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
34extern "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
92typedef 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 */
143typedef 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. */
150typedef 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 */
165typedef 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
294extern int ip_squeue_flag;
295
296/* IP Fragmentation Reassembly Header */
297typedef 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 */
376typedef 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 */
415typedef 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
456struct ill_s;
457
458typedef	void ip_v6intfid_func_t(struct ill_s *, in6_addr_t *);
459typedef void ip_v6mapinfo_func_t(struct ill_s *, uchar_t *, uchar_t *);
460typedef void ip_v4mapinfo_func_t(struct ill_s *, uchar_t *, uchar_t *);
461
462/* IP Mac info structure */
463typedef 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 */
484extern	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
562typedef 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 */
573typedef 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 */
585typedef 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 */
618typedef 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 */
651typedef 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 */
685typedef 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 */
709typedef 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 */
737typedef struct conn_s conn_t;
738
739/*
740 * This is used to match an inbound/outbound datagram with policy.
741 */
742typedef	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 */
808typedef 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
848typedef uintptr_t ip_mac_tx_cookie_t;
849typedef	struct idl_s idl_t;
850typedef	struct idl_tx_list_s idl_tx_list_t;
851
852struct 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
859struct 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 */
875typedef 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
904typedef 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
910typedef 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
918typedef 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 */
931typedef 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 */
1037typedef 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 */
1141typedef 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
1163typedef 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 */
1179enum {
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 */
1195enum {
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 */
1209typedef 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)
1226struct _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
1231typedef 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 */
1242typedef 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 */
1265typedef 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 */
1285typedef 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 */
1294struct dce_s;
1295typedef struct dce_s dce_t;
1296struct ire_s;
1297typedef struct ire_s ire_t;
1298struct ncec_s;
1299typedef struct ncec_s ncec_t;
1300struct nce_s;
1301typedef struct nce_s nce_t;
1302struct ip_recv_attr_s;
1303typedef struct ip_recv_attr_s ip_recv_attr_t;
1304struct ip_xmit_attr_s;
1305typedef struct ip_xmit_attr_s ip_xmit_attr_t;
1306
1307struct tsol_ire_gw_secattr_s;
1308typedef 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 */
1314typedef 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 */
1324typedef uint64_t iaflags_t;
1325
1326/* The ill input function pointer type */
1327typedef void (*pfillinput_t)(mblk_t *, void *, void *, ip_recv_attr_t *,
1328    rtc_t *);
1329
1330/* The ire receive function pointer type */
1331typedef void (*pfirerecv_t)(ire_t *, mblk_t *, void *, ip_recv_attr_t *);
1332
1333/* The ire send and postfrag function pointer types */
1334typedef int (*pfiresend_t)(ire_t *, mblk_t *, void *,
1335    ip_xmit_attr_t *, uint32_t *);
1336typedef 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 */
1348struct	_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 */
1360typedef 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
1374typedef 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 */
1382struct _ill_g_head_s_ {
1383	ill_if_t	*ill_g_list_head;
1384	ill_if_t	*ill_g_list_tail;
1385};
1386
1387typedef 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 */
1422typedef struct ill_hcksum_capab_s ill_hcksum_capab_t;
1423
1424/*
1425 * Per-ill Zero-copy capabilities.
1426 */
1427typedef struct ill_zerocopy_capab_s ill_zerocopy_capab_t;
1428
1429/*
1430 * DLD capbilities.
1431 */
1432typedef struct ill_dld_capab_s ill_dld_capab_t;
1433
1434/*
1435 * Per-ill polling resource map.
1436 */
1437typedef struct ill_rx_ring ill_rx_ring_t;
1438
1439/*
1440 * Per-ill Large Send Offload capabilities.
1441 */
1442typedef 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 */
1514typedef 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 */
1549typedef 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 */
1571typedef 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
1581struct arl_s;
1582
1583/*
1584 * Per-ill capabilities.
1585 */
1586struct ill_hcksum_capab_s {
1587	uint_t	ill_hcksum_version;	/* interface version */
1588	uint_t	ill_hcksum_txflags;	/* capabilities on transmit */
1589};
1590
1591struct ill_zerocopy_capab_s {
1592	uint_t	ill_zerocopy_version;	/* interface version */
1593	uint_t	ill_zerocopy_flags;	/* capabilities */
1594};
1595
1596struct ill_lso_capab_s {
1597	uint_t	ill_lso_flags;		/* capabilities */
1598	uint_t	ill_lso_max;		/* maximum size of payload */
1599};
1600
1601/*
1602 * IP Lower level Structure.
1603 * Instance data structure in ip_open when there is a device below us.
1604 */
1605typedef struct ill_s {
1606	pfillinput_t ill_inputfn;	/* Fast input function selector */
1607	ill_if_t *ill_ifptr;		/* pointer to interface type */
1608	queue_t	*ill_rq;		/* Read queue. */
1609	queue_t	*ill_wq;		/* Write queue. */
1610
1611	int	ill_error;		/* Error value sent up by device. */
1612
1613	ipif_t	*ill_ipif;		/* Interface chain for this ILL. */
1614
1615	uint_t	ill_ipif_up_count;	/* Number of IPIFs currently up. */
1616	uint_t	ill_max_frag;		/* Max IDU from DLPI. */
1617	uint_t	ill_current_frag;	/* Current IDU from DLPI. */
1618	uint_t	ill_mtu;		/* User-specified MTU; SIOCSLIFMTU */
1619	uint_t	ill_mc_mtu;		/* MTU for multi/broadcast */
1620	uint_t	ill_metric;		/* BSD if metric, for compatibility. */
1621	char	*ill_name;		/* Our name. */
1622	uint_t	ill_ipif_dup_count;	/* Number of duplicate addresses. */
1623	uint_t	ill_name_length;	/* Name length, incl. terminator. */
1624	uint_t	ill_net_type;		/* IRE_IF_RESOLVER/IRE_IF_NORESOLVER. */
1625	/*
1626	 * Physical Point of Attachment num.  If DLPI style 1 provider
1627	 * then this is derived from the devname.
1628	 */
1629	uint_t	ill_ppa;
1630	t_uscalar_t	ill_sap;
1631	t_scalar_t	ill_sap_length;	/* Including sign (for position) */
1632	uint_t	ill_phys_addr_length;	/* Excluding the sap. */
1633	uint_t	ill_bcast_addr_length;	/* Only set when the DL provider */
1634					/* supports broadcast. */
1635	t_uscalar_t	ill_mactype;
1636	uint8_t	*ill_frag_ptr;		/* Reassembly state. */
1637	timeout_id_t ill_frag_timer_id; /* timeout id for the frag timer */
1638	ipfb_t	*ill_frag_hash_tbl;	/* Fragment hash list head. */
1639
1640	krwlock_t ill_mcast_lock;	/* Protects multicast state */
1641	kmutex_t ill_mcast_serializer;	/* Serialize across ilg and ilm state */
1642	ilm_t	*ill_ilm;		/* Multicast membership for ill */
1643	uint_t	ill_global_timer;	/* for IGMPv3/MLDv2 general queries */
1644	int	ill_mcast_type;		/* type of router which is querier */
1645					/* on this interface */
1646	uint16_t ill_mcast_v1_time;	/* # slow timeouts since last v1 qry */
1647	uint16_t ill_mcast_v2_time;	/* # slow timeouts since last v2 qry */
1648	uint8_t	ill_mcast_v1_tset;	/* 1 => timer is set; 0 => not set */
1649	uint8_t	ill_mcast_v2_tset;	/* 1 => timer is set; 0 => not set */
1650
1651	uint8_t	ill_mcast_rv;		/* IGMPv3/MLDv2 robustness variable */
1652	int	ill_mcast_qi;		/* IGMPv3/MLDv2 query interval var */
1653
1654	/*
1655	 * All non-NULL cells between 'ill_first_mp_to_free' and
1656	 * 'ill_last_mp_to_free' are freed in ill_delete.
1657	 */
1658#define	ill_first_mp_to_free	ill_bcast_mp
1659	mblk_t	*ill_bcast_mp;		/* DLPI header for broadcasts. */
1660	mblk_t	*ill_unbind_mp;		/* unbind mp from ill_dl_up() */
1661	mblk_t	*ill_promiscoff_mp;	/* for ill_leave_allmulti() */
1662	mblk_t	*ill_dlpi_deferred;	/* b_next chain of control messages */
1663	mblk_t	*ill_dest_addr_mp;	/* mblk which holds ill_dest_addr */
1664	mblk_t	*ill_replumb_mp;	/* replumb mp from ill_replumb() */
1665	mblk_t	*ill_phys_addr_mp;	/* mblk which holds ill_phys_addr */
1666	mblk_t	*ill_mcast_deferred;	/* b_next chain of IGMP/MLD packets */
1667#define	ill_last_mp_to_free	ill_mcast_deferred
1668
1669	cred_t	*ill_credp;		/* opener's credentials */
1670	uint8_t	*ill_phys_addr;		/* ill_phys_addr_mp->b_rptr + off */
1671	uint8_t *ill_dest_addr;		/* ill_dest_addr_mp->b_rptr + off */
1672
1673	uint_t	ill_state_flags;	/* see ILL_* flags above */
1674
1675	/* Following bit fields protected by ipsq_t */
1676	uint_t
1677		ill_needs_attach : 1,
1678		ill_reserved : 1,
1679		ill_isv6 : 1,
1680		ill_dlpi_style_set : 1,
1681
1682		ill_ifname_pending : 1,
1683		ill_logical_down : 1,
1684		ill_dl_up : 1,
1685		ill_up_ipifs : 1,
1686
1687		ill_note_link : 1,	/* supports link-up notification */
1688		ill_capab_reneg : 1, /* capability renegotiation to be done */
1689		ill_dld_capab_inprog : 1, /* direct dld capab call in prog */
1690		ill_need_recover_multicast : 1,
1691
1692		ill_replumbing : 1,
1693		ill_arl_dlpi_pending : 1,
1694		ill_grp_pending : 1,
1695
1696		ill_pad_to_bit_31 : 17;
1697
1698	/* Following bit fields protected by ill_lock */
1699	uint_t
1700		ill_fragtimer_executing : 1,
1701		ill_fragtimer_needrestart : 1,
1702		ill_manual_token : 1,	/* system won't override ill_token */
1703		/*
1704		 * ill_manual_linklocal : system will not change the
1705		 * linklocal whenever ill_token changes.
1706		 */
1707		ill_manual_linklocal : 1,
1708
1709		ill_manual_dst_linklocal : 1, /* same for pt-pt dst linklocal */
1710
1711		ill_mcast_ncec_cleanup : 1,	/* Reaping mcast ncecs. */
1712		ill_pad_bit_31 : 26;
1713
1714	/*
1715	 * Used in SIOCSIFMUXID and SIOCGIFMUXID for 'ifconfig unplumb'.
1716	 */
1717	int	ill_muxid;		/* muxid returned from plink */
1718
1719	/* Used for IP frag reassembly throttling on a per ILL basis.  */
1720	uint_t	ill_ipf_gen;		/* Generation of next fragment queue */
1721	uint_t	ill_frag_count;		/* Count of all reassembly mblk bytes */
1722	uint_t	ill_frag_free_num_pkts;	 /* num of fragmented packets to free */
1723	clock_t	ill_last_frag_clean_time; /* time when frag's were pruned */
1724	int	ill_type;		/* From <net/if_types.h> */
1725	uint_t	ill_dlpi_multicast_state;	/* See below IDS_* */
1726	uint_t	ill_dlpi_fastpath_state;	/* See below IDS_* */
1727
1728	/*
1729	 * Capabilities related fields.
1730	 */
1731	uint_t  ill_dlpi_capab_state;	/* State of capability query, IDCS_* */
1732	uint_t	ill_capab_pending_cnt;
1733	uint64_t ill_capabilities;	/* Enabled capabilities, ILL_CAPAB_* */
1734	ill_hcksum_capab_t *ill_hcksum_capab; /* H/W cksumming capabilities */
1735	ill_zerocopy_capab_t *ill_zerocopy_capab; /* Zero-copy capabilities */
1736	ill_dld_capab_t *ill_dld_capab; /* DLD capabilities */
1737	ill_lso_capab_t	*ill_lso_capab;	/* Large Segment Offload capabilities */
1738	mblk_t	*ill_capab_reset_mp;	/* Preallocated mblk for capab reset */
1739
1740	uint8_t	ill_max_hops;	/* Maximum hops for any logical interface */
1741	uint_t	ill_user_mtu;	/* User-specified MTU via SIOCSLIFLNKINFO */
1742	uint32_t ill_reachable_time;	/* Value for ND algorithm in msec */
1743	uint32_t ill_reachable_retrans_time; /* Value for ND algorithm msec */
1744	uint_t	ill_max_buf;		/* Max # of req to buffer for ND */
1745	in6_addr_t	ill_token;	/* IPv6 interface id */
1746	in6_addr_t	ill_dest_token;	/* Destination IPv6 interface id */
1747	uint_t		ill_token_length;
1748	uint32_t	ill_xmit_count;		/* ndp max multicast xmits */
1749	mib2_ipIfStatsEntry_t	*ill_ip_mib;	/* ver indep. interface mib */
1750	mib2_ipv6IfIcmpEntry_t	*ill_icmp6_mib;	/* Per interface mib */
1751
1752	phyint_t		*ill_phyint;
1753	uint64_t		ill_flags;
1754
1755	kmutex_t	ill_lock;	/* Please see table below */
1756	/*
1757	 * The ill_nd_lla* fields handle the link layer address option
1758	 * from neighbor discovery. This is used for external IPv6
1759	 * address resolution.
1760	 */
1761	mblk_t		*ill_nd_lla_mp;	/* mblk which holds ill_nd_lla */
1762	uint8_t		*ill_nd_lla;	/* Link Layer Address */
1763	uint_t		ill_nd_lla_len;	/* Link Layer Address length */
1764	/*
1765	 * We have 4 phys_addr_req's sent down. This field keeps track
1766	 * of which one is pending.
1767	 */
1768	t_uscalar_t	ill_phys_addr_pend; /* which dl_phys_addr_req pending */
1769	/*
1770	 * Used to save errors that occur during plumbing
1771	 */
1772	uint_t		ill_ifname_pending_err;
1773	avl_node_t	ill_avl_byppa; /* avl node based on ppa */
1774	uint_t		ill_mcast_nces;	/* Number of NCEs that are multicast. */
1775	list_t		ill_nce; /* pointer to nce_s list */
1776	uint_t		ill_refcnt;	/* active refcnt by threads */
1777	uint_t		ill_ire_cnt;	/* ires associated with this ill */
1778	kcondvar_t	ill_cv;
1779	uint_t		ill_ncec_cnt;	/* ncecs associated with this ill */
1780	uint_t		ill_nce_cnt;	/* nces associated with this ill */
1781	uint_t		ill_waiters;	/* threads waiting in ipsq_enter */
1782	/*
1783	 * Contains the upper read queue pointer of the module immediately
1784	 * beneath IP.  This field allows IP to validate sub-capability
1785	 * acknowledgments coming up from downstream.
1786	 */
1787	queue_t		*ill_lmod_rq;	/* read queue pointer of module below */
1788	uint_t		ill_lmod_cnt;	/* number of modules beneath IP */
1789	ip_m_t		*ill_media;	/* media specific params/functions */
1790	t_uscalar_t	ill_dlpi_pending; /* Last DLPI primitive issued */
1791	uint_t		ill_usesrc_ifindex; /* use src addr from this ILL */
1792	struct ill_s	*ill_usesrc_grp_next; /* Next ILL in the usesrc group */
1793	boolean_t	ill_trace_disable;	/* True when alloc fails */
1794	zoneid_t	ill_zoneid;
1795	ip_stack_t	*ill_ipst;	/* Corresponds to a netstack_hold */
1796	uint32_t	ill_dhcpinit;	/* IP_DHCPINIT_IFs for ill */
1797	void		*ill_flownotify_mh; /* Tx flow ctl, mac cb handle */
1798	uint_t		ill_ilm_cnt;    /* ilms referencing this ill */
1799	uint_t		ill_ipallmulti_cnt; /* ip_join_allmulti() calls */
1800	ilm_t		*ill_ipallmulti_ilm;
1801
1802	mblk_t		*ill_saved_ire_mp; /* Allocated for each extra IRE */
1803					/* with ire_ill set so they can */
1804					/* survive the ill going down and up. */
1805	kmutex_t	ill_saved_ire_lock; /* Protects ill_saved_ire_mp, cnt */
1806	uint_t		ill_saved_ire_cnt;	/* # entries */
1807	struct arl_ill_common_s    *ill_common;
1808	ire_t		*ill_ire_multicast; /* IRE_MULTICAST for ill */
1809	clock_t		ill_defend_start;   /* start of 1 hour period */
1810	uint_t		ill_defend_count;   /* # of announce/defends per ill */
1811	/*
1812	 * IPMP fields.
1813	 */
1814	ipmp_illgrp_t	*ill_grp;	/* IPMP group information */
1815	list_node_t	ill_actnode;	/* next active ill in group */
1816	list_node_t	ill_grpnode;	/* next ill in group */
1817	ipif_t		*ill_src_ipif;	/* source address selection rotor */
1818	ipif_t		*ill_move_ipif;	/* ipif awaiting move to new ill */
1819	boolean_t	ill_nom_cast;	/* nominated for mcast/bcast */
1820	uint_t		ill_bound_cnt;	/* # of data addresses bound to ill */
1821	ipif_t		*ill_bound_ipif; /* ipif chain bound to ill */
1822	timeout_id_t	ill_refresh_tid; /* ill refresh retry timeout id */
1823
1824	uint32_t	ill_mrouter_cnt; /* mrouter allmulti joins */
1825	uint32_t	ill_allowed_ips_cnt;
1826	in6_addr_t	*ill_allowed_ips;
1827
1828	/* list of multicast physical addresses joined on this ill */
1829	multiphysaddr_t *ill_mphysaddr_list;
1830} ill_t;
1831
1832/*
1833 * ILL_FREE_OK() means that there are no incoming pointer references
1834 * to the ill.
1835 */
1836#define	ILL_FREE_OK(ill)					\
1837	((ill)->ill_ire_cnt == 0 && (ill)->ill_ilm_cnt == 0 &&	\
1838	(ill)->ill_ncec_cnt == 0 && (ill)->ill_nce_cnt == 0)
1839
1840/*
1841 * An ipif/ill can be marked down only when the ire and ncec references
1842 * to that ipif/ill goes to zero. ILL_DOWN_OK() is a necessary condition
1843 * quiescence checks. See comments above IPIF_DOWN_OK for details
1844 * on why ires and nces are selectively considered for this macro.
1845 */
1846#define	ILL_DOWN_OK(ill)					\
1847	(ill->ill_ire_cnt == 0 && ill->ill_ncec_cnt == 0 &&	\
1848	ill->ill_nce_cnt == 0)
1849
1850/*
1851 * The following table lists the protection levels of the various members
1852 * of the ill_t. Same notation as that used for ipif_t above is used.
1853 *
1854 *				Write			Read
1855 *
1856 * ill_ifptr			ill_g_lock + s		Write once
1857 * ill_rq			ipsq			Write once
1858 * ill_wq			ipsq			Write once
1859 *
1860 * ill_error			ipsq			None
1861 * ill_ipif			ill_g_lock + ipsq	ill_g_lock OR ipsq
1862 * ill_ipif_up_count		ill_lock + ipsq		ill_lock OR ipsq
1863 * ill_max_frag			ill_lock		ill_lock
1864 * ill_current_frag		ill_lock		ill_lock
1865 *
1866 * ill_name			ill_g_lock + ipsq	Write once
1867 * ill_name_length		ill_g_lock + ipsq	Write once
1868 * ill_ndd_name			ipsq			Write once
1869 * ill_net_type			ipsq			Write once
1870 * ill_ppa			ill_g_lock + ipsq	Write once
1871 * ill_sap			ipsq + down ill		Write once
1872 * ill_sap_length		ipsq + down ill		Write once
1873 * ill_phys_addr_length		ipsq + down ill		Write once
1874 *
1875 * ill_bcast_addr_length	ipsq			ipsq
1876 * ill_mactype			ipsq			ipsq
1877 * ill_frag_ptr			ipsq			ipsq
1878 *
1879 * ill_frag_timer_id		ill_lock		ill_lock
1880 * ill_frag_hash_tbl		ipsq			up ill
1881 * ill_ilm			ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1882 * ill_global_timer		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1883 * ill_mcast_type		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1884 * ill_mcast_v1_time		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1885 * ill_mcast_v2_time		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1886 * ill_mcast_v1_tset		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1887 * ill_mcast_v2_tset		ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1888 * ill_mcast_rv			ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1889 * ill_mcast_qi			ill_mcast_lock(WRITER)	ill_mcast_lock(READER)
1890 *
1891 * ill_down_mp			ipsq			ipsq
1892 * ill_dlpi_deferred		ill_lock		ill_lock
1893 * ill_dlpi_pending		ipsq + ill_lock		ipsq or ill_lock or
1894 *							absence of ipsq writer.
1895 * ill_phys_addr_mp		ipsq + down ill		only when ill is up
1896 * ill_mcast_deferred		ill_lock		ill_lock
1897 * ill_phys_addr		ipsq + down ill		only when ill is up
1898 * ill_dest_addr_mp		ipsq + down ill		only when ill is up
1899 * ill_dest_addr		ipsq + down ill		only when ill is up
1900 *
1901 * ill_state_flags		ill_lock		ill_lock
1902 * exclusive bit flags		ipsq_t			ipsq_t
1903 * shared bit flags		ill_lock		ill_lock
1904 *
1905 * ill_muxid			ipsq			Not atomic
1906 *
1907 * ill_ipf_gen			Not atomic
1908 * ill_frag_count		atomics			atomics
1909 * ill_type			ipsq + down ill		only when ill is up
1910 * ill_dlpi_multicast_state	ill_lock		ill_lock
1911 * ill_dlpi_fastpath_state	ill_lock		ill_lock
1912 * ill_dlpi_capab_state		ipsq			ipsq
1913 * ill_max_hops			ipsq			Not atomic
1914 *
1915 * ill_mtu			ill_lock		None
1916 * ill_mc_mtu			ill_lock		None
1917 *
1918 * ill_user_mtu			ipsq + ill_lock		ill_lock
1919 * ill_reachable_time		ipsq + ill_lock		ill_lock
1920 * ill_reachable_retrans_time	ipsq + ill_lock		ill_lock
1921 * ill_max_buf			ipsq + ill_lock		ill_lock
1922 *
1923 * Next 2 fields need ill_lock because of the get ioctls. They should not
1924 * report partially updated results without executing in the ipsq.
1925 * ill_token			ipsq + ill_lock		ill_lock
1926 * ill_token_length		ipsq + ill_lock		ill_lock
1927 * ill_dest_token		ipsq + down ill		only when ill is up
1928 * ill_xmit_count		ipsq + down ill		write once
1929 * ill_ip6_mib			ipsq + down ill		only when ill is up
1930 * ill_icmp6_mib		ipsq + down ill		only when ill is up
1931 *
1932 * ill_phyint			ipsq, ill_g_lock, ill_lock	Any of them
1933 * ill_flags			ill_lock		ill_lock
1934 * ill_nd_lla_mp		ipsq + down ill		only when ill is up
1935 * ill_nd_lla			ipsq + down ill		only when ill is up
1936 * ill_nd_lla_len		ipsq + down ill		only when ill is up
1937 * ill_phys_addr_pend		ipsq + down ill		only when ill is up
1938 * ill_ifname_pending_err	ipsq			ipsq
1939 * ill_avl_byppa		ipsq, ill_g_lock	write once
1940 *
1941 * ill_fastpath_list		ill_lock		ill_lock
1942 * ill_refcnt			ill_lock		ill_lock
1943 * ill_ire_cnt			ill_lock		ill_lock
1944 * ill_cv			ill_lock		ill_lock
1945 * ill_mcast_nces		ill_lock		ill_lock
1946 * ill_ncec_cnt			ill_lock		ill_lock
1947 * ill_nce_cnt			ill_lock		ill_lock
1948 * ill_ilm_cnt			ill_lock		ill_lock
1949 * ill_src_ipif			ill_g_lock		ill_g_lock
1950 * ill_trace			ill_lock		ill_lock
1951 * ill_usesrc_grp_next		ill_g_usesrc_lock	ill_g_usesrc_lock
1952 * ill_dhcpinit			atomics			atomics
1953 * ill_flownotify_mh		write once		write once
1954 * ill_capab_pending_cnt	ipsq			ipsq
1955 * ill_ipallmulti_cnt		ill_lock		ill_lock
1956 * ill_ipallmulti_ilm		ill_lock		ill_lock
1957 * ill_saved_ire_mp		ill_saved_ire_lock	ill_saved_ire_lock
1958 * ill_saved_ire_cnt		ill_saved_ire_lock	ill_saved_ire_lock
1959 * ill_arl			???			???
1960 * ill_ire_multicast		ipsq + quiescent	none
1961 * ill_bound_ipif		ipsq			ipsq
1962 * ill_actnode			ipsq + ipmp_lock	ipsq OR ipmp_lock
1963 * ill_grpnode			ipsq + ill_g_lock	ipsq OR ill_g_lock
1964 * ill_src_ipif			ill_g_lock		ill_g_lock
1965 * ill_move_ipif		ipsq			ipsq
1966 * ill_nom_cast			ipsq			ipsq OR advisory
1967 * ill_refresh_tid		ill_lock		ill_lock
1968 * ill_grp (for IPMP ill)	write once		write once
1969 * ill_grp (for underlying ill)	ipsq + ill_g_lock	ipsq OR ill_g_lock
1970 * ill_grp_pending		ill_mcast_serializer	ill_mcast_serializer
1971 * ill_mrouter_cnt		atomics			atomics
1972 * ill_mphysaddr_list	ill_lock		ill_lock
1973 *
1974 * NOTE: It's OK to make heuristic decisions on an underlying interface
1975 *	 by using IS_UNDER_IPMP() or comparing ill_grp's raw pointer value.
1976 */
1977
1978/*
1979 * For ioctl restart mechanism see ip_reprocess_ioctl()
1980 */
1981struct ip_ioctl_cmd_s;
1982
1983typedef	int (*ifunc_t)(ipif_t *, struct sockaddr_in *, queue_t *, mblk_t *,
1984    struct ip_ioctl_cmd_s *, void *);
1985
1986typedef struct ip_ioctl_cmd_s {
1987	int	ipi_cmd;
1988	size_t	ipi_copyin_size;
1989	uint_t	ipi_flags;
1990	uint_t	ipi_cmd_type;
1991	ifunc_t	ipi_func;
1992	ifunc_t	ipi_func_restart;
1993} ip_ioctl_cmd_t;
1994
1995/*
1996 * ipi_cmd_type:
1997 *
1998 * IF_CMD		1	old style ifreq cmd
1999 * LIF_CMD		2	new style lifreq cmd
2000 * ARP_CMD		3	arpreq cmd
2001 * XARP_CMD		4	xarpreq cmd
2002 * MSFILT_CMD		5	multicast source filter cmd
2003 * MISC_CMD		6	misc cmd (not a more specific one above)
2004 */
2005
2006enum { IF_CMD = 1, LIF_CMD, ARP_CMD, XARP_CMD, MSFILT_CMD, MISC_CMD };
2007
2008#define	IPI_DONTCARE	0	/* For ioctl encoded values that don't matter */
2009
2010/* Flag values in ipi_flags */
2011#define	IPI_PRIV	0x1	/* Root only command */
2012#define	IPI_MODOK	0x2	/* Permitted on mod instance of IP */
2013#define	IPI_WR		0x4	/* Need to grab writer access */
2014#define	IPI_GET_CMD	0x8	/* branch to mi_copyout on success */
2015/*	unused		0x10	*/
2016#define	IPI_NULL_BCONT	0x20	/* ioctl has not data and hence no b_cont */
2017
2018extern ip_ioctl_cmd_t	ip_ndx_ioctl_table[];
2019extern ip_ioctl_cmd_t	ip_misc_ioctl_table[];
2020extern int ip_ndx_ioctl_count;
2021extern int ip_misc_ioctl_count;
2022
2023/* Passed down by ARP to IP during I_PLINK/I_PUNLINK */
2024typedef struct ipmx_s {
2025	char	ipmx_name[LIFNAMSIZ];		/* if name */
2026	uint_t
2027		ipmx_arpdev_stream : 1,		/* This is the arp stream */
2028		ipmx_notused : 31;
2029} ipmx_t;
2030
2031/*
2032 * State for detecting if a driver supports certain features.
2033 * Support for DL_ENABMULTI_REQ uses ill_dlpi_multicast_state.
2034 * Support for DLPI M_DATA fastpath uses ill_dlpi_fastpath_state.
2035 */
2036#define	IDS_UNKNOWN	0	/* No DLPI request sent */
2037#define	IDS_INPROGRESS	1	/* DLPI request sent */
2038#define	IDS_OK		2	/* DLPI request completed successfully */
2039#define	IDS_FAILED	3	/* DLPI request failed */
2040
2041/* Support for DL_CAPABILITY_REQ uses ill_dlpi_capab_state. */
2042enum {
2043	IDCS_UNKNOWN,
2044	IDCS_PROBE_SENT,
2045	IDCS_OK,
2046	IDCS_RESET_SENT,
2047	IDCS_RENEG,
2048	IDCS_FAILED
2049};
2050
2051/* Extended NDP Management Structure */
2052typedef struct ipndp_s {
2053	ndgetf_t	ip_ndp_getf;
2054	ndsetf_t	ip_ndp_setf;
2055	caddr_t		ip_ndp_data;
2056	char		*ip_ndp_name;
2057} ipndp_t;
2058
2059/* IXA Notification types */
2060typedef enum {
2061	IXAN_LSO,	/* LSO capability change */
2062	IXAN_PMTU,	/* PMTU change */
2063	IXAN_ZCOPY	/* ZEROCOPY capability change */
2064} ixa_notify_type_t;
2065
2066typedef uint_t ixa_notify_arg_t;
2067
2068typedef	void	(*ixa_notify_t)(void *, ip_xmit_attr_t *ixa, ixa_notify_type_t,
2069    ixa_notify_arg_t);
2070
2071/*
2072 * Attribute flags that are common to the transmit and receive attributes
2073 */
2074#define	IAF_IS_IPV4		0x80000000	/* ipsec_*_v4 */
2075#define	IAF_TRUSTED_ICMP	0x40000000	/* ipsec_*_icmp_loopback */
2076#define	IAF_NO_LOOP_ZONEID_SET	0x20000000	/* Zone that shouldn't have */
2077						/* a copy */
2078#define	IAF_LOOPBACK_COPY	0x10000000	/* For multi and broadcast */
2079
2080#define	IAF_MASK		0xf0000000	/* Flags that are common */
2081
2082/*
2083 * Transmit side attributes used between the transport protocols and IP as
2084 * well as inside IP. It is also used to cache information in the conn_t i.e.
2085 * replaces conn_ire and the IPsec caching in the conn_t.
2086 */
2087struct ip_xmit_attr_s {
2088	iaflags_t	ixa_flags;	/* IXAF_*. See below */
2089
2090	uint32_t	ixa_free_flags;	/* IXA_FREE_*. See below */
2091	uint32_t	ixa_refcnt;	/* Using atomics */
2092
2093	/*
2094	 * Always initialized independently of ixa_flags settings.
2095	 * Used by ip_xmit so we keep them up front for cache locality.
2096	 */
2097	uint32_t	ixa_xmit_hint;	/* For ECMP and GLD TX ring fanout */
2098	uint_t		ixa_pktlen;	/* Always set. For frag and stats */
2099	zoneid_t	ixa_zoneid;	/* Assumed always set */
2100
2101	/* Always set for conn_ip_output(); might be stale */
2102	/*
2103	 * Since TCP keeps the conn_t around past the process going away
2104	 * we need to use the "notr" (e.g, ire_refhold_notr) for ixa_ire,
2105	 * ixa_nce, and ixa_dce.
2106	 */
2107	ire_t		*ixa_ire;	/* Forwarding table entry */
2108	uint_t		ixa_ire_generation;
2109	nce_t		*ixa_nce;	/* Neighbor cache entry */
2110	dce_t		*ixa_dce;	/* Destination cache entry */
2111	uint_t		ixa_dce_generation;
2112	uint_t		ixa_src_generation;	/* If IXAF_VERIFY_SOURCE */
2113
2114	uint32_t	ixa_src_preferences;	/* prefs for src addr select */
2115	uint32_t	ixa_pmtu;		/* IXAF_VERIFY_PMTU */
2116
2117	/* Set by ULP if IXAF_VERIFY_PMTU; otherwise set by IP */
2118	uint32_t	ixa_fragsize;
2119
2120	int8_t		ixa_use_min_mtu;	/* IXAF_USE_MIN_MTU values */
2121
2122	pfirepostfrag_t	ixa_postfragfn;		/* Set internally in IP */
2123
2124	in6_addr_t	ixa_nexthop_v6;		/* IXAF_NEXTHOP_SET */
2125#define	ixa_nexthop_v4	V4_PART_OF_V6(ixa_nexthop_v6)
2126
2127	zoneid_t	ixa_no_loop_zoneid;	/* IXAF_NO_LOOP_ZONEID_SET */
2128
2129	uint_t		ixa_scopeid;		/* For IPv6 link-locals */
2130
2131	uint_t		ixa_broadcast_ttl;	/* IXAF_BROACAST_TTL_SET */
2132
2133	uint_t		ixa_multicast_ttl;	/* Assumed set for multicast */
2134	uint_t		ixa_multicast_ifindex;	/* Assumed set for multicast */
2135	ipaddr_t	ixa_multicast_ifaddr;	/* Assumed set for multicast */
2136
2137	int		ixa_raw_cksum_offset;	/* If IXAF_SET_RAW_CKSUM */
2138
2139	uint32_t	ixa_ident;		/* For IPv6 fragment header */
2140
2141	uint64_t	ixa_conn_id;		/* Used by DTrace */
2142	/*
2143	 * Cached LSO information.
2144	 */
2145	ill_lso_capab_t	ixa_lso_capab;		/* Valid when IXAF_LSO_CAPAB */
2146
2147	uint64_t	ixa_ipsec_policy_gen;	/* Generation from iph_gen */
2148	/*
2149	 * The following IPsec fields are only initialized when
2150	 * IXAF_IPSEC_SECURE is set. Otherwise they contain garbage.
2151	 */
2152	ipsec_latch_t	*ixa_ipsec_latch;	/* Just the ids */
2153	struct ipsa_s	*ixa_ipsec_ah_sa;	/* Hard reference SA for AH */
2154	struct ipsa_s	*ixa_ipsec_esp_sa;	/* Hard reference SA for ESP */
2155	struct ipsec_policy_s	*ixa_ipsec_policy; /* why are we here? */
2156	struct ipsec_action_s	*ixa_ipsec_action; /* For reflected packets */
2157	ipsa_ref_t	ixa_ipsec_ref[2];	/* Soft reference to SA */
2158						/* 0: ESP, 1: AH */
2159
2160	/*
2161	 * The selectors here are potentially different than the SPD rule's
2162	 * selectors, and we need to have both available for IKEv2.
2163	 *
2164	 * NOTE: "Source" and "Dest" are w.r.t. outbound datagrams.  Ports can
2165	 *	 be zero, and the protocol number is needed to make the ports
2166	 *	 significant.
2167	 */
2168	uint16_t ixa_ipsec_src_port;	/* Source port number of d-gram. */
2169	uint16_t ixa_ipsec_dst_port;	/* Destination port number of d-gram. */
2170	uint8_t  ixa_ipsec_icmp_type;	/* ICMP type of d-gram */
2171	uint8_t  ixa_ipsec_icmp_code;	/* ICMP code of d-gram */
2172
2173	sa_family_t ixa_ipsec_inaf;	/* Inner address family */
2174#define	IXA_MAX_ADDRLEN 4	/* Max addr len. (in 32-bit words) */
2175	uint32_t ixa_ipsec_insrc[IXA_MAX_ADDRLEN];	/* Inner src address */
2176	uint32_t ixa_ipsec_indst[IXA_MAX_ADDRLEN];	/* Inner dest address */
2177	uint8_t  ixa_ipsec_insrcpfx;	/* Inner source prefix */
2178	uint8_t  ixa_ipsec_indstpfx;	/* Inner destination prefix */
2179
2180	uint8_t ixa_ipsec_proto;	/* IP protocol number for d-gram. */
2181
2182	/* Always initialized independently of ixa_flags settings */
2183	uint_t		ixa_ifindex;	/* Assumed always set */
2184	uint16_t	ixa_ip_hdr_length; /* Points to ULP header */
2185	uint8_t		ixa_protocol;	/* Protocol number for ULP cksum */
2186	ts_label_t	*ixa_tsl;	/* Always set. NULL if not TX */
2187	ip_stack_t	*ixa_ipst;	/* Always set */
2188	uint32_t	ixa_extra_ident; /* Set if LSO */
2189	cred_t		*ixa_cred;	/* For getpeerucred */
2190	pid_t		ixa_cpid;	/* For getpeerucred */
2191
2192#ifdef DEBUG
2193	kthread_t	*ixa_curthread;	/* For serialization assert */
2194#endif
2195	squeue_t	*ixa_sqp;	/* Set from conn_sqp as a hint */
2196	uintptr_t	ixa_cookie;	/* cookie to use for tx flow control */
2197
2198	/*
2199	 * Must be set by ULP if any of IXAF_VERIFY_LSO, IXAF_VERIFY_PMTU,
2200	 * or IXAF_VERIFY_ZCOPY is set.
2201	 */
2202	ixa_notify_t	ixa_notify;	/* Registered upcall notify function */
2203	void		*ixa_notify_cookie; /* ULP cookie for ixa_notify */
2204
2205	uint_t		ixa_tcpcleanup;	/* Used by conn_ixa_cleanup */
2206};
2207
2208/*
2209 * Flags to indicate which transmit attributes are set.
2210 * Split into "xxx_SET" ones which indicate that the "xxx" field it set, and
2211 * single flags.
2212 */
2213#define	IXAF_REACH_CONF		0x00000001	/* Reachability confirmation */
2214#define	IXAF_BROADCAST_TTL_SET	0x00000002	/* ixa_broadcast_ttl valid */
2215#define	IXAF_SET_SOURCE		0x00000004	/* Replace if broadcast */
2216#define	IXAF_USE_MIN_MTU	0x00000008	/* IPV6_USE_MIN_MTU */
2217
2218#define	IXAF_DONTFRAG		0x00000010	/* IP*_DONTFRAG */
2219#define	IXAF_VERIFY_PMTU	0x00000020	/* ixa_pmtu/ixa_fragsize set */
2220#define	IXAF_PMTU_DISCOVERY	0x00000040	/* Create/use PMTU state */
2221#define	IXAF_MULTICAST_LOOP	0x00000080	/* IP_MULTICAST_LOOP */
2222
2223#define	IXAF_IPSEC_SECURE	0x00000100	/* Need IPsec processing */
2224#define	IXAF_UCRED_TSL		0x00000200	/* ixa_tsl from SCM_UCRED */
2225#define	IXAF_DONTROUTE		0x00000400	/* SO_DONTROUTE */
2226#define	IXAF_NO_IPSEC		0x00000800	/* Ignore policy */
2227
2228#define	IXAF_PMTU_TOO_SMALL	0x00001000	/* PMTU too small */
2229#define	IXAF_SET_ULP_CKSUM	0x00002000	/* Calculate ULP checksum */
2230#define	IXAF_VERIFY_SOURCE	0x00004000	/* Check that source is ok */
2231#define	IXAF_NEXTHOP_SET	0x00008000	/* ixa_nexthop set */
2232
2233#define	IXAF_PMTU_IPV4_DF	0x00010000	/* Set IPv4 DF */
2234#define	IXAF_NO_DEV_FLOW_CTL	0x00020000	/* Protocol needs no flow ctl */
2235#define	IXAF_NO_TTL_CHANGE	0x00040000	/* Internal to IP */
2236#define	IXAF_IPV6_ADD_FRAGHDR	0x00080000	/* Add fragment header */
2237
2238#define	IXAF_IPSEC_TUNNEL	0x00100000	/* Tunnel mode */
2239#define	IXAF_NO_PFHOOK		0x00200000	/* Skip xmit pfhook */
2240#define	IXAF_NO_TRACE		0x00400000	/* When back from ARP/ND */
2241#define	IXAF_SCOPEID_SET	0x00800000	/* ixa_scopeid set */
2242
2243#define	IXAF_MULTIRT_MULTICAST	0x01000000	/* MULTIRT for multicast */
2244#define	IXAF_NO_HW_CKSUM	0x02000000	/* Force software cksum */
2245#define	IXAF_SET_RAW_CKSUM	0x04000000	/* Use ixa_raw_cksum_offset */
2246#define	IXAF_IPSEC_GLOBAL_POLICY 0x08000000	/* Policy came from global */
2247
2248/* Note the following uses bits 0x10000000 through 0x80000000 */
2249#define	IXAF_IS_IPV4		IAF_IS_IPV4
2250#define	IXAF_TRUSTED_ICMP	IAF_TRUSTED_ICMP
2251#define	IXAF_NO_LOOP_ZONEID_SET	IAF_NO_LOOP_ZONEID_SET
2252#define	IXAF_LOOPBACK_COPY	IAF_LOOPBACK_COPY
2253
2254/* Note: use the upper 32 bits */
2255#define	IXAF_VERIFY_LSO		0x100000000	/* Check LSO capability */
2256#define	IXAF_LSO_CAPAB		0x200000000	/* Capable of LSO */
2257#define	IXAF_VERIFY_ZCOPY	0x400000000	/* Check Zero Copy capability */
2258#define	IXAF_ZCOPY_CAPAB	0x800000000	/* Capable of ZEROCOPY */
2259
2260/*
2261 * The normal flags for sending packets e.g., icmp errors
2262 */
2263#define	IXAF_BASIC_SIMPLE_V4	\
2264	(IXAF_SET_ULP_CKSUM | IXAF_IS_IPV4 | IXAF_VERIFY_SOURCE)
2265#define	IXAF_BASIC_SIMPLE_V6	(IXAF_SET_ULP_CKSUM | IXAF_VERIFY_SOURCE)
2266
2267/*
2268 * Normally these fields do not have a hold. But in some cases they do, for
2269 * instance when we've gone through ip_*_attr_to/from_mblk.
2270 * We use ixa_free_flags to indicate that they have a hold and need to be
2271 * released on cleanup.
2272 */
2273#define	IXA_FREE_CRED		0x00000001	/* ixa_cred needs to be rele */
2274#define	IXA_FREE_TSL		0x00000002	/* ixa_tsl needs to be rele */
2275
2276/*
2277 * Trivial state machine used to synchronize IXA cleanup for TCP connections.
2278 * See conn_ixa_cleanup().
2279 */
2280#define	IXATC_IDLE		0x00000000
2281#define	IXATC_INPROGRESS	0x00000001
2282#define	IXATC_COMPLETE		0x00000002
2283
2284/*
2285 * Simplistic way to set the ixa_xmit_hint for locally generated traffic
2286 * and forwarded traffic. The shift amount are based on the size of the
2287 * structs to discard the low order bits which don't have much if any variation
2288 * (coloring in kmem_cache_alloc might provide some variation).
2289 *
2290 * Basing the locally generated hint on the address of the conn_t means that
2291 * the packets from the same socket/connection do not get reordered.
2292 * Basing the hint for forwarded traffic on the ill_ring_t means that
2293 * packets from the same NIC+ring are likely to use the same outbound ring
2294 * hence we get low contention on the ring in the transmitting driver.
2295 */
2296#define	CONN_TO_XMIT_HINT(connp)	((uint32_t)(((uintptr_t)connp) >> 11))
2297#define	ILL_RING_TO_XMIT_HINT(ring)	((uint32_t)(((uintptr_t)ring) >> 7))
2298
2299/*
2300 * IP set Destination Flags used by function ip_set_destination,
2301 * ip_attr_connect, and conn_connect.
2302 */
2303#define	IPDF_ALLOW_MCBC		0x1	/* Allow multi/broadcast */
2304#define	IPDF_VERIFY_DST		0x2	/* Verify destination addr */
2305#define	IPDF_SELECT_SRC		0x4	/* Select source address */
2306#define	IPDF_LSO		0x8	/* Try LSO */
2307#define	IPDF_IPSEC		0x10	/* Set IPsec policy */
2308#define	IPDF_ZONE_IS_GLOBAL	0x20	/* From conn_zone_is_global */
2309#define	IPDF_ZCOPY		0x40	/* Try ZEROCOPY */
2310#define	IPDF_UNIQUE_DCE		0x80	/* Get a per-destination DCE */
2311
2312/*
2313 * Receive side attributes used between the transport protocols and IP as
2314 * well as inside IP.
2315 */
2316struct ip_recv_attr_s {
2317	iaflags_t	ira_flags;	/* See below */
2318
2319	uint32_t	ira_free_flags;	/* IRA_FREE_*. See below */
2320
2321	/*
2322	 * This is a hint for TCP SYN packets.
2323	 * Always initialized independently of ira_flags settings
2324	 */
2325	squeue_t	*ira_sqp;
2326	ill_rx_ring_t	*ira_ring;	/* Internal to IP */
2327
2328	/* For ip_accept_tcp when IRAF_TARGET_SQP is set */
2329	squeue_t	*ira_target_sqp;
2330	mblk_t		*ira_target_sqp_mp;
2331
2332	/* Always initialized independently of ira_flags settings */
2333	uint32_t	ira_xmit_hint;	/* For ECMP and GLD TX ring fanout */
2334	zoneid_t	ira_zoneid;	/* ALL_ZONES unless local delivery */
2335	uint_t		ira_pktlen;	/* Always set. For frag and stats */
2336	uint16_t	ira_ip_hdr_length; /* Points to ULP header */
2337	uint8_t		ira_protocol;	/* Protocol number for ULP cksum */
2338	uint_t		ira_rifindex;	/* Received ifindex */
2339	uint_t		ira_ruifindex;	/* Received upper ifindex */
2340	ts_label_t	*ira_tsl;	/* Always set. NULL if not TX */
2341	/*
2342	 * ira_rill and ira_ill is set inside IP, but not when conn_recv is
2343	 * called; ULPs should use ira_ruifindex instead.
2344	 */
2345	ill_t		*ira_rill;	/* ill where packet came */
2346	ill_t		*ira_ill;	/* ill where IP address hosted */
2347	cred_t		*ira_cred;	/* For getpeerucred */
2348	pid_t		ira_cpid;	/* For getpeerucred */
2349
2350	/* Used when IRAF_VERIFIED_SRC is set; this source was ok */
2351	ipaddr_t	ira_verified_src;
2352
2353	/*
2354	 * The following IPsec fields are only initialized when
2355	 * IRAF_IPSEC_SECURE is set. Otherwise they contain garbage.
2356	 */
2357	struct ipsec_action_s *ira_ipsec_action; /* how we made it in.. */
2358	struct ipsa_s	*ira_ipsec_ah_sa;	/* SA for AH */
2359	struct ipsa_s	*ira_ipsec_esp_sa;	/* SA for ESP */
2360
2361	ipaddr_t	ira_mroute_tunnel;	/* IRAF_MROUTE_TUNNEL_SET */
2362
2363	zoneid_t	ira_no_loop_zoneid;	/* IRAF_NO_LOOP_ZONEID_SET */
2364
2365	uint32_t	ira_esp_udp_ports;	/* IRAF_ESP_UDP_PORTS */
2366
2367	/*
2368	 * For IP_RECVSLLA and ip_ndp_conflict/find_solicitation.
2369	 * Same size as max for sockaddr_dl
2370	 */
2371#define	IRA_L2SRC_SIZE	244
2372	uint8_t		ira_l2src[IRA_L2SRC_SIZE];	/* If IRAF_L2SRC_SET */
2373
2374	/*
2375	 * Local handle that we use to do lazy setting of ira_l2src.
2376	 * We defer setting l2src until needed but we do before any
2377	 * ip_input pullupmsg or copymsg.
2378	 */
2379	struct mac_header_info_s *ira_mhip;	/* Could be NULL */
2380};
2381
2382/*
2383 * Flags to indicate which receive attributes are set.
2384 */
2385#define	IRAF_SYSTEM_LABELED	0x00000001	/* is_system_labeled() */
2386#define	IRAF_IPV4_OPTIONS	0x00000002	/* Performance */
2387#define	IRAF_MULTICAST		0x00000004	/* Was multicast at L3 */
2388#define	IRAF_BROADCAST		0x00000008	/* Was broadcast at L3 */
2389#define	IRAF_MULTIBROADCAST	(IRAF_MULTICAST|IRAF_BROADCAST)
2390
2391#define	IRAF_LOOPBACK		0x00000010	/* Looped back by IP */
2392#define	IRAF_VERIFY_IP_CKSUM	0x00000020	/* Need to verify IP */
2393#define	IRAF_VERIFY_ULP_CKSUM	0x00000040	/* Need to verify TCP,UDP,etc */
2394#define	IRAF_SCTP_CSUM_ERR	0x00000080	/* sctp pkt has failed chksum */
2395
2396#define	IRAF_IPSEC_SECURE	0x00000100	/* Passed AH and/or ESP */
2397#define	IRAF_DHCP_UNICAST	0x00000200
2398#define	IRAF_IPSEC_DECAPS	0x00000400	/* Was packet decapsulated */
2399					/* from a matching inner packet? */
2400#define	IRAF_TARGET_SQP		0x00000800	/* ira_target_sqp is set */
2401#define	IRAF_VERIFIED_SRC	0x00001000	/* ira_verified_src set */
2402#define	IRAF_RSVP		0x00002000	/* RSVP packet for rsvpd */
2403#define	IRAF_MROUTE_TUNNEL_SET	0x00004000	/* From ip_mroute_decap */
2404#define	IRAF_PIM_REGISTER	0x00008000	/* From register_mforward */
2405
2406#define	IRAF_TX_MAC_EXEMPTABLE	0x00010000	/* Allow MAC_EXEMPT readdown */
2407#define	IRAF_TX_SHARED_ADDR	0x00020000	/* Arrived on ALL_ZONES addr */
2408#define	IRAF_ESP_UDP_PORTS	0x00040000	/* NAT-traversal packet */
2409#define	IRAF_NO_HW_CKSUM	0x00080000	/* Force software cksum */
2410
2411#define	IRAF_ICMP_ERROR		0x00100000	/* Send to conn_recvicmp */
2412#define	IRAF_ROUTER_ALERT	0x00200000	/* IPv6 router alert */
2413#define	IRAF_L2SRC_SET		0x00400000	/* ira_l2src has been set */
2414#define	IRAF_L2SRC_LOOPBACK	0x00800000	/* Came from us */
2415
2416#define	IRAF_L2DST_MULTICAST	0x01000000	/* Multicast at L2 */
2417#define	IRAF_L2DST_BROADCAST	0x02000000	/* Broadcast at L2 */
2418/* Unused 0x04000000 */
2419/* Unused 0x08000000 */
2420
2421/* Below starts with 0x10000000 */
2422#define	IRAF_IS_IPV4		IAF_IS_IPV4
2423#define	IRAF_TRUSTED_ICMP	IAF_TRUSTED_ICMP
2424#define	IRAF_NO_LOOP_ZONEID_SET	IAF_NO_LOOP_ZONEID_SET
2425#define	IRAF_LOOPBACK_COPY	IAF_LOOPBACK_COPY
2426
2427/*
2428 * Normally these fields do not have a hold. But in some cases they do, for
2429 * instance when we've gone through ip_*_attr_to/from_mblk.
2430 * We use ira_free_flags to indicate that they have a hold and need to be
2431 * released on cleanup.
2432 */
2433#define	IRA_FREE_CRED		0x00000001	/* ira_cred needs to be rele */
2434#define	IRA_FREE_TSL		0x00000002	/* ira_tsl needs to be rele */
2435
2436/*
2437 * Optional destination cache entry for path MTU information,
2438 * and ULP metrics.
2439 */
2440struct dce_s {
2441	uint_t		dce_generation;	/* Changed since cached? */
2442	uint_t		dce_flags;	/* See below */
2443	uint_t		dce_ipversion;	/* IPv4/IPv6 version */
2444	uint32_t	dce_pmtu;	/* Path MTU if DCEF_PMTU */
2445	uint32_t	dce_ident;	/* Per destination IP ident. */
2446	iulp_t		dce_uinfo;	/* Metrics if DCEF_UINFO */
2447
2448	struct dce_s	*dce_next;
2449	struct dce_s	**dce_ptpn;
2450	struct dcb_s	*dce_bucket;
2451
2452	union {
2453		in6_addr_t	dceu_v6addr;
2454		ipaddr_t	dceu_v4addr;
2455	} dce_u;
2456#define	dce_v4addr	dce_u.dceu_v4addr
2457#define	dce_v6addr	dce_u.dceu_v6addr
2458	/* Note that for IPv6+IPMP we use the ifindex for the upper interface */
2459	uint_t		dce_ifindex;	/* For IPv6 link-locals */
2460
2461	kmutex_t	dce_lock;
2462	uint_t		dce_refcnt;
2463	uint64_t	dce_last_change_time;	/* Path MTU. In seconds */
2464
2465	ip_stack_t	*dce_ipst;	/* Does not have a netstack_hold */
2466};
2467
2468/*
2469 * Values for dce_generation.
2470 *
2471 * If a DCE has DCE_GENERATION_CONDEMNED, the last dce_refrele should delete
2472 * it.
2473 *
2474 * DCE_GENERATION_VERIFY is never stored in dce_generation but it is
2475 * stored in places that cache DCE (such as ixa_dce_generation).
2476 * It is used as a signal that the cache is stale and needs to be reverified.
2477 */
2478#define	DCE_GENERATION_CONDEMNED	0
2479#define	DCE_GENERATION_VERIFY		1
2480#define	DCE_GENERATION_INITIAL		2
2481#define	DCE_IS_CONDEMNED(dce) \
2482	((dce)->dce_generation == DCE_GENERATION_CONDEMNED)
2483
2484
2485/*
2486 * Values for ips_src_generation.
2487 *
2488 * SRC_GENERATION_VERIFY is never stored in ips_src_generation but it is
2489 * stored in places that cache IREs (ixa_src_generation). It is used as a
2490 * signal that the cache is stale and needs to be reverified.
2491 */
2492#define	SRC_GENERATION_VERIFY		0
2493#define	SRC_GENERATION_INITIAL		1
2494
2495/*
2496 * The kernel stores security attributes of all gateways in a database made
2497 * up of one or more tsol_gcdb_t elements.  Each tsol_gcdb_t contains the
2498 * security-related credentials of the gateway.  More than one gateways may
2499 * share entries in the database.
2500 *
2501 * The tsol_gc_t structure represents the gateway to credential association,
2502 * and refers to an entry in the database.  One or more tsol_gc_t entities are
2503 * grouped together to form one or more tsol_gcgrp_t, each representing the
2504 * list of security attributes specific to the gateway.  A gateway may be
2505 * associated with at most one credentials group.
2506 */
2507struct tsol_gcgrp_s;
2508
2509extern uchar_t	ip6opt_ls;	/* TX IPv6 enabler */
2510
2511/*
2512 * Gateway security credential record.
2513 */
2514typedef struct tsol_gcdb_s {
2515	uint_t		gcdb_refcnt;	/* reference count */
2516	struct rtsa_s	gcdb_attr;	/* security attributes */
2517#define	gcdb_mask	gcdb_attr.rtsa_mask
2518#define	gcdb_doi	gcdb_attr.rtsa_doi
2519#define	gcdb_slrange	gcdb_attr.rtsa_slrange
2520} tsol_gcdb_t;
2521
2522/*
2523 * Gateway to credential association.
2524 */
2525typedef struct tsol_gc_s {
2526	uint_t		gc_refcnt;	/* reference count */
2527	struct tsol_gcgrp_s *gc_grp;	/* pointer to group */
2528	struct tsol_gc_s *gc_prev;	/* previous in list */
2529	struct tsol_gc_s *gc_next;	/* next in list */
2530	tsol_gcdb_t	*gc_db;		/* pointer to actual credentials */
2531} tsol_gc_t;
2532
2533/*
2534 * Gateway credentials group address.
2535 */
2536typedef struct tsol_gcgrp_addr_s {
2537	int		ga_af;		/* address family */
2538	in6_addr_t	ga_addr;	/* IPv4 mapped or IPv6 address */
2539} tsol_gcgrp_addr_t;
2540
2541/*
2542 * Gateway credentials group.
2543 */
2544typedef struct tsol_gcgrp_s {
2545	uint_t		gcgrp_refcnt;	/* reference count */
2546	krwlock_t	gcgrp_rwlock;	/* lock to protect following */
2547	uint_t		gcgrp_count;	/* number of credentials */
2548	tsol_gc_t	*gcgrp_head;	/* first credential in list */
2549	tsol_gc_t	*gcgrp_tail;	/* last credential in list */
2550	tsol_gcgrp_addr_t gcgrp_addr;	/* next-hop gateway address */
2551} tsol_gcgrp_t;
2552
2553extern kmutex_t gcgrp_lock;
2554
2555#define	GC_REFRELE(p) {				\
2556	ASSERT((p)->gc_grp != NULL);		\
2557	rw_enter(&(p)->gc_grp->gcgrp_rwlock, RW_WRITER); \
2558	ASSERT((p)->gc_refcnt > 0);		\
2559	if (--((p)->gc_refcnt) == 0)		\
2560		gc_inactive(p);			\
2561	else					\
2562		rw_exit(&(p)->gc_grp->gcgrp_rwlock); \
2563}
2564
2565#define	GCGRP_REFHOLD(p) {			\
2566	mutex_enter(&gcgrp_lock);		\
2567	++((p)->gcgrp_refcnt);			\
2568	ASSERT((p)->gcgrp_refcnt != 0);		\
2569	mutex_exit(&gcgrp_lock);		\
2570}
2571
2572#define	GCGRP_REFRELE(p) {			\
2573	mutex_enter(&gcgrp_lock);		\
2574	ASSERT((p)->gcgrp_refcnt > 0);		\
2575	if (--((p)->gcgrp_refcnt) == 0)		\
2576		gcgrp_inactive(p);		\
2577	ASSERT(MUTEX_HELD(&gcgrp_lock));	\
2578	mutex_exit(&gcgrp_lock);		\
2579}
2580
2581/*
2582 * IRE gateway security attributes structure, pointed to by tsol_ire_gw_secattr
2583 */
2584struct tsol_tnrhc;
2585
2586struct tsol_ire_gw_secattr_s {
2587	kmutex_t	igsa_lock;	/* lock to protect following */
2588	struct tsol_tnrhc *igsa_rhc;	/* host entry for gateway */
2589	tsol_gc_t	*igsa_gc;	/* for prefix IREs */
2590};
2591
2592void irb_refrele_ftable(irb_t *);
2593
2594extern struct kmem_cache *rt_entry_cache;
2595
2596typedef struct ire4 {
2597	ipaddr_t ire4_mask;		/* Mask for matching this IRE. */
2598	ipaddr_t ire4_addr;		/* Address this IRE represents. */
2599	ipaddr_t ire4_gateway_addr;	/* Gateway including for IRE_ONLINK */
2600	ipaddr_t ire4_setsrc_addr;	/* RTF_SETSRC */
2601} ire4_t;
2602
2603typedef struct ire6 {
2604	in6_addr_t ire6_mask;		/* Mask for matching this IRE. */
2605	in6_addr_t ire6_addr;		/* Address this IRE represents. */
2606	in6_addr_t ire6_gateway_addr;	/* Gateway including for IRE_ONLINK */
2607	in6_addr_t ire6_setsrc_addr;	/* RTF_SETSRC */
2608} ire6_t;
2609
2610typedef union ire_addr {
2611	ire6_t	ire6_u;
2612	ire4_t	ire4_u;
2613} ire_addr_u_t;
2614
2615/*
2616 * Internet Routing Entry
2617 * When we have multiple identical IREs we logically add them by manipulating
2618 * ire_identical_ref and ire_delete first decrements
2619 * that and when it reaches 1 we know it is the last IRE.
2620 * "identical" is defined as being the same for:
2621 * ire_addr, ire_netmask, ire_gateway, ire_ill, ire_zoneid, and ire_type
2622 * For instance, multiple IRE_BROADCASTs for the same subnet number are
2623 * viewed as identical, and so are the IRE_INTERFACEs when there are
2624 * multiple logical interfaces (on the same ill) with the same subnet prefix.
2625 */
2626struct ire_s {
2627	struct	ire_s	*ire_next;	/* The hash chain must be first. */
2628	struct	ire_s	**ire_ptpn;	/* Pointer to previous next. */
2629	uint32_t	ire_refcnt;	/* Number of references */
2630	ill_t		*ire_ill;
2631	uint32_t	ire_identical_ref; /* IRE_INTERFACE, IRE_BROADCAST */
2632	uchar_t		ire_ipversion;	/* IPv4/IPv6 version */
2633	ushort_t	ire_type;	/* Type of IRE */
2634	uint_t		ire_generation;	/* Generation including CONDEMNED */
2635	uint_t	ire_ib_pkt_count;	/* Inbound packets for ire_addr */
2636	uint_t	ire_ob_pkt_count;	/* Outbound packets to ire_addr */
2637	time_t	ire_create_time;	/* Time (in secs) IRE was created. */
2638	uint32_t	ire_flags;	/* flags related to route (RTF_*) */
2639	/*
2640	 * ire_testhidden is TRUE for INTERFACE IREs of IS_UNDER_IPMP(ill)
2641	 * interfaces
2642	 */
2643	boolean_t	ire_testhidden;
2644	pfirerecv_t	ire_recvfn;	/* Receive side handling */
2645	pfiresend_t	ire_sendfn;	/* Send side handling */
2646	pfirepostfrag_t	ire_postfragfn;	/* Bottom end of send handling */
2647
2648	uint_t		ire_masklen;	/* # bits in ire_mask{,_v6} */
2649	ire_addr_u_t	ire_u;		/* IPv4/IPv6 address info. */
2650
2651	irb_t		*ire_bucket;	/* Hash bucket when ire_ptphn is set */
2652	kmutex_t	ire_lock;
2653	clock_t		ire_last_used_time;	/* For IRE_LOCAL reception */
2654	tsol_ire_gw_secattr_t *ire_gw_secattr; /* gateway security attributes */
2655	zoneid_t	ire_zoneid;
2656
2657	/*
2658	 * Cached information of where to send packets that match this route.
2659	 * The ire_dep_* information is used to determine when ire_nce_cache
2660	 * needs to be updated.
2661	 * ire_nce_cache is the fastpath for the Neighbor Cache Entry
2662	 * for IPv6; arp info for IPv4
2663	 * Since this is a cache setup and torn down independently of
2664	 * applications we need to use nce_ref{rele,hold}_notr for it.
2665	 */
2666	nce_t		*ire_nce_cache;
2667
2668	/*
2669	 * Quick check whether the ire_type and ire_masklen indicates
2670	 * that the IRE can have ire_nce_cache set i.e., whether it is
2671	 * IRE_ONLINK and for a single destination.
2672	 */
2673	boolean_t	ire_nce_capable;
2674
2675	/*
2676	 * Dependency tracking so we can safely cache IRE and NCE pointers
2677	 * in offlink and onlink IREs.
2678	 * These are locked under the ips_ire_dep_lock rwlock. Write held
2679	 * when modifying the linkage.
2680	 * ire_dep_parent (Also chain towards IRE for nexthop)
2681	 * ire_dep_parent_generation: ire_generation of ire_dep_parent
2682	 * ire_dep_children (From parent to first child)
2683	 * ire_dep_sib_next (linked list of siblings)
2684	 * ire_dep_sib_ptpn (linked list of siblings)
2685	 *
2686	 * The parent has a ire_refhold on each child, and each child has
2687	 * an ire_refhold on its parent.
2688	 * Since ire_dep_parent is a cache setup and torn down independently of
2689	 * applications we need to use ire_ref{rele,hold}_notr for it.
2690	 */
2691	ire_t		*ire_dep_parent;
2692	ire_t		*ire_dep_children;
2693	ire_t		*ire_dep_sib_next;
2694	ire_t		**ire_dep_sib_ptpn;	/* Pointer to previous next */
2695	uint_t		ire_dep_parent_generation;
2696
2697	uint_t		ire_badcnt;	/* Number of times ND_UNREACHABLE */
2698	uint64_t	ire_last_badcnt;	/* In seconds */
2699
2700	/* ire_defense* and ire_last_used_time are only used on IRE_LOCALs */
2701	uint_t		ire_defense_count;	/* number of ARP conflicts */
2702	uint_t		ire_defense_time;	/* last time defended (secs) */
2703
2704	boolean_t	ire_trace_disable;	/* True when alloc fails */
2705	ip_stack_t	*ire_ipst;	/* Does not have a netstack_hold */
2706	iulp_t		ire_metrics;
2707	/*
2708	 * default and prefix routes that are added without explicitly
2709	 * specifying the interface are termed "unbound" routes, and will
2710	 * have ire_unbound set to true.
2711	 */
2712	boolean_t	ire_unbound;
2713};
2714
2715/* IPv4 compatibility macros */
2716#define	ire_mask		ire_u.ire4_u.ire4_mask
2717#define	ire_addr		ire_u.ire4_u.ire4_addr
2718#define	ire_gateway_addr	ire_u.ire4_u.ire4_gateway_addr
2719#define	ire_setsrc_addr		ire_u.ire4_u.ire4_setsrc_addr
2720
2721#define	ire_mask_v6		ire_u.ire6_u.ire6_mask
2722#define	ire_addr_v6		ire_u.ire6_u.ire6_addr
2723#define	ire_gateway_addr_v6	ire_u.ire6_u.ire6_gateway_addr
2724#define	ire_setsrc_addr_v6	ire_u.ire6_u.ire6_setsrc_addr
2725
2726/*
2727 * Values for ire_generation.
2728 *
2729 * If an IRE is marked with IRE_IS_CONDEMNED, the last walker of
2730 * the bucket should delete this IRE from this bucket.
2731 *
2732 * IRE_GENERATION_VERIFY is never stored in ire_generation but it is
2733 * stored in places that cache IREs (such as ixa_ire_generation and
2734 * ire_dep_parent_generation). It is used as a signal that the cache is
2735 * stale and needs to be reverified.
2736 */
2737#define	IRE_GENERATION_CONDEMNED	0
2738#define	IRE_GENERATION_VERIFY		1
2739#define	IRE_GENERATION_INITIAL		2
2740#define	IRE_IS_CONDEMNED(ire) \
2741	((ire)->ire_generation == IRE_GENERATION_CONDEMNED)
2742
2743/* Convenient typedefs for sockaddrs */
2744typedef	struct sockaddr_in	sin_t;
2745typedef	struct sockaddr_in6	sin6_t;
2746
2747/* Name/Value Descriptor. */
2748typedef struct nv_s {
2749	uint64_t nv_value;
2750	char	*nv_name;
2751} nv_t;
2752
2753#define	ILL_FRAG_HASH(s, i) \
2754	((ntohl(s) ^ ((i) ^ ((i) >> 8))) % ILL_FRAG_HASH_TBL_COUNT)
2755
2756/*
2757 * The MAX number of allowed fragmented packets per hash bucket
2758 * calculation is based on the most common mtu size of 1500. This limit
2759 * will work well for other mtu sizes as well.
2760 */
2761#define	COMMON_IP_MTU 1500
2762#define	MAX_FRAG_MIN 10
2763#define	MAX_FRAG_PKTS(ipst)	\
2764	MAX(MAX_FRAG_MIN, (2 * (ipst->ips_ip_reass_queue_bytes / \
2765	    (COMMON_IP_MTU * ILL_FRAG_HASH_TBL_COUNT))))
2766
2767/*
2768 * Maximum dups allowed per packet.
2769 */
2770extern uint_t ip_max_frag_dups;
2771
2772/*
2773 * Per-packet information for received packets and transmitted.
2774 * Used by the transport protocols when converting between the packet
2775 * and ancillary data and socket options.
2776 *
2777 * Note: This private data structure and related IPPF_* constant
2778 * definitions are exposed to enable compilation of some debugging tools
2779 * like lsof which use struct tcp_t in <inet/tcp.h>. This is intended to be
2780 * a temporary hack and long term alternate interfaces should be defined
2781 * to support the needs of such tools and private definitions moved to
2782 * private headers.
2783 */
2784struct ip_pkt_s {
2785	uint_t		ipp_fields;		/* Which fields are valid */
2786	in6_addr_t	ipp_addr;		/* pktinfo src/dst addr */
2787#define	ipp_addr_v4	V4_PART_OF_V6(ipp_addr)
2788	uint_t		ipp_unicast_hops;	/* IPV6_UNICAST_HOPS, IP_TTL */
2789	uint_t		ipp_hoplimit;		/* IPV6_HOPLIMIT */
2790	uint_t		ipp_hopoptslen;
2791	uint_t		ipp_rthdrdstoptslen;
2792	uint_t		ipp_rthdrlen;
2793	uint_t		ipp_dstoptslen;
2794	uint_t		ipp_fraghdrlen;
2795	ip6_hbh_t	*ipp_hopopts;
2796	ip6_dest_t	*ipp_rthdrdstopts;
2797	ip6_rthdr_t	*ipp_rthdr;
2798	ip6_dest_t	*ipp_dstopts;
2799	ip6_frag_t	*ipp_fraghdr;
2800	uint8_t		ipp_tclass;		/* IPV6_TCLASS */
2801	uint8_t		ipp_type_of_service;	/* IP_TOS */
2802	uint_t		ipp_ipv4_options_len;	/* Len of IPv4 options */
2803	uint8_t		*ipp_ipv4_options;	/* Ptr to IPv4 options */
2804	uint_t		ipp_label_len_v4;	/* Len of TX label for IPv4 */
2805	uint8_t		*ipp_label_v4;		/* TX label for IPv4 */
2806	uint_t		ipp_label_len_v6;	/* Len of TX label for IPv6 */
2807	uint8_t		*ipp_label_v6;		/* TX label for IPv6 */
2808};
2809typedef struct ip_pkt_s ip_pkt_t;
2810
2811extern void ip_pkt_free(ip_pkt_t *);	/* free storage inside ip_pkt_t */
2812extern ipaddr_t ip_pkt_source_route_v4(const ip_pkt_t *);
2813extern in6_addr_t *ip_pkt_source_route_v6(const ip_pkt_t *);
2814extern int ip_pkt_copy(ip_pkt_t *, ip_pkt_t *, int);
2815extern void ip_pkt_source_route_reverse_v4(ip_pkt_t *);
2816
2817/* ipp_fields values */
2818#define	IPPF_ADDR		0x0001	/* Part of in6_pktinfo: src/dst addr */
2819#define	IPPF_HOPLIMIT		0x0002	/* Overrides unicast and multicast */
2820#define	IPPF_TCLASS		0x0004	/* Overrides class in sin6_flowinfo */
2821
2822#define	IPPF_HOPOPTS		0x0010	/* ipp_hopopts set */
2823#define	IPPF_RTHDR		0x0020	/* ipp_rthdr set */
2824#define	IPPF_RTHDRDSTOPTS	0x0040	/* ipp_rthdrdstopts set */
2825#define	IPPF_DSTOPTS		0x0080	/* ipp_dstopts set */
2826
2827#define	IPPF_IPV4_OPTIONS	0x0100	/* ipp_ipv4_options set */
2828#define	IPPF_LABEL_V4		0x0200	/* ipp_label_v4 set */
2829#define	IPPF_LABEL_V6		0x0400	/* ipp_label_v6 set */
2830
2831#define	IPPF_FRAGHDR		0x0800	/* Used for IPsec receive side */
2832
2833/*
2834 * Data structure which is passed to conn_opt_get/set.
2835 * The conn_t is included even though it can be inferred from queue_t.
2836 * setsockopt and getsockopt use conn_ixa and conn_xmit_ipp. However,
2837 * when handling ancillary data we use separate ixa and ipps.
2838 */
2839typedef struct conn_opt_arg_s {
2840	conn_t		*coa_connp;
2841	ip_xmit_attr_t	*coa_ixa;
2842	ip_pkt_t	*coa_ipp;
2843	boolean_t	coa_ancillary;	/* Ancillary data and not setsockopt */
2844	uint_t		coa_changed;	/* See below */
2845} conn_opt_arg_t;
2846
2847/*
2848 * Flags for what changed.
2849 * If we want to be more efficient in the future we can have more fine
2850 * grained flags e.g., a flag for just IP_TOS changing.
2851 * For now we either call ip_set_destination (for "route changed")
2852 * and/or conn_build_hdr_template/conn_prepend_hdr (for "header changed").
2853 */
2854#define	COA_HEADER_CHANGED	0x0001
2855#define	COA_ROUTE_CHANGED	0x0002
2856#define	COA_RCVBUF_CHANGED	0x0004	/* SO_RCVBUF */
2857#define	COA_SNDBUF_CHANGED	0x0008	/* SO_SNDBUF */
2858#define	COA_WROFF_CHANGED	0x0010	/* Header size changed */
2859#define	COA_ICMP_BIND_NEEDED	0x0020
2860#define	COA_OOBINLINE_CHANGED	0x0040
2861
2862#define	TCP_PORTS_OFFSET	0
2863#define	UDP_PORTS_OFFSET	0
2864
2865/*
2866 * lookups return the ill/ipif only if the flags are clear OR Iam writer.
2867 * ill / ipif lookup functions increment the refcnt on the ill / ipif only
2868 * after calling these macros. This ensures that the refcnt on the ipif or
2869 * ill will eventually drop down to zero.
2870 */
2871#define	ILL_LOOKUP_FAILED	1	/* Used as error code */
2872#define	IPIF_LOOKUP_FAILED	2	/* Used as error code */
2873
2874#define	ILL_CAN_LOOKUP(ill)						\
2875	(!((ill)->ill_state_flags & ILL_CONDEMNED) ||			\
2876	IAM_WRITER_ILL(ill))
2877
2878#define	ILL_IS_CONDEMNED(ill)	\
2879	((ill)->ill_state_flags & ILL_CONDEMNED)
2880
2881#define	IPIF_CAN_LOOKUP(ipif)	\
2882	(!((ipif)->ipif_state_flags & IPIF_CONDEMNED) || \
2883	IAM_WRITER_IPIF(ipif))
2884
2885#define	IPIF_IS_CONDEMNED(ipif)	\
2886	((ipif)->ipif_state_flags & IPIF_CONDEMNED)
2887
2888#define	IPIF_IS_CHANGING(ipif)	\
2889	((ipif)->ipif_state_flags & IPIF_CHANGING)
2890
2891/* Macros used to assert that this thread is a writer */
2892#define	IAM_WRITER_IPSQ(ipsq)	((ipsq)->ipsq_xop->ipx_writer == curthread)
2893#define	IAM_WRITER_ILL(ill)	IAM_WRITER_IPSQ((ill)->ill_phyint->phyint_ipsq)
2894#define	IAM_WRITER_IPIF(ipif)	IAM_WRITER_ILL((ipif)->ipif_ill)
2895
2896/*
2897 * Grab ill locks in the proper order. The order is highest addressed
2898 * ill is locked first.
2899 */
2900#define	GRAB_ILL_LOCKS(ill_1, ill_2)				\
2901{								\
2902	if ((ill_1) > (ill_2)) {				\
2903		if (ill_1 != NULL)				\
2904			mutex_enter(&(ill_1)->ill_lock);	\
2905		if (ill_2 != NULL)				\
2906			mutex_enter(&(ill_2)->ill_lock);	\
2907	} else {						\
2908		if (ill_2 != NULL)				\
2909			mutex_enter(&(ill_2)->ill_lock);	\
2910		if (ill_1 != NULL && ill_1 != ill_2)		\
2911			mutex_enter(&(ill_1)->ill_lock);	\
2912	}							\
2913}
2914
2915#define	RELEASE_ILL_LOCKS(ill_1, ill_2)		\
2916{						\
2917	if (ill_1 != NULL)			\
2918		mutex_exit(&(ill_1)->ill_lock);	\
2919	if (ill_2 != NULL && ill_2 != ill_1)	\
2920		mutex_exit(&(ill_2)->ill_lock);	\
2921}
2922
2923/* Get the other protocol instance ill */
2924#define	ILL_OTHER(ill)						\
2925	((ill)->ill_isv6 ? (ill)->ill_phyint->phyint_illv4 :	\
2926	    (ill)->ill_phyint->phyint_illv6)
2927
2928/* ioctl command info: Ioctl properties extracted and stored in here */
2929typedef struct cmd_info_s
2930{
2931	ipif_t  *ci_ipif;	/* ipif associated with [l]ifreq ioctl's */
2932	sin_t	*ci_sin;	/* the sin struct passed down */
2933	sin6_t	*ci_sin6;	/* the sin6_t struct passed down */
2934	struct lifreq *ci_lifr;	/* the lifreq struct passed down */
2935} cmd_info_t;
2936
2937extern struct kmem_cache *ire_cache;
2938
2939extern ipaddr_t	ip_g_all_ones;
2940
2941extern uint_t	ip_loopback_mtu;	/* /etc/system */
2942extern uint_t	ip_loopback_mtuplus;
2943extern uint_t	ip_loopback_mtu_v6plus;
2944
2945extern vmem_t *ip_minor_arena_sa;
2946extern vmem_t *ip_minor_arena_la;
2947
2948/*
2949 * ip_g_forward controls IP forwarding.  It takes two values:
2950 *	0: IP_FORWARD_NEVER	Don't forward packets ever.
2951 *	1: IP_FORWARD_ALWAYS	Forward packets for elsewhere.
2952 *
2953 * RFC1122 says there must be a configuration switch to control forwarding,
2954 * but that the default MUST be to not forward packets ever.  Implicit
2955 * control based on configuration of multiple interfaces MUST NOT be
2956 * implemented (Section 3.1).  SunOS 4.1 did provide the "automatic" capability
2957 * and, in fact, it was the default.  That capability is now provided in the
2958 * /etc/rc2.d/S69inet script.
2959 */
2960
2961#define	ips_ip_respond_to_address_mask_broadcast \
2962					ips_propinfo_tbl[0].prop_cur_bval
2963#define	ips_ip_g_resp_to_echo_bcast	ips_propinfo_tbl[1].prop_cur_bval
2964#define	ips_ip_g_resp_to_echo_mcast	ips_propinfo_tbl[2].prop_cur_bval
2965#define	ips_ip_g_resp_to_timestamp	ips_propinfo_tbl[3].prop_cur_bval
2966#define	ips_ip_g_resp_to_timestamp_bcast ips_propinfo_tbl[4].prop_cur_bval
2967#define	ips_ip_g_send_redirects		ips_propinfo_tbl[5].prop_cur_bval
2968#define	ips_ip_g_forward_directed_bcast	ips_propinfo_tbl[6].prop_cur_bval
2969#define	ips_ip_mrtdebug			ips_propinfo_tbl[7].prop_cur_uval
2970#define	ips_ip_ire_reclaim_fraction	ips_propinfo_tbl[8].prop_cur_uval
2971#define	ips_ip_nce_reclaim_fraction	ips_propinfo_tbl[9].prop_cur_uval
2972#define	ips_ip_dce_reclaim_fraction	ips_propinfo_tbl[10].prop_cur_uval
2973#define	ips_ip_def_ttl			ips_propinfo_tbl[11].prop_cur_uval
2974#define	ips_ip_forward_src_routed	ips_propinfo_tbl[12].prop_cur_bval
2975#define	ips_ip_wroff_extra		ips_propinfo_tbl[13].prop_cur_uval
2976#define	ips_ip_pathmtu_interval		ips_propinfo_tbl[14].prop_cur_uval
2977#define	ips_ip_icmp_return		ips_propinfo_tbl[15].prop_cur_uval
2978#define	ips_ip_path_mtu_discovery	ips_propinfo_tbl[16].prop_cur_bval
2979#define	ips_ip_pmtu_min			ips_propinfo_tbl[17].prop_cur_uval
2980#define	ips_ip_ignore_redirect		ips_propinfo_tbl[18].prop_cur_bval
2981#define	ips_ip_arp_icmp_error		ips_propinfo_tbl[19].prop_cur_bval
2982#define	ips_ip_broadcast_ttl		ips_propinfo_tbl[20].prop_cur_uval
2983#define	ips_ip_icmp_err_interval	ips_propinfo_tbl[21].prop_cur_uval
2984#define	ips_ip_icmp_err_burst		ips_propinfo_tbl[22].prop_cur_uval
2985#define	ips_ip_reass_queue_bytes	ips_propinfo_tbl[23].prop_cur_uval
2986#define	ips_ip_strict_dst_multihoming	ips_propinfo_tbl[24].prop_cur_uval
2987#define	ips_ip_addrs_per_if		ips_propinfo_tbl[25].prop_cur_uval
2988#define	ips_ipsec_override_persocket_policy ips_propinfo_tbl[26].prop_cur_bval
2989#define	ips_icmp_accept_clear_messages	ips_propinfo_tbl[27].prop_cur_bval
2990#define	ips_igmp_accept_clear_messages	ips_propinfo_tbl[28].prop_cur_bval
2991
2992/* IPv6 configuration knobs */
2993#define	ips_delay_first_probe_time	ips_propinfo_tbl[29].prop_cur_uval
2994#define	ips_max_unicast_solicit		ips_propinfo_tbl[30].prop_cur_uval
2995#define	ips_ipv6_def_hops		ips_propinfo_tbl[31].prop_cur_uval
2996#define	ips_ipv6_icmp_return		ips_propinfo_tbl[32].prop_cur_uval
2997#define	ips_ipv6_forward_src_routed	ips_propinfo_tbl[33].prop_cur_bval
2998#define	ips_ipv6_resp_echo_mcast	ips_propinfo_tbl[34].prop_cur_bval
2999#define	ips_ipv6_send_redirects		ips_propinfo_tbl[35].prop_cur_bval
3000#define	ips_ipv6_ignore_redirect	ips_propinfo_tbl[36].prop_cur_bval
3001#define	ips_ipv6_strict_dst_multihoming	ips_propinfo_tbl[37].prop_cur_uval
3002#define	ips_src_check			ips_propinfo_tbl[38].prop_cur_uval
3003#define	ips_ipsec_policy_log_interval	ips_propinfo_tbl[39].prop_cur_uval
3004#define	ips_pim_accept_clear_messages	ips_propinfo_tbl[40].prop_cur_bval
3005#define	ips_ip_ndp_unsolicit_interval	ips_propinfo_tbl[41].prop_cur_uval
3006#define	ips_ip_ndp_unsolicit_count	ips_propinfo_tbl[42].prop_cur_uval
3007#define	ips_ipv6_ignore_home_address_opt ips_propinfo_tbl[43].prop_cur_bval
3008
3009/* Misc IP configuration knobs */
3010#define	ips_ip_policy_mask		ips_propinfo_tbl[44].prop_cur_uval
3011#define	ips_ip_ecmp_behavior		ips_propinfo_tbl[45].prop_cur_uval
3012#define	ips_ip_multirt_ttl		ips_propinfo_tbl[46].prop_cur_uval
3013#define	ips_ip_ire_badcnt_lifetime	ips_propinfo_tbl[47].prop_cur_uval
3014#define	ips_ip_max_temp_idle		ips_propinfo_tbl[48].prop_cur_uval
3015#define	ips_ip_max_temp_defend		ips_propinfo_tbl[49].prop_cur_uval
3016#define	ips_ip_max_defend		ips_propinfo_tbl[50].prop_cur_uval
3017#define	ips_ip_defend_interval		ips_propinfo_tbl[51].prop_cur_uval
3018#define	ips_ip_dup_recovery		ips_propinfo_tbl[52].prop_cur_uval
3019#define	ips_ip_restrict_interzone_loopback ips_propinfo_tbl[53].prop_cur_bval
3020#define	ips_ip_lso_outbound		ips_propinfo_tbl[54].prop_cur_bval
3021#define	ips_igmp_max_version		ips_propinfo_tbl[55].prop_cur_uval
3022#define	ips_mld_max_version		ips_propinfo_tbl[56].prop_cur_uval
3023#define	ips_ip_forwarding		ips_propinfo_tbl[57].prop_cur_bval
3024#define	ips_ipv6_forwarding		ips_propinfo_tbl[58].prop_cur_bval
3025#define	ips_ip_reassembly_timeout	ips_propinfo_tbl[59].prop_cur_uval
3026#define	ips_ipv6_reassembly_timeout	ips_propinfo_tbl[60].prop_cur_uval
3027#define	ips_ip_cgtp_filter		ips_propinfo_tbl[61].prop_cur_bval
3028#define	ips_arp_probe_delay		ips_propinfo_tbl[62].prop_cur_uval
3029#define	ips_arp_fastprobe_delay		ips_propinfo_tbl[63].prop_cur_uval
3030#define	ips_arp_probe_interval		ips_propinfo_tbl[64].prop_cur_uval
3031#define	ips_arp_fastprobe_interval	ips_propinfo_tbl[65].prop_cur_uval
3032#define	ips_arp_probe_count		ips_propinfo_tbl[66].prop_cur_uval
3033#define	ips_arp_fastprobe_count		ips_propinfo_tbl[67].prop_cur_uval
3034#define	ips_ipv4_dad_announce_interval	ips_propinfo_tbl[68].prop_cur_uval
3035#define	ips_ipv6_dad_announce_interval	ips_propinfo_tbl[69].prop_cur_uval
3036#define	ips_arp_defend_interval		ips_propinfo_tbl[70].prop_cur_uval
3037#define	ips_arp_defend_rate		ips_propinfo_tbl[71].prop_cur_uval
3038#define	ips_ndp_defend_interval		ips_propinfo_tbl[72].prop_cur_uval
3039#define	ips_ndp_defend_rate		ips_propinfo_tbl[73].prop_cur_uval
3040#define	ips_arp_defend_period		ips_propinfo_tbl[74].prop_cur_uval
3041#define	ips_ndp_defend_period		ips_propinfo_tbl[75].prop_cur_uval
3042#define	ips_ipv4_icmp_return_pmtu	ips_propinfo_tbl[76].prop_cur_bval
3043#define	ips_ipv6_icmp_return_pmtu	ips_propinfo_tbl[77].prop_cur_bval
3044#define	ips_ip_arp_publish_count	ips_propinfo_tbl[78].prop_cur_uval
3045#define	ips_ip_arp_publish_interval	ips_propinfo_tbl[79].prop_cur_uval
3046#define	ips_ip_strict_src_multihoming	ips_propinfo_tbl[80].prop_cur_uval
3047#define	ips_ipv6_strict_src_multihoming	ips_propinfo_tbl[81].prop_cur_uval
3048#define	ips_ipv6_drop_inbound_icmpv6	ips_propinfo_tbl[82].prop_cur_bval
3049#define	ips_ip_dce_reclaim_threshold	ips_propinfo_tbl[83].prop_cur_uval
3050
3051extern int	dohwcksum;	/* use h/w cksum if supported by the h/w */
3052#ifdef ZC_TEST
3053extern int	noswcksum;
3054#endif
3055
3056extern char	ipif_loopback_name[];
3057
3058extern nv_t	*ire_nv_tbl;
3059
3060extern struct module_info ip_mod_info;
3061
3062#define	HOOKS4_INTERESTED_PHYSICAL_IN(ipst)	\
3063	((ipst)->ips_ip4_physical_in_event.he_interested)
3064#define	HOOKS6_INTERESTED_PHYSICAL_IN(ipst)	\
3065	((ipst)->ips_ip6_physical_in_event.he_interested)
3066#define	HOOKS4_INTERESTED_PHYSICAL_OUT(ipst)	\
3067	((ipst)->ips_ip4_physical_out_event.he_interested)
3068#define	HOOKS6_INTERESTED_PHYSICAL_OUT(ipst)	\
3069	((ipst)->ips_ip6_physical_out_event.he_interested)
3070#define	HOOKS4_INTERESTED_FORWARDING(ipst)	\
3071	((ipst)->ips_ip4_forwarding_event.he_interested)
3072#define	HOOKS6_INTERESTED_FORWARDING(ipst)	\
3073	((ipst)->ips_ip6_forwarding_event.he_interested)
3074#define	HOOKS4_INTERESTED_LOOPBACK_IN(ipst)	\
3075	((ipst)->ips_ip4_loopback_in_event.he_interested)
3076#define	HOOKS6_INTERESTED_LOOPBACK_IN(ipst)	\
3077	((ipst)->ips_ip6_loopback_in_event.he_interested)
3078#define	HOOKS4_INTERESTED_LOOPBACK_OUT(ipst)	\
3079	((ipst)->ips_ip4_loopback_out_event.he_interested)
3080#define	HOOKS6_INTERESTED_LOOPBACK_OUT(ipst)	\
3081	((ipst)->ips_ip6_loopback_out_event.he_interested)
3082/*
3083 * Hooks marcos used inside of ip
3084 * The callers use the above INTERESTED macros first, hence
3085 * the he_interested check is superflous.
3086 */
3087#define	FW_HOOKS(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst, _err) \
3088	if ((_hook).he_interested) {					\
3089		hook_pkt_event_t info;					\
3090									\
3091		_NOTE(CONSTCOND)					\
3092		ASSERT((_ilp != NULL) || (_olp != NULL));		\
3093									\
3094		FW_SET_ILL_INDEX(info.hpe_ifp, (ill_t *)_ilp);		\
3095		FW_SET_ILL_INDEX(info.hpe_ofp, (ill_t *)_olp);		\
3096		info.hpe_protocol = ipst->ips_ipv4_net_data;		\
3097		info.hpe_hdr = _iph;					\
3098		info.hpe_mp = &(_fm);					\
3099		info.hpe_mb = _m;					\
3100		info.hpe_flags = _llm;					\
3101		_err = hook_run(ipst->ips_ipv4_net_data->netd_hooks,	\
3102		    _event, (hook_data_t)&info);			\
3103		if (_err != 0) {					\
3104			ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
3105			    (_hook).he_name, (void *)_fm, (void *)_m));	\
3106			if (_fm != NULL) {				\
3107				freemsg(_fm);				\
3108				_fm = NULL;				\
3109			}						\
3110			_iph = NULL;					\
3111			_m = NULL;					\
3112		} else {						\
3113			_iph = info.hpe_hdr;				\
3114			_m = info.hpe_mb;				\
3115		}							\
3116	}
3117
3118#define	FW_HOOKS6(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst, _err) \
3119	if ((_hook).he_interested) {					\
3120		hook_pkt_event_t info;					\
3121									\
3122		_NOTE(CONSTCOND)					\
3123		ASSERT((_ilp != NULL) || (_olp != NULL));		\
3124									\
3125		FW_SET_ILL_INDEX(info.hpe_ifp, (ill_t *)_ilp);		\
3126		FW_SET_ILL_INDEX(info.hpe_ofp, (ill_t *)_olp);		\
3127		info.hpe_protocol = ipst->ips_ipv6_net_data;		\
3128		info.hpe_hdr = _iph;					\
3129		info.hpe_mp = &(_fm);					\
3130		info.hpe_mb = _m;					\
3131		info.hpe_flags = _llm;					\
3132		_err = hook_run(ipst->ips_ipv6_net_data->netd_hooks,	\
3133		    _event, (hook_data_t)&info);			\
3134		if (_err != 0) {					\
3135			ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
3136			    (_hook).he_name, (void *)_fm, (void *)_m));	\
3137			if (_fm != NULL) {				\
3138				freemsg(_fm);				\
3139				_fm = NULL;				\
3140			}						\
3141			_iph = NULL;					\
3142			_m = NULL;					\
3143		} else {						\
3144			_iph = info.hpe_hdr;				\
3145			_m = info.hpe_mb;				\
3146		}							\
3147	}
3148
3149#define	FW_SET_ILL_INDEX(fp, ill)					\
3150	_NOTE(CONSTCOND)						\
3151	if ((ill) == NULL || (ill)->ill_phyint == NULL) {		\
3152		(fp) = 0;						\
3153		_NOTE(CONSTCOND)					\
3154	} else if (IS_UNDER_IPMP(ill)) {				\
3155		(fp) = ipmp_ill_get_ipmp_ifindex(ill);			\
3156	} else {							\
3157		(fp) = (ill)->ill_phyint->phyint_ifindex;		\
3158	}
3159
3160/*
3161 * Network byte order macros
3162 */
3163#ifdef	_BIG_ENDIAN
3164#define	N_IN_CLASSA_NET		IN_CLASSA_NET
3165#define	N_IN_CLASSD_NET		IN_CLASSD_NET
3166#define	N_INADDR_UNSPEC_GROUP	INADDR_UNSPEC_GROUP
3167#define	N_IN_LOOPBACK_NET	(ipaddr_t)0x7f000000U
3168#else /* _BIG_ENDIAN */
3169#define	N_IN_CLASSA_NET		(ipaddr_t)0x000000ffU
3170#define	N_IN_CLASSD_NET		(ipaddr_t)0x000000f0U
3171#define	N_INADDR_UNSPEC_GROUP	(ipaddr_t)0x000000e0U
3172#define	N_IN_LOOPBACK_NET	(ipaddr_t)0x0000007fU
3173#endif /* _BIG_ENDIAN */
3174#define	CLASSD(addr)	(((addr) & N_IN_CLASSD_NET) == N_INADDR_UNSPEC_GROUP)
3175#define	CLASSE(addr)	(((addr) & N_IN_CLASSD_NET) == N_IN_CLASSD_NET)
3176#define	IP_LOOPBACK_ADDR(addr)			\
3177	(((addr) & N_IN_CLASSA_NET == N_IN_LOOPBACK_NET))
3178
3179extern int	ip_debug;
3180extern uint_t	ip_thread_data;
3181extern krwlock_t ip_thread_rwlock;
3182extern list_t	ip_thread_list;
3183
3184#ifdef IP_DEBUG
3185#include <sys/debug.h>
3186#include <sys/promif.h>
3187
3188#define	ip0dbg(a)	printf a
3189#define	ip1dbg(a)	if (ip_debug > 2) printf a
3190#define	ip2dbg(a)	if (ip_debug > 3) printf a
3191#define	ip3dbg(a)	if (ip_debug > 4) printf a
3192#else
3193#define	ip0dbg(a)	/* */
3194#define	ip1dbg(a)	/* */
3195#define	ip2dbg(a)	/* */
3196#define	ip3dbg(a)	/* */
3197#endif	/* IP_DEBUG */
3198
3199/* Default MAC-layer address string length for mac_colon_addr */
3200#define	MAC_STR_LEN	128
3201
3202struct	mac_header_info_s;
3203
3204extern void	ill_frag_timer(void *);
3205extern ill_t	*ill_first(int, int, ill_walk_context_t *, ip_stack_t *);
3206extern ill_t	*ill_next(ill_walk_context_t *, ill_t *);
3207extern void	ill_frag_timer_start(ill_t *);
3208extern void	ill_nic_event_dispatch(ill_t *, lif_if_t, nic_event_t,
3209    nic_event_data_t, size_t);
3210extern mblk_t	*ip_carve_mp(mblk_t **, ssize_t);
3211extern mblk_t	*ip_dlpi_alloc(size_t, t_uscalar_t);
3212extern mblk_t	*ip_dlnotify_alloc(uint_t, uint_t);
3213extern mblk_t	*ip_dlnotify_alloc2(uint_t, uint_t, uint_t);
3214extern char	*ip_dot_addr(ipaddr_t, char *);
3215extern const char *mac_colon_addr(const uint8_t *, size_t, char *, size_t);
3216extern int	ip_lwput(queue_t *, mblk_t *);
3217extern boolean_t icmp_err_rate_limit(ip_stack_t *);
3218extern void	icmp_frag_needed(mblk_t *, int, ip_recv_attr_t *);
3219extern mblk_t	*icmp_inbound_v4(mblk_t *, ip_recv_attr_t *);
3220extern void	icmp_time_exceeded(mblk_t *, uint8_t, ip_recv_attr_t *);
3221extern void	icmp_unreachable(mblk_t *, uint8_t, ip_recv_attr_t *);
3222extern boolean_t ip_ipsec_policy_inherit(conn_t *, conn_t *, ip_recv_attr_t *);
3223extern void	*ip_pullup(mblk_t *, ssize_t, ip_recv_attr_t *);
3224extern void	ip_setl2src(mblk_t *, ip_recv_attr_t *, ill_t *);
3225extern mblk_t	*ip_check_and_align_header(mblk_t *, uint_t, ip_recv_attr_t *);
3226extern mblk_t	*ip_check_length(mblk_t *, uchar_t *, ssize_t, uint_t, uint_t,
3227    ip_recv_attr_t *);
3228extern mblk_t	*ip_check_optlen(mblk_t *, ipha_t *, uint_t, uint_t,
3229    ip_recv_attr_t *);
3230extern mblk_t	*ip_fix_dbref(mblk_t *, ip_recv_attr_t *);
3231extern uint_t	ip_cksum(mblk_t *, int, uint32_t);
3232extern int	ip_close(queue_t *, int, cred_t *);
3233extern uint16_t	ip_csum_hdr(ipha_t *);
3234extern void	ip_forward_xmit_v4(nce_t *, ill_t *, mblk_t *, ipha_t *,
3235    ip_recv_attr_t *, uint32_t, uint32_t);
3236extern boolean_t ip_forward_options(mblk_t *, ipha_t *, ill_t *,
3237    ip_recv_attr_t *);
3238extern int	ip_fragment_v4(mblk_t *, nce_t *, iaflags_t, uint_t, uint32_t,
3239    uint32_t, zoneid_t, zoneid_t, pfirepostfrag_t postfragfn,
3240    uintptr_t *cookie);
3241extern void	ip_proto_not_sup(mblk_t *, ip_recv_attr_t *);
3242extern void	ip_ire_g_fini(void);
3243extern void	ip_ire_g_init(void);
3244extern void	ip_ire_fini(ip_stack_t *);
3245extern void	ip_ire_init(ip_stack_t *);
3246extern void	ip_mdata_to_mhi(ill_t *, mblk_t *, struct mac_header_info_s *);
3247extern int	ip_openv4(queue_t *q, dev_t *devp, int flag, int sflag,
3248		    cred_t *credp);
3249extern int	ip_openv6(queue_t *q, dev_t *devp, int flag, int sflag,
3250		    cred_t *credp);
3251extern int	ip_reassemble(mblk_t *, ipf_t *, uint_t, boolean_t, ill_t *,
3252    size_t);
3253extern int	ip_rput(queue_t *, mblk_t *);
3254extern void	ip_input(ill_t *, ill_rx_ring_t *, mblk_t *,
3255    struct mac_header_info_s *);
3256extern void	ip_input_v6(ill_t *, ill_rx_ring_t *, mblk_t *,
3257    struct mac_header_info_s *);
3258extern mblk_t	*ip_input_common_v4(ill_t *, ill_rx_ring_t *, mblk_t *,
3259    struct mac_header_info_s *, squeue_t *, mblk_t **, uint_t *);
3260extern mblk_t	*ip_input_common_v6(ill_t *, ill_rx_ring_t *, mblk_t *,
3261    struct mac_header_info_s *, squeue_t *, mblk_t **, uint_t *);
3262extern void	ill_input_full_v4(mblk_t *, void *, void *,
3263    ip_recv_attr_t *, rtc_t *);
3264extern void	ill_input_short_v4(mblk_t *, void *, void *,
3265    ip_recv_attr_t *, rtc_t *);
3266extern void	ill_input_full_v6(mblk_t *, void *, void *,
3267    ip_recv_attr_t *, rtc_t *);
3268extern void	ill_input_short_v6(mblk_t *, void *, void *,
3269    ip_recv_attr_t *, rtc_t *);
3270extern ipaddr_t	ip_input_options(ipha_t *, ipaddr_t, mblk_t *,
3271    ip_recv_attr_t *, int *);
3272extern boolean_t ip_input_local_options(mblk_t *, ipha_t *, ip_recv_attr_t *);
3273extern mblk_t	*ip_input_fragment(mblk_t *, ipha_t *, ip_recv_attr_t *);
3274extern mblk_t	*ip_input_fragment_v6(mblk_t *, ip6_t *, ip6_frag_t *, uint_t,
3275    ip_recv_attr_t *);
3276extern void	ip_input_post_ipsec(mblk_t *, ip_recv_attr_t *);
3277extern void	ip_fanout_v4(mblk_t *, ipha_t *, ip_recv_attr_t *);
3278extern void	ip_fanout_v6(mblk_t *, ip6_t *, ip_recv_attr_t *);
3279extern void	ip_fanout_proto_conn(conn_t *, mblk_t *, ipha_t *, ip6_t *,
3280    ip_recv_attr_t *);
3281extern void	ip_fanout_proto_v4(mblk_t *, ipha_t *, ip_recv_attr_t *);
3282extern void	ip_fanout_send_icmp_v4(mblk_t *, uint_t, uint_t,
3283    ip_recv_attr_t *);
3284extern void	ip_fanout_udp_conn(conn_t *, mblk_t *, ipha_t *, ip6_t *,
3285    ip_recv_attr_t *);
3286extern void	ip_fanout_udp_multi_v4(mblk_t *, ipha_t *, uint16_t, uint16_t,
3287    ip_recv_attr_t *);
3288extern mblk_t	*zero_spi_check(mblk_t *, ip_recv_attr_t *);
3289extern void	ip_build_hdrs_v4(uchar_t *, uint_t, const ip_pkt_t *, uint8_t);
3290extern int	ip_find_hdr_v4(ipha_t *, ip_pkt_t *, boolean_t);
3291extern int	ip_total_hdrs_len_v4(const ip_pkt_t *);
3292
3293extern mblk_t	*ip_accept_tcp(ill_t *, ill_rx_ring_t *, squeue_t *,
3294    mblk_t *, mblk_t **, uint_t *cnt);
3295extern void	ip_rput_dlpi(ill_t *, mblk_t *);
3296extern void	ip_rput_notdata(ill_t *, mblk_t *);
3297
3298extern void	ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *,
3299		    mib2_ipIfStatsEntry_t *);
3300extern void	ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *,
3301		    mib2_ipv6IfIcmpEntry_t *);
3302extern void	ip_rput_other(ipsq_t *, queue_t *, mblk_t *, void *);
3303extern ire_t	*ip_check_multihome(void *, ire_t *, ill_t *);
3304extern void	ip_send_potential_redirect_v4(mblk_t *, ipha_t *, ire_t *,
3305    ip_recv_attr_t *);
3306extern int	ip_set_destination_v4(ipaddr_t *, ipaddr_t, ipaddr_t,
3307    ip_xmit_attr_t *, iulp_t *, uint32_t, uint_t);
3308extern int	ip_set_destination_v6(in6_addr_t *, const in6_addr_t *,
3309    const in6_addr_t *, ip_xmit_attr_t *, iulp_t *, uint32_t, uint_t);
3310
3311extern int	ip_output_simple(mblk_t *, ip_xmit_attr_t *);
3312extern int	ip_output_simple_v4(mblk_t *, ip_xmit_attr_t *);
3313extern int	ip_output_simple_v6(mblk_t *, ip_xmit_attr_t *);
3314extern int	ip_output_options(mblk_t *, ipha_t *, ip_xmit_attr_t *,
3315    ill_t *);
3316extern void	ip_output_local_options(ipha_t *, ip_stack_t *);
3317
3318extern ip_xmit_attr_t *conn_get_ixa(conn_t *, boolean_t);
3319extern ip_xmit_attr_t *conn_get_ixa_tryhard(conn_t *, boolean_t);
3320extern ip_xmit_attr_t *conn_replace_ixa(conn_t *, ip_xmit_attr_t *);
3321extern ip_xmit_attr_t *conn_get_ixa_exclusive(conn_t *);
3322extern ip_xmit_attr_t *ip_xmit_attr_duplicate(ip_xmit_attr_t *);
3323extern void	ip_xmit_attr_replace_tsl(ip_xmit_attr_t *, ts_label_t *);
3324extern void	ip_xmit_attr_restore_tsl(ip_xmit_attr_t *, cred_t *);
3325boolean_t	ip_recv_attr_replace_label(ip_recv_attr_t *, ts_label_t *);
3326extern void	ixa_inactive(ip_xmit_attr_t *);
3327extern void	ixa_refrele(ip_xmit_attr_t *);
3328extern boolean_t ixa_check_drain_insert(conn_t *, ip_xmit_attr_t *);
3329extern void	ixa_cleanup(ip_xmit_attr_t *);
3330extern void	ira_cleanup(ip_recv_attr_t *, boolean_t);
3331extern void	ixa_safe_copy(ip_xmit_attr_t *, ip_xmit_attr_t *);
3332
3333extern int	conn_ip_output(mblk_t *, ip_xmit_attr_t *);
3334extern boolean_t ip_output_verify_local(ip_xmit_attr_t *);
3335extern mblk_t	*ip_output_process_local(mblk_t *, ip_xmit_attr_t *, boolean_t,
3336    boolean_t, conn_t *);
3337
3338extern int	conn_opt_get(conn_opt_arg_t *, t_scalar_t, t_scalar_t,
3339    uchar_t *);
3340extern int	conn_opt_set(conn_opt_arg_t *, t_scalar_t, t_scalar_t, uint_t,
3341    uchar_t *, boolean_t, cred_t *);
3342extern boolean_t	conn_same_as_last_v4(conn_t *, sin_t *);
3343extern boolean_t	conn_same_as_last_v6(conn_t *, sin6_t *);
3344extern int	conn_update_label(const conn_t *, const ip_xmit_attr_t *,
3345    const in6_addr_t *, ip_pkt_t *);
3346
3347extern int	ip_opt_set_multicast_group(conn_t *, t_scalar_t,
3348    uchar_t *, boolean_t, boolean_t);
3349extern int	ip_opt_set_multicast_sources(conn_t *, t_scalar_t,
3350    uchar_t *, boolean_t, boolean_t);
3351extern int	conn_getsockname(conn_t *, struct sockaddr *, uint_t *);
3352extern int	conn_getpeername(conn_t *, struct sockaddr *, uint_t *);
3353
3354extern int	conn_build_hdr_template(conn_t *, uint_t, uint_t,
3355    const in6_addr_t *, const in6_addr_t *, uint32_t);
3356extern mblk_t	*conn_prepend_hdr(ip_xmit_attr_t *, const ip_pkt_t *,
3357    const in6_addr_t *, const in6_addr_t *, uint8_t, uint32_t, uint_t,
3358    mblk_t *, uint_t, uint_t, uint32_t *, int *);
3359extern void	ip_attr_newdst(ip_xmit_attr_t *);
3360extern void	ip_attr_nexthop(const ip_pkt_t *, const ip_xmit_attr_t *,
3361    const in6_addr_t *, in6_addr_t *);
3362extern int	conn_connect(conn_t *, iulp_t *, uint32_t);
3363extern int	ip_attr_connect(const conn_t *, ip_xmit_attr_t *,
3364    const in6_addr_t *, const in6_addr_t *, const in6_addr_t *, in_port_t,
3365    in6_addr_t *, iulp_t *, uint32_t);
3366extern int	conn_inherit_parent(conn_t *, conn_t *);
3367
3368extern void	conn_ixa_cleanup(conn_t *connp, void *arg);
3369
3370extern boolean_t conn_wantpacket(conn_t *, ip_recv_attr_t *, ipha_t *);
3371extern uint_t	ip_type_v4(ipaddr_t, ip_stack_t *);
3372extern uint_t	ip_type_v6(const in6_addr_t *, ip_stack_t *);
3373
3374extern int	ip_wput_nondata(queue_t *, mblk_t *);
3375extern int	ip_wsrv(queue_t *);
3376extern char	*ip_nv_lookup(nv_t *, int);
3377extern boolean_t ip_local_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3378extern boolean_t ip_remote_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3379extern ipaddr_t ip_massage_options(ipha_t *, netstack_t *);
3380extern ipaddr_t ip_net_mask(ipaddr_t);
3381extern void	arp_bringup_done(ill_t *, int);
3382extern void	arp_replumb_done(ill_t *, int);
3383
3384extern struct qinit iprinitv6;
3385
3386extern void	ipmp_init(ip_stack_t *);
3387extern void	ipmp_destroy(ip_stack_t *);
3388extern ipmp_grp_t *ipmp_grp_create(const char *, phyint_t *);
3389extern void	ipmp_grp_destroy(ipmp_grp_t *);
3390extern void	ipmp_grp_info(const ipmp_grp_t *, lifgroupinfo_t *);
3391extern int	ipmp_grp_rename(ipmp_grp_t *, const char *);
3392extern ipmp_grp_t *ipmp_grp_lookup(const char *, ip_stack_t *);
3393extern int	ipmp_grp_vet_phyint(ipmp_grp_t *, phyint_t *);
3394extern ipmp_illgrp_t *ipmp_illgrp_create(ill_t *);
3395extern void	ipmp_illgrp_destroy(ipmp_illgrp_t *);
3396extern ill_t	*ipmp_illgrp_add_ipif(ipmp_illgrp_t *, ipif_t *);
3397extern void	ipmp_illgrp_del_ipif(ipmp_illgrp_t *, ipif_t *);
3398extern ill_t	*ipmp_illgrp_next_ill(ipmp_illgrp_t *);
3399extern ill_t	*ipmp_illgrp_hold_next_ill(ipmp_illgrp_t *);
3400extern ill_t	*ipmp_illgrp_hold_cast_ill(ipmp_illgrp_t *);
3401extern ill_t	*ipmp_illgrp_ipmp_ill(ipmp_illgrp_t *);
3402extern void	ipmp_illgrp_refresh_mtu(ipmp_illgrp_t *);
3403extern ipmp_arpent_t *ipmp_illgrp_create_arpent(ipmp_illgrp_t *,
3404    boolean_t, ipaddr_t, uchar_t *, size_t, uint16_t);
3405extern void	ipmp_illgrp_destroy_arpent(ipmp_illgrp_t *, ipmp_arpent_t *);
3406extern ipmp_arpent_t *ipmp_illgrp_lookup_arpent(ipmp_illgrp_t *, ipaddr_t *);
3407extern void	ipmp_illgrp_refresh_arpent(ipmp_illgrp_t *);
3408extern void	ipmp_illgrp_mark_arpent(ipmp_illgrp_t *, ipmp_arpent_t *);
3409extern ill_t	*ipmp_illgrp_find_ill(ipmp_illgrp_t *, uchar_t *, uint_t);
3410extern void	ipmp_illgrp_link_grp(ipmp_illgrp_t *, ipmp_grp_t *);
3411extern int	ipmp_illgrp_unlink_grp(ipmp_illgrp_t *);
3412extern uint_t	ipmp_ill_get_ipmp_ifindex(const ill_t *);
3413extern void	ipmp_ill_join_illgrp(ill_t *, ipmp_illgrp_t *);
3414extern void	ipmp_ill_leave_illgrp(ill_t *);
3415extern ill_t	*ipmp_ill_hold_ipmp_ill(ill_t *);
3416extern ill_t	*ipmp_ill_hold_xmit_ill(ill_t *, boolean_t);
3417extern boolean_t ipmp_ill_is_active(ill_t *);
3418extern void	ipmp_ill_refresh_active(ill_t *);
3419extern void	ipmp_phyint_join_grp(phyint_t *, ipmp_grp_t *);
3420extern void	ipmp_phyint_leave_grp(phyint_t *);
3421extern void	ipmp_phyint_refresh_active(phyint_t *);
3422extern ill_t	*ipmp_ipif_bound_ill(const ipif_t *);
3423extern ill_t	*ipmp_ipif_hold_bound_ill(const ipif_t *);
3424extern boolean_t ipmp_ipif_is_dataaddr(const ipif_t *);
3425extern boolean_t ipmp_ipif_is_stubaddr(const ipif_t *);
3426extern boolean_t ipmp_packet_is_probe(mblk_t *, ill_t *);
3427extern void	ipmp_ncec_delete_nce(ncec_t *);
3428extern void	ipmp_ncec_refresh_nce(ncec_t *);
3429
3430extern void	conn_drain_insert(conn_t *, idl_tx_list_t *);
3431extern void	conn_setqfull(conn_t *, boolean_t *);
3432extern void	conn_clrqfull(conn_t *, boolean_t *);
3433extern int	conn_ipsec_length(conn_t *);
3434extern ipaddr_t	ip_get_dst(ipha_t *);
3435extern uint_t	ip_get_pmtu(ip_xmit_attr_t *);
3436extern uint_t	ip_get_base_mtu(ill_t *, ire_t *);
3437extern mblk_t *ip_output_attach_policy(mblk_t *, ipha_t *, ip6_t *,
3438    const conn_t *, ip_xmit_attr_t *);
3439extern int	ipsec_out_extra_length(ip_xmit_attr_t *);
3440extern int	ipsec_out_process(mblk_t *, ip_xmit_attr_t *);
3441extern int	ip_output_post_ipsec(mblk_t *, ip_xmit_attr_t *);
3442extern void	ipsec_out_to_in(ip_xmit_attr_t *, ill_t *ill,
3443    ip_recv_attr_t *);
3444
3445extern void	ire_cleanup(ire_t *);
3446extern void	ire_inactive(ire_t *);
3447extern boolean_t irb_inactive(irb_t *);
3448extern ire_t	*ire_unlink(irb_t *);
3449
3450#ifdef DEBUG
3451extern	boolean_t th_trace_ref(const void *, ip_stack_t *);
3452extern	void	th_trace_unref(const void *);
3453extern	void	th_trace_cleanup(const void *, boolean_t);
3454extern	void	ire_trace_ref(ire_t *);
3455extern	void	ire_untrace_ref(ire_t *);
3456#endif
3457
3458extern int	ip_srcid_insert(const in6_addr_t *, zoneid_t, ip_stack_t *);
3459extern int	ip_srcid_remove(const in6_addr_t *, zoneid_t, ip_stack_t *);
3460extern boolean_t ip_srcid_find_id(uint_t, in6_addr_t *, zoneid_t, boolean_t,
3461    netstack_t *);
3462extern uint_t	ip_srcid_find_addr(const in6_addr_t *, zoneid_t, netstack_t *);
3463
3464extern uint8_t	ipoptp_next(ipoptp_t *);
3465extern uint8_t	ipoptp_first(ipoptp_t *, ipha_t *);
3466extern int	ip_opt_get_user(conn_t *, uchar_t *);
3467extern int	ipsec_req_from_conn(conn_t *, ipsec_req_t *, int);
3468extern int	ip_snmp_get(queue_t *q, mblk_t *mctl, int level, boolean_t);
3469extern int	ip_snmp_set(queue_t *q, int, int, uchar_t *, int);
3470extern void	ip_process_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3471extern void	ip_quiesce_conn(conn_t *);
3472extern  void    ip_reprocess_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3473extern void	ip_ioctl_finish(queue_t *, mblk_t *, int, int, ipsq_t *);
3474
3475extern boolean_t ip_cmpbuf(const void *, uint_t, boolean_t, const void *,
3476    uint_t);
3477extern boolean_t ip_allocbuf(void **, uint_t *, boolean_t, const void *,
3478    uint_t);
3479extern void	ip_savebuf(void **, uint_t *, boolean_t, const void *, uint_t);
3480
3481extern boolean_t	ipsq_pending_mp_cleanup(ill_t *, conn_t *);
3482extern void	conn_ioctl_cleanup(conn_t *);
3483
3484extern void	ip_unbind(conn_t *);
3485
3486extern void tnet_init(void);
3487extern void tnet_fini(void);
3488
3489/*
3490 * Hook functions to enable cluster networking
3491 * On non-clustered systems these vectors must always be NULL.
3492 */
3493extern int (*cl_inet_isclusterwide)(netstackid_t stack_id, uint8_t protocol,
3494    sa_family_t addr_family, uint8_t *laddrp, void *args);
3495extern uint32_t (*cl_inet_ipident)(netstackid_t stack_id, uint8_t protocol,
3496    sa_family_t addr_family, uint8_t *laddrp, uint8_t *faddrp,
3497    void *args);
3498extern int (*cl_inet_connect2)(netstackid_t stack_id, uint8_t protocol,
3499    boolean_t is_outgoing, sa_family_t addr_family, uint8_t *laddrp,
3500    in_port_t lport, uint8_t *faddrp, in_port_t fport, void *args);
3501extern void (*cl_inet_getspi)(netstackid_t, uint8_t, uint8_t *, size_t,
3502    void *);
3503extern void (*cl_inet_getspi)(netstackid_t stack_id, uint8_t protocol,
3504    uint8_t *ptr, size_t len, void *args);
3505extern int (*cl_inet_checkspi)(netstackid_t stack_id, uint8_t protocol,
3506    uint32_t spi, void *args);
3507extern void (*cl_inet_deletespi)(netstackid_t stack_id, uint8_t protocol,
3508    uint32_t spi, void *args);
3509extern void (*cl_inet_idlesa)(netstackid_t, uint8_t, uint32_t,
3510    sa_family_t, in6_addr_t, in6_addr_t, void *);
3511
3512
3513/* Hooks for CGTP (multirt routes) filtering module */
3514#define	CGTP_FILTER_REV_1	1
3515#define	CGTP_FILTER_REV_2	2
3516#define	CGTP_FILTER_REV_3	3
3517#define	CGTP_FILTER_REV		CGTP_FILTER_REV_3
3518
3519/* cfo_filter and cfo_filter_v6 hooks return values */
3520#define	CGTP_IP_PKT_NOT_CGTP	0
3521#define	CGTP_IP_PKT_PREMIUM	1
3522#define	CGTP_IP_PKT_DUPLICATE	2
3523
3524/* Version 3 of the filter interface */
3525typedef struct cgtp_filter_ops {
3526	int	cfo_filter_rev;			/* CGTP_FILTER_REV_3 */
3527	int	(*cfo_change_state)(netstackid_t, int);
3528	int	(*cfo_add_dest_v4)(netstackid_t, ipaddr_t, ipaddr_t,
3529		    ipaddr_t, ipaddr_t);
3530	int	(*cfo_del_dest_v4)(netstackid_t, ipaddr_t, ipaddr_t);
3531	int	(*cfo_add_dest_v6)(netstackid_t, in6_addr_t *, in6_addr_t *,
3532		    in6_addr_t *, in6_addr_t *);
3533	int	(*cfo_del_dest_v6)(netstackid_t, in6_addr_t *, in6_addr_t *);
3534	int	(*cfo_filter)(netstackid_t, uint_t, mblk_t *);
3535	int	(*cfo_filter_v6)(netstackid_t, uint_t, ip6_t *,
3536		    ip6_frag_t *);
3537} cgtp_filter_ops_t;
3538
3539#define	CGTP_MCAST_SUCCESS	1
3540
3541/*
3542 * The separate CGTP module needs this global symbol so that it
3543 * can check the version and determine whether to use the old or the new
3544 * version of the filtering interface.
3545 */
3546extern int	ip_cgtp_filter_rev;
3547
3548extern int	ip_cgtp_filter_supported(void);
3549extern int	ip_cgtp_filter_register(netstackid_t, cgtp_filter_ops_t *);
3550extern int	ip_cgtp_filter_unregister(netstackid_t);
3551extern int	ip_cgtp_filter_is_registered(netstackid_t);
3552
3553/*
3554 * rr_ring_state cycles in the order shown below from RR_FREE through
3555 * RR_FREE_IN_PROG and  back to RR_FREE.
3556 */
3557typedef enum {
3558	RR_FREE,			/* Free slot */
3559	RR_SQUEUE_UNBOUND,		/* Ring's squeue is unbound */
3560	RR_SQUEUE_BIND_INPROG,		/* Ring's squeue bind in progress */
3561	RR_SQUEUE_BOUND,		/* Ring's squeue bound to cpu */
3562	RR_FREE_INPROG			/* Ring is being freed */
3563} ip_ring_state_t;
3564
3565#define	ILL_MAX_RINGS		256	/* Max num of rx rings we can manage */
3566#define	ILL_POLLING		0x01	/* Polling in use */
3567
3568/*
3569 * These functions pointer types are exported by the mac/dls layer.
3570 * we need to duplicate the definitions here because we cannot
3571 * include mac/dls header files here.
3572 */
3573typedef boolean_t		(*ip_mac_intr_disable_t)(void *);
3574typedef int			(*ip_mac_intr_enable_t)(void *);
3575typedef ip_mac_tx_cookie_t	(*ip_dld_tx_t)(void *, mblk_t *,
3576    uint64_t, uint16_t);
3577typedef	void			(*ip_flow_enable_t)(void *, ip_mac_tx_cookie_t);
3578typedef void			*(*ip_dld_callb_t)(void *,
3579    ip_flow_enable_t, void *);
3580typedef boolean_t		(*ip_dld_fctl_t)(void *, ip_mac_tx_cookie_t);
3581typedef int			(*ip_capab_func_t)(void *, uint_t,
3582    void *, uint_t);
3583
3584/*
3585 * POLLING README
3586 * sq_get_pkts() is called to pick packets from softring in poll mode. It
3587 * calls rr_rx to get the chain and process it with rr_ip_accept.
3588 * rr_rx = mac_soft_ring_poll() to pick packets
3589 * rr_ip_accept = ip_accept_tcp() to process packets
3590 */
3591
3592/*
3593 * XXX: With protocol, service specific squeues, they will have
3594 * specific acceptor functions.
3595 */
3596typedef	mblk_t *(*ip_mac_rx_t)(void *, size_t);
3597typedef mblk_t *(*ip_accept_t)(ill_t *, ill_rx_ring_t *,
3598    squeue_t *, mblk_t *, mblk_t **, uint_t *);
3599
3600/*
3601 * rr_intr_enable, rr_intr_disable, rr_rx_handle, rr_rx:
3602 * May be accessed while in the squeue AND after checking that SQS_POLL_CAPAB
3603 * is set.
3604 *
3605 * rr_ring_state: Protected by ill_lock.
3606 */
3607struct ill_rx_ring {
3608	ip_mac_intr_disable_t	rr_intr_disable; /* Interrupt disabling func */
3609	ip_mac_intr_enable_t	rr_intr_enable;	/* Interrupt enabling func */
3610	void			*rr_intr_handle; /* Handle interrupt funcs */
3611	ip_mac_rx_t		rr_rx;		/* Driver receive function */
3612	ip_accept_t		rr_ip_accept;	/* IP accept function */
3613	void			*rr_rx_handle;	/* Handle for Rx ring */
3614	squeue_t		*rr_sqp; /* Squeue the ring is bound to */
3615	ill_t			*rr_ill;	/* back pointer to ill */
3616	ip_ring_state_t		rr_ring_state;	/* State of this ring */
3617};
3618
3619/*
3620 * IP - DLD direct function call capability
3621 * Suffixes, df - dld function, dh - dld handle,
3622 * cf - client (IP) function, ch - client handle
3623 */
3624typedef struct ill_dld_direct_s {		/* DLD provided driver Tx */
3625	ip_dld_tx_t		idd_tx_df;	/* str_mdata_fastpath_put */
3626	void			*idd_tx_dh;	/* dld_str_t *dsp */
3627	ip_dld_callb_t		idd_tx_cb_df;	/* mac_tx_srs_notify */
3628	void			*idd_tx_cb_dh;	/* mac_client_handle_t *mch */
3629	ip_dld_fctl_t		idd_tx_fctl_df;	/* mac_tx_is_flow_blocked */
3630	void			*idd_tx_fctl_dh;	/* mac_client_handle */
3631} ill_dld_direct_t;
3632
3633/* IP - DLD polling capability */
3634typedef struct ill_dld_poll_s {
3635	ill_rx_ring_t		idp_ring_tbl[ILL_MAX_RINGS];
3636} ill_dld_poll_t;
3637
3638/* Describes ill->ill_dld_capab */
3639struct ill_dld_capab_s {
3640	ip_capab_func_t		idc_capab_df;	/* dld_capab_func */
3641	void			*idc_capab_dh;	/* dld_str_t *dsp */
3642	ill_dld_direct_t	idc_direct;
3643	ill_dld_poll_t		idc_poll;
3644};
3645
3646/*
3647 * IP squeues exports
3648 */
3649extern boolean_t	ip_squeue_fanout;
3650
3651#define	IP_SQUEUE_GET(hint) ip_squeue_random(hint)
3652
3653extern void ip_squeue_init(void (*)(squeue_t *));
3654extern squeue_t	*ip_squeue_random(uint_t);
3655extern squeue_t *ip_squeue_get(ill_rx_ring_t *);
3656extern squeue_t *ip_squeue_getfree(pri_t);
3657extern int ip_squeue_cpu_move(squeue_t *, processorid_t);
3658extern void *ip_squeue_add_ring(ill_t *, void *);
3659extern void ip_squeue_bind_ring(ill_t *, ill_rx_ring_t *, processorid_t);
3660extern void ip_squeue_clean_ring(ill_t *, ill_rx_ring_t *);
3661extern void ip_squeue_quiesce_ring(ill_t *, ill_rx_ring_t *);
3662extern void ip_squeue_restart_ring(ill_t *, ill_rx_ring_t *);
3663extern void ip_squeue_clean_all(ill_t *);
3664extern boolean_t	ip_source_routed(ipha_t *, ip_stack_t *);
3665
3666extern int tcp_wput(queue_t *, mblk_t *);
3667
3668extern int	ip_fill_mtuinfo(conn_t *, ip_xmit_attr_t *,
3669    struct ip6_mtuinfo *);
3670extern hook_t *ipobs_register_hook(netstack_t *, pfv_t);
3671extern void ipobs_unregister_hook(netstack_t *, hook_t *);
3672extern void ipobs_hook(mblk_t *, int, zoneid_t, zoneid_t, const ill_t *,
3673    ip_stack_t *);
3674typedef void    (*ipsq_func_t)(ipsq_t *, queue_t *, mblk_t *, void *);
3675
3676extern void	dce_g_init(void);
3677extern void	dce_g_destroy(void);
3678extern void	dce_stack_init(ip_stack_t *);
3679extern void	dce_stack_destroy(ip_stack_t *);
3680extern void	dce_cleanup(uint_t, ip_stack_t *);
3681extern dce_t	*dce_get_default(ip_stack_t *);
3682extern dce_t	*dce_lookup_pkt(mblk_t *, ip_xmit_attr_t *, uint_t *);
3683extern dce_t	*dce_lookup_v4(ipaddr_t, ip_stack_t *, uint_t *);
3684extern dce_t	*dce_lookup_v6(const in6_addr_t *, uint_t, ip_stack_t *,
3685    uint_t *);
3686extern dce_t	*dce_lookup_and_add_v4(ipaddr_t, ip_stack_t *);
3687extern dce_t	*dce_lookup_and_add_v6(const in6_addr_t *, uint_t,
3688    ip_stack_t *);
3689extern int	dce_update_uinfo_v4(ipaddr_t, iulp_t *, ip_stack_t *);
3690extern int	dce_update_uinfo_v6(const in6_addr_t *, uint_t, iulp_t *,
3691    ip_stack_t *);
3692extern int	dce_update_uinfo(const in6_addr_t *, uint_t, iulp_t *,
3693    ip_stack_t *);
3694extern void	dce_increment_generation(dce_t *);
3695extern void	dce_increment_all_generations(boolean_t, ip_stack_t *);
3696extern void	dce_refrele(dce_t *);
3697extern void	dce_refhold(dce_t *);
3698extern void	dce_refrele_notr(dce_t *);
3699extern void	dce_refhold_notr(dce_t *);
3700mblk_t		*ip_snmp_get_mib2_ip_dce(queue_t *, mblk_t *, ip_stack_t *ipst);
3701
3702extern ip_laddr_t ip_laddr_verify_v4(ipaddr_t, zoneid_t,
3703    ip_stack_t *, boolean_t);
3704extern ip_laddr_t ip_laddr_verify_v6(const in6_addr_t *, zoneid_t,
3705    ip_stack_t *, boolean_t, uint_t);
3706extern int	ip_laddr_fanout_insert(conn_t *);
3707
3708extern boolean_t ip_verify_src(mblk_t *, ip_xmit_attr_t *, uint_t *);
3709extern int	ip_verify_ire(mblk_t *, ip_xmit_attr_t *);
3710
3711extern mblk_t	*ip_xmit_attr_to_mblk(ip_xmit_attr_t *);
3712extern boolean_t ip_xmit_attr_from_mblk(mblk_t *, ip_xmit_attr_t *);
3713extern mblk_t	*ip_xmit_attr_free_mblk(mblk_t *);
3714extern mblk_t	*ip_recv_attr_to_mblk(ip_recv_attr_t *);
3715extern boolean_t ip_recv_attr_from_mblk(mblk_t *, ip_recv_attr_t *);
3716extern mblk_t	*ip_recv_attr_free_mblk(mblk_t *);
3717extern boolean_t ip_recv_attr_is_mblk(mblk_t *);
3718
3719#ifdef	__PRAGMA_REDEFINE_EXTNAME
3720#pragma	redefine_extname inet_ntop _inet_ntop
3721#pragma	redefine_extname inet_pton _inet_pton
3722#else
3723#define	inet_ntop _inet_ntop
3724#define	inet_pton _inet_pton
3725#endif	/* __PRAGMA_REDEFINE_EXTNAME */
3726
3727extern char	*inet_ntop(int, const void *, char *, int);
3728extern int	inet_pton(int, char *, void *);
3729
3730/*
3731 * Squeue tags. Tags only need to be unique when the callback function is the
3732 * same to distinguish between different calls, but we use unique tags for
3733 * convenience anyway.
3734 */
3735#define	SQTAG_IP_INPUT			1
3736#define	SQTAG_TCP_INPUT_ICMP_ERR	2
3737#define	SQTAG_TCP6_INPUT_ICMP_ERR	3
3738#define	SQTAG_IP_TCP_INPUT		4
3739#define	SQTAG_IP6_TCP_INPUT		5
3740#define	SQTAG_IP_TCP_CLOSE		6
3741#define	SQTAG_TCP_OUTPUT		7
3742#define	SQTAG_TCP_TIMER			8
3743#define	SQTAG_TCP_TIMEWAIT		9
3744#define	SQTAG_TCP_ACCEPT_FINISH		10
3745#define	SQTAG_TCP_ACCEPT_FINISH_Q0	11
3746#define	SQTAG_TCP_ACCEPT_PENDING	12
3747#define	SQTAG_TCP_LISTEN_DISCON		13
3748#define	SQTAG_TCP_CONN_REQ_1		14
3749#define	SQTAG_TCP_EAGER_BLOWOFF		15
3750#define	SQTAG_TCP_EAGER_CLEANUP		16
3751#define	SQTAG_TCP_EAGER_CLEANUP_Q0	17
3752#define	SQTAG_TCP_CONN_IND		18
3753#define	SQTAG_TCP_RSRV			19
3754#define	SQTAG_TCP_ABORT_BUCKET		20
3755#define	SQTAG_TCP_REINPUT		21
3756#define	SQTAG_TCP_REINPUT_EAGER		22
3757#define	SQTAG_TCP_INPUT_MCTL		23
3758#define	SQTAG_TCP_RPUTOTHER		24
3759#define	SQTAG_IP_PROTO_AGAIN		25
3760#define	SQTAG_IP_FANOUT_TCP		26
3761#define	SQTAG_IPSQ_CLEAN_RING		27
3762#define	SQTAG_TCP_WPUT_OTHER		28
3763#define	SQTAG_TCP_CONN_REQ_UNBOUND	29
3764#define	SQTAG_TCP_SEND_PENDING		30
3765#define	SQTAG_BIND_RETRY		31
3766#define	SQTAG_UDP_FANOUT		32
3767#define	SQTAG_UDP_INPUT			33
3768#define	SQTAG_UDP_WPUT			34
3769#define	SQTAG_UDP_OUTPUT		35
3770#define	SQTAG_TCP_KSSL_INPUT		36
3771#define	SQTAG_TCP_DROP_Q0		37
3772#define	SQTAG_TCP_CONN_REQ_2		38
3773#define	SQTAG_IP_INPUT_RX_RING		39
3774#define	SQTAG_SQUEUE_CHANGE		40
3775#define	SQTAG_CONNECT_FINISH		41
3776#define	SQTAG_SYNCHRONOUS_OP		42
3777#define	SQTAG_TCP_SHUTDOWN_OUTPUT	43
3778#define	SQTAG_TCP_IXA_CLEANUP		44
3779#define	SQTAG_TCP_SEND_SYNACK		45
3780
3781extern sin_t	sin_null;	/* Zero address for quick clears */
3782extern sin6_t	sin6_null;	/* Zero address for quick clears */
3783
3784#endif	/* _KERNEL */
3785
3786#ifdef	__cplusplus
3787}
3788#endif
3789
3790#endif	/* _INET_IP_H */
3791