xref: /illumos-gate/usr/src/uts/common/inet/ip.h (revision 241c90a0)
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) 1990 Mentat Inc.
24  * Copyright (c) 1991, 2010, Oracle and/or its affiliates. All rights reserved.
25  * Copyright 2017 Nexenta Systems, Inc.
26  * Copyright 2017 OmniTI Computer Consulting, Inc. All rights reserved.
27  * Copyright 2019, Joyent, Inc.
28  */
29 
30 #ifndef	_INET_IP_H
31 #define	_INET_IP_H
32 
33 #ifdef	__cplusplus
34 extern "C" {
35 #endif
36 
37 #include <sys/isa_defs.h>
38 #include <sys/types.h>
39 #include <inet/mib2.h>
40 #include <inet/nd.h>
41 #include <sys/atomic.h>
42 #include <net/if_dl.h>
43 #include <net/if.h>
44 #include <netinet/ip.h>
45 #include <netinet/igmp.h>
46 #include <sys/neti.h>
47 #include <sys/hook.h>
48 #include <sys/hook_event.h>
49 #include <sys/hook_impl.h>
50 #include <inet/ip_stack.h>
51 
52 #ifdef _KERNEL
53 #include <netinet/ip6.h>
54 #include <sys/avl.h>
55 #include <sys/list.h>
56 #include <sys/vmem.h>
57 #include <sys/squeue.h>
58 #include <net/route.h>
59 #include <sys/systm.h>
60 #include <net/radix.h>
61 #include <sys/modhash.h>
62 
63 #ifdef DEBUG
64 #define	CONN_DEBUG
65 #endif
66 
67 #define	IP_DEBUG
68 /*
69  * The mt-streams(9F) flags for the IP module; put here so that other
70  * "drivers" that are actually IP (e.g., ICMP, UDP) can use the same set
71  * of flags.
72  */
73 #define	IP_DEVMTFLAGS D_MP
74 #endif	/* _KERNEL */
75 
76 #define	IP_MOD_NAME	"ip"
77 #define	IP_DEV_NAME	"/dev/ip"
78 #define	IP6_DEV_NAME	"/dev/ip6"
79 
80 #define	UDP_MOD_NAME	"udp"
81 #define	UDP_DEV_NAME	"/dev/udp"
82 #define	UDP6_DEV_NAME	"/dev/udp6"
83 
84 #define	TCP_MOD_NAME	"tcp"
85 #define	TCP_DEV_NAME	"/dev/tcp"
86 #define	TCP6_DEV_NAME	"/dev/tcp6"
87 
88 #define	SCTP_MOD_NAME	"sctp"
89 
90 #ifndef	_IPADDR_T
91 #define	_IPADDR_T
92 typedef uint32_t ipaddr_t;
93 #endif
94 
95 /* Number of bits in an address */
96 #define	IP_ABITS		32
97 #define	IPV4_ABITS		IP_ABITS
98 #define	IPV6_ABITS		128
99 #define	IP_MAX_HW_LEN	40
100 
101 #define	IP_HOST_MASK		(ipaddr_t)0xffffffffU
102 
103 #define	IP_CSUM(mp, off, sum)		(~ip_cksum(mp, off, sum) & 0xFFFF)
104 #define	IP_CSUM_PARTIAL(mp, off, sum)	ip_cksum(mp, off, sum)
105 #define	IP_BCSUM_PARTIAL(bp, len, sum)	bcksum(bp, len, sum)
106 
107 #define	ILL_FRAG_HASH_TBL_COUNT	((unsigned int)64)
108 #define	ILL_FRAG_HASH_TBL_SIZE	(ILL_FRAG_HASH_TBL_COUNT * sizeof (ipfb_t))
109 
110 #define	IPV4_ADDR_LEN			4
111 #define	IP_ADDR_LEN			IPV4_ADDR_LEN
112 #define	IP_ARP_PROTO_TYPE		0x0800
113 
114 #define	IPV4_VERSION			4
115 #define	IP_VERSION			IPV4_VERSION
116 #define	IP_SIMPLE_HDR_LENGTH_IN_WORDS	5
117 #define	IP_SIMPLE_HDR_LENGTH		20
118 #define	IP_MAX_HDR_LENGTH		60
119 
120 #define	IP_MAX_OPT_LENGTH (IP_MAX_HDR_LENGTH-IP_SIMPLE_HDR_LENGTH)
121 
122 #define	IP_MIN_MTU			(IP_MAX_HDR_LENGTH + 8)	/* 68 bytes */
123 
124 /*
125  * XXX IP_MAXPACKET is defined in <netinet/ip.h> as well. At some point the
126  * 2 files should be cleaned up to remove all redundant definitions.
127  */
128 #define	IP_MAXPACKET			65535
129 #define	IP_SIMPLE_HDR_VERSION \
130 	((IP_VERSION << 4) | IP_SIMPLE_HDR_LENGTH_IN_WORDS)
131 
132 #define	UDPH_SIZE			8
133 
134 /*
135  * Constants and type definitions to support IP IOCTL commands
136  */
137 #define	IP_IOCTL			(('i'<<8)|'p')
138 #define	IP_IOC_IRE_DELETE		4
139 #define	IP_IOC_IRE_DELETE_NO_REPLY	5
140 #define	IP_IOC_RTS_REQUEST		7
141 
142 /* Common definitions used by IP IOCTL data structures */
143 typedef struct ipllcmd_s {
144 	uint_t	ipllc_cmd;
145 	uint_t	ipllc_name_offset;
146 	uint_t	ipllc_name_length;
147 } ipllc_t;
148 
149 /* IP IRE Delete Command Structure. */
150 typedef struct ipid_s {
151 	ipllc_t	ipid_ipllc;
152 	uint_t	ipid_ire_type;
153 	uint_t	ipid_addr_offset;
154 	uint_t	ipid_addr_length;
155 	uint_t	ipid_mask_offset;
156 	uint_t	ipid_mask_length;
157 } ipid_t;
158 
159 #define	ipid_cmd		ipid_ipllc.ipllc_cmd
160 
161 #ifdef _KERNEL
162 /*
163  * Temporary state for ip options parser.
164  */
165 typedef struct ipoptp_s
166 {
167 	uint8_t		*ipoptp_next;	/* next option to look at */
168 	uint8_t		*ipoptp_end;	/* end of options */
169 	uint8_t		*ipoptp_cur;	/* start of current option */
170 	uint8_t		ipoptp_len;	/* length of current option */
171 	uint32_t	ipoptp_flags;
172 } ipoptp_t;
173 
174 /*
175  * Flag(s) for ipoptp_flags
176  */
177 #define	IPOPTP_ERROR	0x00000001
178 #endif	/* _KERNEL */
179 
180 /* Controls forwarding of IP packets, set via ipadm(1M)/ndd(1M) */
181 #define	IP_FORWARD_NEVER	0
182 #define	IP_FORWARD_ALWAYS	1
183 
184 #define	WE_ARE_FORWARDING(ipst)	((ipst)->ips_ip_forwarding == IP_FORWARD_ALWAYS)
185 
186 #define	IPH_HDR_LENGTH(ipha)						\
187 	((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length & 0xF) << 2)
188 
189 #define	IPH_HDR_VERSION(ipha)						\
190 	((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length) >> 4)
191 
192 #ifdef _KERNEL
193 /*
194  * IP reassembly macros.  We hide starting and ending offsets in b_next and
195  * b_prev of messages on the reassembly queue.	The messages are chained using
196  * b_cont.  These macros are used in ip_reassemble() so we don't have to see
197  * the ugly casts and assignments.
198  * Note that the offsets are <= 64k i.e. a uint_t is sufficient to represent
199  * them.
200  */
201 #define	IP_REASS_START(mp)		((uint_t)(uintptr_t)((mp)->b_next))
202 #define	IP_REASS_SET_START(mp, u)	\
203 	((mp)->b_next = (mblk_t *)(uintptr_t)(u))
204 #define	IP_REASS_END(mp)		((uint_t)(uintptr_t)((mp)->b_prev))
205 #define	IP_REASS_SET_END(mp, u)		\
206 	((mp)->b_prev = (mblk_t *)(uintptr_t)(u))
207 
208 #define	IP_REASS_COMPLETE	0x1
209 #define	IP_REASS_PARTIAL	0x2
210 #define	IP_REASS_FAILED		0x4
211 
212 /*
213  * Test to determine whether this is a module instance of IP or a
214  * driver instance of IP.
215  */
216 #define	CONN_Q(q)	(WR(q)->q_next == NULL)
217 
218 #define	Q_TO_CONN(q)	((conn_t *)(q)->q_ptr)
219 #define	Q_TO_TCP(q)	(Q_TO_CONN((q))->conn_tcp)
220 #define	Q_TO_UDP(q)	(Q_TO_CONN((q))->conn_udp)
221 #define	Q_TO_ICMP(q)	(Q_TO_CONN((q))->conn_icmp)
222 #define	Q_TO_RTS(q)	(Q_TO_CONN((q))->conn_rts)
223 
224 #define	CONNP_TO_WQ(connp)	((connp)->conn_wq)
225 #define	CONNP_TO_RQ(connp)	((connp)->conn_rq)
226 
227 #define	GRAB_CONN_LOCK(q)	{				\
228 	if (q != NULL && CONN_Q(q))				\
229 		mutex_enter(&(Q_TO_CONN(q))->conn_lock);	\
230 }
231 
232 #define	RELEASE_CONN_LOCK(q)	{				\
233 	if (q != NULL && CONN_Q(q))				\
234 		mutex_exit(&(Q_TO_CONN(q))->conn_lock);		\
235 }
236 
237 /*
238  * Ref counter macros for ioctls. This provides a guard for TCP to stop
239  * tcp_close from removing the rq/wq whilst an ioctl is still in flight on the
240  * stream. The ioctl could have been queued on e.g. an ipsq. tcp_close will wait
241  * until the ioctlref count is zero before proceeding.
242  * Ideally conn_oper_pending_ill would be used for this purpose. However, in the
243  * case where an ioctl is aborted or interrupted, it can be cleared prematurely.
244  * There are also some race possibilities between ip and the stream head which
245  * can also end up with conn_oper_pending_ill being cleared prematurely. So, to
246  * avoid these situations, we use a dedicated ref counter for ioctls which is
247  * used in addition to and in parallel with the normal conn_ref count.
248  */
249 #define	CONN_INC_IOCTLREF_LOCKED(connp)	{			\
250 	ASSERT(MUTEX_HELD(&(connp)->conn_lock));		\
251 	DTRACE_PROBE1(conn__inc__ioctlref, conn_t *, (connp));	\
252 	(connp)->conn_ioctlref++;				\
253 	mutex_exit(&(connp)->conn_lock);			\
254 }
255 
256 #define	CONN_INC_IOCTLREF(connp)	{			\
257 	mutex_enter(&(connp)->conn_lock);			\
258 	CONN_INC_IOCTLREF_LOCKED(connp);			\
259 }
260 
261 #define	CONN_DEC_IOCTLREF(connp)	{			\
262 	mutex_enter(&(connp)->conn_lock);			\
263 	DTRACE_PROBE1(conn__dec__ioctlref, conn_t *, (connp));	\
264 	/* Make sure conn_ioctlref will not underflow. */	\
265 	ASSERT((connp)->conn_ioctlref != 0);			\
266 	if ((--(connp)->conn_ioctlref == 0) &&			\
267 	    ((connp)->conn_state_flags & CONN_CLOSING)) {	\
268 		cv_broadcast(&(connp)->conn_cv);		\
269 	}							\
270 	mutex_exit(&(connp)->conn_lock);			\
271 }
272 
273 
274 /*
275  * Complete the pending operation. Usually an ioctl. Can also
276  * be a bind or option management request that got enqueued
277  * in an ipsq_t. Called on completion of the operation.
278  */
279 #define	CONN_OPER_PENDING_DONE(connp)	{			\
280 	mutex_enter(&(connp)->conn_lock);			\
281 	(connp)->conn_oper_pending_ill = NULL;			\
282 	cv_broadcast(&(connp)->conn_refcv);			\
283 	mutex_exit(&(connp)->conn_lock);			\
284 	CONN_DEC_REF(connp);					\
285 }
286 
287 /*
288  * Values for squeue switch:
289  */
290 #define	IP_SQUEUE_ENTER_NODRAIN	1
291 #define	IP_SQUEUE_ENTER	2
292 #define	IP_SQUEUE_FILL 3
293 
294 extern int ip_squeue_flag;
295 
296 /* IP Fragmentation Reassembly Header */
297 typedef struct ipf_s {
298 	struct ipf_s	*ipf_hash_next;
299 	struct ipf_s	**ipf_ptphn;	/* Pointer to previous hash next. */
300 	uint32_t	ipf_ident;	/* Ident to match. */
301 	uint8_t		ipf_protocol;	/* Protocol to match. */
302 	uchar_t		ipf_last_frag_seen : 1;	/* Last fragment seen ? */
303 	time_t		ipf_timestamp;	/* Reassembly start time. */
304 	mblk_t		*ipf_mp;	/* mblk we live in. */
305 	mblk_t		*ipf_tail_mp;	/* Frag queue tail pointer. */
306 	int		ipf_hole_cnt;	/* Number of holes (hard-case). */
307 	int		ipf_end;	/* Tail end offset (0 -> hard-case). */
308 	uint_t		ipf_gen;	/* Frag queue generation */
309 	size_t		ipf_count;	/* Count of bytes used by frag */
310 	uint_t		ipf_nf_hdr_len; /* Length of nonfragmented header */
311 	in6_addr_t	ipf_v6src;	/* IPv6 source address */
312 	in6_addr_t	ipf_v6dst;	/* IPv6 dest address */
313 	uint_t		ipf_prev_nexthdr_offset; /* Offset for nexthdr value */
314 	uint8_t		ipf_ecn;	/* ECN info for the fragments */
315 	uint8_t		ipf_num_dups;	/* Number of times dup frags recvd */
316 	uint16_t	ipf_checksum_flags; /* Hardware checksum flags */
317 	uint32_t	ipf_checksum;	/* Partial checksum of fragment data */
318 } ipf_t;
319 
320 /*
321  * IPv4 Fragments
322  */
323 #define	IS_V4_FRAGMENT(ipha_fragment_offset_and_flags)			\
324 	(((ntohs(ipha_fragment_offset_and_flags) & IPH_OFFSET) != 0) ||	\
325 	((ntohs(ipha_fragment_offset_and_flags) & IPH_MF) != 0))
326 
327 #define	ipf_src	V4_PART_OF_V6(ipf_v6src)
328 #define	ipf_dst	V4_PART_OF_V6(ipf_v6dst)
329 
330 #endif /* _KERNEL */
331 
332 /* ICMP types */
333 #define	ICMP_ECHO_REPLY			0
334 #define	ICMP_DEST_UNREACHABLE		3
335 #define	ICMP_SOURCE_QUENCH		4
336 #define	ICMP_REDIRECT			5
337 #define	ICMP_ECHO_REQUEST		8
338 #define	ICMP_ROUTER_ADVERTISEMENT	9
339 #define	ICMP_ROUTER_SOLICITATION	10
340 #define	ICMP_TIME_EXCEEDED		11
341 #define	ICMP_PARAM_PROBLEM		12
342 #define	ICMP_TIME_STAMP_REQUEST		13
343 #define	ICMP_TIME_STAMP_REPLY		14
344 #define	ICMP_INFO_REQUEST		15
345 #define	ICMP_INFO_REPLY			16
346 #define	ICMP_ADDRESS_MASK_REQUEST	17
347 #define	ICMP_ADDRESS_MASK_REPLY		18
348 
349 /* Evaluates to true if the ICMP type is an ICMP error */
350 #define	ICMP_IS_ERROR(type)	(		\
351 	(type) == ICMP_DEST_UNREACHABLE ||	\
352 	(type) == ICMP_SOURCE_QUENCH ||		\
353 	(type) == ICMP_TIME_EXCEEDED ||		\
354 	(type) == ICMP_PARAM_PROBLEM)
355 
356 /* ICMP_TIME_EXCEEDED codes */
357 #define	ICMP_TTL_EXCEEDED		0
358 #define	ICMP_REASSEMBLY_TIME_EXCEEDED	1
359 
360 /* ICMP_DEST_UNREACHABLE codes */
361 #define	ICMP_NET_UNREACHABLE		0
362 #define	ICMP_HOST_UNREACHABLE		1
363 #define	ICMP_PROTOCOL_UNREACHABLE	2
364 #define	ICMP_PORT_UNREACHABLE		3
365 #define	ICMP_FRAGMENTATION_NEEDED	4
366 #define	ICMP_SOURCE_ROUTE_FAILED	5
367 #define	ICMP_DEST_NET_UNKNOWN		6
368 #define	ICMP_DEST_HOST_UNKNOWN		7
369 #define	ICMP_SRC_HOST_ISOLATED		8
370 #define	ICMP_DEST_NET_UNREACH_ADMIN	9
371 #define	ICMP_DEST_HOST_UNREACH_ADMIN	10
372 #define	ICMP_DEST_NET_UNREACH_TOS	11
373 #define	ICMP_DEST_HOST_UNREACH_TOS	12
374 
375 /* ICMP Header Structure */
376 typedef struct icmph_s {
377 	uint8_t		icmph_type;
378 	uint8_t		icmph_code;
379 	uint16_t	icmph_checksum;
380 	union {
381 		struct { /* ECHO request/response structure */
382 			uint16_t	u_echo_ident;
383 			uint16_t	u_echo_seqnum;
384 		} u_echo;
385 		struct { /* Destination unreachable structure */
386 			uint16_t	u_du_zero;
387 			uint16_t	u_du_mtu;
388 		} u_du;
389 		struct { /* Parameter problem structure */
390 			uint8_t		u_pp_ptr;
391 			uint8_t		u_pp_rsvd[3];
392 		} u_pp;
393 		struct { /* Redirect structure */
394 			ipaddr_t	u_rd_gateway;
395 		} u_rd;
396 	} icmph_u;
397 } icmph_t;
398 
399 #define	icmph_echo_ident	icmph_u.u_echo.u_echo_ident
400 #define	icmph_echo_seqnum	icmph_u.u_echo.u_echo_seqnum
401 #define	icmph_du_zero		icmph_u.u_du.u_du_zero
402 #define	icmph_du_mtu		icmph_u.u_du.u_du_mtu
403 #define	icmph_pp_ptr		icmph_u.u_pp.u_pp_ptr
404 #define	icmph_rd_gateway	icmph_u.u_rd.u_rd_gateway
405 
406 #define	ICMPH_SIZE	8
407 
408 /*
409  * Minimum length of transport layer header included in an ICMP error
410  * message for it to be considered valid.
411  */
412 #define	ICMP_MIN_TP_HDR_LEN	8
413 
414 /* Aligned IP header */
415 typedef struct ipha_s {
416 	uint8_t		ipha_version_and_hdr_length;
417 	uint8_t		ipha_type_of_service;
418 	uint16_t	ipha_length;
419 	uint16_t	ipha_ident;
420 	uint16_t	ipha_fragment_offset_and_flags;
421 	uint8_t		ipha_ttl;
422 	uint8_t		ipha_protocol;
423 	uint16_t	ipha_hdr_checksum;
424 	ipaddr_t	ipha_src;
425 	ipaddr_t	ipha_dst;
426 } ipha_t;
427 
428 /*
429  * IP Flags
430  *
431  * Some of these constant names are copied for the DTrace IP provider in
432  * usr/src/lib/libdtrace/common/{ip.d.in, ip.sed.in}, which should be kept
433  * in sync.
434  */
435 #define	IPH_DF		0x4000	/* Don't fragment */
436 #define	IPH_MF		0x2000	/* More fragments to come */
437 #define	IPH_OFFSET	0x1FFF	/* Where the offset lives */
438 
439 /* Byte-order specific values */
440 #ifdef	_BIG_ENDIAN
441 #define	IPH_DF_HTONS	0x4000	/* Don't fragment */
442 #define	IPH_MF_HTONS	0x2000	/* More fragments to come */
443 #define	IPH_OFFSET_HTONS 0x1FFF	/* Where the offset lives */
444 #else
445 #define	IPH_DF_HTONS	0x0040	/* Don't fragment */
446 #define	IPH_MF_HTONS	0x0020	/* More fragments to come */
447 #define	IPH_OFFSET_HTONS 0xFF1F	/* Where the offset lives */
448 #endif
449 
450 /* ECN code points for IPv4 TOS byte and IPv6 traffic class octet. */
451 #define	IPH_ECN_NECT	0x0	/* Not ECN-Capable Transport */
452 #define	IPH_ECN_ECT1	0x1	/* ECN-Capable Transport, ECT(1) */
453 #define	IPH_ECN_ECT0	0x2	/* ECN-Capable Transport, ECT(0) */
454 #define	IPH_ECN_CE	0x3	/* ECN-Congestion Experienced (CE) */
455 
456 struct ill_s;
457 
458 typedef	void ip_v6intfid_func_t(struct ill_s *, in6_addr_t *);
459 typedef void ip_v6mapinfo_func_t(struct ill_s *, uchar_t *, uchar_t *);
460 typedef void ip_v4mapinfo_func_t(struct ill_s *, uchar_t *, uchar_t *);
461 
462 /* IP Mac info structure */
463 typedef struct ip_m_s {
464 	t_uscalar_t		ip_m_mac_type;	/* From <sys/dlpi.h> */
465 	int			ip_m_type;	/* From <net/if_types.h> */
466 	t_uscalar_t		ip_m_ipv4sap;
467 	t_uscalar_t		ip_m_ipv6sap;
468 	ip_v4mapinfo_func_t	*ip_m_v4mapping;
469 	ip_v6mapinfo_func_t	*ip_m_v6mapping;
470 	ip_v6intfid_func_t	*ip_m_v6intfid;
471 	ip_v6intfid_func_t	*ip_m_v6destintfid;
472 } ip_m_t;
473 
474 /*
475  * The following functions attempt to reduce the link layer dependency
476  * of the IP stack. The current set of link specific operations are:
477  * a. map from IPv4 class D (224.0/4) multicast address range or the
478  * IPv6 multicast address range (ff00::/8) to the link layer multicast
479  * address.
480  * b. derive the default IPv6 interface identifier from the interface.
481  * c. derive the default IPv6 destination interface identifier from
482  * the interface (point-to-point only).
483  */
484 extern	void ip_mcast_mapping(struct ill_s *, uchar_t *, uchar_t *);
485 /* ip_m_v6*intfid return void and are never NULL */
486 #define	MEDIA_V6INTFID(ip_m, ill, v6ptr) (ip_m)->ip_m_v6intfid(ill, v6ptr)
487 #define	MEDIA_V6DESTINTFID(ip_m, ill, v6ptr) \
488 	(ip_m)->ip_m_v6destintfid(ill, v6ptr)
489 
490 /* Router entry types */
491 #define	IRE_BROADCAST		0x0001	/* Route entry for broadcast address */
492 #define	IRE_DEFAULT		0x0002	/* Route entry for default gateway */
493 #define	IRE_LOCAL		0x0004	/* Route entry for local address */
494 #define	IRE_LOOPBACK		0x0008	/* Route entry for loopback address */
495 #define	IRE_PREFIX		0x0010	/* Route entry for prefix routes */
496 #ifndef _KERNEL
497 /* Keep so user-level still compiles */
498 #define	IRE_CACHE		0x0020	/* Cached Route entry */
499 #endif
500 #define	IRE_IF_NORESOLVER	0x0040	/* Route entry for local interface */
501 					/* net without any address mapping. */
502 #define	IRE_IF_RESOLVER		0x0080	/* Route entry for local interface */
503 					/* net with resolver. */
504 #define	IRE_HOST		0x0100	/* Host route entry */
505 /* Keep so user-level still compiles */
506 #define	IRE_HOST_REDIRECT	0x0200	/* only used for T_SVR4_OPTMGMT_REQ */
507 #define	IRE_IF_CLONE		0x0400	/* Per host clone of IRE_IF */
508 #define	IRE_MULTICAST		0x0800	/* Special - not in table */
509 #define	IRE_NOROUTE		0x1000	/* Special - not in table */
510 
511 #define	IRE_INTERFACE		(IRE_IF_NORESOLVER | IRE_IF_RESOLVER)
512 
513 #define	IRE_IF_ALL		(IRE_IF_NORESOLVER | IRE_IF_RESOLVER | \
514 				    IRE_IF_CLONE)
515 #define	IRE_OFFSUBNET		(IRE_DEFAULT | IRE_PREFIX | IRE_HOST)
516 #define	IRE_OFFLINK		IRE_OFFSUBNET
517 /*
518  * Note that we view IRE_NOROUTE as ONLINK since we can "send" to them without
519  * going through a router; the result of sending will be an error/icmp error.
520  */
521 #define	IRE_ONLINK		(IRE_IF_ALL|IRE_LOCAL|IRE_LOOPBACK| \
522 				    IRE_BROADCAST|IRE_MULTICAST|IRE_NOROUTE)
523 
524 /* Arguments to ire_flush_cache() */
525 #define	IRE_FLUSH_DELETE	0
526 #define	IRE_FLUSH_ADD		1
527 #define	IRE_FLUSH_GWCHANGE	2
528 
529 /*
530  * Flags to ire_route_recursive
531  */
532 #define	IRR_NONE		0
533 #define	IRR_ALLOCATE		1	/* OK to allocate IRE_IF_CLONE */
534 #define	IRR_INCOMPLETE		2	/* OK to return incomplete chain */
535 
536 /*
537  * Open/close synchronization flags.
538  * These are kept in a separate field in the conn and the synchronization
539  * depends on the atomic 32 bit access to that field.
540  */
541 #define	CONN_CLOSING		0x01	/* ip_close waiting for ip_wsrv */
542 #define	CONN_CONDEMNED		0x02	/* conn is closing, no more refs */
543 #define	CONN_INCIPIENT		0x04	/* conn not yet visible, no refs */
544 #define	CONN_QUIESCED		0x08	/* conn is now quiescent */
545 #define	CONN_UPDATE_ILL		0x10	/* conn_update_ill in progress */
546 
547 /*
548  * Flags for dce_flags field. Specifies which information has been set.
549  * dce_ident is always present, but the other ones are identified by the flags.
550  */
551 #define	DCEF_DEFAULT		0x0001	/* Default DCE - no pmtu or uinfo */
552 #define	DCEF_PMTU		0x0002	/* Different than interface MTU */
553 #define	DCEF_UINFO		0x0004	/* dce_uinfo set */
554 #define	DCEF_TOO_SMALL_PMTU	0x0008	/* Smaller than IPv4 MIN */
555 
556 #ifdef _KERNEL
557 /*
558  * Extra structures need for per-src-addr filtering (IGMPv3/MLDv2)
559  */
560 #define	MAX_FILTER_SIZE	64
561 
562 typedef struct slist_s {
563 	int		sl_numsrc;
564 	in6_addr_t	sl_addr[MAX_FILTER_SIZE];
565 } slist_t;
566 
567 /*
568  * Following struct is used to maintain retransmission state for
569  * a multicast group.  One rtx_state_t struct is an in-line field
570  * of the ilm_t struct; the slist_ts in the rtx_state_t struct are
571  * alloc'd as needed.
572  */
573 typedef struct rtx_state_s {
574 	uint_t		rtx_timer;	/* retrans timer */
575 	int		rtx_cnt;	/* retrans count */
576 	int		rtx_fmode_cnt;	/* retrans count for fmode change */
577 	slist_t		*rtx_allow;
578 	slist_t		*rtx_block;
579 } rtx_state_t;
580 
581 /*
582  * Used to construct list of multicast address records that will be
583  * sent in a single listener report.
584  */
585 typedef struct mrec_s {
586 	struct mrec_s	*mrec_next;
587 	uint8_t		mrec_type;
588 	uint8_t		mrec_auxlen;	/* currently unused */
589 	in6_addr_t	mrec_group;
590 	slist_t		mrec_srcs;
591 } mrec_t;
592 
593 /* Group membership list per upper conn */
594 
595 /*
596  * We record the multicast information from the socket option in
597  * ilg_ifaddr/ilg_ifindex. This allows rejoining the group in the case when
598  * the ifaddr (or ifindex) disappears and later reappears, potentially on
599  * a different ill. The IPv6 multicast socket options and ioctls all specify
600  * the interface using an ifindex. For IPv4 some socket options/ioctls use
601  * the interface address and others use the index. We record here the method
602  * that was actually used (and leave the other of ilg_ifaddr or ilg_ifindex)
603  * at zero so that we can rejoin the way the application intended.
604  *
605  * We track the ill on which we will or already have joined an ilm using
606  * ilg_ill. When we have succeeded joining the ilm and have a refhold on it
607  * then we set ilg_ilm. Thus intentionally there is a window where ilg_ill is
608  * set and ilg_ilm is not set. This allows clearing ilg_ill as a signal that
609  * the ill is being unplumbed and the ilm should be discarded.
610  *
611  * ilg records the state of multicast memberships of a socket end point.
612  * ilm records the state of multicast memberships with the driver and is
613  * maintained per interface.
614  *
615  * The ilg state is protected by conn_ilg_lock.
616  * The ilg will not be freed until ilg_refcnt drops to zero.
617  */
618 typedef struct ilg_s {
619 	struct ilg_s	*ilg_next;
620 	struct ilg_s	**ilg_ptpn;
621 	struct conn_s	*ilg_connp;	/* Back pointer to get lock */
622 	in6_addr_t	ilg_v6group;
623 	ipaddr_t	ilg_ifaddr;	/* For some IPv4 cases */
624 	uint_t		ilg_ifindex;	/* IPv6 and some other IPv4 cases */
625 	struct ill_s	*ilg_ill;	/* Where ilm is joined. No refhold */
626 	struct ilm_s	*ilg_ilm;	/* With ilm_refhold */
627 	uint_t		ilg_refcnt;
628 	mcast_record_t	ilg_fmode;	/* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */
629 	slist_t		*ilg_filter;
630 	boolean_t	ilg_condemned;	/* Conceptually deleted */
631 } ilg_t;
632 
633 /*
634  * Multicast address list entry for ill.
635  * ilm_ill is used by IPv4 and IPv6
636  *
637  * The ilm state (and other multicast state on the ill) is protected by
638  * ill_mcast_lock. Operations that change state on both an ilg and ilm
639  * in addition use ill_mcast_serializer to ensure that we can't have
640  * interleaving between e.g., add and delete operations for the same conn_t,
641  * group, and ill. The ill_mcast_serializer is also used to ensure that
642  * multicast group joins do not occur on an interface that is in the process
643  * of joining an IPMP group.
644  *
645  * The comment below (and for other netstack_t references) refers
646  * to the fact that we only do netstack_hold in particular cases,
647  * such as the references from open endpoints (ill_t and conn_t's
648  * pointers). Internally within IP we rely on IP's ability to cleanup e.g.
649  * ire_t's when an ill goes away.
650  */
651 typedef struct ilm_s {
652 	in6_addr_t	ilm_v6addr;
653 	int		ilm_refcnt;
654 	uint_t		ilm_timer;	/* IGMP/MLD query resp timer, in msec */
655 	struct ilm_s	*ilm_next;	/* Linked list for each ill */
656 	uint_t		ilm_state;	/* state of the membership */
657 	struct ill_s	*ilm_ill;	/* Back pointer to ill - ill_ilm_cnt */
658 	zoneid_t	ilm_zoneid;
659 	int		ilm_no_ilg_cnt;	/* number of joins w/ no ilg */
660 	mcast_record_t	ilm_fmode;	/* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */
661 	slist_t		*ilm_filter;	/* source filter list */
662 	slist_t		*ilm_pendsrcs;	/* relevant src addrs for pending req */
663 	rtx_state_t	ilm_rtx;	/* SCR retransmission state */
664 	ipaddr_t	ilm_ifaddr;	/* For IPv4 netstat */
665 	ip_stack_t	*ilm_ipst;	/* Does not have a netstack_hold */
666 } ilm_t;
667 
668 #define	ilm_addr	V4_PART_OF_V6(ilm_v6addr)
669 
670 /*
671  * Soft reference to an IPsec SA.
672  *
673  * On relative terms, conn's can be persistent (living as long as the
674  * processes which create them), while SA's are ephemeral (dying when
675  * they hit their time-based or byte-based lifetimes).
676  *
677  * We could hold a hard reference to an SA from an ipsec_latch_t,
678  * but this would cause expired SA's to linger for a potentially
679  * unbounded time.
680  *
681  * Instead, we remember the hash bucket number and bucket generation
682  * in addition to the pointer.  The bucket generation is incremented on
683  * each deletion.
684  */
685 typedef struct ipsa_ref_s
686 {
687 	struct ipsa_s	*ipsr_sa;
688 	struct isaf_s	*ipsr_bucket;
689 	uint64_t	ipsr_gen;
690 } ipsa_ref_t;
691 
692 /*
693  * IPsec "latching" state.
694  *
695  * In the presence of IPsec policy, fully-bound conn's bind a connection
696  * to more than just the 5-tuple, but also a specific IPsec action and
697  * identity-pair.
698  * The identity pair is accessed from both the receive and transmit side
699  * hence it is maintained in the ipsec_latch_t structure. conn_latch and
700  * ixa_ipsec_latch points to it.
701  * The policy and actions are stored in conn_latch_in_policy and
702  * conn_latch_in_action for the inbound side, and in ixa_ipsec_policy and
703  * ixa_ipsec_action for the transmit side.
704  *
705  * As an optimization, we also cache soft references to IPsec SA's in
706  * ip_xmit_attr_t so that we can fast-path around most of the work needed for
707  * outbound IPsec SA selection.
708  */
709 typedef struct ipsec_latch_s
710 {
711 	kmutex_t	ipl_lock;
712 	uint32_t	ipl_refcnt;
713 
714 	struct ipsid_s	*ipl_local_cid;
715 	struct ipsid_s	*ipl_remote_cid;
716 	unsigned int
717 			ipl_ids_latched : 1,
718 
719 			ipl_pad_to_bit_31 : 31;
720 } ipsec_latch_t;
721 
722 #define	IPLATCH_REFHOLD(ipl) { \
723 	atomic_inc_32(&(ipl)->ipl_refcnt);		\
724 	ASSERT((ipl)->ipl_refcnt != 0);			\
725 }
726 
727 #define	IPLATCH_REFRELE(ipl) {				\
728 	ASSERT((ipl)->ipl_refcnt != 0);				\
729 	membar_exit();						\
730 	if (atomic_dec_32_nv(&(ipl)->ipl_refcnt) == 0)	\
731 		iplatch_free(ipl);				\
732 }
733 
734 /*
735  * peer identity structure.
736  */
737 typedef struct conn_s conn_t;
738 
739 /*
740  * This is used to match an inbound/outbound datagram with policy.
741  */
742 typedef	struct ipsec_selector {
743 	in6_addr_t	ips_local_addr_v6;
744 	in6_addr_t	ips_remote_addr_v6;
745 	uint16_t	ips_local_port;
746 	uint16_t	ips_remote_port;
747 	uint8_t		ips_icmp_type;
748 	uint8_t		ips_icmp_code;
749 	uint8_t		ips_protocol;
750 	uint8_t		ips_isv4 : 1,
751 			ips_is_icmp_inv_acq: 1;
752 } ipsec_selector_t;
753 
754 /*
755  * Note that we put v4 addresses in the *first* 32-bit word of the
756  * selector rather than the last to simplify the prefix match/mask code
757  * in spd.c
758  */
759 #define	ips_local_addr_v4 ips_local_addr_v6.s6_addr32[0]
760 #define	ips_remote_addr_v4 ips_remote_addr_v6.s6_addr32[0]
761 
762 /* Values used in IP by IPSEC Code */
763 #define		IPSEC_OUTBOUND		B_TRUE
764 #define		IPSEC_INBOUND		B_FALSE
765 
766 /*
767  * There are two variants in policy failures. The packet may come in
768  * secure when not needed (IPSEC_POLICY_???_NOT_NEEDED) or it may not
769  * have the desired level of protection (IPSEC_POLICY_MISMATCH).
770  */
771 #define	IPSEC_POLICY_NOT_NEEDED		0
772 #define	IPSEC_POLICY_MISMATCH		1
773 #define	IPSEC_POLICY_AUTH_NOT_NEEDED	2
774 #define	IPSEC_POLICY_ENCR_NOT_NEEDED	3
775 #define	IPSEC_POLICY_SE_NOT_NEEDED	4
776 #define	IPSEC_POLICY_MAX		5	/* Always max + 1. */
777 
778 /*
779  * Check with IPSEC inbound policy if
780  *
781  * 1) per-socket policy is present - indicated by conn_in_enforce_policy.
782  * 2) Or if we have not cached policy on the conn and the global policy is
783  *    non-empty.
784  */
785 #define	CONN_INBOUND_POLICY_PRESENT(connp, ipss)	\
786 	((connp)->conn_in_enforce_policy ||		\
787 	(!((connp)->conn_policy_cached) &&		\
788 	(ipss)->ipsec_inbound_v4_policy_present))
789 
790 #define	CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)	\
791 	((connp)->conn_in_enforce_policy ||		\
792 	(!(connp)->conn_policy_cached &&		\
793 	(ipss)->ipsec_inbound_v6_policy_present))
794 
795 #define	CONN_OUTBOUND_POLICY_PRESENT(connp, ipss)	\
796 	((connp)->conn_out_enforce_policy ||		\
797 	(!((connp)->conn_policy_cached) &&		\
798 	(ipss)->ipsec_outbound_v4_policy_present))
799 
800 #define	CONN_OUTBOUND_POLICY_PRESENT_V6(connp, ipss)	\
801 	((connp)->conn_out_enforce_policy ||		\
802 	(!(connp)->conn_policy_cached &&		\
803 	(ipss)->ipsec_outbound_v6_policy_present))
804 
805 /*
806  * Information cached in IRE for upper layer protocol (ULP).
807  */
808 typedef struct iulp_s {
809 	boolean_t	iulp_set;	/* Is any metric set? */
810 	uint32_t	iulp_ssthresh;	/* Slow start threshold (TCP). */
811 	clock_t		iulp_rtt;	/* Guestimate in millisecs. */
812 	clock_t		iulp_rtt_sd;	/* Cached value of RTT variance. */
813 	uint32_t	iulp_spipe;	/* Send pipe size. */
814 	uint32_t	iulp_rpipe;	/* Receive pipe size. */
815 	uint32_t	iulp_rtomax;	/* Max round trip timeout. */
816 	uint32_t	iulp_sack;	/* Use SACK option (TCP)? */
817 	uint32_t	iulp_mtu;	/* Setable with routing sockets */
818 
819 	uint32_t
820 		iulp_tstamp_ok : 1,	/* Use timestamp option (TCP)? */
821 		iulp_wscale_ok : 1,	/* Use window scale option (TCP)? */
822 		iulp_ecn_ok : 1,	/* Enable ECN (for TCP)? */
823 		iulp_pmtud_ok : 1,	/* Enable PMTUd? */
824 
825 		/* These three are passed out by ip_set_destination */
826 		iulp_localnet: 1,	/* IRE_ONLINK */
827 		iulp_loopback: 1,	/* IRE_LOOPBACK */
828 		iulp_local: 1,		/* IRE_LOCAL */
829 
830 		iulp_not_used : 25;
831 } iulp_t;
832 
833 /*
834  * The conn drain list structure (idl_t), protected by idl_lock.  Each conn_t
835  * inserted in the list points back at this idl_t using conn_idl, and is
836  * chained by conn_drain_next and conn_drain_prev, which are also protected by
837  * idl_lock.  When flow control is relieved, either ip_wsrv() (STREAMS) or
838  * ill_flow_enable() (non-STREAMS) will call conn_drain().
839  *
840  * The conn drain list, idl_t, itself is part of tx cookie list structure.
841  * A tx cookie list points to a blocked Tx ring and contains the list of
842  * all conn's that are blocked due to the flow-controlled Tx ring (via
843  * the idl drain list). Note that a link can have multiple Tx rings. The
844  * drain list will store the conn's blocked due to Tx ring being flow
845  * controlled.
846  */
847 
848 typedef uintptr_t ip_mac_tx_cookie_t;
849 typedef	struct idl_s idl_t;
850 typedef	struct idl_tx_list_s idl_tx_list_t;
851 
852 struct idl_tx_list_s {
853 	ip_mac_tx_cookie_t	txl_cookie;
854 	kmutex_t		txl_lock;	/* Lock for this list */
855 	idl_t			*txl_drain_list;
856 	int			txl_drain_index;
857 };
858 
859 struct idl_s {
860 	conn_t		*idl_conn;		/* Head of drain list */
861 	kmutex_t	idl_lock;		/* Lock for this list */
862 	idl_tx_list_t	*idl_itl;
863 };
864 
865 /*
866  * Interface route structure which holds the necessary information to recreate
867  * routes that are tied to an interface i.e. have ire_ill set.
868  *
869  * These routes which were initially created via a routing socket or via the
870  * SIOCADDRT ioctl may be gateway routes (RTF_GATEWAY being set) or may be
871  * traditional interface routes.  When an ill comes back up after being
872  * down, this information will be used to recreate the routes.  These
873  * are part of an mblk_t chain that hangs off of the ILL (ill_saved_ire_mp).
874  */
875 typedef struct ifrt_s {
876 	ushort_t	ifrt_type;		/* Type of IRE */
877 	in6_addr_t	ifrt_v6addr;		/* Address IRE represents. */
878 	in6_addr_t	ifrt_v6gateway_addr;	/* Gateway if IRE_OFFLINK */
879 	in6_addr_t	ifrt_v6setsrc_addr;	/* Src addr if RTF_SETSRC */
880 	in6_addr_t	ifrt_v6mask;		/* Mask for matching IRE. */
881 	uint32_t	ifrt_flags;		/* flags related to route */
882 	iulp_t		ifrt_metrics;		/* Routing socket metrics */
883 	zoneid_t	ifrt_zoneid;		/* zoneid for route */
884 } ifrt_t;
885 
886 #define	ifrt_addr		V4_PART_OF_V6(ifrt_v6addr)
887 #define	ifrt_gateway_addr	V4_PART_OF_V6(ifrt_v6gateway_addr)
888 #define	ifrt_mask		V4_PART_OF_V6(ifrt_v6mask)
889 #define	ifrt_setsrc_addr	V4_PART_OF_V6(ifrt_v6setsrc_addr)
890 
891 /* Number of IP addresses that can be hosted on a physical interface */
892 #define	MAX_ADDRS_PER_IF	8192
893 /*
894  * Number of Source addresses to be considered for source address
895  * selection. Used by ipif_select_source_v4/v6.
896  */
897 #define	MAX_IPIF_SELECT_SOURCE	50
898 
899 #ifdef IP_DEBUG
900 /*
901  * Trace refholds and refreles for debugging.
902  */
903 #define	TR_STACK_DEPTH	14
904 typedef struct tr_buf_s {
905 	int	tr_depth;
906 	clock_t	tr_time;
907 	pc_t	tr_stack[TR_STACK_DEPTH];
908 } tr_buf_t;
909 
910 typedef struct th_trace_s {
911 	int		th_refcnt;
912 	uint_t		th_trace_lastref;
913 	kthread_t	*th_id;
914 #define	TR_BUF_MAX	38
915 	tr_buf_t	th_trbuf[TR_BUF_MAX];
916 } th_trace_t;
917 
918 typedef struct th_hash_s {
919 	list_node_t	thh_link;
920 	mod_hash_t	*thh_hash;
921 	ip_stack_t	*thh_ipst;
922 } th_hash_t;
923 #endif
924 
925 /* The following are ipif_state_flags */
926 #define	IPIF_CONDEMNED		0x1	/* The ipif is being removed */
927 #define	IPIF_CHANGING		0x2	/* A critcal ipif field is changing */
928 #define	IPIF_SET_LINKLOCAL	0x10	/* transient flag during bringup */
929 
930 /* IP interface structure, one per local address */
931 typedef struct ipif_s {
932 	struct	ipif_s	*ipif_next;
933 	struct	ill_s	*ipif_ill;	/* Back pointer to our ill */
934 	int	ipif_id;		/* Logical unit number */
935 	in6_addr_t ipif_v6lcl_addr;	/* Local IP address for this if. */
936 	in6_addr_t ipif_v6subnet;	/* Subnet prefix for this if. */
937 	in6_addr_t ipif_v6net_mask;	/* Net mask for this interface. */
938 	in6_addr_t ipif_v6brd_addr;	/* Broadcast addr for this interface. */
939 	in6_addr_t ipif_v6pp_dst_addr;	/* Point-to-point dest address. */
940 	uint64_t ipif_flags;		/* Interface flags. */
941 	uint_t	ipif_ire_type;		/* IRE_LOCAL or IRE_LOOPBACK */
942 
943 	/*
944 	 * The packet count in the ipif contain the sum of the
945 	 * packet counts in dead IRE_LOCAL/LOOPBACK for this ipif.
946 	 */
947 	uint_t	ipif_ib_pkt_count;	/* Inbound packets for our dead IREs */
948 
949 	/* Exclusive bit fields, protected by ipsq_t */
950 	unsigned int
951 		ipif_was_up : 1,	/* ipif was up before */
952 		ipif_addr_ready : 1,	/* DAD is done */
953 		ipif_was_dup : 1,	/* DAD had failed */
954 		ipif_added_nce : 1,	/* nce added for local address */
955 
956 		ipif_pad_to_31 : 28;
957 
958 	ilm_t	*ipif_allhosts_ilm;	/* For all-nodes join */
959 	ilm_t	*ipif_solmulti_ilm;	/* For IPv6 solicited multicast join */
960 
961 	uint_t	ipif_seqid;		/* unique index across all ills */
962 	uint_t	ipif_state_flags;	/* See IPIF_* flag defs above */
963 	uint_t	ipif_refcnt;		/* active consistent reader cnt */
964 
965 	zoneid_t ipif_zoneid;		/* zone ID number */
966 	timeout_id_t ipif_recovery_id;	/* Timer for DAD recovery */
967 	boolean_t ipif_trace_disable;	/* True when alloc fails */
968 	/*
969 	 * For an IPMP interface, ipif_bound_ill tracks the ill whose hardware
970 	 * information this ipif is associated with via ARP/NDP.  We can use
971 	 * an ill pointer (rather than an index) because only ills that are
972 	 * part of a group will be pointed to, and an ill cannot disappear
973 	 * while it's in a group.
974 	 */
975 	struct ill_s    *ipif_bound_ill;
976 	struct ipif_s   *ipif_bound_next; /* bound ipif chain */
977 	boolean_t	ipif_bound;	/* B_TRUE if we successfully bound */
978 
979 	struct ire_s	*ipif_ire_local; /* Our IRE_LOCAL or LOOPBACK */
980 	struct ire_s	*ipif_ire_if;	 /* Our IRE_INTERFACE */
981 } ipif_t;
982 
983 /*
984  * The following table lists the protection levels of the various members
985  * of the ipif_t. The following notation is used.
986  *
987  * Write once - Written to only once at the time of bringing up
988  * the interface and can be safely read after the bringup without any lock.
989  *
990  * ipsq - Need to execute in the ipsq to perform the indicated access.
991  *
992  * ill_lock - Need to hold this mutex to perform the indicated access.
993  *
994  * ill_g_lock - Need to hold this rw lock as reader/writer for read access or
995  * write access respectively.
996  *
997  * down ill - Written to only when the ill is down (i.e all ipifs are down)
998  * up ill - Read only when the ill is up (i.e. at least 1 ipif is up)
999  *
1000  *		 Table of ipif_t members and their protection
1001  *
1002  * ipif_next		ipsq + ill_lock +	ipsq OR ill_lock OR
1003  *			ill_g_lock		ill_g_lock
1004  * ipif_ill		ipsq + down ipif	write once
1005  * ipif_id		ipsq + down ipif	write once
1006  * ipif_v6lcl_addr	ipsq + down ipif	up ipif
1007  * ipif_v6subnet	ipsq + down ipif	up ipif
1008  * ipif_v6net_mask	ipsq + down ipif	up ipif
1009  *
1010  * ipif_v6brd_addr
1011  * ipif_v6pp_dst_addr
1012  * ipif_flags		ill_lock		ill_lock
1013  * ipif_ire_type	ipsq + down ill		up ill
1014  *
1015  * ipif_ib_pkt_count	Approx
1016  *
1017  * bit fields		ill_lock		ill_lock
1018  *
1019  * ipif_allhosts_ilm	ipsq			ipsq
1020  * ipif_solmulti_ilm	ipsq			ipsq
1021  *
1022  * ipif_seqid		ipsq			Write once
1023  *
1024  * ipif_state_flags	ill_lock		ill_lock
1025  * ipif_refcnt		ill_lock		ill_lock
1026  * ipif_bound_ill	ipsq + ipmp_lock	ipsq OR ipmp_lock
1027  * ipif_bound_next	ipsq			ipsq
1028  * ipif_bound		ipsq			ipsq
1029  *
1030  * ipif_ire_local	ipsq + ips_ill_g_lock	ipsq OR ips_ill_g_lock
1031  * ipif_ire_if		ipsq + ips_ill_g_lock	ipsq OR ips_ill_g_lock
1032  */
1033 
1034 /*
1035  * Return values from ip_laddr_verify_{v4,v6}
1036  */
1037 typedef enum { IPVL_UNICAST_UP, IPVL_UNICAST_DOWN, IPVL_MCAST, IPVL_BCAST,
1038 	    IPVL_BAD} ip_laddr_t;
1039 
1040 
1041 #define	IP_TR_HASH(tid)	((((uintptr_t)tid) >> 6) & (IP_TR_HASH_MAX - 1))
1042 
1043 #ifdef DEBUG
1044 #define	IPIF_TRACE_REF(ipif)	ipif_trace_ref(ipif)
1045 #define	ILL_TRACE_REF(ill)	ill_trace_ref(ill)
1046 #define	IPIF_UNTRACE_REF(ipif)	ipif_untrace_ref(ipif)
1047 #define	ILL_UNTRACE_REF(ill)	ill_untrace_ref(ill)
1048 #else
1049 #define	IPIF_TRACE_REF(ipif)
1050 #define	ILL_TRACE_REF(ill)
1051 #define	IPIF_UNTRACE_REF(ipif)
1052 #define	ILL_UNTRACE_REF(ill)
1053 #endif
1054 
1055 /* IPv4 compatibility macros */
1056 #define	ipif_lcl_addr		V4_PART_OF_V6(ipif_v6lcl_addr)
1057 #define	ipif_subnet		V4_PART_OF_V6(ipif_v6subnet)
1058 #define	ipif_net_mask		V4_PART_OF_V6(ipif_v6net_mask)
1059 #define	ipif_brd_addr		V4_PART_OF_V6(ipif_v6brd_addr)
1060 #define	ipif_pp_dst_addr	V4_PART_OF_V6(ipif_v6pp_dst_addr)
1061 
1062 /* Macros for easy backreferences to the ill. */
1063 #define	ipif_isv6		ipif_ill->ill_isv6
1064 
1065 #define	SIOCLIFADDR_NDX 112	/* ndx of SIOCLIFADDR in the ndx ioctl table */
1066 
1067 /*
1068  * mode value for ip_ioctl_finish for finishing an ioctl
1069  */
1070 #define	CONN_CLOSE	1		/* No mi_copy */
1071 #define	COPYOUT		2		/* do an mi_copyout if needed */
1072 #define	NO_COPYOUT	3		/* do an mi_copy_done */
1073 #define	IPI2MODE(ipi)	((ipi)->ipi_flags & IPI_GET_CMD ? COPYOUT : NO_COPYOUT)
1074 
1075 /*
1076  * The IP-MT design revolves around the serialization objects ipsq_t (IPSQ)
1077  * and ipxop_t (exclusive operation or "xop").  Becoming "writer" on an IPSQ
1078  * ensures that no other threads can become "writer" on any IPSQs sharing that
1079  * IPSQ's xop until the writer thread is done.
1080  *
1081  * Each phyint points to one IPSQ that remains fixed over the phyint's life.
1082  * Each IPSQ points to one xop that can change over the IPSQ's life.  If a
1083  * phyint is *not* in an IPMP group, then its IPSQ will refer to the IPSQ's
1084  * "own" xop (ipsq_ownxop).  If a phyint *is* part of an IPMP group, then its
1085  * IPSQ will refer to the "group" xop, which is shorthand for the xop of the
1086  * IPSQ of the IPMP meta-interface's phyint.  Thus, all phyints that are part
1087  * of the same IPMP group will have their IPSQ's point to the group xop, and
1088  * thus becoming "writer" on any phyint in the group will prevent any other
1089  * writer on any other phyint in the group.  All IPSQs sharing the same xop
1090  * are chained together through ipsq_next (in the degenerate common case,
1091  * ipsq_next simply refers to itself).  Note that the group xop is guaranteed
1092  * to exist at least as long as there are members in the group, since the IPMP
1093  * meta-interface can only be destroyed if the group is empty.
1094  *
1095  * Incoming exclusive operation requests are enqueued on the IPSQ they arrived
1096  * on rather than the xop.  This makes switching xop's (as would happen when a
1097  * phyint leaves an IPMP group) simple, because after the phyint leaves the
1098  * group, any operations enqueued on its IPSQ can be safely processed with
1099  * respect to its new xop, and any operations enqueued on the IPSQs of its
1100  * former group can be processed with respect to their existing group xop.
1101  * Even so, switching xops is a subtle dance; see ipsq_dq() for details.
1102  *
1103  * An IPSQ's "own" xop is embedded within the IPSQ itself since they have have
1104  * identical lifetimes, and because doing so simplifies pointer management.
1105  * While each phyint and IPSQ point to each other, it is not possible to free
1106  * the IPSQ when the phyint is freed, since we may still *inside* the IPSQ
1107  * when the phyint is being freed.  Thus, ipsq_phyint is set to NULL when the
1108  * phyint is freed, and the IPSQ free is later done in ipsq_exit().
1109  *
1110  * ipsq_t synchronization:	read			write
1111  *
1112  *	ipsq_xopq_mphead	ipx_lock		ipx_lock
1113  *	ipsq_xopq_mptail	ipx_lock		ipx_lock
1114  *	ipsq_xop_switch_mp	ipsq_lock		ipsq_lock
1115  *	ipsq_phyint		write once		write once
1116  *	ipsq_next		RW_READER ill_g_lock	RW_WRITER ill_g_lock
1117  *	ipsq_xop		ipsq_lock or ipsq	ipsq_lock + ipsq
1118  *	ipsq_swxop		ipsq			ipsq
1119  *	ipsq_ownxop		see ipxop_t		see ipxop_t
1120  *	ipsq_ipst		write once		write once
1121  *
1122  * ipxop_t synchronization:     read			write
1123  *
1124  *	ipx_writer		ipx_lock		ipx_lock
1125  *	ipx_xop_queued		ipx_lock		ipx_lock
1126  *	ipx_mphead		ipx_lock		ipx_lock
1127  *	ipx_mptail		ipx_lock		ipx_lock
1128  *	ipx_ipsq		write once		write once
1129  *	ips_ipsq_queued		ipx_lock		ipx_lock
1130  *	ipx_waitfor		ipsq or ipx_lock	ipsq + ipx_lock
1131  *	ipx_reentry_cnt		ipsq or ipx_lock	ipsq + ipx_lock
1132  *	ipx_current_done	ipsq			ipsq
1133  *	ipx_current_ioctl	ipsq			ipsq
1134  *	ipx_current_ipif	ipsq or ipx_lock	ipsq + ipx_lock
1135  *	ipx_pending_ipif	ipsq or ipx_lock	ipsq + ipx_lock
1136  *	ipx_pending_mp		ipsq or ipx_lock	ipsq + ipx_lock
1137  *	ipx_forced		ipsq			ipsq
1138  *	ipx_depth		ipsq			ipsq
1139  *	ipx_stack		ipsq			ipsq
1140  */
1141 typedef struct ipxop_s {
1142 	kmutex_t	ipx_lock;	/* see above */
1143 	kthread_t	*ipx_writer;	/* current owner */
1144 	mblk_t		*ipx_mphead;	/* messages tied to this op */
1145 	mblk_t		*ipx_mptail;
1146 	struct ipsq_s	*ipx_ipsq;	/* associated ipsq */
1147 	boolean_t	ipx_ipsq_queued; /* ipsq using xop has queued op */
1148 	int		ipx_waitfor;	/* waiting; values encoded below */
1149 	int		ipx_reentry_cnt;
1150 	boolean_t	ipx_current_done;  /* is the current operation done? */
1151 	int		ipx_current_ioctl; /* current ioctl, or 0 if no ioctl */
1152 	ipif_t		*ipx_current_ipif; /* ipif for current op */
1153 	ipif_t		*ipx_pending_ipif; /* ipif for ipx_pending_mp */
1154 	mblk_t		*ipx_pending_mp; /* current ioctl mp while waiting */
1155 	boolean_t	ipx_forced;			/* debugging aid */
1156 #ifdef DEBUG
1157 	int		ipx_depth;			/* debugging aid */
1158 #define	IPX_STACK_DEPTH	15
1159 	pc_t		ipx_stack[IPX_STACK_DEPTH];	/* debugging aid */
1160 #endif
1161 } ipxop_t;
1162 
1163 typedef struct ipsq_s {
1164 	kmutex_t ipsq_lock;		/* see above */
1165 	mblk_t	*ipsq_switch_mp;	/* op to handle right after switch */
1166 	mblk_t	*ipsq_xopq_mphead;	/* list of excl ops (mostly ioctls) */
1167 	mblk_t	*ipsq_xopq_mptail;
1168 	struct phyint	*ipsq_phyint;	/* associated phyint */
1169 	struct ipsq_s	*ipsq_next;	/* next ipsq sharing ipsq_xop */
1170 	struct ipxop_s	*ipsq_xop;	/* current xop synchronization info */
1171 	struct ipxop_s	*ipsq_swxop;	/* switch xop to on ipsq_exit() */
1172 	struct ipxop_s	ipsq_ownxop;	/* our own xop (may not be in-use) */
1173 	ip_stack_t	*ipsq_ipst;	/* does not have a netstack_hold */
1174 } ipsq_t;
1175 
1176 /*
1177  * ipx_waitfor values:
1178  */
1179 enum {
1180 	IPIF_DOWN = 1,	/* ipif_down() waiting for refcnts to drop */
1181 	ILL_DOWN,	/* ill_down() waiting for refcnts to drop */
1182 	IPIF_FREE,	/* ipif_free() waiting for refcnts to drop */
1183 	ILL_FREE	/* ill unplumb waiting for refcnts to drop */
1184 };
1185 
1186 /* Operation types for ipsq_try_enter() */
1187 #define	CUR_OP 0	/* request writer within current operation */
1188 #define	NEW_OP 1	/* request writer for a new operation */
1189 #define	SWITCH_OP 2	/* request writer once IPSQ XOP switches */
1190 
1191 /*
1192  * Kstats tracked on each IPMP meta-interface.  Order here must match
1193  * ipmp_kstats[] in ip/ipmp.c.
1194  */
1195 enum {
1196 	IPMP_KSTAT_OBYTES,	IPMP_KSTAT_OBYTES64,	IPMP_KSTAT_RBYTES,
1197 	IPMP_KSTAT_RBYTES64,	IPMP_KSTAT_OPACKETS,	IPMP_KSTAT_OPACKETS64,
1198 	IPMP_KSTAT_OERRORS,	IPMP_KSTAT_IPACKETS,	IPMP_KSTAT_IPACKETS64,
1199 	IPMP_KSTAT_IERRORS,	IPMP_KSTAT_MULTIRCV,	IPMP_KSTAT_MULTIXMT,
1200 	IPMP_KSTAT_BRDCSTRCV,	IPMP_KSTAT_BRDCSTXMT,	IPMP_KSTAT_LINK_UP,
1201 	IPMP_KSTAT_MAX		/* keep last */
1202 };
1203 
1204 /*
1205  * phyint represents state that is common to both IPv4 and IPv6 interfaces.
1206  * There is a separate ill_t representing IPv4 and IPv6 which has a
1207  * backpointer to the phyint structure for accessing common state.
1208  */
1209 typedef struct phyint {
1210 	struct ill_s	*phyint_illv4;
1211 	struct ill_s	*phyint_illv6;
1212 	uint_t		phyint_ifindex;		/* SIOCSLIFINDEX */
1213 	uint64_t	phyint_flags;
1214 	avl_node_t	phyint_avl_by_index;	/* avl tree by index */
1215 	avl_node_t	phyint_avl_by_name;	/* avl tree by name */
1216 	kmutex_t	phyint_lock;
1217 	struct ipsq_s	*phyint_ipsq;		/* back pointer to ipsq */
1218 	struct ipmp_grp_s *phyint_grp;		/* associated IPMP group */
1219 	char		phyint_name[LIFNAMSIZ];	/* physical interface name */
1220 	uint64_t	phyint_kstats0[IPMP_KSTAT_MAX];	/* baseline kstats */
1221 } phyint_t;
1222 
1223 #define	CACHE_ALIGN_SIZE 64
1224 #define	CACHE_ALIGN(align_struct)	P2ROUNDUP(sizeof (struct align_struct),\
1225 							CACHE_ALIGN_SIZE)
1226 struct _phyint_list_s_ {
1227 	avl_tree_t	phyint_list_avl_by_index;	/* avl tree by index */
1228 	avl_tree_t	phyint_list_avl_by_name;	/* avl tree by name */
1229 };
1230 
1231 typedef union phyint_list_u {
1232 	struct	_phyint_list_s_ phyint_list_s;
1233 	char	phyint_list_filler[CACHE_ALIGN(_phyint_list_s_)];
1234 } phyint_list_t;
1235 
1236 #define	phyint_list_avl_by_index	phyint_list_s.phyint_list_avl_by_index
1237 #define	phyint_list_avl_by_name		phyint_list_s.phyint_list_avl_by_name
1238 
1239 /*
1240  * Fragmentation hash bucket
1241  */
1242 typedef struct ipfb_s {
1243 	struct ipf_s	*ipfb_ipf;	/* List of ... */
1244 	size_t		ipfb_count;	/* Count of bytes used by frag(s) */
1245 	kmutex_t	ipfb_lock;	/* Protect all ipf in list */
1246 	uint_t		ipfb_frag_pkts; /* num of distinct fragmented pkts */
1247 } ipfb_t;
1248 
1249 /*
1250  * IRE bucket structure. Usually there is an array of such structures,
1251  * each pointing to a linked list of ires. irb_refcnt counts the number
1252  * of walkers of a given hash bucket. Usually the reference count is
1253  * bumped up if the walker wants no IRES to be DELETED while walking the
1254  * list. Bumping up does not PREVENT ADDITION. This allows walking a given
1255  * hash bucket without stumbling up on a free pointer.
1256  *
1257  * irb_t structures in ip_ftable are dynamically allocated and freed.
1258  * In order to identify the irb_t structures that can be safely kmem_free'd
1259  * we need to ensure that
1260  *  - the irb_refcnt is quiescent, indicating no other walkers,
1261  *  - no other threads or ire's are holding references to the irb,
1262  *	i.e., irb_nire == 0,
1263  *  - there are no active ire's in the bucket, i.e., irb_ire_cnt == 0
1264  */
1265 typedef struct irb {
1266 	struct ire_s	*irb_ire;	/* First ire in this bucket */
1267 					/* Should be first in this struct */
1268 	krwlock_t	irb_lock;	/* Protect this bucket */
1269 	uint_t		irb_refcnt;	/* Protected by irb_lock */
1270 	uchar_t		irb_marks;	/* CONDEMNED ires in this bucket ? */
1271 #define	IRB_MARK_CONDEMNED	0x0001	/* Contains some IRE_IS_CONDEMNED */
1272 #define	IRB_MARK_DYNAMIC	0x0002	/* Dynamically allocated */
1273 	/* Once IPv6 uses radix then IRB_MARK_DYNAMIC will be always be set */
1274 	uint_t		irb_ire_cnt;	/* Num of active IRE in this bucket */
1275 	int		irb_nire;	/* Num of ftable ire's that ref irb */
1276 	ip_stack_t	*irb_ipst;	/* Does not have a netstack_hold */
1277 } irb_t;
1278 
1279 /*
1280  * This is the structure used to store the multicast physical addresses
1281  * that an interface has joined.
1282  * The refcnt keeps track of the number of multicast IP addresses mapping
1283  * to a physical multicast address.
1284  */
1285 typedef struct multiphysaddr_s {
1286 	struct	multiphysaddr_s  *mpa_next;
1287 	char	mpa_addr[IP_MAX_HW_LEN];
1288 	int	mpa_refcnt;
1289 } multiphysaddr_t;
1290 
1291 #define	IRB2RT(irb)	(rt_t *)((caddr_t)(irb) - offsetof(rt_t, rt_irb))
1292 
1293 /* Forward declarations */
1294 struct dce_s;
1295 typedef struct dce_s dce_t;
1296 struct ire_s;
1297 typedef struct ire_s ire_t;
1298 struct ncec_s;
1299 typedef struct ncec_s ncec_t;
1300 struct nce_s;
1301 typedef struct nce_s nce_t;
1302 struct ip_recv_attr_s;
1303 typedef struct ip_recv_attr_s ip_recv_attr_t;
1304 struct ip_xmit_attr_s;
1305 typedef struct ip_xmit_attr_s ip_xmit_attr_t;
1306 
1307 struct tsol_ire_gw_secattr_s;
1308 typedef struct tsol_ire_gw_secattr_s tsol_ire_gw_secattr_t;
1309 
1310 /*
1311  * This is a structure for a one-element route cache that is passed
1312  * by reference between ip_input and ill_inputfn.
1313  */
1314 typedef struct {
1315 	ire_t		*rtc_ire;
1316 	ipaddr_t	rtc_ipaddr;
1317 	in6_addr_t	rtc_ip6addr;
1318 } rtc_t;
1319 
1320 /*
1321  * Note: Temporarily use 64 bits, and will probably go back to 32 bits after
1322  * more cleanup work is done.
1323  */
1324 typedef uint64_t iaflags_t;
1325 
1326 /* The ill input function pointer type */
1327 typedef void (*pfillinput_t)(mblk_t *, void *, void *, ip_recv_attr_t *,
1328     rtc_t *);
1329 
1330 /* The ire receive function pointer type */
1331 typedef void (*pfirerecv_t)(ire_t *, mblk_t *, void *, ip_recv_attr_t *);
1332 
1333 /* The ire send and postfrag function pointer types */
1334 typedef int (*pfiresend_t)(ire_t *, mblk_t *, void *,
1335     ip_xmit_attr_t *, uint32_t *);
1336 typedef int (*pfirepostfrag_t)(mblk_t *, nce_t *, iaflags_t, uint_t, uint32_t,
1337     zoneid_t, zoneid_t, uintptr_t *);
1338 
1339 
1340 #define	IP_V4_G_HEAD	0
1341 #define	IP_V6_G_HEAD	1
1342 
1343 #define	MAX_G_HEADS	2
1344 
1345 /*
1346  * unpadded ill_if structure
1347  */
1348 struct	_ill_if_s_ {
1349 	union ill_if_u	*illif_next;
1350 	union ill_if_u	*illif_prev;
1351 	avl_tree_t	illif_avl_by_ppa;	/* AVL tree sorted on ppa */
1352 	vmem_t		*illif_ppa_arena;	/* ppa index space */
1353 	uint16_t	illif_mcast_v1;		/* hints for		  */
1354 	uint16_t	illif_mcast_v2;		/* [igmp|mld]_slowtimo	  */
1355 	int		illif_name_len;		/* name length */
1356 	char		illif_name[LIFNAMSIZ];	/* name of interface type */
1357 };
1358 
1359 /* cache aligned ill_if structure */
1360 typedef union	ill_if_u {
1361 	struct	_ill_if_s_ ill_if_s;
1362 	char	illif_filler[CACHE_ALIGN(_ill_if_s_)];
1363 } ill_if_t;
1364 
1365 #define	illif_next		ill_if_s.illif_next
1366 #define	illif_prev		ill_if_s.illif_prev
1367 #define	illif_avl_by_ppa	ill_if_s.illif_avl_by_ppa
1368 #define	illif_ppa_arena		ill_if_s.illif_ppa_arena
1369 #define	illif_mcast_v1		ill_if_s.illif_mcast_v1
1370 #define	illif_mcast_v2		ill_if_s.illif_mcast_v2
1371 #define	illif_name		ill_if_s.illif_name
1372 #define	illif_name_len		ill_if_s.illif_name_len
1373 
1374 typedef struct ill_walk_context_s {
1375 	int	ctx_current_list; /* current list being searched */
1376 	int	ctx_last_list;	 /* last list to search */
1377 } ill_walk_context_t;
1378 
1379 /*
1380  * ill_g_heads structure, one for IPV4 and one for IPV6
1381  */
1382 struct _ill_g_head_s_ {
1383 	ill_if_t	*ill_g_list_head;
1384 	ill_if_t	*ill_g_list_tail;
1385 };
1386 
1387 typedef union ill_g_head_u {
1388 	struct _ill_g_head_s_ ill_g_head_s;
1389 	char	ill_g_head_filler[CACHE_ALIGN(_ill_g_head_s_)];
1390 } ill_g_head_t;
1391 
1392 #define	ill_g_list_head	ill_g_head_s.ill_g_list_head
1393 #define	ill_g_list_tail	ill_g_head_s.ill_g_list_tail
1394 
1395 #define	IP_V4_ILL_G_LIST(ipst)	\
1396 	(ipst)->ips_ill_g_heads[IP_V4_G_HEAD].ill_g_list_head
1397 #define	IP_V6_ILL_G_LIST(ipst)	\
1398 	(ipst)->ips_ill_g_heads[IP_V6_G_HEAD].ill_g_list_head
1399 #define	IP_VX_ILL_G_LIST(i, ipst)	\
1400 	(ipst)->ips_ill_g_heads[i].ill_g_list_head
1401 
1402 #define	ILL_START_WALK_V4(ctx_ptr, ipst)	\
1403 	ill_first(IP_V4_G_HEAD, IP_V4_G_HEAD, ctx_ptr, ipst)
1404 #define	ILL_START_WALK_V6(ctx_ptr, ipst)	\
1405 	ill_first(IP_V6_G_HEAD, IP_V6_G_HEAD, ctx_ptr, ipst)
1406 #define	ILL_START_WALK_ALL(ctx_ptr, ipst)	\
1407 	ill_first(MAX_G_HEADS, MAX_G_HEADS, ctx_ptr, ipst)
1408 
1409 /*
1410  * Capabilities, possible flags for ill_capabilities.
1411  */
1412 #define	ILL_CAPAB_LSO		0x04		/* Large Send Offload */
1413 #define	ILL_CAPAB_HCKSUM	0x08		/* Hardware checksumming */
1414 #define	ILL_CAPAB_ZEROCOPY	0x10		/* Zero-copy */
1415 #define	ILL_CAPAB_DLD		0x20		/* DLD capabilities */
1416 #define	ILL_CAPAB_DLD_POLL	0x40		/* Polling */
1417 #define	ILL_CAPAB_DLD_DIRECT	0x80		/* Direct function call */
1418 
1419 /*
1420  * Per-ill Hardware Checksumming capbilities.
1421  */
1422 typedef struct ill_hcksum_capab_s ill_hcksum_capab_t;
1423 
1424 /*
1425  * Per-ill Zero-copy capabilities.
1426  */
1427 typedef struct ill_zerocopy_capab_s ill_zerocopy_capab_t;
1428 
1429 /*
1430  * DLD capbilities.
1431  */
1432 typedef struct ill_dld_capab_s ill_dld_capab_t;
1433 
1434 /*
1435  * Per-ill polling resource map.
1436  */
1437 typedef struct ill_rx_ring ill_rx_ring_t;
1438 
1439 /*
1440  * Per-ill Large Send Offload capabilities.
1441  */
1442 typedef struct ill_lso_capab_s ill_lso_capab_t;
1443 
1444 /* The following are ill_state_flags */
1445 #define	ILL_LL_SUBNET_PENDING	0x01	/* Waiting for DL_INFO_ACK from drv */
1446 #define	ILL_CONDEMNED		0x02	/* No more new ref's to the ILL */
1447 #define	ILL_DL_UNBIND_IN_PROGRESS	0x04	/* UNBIND_REQ is sent */
1448 /*
1449  * ILL_DOWN_IN_PROGRESS is set to ensure the following:
1450  * - no packets are sent to the driver after the DL_UNBIND_REQ is sent,
1451  * - no longstanding references will be acquired on objects that are being
1452  *   brought down.
1453  */
1454 #define	ILL_DOWN_IN_PROGRESS	0x08
1455 
1456 /* Is this an ILL whose source address is used by other ILL's ? */
1457 #define	IS_USESRC_ILL(ill)			\
1458 	(((ill)->ill_usesrc_ifindex == 0) &&	\
1459 	((ill)->ill_usesrc_grp_next != NULL))
1460 
1461 /* Is this a client/consumer of the usesrc ILL ? */
1462 #define	IS_USESRC_CLI_ILL(ill)			\
1463 	(((ill)->ill_usesrc_ifindex != 0) &&	\
1464 	((ill)->ill_usesrc_grp_next != NULL))
1465 
1466 /* Is this an virtual network interface (vni) ILL ? */
1467 #define	IS_VNI(ill)							\
1468 	(((ill)->ill_phyint->phyint_flags & (PHYI_LOOPBACK|PHYI_VIRTUAL)) == \
1469 	PHYI_VIRTUAL)
1470 
1471 /* Is this a loopback ILL? */
1472 #define	IS_LOOPBACK(ill) \
1473 	((ill)->ill_phyint->phyint_flags & PHYI_LOOPBACK)
1474 
1475 /* Is this an IPMP meta-interface ILL? */
1476 #define	IS_IPMP(ill)							\
1477 	((ill)->ill_phyint->phyint_flags & PHYI_IPMP)
1478 
1479 /* Is this ILL under an IPMP meta-interface? (aka "in a group?") */
1480 #define	IS_UNDER_IPMP(ill)						\
1481 	((ill)->ill_grp != NULL && !IS_IPMP(ill))
1482 
1483 /* Is ill1 in the same illgrp as ill2? */
1484 #define	IS_IN_SAME_ILLGRP(ill1, ill2)					\
1485 	((ill1)->ill_grp != NULL && ((ill1)->ill_grp == (ill2)->ill_grp))
1486 
1487 /* Is ill1 on the same LAN as ill2? */
1488 #define	IS_ON_SAME_LAN(ill1, ill2)					\
1489 	((ill1) == (ill2) || IS_IN_SAME_ILLGRP(ill1, ill2))
1490 
1491 #define	ILL_OTHER(ill)							\
1492 	((ill)->ill_isv6 ? (ill)->ill_phyint->phyint_illv4 :		\
1493 	    (ill)->ill_phyint->phyint_illv6)
1494 
1495 /*
1496  * IPMP group ILL state structure -- up to two per IPMP group (V4 and V6).
1497  * Created when the V4 and/or V6 IPMP meta-interface is I_PLINK'd.  It is
1498  * guaranteed to persist while there are interfaces of that type in the group.
1499  * In general, most fields are accessed outside of the IPSQ (e.g., in the
1500  * datapath), and thus use locks in addition to the IPSQ for protection.
1501  *
1502  * synchronization:		read			write
1503  *
1504  *	ig_if			ipsq or ill_g_lock	ipsq and ill_g_lock
1505  *	ig_actif		ipsq or ipmp_lock	ipsq and ipmp_lock
1506  *	ig_nactif		ipsq or ipmp_lock	ipsq and ipmp_lock
1507  *	ig_next_ill		ipsq or ipmp_lock	ipsq and ipmp_lock
1508  *	ig_ipmp_ill		write once		write once
1509  *	ig_cast_ill		ipsq or ipmp_lock	ipsq and ipmp_lock
1510  *	ig_arpent		ipsq			ipsq
1511  *	ig_mtu			ipsq			ipsq
1512  *	ig_mc_mtu		ipsq			ipsq
1513  */
1514 typedef struct ipmp_illgrp_s {
1515 	list_t		ig_if;		/* list of all interfaces */
1516 	list_t		ig_actif;	/* list of active interfaces */
1517 	uint_t		ig_nactif;	/* number of active interfaces */
1518 	struct ill_s	*ig_next_ill;	/* next active interface to use */
1519 	struct ill_s	*ig_ipmp_ill;	/* backpointer to IPMP meta-interface */
1520 	struct ill_s	*ig_cast_ill;	/* nominated ill for multi/broadcast */
1521 	list_t		ig_arpent;	/* list of ARP entries */
1522 	uint_t		ig_mtu;		/* ig_ipmp_ill->ill_mtu */
1523 	uint_t		ig_mc_mtu;	/* ig_ipmp_ill->ill_mc_mtu */
1524 } ipmp_illgrp_t;
1525 
1526 /*
1527  * IPMP group state structure -- one per IPMP group.  Created when the
1528  * IPMP meta-interface is plumbed; it is guaranteed to persist while there
1529  * are interfaces in it.
1530  *
1531  * ipmp_grp_t synchronization:		read			write
1532  *
1533  *	gr_name				ipmp_lock		ipmp_lock
1534  *	gr_ifname			write once		write once
1535  *	gr_mactype			ipmp_lock		ipmp_lock
1536  *	gr_phyint			write once		write once
1537  *	gr_nif				ipmp_lock		ipmp_lock
1538  *	gr_nactif			ipsq			ipsq
1539  *	gr_v4				ipmp_lock		ipmp_lock
1540  *	gr_v6				ipmp_lock		ipmp_lock
1541  *	gr_nv4				ipmp_lock		ipmp_lock
1542  *	gr_nv6				ipmp_lock		ipmp_lock
1543  *	gr_pendv4			ipmp_lock		ipmp_lock
1544  *	gr_pendv6			ipmp_lock		ipmp_lock
1545  *	gr_linkdownmp			ipsq			ipsq
1546  *	gr_ksp				ipmp_lock		ipmp_lock
1547  *	gr_kstats0			atomic			atomic
1548  */
1549 typedef struct ipmp_grp_s {
1550 	char		gr_name[LIFGRNAMSIZ];	/* group name */
1551 	char		gr_ifname[LIFNAMSIZ];	/* interface name */
1552 	t_uscalar_t	gr_mactype;	/* DLPI mactype of group */
1553 	phyint_t	*gr_phyint;	/* IPMP group phyint */
1554 	uint_t		gr_nif;		/* number of interfaces in group */
1555 	uint_t		gr_nactif;	/* number of active interfaces */
1556 	ipmp_illgrp_t	*gr_v4;		/* V4 group information */
1557 	ipmp_illgrp_t	*gr_v6;		/* V6 group information */
1558 	uint_t		gr_nv4;		/* number of ills in V4 group */
1559 	uint_t		gr_nv6;		/* number of ills in V6 group */
1560 	uint_t		gr_pendv4;	/* number of pending ills in V4 group */
1561 	uint_t		gr_pendv6;	/* number of pending ills in V6 group */
1562 	mblk_t		*gr_linkdownmp;	/* message used to bring link down */
1563 	kstat_t		*gr_ksp;	/* group kstat pointer */
1564 	uint64_t	gr_kstats0[IPMP_KSTAT_MAX]; /* baseline group kstats */
1565 } ipmp_grp_t;
1566 
1567 /*
1568  * IPMP ARP entry -- one per SIOCS*ARP entry tied to the group.  Used to keep
1569  * ARP up-to-date as the active set of interfaces in the group changes.
1570  */
1571 typedef struct ipmp_arpent_s {
1572 	ipaddr_t	ia_ipaddr;	/* IP address for this entry */
1573 	boolean_t	ia_proxyarp;	/* proxy ARP entry? */
1574 	boolean_t	ia_notified;	/* ARP notified about this entry? */
1575 	list_node_t	ia_node;	/* next ARP entry in list */
1576 	uint16_t	ia_flags;	/* nce_flags for the address */
1577 	size_t		ia_lladdr_len;
1578 	uchar_t		*ia_lladdr;
1579 } ipmp_arpent_t;
1580 
1581 struct arl_s;
1582 
1583 /*
1584  * Per-ill capabilities.
1585  */
1586 struct ill_hcksum_capab_s {
1587 	uint_t	ill_hcksum_version;	/* interface version */
1588 	uint_t	ill_hcksum_txflags;	/* capabilities on transmit */
1589 };
1590 
1591 struct ill_zerocopy_capab_s {
1592 	uint_t	ill_zerocopy_version;	/* interface version */
1593 	uint_t	ill_zerocopy_flags;	/* capabilities */
1594 };
1595 
1596 struct ill_lso_capab_s {
1597 	uint_t	ill_lso_flags;		/* capabilities */
1598 	uint_t	ill_lso_max_tcpv4;	/* maximum size of payload */
1599 	uint_t	ill_lso_max_tcpv6;	/* maximum size of payload */
1600 };
1601 
1602 /*
1603  * IP Lower level Structure.
1604  * Instance data structure in ip_open when there is a device below us.
1605  */
1606 typedef struct ill_s {
1607 	pfillinput_t ill_inputfn;	/* Fast input function selector */
1608 	ill_if_t *ill_ifptr;		/* pointer to interface type */
1609 	queue_t	*ill_rq;		/* Read queue. */
1610 	queue_t	*ill_wq;		/* Write queue. */
1611 
1612 	int	ill_error;		/* Error value sent up by device. */
1613 
1614 	ipif_t	*ill_ipif;		/* Interface chain for this ILL. */
1615 
1616 	uint_t	ill_ipif_up_count;	/* Number of IPIFs currently up. */
1617 	uint_t	ill_max_frag;		/* Max IDU from DLPI. */
1618 	uint_t	ill_current_frag;	/* Current IDU from DLPI. */
1619 	uint_t	ill_mtu;		/* User-specified MTU; SIOCSLIFMTU */
1620 	uint_t	ill_mc_mtu;		/* MTU for multi/broadcast */
1621 	uint_t	ill_metric;		/* BSD if metric, for compatibility. */
1622 	char	*ill_name;		/* Our name. */
1623 	uint_t	ill_ipif_dup_count;	/* Number of duplicate addresses. */
1624 	uint_t	ill_name_length;	/* Name length, incl. terminator. */
1625 	uint_t	ill_net_type;		/* IRE_IF_RESOLVER/IRE_IF_NORESOLVER. */
1626 	/*
1627 	 * Physical Point of Attachment num.  If DLPI style 1 provider
1628 	 * then this is derived from the devname.
1629 	 */
1630 	uint_t	ill_ppa;
1631 	t_uscalar_t	ill_sap;
1632 	t_scalar_t	ill_sap_length;	/* Including sign (for position) */
1633 	uint_t	ill_phys_addr_length;	/* Excluding the sap. */
1634 	uint_t	ill_bcast_addr_length;	/* Only set when the DL provider */
1635 					/* supports broadcast. */
1636 	t_uscalar_t	ill_mactype;
1637 	uint8_t	*ill_frag_ptr;		/* Reassembly state. */
1638 	timeout_id_t ill_frag_timer_id; /* timeout id for the frag timer */
1639 	ipfb_t	*ill_frag_hash_tbl;	/* Fragment hash list head. */
1640 
1641 	krwlock_t ill_mcast_lock;	/* Protects multicast state */
1642 	kmutex_t ill_mcast_serializer;	/* Serialize across ilg and ilm state */
1643 	ilm_t	*ill_ilm;		/* Multicast membership for ill */
1644 	uint_t	ill_global_timer;	/* for IGMPv3/MLDv2 general queries */
1645 	int	ill_mcast_type;		/* type of router which is querier */
1646 					/* on this interface */
1647 	uint16_t ill_mcast_v1_time;	/* # slow timeouts since last v1 qry */
1648 	uint16_t ill_mcast_v2_time;	/* # slow timeouts since last v2 qry */
1649 	uint8_t	ill_mcast_v1_tset;	/* 1 => timer is set; 0 => not set */
1650 	uint8_t	ill_mcast_v2_tset;	/* 1 => timer is set; 0 => not set */
1651 
1652 	uint8_t	ill_mcast_rv;		/* IGMPv3/MLDv2 robustness variable */
1653 	int	ill_mcast_qi;		/* IGMPv3/MLDv2 query interval var */
1654 
1655 	/*
1656 	 * All non-NULL cells between 'ill_first_mp_to_free' and
1657 	 * 'ill_last_mp_to_free' are freed in ill_delete.
1658 	 */
1659 #define	ill_first_mp_to_free	ill_bcast_mp
1660 	mblk_t	*ill_bcast_mp;		/* DLPI header for broadcasts. */
1661 	mblk_t	*ill_unbind_mp;		/* unbind mp from ill_dl_up() */
1662 	mblk_t	*ill_promiscoff_mp;	/* for ill_leave_allmulti() */
1663 	mblk_t	*ill_dlpi_deferred;	/* b_next chain of control messages */
1664 	mblk_t	*ill_dest_addr_mp;	/* mblk which holds ill_dest_addr */
1665 	mblk_t	*ill_replumb_mp;	/* replumb mp from ill_replumb() */
1666 	mblk_t	*ill_phys_addr_mp;	/* mblk which holds ill_phys_addr */
1667 	mblk_t	*ill_mcast_deferred;	/* b_next chain of IGMP/MLD packets */
1668 #define	ill_last_mp_to_free	ill_mcast_deferred
1669 
1670 	cred_t	*ill_credp;		/* opener's credentials */
1671 	uint8_t	*ill_phys_addr;		/* ill_phys_addr_mp->b_rptr + off */
1672 	uint8_t *ill_dest_addr;		/* ill_dest_addr_mp->b_rptr + off */
1673 
1674 	uint_t	ill_state_flags;	/* see ILL_* flags above */
1675 
1676 	/* Following bit fields protected by ipsq_t */
1677 	uint_t
1678 		ill_needs_attach : 1,
1679 		ill_reserved : 1,
1680 		ill_isv6 : 1,
1681 		ill_dlpi_style_set : 1,
1682 
1683 		ill_ifname_pending : 1,
1684 		ill_logical_down : 1,
1685 		ill_dl_up : 1,
1686 		ill_up_ipifs : 1,
1687 
1688 		ill_note_link : 1,	/* supports link-up notification */
1689 		ill_capab_reneg : 1, /* capability renegotiation to be done */
1690 		ill_dld_capab_inprog : 1, /* direct dld capab call in prog */
1691 		ill_need_recover_multicast : 1,
1692 
1693 		ill_replumbing : 1,
1694 		ill_arl_dlpi_pending : 1,
1695 		ill_grp_pending : 1,
1696 
1697 		ill_pad_to_bit_31 : 17;
1698 
1699 	/* Following bit fields protected by ill_lock */
1700 	uint_t
1701 		ill_fragtimer_executing : 1,
1702 		ill_fragtimer_needrestart : 1,
1703 		ill_manual_token : 1,	/* system won't override ill_token */
1704 		/*
1705 		 * ill_manual_linklocal : system will not change the
1706 		 * linklocal whenever ill_token changes.
1707 		 */
1708 		ill_manual_linklocal : 1,
1709 
1710 		ill_manual_dst_linklocal : 1, /* same for pt-pt dst linklocal */
1711 
1712 		ill_mcast_ncec_cleanup : 1,	/* Reaping mcast ncecs. */
1713 		ill_pad_bit_31 : 26;
1714 
1715 	/*
1716 	 * Used in SIOCSIFMUXID and SIOCGIFMUXID for 'ifconfig unplumb'.
1717 	 */
1718 	int	ill_muxid;		/* muxid returned from plink */
1719 
1720 	/* Used for IP frag reassembly throttling on a per ILL basis.  */
1721 	uint_t	ill_ipf_gen;		/* Generation of next fragment queue */
1722 	uint_t	ill_frag_count;		/* Count of all reassembly mblk bytes */
1723 	uint_t	ill_frag_free_num_pkts;	 /* num of fragmented packets to free */
1724 	clock_t	ill_last_frag_clean_time; /* time when frag's were pruned */
1725 	int	ill_type;		/* From <net/if_types.h> */
1726 	uint_t	ill_dlpi_multicast_state;	/* See below IDS_* */
1727 	uint_t	ill_dlpi_fastpath_state;	/* See below IDS_* */
1728 
1729 	/*
1730 	 * Capabilities related fields.
1731 	 */
1732 	uint_t  ill_dlpi_capab_state;	/* State of capability query, IDCS_* */
1733 	uint_t	ill_capab_pending_cnt;
1734 	uint64_t ill_capabilities;	/* Enabled capabilities, ILL_CAPAB_* */
1735 	ill_hcksum_capab_t *ill_hcksum_capab; /* H/W cksumming capabilities */
1736 	ill_zerocopy_capab_t *ill_zerocopy_capab; /* Zero-copy capabilities */
1737 	ill_dld_capab_t *ill_dld_capab; /* DLD capabilities */
1738 	ill_lso_capab_t	*ill_lso_capab;	/* Large Segment Offload capabilities */
1739 	mblk_t	*ill_capab_reset_mp;	/* Preallocated mblk for capab reset */
1740 
1741 	uint8_t	ill_max_hops;	/* Maximum hops for any logical interface */
1742 	uint_t	ill_user_mtu;	/* User-specified MTU via SIOCSLIFLNKINFO */
1743 	uint32_t ill_reachable_time;	/* Value for ND algorithm in msec */
1744 	uint32_t ill_reachable_retrans_time; /* Value for ND algorithm msec */
1745 	uint_t	ill_max_buf;		/* Max # of req to buffer for ND */
1746 	in6_addr_t	ill_token;	/* IPv6 interface id */
1747 	in6_addr_t	ill_dest_token;	/* Destination IPv6 interface id */
1748 	uint_t		ill_token_length;
1749 	uint32_t	ill_xmit_count;		/* ndp max multicast xmits */
1750 	mib2_ipIfStatsEntry_t	*ill_ip_mib;	/* ver indep. interface mib */
1751 	mib2_ipv6IfIcmpEntry_t	*ill_icmp6_mib;	/* Per interface mib */
1752 
1753 	phyint_t		*ill_phyint;
1754 	uint64_t		ill_flags;
1755 
1756 	kmutex_t	ill_lock;	/* Please see table below */
1757 	/*
1758 	 * The ill_nd_lla* fields handle the link layer address option
1759 	 * from neighbor discovery. This is used for external IPv6
1760 	 * address resolution.
1761 	 */
1762 	mblk_t		*ill_nd_lla_mp;	/* mblk which holds ill_nd_lla */
1763 	uint8_t		*ill_nd_lla;	/* Link Layer Address */
1764 	uint_t		ill_nd_lla_len;	/* Link Layer Address length */
1765 	/*
1766 	 * We have 4 phys_addr_req's sent down. This field keeps track
1767 	 * of which one is pending.
1768 	 */
1769 	t_uscalar_t	ill_phys_addr_pend; /* which dl_phys_addr_req pending */
1770 	/*
1771 	 * Used to save errors that occur during plumbing
1772 	 */
1773 	uint_t		ill_ifname_pending_err;
1774 	avl_node_t	ill_avl_byppa; /* avl node based on ppa */
1775 	uint_t		ill_mcast_nces;	/* Number of NCEs that are multicast. */
1776 	list_t		ill_nce; /* pointer to nce_s list */
1777 	uint_t		ill_refcnt;	/* active refcnt by threads */
1778 	uint_t		ill_ire_cnt;	/* ires associated with this ill */
1779 	kcondvar_t	ill_cv;
1780 	uint_t		ill_ncec_cnt;	/* ncecs associated with this ill */
1781 	uint_t		ill_nce_cnt;	/* nces associated with this ill */
1782 	uint_t		ill_waiters;	/* threads waiting in ipsq_enter */
1783 	/*
1784 	 * Contains the upper read queue pointer of the module immediately
1785 	 * beneath IP.  This field allows IP to validate sub-capability
1786 	 * acknowledgments coming up from downstream.
1787 	 */
1788 	queue_t		*ill_lmod_rq;	/* read queue pointer of module below */
1789 	uint_t		ill_lmod_cnt;	/* number of modules beneath IP */
1790 	ip_m_t		*ill_media;	/* media specific params/functions */
1791 	t_uscalar_t	ill_dlpi_pending; /* Last DLPI primitive issued */
1792 	uint_t		ill_usesrc_ifindex; /* use src addr from this ILL */
1793 	struct ill_s	*ill_usesrc_grp_next; /* Next ILL in the usesrc group */
1794 	boolean_t	ill_trace_disable;	/* True when alloc fails */
1795 	zoneid_t	ill_zoneid;
1796 	ip_stack_t	*ill_ipst;	/* Corresponds to a netstack_hold */
1797 	uint32_t	ill_dhcpinit;	/* IP_DHCPINIT_IFs for ill */
1798 	void		*ill_flownotify_mh; /* Tx flow ctl, mac cb handle */
1799 	uint_t		ill_ilm_cnt;    /* ilms referencing this ill */
1800 	uint_t		ill_ipallmulti_cnt; /* ip_join_allmulti() calls */
1801 	ilm_t		*ill_ipallmulti_ilm;
1802 
1803 	mblk_t		*ill_saved_ire_mp; /* Allocated for each extra IRE */
1804 					/* with ire_ill set so they can */
1805 					/* survive the ill going down and up. */
1806 	kmutex_t	ill_saved_ire_lock; /* Protects ill_saved_ire_mp, cnt */
1807 	uint_t		ill_saved_ire_cnt;	/* # entries */
1808 	struct arl_ill_common_s    *ill_common;
1809 	ire_t		*ill_ire_multicast; /* IRE_MULTICAST for ill */
1810 	clock_t		ill_defend_start;   /* start of 1 hour period */
1811 	uint_t		ill_defend_count;   /* # of announce/defends per ill */
1812 	/*
1813 	 * IPMP fields.
1814 	 */
1815 	ipmp_illgrp_t	*ill_grp;	/* IPMP group information */
1816 	list_node_t	ill_actnode;	/* next active ill in group */
1817 	list_node_t	ill_grpnode;	/* next ill in group */
1818 	ipif_t		*ill_src_ipif;	/* source address selection rotor */
1819 	ipif_t		*ill_move_ipif;	/* ipif awaiting move to new ill */
1820 	boolean_t	ill_nom_cast;	/* nominated for mcast/bcast */
1821 	uint_t		ill_bound_cnt;	/* # of data addresses bound to ill */
1822 	ipif_t		*ill_bound_ipif; /* ipif chain bound to ill */
1823 	timeout_id_t	ill_refresh_tid; /* ill refresh retry timeout id */
1824 
1825 	uint32_t	ill_mrouter_cnt; /* mrouter allmulti joins */
1826 	uint32_t	ill_allowed_ips_cnt;
1827 	in6_addr_t	*ill_allowed_ips;
1828 
1829 	/* list of multicast physical addresses joined on this ill */
1830 	multiphysaddr_t *ill_mphysaddr_list;
1831 } ill_t;
1832 
1833 /*
1834  * ILL_FREE_OK() means that there are no incoming pointer references
1835  * to the ill.
1836  */
1837 #define	ILL_FREE_OK(ill)					\
1838 	((ill)->ill_ire_cnt == 0 && (ill)->ill_ilm_cnt == 0 &&	\
1839 	(ill)->ill_ncec_cnt == 0 && (ill)->ill_nce_cnt == 0)
1840 
1841 /*
1842  * An ipif/ill can be marked down only when the ire and ncec references
1843  * to that ipif/ill goes to zero. ILL_DOWN_OK() is a necessary condition
1844  * quiescence checks. See comments above IPIF_DOWN_OK for details
1845  * on why ires and nces are selectively considered for this macro.
1846  */
1847 #define	ILL_DOWN_OK(ill)					\
1848 	(ill->ill_ire_cnt == 0 && ill->ill_ncec_cnt == 0 &&	\
1849 	ill->ill_nce_cnt == 0)
1850 
1851 /*
1852  * The following table lists the protection levels of the various members
1853  * of the ill_t. Same notation as that used for ipif_t above is used.
1854  *
1855  *				Write			Read
1856  *
1857  * ill_ifptr			ill_g_lock + s		Write once
1858  * ill_rq			ipsq			Write once
1859  * ill_wq			ipsq			Write once
1860  *
1861  * ill_error			ipsq			None
1862  * ill_ipif			ill_g_lock + ipsq	ill_g_lock OR ipsq
1863  * ill_ipif_up_count		ill_lock + ipsq		ill_lock OR ipsq
1864  * ill_max_frag			ill_lock		ill_lock
1865  * ill_current_frag		ill_lock		ill_lock
1866  *
1867  * ill_name			ill_g_lock + ipsq	Write once
1868  * ill_name_length		ill_g_lock + ipsq	Write once
1869  * ill_ndd_name			ipsq			Write once
1870  * ill_net_type			ipsq			Write once
1871  * ill_ppa			ill_g_lock + ipsq	Write once
1872  * ill_sap			ipsq + down ill		Write once
1873  * ill_sap_length		ipsq + down ill		Write once
1874  * ill_phys_addr_length		ipsq + down ill		Write once
1875  *
1876  * ill_bcast_addr_length	ipsq			ipsq
1877  * ill_mactype			ipsq			ipsq
1878  * ill_frag_ptr			ipsq			ipsq
1879  *
1880  * ill_frag_timer_id		ill_lock		ill_lock
1881  * ill_frag_hash_tbl		ipsq			up ill
1882  * ill_ilm			ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1883  * ill_global_timer		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1884  * ill_mcast_type		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1885  * ill_mcast_v1_time		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1886  * ill_mcast_v2_time		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1887  * ill_mcast_v1_tset		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1888  * ill_mcast_v2_tset		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1889  * ill_mcast_rv			ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1890  * ill_mcast_qi			ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1891  *
1892  * ill_down_mp			ipsq			ipsq
1893  * ill_dlpi_deferred		ill_lock		ill_lock
1894  * ill_dlpi_pending		ipsq + ill_lock		ipsq or ill_lock or
1895  *							absence of ipsq writer.
1896  * ill_phys_addr_mp		ipsq + down ill		only when ill is up
1897  * ill_mcast_deferred		ill_lock		ill_lock
1898  * ill_phys_addr		ipsq + down ill		only when ill is up
1899  * ill_dest_addr_mp		ipsq + down ill		only when ill is up
1900  * ill_dest_addr		ipsq + down ill		only when ill is up
1901  *
1902  * ill_state_flags		ill_lock		ill_lock
1903  * exclusive bit flags		ipsq_t			ipsq_t
1904  * shared bit flags		ill_lock		ill_lock
1905  *
1906  * ill_muxid			ipsq			Not atomic
1907  *
1908  * ill_ipf_gen			Not atomic
1909  * ill_frag_count		atomics			atomics
1910  * ill_type			ipsq + down ill		only when ill is up
1911  * ill_dlpi_multicast_state	ill_lock		ill_lock
1912  * ill_dlpi_fastpath_state	ill_lock		ill_lock
1913  * ill_dlpi_capab_state		ipsq			ipsq
1914  * ill_max_hops			ipsq			Not atomic
1915  *
1916  * ill_mtu			ill_lock		None
1917  * ill_mc_mtu			ill_lock		None
1918  *
1919  * ill_user_mtu			ipsq + ill_lock		ill_lock
1920  * ill_reachable_time		ipsq + ill_lock		ill_lock
1921  * ill_reachable_retrans_time	ipsq + ill_lock		ill_lock
1922  * ill_max_buf			ipsq + ill_lock		ill_lock
1923  *
1924  * Next 2 fields need ill_lock because of the get ioctls. They should not
1925  * report partially updated results without executing in the ipsq.
1926  * ill_token			ipsq + ill_lock		ill_lock
1927  * ill_token_length		ipsq + ill_lock		ill_lock
1928  * ill_dest_token		ipsq + down ill		only when ill is up
1929  * ill_xmit_count		ipsq + down ill		write once
1930  * ill_ip6_mib			ipsq + down ill		only when ill is up
1931  * ill_icmp6_mib		ipsq + down ill		only when ill is up
1932  *
1933  * ill_phyint			ipsq, ill_g_lock, ill_lock	Any of them
1934  * ill_flags			ill_lock		ill_lock
1935  * ill_nd_lla_mp		ipsq + down ill		only when ill is up
1936  * ill_nd_lla			ipsq + down ill		only when ill is up
1937  * ill_nd_lla_len		ipsq + down ill		only when ill is up
1938  * ill_phys_addr_pend		ipsq + down ill		only when ill is up
1939  * ill_ifname_pending_err	ipsq			ipsq
1940  * ill_avl_byppa		ipsq, ill_g_lock	write once
1941  *
1942  * ill_fastpath_list		ill_lock		ill_lock
1943  * ill_refcnt			ill_lock		ill_lock
1944  * ill_ire_cnt			ill_lock		ill_lock
1945  * ill_cv			ill_lock		ill_lock
1946  * ill_mcast_nces		ill_lock		ill_lock
1947  * ill_ncec_cnt			ill_lock		ill_lock
1948  * ill_nce_cnt			ill_lock		ill_lock
1949  * ill_ilm_cnt			ill_lock		ill_lock
1950  * ill_src_ipif			ill_g_lock		ill_g_lock
1951  * ill_trace			ill_lock		ill_lock
1952  * ill_usesrc_grp_next		ill_g_usesrc_lock	ill_g_usesrc_lock
1953  * ill_dhcpinit			atomics			atomics
1954  * ill_flownotify_mh		write once		write once
1955  * ill_capab_pending_cnt	ipsq			ipsq
1956  * ill_ipallmulti_cnt		ill_lock		ill_lock
1957  * ill_ipallmulti_ilm		ill_lock		ill_lock
1958  * ill_saved_ire_mp		ill_saved_ire_lock	ill_saved_ire_lock
1959  * ill_saved_ire_cnt		ill_saved_ire_lock	ill_saved_ire_lock
1960  * ill_arl			???			???
1961  * ill_ire_multicast		ipsq + quiescent	none
1962  * ill_bound_ipif		ipsq			ipsq
1963  * ill_actnode			ipsq + ipmp_lock	ipsq OR ipmp_lock
1964  * ill_grpnode			ipsq + ill_g_lock	ipsq OR ill_g_lock
1965  * ill_src_ipif			ill_g_lock		ill_g_lock
1966  * ill_move_ipif		ipsq			ipsq
1967  * ill_nom_cast			ipsq			ipsq OR advisory
1968  * ill_refresh_tid		ill_lock		ill_lock
1969  * ill_grp (for IPMP ill)	write once		write once
1970  * ill_grp (for underlying ill)	ipsq + ill_g_lock	ipsq OR ill_g_lock
1971  * ill_grp_pending		ill_mcast_serializer	ill_mcast_serializer
1972  * ill_mrouter_cnt		atomics			atomics
1973  * ill_mphysaddr_list	ill_lock		ill_lock
1974  *
1975  * NOTE: It's OK to make heuristic decisions on an underlying interface
1976  *	 by using IS_UNDER_IPMP() or comparing ill_grp's raw pointer value.
1977  */
1978 
1979 /*
1980  * For ioctl restart mechanism see ip_reprocess_ioctl()
1981  */
1982 struct ip_ioctl_cmd_s;
1983 
1984 typedef	int (*ifunc_t)(ipif_t *, struct sockaddr_in *, queue_t *, mblk_t *,
1985     struct ip_ioctl_cmd_s *, void *);
1986 
1987 typedef struct ip_ioctl_cmd_s {
1988 	int	ipi_cmd;
1989 	size_t	ipi_copyin_size;
1990 	uint_t	ipi_flags;
1991 	uint_t	ipi_cmd_type;
1992 	ifunc_t	ipi_func;
1993 	ifunc_t	ipi_func_restart;
1994 } ip_ioctl_cmd_t;
1995 
1996 /*
1997  * ipi_cmd_type:
1998  *
1999  * IF_CMD		1	old style ifreq cmd
2000  * LIF_CMD		2	new style lifreq cmd
2001  * ARP_CMD		3	arpreq cmd
2002  * XARP_CMD		4	xarpreq cmd
2003  * MSFILT_CMD		5	multicast source filter cmd
2004  * MISC_CMD		6	misc cmd (not a more specific one above)
2005  */
2006 
2007 enum { IF_CMD = 1, LIF_CMD, ARP_CMD, XARP_CMD, MSFILT_CMD, MISC_CMD };
2008 
2009 #define	IPI_DONTCARE	0	/* For ioctl encoded values that don't matter */
2010 
2011 /* Flag values in ipi_flags */
2012 #define	IPI_PRIV	0x1	/* Root only command */
2013 #define	IPI_MODOK	0x2	/* Permitted on mod instance of IP */
2014 #define	IPI_WR		0x4	/* Need to grab writer access */
2015 #define	IPI_GET_CMD	0x8	/* branch to mi_copyout on success */
2016 /*	unused		0x10	*/
2017 #define	IPI_NULL_BCONT	0x20	/* ioctl has not data and hence no b_cont */
2018 
2019 extern ip_ioctl_cmd_t	ip_ndx_ioctl_table[];
2020 extern ip_ioctl_cmd_t	ip_misc_ioctl_table[];
2021 extern int ip_ndx_ioctl_count;
2022 extern int ip_misc_ioctl_count;
2023 
2024 /* Passed down by ARP to IP during I_PLINK/I_PUNLINK */
2025 typedef struct ipmx_s {
2026 	char	ipmx_name[LIFNAMSIZ];		/* if name */
2027 	uint_t
2028 		ipmx_arpdev_stream : 1,		/* This is the arp stream */
2029 		ipmx_notused : 31;
2030 } ipmx_t;
2031 
2032 /*
2033  * State for detecting if a driver supports certain features.
2034  * Support for DL_ENABMULTI_REQ uses ill_dlpi_multicast_state.
2035  * Support for DLPI M_DATA fastpath uses ill_dlpi_fastpath_state.
2036  */
2037 #define	IDS_UNKNOWN	0	/* No DLPI request sent */
2038 #define	IDS_INPROGRESS	1	/* DLPI request sent */
2039 #define	IDS_OK		2	/* DLPI request completed successfully */
2040 #define	IDS_FAILED	3	/* DLPI request failed */
2041 
2042 /* Support for DL_CAPABILITY_REQ uses ill_dlpi_capab_state. */
2043 enum {
2044 	IDCS_UNKNOWN,
2045 	IDCS_PROBE_SENT,
2046 	IDCS_OK,
2047 	IDCS_RESET_SENT,
2048 	IDCS_RENEG,
2049 	IDCS_FAILED
2050 };
2051 
2052 /* Extended NDP Management Structure */
2053 typedef struct ipndp_s {
2054 	ndgetf_t	ip_ndp_getf;
2055 	ndsetf_t	ip_ndp_setf;
2056 	caddr_t		ip_ndp_data;
2057 	char		*ip_ndp_name;
2058 } ipndp_t;
2059 
2060 /* IXA Notification types */
2061 typedef enum {
2062 	IXAN_LSO,	/* LSO capability change */
2063 	IXAN_PMTU,	/* PMTU change */
2064 	IXAN_ZCOPY	/* ZEROCOPY capability change */
2065 } ixa_notify_type_t;
2066 
2067 typedef uint_t ixa_notify_arg_t;
2068 
2069 typedef	void	(*ixa_notify_t)(void *, ip_xmit_attr_t *ixa, ixa_notify_type_t,
2070     ixa_notify_arg_t);
2071 
2072 /*
2073  * Attribute flags that are common to the transmit and receive attributes
2074  */
2075 #define	IAF_IS_IPV4		0x80000000	/* ipsec_*_v4 */
2076 #define	IAF_TRUSTED_ICMP	0x40000000	/* ipsec_*_icmp_loopback */
2077 #define	IAF_NO_LOOP_ZONEID_SET	0x20000000	/* Zone that shouldn't have */
2078 						/* a copy */
2079 #define	IAF_LOOPBACK_COPY	0x10000000	/* For multi and broadcast */
2080 
2081 #define	IAF_MASK		0xf0000000	/* Flags that are common */
2082 
2083 /*
2084  * Transmit side attributes used between the transport protocols and IP as
2085  * well as inside IP. It is also used to cache information in the conn_t i.e.
2086  * replaces conn_ire and the IPsec caching in the conn_t.
2087  */
2088 struct ip_xmit_attr_s {
2089 	iaflags_t	ixa_flags;	/* IXAF_*. See below */
2090 
2091 	uint32_t	ixa_free_flags;	/* IXA_FREE_*. See below */
2092 	uint32_t	ixa_refcnt;	/* Using atomics */
2093 
2094 	/*
2095 	 * Always initialized independently of ixa_flags settings.
2096 	 * Used by ip_xmit so we keep them up front for cache locality.
2097 	 */
2098 	uint32_t	ixa_xmit_hint;	/* For ECMP and GLD TX ring fanout */
2099 	uint_t		ixa_pktlen;	/* Always set. For frag and stats */
2100 	zoneid_t	ixa_zoneid;	/* Assumed always set */
2101 
2102 	/* Always set for conn_ip_output(); might be stale */
2103 	/*
2104 	 * Since TCP keeps the conn_t around past the process going away
2105 	 * we need to use the "notr" (e.g, ire_refhold_notr) for ixa_ire,
2106 	 * ixa_nce, and ixa_dce.
2107 	 */
2108 	ire_t		*ixa_ire;	/* Forwarding table entry */
2109 	uint_t		ixa_ire_generation;
2110 	nce_t		*ixa_nce;	/* Neighbor cache entry */
2111 	dce_t		*ixa_dce;	/* Destination cache entry */
2112 	uint_t		ixa_dce_generation;
2113 	uint_t		ixa_src_generation;	/* If IXAF_VERIFY_SOURCE */
2114 
2115 	uint32_t	ixa_src_preferences;	/* prefs for src addr select */
2116 	uint32_t	ixa_pmtu;		/* IXAF_VERIFY_PMTU */
2117 
2118 	/* Set by ULP if IXAF_VERIFY_PMTU; otherwise set by IP */
2119 	uint32_t	ixa_fragsize;
2120 
2121 	int8_t		ixa_use_min_mtu;	/* IXAF_USE_MIN_MTU values */
2122 
2123 	pfirepostfrag_t	ixa_postfragfn;		/* Set internally in IP */
2124 
2125 	in6_addr_t	ixa_nexthop_v6;		/* IXAF_NEXTHOP_SET */
2126 #define	ixa_nexthop_v4	V4_PART_OF_V6(ixa_nexthop_v6)
2127 
2128 	zoneid_t	ixa_no_loop_zoneid;	/* IXAF_NO_LOOP_ZONEID_SET */
2129 
2130 	uint_t		ixa_scopeid;		/* For IPv6 link-locals */
2131 
2132 	uint_t		ixa_broadcast_ttl;	/* IXAF_BROACAST_TTL_SET */
2133 
2134 	uint_t		ixa_multicast_ttl;	/* Assumed set for multicast */
2135 	uint_t		ixa_multicast_ifindex;	/* Assumed set for multicast */
2136 	ipaddr_t	ixa_multicast_ifaddr;	/* Assumed set for multicast */
2137 
2138 	int		ixa_raw_cksum_offset;	/* If IXAF_SET_RAW_CKSUM */
2139 
2140 	uint32_t	ixa_ident;		/* For IPv6 fragment header */
2141 
2142 	uint64_t	ixa_conn_id;		/* Used by DTrace */
2143 	/*
2144 	 * Cached LSO information.
2145 	 */
2146 	ill_lso_capab_t	ixa_lso_capab;		/* Valid when IXAF_LSO_CAPAB */
2147 
2148 	uint64_t	ixa_ipsec_policy_gen;	/* Generation from iph_gen */
2149 	/*
2150 	 * The following IPsec fields are only initialized when
2151 	 * IXAF_IPSEC_SECURE is set. Otherwise they contain garbage.
2152 	 */
2153 	ipsec_latch_t	*ixa_ipsec_latch;	/* Just the ids */
2154 	struct ipsa_s	*ixa_ipsec_ah_sa;	/* Hard reference SA for AH */
2155 	struct ipsa_s	*ixa_ipsec_esp_sa;	/* Hard reference SA for ESP */
2156 	struct ipsec_policy_s	*ixa_ipsec_policy; /* why are we here? */
2157 	struct ipsec_action_s	*ixa_ipsec_action; /* For reflected packets */
2158 	ipsa_ref_t	ixa_ipsec_ref[2];	/* Soft reference to SA */
2159 						/* 0: ESP, 1: AH */
2160 
2161 	/*
2162 	 * The selectors here are potentially different than the SPD rule's
2163 	 * selectors, and we need to have both available for IKEv2.
2164 	 *
2165 	 * NOTE: "Source" and "Dest" are w.r.t. outbound datagrams.  Ports can
2166 	 *	 be zero, and the protocol number is needed to make the ports
2167 	 *	 significant.
2168 	 */
2169 	uint16_t ixa_ipsec_src_port;	/* Source port number of d-gram. */
2170 	uint16_t ixa_ipsec_dst_port;	/* Destination port number of d-gram. */
2171 	uint8_t  ixa_ipsec_icmp_type;	/* ICMP type of d-gram */
2172 	uint8_t  ixa_ipsec_icmp_code;	/* ICMP code of d-gram */
2173 
2174 	sa_family_t ixa_ipsec_inaf;	/* Inner address family */
2175 #define	IXA_MAX_ADDRLEN 4	/* Max addr len. (in 32-bit words) */
2176 	uint32_t ixa_ipsec_insrc[IXA_MAX_ADDRLEN];	/* Inner src address */
2177 	uint32_t ixa_ipsec_indst[IXA_MAX_ADDRLEN];	/* Inner dest address */
2178 	uint8_t  ixa_ipsec_insrcpfx;	/* Inner source prefix */
2179 	uint8_t  ixa_ipsec_indstpfx;	/* Inner destination prefix */
2180 
2181 	uint8_t ixa_ipsec_proto;	/* IP protocol number for d-gram. */
2182 
2183 	/* Always initialized independently of ixa_flags settings */
2184 	uint_t		ixa_ifindex;	/* Assumed always set */
2185 	uint16_t	ixa_ip_hdr_length; /* Points to ULP header */
2186 	uint8_t		ixa_protocol;	/* Protocol number for ULP cksum */
2187 	ts_label_t	*ixa_tsl;	/* Always set. NULL if not TX */
2188 	ip_stack_t	*ixa_ipst;	/* Always set */
2189 	uint32_t	ixa_extra_ident; /* Set if LSO */
2190 	cred_t		*ixa_cred;	/* For getpeerucred */
2191 	pid_t		ixa_cpid;	/* For getpeerucred */
2192 
2193 #ifdef DEBUG
2194 	kthread_t	*ixa_curthread;	/* For serialization assert */
2195 #endif
2196 	squeue_t	*ixa_sqp;	/* Set from conn_sqp as a hint */
2197 	uintptr_t	ixa_cookie;	/* cookie to use for tx flow control */
2198 
2199 	/*
2200 	 * Must be set by ULP if any of IXAF_VERIFY_LSO, IXAF_VERIFY_PMTU,
2201 	 * or IXAF_VERIFY_ZCOPY is set.
2202 	 */
2203 	ixa_notify_t	ixa_notify;	/* Registered upcall notify function */
2204 	void		*ixa_notify_cookie; /* ULP cookie for ixa_notify */
2205 
2206 	uint_t		ixa_tcpcleanup;	/* Used by conn_ixa_cleanup */
2207 };
2208 
2209 /*
2210  * Flags to indicate which transmit attributes are set.
2211  * Split into "xxx_SET" ones which indicate that the "xxx" field it set, and
2212  * single flags.
2213  */
2214 #define	IXAF_REACH_CONF		0x00000001	/* Reachability confirmation */
2215 #define	IXAF_BROADCAST_TTL_SET	0x00000002	/* ixa_broadcast_ttl valid */
2216 #define	IXAF_SET_SOURCE		0x00000004	/* Replace if broadcast */
2217 #define	IXAF_USE_MIN_MTU	0x00000008	/* IPV6_USE_MIN_MTU */
2218 
2219 #define	IXAF_DONTFRAG		0x00000010	/* IP*_DONTFRAG */
2220 #define	IXAF_VERIFY_PMTU	0x00000020	/* ixa_pmtu/ixa_fragsize set */
2221 #define	IXAF_PMTU_DISCOVERY	0x00000040	/* Create/use PMTU state */
2222 #define	IXAF_MULTICAST_LOOP	0x00000080	/* IP_MULTICAST_LOOP */
2223 
2224 #define	IXAF_IPSEC_SECURE	0x00000100	/* Need IPsec processing */
2225 #define	IXAF_UCRED_TSL		0x00000200	/* ixa_tsl from SCM_UCRED */
2226 #define	IXAF_DONTROUTE		0x00000400	/* SO_DONTROUTE */
2227 #define	IXAF_NO_IPSEC		0x00000800	/* Ignore policy */
2228 
2229 #define	IXAF_PMTU_TOO_SMALL	0x00001000	/* PMTU too small */
2230 #define	IXAF_SET_ULP_CKSUM	0x00002000	/* Calculate ULP checksum */
2231 #define	IXAF_VERIFY_SOURCE	0x00004000	/* Check that source is ok */
2232 #define	IXAF_NEXTHOP_SET	0x00008000	/* ixa_nexthop set */
2233 
2234 #define	IXAF_PMTU_IPV4_DF	0x00010000	/* Set IPv4 DF */
2235 #define	IXAF_NO_DEV_FLOW_CTL	0x00020000	/* Protocol needs no flow ctl */
2236 #define	IXAF_NO_TTL_CHANGE	0x00040000	/* Internal to IP */
2237 #define	IXAF_IPV6_ADD_FRAGHDR	0x00080000	/* Add fragment header */
2238 
2239 #define	IXAF_IPSEC_TUNNEL	0x00100000	/* Tunnel mode */
2240 #define	IXAF_NO_PFHOOK		0x00200000	/* Skip xmit pfhook */
2241 #define	IXAF_NO_TRACE		0x00400000	/* When back from ARP/ND */
2242 #define	IXAF_SCOPEID_SET	0x00800000	/* ixa_scopeid set */
2243 
2244 #define	IXAF_MULTIRT_MULTICAST	0x01000000	/* MULTIRT for multicast */
2245 #define	IXAF_NO_HW_CKSUM	0x02000000	/* Force software cksum */
2246 #define	IXAF_SET_RAW_CKSUM	0x04000000	/* Use ixa_raw_cksum_offset */
2247 #define	IXAF_IPSEC_GLOBAL_POLICY 0x08000000	/* Policy came from global */
2248 
2249 /* Note the following uses bits 0x10000000 through 0x80000000 */
2250 #define	IXAF_IS_IPV4		IAF_IS_IPV4
2251 #define	IXAF_TRUSTED_ICMP	IAF_TRUSTED_ICMP
2252 #define	IXAF_NO_LOOP_ZONEID_SET	IAF_NO_LOOP_ZONEID_SET
2253 #define	IXAF_LOOPBACK_COPY	IAF_LOOPBACK_COPY
2254 
2255 /* Note: use the upper 32 bits */
2256 #define	IXAF_VERIFY_LSO		0x100000000	/* Check LSO capability */
2257 #define	IXAF_LSO_CAPAB		0x200000000	/* Capable of LSO */
2258 #define	IXAF_VERIFY_ZCOPY	0x400000000	/* Check Zero Copy capability */
2259 #define	IXAF_ZCOPY_CAPAB	0x800000000	/* Capable of ZEROCOPY */
2260 
2261 /*
2262  * The normal flags for sending packets e.g., icmp errors
2263  */
2264 #define	IXAF_BASIC_SIMPLE_V4	\
2265 	(IXAF_SET_ULP_CKSUM | IXAF_IS_IPV4 | IXAF_VERIFY_SOURCE)
2266 #define	IXAF_BASIC_SIMPLE_V6	(IXAF_SET_ULP_CKSUM | IXAF_VERIFY_SOURCE)
2267 
2268 /*
2269  * Normally these fields do not have a hold. But in some cases they do, for
2270  * instance when we've gone through ip_*_attr_to/from_mblk.
2271  * We use ixa_free_flags to indicate that they have a hold and need to be
2272  * released on cleanup.
2273  */
2274 #define	IXA_FREE_CRED		0x00000001	/* ixa_cred needs to be rele */
2275 #define	IXA_FREE_TSL		0x00000002	/* ixa_tsl needs to be rele */
2276 
2277 /*
2278  * Trivial state machine used to synchronize IXA cleanup for TCP connections.
2279  * See conn_ixa_cleanup().
2280  */
2281 #define	IXATC_IDLE		0x00000000
2282 #define	IXATC_INPROGRESS	0x00000001
2283 #define	IXATC_COMPLETE		0x00000002
2284 
2285 /*
2286  * Simplistic way to set the ixa_xmit_hint for locally generated traffic
2287  * and forwarded traffic. The shift amount are based on the size of the
2288  * structs to discard the low order bits which don't have much if any variation
2289  * (coloring in kmem_cache_alloc might provide some variation).
2290  *
2291  * Basing the locally generated hint on the address of the conn_t means that
2292  * the packets from the same socket/connection do not get reordered.
2293  * Basing the hint for forwarded traffic on the ill_ring_t means that
2294  * packets from the same NIC+ring are likely to use the same outbound ring
2295  * hence we get low contention on the ring in the transmitting driver.
2296  */
2297 #define	CONN_TO_XMIT_HINT(connp)	((uint32_t)(((uintptr_t)connp) >> 11))
2298 #define	ILL_RING_TO_XMIT_HINT(ring)	((uint32_t)(((uintptr_t)ring) >> 7))
2299 
2300 /*
2301  * IP set Destination Flags used by function ip_set_destination,
2302  * ip_attr_connect, and conn_connect.
2303  */
2304 #define	IPDF_ALLOW_MCBC		0x1	/* Allow multi/broadcast */
2305 #define	IPDF_VERIFY_DST		0x2	/* Verify destination addr */
2306 #define	IPDF_SELECT_SRC		0x4	/* Select source address */
2307 #define	IPDF_LSO		0x8	/* Try LSO */
2308 #define	IPDF_IPSEC		0x10	/* Set IPsec policy */
2309 #define	IPDF_ZONE_IS_GLOBAL	0x20	/* From conn_zone_is_global */
2310 #define	IPDF_ZCOPY		0x40	/* Try ZEROCOPY */
2311 #define	IPDF_UNIQUE_DCE		0x80	/* Get a per-destination DCE */
2312 
2313 /*
2314  * Receive side attributes used between the transport protocols and IP as
2315  * well as inside IP.
2316  */
2317 struct ip_recv_attr_s {
2318 	iaflags_t	ira_flags;	/* See below */
2319 
2320 	uint32_t	ira_free_flags;	/* IRA_FREE_*. See below */
2321 
2322 	/*
2323 	 * This is a hint for TCP SYN packets.
2324 	 * Always initialized independently of ira_flags settings
2325 	 */
2326 	squeue_t	*ira_sqp;
2327 	ill_rx_ring_t	*ira_ring;	/* Internal to IP */
2328 
2329 	/* For ip_accept_tcp when IRAF_TARGET_SQP is set */
2330 	squeue_t	*ira_target_sqp;
2331 	mblk_t		*ira_target_sqp_mp;
2332 
2333 	/* Always initialized independently of ira_flags settings */
2334 	uint32_t	ira_xmit_hint;	/* For ECMP and GLD TX ring fanout */
2335 	zoneid_t	ira_zoneid;	/* ALL_ZONES unless local delivery */
2336 	uint_t		ira_pktlen;	/* Always set. For frag and stats */
2337 	uint16_t	ira_ip_hdr_length; /* Points to ULP header */
2338 	uint8_t		ira_protocol;	/* Protocol number for ULP cksum */
2339 	uint_t		ira_rifindex;	/* Received ifindex */
2340 	uint_t		ira_ruifindex;	/* Received upper ifindex */
2341 	ts_label_t	*ira_tsl;	/* Always set. NULL if not TX */
2342 	/*
2343 	 * ira_rill and ira_ill is set inside IP, but not when conn_recv is
2344 	 * called; ULPs should use ira_ruifindex instead.
2345 	 */
2346 	ill_t		*ira_rill;	/* ill where packet came */
2347 	ill_t		*ira_ill;	/* ill where IP address hosted */
2348 	cred_t		*ira_cred;	/* For getpeerucred */
2349 	pid_t		ira_cpid;	/* For getpeerucred */
2350 
2351 	/* Used when IRAF_VERIFIED_SRC is set; this source was ok */
2352 	ipaddr_t	ira_verified_src;
2353 
2354 	/*
2355 	 * The following IPsec fields are only initialized when
2356 	 * IRAF_IPSEC_SECURE is set. Otherwise they contain garbage.
2357 	 */
2358 	struct ipsec_action_s *ira_ipsec_action; /* how we made it in.. */
2359 	struct ipsa_s	*ira_ipsec_ah_sa;	/* SA for AH */
2360 	struct ipsa_s	*ira_ipsec_esp_sa;	/* SA for ESP */
2361 
2362 	ipaddr_t	ira_mroute_tunnel;	/* IRAF_MROUTE_TUNNEL_SET */
2363 
2364 	zoneid_t	ira_no_loop_zoneid;	/* IRAF_NO_LOOP_ZONEID_SET */
2365 
2366 	uint32_t	ira_esp_udp_ports;	/* IRAF_ESP_UDP_PORTS */
2367 
2368 	/*
2369 	 * For IP_RECVSLLA and ip_ndp_conflict/find_solicitation.
2370 	 * Same size as max for sockaddr_dl
2371 	 */
2372 #define	IRA_L2SRC_SIZE	244
2373 	uint8_t		ira_l2src[IRA_L2SRC_SIZE];	/* If IRAF_L2SRC_SET */
2374 
2375 	/*
2376 	 * Local handle that we use to do lazy setting of ira_l2src.
2377 	 * We defer setting l2src until needed but we do before any
2378 	 * ip_input pullupmsg or copymsg.
2379 	 */
2380 	struct mac_header_info_s *ira_mhip;	/* Could be NULL */
2381 };
2382 
2383 /*
2384  * Flags to indicate which receive attributes are set.
2385  */
2386 #define	IRAF_SYSTEM_LABELED	0x00000001	/* is_system_labeled() */
2387 #define	IRAF_IPV4_OPTIONS	0x00000002	/* Performance */
2388 #define	IRAF_MULTICAST		0x00000004	/* Was multicast at L3 */
2389 #define	IRAF_BROADCAST		0x00000008	/* Was broadcast at L3 */
2390 #define	IRAF_MULTIBROADCAST	(IRAF_MULTICAST|IRAF_BROADCAST)
2391 
2392 #define	IRAF_LOOPBACK		0x00000010	/* Looped back by IP */
2393 #define	IRAF_VERIFY_IP_CKSUM	0x00000020	/* Need to verify IP */
2394 #define	IRAF_VERIFY_ULP_CKSUM	0x00000040	/* Need to verify TCP,UDP,etc */
2395 #define	IRAF_SCTP_CSUM_ERR	0x00000080	/* sctp pkt has failed chksum */
2396 
2397 #define	IRAF_IPSEC_SECURE	0x00000100	/* Passed AH and/or ESP */
2398 #define	IRAF_DHCP_UNICAST	0x00000200
2399 #define	IRAF_IPSEC_DECAPS	0x00000400	/* Was packet decapsulated */
2400 					/* from a matching inner packet? */
2401 #define	IRAF_TARGET_SQP		0x00000800	/* ira_target_sqp is set */
2402 #define	IRAF_VERIFIED_SRC	0x00001000	/* ira_verified_src set */
2403 #define	IRAF_RSVP		0x00002000	/* RSVP packet for rsvpd */
2404 #define	IRAF_MROUTE_TUNNEL_SET	0x00004000	/* From ip_mroute_decap */
2405 #define	IRAF_PIM_REGISTER	0x00008000	/* From register_mforward */
2406 
2407 #define	IRAF_TX_MAC_EXEMPTABLE	0x00010000	/* Allow MAC_EXEMPT readdown */
2408 #define	IRAF_TX_SHARED_ADDR	0x00020000	/* Arrived on ALL_ZONES addr */
2409 #define	IRAF_ESP_UDP_PORTS	0x00040000	/* NAT-traversal packet */
2410 #define	IRAF_NO_HW_CKSUM	0x00080000	/* Force software cksum */
2411 
2412 #define	IRAF_ICMP_ERROR		0x00100000	/* Send to conn_recvicmp */
2413 #define	IRAF_ROUTER_ALERT	0x00200000	/* IPv6 router alert */
2414 #define	IRAF_L2SRC_SET		0x00400000	/* ira_l2src has been set */
2415 #define	IRAF_L2SRC_LOOPBACK	0x00800000	/* Came from us */
2416 
2417 #define	IRAF_L2DST_MULTICAST	0x01000000	/* Multicast at L2 */
2418 #define	IRAF_L2DST_BROADCAST	0x02000000	/* Broadcast at L2 */
2419 /* Unused 0x04000000 */
2420 /* Unused 0x08000000 */
2421 
2422 /* Below starts with 0x10000000 */
2423 #define	IRAF_IS_IPV4		IAF_IS_IPV4
2424 #define	IRAF_TRUSTED_ICMP	IAF_TRUSTED_ICMP
2425 #define	IRAF_NO_LOOP_ZONEID_SET	IAF_NO_LOOP_ZONEID_SET
2426 #define	IRAF_LOOPBACK_COPY	IAF_LOOPBACK_COPY
2427 
2428 /*
2429  * Normally these fields do not have a hold. But in some cases they do, for
2430  * instance when we've gone through ip_*_attr_to/from_mblk.
2431  * We use ira_free_flags to indicate that they have a hold and need to be
2432  * released on cleanup.
2433  */
2434 #define	IRA_FREE_CRED		0x00000001	/* ira_cred needs to be rele */
2435 #define	IRA_FREE_TSL		0x00000002	/* ira_tsl needs to be rele */
2436 
2437 /*
2438  * Optional destination cache entry for path MTU information,
2439  * and ULP metrics.
2440  */
2441 struct dce_s {
2442 	uint_t		dce_generation;	/* Changed since cached? */
2443 	uint_t		dce_flags;	/* See below */
2444 	uint_t		dce_ipversion;	/* IPv4/IPv6 version */
2445 	uint32_t	dce_pmtu;	/* Path MTU if DCEF_PMTU */
2446 	uint32_t	dce_ident;	/* Per destination IP ident. */
2447 	iulp_t		dce_uinfo;	/* Metrics if DCEF_UINFO */
2448 
2449 	struct dce_s	*dce_next;
2450 	struct dce_s	**dce_ptpn;
2451 	struct dcb_s	*dce_bucket;
2452 
2453 	union {
2454 		in6_addr_t	dceu_v6addr;
2455 		ipaddr_t	dceu_v4addr;
2456 	} dce_u;
2457 #define	dce_v4addr	dce_u.dceu_v4addr
2458 #define	dce_v6addr	dce_u.dceu_v6addr
2459 	/* Note that for IPv6+IPMP we use the ifindex for the upper interface */
2460 	uint_t		dce_ifindex;	/* For IPv6 link-locals */
2461 
2462 	kmutex_t	dce_lock;
2463 	uint_t		dce_refcnt;
2464 	uint64_t	dce_last_change_time;	/* Path MTU. In seconds */
2465 
2466 	ip_stack_t	*dce_ipst;	/* Does not have a netstack_hold */
2467 };
2468 
2469 /*
2470  * Values for dce_generation.
2471  *
2472  * If a DCE has DCE_GENERATION_CONDEMNED, the last dce_refrele should delete
2473  * it.
2474  *
2475  * DCE_GENERATION_VERIFY is never stored in dce_generation but it is
2476  * stored in places that cache DCE (such as ixa_dce_generation).
2477  * It is used as a signal that the cache is stale and needs to be reverified.
2478  */
2479 #define	DCE_GENERATION_CONDEMNED	0
2480 #define	DCE_GENERATION_VERIFY		1
2481 #define	DCE_GENERATION_INITIAL		2
2482 #define	DCE_IS_CONDEMNED(dce) \
2483 	((dce)->dce_generation == DCE_GENERATION_CONDEMNED)
2484 
2485 
2486 /*
2487  * Values for ips_src_generation.
2488  *
2489  * SRC_GENERATION_VERIFY is never stored in ips_src_generation but it is
2490  * stored in places that cache IREs (ixa_src_generation). It is used as a
2491  * signal that the cache is stale and needs to be reverified.
2492  */
2493 #define	SRC_GENERATION_VERIFY		0
2494 #define	SRC_GENERATION_INITIAL		1
2495 
2496 /*
2497  * The kernel stores security attributes of all gateways in a database made
2498  * up of one or more tsol_gcdb_t elements.  Each tsol_gcdb_t contains the
2499  * security-related credentials of the gateway.  More than one gateways may
2500  * share entries in the database.
2501  *
2502  * The tsol_gc_t structure represents the gateway to credential association,
2503  * and refers to an entry in the database.  One or more tsol_gc_t entities are
2504  * grouped together to form one or more tsol_gcgrp_t, each representing the
2505  * list of security attributes specific to the gateway.  A gateway may be
2506  * associated with at most one credentials group.
2507  */
2508 struct tsol_gcgrp_s;
2509 
2510 extern uchar_t	ip6opt_ls;	/* TX IPv6 enabler */
2511 
2512 /*
2513  * Gateway security credential record.
2514  */
2515 typedef struct tsol_gcdb_s {
2516 	uint_t		gcdb_refcnt;	/* reference count */
2517 	struct rtsa_s	gcdb_attr;	/* security attributes */
2518 #define	gcdb_mask	gcdb_attr.rtsa_mask
2519 #define	gcdb_doi	gcdb_attr.rtsa_doi
2520 #define	gcdb_slrange	gcdb_attr.rtsa_slrange
2521 } tsol_gcdb_t;
2522 
2523 /*
2524  * Gateway to credential association.
2525  */
2526 typedef struct tsol_gc_s {
2527 	uint_t		gc_refcnt;	/* reference count */
2528 	struct tsol_gcgrp_s *gc_grp;	/* pointer to group */
2529 	struct tsol_gc_s *gc_prev;	/* previous in list */
2530 	struct tsol_gc_s *gc_next;	/* next in list */
2531 	tsol_gcdb_t	*gc_db;		/* pointer to actual credentials */
2532 } tsol_gc_t;
2533 
2534 /*
2535  * Gateway credentials group address.
2536  */
2537 typedef struct tsol_gcgrp_addr_s {
2538 	int		ga_af;		/* address family */
2539 	in6_addr_t	ga_addr;	/* IPv4 mapped or IPv6 address */
2540 } tsol_gcgrp_addr_t;
2541 
2542 /*
2543  * Gateway credentials group.
2544  */
2545 typedef struct tsol_gcgrp_s {
2546 	uint_t		gcgrp_refcnt;	/* reference count */
2547 	krwlock_t	gcgrp_rwlock;	/* lock to protect following */
2548 	uint_t		gcgrp_count;	/* number of credentials */
2549 	tsol_gc_t	*gcgrp_head;	/* first credential in list */
2550 	tsol_gc_t	*gcgrp_tail;	/* last credential in list */
2551 	tsol_gcgrp_addr_t gcgrp_addr;	/* next-hop gateway address */
2552 } tsol_gcgrp_t;
2553 
2554 extern kmutex_t gcgrp_lock;
2555 
2556 #define	GC_REFRELE(p) {				\
2557 	ASSERT((p)->gc_grp != NULL);		\
2558 	rw_enter(&(p)->gc_grp->gcgrp_rwlock, RW_WRITER); \
2559 	ASSERT((p)->gc_refcnt > 0);		\
2560 	if (--((p)->gc_refcnt) == 0)		\
2561 		gc_inactive(p);			\
2562 	else					\
2563 		rw_exit(&(p)->gc_grp->gcgrp_rwlock); \
2564 }
2565 
2566 #define	GCGRP_REFHOLD(p) {			\
2567 	mutex_enter(&gcgrp_lock);		\
2568 	++((p)->gcgrp_refcnt);			\
2569 	ASSERT((p)->gcgrp_refcnt != 0);		\
2570 	mutex_exit(&gcgrp_lock);		\
2571 }
2572 
2573 #define	GCGRP_REFRELE(p) {			\
2574 	mutex_enter(&gcgrp_lock);		\
2575 	ASSERT((p)->gcgrp_refcnt > 0);		\
2576 	if (--((p)->gcgrp_refcnt) == 0)		\
2577 		gcgrp_inactive(p);		\
2578 	ASSERT(MUTEX_HELD(&gcgrp_lock));	\
2579 	mutex_exit(&gcgrp_lock);		\
2580 }
2581 
2582 /*
2583  * IRE gateway security attributes structure, pointed to by tsol_ire_gw_secattr
2584  */
2585 struct tsol_tnrhc;
2586 
2587 struct tsol_ire_gw_secattr_s {
2588 	kmutex_t	igsa_lock;	/* lock to protect following */
2589 	struct tsol_tnrhc *igsa_rhc;	/* host entry for gateway */
2590 	tsol_gc_t	*igsa_gc;	/* for prefix IREs */
2591 };
2592 
2593 void irb_refrele_ftable(irb_t *);
2594 
2595 extern struct kmem_cache *rt_entry_cache;
2596 
2597 typedef struct ire4 {
2598 	ipaddr_t ire4_mask;		/* Mask for matching this IRE. */
2599 	ipaddr_t ire4_addr;		/* Address this IRE represents. */
2600 	ipaddr_t ire4_gateway_addr;	/* Gateway including for IRE_ONLINK */
2601 	ipaddr_t ire4_setsrc_addr;	/* RTF_SETSRC */
2602 } ire4_t;
2603 
2604 typedef struct ire6 {
2605 	in6_addr_t ire6_mask;		/* Mask for matching this IRE. */
2606 	in6_addr_t ire6_addr;		/* Address this IRE represents. */
2607 	in6_addr_t ire6_gateway_addr;	/* Gateway including for IRE_ONLINK */
2608 	in6_addr_t ire6_setsrc_addr;	/* RTF_SETSRC */
2609 } ire6_t;
2610 
2611 typedef union ire_addr {
2612 	ire6_t	ire6_u;
2613 	ire4_t	ire4_u;
2614 } ire_addr_u_t;
2615 
2616 /*
2617  * Internet Routing Entry
2618  * When we have multiple identical IREs we logically add them by manipulating
2619  * ire_identical_ref and ire_delete first decrements
2620  * that and when it reaches 1 we know it is the last IRE.
2621  * "identical" is defined as being the same for:
2622  * ire_addr, ire_netmask, ire_gateway, ire_ill, ire_zoneid, and ire_type
2623  * For instance, multiple IRE_BROADCASTs for the same subnet number are
2624  * viewed as identical, and so are the IRE_INTERFACEs when there are
2625  * multiple logical interfaces (on the same ill) with the same subnet prefix.
2626  */
2627 struct ire_s {
2628 	struct	ire_s	*ire_next;	/* The hash chain must be first. */
2629 	struct	ire_s	**ire_ptpn;	/* Pointer to previous next. */
2630 	uint32_t	ire_refcnt;	/* Number of references */
2631 	ill_t		*ire_ill;
2632 	uint32_t	ire_identical_ref; /* IRE_INTERFACE, IRE_BROADCAST */
2633 	uchar_t		ire_ipversion;	/* IPv4/IPv6 version */
2634 	ushort_t	ire_type;	/* Type of IRE */
2635 	uint_t		ire_generation;	/* Generation including CONDEMNED */
2636 	uint_t	ire_ib_pkt_count;	/* Inbound packets for ire_addr */
2637 	uint_t	ire_ob_pkt_count;	/* Outbound packets to ire_addr */
2638 	time_t	ire_create_time;	/* Time (in secs) IRE was created. */
2639 	uint32_t	ire_flags;	/* flags related to route (RTF_*) */
2640 	/*
2641 	 * ire_testhidden is TRUE for INTERFACE IREs of IS_UNDER_IPMP(ill)
2642 	 * interfaces
2643 	 */
2644 	boolean_t	ire_testhidden;
2645 	pfirerecv_t	ire_recvfn;	/* Receive side handling */
2646 	pfiresend_t	ire_sendfn;	/* Send side handling */
2647 	pfirepostfrag_t	ire_postfragfn;	/* Bottom end of send handling */
2648 
2649 	uint_t		ire_masklen;	/* # bits in ire_mask{,_v6} */
2650 	ire_addr_u_t	ire_u;		/* IPv4/IPv6 address info. */
2651 
2652 	irb_t		*ire_bucket;	/* Hash bucket when ire_ptphn is set */
2653 	kmutex_t	ire_lock;
2654 	clock_t		ire_last_used_time;	/* For IRE_LOCAL reception */
2655 	tsol_ire_gw_secattr_t *ire_gw_secattr; /* gateway security attributes */
2656 	zoneid_t	ire_zoneid;
2657 
2658 	/*
2659 	 * Cached information of where to send packets that match this route.
2660 	 * The ire_dep_* information is used to determine when ire_nce_cache
2661 	 * needs to be updated.
2662 	 * ire_nce_cache is the fastpath for the Neighbor Cache Entry
2663 	 * for IPv6; arp info for IPv4
2664 	 * Since this is a cache setup and torn down independently of
2665 	 * applications we need to use nce_ref{rele,hold}_notr for it.
2666 	 */
2667 	nce_t		*ire_nce_cache;
2668 
2669 	/*
2670 	 * Quick check whether the ire_type and ire_masklen indicates
2671 	 * that the IRE can have ire_nce_cache set i.e., whether it is
2672 	 * IRE_ONLINK and for a single destination.
2673 	 */
2674 	boolean_t	ire_nce_capable;
2675 
2676 	/*
2677 	 * Dependency tracking so we can safely cache IRE and NCE pointers
2678 	 * in offlink and onlink IREs.
2679 	 * These are locked under the ips_ire_dep_lock rwlock. Write held
2680 	 * when modifying the linkage.
2681 	 * ire_dep_parent (Also chain towards IRE for nexthop)
2682 	 * ire_dep_parent_generation: ire_generation of ire_dep_parent
2683 	 * ire_dep_children (From parent to first child)
2684 	 * ire_dep_sib_next (linked list of siblings)
2685 	 * ire_dep_sib_ptpn (linked list of siblings)
2686 	 *
2687 	 * The parent has a ire_refhold on each child, and each child has
2688 	 * an ire_refhold on its parent.
2689 	 * Since ire_dep_parent is a cache setup and torn down independently of
2690 	 * applications we need to use ire_ref{rele,hold}_notr for it.
2691 	 */
2692 	ire_t		*ire_dep_parent;
2693 	ire_t		*ire_dep_children;
2694 	ire_t		*ire_dep_sib_next;
2695 	ire_t		**ire_dep_sib_ptpn;	/* Pointer to previous next */
2696 	uint_t		ire_dep_parent_generation;
2697 
2698 	uint_t		ire_badcnt;	/* Number of times ND_UNREACHABLE */
2699 	uint64_t	ire_last_badcnt;	/* In seconds */
2700 
2701 	/* ire_defense* and ire_last_used_time are only used on IRE_LOCALs */
2702 	uint_t		ire_defense_count;	/* number of ARP conflicts */
2703 	uint_t		ire_defense_time;	/* last time defended (secs) */
2704 
2705 	boolean_t	ire_trace_disable;	/* True when alloc fails */
2706 	ip_stack_t	*ire_ipst;	/* Does not have a netstack_hold */
2707 	iulp_t		ire_metrics;
2708 	/*
2709 	 * default and prefix routes that are added without explicitly
2710 	 * specifying the interface are termed "unbound" routes, and will
2711 	 * have ire_unbound set to true.
2712 	 */
2713 	boolean_t	ire_unbound;
2714 };
2715 
2716 /* IPv4 compatibility macros */
2717 #define	ire_mask		ire_u.ire4_u.ire4_mask
2718 #define	ire_addr		ire_u.ire4_u.ire4_addr
2719 #define	ire_gateway_addr	ire_u.ire4_u.ire4_gateway_addr
2720 #define	ire_setsrc_addr		ire_u.ire4_u.ire4_setsrc_addr
2721 
2722 #define	ire_mask_v6		ire_u.ire6_u.ire6_mask
2723 #define	ire_addr_v6		ire_u.ire6_u.ire6_addr
2724 #define	ire_gateway_addr_v6	ire_u.ire6_u.ire6_gateway_addr
2725 #define	ire_setsrc_addr_v6	ire_u.ire6_u.ire6_setsrc_addr
2726 
2727 /*
2728  * Values for ire_generation.
2729  *
2730  * If an IRE is marked with IRE_IS_CONDEMNED, the last walker of
2731  * the bucket should delete this IRE from this bucket.
2732  *
2733  * IRE_GENERATION_VERIFY is never stored in ire_generation but it is
2734  * stored in places that cache IREs (such as ixa_ire_generation and
2735  * ire_dep_parent_generation). It is used as a signal that the cache is
2736  * stale and needs to be reverified.
2737  */
2738 #define	IRE_GENERATION_CONDEMNED	0
2739 #define	IRE_GENERATION_VERIFY		1
2740 #define	IRE_GENERATION_INITIAL		2
2741 #define	IRE_IS_CONDEMNED(ire) \
2742 	((ire)->ire_generation == IRE_GENERATION_CONDEMNED)
2743 
2744 /* Convenient typedefs for sockaddrs */
2745 typedef	struct sockaddr_in	sin_t;
2746 typedef	struct sockaddr_in6	sin6_t;
2747 
2748 /* Name/Value Descriptor. */
2749 typedef struct nv_s {
2750 	uint64_t nv_value;
2751 	char	*nv_name;
2752 } nv_t;
2753 
2754 #define	ILL_FRAG_HASH(s, i) \
2755 	((ntohl(s) ^ ((i) ^ ((i) >> 8))) % ILL_FRAG_HASH_TBL_COUNT)
2756 
2757 /*
2758  * The MAX number of allowed fragmented packets per hash bucket
2759  * calculation is based on the most common mtu size of 1500. This limit
2760  * will work well for other mtu sizes as well.
2761  */
2762 #define	COMMON_IP_MTU 1500
2763 #define	MAX_FRAG_MIN 10
2764 #define	MAX_FRAG_PKTS(ipst)	\
2765 	MAX(MAX_FRAG_MIN, (2 * (ipst->ips_ip_reass_queue_bytes / \
2766 	    (COMMON_IP_MTU * ILL_FRAG_HASH_TBL_COUNT))))
2767 
2768 /*
2769  * Maximum dups allowed per packet.
2770  */
2771 extern uint_t ip_max_frag_dups;
2772 
2773 /*
2774  * Per-packet information for received packets and transmitted.
2775  * Used by the transport protocols when converting between the packet
2776  * and ancillary data and socket options.
2777  *
2778  * Note: This private data structure and related IPPF_* constant
2779  * definitions are exposed to enable compilation of some debugging tools
2780  * like lsof which use struct tcp_t in <inet/tcp.h>. This is intended to be
2781  * a temporary hack and long term alternate interfaces should be defined
2782  * to support the needs of such tools and private definitions moved to
2783  * private headers.
2784  */
2785 struct ip_pkt_s {
2786 	uint_t		ipp_fields;		/* Which fields are valid */
2787 	in6_addr_t	ipp_addr;		/* pktinfo src/dst addr */
2788 #define	ipp_addr_v4	V4_PART_OF_V6(ipp_addr)
2789 	uint_t		ipp_unicast_hops;	/* IPV6_UNICAST_HOPS, IP_TTL */
2790 	uint_t		ipp_hoplimit;		/* IPV6_HOPLIMIT */
2791 	uint_t		ipp_hopoptslen;
2792 	uint_t		ipp_rthdrdstoptslen;
2793 	uint_t		ipp_rthdrlen;
2794 	uint_t		ipp_dstoptslen;
2795 	uint_t		ipp_fraghdrlen;
2796 	ip6_hbh_t	*ipp_hopopts;
2797 	ip6_dest_t	*ipp_rthdrdstopts;
2798 	ip6_rthdr_t	*ipp_rthdr;
2799 	ip6_dest_t	*ipp_dstopts;
2800 	ip6_frag_t	*ipp_fraghdr;
2801 	uint8_t		ipp_tclass;		/* IPV6_TCLASS */
2802 	uint8_t		ipp_type_of_service;	/* IP_TOS */
2803 	uint_t		ipp_ipv4_options_len;	/* Len of IPv4 options */
2804 	uint8_t		*ipp_ipv4_options;	/* Ptr to IPv4 options */
2805 	uint_t		ipp_label_len_v4;	/* Len of TX label for IPv4 */
2806 	uint8_t		*ipp_label_v4;		/* TX label for IPv4 */
2807 	uint_t		ipp_label_len_v6;	/* Len of TX label for IPv6 */
2808 	uint8_t		*ipp_label_v6;		/* TX label for IPv6 */
2809 };
2810 typedef struct ip_pkt_s ip_pkt_t;
2811 
2812 extern void ip_pkt_free(ip_pkt_t *);	/* free storage inside ip_pkt_t */
2813 extern ipaddr_t ip_pkt_source_route_v4(const ip_pkt_t *);
2814 extern in6_addr_t *ip_pkt_source_route_v6(const ip_pkt_t *);
2815 extern int ip_pkt_copy(ip_pkt_t *, ip_pkt_t *, int);
2816 extern void ip_pkt_source_route_reverse_v4(ip_pkt_t *);
2817 
2818 /* ipp_fields values */
2819 #define	IPPF_ADDR		0x0001	/* Part of in6_pktinfo: src/dst addr */
2820 #define	IPPF_HOPLIMIT		0x0002	/* Overrides unicast and multicast */
2821 #define	IPPF_TCLASS		0x0004	/* Overrides class in sin6_flowinfo */
2822 
2823 #define	IPPF_HOPOPTS		0x0010	/* ipp_hopopts set */
2824 #define	IPPF_RTHDR		0x0020	/* ipp_rthdr set */
2825 #define	IPPF_RTHDRDSTOPTS	0x0040	/* ipp_rthdrdstopts set */
2826 #define	IPPF_DSTOPTS		0x0080	/* ipp_dstopts set */
2827 
2828 #define	IPPF_IPV4_OPTIONS	0x0100	/* ipp_ipv4_options set */
2829 #define	IPPF_LABEL_V4		0x0200	/* ipp_label_v4 set */
2830 #define	IPPF_LABEL_V6		0x0400	/* ipp_label_v6 set */
2831 
2832 #define	IPPF_FRAGHDR		0x0800	/* Used for IPsec receive side */
2833 
2834 /*
2835  * Data structure which is passed to conn_opt_get/set.
2836  * The conn_t is included even though it can be inferred from queue_t.
2837  * setsockopt and getsockopt use conn_ixa and conn_xmit_ipp. However,
2838  * when handling ancillary data we use separate ixa and ipps.
2839  */
2840 typedef struct conn_opt_arg_s {
2841 	conn_t		*coa_connp;
2842 	ip_xmit_attr_t	*coa_ixa;
2843 	ip_pkt_t	*coa_ipp;
2844 	boolean_t	coa_ancillary;	/* Ancillary data and not setsockopt */
2845 	uint_t		coa_changed;	/* See below */
2846 } conn_opt_arg_t;
2847 
2848 /*
2849  * Flags for what changed.
2850  * If we want to be more efficient in the future we can have more fine
2851  * grained flags e.g., a flag for just IP_TOS changing.
2852  * For now we either call ip_set_destination (for "route changed")
2853  * and/or conn_build_hdr_template/conn_prepend_hdr (for "header changed").
2854  */
2855 #define	COA_HEADER_CHANGED	0x0001
2856 #define	COA_ROUTE_CHANGED	0x0002
2857 #define	COA_RCVBUF_CHANGED	0x0004	/* SO_RCVBUF */
2858 #define	COA_SNDBUF_CHANGED	0x0008	/* SO_SNDBUF */
2859 #define	COA_WROFF_CHANGED	0x0010	/* Header size changed */
2860 #define	COA_ICMP_BIND_NEEDED	0x0020
2861 #define	COA_OOBINLINE_CHANGED	0x0040
2862 
2863 #define	TCP_PORTS_OFFSET	0
2864 #define	UDP_PORTS_OFFSET	0
2865 
2866 /*
2867  * lookups return the ill/ipif only if the flags are clear OR Iam writer.
2868  * ill / ipif lookup functions increment the refcnt on the ill / ipif only
2869  * after calling these macros. This ensures that the refcnt on the ipif or
2870  * ill will eventually drop down to zero.
2871  */
2872 #define	ILL_LOOKUP_FAILED	1	/* Used as error code */
2873 #define	IPIF_LOOKUP_FAILED	2	/* Used as error code */
2874 
2875 #define	ILL_CAN_LOOKUP(ill)						\
2876 	(!((ill)->ill_state_flags & ILL_CONDEMNED) ||			\
2877 	IAM_WRITER_ILL(ill))
2878 
2879 #define	ILL_IS_CONDEMNED(ill)	\
2880 	((ill)->ill_state_flags & ILL_CONDEMNED)
2881 
2882 #define	IPIF_CAN_LOOKUP(ipif)	\
2883 	(!((ipif)->ipif_state_flags & IPIF_CONDEMNED) || \
2884 	IAM_WRITER_IPIF(ipif))
2885 
2886 #define	IPIF_IS_CONDEMNED(ipif)	\
2887 	((ipif)->ipif_state_flags & IPIF_CONDEMNED)
2888 
2889 #define	IPIF_IS_CHANGING(ipif)	\
2890 	((ipif)->ipif_state_flags & IPIF_CHANGING)
2891 
2892 /* Macros used to assert that this thread is a writer */
2893 #define	IAM_WRITER_IPSQ(ipsq)	((ipsq)->ipsq_xop->ipx_writer == curthread)
2894 #define	IAM_WRITER_ILL(ill)	IAM_WRITER_IPSQ((ill)->ill_phyint->phyint_ipsq)
2895 #define	IAM_WRITER_IPIF(ipif)	IAM_WRITER_ILL((ipif)->ipif_ill)
2896 
2897 /*
2898  * Grab ill locks in the proper order. The order is highest addressed
2899  * ill is locked first.
2900  */
2901 #define	GRAB_ILL_LOCKS(ill_1, ill_2)				\
2902 {								\
2903 	if ((ill_1) > (ill_2)) {				\
2904 		if (ill_1 != NULL)				\
2905 			mutex_enter(&(ill_1)->ill_lock);	\
2906 		if (ill_2 != NULL)				\
2907 			mutex_enter(&(ill_2)->ill_lock);	\
2908 	} else {						\
2909 		if (ill_2 != NULL)				\
2910 			mutex_enter(&(ill_2)->ill_lock);	\
2911 		if (ill_1 != NULL && ill_1 != ill_2)		\
2912 			mutex_enter(&(ill_1)->ill_lock);	\
2913 	}							\
2914 }
2915 
2916 #define	RELEASE_ILL_LOCKS(ill_1, ill_2)		\
2917 {						\
2918 	if (ill_1 != NULL)			\
2919 		mutex_exit(&(ill_1)->ill_lock);	\
2920 	if (ill_2 != NULL && ill_2 != ill_1)	\
2921 		mutex_exit(&(ill_2)->ill_lock);	\
2922 }
2923 
2924 /* Get the other protocol instance ill */
2925 #define	ILL_OTHER(ill)						\
2926 	((ill)->ill_isv6 ? (ill)->ill_phyint->phyint_illv4 :	\
2927 	    (ill)->ill_phyint->phyint_illv6)
2928 
2929 /* ioctl command info: Ioctl properties extracted and stored in here */
2930 typedef struct cmd_info_s
2931 {
2932 	ipif_t  *ci_ipif;	/* ipif associated with [l]ifreq ioctl's */
2933 	sin_t	*ci_sin;	/* the sin struct passed down */
2934 	sin6_t	*ci_sin6;	/* the sin6_t struct passed down */
2935 	struct lifreq *ci_lifr;	/* the lifreq struct passed down */
2936 } cmd_info_t;
2937 
2938 extern struct kmem_cache *ire_cache;
2939 
2940 extern ipaddr_t	ip_g_all_ones;
2941 
2942 extern uint_t	ip_loopback_mtu;	/* /etc/system */
2943 extern uint_t	ip_loopback_mtuplus;
2944 extern uint_t	ip_loopback_mtu_v6plus;
2945 
2946 extern vmem_t *ip_minor_arena_sa;
2947 extern vmem_t *ip_minor_arena_la;
2948 
2949 /*
2950  * ip_g_forward controls IP forwarding.  It takes two values:
2951  *	0: IP_FORWARD_NEVER	Don't forward packets ever.
2952  *	1: IP_FORWARD_ALWAYS	Forward packets for elsewhere.
2953  *
2954  * RFC1122 says there must be a configuration switch to control forwarding,
2955  * but that the default MUST be to not forward packets ever.  Implicit
2956  * control based on configuration of multiple interfaces MUST NOT be
2957  * implemented (Section 3.1).  SunOS 4.1 did provide the "automatic" capability
2958  * and, in fact, it was the default.  That capability is now provided in the
2959  * /etc/rc2.d/S69inet script.
2960  */
2961 
2962 #define	ips_ip_respond_to_address_mask_broadcast \
2963 					ips_propinfo_tbl[0].prop_cur_bval
2964 #define	ips_ip_g_resp_to_echo_bcast	ips_propinfo_tbl[1].prop_cur_bval
2965 #define	ips_ip_g_resp_to_echo_mcast	ips_propinfo_tbl[2].prop_cur_bval
2966 #define	ips_ip_g_resp_to_timestamp	ips_propinfo_tbl[3].prop_cur_bval
2967 #define	ips_ip_g_resp_to_timestamp_bcast ips_propinfo_tbl[4].prop_cur_bval
2968 #define	ips_ip_g_send_redirects		ips_propinfo_tbl[5].prop_cur_bval
2969 #define	ips_ip_g_forward_directed_bcast	ips_propinfo_tbl[6].prop_cur_bval
2970 #define	ips_ip_mrtdebug			ips_propinfo_tbl[7].prop_cur_uval
2971 #define	ips_ip_ire_reclaim_fraction	ips_propinfo_tbl[8].prop_cur_uval
2972 #define	ips_ip_nce_reclaim_fraction	ips_propinfo_tbl[9].prop_cur_uval
2973 #define	ips_ip_dce_reclaim_fraction	ips_propinfo_tbl[10].prop_cur_uval
2974 #define	ips_ip_def_ttl			ips_propinfo_tbl[11].prop_cur_uval
2975 #define	ips_ip_forward_src_routed	ips_propinfo_tbl[12].prop_cur_bval
2976 #define	ips_ip_wroff_extra		ips_propinfo_tbl[13].prop_cur_uval
2977 #define	ips_ip_pathmtu_interval		ips_propinfo_tbl[14].prop_cur_uval
2978 #define	ips_ip_icmp_return		ips_propinfo_tbl[15].prop_cur_uval
2979 #define	ips_ip_path_mtu_discovery	ips_propinfo_tbl[16].prop_cur_bval
2980 #define	ips_ip_pmtu_min			ips_propinfo_tbl[17].prop_cur_uval
2981 #define	ips_ip_ignore_redirect		ips_propinfo_tbl[18].prop_cur_bval
2982 #define	ips_ip_arp_icmp_error		ips_propinfo_tbl[19].prop_cur_bval
2983 #define	ips_ip_broadcast_ttl		ips_propinfo_tbl[20].prop_cur_uval
2984 #define	ips_ip_icmp_err_interval	ips_propinfo_tbl[21].prop_cur_uval
2985 #define	ips_ip_icmp_err_burst		ips_propinfo_tbl[22].prop_cur_uval
2986 #define	ips_ip_reass_queue_bytes	ips_propinfo_tbl[23].prop_cur_uval
2987 #define	ips_ip_strict_dst_multihoming	ips_propinfo_tbl[24].prop_cur_uval
2988 #define	ips_ip_addrs_per_if		ips_propinfo_tbl[25].prop_cur_uval
2989 #define	ips_ipsec_override_persocket_policy ips_propinfo_tbl[26].prop_cur_bval
2990 #define	ips_icmp_accept_clear_messages	ips_propinfo_tbl[27].prop_cur_bval
2991 #define	ips_igmp_accept_clear_messages	ips_propinfo_tbl[28].prop_cur_bval
2992 
2993 /* IPv6 configuration knobs */
2994 #define	ips_delay_first_probe_time	ips_propinfo_tbl[29].prop_cur_uval
2995 #define	ips_max_unicast_solicit		ips_propinfo_tbl[30].prop_cur_uval
2996 #define	ips_ipv6_def_hops		ips_propinfo_tbl[31].prop_cur_uval
2997 #define	ips_ipv6_icmp_return		ips_propinfo_tbl[32].prop_cur_uval
2998 #define	ips_ipv6_forward_src_routed	ips_propinfo_tbl[33].prop_cur_bval
2999 #define	ips_ipv6_resp_echo_mcast	ips_propinfo_tbl[34].prop_cur_bval
3000 #define	ips_ipv6_send_redirects		ips_propinfo_tbl[35].prop_cur_bval
3001 #define	ips_ipv6_ignore_redirect	ips_propinfo_tbl[36].prop_cur_bval
3002 #define	ips_ipv6_strict_dst_multihoming	ips_propinfo_tbl[37].prop_cur_uval
3003 #define	ips_src_check			ips_propinfo_tbl[38].prop_cur_uval
3004 #define	ips_ipsec_policy_log_interval	ips_propinfo_tbl[39].prop_cur_uval
3005 #define	ips_pim_accept_clear_messages	ips_propinfo_tbl[40].prop_cur_bval
3006 #define	ips_ip_ndp_unsolicit_interval	ips_propinfo_tbl[41].prop_cur_uval
3007 #define	ips_ip_ndp_unsolicit_count	ips_propinfo_tbl[42].prop_cur_uval
3008 #define	ips_ipv6_ignore_home_address_opt ips_propinfo_tbl[43].prop_cur_bval
3009 
3010 /* Misc IP configuration knobs */
3011 #define	ips_ip_policy_mask		ips_propinfo_tbl[44].prop_cur_uval
3012 #define	ips_ip_ecmp_behavior		ips_propinfo_tbl[45].prop_cur_uval
3013 #define	ips_ip_multirt_ttl		ips_propinfo_tbl[46].prop_cur_uval
3014 #define	ips_ip_ire_badcnt_lifetime	ips_propinfo_tbl[47].prop_cur_uval
3015 #define	ips_ip_max_temp_idle		ips_propinfo_tbl[48].prop_cur_uval
3016 #define	ips_ip_max_temp_defend		ips_propinfo_tbl[49].prop_cur_uval
3017 #define	ips_ip_max_defend		ips_propinfo_tbl[50].prop_cur_uval
3018 #define	ips_ip_defend_interval		ips_propinfo_tbl[51].prop_cur_uval
3019 #define	ips_ip_dup_recovery		ips_propinfo_tbl[52].prop_cur_uval
3020 #define	ips_ip_restrict_interzone_loopback ips_propinfo_tbl[53].prop_cur_bval
3021 #define	ips_ip_lso_outbound		ips_propinfo_tbl[54].prop_cur_bval
3022 #define	ips_igmp_max_version		ips_propinfo_tbl[55].prop_cur_uval
3023 #define	ips_mld_max_version		ips_propinfo_tbl[56].prop_cur_uval
3024 #define	ips_ip_forwarding		ips_propinfo_tbl[57].prop_cur_bval
3025 #define	ips_ipv6_forwarding		ips_propinfo_tbl[58].prop_cur_bval
3026 #define	ips_ip_reassembly_timeout	ips_propinfo_tbl[59].prop_cur_uval
3027 #define	ips_ipv6_reassembly_timeout	ips_propinfo_tbl[60].prop_cur_uval
3028 #define	ips_ip_cgtp_filter		ips_propinfo_tbl[61].prop_cur_bval
3029 #define	ips_arp_probe_delay		ips_propinfo_tbl[62].prop_cur_uval
3030 #define	ips_arp_fastprobe_delay		ips_propinfo_tbl[63].prop_cur_uval
3031 #define	ips_arp_probe_interval		ips_propinfo_tbl[64].prop_cur_uval
3032 #define	ips_arp_fastprobe_interval	ips_propinfo_tbl[65].prop_cur_uval
3033 #define	ips_arp_probe_count		ips_propinfo_tbl[66].prop_cur_uval
3034 #define	ips_arp_fastprobe_count		ips_propinfo_tbl[67].prop_cur_uval
3035 #define	ips_ipv4_dad_announce_interval	ips_propinfo_tbl[68].prop_cur_uval
3036 #define	ips_ipv6_dad_announce_interval	ips_propinfo_tbl[69].prop_cur_uval
3037 #define	ips_arp_defend_interval		ips_propinfo_tbl[70].prop_cur_uval
3038 #define	ips_arp_defend_rate		ips_propinfo_tbl[71].prop_cur_uval
3039 #define	ips_ndp_defend_interval		ips_propinfo_tbl[72].prop_cur_uval
3040 #define	ips_ndp_defend_rate		ips_propinfo_tbl[73].prop_cur_uval
3041 #define	ips_arp_defend_period		ips_propinfo_tbl[74].prop_cur_uval
3042 #define	ips_ndp_defend_period		ips_propinfo_tbl[75].prop_cur_uval
3043 #define	ips_ipv4_icmp_return_pmtu	ips_propinfo_tbl[76].prop_cur_bval
3044 #define	ips_ipv6_icmp_return_pmtu	ips_propinfo_tbl[77].prop_cur_bval
3045 #define	ips_ip_arp_publish_count	ips_propinfo_tbl[78].prop_cur_uval
3046 #define	ips_ip_arp_publish_interval	ips_propinfo_tbl[79].prop_cur_uval
3047 #define	ips_ip_strict_src_multihoming	ips_propinfo_tbl[80].prop_cur_uval
3048 #define	ips_ipv6_strict_src_multihoming	ips_propinfo_tbl[81].prop_cur_uval
3049 #define	ips_ipv6_drop_inbound_icmpv6	ips_propinfo_tbl[82].prop_cur_bval
3050 #define	ips_ip_dce_reclaim_threshold	ips_propinfo_tbl[83].prop_cur_uval
3051 
3052 extern int	dohwcksum;	/* use h/w cksum if supported by the h/w */
3053 #ifdef ZC_TEST
3054 extern int	noswcksum;
3055 #endif
3056 
3057 extern char	ipif_loopback_name[];
3058 
3059 extern nv_t	*ire_nv_tbl;
3060 
3061 extern struct module_info ip_mod_info;
3062 
3063 #define	HOOKS4_INTERESTED_PHYSICAL_IN(ipst)	\
3064 	((ipst)->ips_ip4_physical_in_event.he_interested)
3065 #define	HOOKS6_INTERESTED_PHYSICAL_IN(ipst)	\
3066 	((ipst)->ips_ip6_physical_in_event.he_interested)
3067 #define	HOOKS4_INTERESTED_PHYSICAL_OUT(ipst)	\
3068 	((ipst)->ips_ip4_physical_out_event.he_interested)
3069 #define	HOOKS6_INTERESTED_PHYSICAL_OUT(ipst)	\
3070 	((ipst)->ips_ip6_physical_out_event.he_interested)
3071 #define	HOOKS4_INTERESTED_FORWARDING(ipst)	\
3072 	((ipst)->ips_ip4_forwarding_event.he_interested)
3073 #define	HOOKS6_INTERESTED_FORWARDING(ipst)	\
3074 	((ipst)->ips_ip6_forwarding_event.he_interested)
3075 #define	HOOKS4_INTERESTED_LOOPBACK_IN(ipst)	\
3076 	((ipst)->ips_ip4_loopback_in_event.he_interested)
3077 #define	HOOKS6_INTERESTED_LOOPBACK_IN(ipst)	\
3078 	((ipst)->ips_ip6_loopback_in_event.he_interested)
3079 #define	HOOKS4_INTERESTED_LOOPBACK_OUT(ipst)	\
3080 	((ipst)->ips_ip4_loopback_out_event.he_interested)
3081 #define	HOOKS6_INTERESTED_LOOPBACK_OUT(ipst)	\
3082 	((ipst)->ips_ip6_loopback_out_event.he_interested)
3083 /*
3084  * Hooks marcos used inside of ip
3085  * The callers use the above INTERESTED macros first, hence
3086  * the he_interested check is superflous.
3087  */
3088 #define	FW_HOOKS(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst, _err) \
3089 	if ((_hook).he_interested) {					\
3090 		hook_pkt_event_t info;					\
3091 									\
3092 		_NOTE(CONSTCOND)					\
3093 		ASSERT((_ilp != NULL) || (_olp != NULL));		\
3094 									\
3095 		FW_SET_ILL_INDEX(info.hpe_ifp, (ill_t *)_ilp);		\
3096 		FW_SET_ILL_INDEX(info.hpe_ofp, (ill_t *)_olp);		\
3097 		info.hpe_protocol = ipst->ips_ipv4_net_data;		\
3098 		info.hpe_hdr = _iph;					\
3099 		info.hpe_mp = &(_fm);					\
3100 		info.hpe_mb = _m;					\
3101 		info.hpe_flags = _llm;					\
3102 		_err = hook_run(ipst->ips_ipv4_net_data->netd_hooks,	\
3103 		    _event, (hook_data_t)&info);			\
3104 		if (_err != 0) {					\
3105 			ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
3106 			    (_hook).he_name, (void *)_fm, (void *)_m));	\
3107 			if (_fm != NULL) {				\
3108 				freemsg(_fm);				\
3109 				_fm = NULL;				\
3110 			}						\
3111 			_iph = NULL;					\
3112 			_m = NULL;					\
3113 		} else {						\
3114 			_iph = info.hpe_hdr;				\
3115 			_m = info.hpe_mb;				\
3116 		}							\
3117 	}
3118 
3119 #define	FW_HOOKS6(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst, _err) \
3120 	if ((_hook).he_interested) {					\
3121 		hook_pkt_event_t info;					\
3122 									\
3123 		_NOTE(CONSTCOND)					\
3124 		ASSERT((_ilp != NULL) || (_olp != NULL));		\
3125 									\
3126 		FW_SET_ILL_INDEX(info.hpe_ifp, (ill_t *)_ilp);		\
3127 		FW_SET_ILL_INDEX(info.hpe_ofp, (ill_t *)_olp);		\
3128 		info.hpe_protocol = ipst->ips_ipv6_net_data;		\
3129 		info.hpe_hdr = _iph;					\
3130 		info.hpe_mp = &(_fm);					\
3131 		info.hpe_mb = _m;					\
3132 		info.hpe_flags = _llm;					\
3133 		_err = hook_run(ipst->ips_ipv6_net_data->netd_hooks,	\
3134 		    _event, (hook_data_t)&info);			\
3135 		if (_err != 0) {					\
3136 			ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
3137 			    (_hook).he_name, (void *)_fm, (void *)_m));	\
3138 			if (_fm != NULL) {				\
3139 				freemsg(_fm);				\
3140 				_fm = NULL;				\
3141 			}						\
3142 			_iph = NULL;					\
3143 			_m = NULL;					\
3144 		} else {						\
3145 			_iph = info.hpe_hdr;				\
3146 			_m = info.hpe_mb;				\
3147 		}							\
3148 	}
3149 
3150 #define	FW_SET_ILL_INDEX(fp, ill)					\
3151 	_NOTE(CONSTCOND)						\
3152 	if ((ill) == NULL || (ill)->ill_phyint == NULL) {		\
3153 		(fp) = 0;						\
3154 		_NOTE(CONSTCOND)					\
3155 	} else if (IS_UNDER_IPMP(ill)) {				\
3156 		(fp) = ipmp_ill_get_ipmp_ifindex(ill);			\
3157 	} else {							\
3158 		(fp) = (ill)->ill_phyint->phyint_ifindex;		\
3159 	}
3160 
3161 /*
3162  * Network byte order macros
3163  */
3164 #ifdef	_BIG_ENDIAN
3165 #define	N_IN_CLASSA_NET		IN_CLASSA_NET
3166 #define	N_IN_CLASSD_NET		IN_CLASSD_NET
3167 #define	N_INADDR_UNSPEC_GROUP	INADDR_UNSPEC_GROUP
3168 #define	N_IN_LOOPBACK_NET	(ipaddr_t)0x7f000000U
3169 #else /* _BIG_ENDIAN */
3170 #define	N_IN_CLASSA_NET		(ipaddr_t)0x000000ffU
3171 #define	N_IN_CLASSD_NET		(ipaddr_t)0x000000f0U
3172 #define	N_INADDR_UNSPEC_GROUP	(ipaddr_t)0x000000e0U
3173 #define	N_IN_LOOPBACK_NET	(ipaddr_t)0x0000007fU
3174 #endif /* _BIG_ENDIAN */
3175 #define	CLASSD(addr)	(((addr) & N_IN_CLASSD_NET) == N_INADDR_UNSPEC_GROUP)
3176 #define	CLASSE(addr)	(((addr) & N_IN_CLASSD_NET) == N_IN_CLASSD_NET)
3177 #define	IP_LOOPBACK_ADDR(addr)			\
3178 	(((addr) & N_IN_CLASSA_NET == N_IN_LOOPBACK_NET))
3179 
3180 extern int	ip_debug;
3181 extern uint_t	ip_thread_data;
3182 extern krwlock_t ip_thread_rwlock;
3183 extern list_t	ip_thread_list;
3184 
3185 #ifdef IP_DEBUG
3186 #include <sys/debug.h>
3187 #include <sys/promif.h>
3188 
3189 #define	ip0dbg(a)	printf a
3190 #define	ip1dbg(a)	if (ip_debug > 2) printf a
3191 #define	ip2dbg(a)	if (ip_debug > 3) printf a
3192 #define	ip3dbg(a)	if (ip_debug > 4) printf a
3193 #else
3194 #define	ip0dbg(a)	/* */
3195 #define	ip1dbg(a)	/* */
3196 #define	ip2dbg(a)	/* */
3197 #define	ip3dbg(a)	/* */
3198 #endif	/* IP_DEBUG */
3199 
3200 /* Default MAC-layer address string length for mac_colon_addr */
3201 #define	MAC_STR_LEN	128
3202 
3203 struct	mac_header_info_s;
3204 
3205 extern void	ill_frag_timer(void *);
3206 extern ill_t	*ill_first(int, int, ill_walk_context_t *, ip_stack_t *);
3207 extern ill_t	*ill_next(ill_walk_context_t *, ill_t *);
3208 extern void	ill_frag_timer_start(ill_t *);
3209 extern void	ill_nic_event_dispatch(ill_t *, lif_if_t, nic_event_t,
3210     nic_event_data_t, size_t);
3211 extern mblk_t	*ip_carve_mp(mblk_t **, ssize_t);
3212 extern mblk_t	*ip_dlpi_alloc(size_t, t_uscalar_t);
3213 extern mblk_t	*ip_dlnotify_alloc(uint_t, uint_t);
3214 extern mblk_t	*ip_dlnotify_alloc2(uint_t, uint_t, uint_t);
3215 extern char	*ip_dot_addr(ipaddr_t, char *);
3216 extern const char *mac_colon_addr(const uint8_t *, size_t, char *, size_t);
3217 extern int	ip_lwput(queue_t *, mblk_t *);
3218 extern boolean_t icmp_err_rate_limit(ip_stack_t *);
3219 extern void	icmp_frag_needed(mblk_t *, int, ip_recv_attr_t *);
3220 extern mblk_t	*icmp_inbound_v4(mblk_t *, ip_recv_attr_t *);
3221 extern void	icmp_time_exceeded(mblk_t *, uint8_t, ip_recv_attr_t *);
3222 extern void	icmp_unreachable(mblk_t *, uint8_t, ip_recv_attr_t *);
3223 extern boolean_t ip_ipsec_policy_inherit(conn_t *, conn_t *, ip_recv_attr_t *);
3224 extern void	*ip_pullup(mblk_t *, ssize_t, ip_recv_attr_t *);
3225 extern void	ip_setl2src(mblk_t *, ip_recv_attr_t *, ill_t *);
3226 extern mblk_t	*ip_check_and_align_header(mblk_t *, uint_t, ip_recv_attr_t *);
3227 extern mblk_t	*ip_check_length(mblk_t *, uchar_t *, ssize_t, uint_t, uint_t,
3228     ip_recv_attr_t *);
3229 extern mblk_t	*ip_check_optlen(mblk_t *, ipha_t *, uint_t, uint_t,
3230     ip_recv_attr_t *);
3231 extern mblk_t	*ip_fix_dbref(mblk_t *, ip_recv_attr_t *);
3232 extern uint_t	ip_cksum(mblk_t *, int, uint32_t);
3233 extern int	ip_close(queue_t *, int, cred_t *);
3234 extern uint16_t	ip_csum_hdr(ipha_t *);
3235 extern void	ip_forward_xmit_v4(nce_t *, ill_t *, mblk_t *, ipha_t *,
3236     ip_recv_attr_t *, uint32_t, uint32_t);
3237 extern boolean_t ip_forward_options(mblk_t *, ipha_t *, ill_t *,
3238     ip_recv_attr_t *);
3239 extern int	ip_fragment_v4(mblk_t *, nce_t *, iaflags_t, uint_t, uint32_t,
3240     uint32_t, zoneid_t, zoneid_t, pfirepostfrag_t postfragfn,
3241     uintptr_t *cookie);
3242 extern void	ip_proto_not_sup(mblk_t *, ip_recv_attr_t *);
3243 extern void	ip_ire_g_fini(void);
3244 extern void	ip_ire_g_init(void);
3245 extern void	ip_ire_fini(ip_stack_t *);
3246 extern void	ip_ire_init(ip_stack_t *);
3247 extern void	ip_mdata_to_mhi(ill_t *, mblk_t *, struct mac_header_info_s *);
3248 extern int	ip_openv4(queue_t *q, dev_t *devp, int flag, int sflag,
3249 		    cred_t *credp);
3250 extern int	ip_openv6(queue_t *q, dev_t *devp, int flag, int sflag,
3251 		    cred_t *credp);
3252 extern int	ip_reassemble(mblk_t *, ipf_t *, uint_t, boolean_t, ill_t *,
3253     size_t);
3254 extern int	ip_rput(queue_t *, mblk_t *);
3255 extern void	ip_input(ill_t *, ill_rx_ring_t *, mblk_t *,
3256     struct mac_header_info_s *);
3257 extern void	ip_input_v6(ill_t *, ill_rx_ring_t *, mblk_t *,
3258     struct mac_header_info_s *);
3259 extern mblk_t	*ip_input_common_v4(ill_t *, ill_rx_ring_t *, mblk_t *,
3260     struct mac_header_info_s *, squeue_t *, mblk_t **, uint_t *);
3261 extern mblk_t	*ip_input_common_v6(ill_t *, ill_rx_ring_t *, mblk_t *,
3262     struct mac_header_info_s *, squeue_t *, mblk_t **, uint_t *);
3263 extern void	ill_input_full_v4(mblk_t *, void *, void *,
3264     ip_recv_attr_t *, rtc_t *);
3265 extern void	ill_input_short_v4(mblk_t *, void *, void *,
3266     ip_recv_attr_t *, rtc_t *);
3267 extern void	ill_input_full_v6(mblk_t *, void *, void *,
3268     ip_recv_attr_t *, rtc_t *);
3269 extern void	ill_input_short_v6(mblk_t *, void *, void *,
3270     ip_recv_attr_t *, rtc_t *);
3271 extern ipaddr_t	ip_input_options(ipha_t *, ipaddr_t, mblk_t *,
3272     ip_recv_attr_t *, int *);
3273 extern boolean_t ip_input_local_options(mblk_t *, ipha_t *, ip_recv_attr_t *);
3274 extern mblk_t	*ip_input_fragment(mblk_t *, ipha_t *, ip_recv_attr_t *);
3275 extern mblk_t	*ip_input_fragment_v6(mblk_t *, ip6_t *, ip6_frag_t *, uint_t,
3276     ip_recv_attr_t *);
3277 extern void	ip_input_post_ipsec(mblk_t *, ip_recv_attr_t *);
3278 extern void	ip_fanout_v4(mblk_t *, ipha_t *, ip_recv_attr_t *);
3279 extern void	ip_fanout_v6(mblk_t *, ip6_t *, ip_recv_attr_t *);
3280 extern void	ip_fanout_proto_conn(conn_t *, mblk_t *, ipha_t *, ip6_t *,
3281     ip_recv_attr_t *);
3282 extern void	ip_fanout_proto_v4(mblk_t *, ipha_t *, ip_recv_attr_t *);
3283 extern void	ip_fanout_send_icmp_v4(mblk_t *, uint_t, uint_t,
3284     ip_recv_attr_t *);
3285 extern void	ip_fanout_udp_conn(conn_t *, mblk_t *, ipha_t *, ip6_t *,
3286     ip_recv_attr_t *);
3287 extern void	ip_fanout_udp_multi_v4(mblk_t *, ipha_t *, uint16_t, uint16_t,
3288     ip_recv_attr_t *);
3289 extern mblk_t	*zero_spi_check(mblk_t *, ip_recv_attr_t *);
3290 extern void	ip_build_hdrs_v4(uchar_t *, uint_t, const ip_pkt_t *, uint8_t);
3291 extern int	ip_find_hdr_v4(ipha_t *, ip_pkt_t *, boolean_t);
3292 extern int	ip_total_hdrs_len_v4(const ip_pkt_t *);
3293 
3294 extern mblk_t	*ip_accept_tcp(ill_t *, ill_rx_ring_t *, squeue_t *,
3295     mblk_t *, mblk_t **, uint_t *cnt);
3296 extern void	ip_rput_dlpi(ill_t *, mblk_t *);
3297 extern void	ip_rput_notdata(ill_t *, mblk_t *);
3298 
3299 extern void	ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *,
3300 		    mib2_ipIfStatsEntry_t *);
3301 extern void	ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *,
3302 		    mib2_ipv6IfIcmpEntry_t *);
3303 extern void	ip_rput_other(ipsq_t *, queue_t *, mblk_t *, void *);
3304 extern ire_t	*ip_check_multihome(void *, ire_t *, ill_t *);
3305 extern void	ip_send_potential_redirect_v4(mblk_t *, ipha_t *, ire_t *,
3306     ip_recv_attr_t *);
3307 extern int	ip_set_destination_v4(ipaddr_t *, ipaddr_t, ipaddr_t,
3308     ip_xmit_attr_t *, iulp_t *, uint32_t, uint_t);
3309 extern int	ip_set_destination_v6(in6_addr_t *, const in6_addr_t *,
3310     const in6_addr_t *, ip_xmit_attr_t *, iulp_t *, uint32_t, uint_t);
3311 
3312 extern int	ip_output_simple(mblk_t *, ip_xmit_attr_t *);
3313 extern int	ip_output_simple_v4(mblk_t *, ip_xmit_attr_t *);
3314 extern int	ip_output_simple_v6(mblk_t *, ip_xmit_attr_t *);
3315 extern int	ip_output_options(mblk_t *, ipha_t *, ip_xmit_attr_t *,
3316     ill_t *);
3317 extern void	ip_output_local_options(ipha_t *, ip_stack_t *);
3318 
3319 extern ip_xmit_attr_t *conn_get_ixa(conn_t *, boolean_t);
3320 extern ip_xmit_attr_t *conn_get_ixa_tryhard(conn_t *, boolean_t);
3321 extern ip_xmit_attr_t *conn_replace_ixa(conn_t *, ip_xmit_attr_t *);
3322 extern ip_xmit_attr_t *conn_get_ixa_exclusive(conn_t *);
3323 extern ip_xmit_attr_t *ip_xmit_attr_duplicate(ip_xmit_attr_t *);
3324 extern void	ip_xmit_attr_replace_tsl(ip_xmit_attr_t *, ts_label_t *);
3325 extern void	ip_xmit_attr_restore_tsl(ip_xmit_attr_t *, cred_t *);
3326 boolean_t	ip_recv_attr_replace_label(ip_recv_attr_t *, ts_label_t *);
3327 extern void	ixa_inactive(ip_xmit_attr_t *);
3328 extern void	ixa_refrele(ip_xmit_attr_t *);
3329 extern boolean_t ixa_check_drain_insert(conn_t *, ip_xmit_attr_t *);
3330 extern void	ixa_cleanup(ip_xmit_attr_t *);
3331 extern void	ira_cleanup(ip_recv_attr_t *, boolean_t);
3332 extern void	ixa_safe_copy(ip_xmit_attr_t *, ip_xmit_attr_t *);
3333 
3334 extern int	conn_ip_output(mblk_t *, ip_xmit_attr_t *);
3335 extern boolean_t ip_output_verify_local(ip_xmit_attr_t *);
3336 extern mblk_t	*ip_output_process_local(mblk_t *, ip_xmit_attr_t *, boolean_t,
3337     boolean_t, conn_t *);
3338 
3339 extern int	conn_opt_get(conn_opt_arg_t *, t_scalar_t, t_scalar_t,
3340     uchar_t *);
3341 extern int	conn_opt_set(conn_opt_arg_t *, t_scalar_t, t_scalar_t, uint_t,
3342     uchar_t *, boolean_t, cred_t *);
3343 extern boolean_t	conn_same_as_last_v4(conn_t *, sin_t *);
3344 extern boolean_t	conn_same_as_last_v6(conn_t *, sin6_t *);
3345 extern int	conn_update_label(const conn_t *, const ip_xmit_attr_t *,
3346     const in6_addr_t *, ip_pkt_t *);
3347 
3348 extern int	ip_opt_set_multicast_group(conn_t *, t_scalar_t,
3349     uchar_t *, boolean_t, boolean_t);
3350 extern int	ip_opt_set_multicast_sources(conn_t *, t_scalar_t,
3351     uchar_t *, boolean_t, boolean_t);
3352 extern int	conn_getsockname(conn_t *, struct sockaddr *, uint_t *);
3353 extern int	conn_getpeername(conn_t *, struct sockaddr *, uint_t *);
3354 
3355 extern int	conn_build_hdr_template(conn_t *, uint_t, uint_t,
3356     const in6_addr_t *, const in6_addr_t *, uint32_t);
3357 extern mblk_t	*conn_prepend_hdr(ip_xmit_attr_t *, const ip_pkt_t *,
3358     const in6_addr_t *, const in6_addr_t *, uint8_t, uint32_t, uint_t,
3359     mblk_t *, uint_t, uint_t, uint32_t *, int *);
3360 extern void	ip_attr_newdst(ip_xmit_attr_t *);
3361 extern void	ip_attr_nexthop(const ip_pkt_t *, const ip_xmit_attr_t *,
3362     const in6_addr_t *, in6_addr_t *);
3363 extern int	conn_connect(conn_t *, iulp_t *, uint32_t);
3364 extern int	ip_attr_connect(const conn_t *, ip_xmit_attr_t *,
3365     const in6_addr_t *, const in6_addr_t *, const in6_addr_t *, in_port_t,
3366     in6_addr_t *, iulp_t *, uint32_t);
3367 extern int	conn_inherit_parent(conn_t *, conn_t *);
3368 
3369 extern void	conn_ixa_cleanup(conn_t *connp, void *arg);
3370 
3371 extern boolean_t conn_wantpacket(conn_t *, ip_recv_attr_t *, ipha_t *);
3372 extern uint_t	ip_type_v4(ipaddr_t, ip_stack_t *);
3373 extern uint_t	ip_type_v6(const in6_addr_t *, ip_stack_t *);
3374 
3375 extern int	ip_wput_nondata(queue_t *, mblk_t *);
3376 extern int	ip_wsrv(queue_t *);
3377 extern char	*ip_nv_lookup(nv_t *, int);
3378 extern boolean_t ip_local_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3379 extern boolean_t ip_remote_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3380 extern ipaddr_t ip_massage_options(ipha_t *, netstack_t *);
3381 extern ipaddr_t ip_net_mask(ipaddr_t);
3382 extern void	arp_bringup_done(ill_t *, int);
3383 extern void	arp_replumb_done(ill_t *, int);
3384 
3385 extern struct qinit iprinitv6;
3386 
3387 extern void	ipmp_init(ip_stack_t *);
3388 extern void	ipmp_destroy(ip_stack_t *);
3389 extern ipmp_grp_t *ipmp_grp_create(const char *, phyint_t *);
3390 extern void	ipmp_grp_destroy(ipmp_grp_t *);
3391 extern void	ipmp_grp_info(const ipmp_grp_t *, lifgroupinfo_t *);
3392 extern int	ipmp_grp_rename(ipmp_grp_t *, const char *);
3393 extern ipmp_grp_t *ipmp_grp_lookup(const char *, ip_stack_t *);
3394 extern int	ipmp_grp_vet_phyint(ipmp_grp_t *, phyint_t *);
3395 extern ipmp_illgrp_t *ipmp_illgrp_create(ill_t *);
3396 extern void	ipmp_illgrp_destroy(ipmp_illgrp_t *);
3397 extern ill_t	*ipmp_illgrp_add_ipif(ipmp_illgrp_t *, ipif_t *);
3398 extern void	ipmp_illgrp_del_ipif(ipmp_illgrp_t *, ipif_t *);
3399 extern ill_t	*ipmp_illgrp_next_ill(ipmp_illgrp_t *);
3400 extern ill_t	*ipmp_illgrp_hold_next_ill(ipmp_illgrp_t *);
3401 extern ill_t	*ipmp_illgrp_hold_cast_ill(ipmp_illgrp_t *);
3402 extern ill_t	*ipmp_illgrp_ipmp_ill(ipmp_illgrp_t *);
3403 extern void	ipmp_illgrp_refresh_mtu(ipmp_illgrp_t *);
3404 extern ipmp_arpent_t *ipmp_illgrp_create_arpent(ipmp_illgrp_t *,
3405     boolean_t, ipaddr_t, uchar_t *, size_t, uint16_t);
3406 extern void	ipmp_illgrp_destroy_arpent(ipmp_illgrp_t *, ipmp_arpent_t *);
3407 extern ipmp_arpent_t *ipmp_illgrp_lookup_arpent(ipmp_illgrp_t *, ipaddr_t *);
3408 extern void	ipmp_illgrp_refresh_arpent(ipmp_illgrp_t *);
3409 extern void	ipmp_illgrp_mark_arpent(ipmp_illgrp_t *, ipmp_arpent_t *);
3410 extern ill_t	*ipmp_illgrp_find_ill(ipmp_illgrp_t *, uchar_t *, uint_t);
3411 extern void	ipmp_illgrp_link_grp(ipmp_illgrp_t *, ipmp_grp_t *);
3412 extern int	ipmp_illgrp_unlink_grp(ipmp_illgrp_t *);
3413 extern uint_t	ipmp_ill_get_ipmp_ifindex(const ill_t *);
3414 extern void	ipmp_ill_join_illgrp(ill_t *, ipmp_illgrp_t *);
3415 extern void	ipmp_ill_leave_illgrp(ill_t *);
3416 extern ill_t	*ipmp_ill_hold_ipmp_ill(ill_t *);
3417 extern ill_t	*ipmp_ill_hold_xmit_ill(ill_t *, boolean_t);
3418 extern boolean_t ipmp_ill_is_active(ill_t *);
3419 extern void	ipmp_ill_refresh_active(ill_t *);
3420 extern void	ipmp_phyint_join_grp(phyint_t *, ipmp_grp_t *);
3421 extern void	ipmp_phyint_leave_grp(phyint_t *);
3422 extern void	ipmp_phyint_refresh_active(phyint_t *);
3423 extern ill_t	*ipmp_ipif_bound_ill(const ipif_t *);
3424 extern ill_t	*ipmp_ipif_hold_bound_ill(const ipif_t *);
3425 extern boolean_t ipmp_ipif_is_dataaddr(const ipif_t *);
3426 extern boolean_t ipmp_ipif_is_stubaddr(const ipif_t *);
3427 extern boolean_t ipmp_packet_is_probe(mblk_t *, ill_t *);
3428 extern void	ipmp_ncec_delete_nce(ncec_t *);
3429 extern void	ipmp_ncec_refresh_nce(ncec_t *);
3430 
3431 extern void	conn_drain_insert(conn_t *, idl_tx_list_t *);
3432 extern void	conn_setqfull(conn_t *, boolean_t *);
3433 extern void	conn_clrqfull(conn_t *, boolean_t *);
3434 extern int	conn_ipsec_length(conn_t *);
3435 extern ipaddr_t	ip_get_dst(ipha_t *);
3436 extern uint_t	ip_get_pmtu(ip_xmit_attr_t *);
3437 extern uint_t	ip_get_base_mtu(ill_t *, ire_t *);
3438 extern mblk_t *ip_output_attach_policy(mblk_t *, ipha_t *, ip6_t *,
3439     const conn_t *, ip_xmit_attr_t *);
3440 extern int	ipsec_out_extra_length(ip_xmit_attr_t *);
3441 extern int	ipsec_out_process(mblk_t *, ip_xmit_attr_t *);
3442 extern int	ip_output_post_ipsec(mblk_t *, ip_xmit_attr_t *);
3443 extern void	ipsec_out_to_in(ip_xmit_attr_t *, ill_t *ill,
3444     ip_recv_attr_t *);
3445 
3446 extern void	ire_cleanup(ire_t *);
3447 extern void	ire_inactive(ire_t *);
3448 extern boolean_t irb_inactive(irb_t *);
3449 extern ire_t	*ire_unlink(irb_t *);
3450 
3451 #ifdef DEBUG
3452 extern	boolean_t th_trace_ref(const void *, ip_stack_t *);
3453 extern	void	th_trace_unref(const void *);
3454 extern	void	th_trace_cleanup(const void *, boolean_t);
3455 extern	void	ire_trace_ref(ire_t *);
3456 extern	void	ire_untrace_ref(ire_t *);
3457 #endif
3458 
3459 extern int	ip_srcid_insert(const in6_addr_t *, zoneid_t, ip_stack_t *);
3460 extern int	ip_srcid_remove(const in6_addr_t *, zoneid_t, ip_stack_t *);
3461 extern boolean_t ip_srcid_find_id(uint_t, in6_addr_t *, zoneid_t, boolean_t,
3462     netstack_t *);
3463 extern uint_t	ip_srcid_find_addr(const in6_addr_t *, zoneid_t, netstack_t *);
3464 
3465 extern uint8_t	ipoptp_next(ipoptp_t *);
3466 extern uint8_t	ipoptp_first(ipoptp_t *, ipha_t *);
3467 extern int	ip_opt_get_user(conn_t *, uchar_t *);
3468 extern int	ipsec_req_from_conn(conn_t *, ipsec_req_t *, int);
3469 extern int	ip_snmp_get(queue_t *q, mblk_t *mctl, int level, boolean_t);
3470 extern int	ip_snmp_set(queue_t *q, int, int, uchar_t *, int);
3471 extern void	ip_process_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3472 extern void	ip_quiesce_conn(conn_t *);
3473 extern  void    ip_reprocess_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3474 extern void	ip_ioctl_finish(queue_t *, mblk_t *, int, int, ipsq_t *);
3475 
3476 extern boolean_t ip_cmpbuf(const void *, uint_t, boolean_t, const void *,
3477     uint_t);
3478 extern boolean_t ip_allocbuf(void **, uint_t *, boolean_t, const void *,
3479     uint_t);
3480 extern void	ip_savebuf(void **, uint_t *, boolean_t, const void *, uint_t);
3481 
3482 extern boolean_t	ipsq_pending_mp_cleanup(ill_t *, conn_t *);
3483 extern void	conn_ioctl_cleanup(conn_t *);
3484 
3485 extern void	ip_unbind(conn_t *);
3486 
3487 extern void tnet_init(void);
3488 extern void tnet_fini(void);
3489 
3490 /*
3491  * Hook functions to enable cluster networking
3492  * On non-clustered systems these vectors must always be NULL.
3493  */
3494 extern int (*cl_inet_isclusterwide)(netstackid_t stack_id, uint8_t protocol,
3495     sa_family_t addr_family, uint8_t *laddrp, void *args);
3496 extern uint32_t (*cl_inet_ipident)(netstackid_t stack_id, uint8_t protocol,
3497     sa_family_t addr_family, uint8_t *laddrp, uint8_t *faddrp,
3498     void *args);
3499 extern int (*cl_inet_connect2)(netstackid_t stack_id, uint8_t protocol,
3500     boolean_t is_outgoing, sa_family_t addr_family, uint8_t *laddrp,
3501     in_port_t lport, uint8_t *faddrp, in_port_t fport, void *args);
3502 extern void (*cl_inet_getspi)(netstackid_t, uint8_t, uint8_t *, size_t,
3503     void *);
3504 extern void (*cl_inet_getspi)(netstackid_t stack_id, uint8_t protocol,
3505     uint8_t *ptr, size_t len, void *args);
3506 extern int (*cl_inet_checkspi)(netstackid_t stack_id, uint8_t protocol,
3507     uint32_t spi, void *args);
3508 extern void (*cl_inet_deletespi)(netstackid_t stack_id, uint8_t protocol,
3509     uint32_t spi, void *args);
3510 extern void (*cl_inet_idlesa)(netstackid_t, uint8_t, uint32_t,
3511     sa_family_t, in6_addr_t, in6_addr_t, void *);
3512 
3513 
3514 /* Hooks for CGTP (multirt routes) filtering module */
3515 #define	CGTP_FILTER_REV_1	1
3516 #define	CGTP_FILTER_REV_2	2
3517 #define	CGTP_FILTER_REV_3	3
3518 #define	CGTP_FILTER_REV		CGTP_FILTER_REV_3
3519 
3520 /* cfo_filter and cfo_filter_v6 hooks return values */
3521 #define	CGTP_IP_PKT_NOT_CGTP	0
3522 #define	CGTP_IP_PKT_PREMIUM	1
3523 #define	CGTP_IP_PKT_DUPLICATE	2
3524 
3525 /* Version 3 of the filter interface */
3526 typedef struct cgtp_filter_ops {
3527 	int	cfo_filter_rev;			/* CGTP_FILTER_REV_3 */
3528 	int	(*cfo_change_state)(netstackid_t, int);
3529 	int	(*cfo_add_dest_v4)(netstackid_t, ipaddr_t, ipaddr_t,
3530 		    ipaddr_t, ipaddr_t);
3531 	int	(*cfo_del_dest_v4)(netstackid_t, ipaddr_t, ipaddr_t);
3532 	int	(*cfo_add_dest_v6)(netstackid_t, in6_addr_t *, in6_addr_t *,
3533 		    in6_addr_t *, in6_addr_t *);
3534 	int	(*cfo_del_dest_v6)(netstackid_t, in6_addr_t *, in6_addr_t *);
3535 	int	(*cfo_filter)(netstackid_t, uint_t, mblk_t *);
3536 	int	(*cfo_filter_v6)(netstackid_t, uint_t, ip6_t *,
3537 		    ip6_frag_t *);
3538 } cgtp_filter_ops_t;
3539 
3540 #define	CGTP_MCAST_SUCCESS	1
3541 
3542 /*
3543  * The separate CGTP module needs this global symbol so that it
3544  * can check the version and determine whether to use the old or the new
3545  * version of the filtering interface.
3546  */
3547 extern int	ip_cgtp_filter_rev;
3548 
3549 extern int	ip_cgtp_filter_supported(void);
3550 extern int	ip_cgtp_filter_register(netstackid_t, cgtp_filter_ops_t *);
3551 extern int	ip_cgtp_filter_unregister(netstackid_t);
3552 extern int	ip_cgtp_filter_is_registered(netstackid_t);
3553 
3554 /*
3555  * rr_ring_state cycles in the order shown below from RR_FREE through
3556  * RR_FREE_IN_PROG and  back to RR_FREE.
3557  */
3558 typedef enum {
3559 	RR_FREE,			/* Free slot */
3560 	RR_SQUEUE_UNBOUND,		/* Ring's squeue is unbound */
3561 	RR_SQUEUE_BIND_INPROG,		/* Ring's squeue bind in progress */
3562 	RR_SQUEUE_BOUND,		/* Ring's squeue bound to cpu */
3563 	RR_FREE_INPROG			/* Ring is being freed */
3564 } ip_ring_state_t;
3565 
3566 #define	ILL_MAX_RINGS		256	/* Max num of rx rings we can manage */
3567 #define	ILL_POLLING		0x01	/* Polling in use */
3568 
3569 /*
3570  * These functions pointer types are exported by the mac/dls layer.
3571  * we need to duplicate the definitions here because we cannot
3572  * include mac/dls header files here.
3573  */
3574 typedef boolean_t		(*ip_mac_intr_disable_t)(void *);
3575 typedef int			(*ip_mac_intr_enable_t)(void *);
3576 typedef ip_mac_tx_cookie_t	(*ip_dld_tx_t)(void *, mblk_t *,
3577     uint64_t, uint16_t);
3578 typedef	void			(*ip_flow_enable_t)(void *, ip_mac_tx_cookie_t);
3579 typedef void			*(*ip_dld_callb_t)(void *,
3580     ip_flow_enable_t, void *);
3581 typedef boolean_t		(*ip_dld_fctl_t)(void *, ip_mac_tx_cookie_t);
3582 typedef int			(*ip_capab_func_t)(void *, uint_t,
3583     void *, uint_t);
3584 
3585 /*
3586  * POLLING README
3587  * sq_get_pkts() is called to pick packets from softring in poll mode. It
3588  * calls rr_rx to get the chain and process it with rr_ip_accept.
3589  * rr_rx = mac_soft_ring_poll() to pick packets
3590  * rr_ip_accept = ip_accept_tcp() to process packets
3591  */
3592 
3593 /*
3594  * XXX: With protocol, service specific squeues, they will have
3595  * specific acceptor functions.
3596  */
3597 typedef	mblk_t *(*ip_mac_rx_t)(void *, size_t);
3598 typedef mblk_t *(*ip_accept_t)(ill_t *, ill_rx_ring_t *,
3599     squeue_t *, mblk_t *, mblk_t **, uint_t *);
3600 
3601 /*
3602  * rr_intr_enable, rr_intr_disable, rr_rx_handle, rr_rx:
3603  * May be accessed while in the squeue AND after checking that SQS_POLL_CAPAB
3604  * is set.
3605  *
3606  * rr_ring_state: Protected by ill_lock.
3607  */
3608 struct ill_rx_ring {
3609 	ip_mac_intr_disable_t	rr_intr_disable; /* Interrupt disabling func */
3610 	ip_mac_intr_enable_t	rr_intr_enable;	/* Interrupt enabling func */
3611 	void			*rr_intr_handle; /* Handle interrupt funcs */
3612 	ip_mac_rx_t		rr_rx;		/* Driver receive function */
3613 	ip_accept_t		rr_ip_accept;	/* IP accept function */
3614 	void			*rr_rx_handle;	/* Handle for Rx ring */
3615 	squeue_t		*rr_sqp; /* Squeue the ring is bound to */
3616 	ill_t			*rr_ill;	/* back pointer to ill */
3617 	ip_ring_state_t		rr_ring_state;	/* State of this ring */
3618 };
3619 
3620 /*
3621  * IP - DLD direct function call capability
3622  * Suffixes, df - dld function, dh - dld handle,
3623  * cf - client (IP) function, ch - client handle
3624  */
3625 typedef struct ill_dld_direct_s {		/* DLD provided driver Tx */
3626 	ip_dld_tx_t		idd_tx_df;	/* str_mdata_fastpath_put */
3627 	void			*idd_tx_dh;	/* dld_str_t *dsp */
3628 	ip_dld_callb_t		idd_tx_cb_df;	/* mac_tx_srs_notify */
3629 	void			*idd_tx_cb_dh;	/* mac_client_handle_t *mch */
3630 	ip_dld_fctl_t		idd_tx_fctl_df;	/* mac_tx_is_flow_blocked */
3631 	void			*idd_tx_fctl_dh;	/* mac_client_handle */
3632 } ill_dld_direct_t;
3633 
3634 /* IP - DLD polling capability */
3635 typedef struct ill_dld_poll_s {
3636 	ill_rx_ring_t		idp_ring_tbl[ILL_MAX_RINGS];
3637 } ill_dld_poll_t;
3638 
3639 /* Describes ill->ill_dld_capab */
3640 struct ill_dld_capab_s {
3641 	ip_capab_func_t		idc_capab_df;	/* dld_capab_func */
3642 	void			*idc_capab_dh;	/* dld_str_t *dsp */
3643 	ill_dld_direct_t	idc_direct;
3644 	ill_dld_poll_t		idc_poll;
3645 };
3646 
3647 /*
3648  * IP squeues exports
3649  */
3650 extern boolean_t	ip_squeue_fanout;
3651 
3652 #define	IP_SQUEUE_GET(hint) ip_squeue_random(hint)
3653 
3654 extern void ip_squeue_init(void (*)(squeue_t *));
3655 extern squeue_t	*ip_squeue_random(uint_t);
3656 extern squeue_t *ip_squeue_get(ill_rx_ring_t *);
3657 extern squeue_t *ip_squeue_getfree(pri_t);
3658 extern int ip_squeue_cpu_move(squeue_t *, processorid_t);
3659 extern void *ip_squeue_add_ring(ill_t *, void *);
3660 extern void ip_squeue_bind_ring(ill_t *, ill_rx_ring_t *, processorid_t);
3661 extern void ip_squeue_clean_ring(ill_t *, ill_rx_ring_t *);
3662 extern void ip_squeue_quiesce_ring(ill_t *, ill_rx_ring_t *);
3663 extern void ip_squeue_restart_ring(ill_t *, ill_rx_ring_t *);
3664 extern void ip_squeue_clean_all(ill_t *);
3665 extern boolean_t	ip_source_routed(ipha_t *, ip_stack_t *);
3666 
3667 extern int tcp_wput(queue_t *, mblk_t *);
3668 
3669 extern int	ip_fill_mtuinfo(conn_t *, ip_xmit_attr_t *,
3670     struct ip6_mtuinfo *);
3671 extern hook_t *ipobs_register_hook(netstack_t *, pfv_t);
3672 extern void ipobs_unregister_hook(netstack_t *, hook_t *);
3673 extern void ipobs_hook(mblk_t *, int, zoneid_t, zoneid_t, const ill_t *,
3674     ip_stack_t *);
3675 typedef void    (*ipsq_func_t)(ipsq_t *, queue_t *, mblk_t *, void *);
3676 
3677 extern void	dce_g_init(void);
3678 extern void	dce_g_destroy(void);
3679 extern void	dce_stack_init(ip_stack_t *);
3680 extern void	dce_stack_destroy(ip_stack_t *);
3681 extern void	dce_cleanup(uint_t, ip_stack_t *);
3682 extern dce_t	*dce_get_default(ip_stack_t *);
3683 extern dce_t	*dce_lookup_pkt(mblk_t *, ip_xmit_attr_t *, uint_t *);
3684 extern dce_t	*dce_lookup_v4(ipaddr_t, ip_stack_t *, uint_t *);
3685 extern dce_t	*dce_lookup_v6(const in6_addr_t *, uint_t, ip_stack_t *,
3686     uint_t *);
3687 extern dce_t	*dce_lookup_and_add_v4(ipaddr_t, ip_stack_t *);
3688 extern dce_t	*dce_lookup_and_add_v6(const in6_addr_t *, uint_t,
3689     ip_stack_t *);
3690 extern int	dce_update_uinfo_v4(ipaddr_t, iulp_t *, ip_stack_t *);
3691 extern int	dce_update_uinfo_v6(const in6_addr_t *, uint_t, iulp_t *,
3692     ip_stack_t *);
3693 extern int	dce_update_uinfo(const in6_addr_t *, uint_t, iulp_t *,
3694     ip_stack_t *);
3695 extern void	dce_increment_generation(dce_t *);
3696 extern void	dce_increment_all_generations(boolean_t, ip_stack_t *);
3697 extern void	dce_refrele(dce_t *);
3698 extern void	dce_refhold(dce_t *);
3699 extern void	dce_refrele_notr(dce_t *);
3700 extern void	dce_refhold_notr(dce_t *);
3701 mblk_t		*ip_snmp_get_mib2_ip_dce(queue_t *, mblk_t *, ip_stack_t *ipst);
3702 
3703 extern ip_laddr_t ip_laddr_verify_v4(ipaddr_t, zoneid_t,
3704     ip_stack_t *, boolean_t);
3705 extern ip_laddr_t ip_laddr_verify_v6(const in6_addr_t *, zoneid_t,
3706     ip_stack_t *, boolean_t, uint_t);
3707 extern int	ip_laddr_fanout_insert(conn_t *);
3708 
3709 extern boolean_t ip_verify_src(mblk_t *, ip_xmit_attr_t *, uint_t *);
3710 extern int	ip_verify_ire(mblk_t *, ip_xmit_attr_t *);
3711 
3712 extern mblk_t	*ip_xmit_attr_to_mblk(ip_xmit_attr_t *);
3713 extern boolean_t ip_xmit_attr_from_mblk(mblk_t *, ip_xmit_attr_t *);
3714 extern mblk_t	*ip_xmit_attr_free_mblk(mblk_t *);
3715 extern mblk_t	*ip_recv_attr_to_mblk(ip_recv_attr_t *);
3716 extern boolean_t ip_recv_attr_from_mblk(mblk_t *, ip_recv_attr_t *);
3717 extern mblk_t	*ip_recv_attr_free_mblk(mblk_t *);
3718 extern boolean_t ip_recv_attr_is_mblk(mblk_t *);
3719 
3720 #ifdef	__PRAGMA_REDEFINE_EXTNAME
3721 #pragma	redefine_extname inet_ntop _inet_ntop
3722 #pragma	redefine_extname inet_pton _inet_pton
3723 #else
3724 #define	inet_ntop _inet_ntop
3725 #define	inet_pton _inet_pton
3726 #endif	/* __PRAGMA_REDEFINE_EXTNAME */
3727 
3728 extern char	*inet_ntop(int, const void *, char *, int);
3729 extern int	inet_pton(int, char *, void *);
3730 
3731 /*
3732  * Squeue tags. Tags only need to be unique when the callback function is the
3733  * same to distinguish between different calls, but we use unique tags for
3734  * convenience anyway.
3735  */
3736 #define	SQTAG_IP_INPUT			1
3737 #define	SQTAG_TCP_INPUT_ICMP_ERR	2
3738 #define	SQTAG_TCP6_INPUT_ICMP_ERR	3
3739 #define	SQTAG_IP_TCP_INPUT		4
3740 #define	SQTAG_IP6_TCP_INPUT		5
3741 #define	SQTAG_IP_TCP_CLOSE		6
3742 #define	SQTAG_TCP_OUTPUT		7
3743 #define	SQTAG_TCP_TIMER			8
3744 #define	SQTAG_TCP_TIMEWAIT		9
3745 #define	SQTAG_TCP_ACCEPT_FINISH		10
3746 #define	SQTAG_TCP_ACCEPT_FINISH_Q0	11
3747 #define	SQTAG_TCP_ACCEPT_PENDING	12
3748 #define	SQTAG_TCP_LISTEN_DISCON		13
3749 #define	SQTAG_TCP_CONN_REQ_1		14
3750 #define	SQTAG_TCP_EAGER_BLOWOFF		15
3751 #define	SQTAG_TCP_EAGER_CLEANUP		16
3752 #define	SQTAG_TCP_EAGER_CLEANUP_Q0	17
3753 #define	SQTAG_TCP_CONN_IND		18
3754 #define	SQTAG_TCP_RSRV			19
3755 #define	SQTAG_TCP_ABORT_BUCKET		20
3756 #define	SQTAG_TCP_REINPUT		21
3757 #define	SQTAG_TCP_REINPUT_EAGER		22
3758 #define	SQTAG_TCP_INPUT_MCTL		23
3759 #define	SQTAG_TCP_RPUTOTHER		24
3760 #define	SQTAG_IP_PROTO_AGAIN		25
3761 #define	SQTAG_IP_FANOUT_TCP		26
3762 #define	SQTAG_IPSQ_CLEAN_RING		27
3763 #define	SQTAG_TCP_WPUT_OTHER		28
3764 #define	SQTAG_TCP_CONN_REQ_UNBOUND	29
3765 #define	SQTAG_TCP_SEND_PENDING		30
3766 #define	SQTAG_BIND_RETRY		31
3767 #define	SQTAG_UDP_FANOUT		32
3768 #define	SQTAG_UDP_INPUT			33
3769 #define	SQTAG_UDP_WPUT			34
3770 #define	SQTAG_UDP_OUTPUT		35
3771 #define	SQTAG_TCP_KSSL_INPUT		36
3772 #define	SQTAG_TCP_DROP_Q0		37
3773 #define	SQTAG_TCP_CONN_REQ_2		38
3774 #define	SQTAG_IP_INPUT_RX_RING		39
3775 #define	SQTAG_SQUEUE_CHANGE		40
3776 #define	SQTAG_CONNECT_FINISH		41
3777 #define	SQTAG_SYNCHRONOUS_OP		42
3778 #define	SQTAG_TCP_SHUTDOWN_OUTPUT	43
3779 #define	SQTAG_TCP_IXA_CLEANUP		44
3780 #define	SQTAG_TCP_SEND_SYNACK		45
3781 
3782 extern sin_t	sin_null;	/* Zero address for quick clears */
3783 extern sin6_t	sin6_null;	/* Zero address for quick clears */
3784 
3785 #endif	/* _KERNEL */
3786 
3787 #ifdef	__cplusplus
3788 }
3789 #endif
3790 
3791 #endif	/* _INET_IP_H */
3792