero address for quick clears */
3748 extern sin6_t	sin6_null;	/* Zero address for quick clears */
3749 
3750 #endif	/* _KERNEL */
3751 
3752 #ifdef	__cplusplus
3753 }
3754 #endif
3755 
3756 #endif	/* _INET_IP_H */
3757