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 (c) 2012, Joyent, Inc. All rights reserved.
26 * Copyright 2014 Nexenta Systems, Inc. All rights reserved.
27 */
28
29 #ifndef _INET_IP_H
30 #define _INET_IP_H
31
32 #ifdef __cplusplus
33 extern "C" {
34 #endif
35
36 #include <sys/isa_defs.h>
37 #include <sys/types.h>
38 #include <inet/mib2.h>
39 #include <inet/nd.h>
40 #include <sys/atomic.h>
41 #include <net/if_dl.h>
42 #include <net/if.h>
43 #include <netinet/ip.h>
44 #include <netinet/igmp.h>
45 #include <sys/neti.h>
46 #include <sys/hook.h>
47 #include <sys/hook_event.h>
48 #include <sys/hook_impl.h>
49 #include <inet/ip_stack.h>
50
51 #ifdef _KERNEL
52 #include <netinet/ip6.h>
53 #include <sys/avl.h>
54 #include <sys/list.h>
55 #include <sys/vmem.h>
56 #include <sys/squeue.h>
57 #include <net/route.h>
58 #include <sys/systm.h>
59 #include <net/radix.h>
60 #include <sys/modhash.h>
61
62 #ifdef DEBUG
63 #define CONN_DEBUG
64 #endif
65
66 #define IP_DEBUG
67 /*
68 * The mt-streams(9F) flags for the IP module; put here so that other
69 * "drivers" that are actually IP (e.g., ICMP, UDP) can use the same set
70 * of flags.
71 */
72 #define IP_DEVMTFLAGS D_MP
73 #endif /* _KERNEL */
74
75 #define IP_MOD_NAME "ip"
76 #define IP_DEV_NAME "/dev/ip"
77 #define IP6_DEV_NAME "/dev/ip6"
78
79 #define UDP_MOD_NAME "udp"
80 #define UDP_DEV_NAME "/dev/udp"
81 #define UDP6_DEV_NAME "/dev/udp6"
82
83 #define TCP_MOD_NAME "tcp"
84 #define TCP_DEV_NAME "/dev/tcp"
85 #define TCP6_DEV_NAME "/dev/tcp6"
86
87 #define SCTP_MOD_NAME "sctp"
88
89 #ifndef _IPADDR_T
90 #define _IPADDR_T
91 typedef uint32_t ipaddr_t;
92 #endif
93
94 /* Number of bits in an address */
95 #define IP_ABITS 32
96 #define IPV4_ABITS IP_ABITS
97 #define IPV6_ABITS 128
98 #define IP_MAX_HW_LEN 40
99
100 #define IP_HOST_MASK (ipaddr_t)0xffffffffU
101
102 #define IP_CSUM(mp, off, sum) (~ip_cksum(mp, off, sum) & 0xFFFF)
103 #define IP_CSUM_PARTIAL(mp, off, sum) ip_cksum(mp, off, sum)
104 #define IP_BCSUM_PARTIAL(bp, len, sum) bcksum(bp, len, sum)
105
106 #define ILL_FRAG_HASH_TBL_COUNT ((unsigned int)64)
107 #define ILL_FRAG_HASH_TBL_SIZE (ILL_FRAG_HASH_TBL_COUNT * sizeof (ipfb_t))
108
109 #define IPV4_ADDR_LEN 4
110 #define IP_ADDR_LEN IPV4_ADDR_LEN
111 #define IP_ARP_PROTO_TYPE 0x0800
112
113 #define IPV4_VERSION 4
114 #define IP_VERSION IPV4_VERSION
115 #define IP_SIMPLE_HDR_LENGTH_IN_WORDS 5
116 #define IP_SIMPLE_HDR_LENGTH 20
117 #define IP_MAX_HDR_LENGTH 60
118
119 #define IP_MAX_OPT_LENGTH (IP_MAX_HDR_LENGTH-IP_SIMPLE_HDR_LENGTH)
120
121 #define IP_MIN_MTU (IP_MAX_HDR_LENGTH + 8) /* 68 bytes */
122
123 /*
124 * XXX IP_MAXPACKET is defined in <netinet/ip.h> as well. At some point the
125 * 2 files should be cleaned up to remove all redundant definitions.
126 */
127 #define IP_MAXPACKET 65535
128 #define IP_SIMPLE_HDR_VERSION \
129 ((IP_VERSION << 4) | IP_SIMPLE_HDR_LENGTH_IN_WORDS)
130
131 #define UDPH_SIZE 8
132
133 /*
134 * Constants and type definitions to support IP IOCTL commands
135 */
136 #define IP_IOCTL (('i'<<8)|'p')
137 #define IP_IOC_IRE_DELETE 4
138 #define IP_IOC_IRE_DELETE_NO_REPLY 5
139 #define IP_IOC_RTS_REQUEST 7
140
141 /* Common definitions used by IP IOCTL data structures */
142 typedef struct ipllcmd_s {
143 uint_t ipllc_cmd;
144 uint_t ipllc_name_offset;
145 uint_t ipllc_name_length;
146 } ipllc_t;
147
148 /* IP IRE Delete Command Structure. */
149 typedef struct ipid_s {
150 ipllc_t ipid_ipllc;
151 uint_t ipid_ire_type;
152 uint_t ipid_addr_offset;
153 uint_t ipid_addr_length;
154 uint_t ipid_mask_offset;
155 uint_t ipid_mask_length;
156 } ipid_t;
157
158 #define ipid_cmd ipid_ipllc.ipllc_cmd
159
160 #ifdef _KERNEL
161 /*
162 * Temporary state for ip options parser.
163 */
164 typedef struct ipoptp_s
165 {
166 uint8_t *ipoptp_next; /* next option to look at */
167 uint8_t *ipoptp_end; /* end of options */
168 uint8_t *ipoptp_cur; /* start of current option */
169 uint8_t ipoptp_len; /* length of current option */
170 uint32_t ipoptp_flags;
171 } ipoptp_t;
172
173 /*
174 * Flag(s) for ipoptp_flags
175 */
176 #define IPOPTP_ERROR 0x00000001
177 #endif /* _KERNEL */
178
179 /* Controls forwarding of IP packets, set via ipadm(1M)/ndd(1M) */
180 #define IP_FORWARD_NEVER 0
181 #define IP_FORWARD_ALWAYS 1
182
183 #define WE_ARE_FORWARDING(ipst) ((ipst)->ips_ip_forwarding == IP_FORWARD_ALWAYS)
184
185 #define IPH_HDR_LENGTH(ipha) \
186 ((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length & 0xF) << 2)
187
188 #define IPH_HDR_VERSION(ipha) \
189 ((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length) >> 4)
190
191 #ifdef _KERNEL
192 /*
193 * IP reassembly macros. We hide starting and ending offsets in b_next and
194 * b_prev of messages on the reassembly queue. The messages are chained using
195 * b_cont. These macros are used in ip_reassemble() so we don't have to see
196 * the ugly casts and assignments.
197 * Note that the offsets are <= 64k i.e. a uint_t is sufficient to represent
198 * them.
199 */
200 #define IP_REASS_START(mp) ((uint_t)(uintptr_t)((mp)->b_next))
201 #define IP_REASS_SET_START(mp, u) \
202 ((mp)->b_next = (mblk_t *)(uintptr_t)(u))
203 #define IP_REASS_END(mp) ((uint_t)(uintptr_t)((mp)->b_prev))
204 #define IP_REASS_SET_END(mp, u) \
205 ((mp)->b_prev = (mblk_t *)(uintptr_t)(u))
206
207 #define IP_REASS_COMPLETE 0x1
208 #define IP_REASS_PARTIAL 0x2
209 #define IP_REASS_FAILED 0x4
210
211 /*
212 * Test to determine whether this is a module instance of IP or a
213 * driver instance of IP.
214 */
215 #define CONN_Q(q) (WR(q)->q_next == NULL)
216
217 #define Q_TO_CONN(q) ((conn_t *)(q)->q_ptr)
218 #define Q_TO_TCP(q) (Q_TO_CONN((q))->conn_tcp)
219 #define Q_TO_UDP(q) (Q_TO_CONN((q))->conn_udp)
220 #define Q_TO_ICMP(q) (Q_TO_CONN((q))->conn_icmp)
221 #define Q_TO_RTS(q) (Q_TO_CONN((q))->conn_rts)
222
223 #define CONNP_TO_WQ(connp) ((connp)->conn_wq)
224 #define CONNP_TO_RQ(connp) ((connp)->conn_rq)
225
226 #define GRAB_CONN_LOCK(q) { \
227 if (q != NULL && CONN_Q(q)) \
228 mutex_enter(&(Q_TO_CONN(q))->conn_lock); \
229 }
230
231 #define RELEASE_CONN_LOCK(q) { \
232 if (q != NULL && CONN_Q(q)) \
233 mutex_exit(&(Q_TO_CONN(q))->conn_lock); \
234 }
235
236 /*
237 * Ref counter macros for ioctls. This provides a guard for TCP to stop
238 * tcp_close from removing the rq/wq whilst an ioctl is still in flight on the
239 * stream. The ioctl could have been queued on e.g. an ipsq. tcp_close will wait
240 * until the ioctlref count is zero before proceeding.
241 * Ideally conn_oper_pending_ill would be used for this purpose. However, in the
242 * case where an ioctl is aborted or interrupted, it can be cleared prematurely.
243 * There are also some race possibilities between ip and the stream head which
244 * can also end up with conn_oper_pending_ill being cleared prematurely. So, to
245 * avoid these situations, we use a dedicated ref counter for ioctls which is
246 * used in addition to and in parallel with the normal conn_ref count.
247 */
248 #define CONN_INC_IOCTLREF_LOCKED(connp) { \
249 ASSERT(MUTEX_HELD(&(connp)->conn_lock)); \
250 DTRACE_PROBE1(conn__inc__ioctlref, conn_t *, (connp)); \
251 (connp)->conn_ioctlref++; \
252 mutex_exit(&(connp)->conn_lock); \
253 }
254
255 #define CONN_INC_IOCTLREF(connp) { \
256 mutex_enter(&(connp)->conn_lock); \
257 CONN_INC_IOCTLREF_LOCKED(connp); \
258 }
259
260 #define CONN_DEC_IOCTLREF(connp) { \
261 mutex_enter(&(connp)->conn_lock); \
262 DTRACE_PROBE1(conn__dec__ioctlref, conn_t *, (connp)); \
263 /* Make sure conn_ioctlref will not underflow. */ \
264 ASSERT((connp)->conn_ioctlref != 0); \
265 if ((--(connp)->conn_ioctlref == 0) && \
266 ((connp)->conn_state_flags & CONN_CLOSING)) { \
267 cv_broadcast(&(connp)->conn_cv); \
268 } \
269 mutex_exit(&(connp)->conn_lock); \
270 }
271
272
273 /*
274 * Complete the pending operation. Usually an ioctl. Can also
275 * be a bind or option management request that got enqueued
276 * in an ipsq_t. Called on completion of the operation.
277 */
278 #define CONN_OPER_PENDING_DONE(connp) { \
279 mutex_enter(&(connp)->conn_lock); \
280 (connp)->conn_oper_pending_ill = NULL; \
281 cv_broadcast(&(connp)->conn_refcv); \
282 mutex_exit(&(connp)->conn_lock); \
283 CONN_DEC_REF(connp); \
284 }
285
286 /*
287 * Values for squeue switch:
288 */
289 #define IP_SQUEUE_ENTER_NODRAIN 1
290 #define IP_SQUEUE_ENTER 2
291 #define IP_SQUEUE_FILL 3
292
293 extern int ip_squeue_flag;
294
295 /* IP Fragmentation Reassembly Header */
296 typedef struct ipf_s {
297 struct ipf_s *ipf_hash_next;
298 struct ipf_s **ipf_ptphn; /* Pointer to previous hash next. */
299 uint32_t ipf_ident; /* Ident to match. */
300 uint8_t ipf_protocol; /* Protocol to match. */
301 uchar_t ipf_last_frag_seen : 1; /* Last fragment seen ? */
302 time_t ipf_timestamp; /* Reassembly start time. */
303 mblk_t *ipf_mp; /* mblk we live in. */
304 mblk_t *ipf_tail_mp; /* Frag queue tail pointer. */
305 int ipf_hole_cnt; /* Number of holes (hard-case). */
306 int ipf_end; /* Tail end offset (0 -> hard-case). */
307 uint_t ipf_gen; /* Frag queue generation */
308 size_t ipf_count; /* Count of bytes used by frag */
309 uint_t ipf_nf_hdr_len; /* Length of nonfragmented header */
310 in6_addr_t ipf_v6src; /* IPv6 source address */
311 in6_addr_t ipf_v6dst; /* IPv6 dest address */
312 uint_t ipf_prev_nexthdr_offset; /* Offset for nexthdr value */
313 uint8_t ipf_ecn; /* ECN info for the fragments */
314 uint8_t ipf_num_dups; /* Number of times dup frags recvd */
315 uint16_t ipf_checksum_flags; /* Hardware checksum flags */
316 uint32_t ipf_checksum; /* Partial checksum of fragment data */
317 } ipf_t;
318
319 /*
320 * IPv4 Fragments
321 */
322 #define IS_V4_FRAGMENT(ipha_fragment_offset_and_flags) \
323 (((ntohs(ipha_fragment_offset_and_flags) & IPH_OFFSET) != 0) || \
324 ((ntohs(ipha_fragment_offset_and_flags) & IPH_MF) != 0))
325
326 #define ipf_src V4_PART_OF_V6(ipf_v6src)
327 #define ipf_dst V4_PART_OF_V6(ipf_v6dst)
328
329 #endif /* _KERNEL */
330
331 /* ICMP types */
332 #define ICMP_ECHO_REPLY 0
333 #define ICMP_DEST_UNREACHABLE 3
334 #define ICMP_SOURCE_QUENCH 4
335 #define ICMP_REDIRECT 5
336 #define ICMP_ECHO_REQUEST 8
337 #define ICMP_ROUTER_ADVERTISEMENT 9
338 #define ICMP_ROUTER_SOLICITATION 10
339 #define ICMP_TIME_EXCEEDED 11
340 #define ICMP_PARAM_PROBLEM 12
341 #define ICMP_TIME_STAMP_REQUEST 13
342 #define ICMP_TIME_STAMP_REPLY 14
343 #define ICMP_INFO_REQUEST 15
344 #define ICMP_INFO_REPLY 16
345 #define ICMP_ADDRESS_MASK_REQUEST 17
346 #define ICMP_ADDRESS_MASK_REPLY 18
347
348 /* Evaluates to true if the ICMP type is an ICMP error */
349 #define ICMP_IS_ERROR(type) ( \
350 (type) == ICMP_DEST_UNREACHABLE || \
351 (type) == ICMP_SOURCE_QUENCH || \
352 (type) == ICMP_TIME_EXCEEDED || \
353 (type) == ICMP_PARAM_PROBLEM)
354
355 /* ICMP_TIME_EXCEEDED codes */
356 #define ICMP_TTL_EXCEEDED 0
357 #define ICMP_REASSEMBLY_TIME_EXCEEDED 1
358
359 /* ICMP_DEST_UNREACHABLE codes */
360 #define ICMP_NET_UNREACHABLE 0
361 #define ICMP_HOST_UNREACHABLE 1
362 #define ICMP_PROTOCOL_UNREACHABLE 2
363 #define ICMP_PORT_UNREACHABLE 3
364 #define ICMP_FRAGMENTATION_NEEDED 4
365 #define ICMP_SOURCE_ROUTE_FAILED 5
366 #define ICMP_DEST_NET_UNKNOWN 6
367 #define ICMP_DEST_HOST_UNKNOWN 7
368 #define ICMP_SRC_HOST_ISOLATED 8
369 #define ICMP_DEST_NET_UNREACH_ADMIN 9
370 #define ICMP_DEST_HOST_UNREACH_ADMIN 10
371 #define ICMP_DEST_NET_UNREACH_TOS 11
372 #define ICMP_DEST_HOST_UNREACH_TOS 12
373
374 /* ICMP Header Structure */
375 typedef struct icmph_s {
376 uint8_t icmph_type;
377 uint8_t icmph_code;
378 uint16_t icmph_checksum;
379 union {
380 struct { /* ECHO request/response structure */
381 uint16_t u_echo_ident;
382 uint16_t u_echo_seqnum;
383 } u_echo;
384 struct { /* Destination unreachable structure */
385 uint16_t u_du_zero;
386 uint16_t u_du_mtu;
387 } u_du;
388 struct { /* Parameter problem structure */
389 uint8_t u_pp_ptr;
390 uint8_t u_pp_rsvd[3];
391 } u_pp;
392 struct { /* Redirect structure */
393 ipaddr_t u_rd_gateway;
394 } u_rd;
395 } icmph_u;
396 } icmph_t;
397
398 #define icmph_echo_ident icmph_u.u_echo.u_echo_ident
399 #define icmph_echo_seqnum icmph_u.u_echo.u_echo_seqnum
400 #define icmph_du_zero icmph_u.u_du.u_du_zero
401 #define icmph_du_mtu icmph_u.u_du.u_du_mtu
402 #define icmph_pp_ptr icmph_u.u_pp.u_pp_ptr
403 #define icmph_rd_gateway icmph_u.u_rd.u_rd_gateway
404
405 #define ICMPH_SIZE 8
406
407 /*
408 * Minimum length of transport layer header included in an ICMP error
409 * message for it to be considered valid.
410 */
411 #define ICMP_MIN_TP_HDR_LEN 8
412
413 /* Aligned IP header */
414 typedef struct ipha_s {
415 uint8_t ipha_version_and_hdr_length;
416 uint8_t ipha_type_of_service;
417 uint16_t ipha_length;
418 uint16_t ipha_ident;
419 uint16_t ipha_fragment_offset_and_flags;
420 uint8_t ipha_ttl;
421 uint8_t ipha_protocol;
422 uint16_t ipha_hdr_checksum;
423 ipaddr_t ipha_src;
424 ipaddr_t ipha_dst;
425 } ipha_t;
426
427 /*
428 * IP Flags
429 *
430 * Some of these constant names are copied for the DTrace IP provider in
431 * usr/src/lib/libdtrace/common/{ip.d.in, ip.sed.in}, which should be kept
432 * in sync.
433 */
434 #define IPH_DF 0x4000 /* Don't fragment */
435 #define IPH_MF 0x2000 /* More fragments to come */
436 #define IPH_OFFSET 0x1FFF /* Where the offset lives */
437
438 /* Byte-order specific values */
439 #ifdef _BIG_ENDIAN
440 #define IPH_DF_HTONS 0x4000 /* Don't fragment */
441 #define IPH_MF_HTONS 0x2000 /* More fragments to come */
442 #define IPH_OFFSET_HTONS 0x1FFF /* Where the offset lives */
443 #else
444 #define IPH_DF_HTONS 0x0040 /* Don't fragment */
445 #define IPH_MF_HTONS 0x0020 /* More fragments to come */
446 #define IPH_OFFSET_HTONS 0xFF1F /* Where the offset lives */
447 #endif
448
449 /* ECN code points for IPv4 TOS byte and IPv6 traffic class octet. */
450 #define IPH_ECN_NECT 0x0 /* Not ECN-Capable Transport */
451 #define IPH_ECN_ECT1 0x1 /* ECN-Capable Transport, ECT(1) */
452 #define IPH_ECN_ECT0 0x2 /* ECN-Capable Transport, ECT(0) */
453 #define IPH_ECN_CE 0x3 /* ECN-Congestion Experienced (CE) */
454
455 struct ill_s;
456
457 typedef void ip_v6intfid_func_t(struct ill_s *, in6_addr_t *);
458 typedef void ip_v6mapinfo_func_t(struct ill_s *, uchar_t *, uchar_t *);
459 typedef void ip_v4mapinfo_func_t(struct ill_s *, uchar_t *, uchar_t *);
460
461 /* IP Mac info structure */
462 typedef struct ip_m_s {
463 t_uscalar_t ip_m_mac_type; /* From <sys/dlpi.h> */
464 int ip_m_type; /* From <net/if_types.h> */
465 t_uscalar_t ip_m_ipv4sap;
466 t_uscalar_t ip_m_ipv6sap;
467 ip_v4mapinfo_func_t *ip_m_v4mapping;
468 ip_v6mapinfo_func_t *ip_m_v6mapping;
469 ip_v6intfid_func_t *ip_m_v6intfid;
470 ip_v6intfid_func_t *ip_m_v6destintfid;
471 } ip_m_t;
472
473 /*
474 * The following functions attempt to reduce the link layer dependency
475 * of the IP stack. The current set of link specific operations are:
476 * a. map from IPv4 class D (224.0/4) multicast address range or the
477 * IPv6 multicast address range (ff00::/8) to the link layer multicast
478 * address.
479 * b. derive the default IPv6 interface identifier from the interface.
480 * c. derive the default IPv6 destination interface identifier from
481 * the interface (point-to-point only).
482 */
483 extern void ip_mcast_mapping(struct ill_s *, uchar_t *, uchar_t *);
484 /* ip_m_v6*intfid return void and are never NULL */
485 #define MEDIA_V6INTFID(ip_m, ill, v6ptr) (ip_m)->ip_m_v6intfid(ill, v6ptr)
486 #define MEDIA_V6DESTINTFID(ip_m, ill, v6ptr) \
487 (ip_m)->ip_m_v6destintfid(ill, v6ptr)
488
489 /* Router entry types */
490 #define IRE_BROADCAST 0x0001 /* Route entry for broadcast address */
491 #define IRE_DEFAULT 0x0002 /* Route entry for default gateway */
492 #define IRE_LOCAL 0x0004 /* Route entry for local address */
493 #define IRE_LOOPBACK 0x0008 /* Route entry for loopback address */
494 #define IRE_PREFIX 0x0010 /* Route entry for prefix routes */
495 #ifndef _KERNEL
496 /* Keep so user-level still compiles */
497 #define IRE_CACHE 0x0020 /* Cached Route entry */
498 #endif
499 #define IRE_IF_NORESOLVER 0x0040 /* Route entry for local interface */
500 /* net without any address mapping. */
501 #define IRE_IF_RESOLVER 0x0080 /* Route entry for local interface */
502 /* net with resolver. */
503 #define IRE_HOST 0x0100 /* Host route entry */
504 /* Keep so user-level still compiles */
505 #define IRE_HOST_REDIRECT 0x0200 /* only used for T_SVR4_OPTMGMT_REQ */
506 #define IRE_IF_CLONE 0x0400 /* Per host clone of IRE_IF */
507 #define IRE_MULTICAST 0x0800 /* Special - not in table */
508 #define IRE_NOROUTE 0x1000 /* Special - not in table */
509
510 #define IRE_INTERFACE (IRE_IF_NORESOLVER | IRE_IF_RESOLVER)
511
512 #define IRE_IF_ALL (IRE_IF_NORESOLVER | IRE_IF_RESOLVER | \
513 IRE_IF_CLONE)
514 #define IRE_OFFSUBNET (IRE_DEFAULT | IRE_PREFIX | IRE_HOST)
515 #define IRE_OFFLINK IRE_OFFSUBNET
516 /*
517 * Note that we view IRE_NOROUTE as ONLINK since we can "send" to them without
518 * going through a router; the result of sending will be an error/icmp error.
519 */
520 #define IRE_ONLINK (IRE_IF_ALL|IRE_LOCAL|IRE_LOOPBACK| \
521 IRE_BROADCAST|IRE_MULTICAST|IRE_NOROUTE)
522
523 /* Arguments to ire_flush_cache() */
524 #define IRE_FLUSH_DELETE 0
525 #define IRE_FLUSH_ADD 1
526 #define IRE_FLUSH_GWCHANGE 2
527
528 /*
529 * Flags to ire_route_recursive
530 */
531 #define IRR_NONE 0
532 #define IRR_ALLOCATE 1 /* OK to allocate IRE_IF_CLONE */
533 #define IRR_INCOMPLETE 2 /* OK to return incomplete chain */
534
535 /*
536 * Open/close synchronization flags.
537 * These are kept in a separate field in the conn and the synchronization
538 * depends on the atomic 32 bit access to that field.
539 */
540 #define CONN_CLOSING 0x01 /* ip_close waiting for ip_wsrv */
541 #define CONN_CONDEMNED 0x02 /* conn is closing, no more refs */
542 #define CONN_INCIPIENT 0x04 /* conn not yet visible, no refs */
543 #define CONN_QUIESCED 0x08 /* conn is now quiescent */
544 #define CONN_UPDATE_ILL 0x10 /* conn_update_ill in progress */
545
546 /*
547 * Flags for dce_flags field. Specifies which information has been set.
548 * dce_ident is always present, but the other ones are identified by the flags.
549 */
550 #define DCEF_DEFAULT 0x0001 /* Default DCE - no pmtu or uinfo */
551 #define DCEF_PMTU 0x0002 /* Different than interface MTU */
552 #define DCEF_UINFO 0x0004 /* dce_uinfo set */
553 #define DCEF_TOO_SMALL_PMTU 0x0008 /* Smaller than IPv4/IPv6 MIN */
554
555 #ifdef _KERNEL
556 /*
557 * Extra structures need for per-src-addr filtering (IGMPv3/MLDv2)
558 */
559 #define MAX_FILTER_SIZE 64
560
561 typedef struct slist_s {
562 int sl_numsrc;
563 in6_addr_t sl_addr[MAX_FILTER_SIZE];
564 } slist_t;
565
566 /*
567 * Following struct is used to maintain retransmission state for
568 * a multicast group. One rtx_state_t struct is an in-line field
569 * of the ilm_t struct; the slist_ts in the rtx_state_t struct are
570 * alloc'd as needed.
571 */
572 typedef struct rtx_state_s {
573 uint_t rtx_timer; /* retrans timer */
574 int rtx_cnt; /* retrans count */
575 int rtx_fmode_cnt; /* retrans count for fmode change */
576 slist_t *rtx_allow;
577 slist_t *rtx_block;
578 } rtx_state_t;
579
580 /*
581 * Used to construct list of multicast address records that will be
582 * sent in a single listener report.
583 */
584 typedef struct mrec_s {
585 struct mrec_s *mrec_next;
586 uint8_t mrec_type;
587 uint8_t mrec_auxlen; /* currently unused */
588 in6_addr_t mrec_group;
589 slist_t mrec_srcs;
590 } mrec_t;
591
592 /* Group membership list per upper conn */
593
594 /*
595 * We record the multicast information from the socket option in
596 * ilg_ifaddr/ilg_ifindex. This allows rejoining the group in the case when
597 * the ifaddr (or ifindex) disappears and later reappears, potentially on
598 * a different ill. The IPv6 multicast socket options and ioctls all specify
599 * the interface using an ifindex. For IPv4 some socket options/ioctls use
600 * the interface address and others use the index. We record here the method
601 * that was actually used (and leave the other of ilg_ifaddr or ilg_ifindex)
602 * at zero so that we can rejoin the way the application intended.
603 *
604 * We track the ill on which we will or already have joined an ilm using
605 * ilg_ill. When we have succeeded joining the ilm and have a refhold on it
606 * then we set ilg_ilm. Thus intentionally there is a window where ilg_ill is
607 * set and ilg_ilm is not set. This allows clearing ilg_ill as a signal that
608 * the ill is being unplumbed and the ilm should be discarded.
609 *
610 * ilg records the state of multicast memberships of a socket end point.
611 * ilm records the state of multicast memberships with the driver and is
612 * maintained per interface.
613 *
614 * The ilg state is protected by conn_ilg_lock.
615 * The ilg will not be freed until ilg_refcnt drops to zero.
616 */
617 typedef struct ilg_s {
618 struct ilg_s *ilg_next;
619 struct ilg_s **ilg_ptpn;
620 struct conn_s *ilg_connp; /* Back pointer to get lock */
621 in6_addr_t ilg_v6group;
622 ipaddr_t ilg_ifaddr; /* For some IPv4 cases */
623 uint_t ilg_ifindex; /* IPv6 and some other IPv4 cases */
624 struct ill_s *ilg_ill; /* Where ilm is joined. No refhold */
625 struct ilm_s *ilg_ilm; /* With ilm_refhold */
626 uint_t ilg_refcnt;
627 mcast_record_t ilg_fmode; /* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */
628 slist_t *ilg_filter;
629 boolean_t ilg_condemned; /* Conceptually deleted */
630 } ilg_t;
631
632 /*
633 * Multicast address list entry for ill.
634 * ilm_ill is used by IPv4 and IPv6
635 *
636 * The ilm state (and other multicast state on the ill) is protected by
637 * ill_mcast_lock. Operations that change state on both an ilg and ilm
638 * in addition use ill_mcast_serializer to ensure that we can't have
639 * interleaving between e.g., add and delete operations for the same conn_t,
640 * group, and ill. The ill_mcast_serializer is also used to ensure that
641 * multicast group joins do not occur on an interface that is in the process
642 * of joining an IPMP group.
643 *
644 * The comment below (and for other netstack_t references) refers
645 * to the fact that we only do netstack_hold in particular cases,
646 * such as the references from open endpoints (ill_t and conn_t's
647 * pointers). Internally within IP we rely on IP's ability to cleanup e.g.
648 * ire_t's when an ill goes away.
649 */
650 typedef struct ilm_s {
651 in6_addr_t ilm_v6addr;
652 int ilm_refcnt;
653 uint_t ilm_timer; /* IGMP/MLD query resp timer, in msec */
654 struct ilm_s *ilm_next; /* Linked list for each ill */
655 uint_t ilm_state; /* state of the membership */
656 struct ill_s *ilm_ill; /* Back pointer to ill - ill_ilm_cnt */
657 zoneid_t ilm_zoneid;
658 int ilm_no_ilg_cnt; /* number of joins w/ no ilg */
659 mcast_record_t ilm_fmode; /* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */
660 slist_t *ilm_filter; /* source filter list */
661 slist_t *ilm_pendsrcs; /* relevant src addrs for pending req */
662 rtx_state_t ilm_rtx; /* SCR retransmission state */
663 ipaddr_t ilm_ifaddr; /* For IPv4 netstat */
664 ip_stack_t *ilm_ipst; /* Does not have a netstack_hold */
665 } ilm_t;
666
667 #define ilm_addr V4_PART_OF_V6(ilm_v6addr)
668
669 /*
670 * Soft reference to an IPsec SA.
671 *
672 * On relative terms, conn's can be persistent (living as long as the
673 * processes which create them), while SA's are ephemeral (dying when
674 * they hit their time-based or byte-based lifetimes).
675 *
676 * We could hold a hard reference to an SA from an ipsec_latch_t,
677 * but this would cause expired SA's to linger for a potentially
678 * unbounded time.
679 *
680 * Instead, we remember the hash bucket number and bucket generation
681 * in addition to the pointer. The bucket generation is incremented on
682 * each deletion.
683 */
684 typedef struct ipsa_ref_s
685 {
686 struct ipsa_s *ipsr_sa;
687 struct isaf_s *ipsr_bucket;
688 uint64_t ipsr_gen;
689 } ipsa_ref_t;
690
691 /*
692 * IPsec "latching" state.
693 *
694 * In the presence of IPsec policy, fully-bound conn's bind a connection
695 * to more than just the 5-tuple, but also a specific IPsec action and
696 * identity-pair.
697 * The identity pair is accessed from both the receive and transmit side
698 * hence it is maintained in the ipsec_latch_t structure. conn_latch and
699 * ixa_ipsec_latch points to it.
700 * The policy and actions are stored in conn_latch_in_policy and
701 * conn_latch_in_action for the inbound side, and in ixa_ipsec_policy and
702 * ixa_ipsec_action for the transmit side.
703 *
704 * As an optimization, we also cache soft references to IPsec SA's in
705 * ip_xmit_attr_t so that we can fast-path around most of the work needed for
706 * outbound IPsec SA selection.
707 */
708 typedef struct ipsec_latch_s
709 {
710 kmutex_t ipl_lock;
711 uint32_t ipl_refcnt;
712
713 struct ipsid_s *ipl_local_cid;
714 struct ipsid_s *ipl_remote_cid;
715 unsigned int
716 ipl_ids_latched : 1,
717
718 ipl_pad_to_bit_31 : 31;
719 } ipsec_latch_t;
720
721 #define IPLATCH_REFHOLD(ipl) { \
722 atomic_add_32(&(ipl)->ipl_refcnt, 1); \
723 ASSERT((ipl)->ipl_refcnt != 0); \
724 }
725
726 #define IPLATCH_REFRELE(ipl) { \
727 ASSERT((ipl)->ipl_refcnt != 0); \
728 membar_exit(); \
729 if (atomic_add_32_nv(&(ipl)->ipl_refcnt, -1) == 0) \
730 iplatch_free(ipl); \
731 }
732
733 /*
734 * peer identity structure.
735 */
736 typedef struct conn_s conn_t;
737
738 /*
739 * This is used to match an inbound/outbound datagram with policy.
740 */
741 typedef struct ipsec_selector {
742 in6_addr_t ips_local_addr_v6;
743 in6_addr_t ips_remote_addr_v6;
744 uint16_t ips_local_port;
745 uint16_t ips_remote_port;
746 uint8_t ips_icmp_type;
747 uint8_t ips_icmp_code;
748 uint8_t ips_protocol;
749 uint8_t ips_isv4 : 1,
750 ips_is_icmp_inv_acq: 1;
751 } ipsec_selector_t;
752
753 /*
754 * Note that we put v4 addresses in the *first* 32-bit word of the
755 * selector rather than the last to simplify the prefix match/mask code
756 * in spd.c
757 */
758 #define ips_local_addr_v4 ips_local_addr_v6.s6_addr32[0]
759 #define ips_remote_addr_v4 ips_remote_addr_v6.s6_addr32[0]
760
761 /* Values used in IP by IPSEC Code */
762 #define IPSEC_OUTBOUND B_TRUE
763 #define IPSEC_INBOUND B_FALSE
764
765 /*
766 * There are two variants in policy failures. The packet may come in
767 * secure when not needed (IPSEC_POLICY_???_NOT_NEEDED) or it may not
768 * have the desired level of protection (IPSEC_POLICY_MISMATCH).
769 */
770 #define IPSEC_POLICY_NOT_NEEDED 0
771 #define IPSEC_POLICY_MISMATCH 1
772 #define IPSEC_POLICY_AUTH_NOT_NEEDED 2
773 #define IPSEC_POLICY_ENCR_NOT_NEEDED 3
774 #define IPSEC_POLICY_SE_NOT_NEEDED 4
775 #define IPSEC_POLICY_MAX 5 /* Always max + 1. */
776
777 /*
778 * Check with IPSEC inbound policy if
779 *
780 * 1) per-socket policy is present - indicated by conn_in_enforce_policy.
781 * 2) Or if we have not cached policy on the conn and the global policy is
782 * non-empty.
783 */
784 #define CONN_INBOUND_POLICY_PRESENT(connp, ipss) \
785 ((connp)->conn_in_enforce_policy || \
786 (!((connp)->conn_policy_cached) && \
787 (ipss)->ipsec_inbound_v4_policy_present))
788
789 #define CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss) \
790 ((connp)->conn_in_enforce_policy || \
791 (!(connp)->conn_policy_cached && \
792 (ipss)->ipsec_inbound_v6_policy_present))
793
794 #define CONN_OUTBOUND_POLICY_PRESENT(connp, ipss) \
795 ((connp)->conn_out_enforce_policy || \
796 (!((connp)->conn_policy_cached) && \
797 (ipss)->ipsec_outbound_v4_policy_present))
798
799 #define CONN_OUTBOUND_POLICY_PRESENT_V6(connp, ipss) \
800 ((connp)->conn_out_enforce_policy || \
801 (!(connp)->conn_policy_cached && \
802 (ipss)->ipsec_outbound_v6_policy_present))
803
804 /*
805 * Information cached in IRE for upper layer protocol (ULP).
806 */
807 typedef struct iulp_s {
808 boolean_t iulp_set; /* Is any metric set? */
809 uint32_t iulp_ssthresh; /* Slow start threshold (TCP). */
810 clock_t iulp_rtt; /* Guestimate in millisecs. */
811 clock_t iulp_rtt_sd; /* Cached value of RTT variance. */
812 uint32_t iulp_spipe; /* Send pipe size. */
813 uint32_t iulp_rpipe; /* Receive pipe size. */
814 uint32_t iulp_rtomax; /* Max round trip timeout. */
815 uint32_t iulp_sack; /* Use SACK option (TCP)? */
816 uint32_t iulp_mtu; /* Setable with routing sockets */
817
818 uint32_t
819 iulp_tstamp_ok : 1, /* Use timestamp option (TCP)? */
820 iulp_wscale_ok : 1, /* Use window scale option (TCP)? */
821 iulp_ecn_ok : 1, /* Enable ECN (for TCP)? */
822 iulp_pmtud_ok : 1, /* Enable PMTUd? */
823
824 /* These three are passed out by ip_set_destination */
825 iulp_localnet: 1, /* IRE_ONLINK */
826 iulp_loopback: 1, /* IRE_LOOPBACK */
827 iulp_local: 1, /* IRE_LOCAL */
828
829 iulp_not_used : 25;
830 } iulp_t;
831
832 /*
833 * The conn drain list structure (idl_t), protected by idl_lock. Each conn_t
834 * inserted in the list points back at this idl_t using conn_idl, and is
835 * chained by conn_drain_next and conn_drain_prev, which are also protected by
836 * idl_lock. When flow control is relieved, either ip_wsrv() (STREAMS) or
837 * ill_flow_enable() (non-STREAMS) will call conn_drain().
838 *
839 * The conn drain list, idl_t, itself is part of tx cookie list structure.
840 * A tx cookie list points to a blocked Tx ring and contains the list of
841 * all conn's that are blocked due to the flow-controlled Tx ring (via
842 * the idl drain list). Note that a link can have multiple Tx rings. The
843 * drain list will store the conn's blocked due to Tx ring being flow
844 * controlled.
845 */
846
847 typedef uintptr_t ip_mac_tx_cookie_t;
848 typedef struct idl_s idl_t;
849 typedef struct idl_tx_list_s idl_tx_list_t;
850
851 struct idl_tx_list_s {
852 ip_mac_tx_cookie_t txl_cookie;
853 kmutex_t txl_lock; /* Lock for this list */
854 idl_t *txl_drain_list;
855 int txl_drain_index;
856 };
857
858 struct idl_s {
859 conn_t *idl_conn; /* Head of drain list */
860 kmutex_t idl_lock; /* Lock for this list */
861 idl_tx_list_t *idl_itl;
862 };
863
864 /*
865 * Interface route structure which holds the necessary information to recreate
866 * routes that are tied to an interface i.e. have ire_ill set.
867 *
868 * These routes which were initially created via a routing socket or via the
869 * SIOCADDRT ioctl may be gateway routes (RTF_GATEWAY being set) or may be
870 * traditional interface routes. When an ill comes back up after being
871 * down, this information will be used to recreate the routes. These
872 * are part of an mblk_t chain that hangs off of the ILL (ill_saved_ire_mp).
873 */
874 typedef struct ifrt_s {
875 ushort_t ifrt_type; /* Type of IRE */
876 in6_addr_t ifrt_v6addr; /* Address IRE represents. */
877 in6_addr_t ifrt_v6gateway_addr; /* Gateway if IRE_OFFLINK */
878 in6_addr_t ifrt_v6setsrc_addr; /* Src addr if RTF_SETSRC */
879 in6_addr_t ifrt_v6mask; /* Mask for matching IRE. */
880 uint32_t ifrt_flags; /* flags related to route */
881 iulp_t ifrt_metrics; /* Routing socket metrics */
882 zoneid_t ifrt_zoneid; /* zoneid for route */
883 } ifrt_t;
884
885 #define ifrt_addr V4_PART_OF_V6(ifrt_v6addr)
886 #define ifrt_gateway_addr V4_PART_OF_V6(ifrt_v6gateway_addr)
887 #define ifrt_mask V4_PART_OF_V6(ifrt_v6mask)
888 #define ifrt_setsrc_addr V4_PART_OF_V6(ifrt_v6setsrc_addr)
889
890 /* Number of IP addresses that can be hosted on a physical interface */
891 #define MAX_ADDRS_PER_IF 8192
892 /*
893 * Number of Source addresses to be considered for source address
894 * selection. Used by ipif_select_source_v4/v6.
895 */
896 #define MAX_IPIF_SELECT_SOURCE 50
897
898 #ifdef IP_DEBUG
899 /*
900 * Trace refholds and refreles for debugging.
901 */
902 #define TR_STACK_DEPTH 14
903 typedef struct tr_buf_s {
904 int tr_depth;
905 clock_t tr_time;
906 pc_t tr_stack[TR_STACK_DEPTH];
907 } tr_buf_t;
908
909 typedef struct th_trace_s {
910 int th_refcnt;
911 uint_t th_trace_lastref;
912 kthread_t *th_id;
913 #define TR_BUF_MAX 38
914 tr_buf_t th_trbuf[TR_BUF_MAX];
915 } th_trace_t;
916
917 typedef struct th_hash_s {
918 list_node_t thh_link;
919 mod_hash_t *thh_hash;
920 ip_stack_t *thh_ipst;
921 } th_hash_t;
922 #endif
923
924 /* The following are ipif_state_flags */
925 #define IPIF_CONDEMNED 0x1 /* The ipif is being removed */
926 #define IPIF_CHANGING 0x2 /* A critcal ipif field is changing */
927 #define IPIF_SET_LINKLOCAL 0x10 /* transient flag during bringup */
928
929 /* IP interface structure, one per local address */
930 typedef struct ipif_s {
931 struct ipif_s *ipif_next;
932 struct ill_s *ipif_ill; /* Back pointer to our ill */
933 int ipif_id; /* Logical unit number */
934 in6_addr_t ipif_v6lcl_addr; /* Local IP address for this if. */
935 in6_addr_t ipif_v6subnet; /* Subnet prefix for this if. */
936 in6_addr_t ipif_v6net_mask; /* Net mask for this interface. */
937 in6_addr_t ipif_v6brd_addr; /* Broadcast addr for this interface. */
938 in6_addr_t ipif_v6pp_dst_addr; /* Point-to-point dest address. */
939 uint64_t ipif_flags; /* Interface flags. */
940 uint_t ipif_ire_type; /* IRE_LOCAL or IRE_LOOPBACK */
941
942 /*
943 * The packet count in the ipif contain the sum of the
944 * packet counts in dead IRE_LOCAL/LOOPBACK for this ipif.
945 */
946 uint_t ipif_ib_pkt_count; /* Inbound packets for our dead IREs */
947
948 /* Exclusive bit fields, protected by ipsq_t */
949 unsigned int
950 ipif_was_up : 1, /* ipif was up before */
951 ipif_addr_ready : 1, /* DAD is done */
952 ipif_was_dup : 1, /* DAD had failed */
953 ipif_added_nce : 1, /* nce added for local address */
954
955 ipif_pad_to_31 : 28;
956
957 ilm_t *ipif_allhosts_ilm; /* For all-nodes join */
958 ilm_t *ipif_solmulti_ilm; /* For IPv6 solicited multicast join */
959
960 uint_t ipif_seqid; /* unique index across all ills */
961 uint_t ipif_state_flags; /* See IPIF_* flag defs above */
962 uint_t ipif_refcnt; /* active consistent reader cnt */
963
964 zoneid_t ipif_zoneid; /* zone ID number */
965 timeout_id_t ipif_recovery_id; /* Timer for DAD recovery */
966 boolean_t ipif_trace_disable; /* True when alloc fails */
967 /*
968 * For an IPMP interface, ipif_bound_ill tracks the ill whose hardware
969 * information this ipif is associated with via ARP/NDP. We can use
970 * an ill pointer (rather than an index) because only ills that are
971 * part of a group will be pointed to, and an ill cannot disappear
972 * while it's in a group.
973 */
974 struct ill_s *ipif_bound_ill;
975 struct ipif_s *ipif_bound_next; /* bound ipif chain */
976 boolean_t ipif_bound; /* B_TRUE if we successfully bound */
977
978 struct ire_s *ipif_ire_local; /* Our IRE_LOCAL or LOOPBACK */
979 struct ire_s *ipif_ire_if; /* Our IRE_INTERFACE */
980 } ipif_t;
981
982 /*
983 * The following table lists the protection levels of the various members
984 * of the ipif_t. The following notation is used.
985 *
986 * Write once - Written to only once at the time of bringing up
987 * the interface and can be safely read after the bringup without any lock.
988 *
989 * ipsq - Need to execute in the ipsq to perform the indicated access.
990 *
991 * ill_lock - Need to hold this mutex to perform the indicated access.
992 *
993 * ill_g_lock - Need to hold this rw lock as reader/writer for read access or
994 * write access respectively.
995 *
996 * down ill - Written to only when the ill is down (i.e all ipifs are down)
997 * up ill - Read only when the ill is up (i.e. at least 1 ipif is up)
998 *
999 * Table of ipif_t members and their protection
1000 *
1001 * ipif_next ipsq + ill_lock + ipsq OR ill_lock OR
1002 * ill_g_lock ill_g_lock
1003 * ipif_ill ipsq + down ipif write once
1004 * ipif_id ipsq + down ipif write once
1005 * ipif_v6lcl_addr ipsq + down ipif up ipif
1006 * ipif_v6subnet ipsq + down ipif up ipif
1007 * ipif_v6net_mask ipsq + down ipif up ipif
1008 *
1009 * ipif_v6brd_addr
1010 * ipif_v6pp_dst_addr
1011 * ipif_flags ill_lock ill_lock
1012 * ipif_ire_type ipsq + down ill up ill
1013 *
1014 * ipif_ib_pkt_count Approx
1015 *
1016 * bit fields ill_lock ill_lock
1017 *
1018 * ipif_allhosts_ilm ipsq ipsq
1019 * ipif_solmulti_ilm ipsq ipsq
1020 *
1021 * ipif_seqid ipsq Write once
1022 *
1023 * ipif_state_flags ill_lock ill_lock
1024 * ipif_refcnt ill_lock ill_lock
1025 * ipif_bound_ill ipsq + ipmp_lock ipsq OR ipmp_lock
1026 * ipif_bound_next ipsq ipsq
1027 * ipif_bound ipsq ipsq
1028 *
1029 * ipif_ire_local ipsq + ips_ill_g_lock ipsq OR ips_ill_g_lock
1030 * ipif_ire_if ipsq + ips_ill_g_lock ipsq OR ips_ill_g_lock
1031 */
1032
1033 /*
1034 * Return values from ip_laddr_verify_{v4,v6}
1035 */
1036 typedef enum { IPVL_UNICAST_UP, IPVL_UNICAST_DOWN, IPVL_MCAST, IPVL_BCAST,
1037 IPVL_BAD} ip_laddr_t;
1038
1039
1040 #define IP_TR_HASH(tid) ((((uintptr_t)tid) >> 6) & (IP_TR_HASH_MAX - 1))
1041
1042 #ifdef DEBUG
1043 #define IPIF_TRACE_REF(ipif) ipif_trace_ref(ipif)
1044 #define ILL_TRACE_REF(ill) ill_trace_ref(ill)
1045 #define IPIF_UNTRACE_REF(ipif) ipif_untrace_ref(ipif)
1046 #define ILL_UNTRACE_REF(ill) ill_untrace_ref(ill)
1047 #else
1048 #define IPIF_TRACE_REF(ipif)
1049 #define ILL_TRACE_REF(ill)
1050 #define IPIF_UNTRACE_REF(ipif)
1051 #define ILL_UNTRACE_REF(ill)
1052 #endif
1053
1054 /* IPv4 compatibility macros */
1055 #define ipif_lcl_addr V4_PART_OF_V6(ipif_v6lcl_addr)
1056 #define ipif_subnet V4_PART_OF_V6(ipif_v6subnet)
1057 #define ipif_net_mask V4_PART_OF_V6(ipif_v6net_mask)
1058 #define ipif_brd_addr V4_PART_OF_V6(ipif_v6brd_addr)
1059 #define ipif_pp_dst_addr V4_PART_OF_V6(ipif_v6pp_dst_addr)
1060
1061 /* Macros for easy backreferences to the ill. */
1062 #define ipif_isv6 ipif_ill->ill_isv6
1063
1064 #define SIOCLIFADDR_NDX 112 /* ndx of SIOCLIFADDR in the ndx ioctl table */
1065
1066 /*
1067 * mode value for ip_ioctl_finish for finishing an ioctl
1068 */
1069 #define CONN_CLOSE 1 /* No mi_copy */
1070 #define COPYOUT 2 /* do an mi_copyout if needed */
1071 #define NO_COPYOUT 3 /* do an mi_copy_done */
1072 #define IPI2MODE(ipi) ((ipi)->ipi_flags & IPI_GET_CMD ? COPYOUT : NO_COPYOUT)
1073
1074 /*
1075 * The IP-MT design revolves around the serialization objects ipsq_t (IPSQ)
1076 * and ipxop_t (exclusive operation or "xop"). Becoming "writer" on an IPSQ
1077 * ensures that no other threads can become "writer" on any IPSQs sharing that
1078 * IPSQ's xop until the writer thread is done.
1079 *
1080 * Each phyint points to one IPSQ that remains fixed over the phyint's life.
1081 * Each IPSQ points to one xop that can change over the IPSQ's life. If a
1082 * phyint is *not* in an IPMP group, then its IPSQ will refer to the IPSQ's
1083 * "own" xop (ipsq_ownxop). If a phyint *is* part of an IPMP group, then its
1084 * IPSQ will refer to the "group" xop, which is shorthand for the xop of the
1085 * IPSQ of the IPMP meta-interface's phyint. Thus, all phyints that are part
1086 * of the same IPMP group will have their IPSQ's point to the group xop, and
1087 * thus becoming "writer" on any phyint in the group will prevent any other
1088 * writer on any other phyint in the group. All IPSQs sharing the same xop
1089 * are chained together through ipsq_next (in the degenerate common case,
1090 * ipsq_next simply refers to itself). Note that the group xop is guaranteed
1091 * to exist at least as long as there are members in the group, since the IPMP
1092 * meta-interface can only be destroyed if the group is empty.
1093 *
1094 * Incoming exclusive operation requests are enqueued on the IPSQ they arrived
1095 * on rather than the xop. This makes switching xop's (as would happen when a
1096 * phyint leaves an IPMP group) simple, because after the phyint leaves the
1097 * group, any operations enqueued on its IPSQ can be safely processed with
1098 * respect to its new xop, and any operations enqueued on the IPSQs of its
1099 * former group can be processed with respect to their existing group xop.
1100 * Even so, switching xops is a subtle dance; see ipsq_dq() for details.
1101 *
1102 * An IPSQ's "own" xop is embedded within the IPSQ itself since they have have
1103 * identical lifetimes, and because doing so simplifies pointer management.
1104 * While each phyint and IPSQ point to each other, it is not possible to free
1105 * the IPSQ when the phyint is freed, since we may still *inside* the IPSQ
1106 * when the phyint is being freed. Thus, ipsq_phyint is set to NULL when the
1107 * phyint is freed, and the IPSQ free is later done in ipsq_exit().
1108 *
1109 * ipsq_t synchronization: read write
1110 *
1111 * ipsq_xopq_mphead ipx_lock ipx_lock
1112 * ipsq_xopq_mptail ipx_lock ipx_lock
1113 * ipsq_xop_switch_mp ipsq_lock ipsq_lock
1114 * ipsq_phyint write once write once
1115 * ipsq_next RW_READER ill_g_lock RW_WRITER ill_g_lock
1116 * ipsq_xop ipsq_lock or ipsq ipsq_lock + ipsq
1117 * ipsq_swxop ipsq ipsq
1118 * ipsq_ownxop see ipxop_t see ipxop_t
1119 * ipsq_ipst write once write once
1120 *
1121 * ipxop_t synchronization: read write
1122 *
1123 * ipx_writer ipx_lock ipx_lock
1124 * ipx_xop_queued ipx_lock ipx_lock
1125 * ipx_mphead ipx_lock ipx_lock
1126 * ipx_mptail ipx_lock ipx_lock
1127 * ipx_ipsq write once write once
1128 * ips_ipsq_queued ipx_lock ipx_lock
1129 * ipx_waitfor ipsq or ipx_lock ipsq + ipx_lock
1130 * ipx_reentry_cnt ipsq or ipx_lock ipsq + ipx_lock
1131 * ipx_current_done ipsq ipsq
1132 * ipx_current_ioctl ipsq ipsq
1133 * ipx_current_ipif ipsq or ipx_lock ipsq + ipx_lock
1134 * ipx_pending_ipif ipsq or ipx_lock ipsq + ipx_lock
1135 * ipx_pending_mp ipsq or ipx_lock ipsq + ipx_lock
1136 * ipx_forced ipsq ipsq
1137 * ipx_depth ipsq ipsq
1138 * ipx_stack ipsq ipsq
1139 */
1140 typedef struct ipxop_s {
1141 kmutex_t ipx_lock; /* see above */
1142 kthread_t *ipx_writer; /* current owner */
1143 mblk_t *ipx_mphead; /* messages tied to this op */
1144 mblk_t *ipx_mptail;
1145 struct ipsq_s *ipx_ipsq; /* associated ipsq */
1146 boolean_t ipx_ipsq_queued; /* ipsq using xop has queued op */
1147 int ipx_waitfor; /* waiting; values encoded below */
1148 int ipx_reentry_cnt;
1149 boolean_t ipx_current_done; /* is the current operation done? */
1150 int ipx_current_ioctl; /* current ioctl, or 0 if no ioctl */
1151 ipif_t *ipx_current_ipif; /* ipif for current op */
1152 ipif_t *ipx_pending_ipif; /* ipif for ipx_pending_mp */
1153 mblk_t *ipx_pending_mp; /* current ioctl mp while waiting */
1154 boolean_t ipx_forced; /* debugging aid */
1155 #ifdef DEBUG
1156 int ipx_depth; /* debugging aid */
1157 #define IPX_STACK_DEPTH 15
1158 pc_t ipx_stack[IPX_STACK_DEPTH]; /* debugging aid */
1159 #endif
1160 } ipxop_t;
1161
1162 typedef struct ipsq_s {
1163 kmutex_t ipsq_lock; /* see above */
1164 mblk_t *ipsq_switch_mp; /* op to handle right after switch */
1165 mblk_t *ipsq_xopq_mphead; /* list of excl ops (mostly ioctls) */
1166 mblk_t *ipsq_xopq_mptail;
1167 struct phyint *ipsq_phyint; /* associated phyint */
1168 struct ipsq_s *ipsq_next; /* next ipsq sharing ipsq_xop */
1169 struct ipxop_s *ipsq_xop; /* current xop synchronization info */
1170 struct ipxop_s *ipsq_swxop; /* switch xop to on ipsq_exit() */
1171 struct ipxop_s ipsq_ownxop; /* our own xop (may not be in-use) */
1172 ip_stack_t *ipsq_ipst; /* does not have a netstack_hold */
1173 } ipsq_t;
1174
1175 /*
1176 * ipx_waitfor values:
1177 */
1178 enum {
1179 IPIF_DOWN = 1, /* ipif_down() waiting for refcnts to drop */
1180 ILL_DOWN, /* ill_down() waiting for refcnts to drop */
1181 IPIF_FREE, /* ipif_free() waiting for refcnts to drop */
1182 ILL_FREE /* ill unplumb waiting for refcnts to drop */
1183 };
1184
1185 /* Operation types for ipsq_try_enter() */
1186 #define CUR_OP 0 /* request writer within current operation */
1187 #define NEW_OP 1 /* request writer for a new operation */
1188 #define SWITCH_OP 2 /* request writer once IPSQ XOP switches */
1189
1190 /*
1191 * Kstats tracked on each IPMP meta-interface. Order here must match
1192 * ipmp_kstats[] in ip/ipmp.c.
1193 */
1194 enum {
1195 IPMP_KSTAT_OBYTES, IPMP_KSTAT_OBYTES64, IPMP_KSTAT_RBYTES,
1196 IPMP_KSTAT_RBYTES64, IPMP_KSTAT_OPACKETS, IPMP_KSTAT_OPACKETS64,
1197 IPMP_KSTAT_OERRORS, IPMP_KSTAT_IPACKETS, IPMP_KSTAT_IPACKETS64,
1198 IPMP_KSTAT_IERRORS, IPMP_KSTAT_MULTIRCV, IPMP_KSTAT_MULTIXMT,
1199 IPMP_KSTAT_BRDCSTRCV, IPMP_KSTAT_BRDCSTXMT, IPMP_KSTAT_LINK_UP,
1200 IPMP_KSTAT_MAX /* keep last */
1201 };
1202
1203 /*
1204 * phyint represents state that is common to both IPv4 and IPv6 interfaces.
1205 * There is a separate ill_t representing IPv4 and IPv6 which has a
1206 * backpointer to the phyint structure for accessing common state.
1207 */
1208 typedef struct phyint {
1209 struct ill_s *phyint_illv4;
1210 struct ill_s *phyint_illv6;
1211 uint_t phyint_ifindex; /* SIOCSLIFINDEX */
1212 uint64_t phyint_flags;
1213 avl_node_t phyint_avl_by_index; /* avl tree by index */
1214 avl_node_t phyint_avl_by_name; /* avl tree by name */
1215 kmutex_t phyint_lock;
1216 struct ipsq_s *phyint_ipsq; /* back pointer to ipsq */
1217 struct ipmp_grp_s *phyint_grp; /* associated IPMP group */
1218 char phyint_name[LIFNAMSIZ]; /* physical interface name */
1219 uint64_t phyint_kstats0[IPMP_KSTAT_MAX]; /* baseline kstats */
1220 } phyint_t;
1221
1222 #define CACHE_ALIGN_SIZE 64
1223 #define CACHE_ALIGN(align_struct) P2ROUNDUP(sizeof (struct align_struct),\
1224 CACHE_ALIGN_SIZE)
1225 struct _phyint_list_s_ {
1226 avl_tree_t phyint_list_avl_by_index; /* avl tree by index */
1227 avl_tree_t phyint_list_avl_by_name; /* avl tree by name */
1228 };
1229
1230 typedef union phyint_list_u {
1231 struct _phyint_list_s_ phyint_list_s;
1232 char phyint_list_filler[CACHE_ALIGN(_phyint_list_s_)];
1233 } phyint_list_t;
1234
1235 #define phyint_list_avl_by_index phyint_list_s.phyint_list_avl_by_index
1236 #define phyint_list_avl_by_name phyint_list_s.phyint_list_avl_by_name
1237
1238 /*
1239 * Fragmentation hash bucket
1240 */
1241 typedef struct ipfb_s {
1242 struct ipf_s *ipfb_ipf; /* List of ... */
1243 size_t ipfb_count; /* Count of bytes used by frag(s) */
1244 kmutex_t ipfb_lock; /* Protect all ipf in list */
1245 uint_t ipfb_frag_pkts; /* num of distinct fragmented pkts */
1246 } ipfb_t;
1247
1248 /*
1249 * IRE bucket structure. Usually there is an array of such structures,
1250 * each pointing to a linked list of ires. irb_refcnt counts the number
1251 * of walkers of a given hash bucket. Usually the reference count is
1252 * bumped up if the walker wants no IRES to be DELETED while walking the
1253 * list. Bumping up does not PREVENT ADDITION. This allows walking a given
1254 * hash bucket without stumbling up on a free pointer.
1255 *
1256 * irb_t structures in ip_ftable are dynamically allocated and freed.
1257 * In order to identify the irb_t structures that can be safely kmem_free'd
1258 * we need to ensure that
1259 * - the irb_refcnt is quiescent, indicating no other walkers,
1260 * - no other threads or ire's are holding references to the irb,
1261 * i.e., irb_nire == 0,
1262 * - there are no active ire's in the bucket, i.e., irb_ire_cnt == 0
1263 */
1264 typedef struct irb {
1265 struct ire_s *irb_ire; /* First ire in this bucket */
1266 /* Should be first in this struct */
1267 krwlock_t irb_lock; /* Protect this bucket */
1268 uint_t irb_refcnt; /* Protected by irb_lock */
1269 uchar_t irb_marks; /* CONDEMNED ires in this bucket ? */
1270 #define IRB_MARK_CONDEMNED 0x0001 /* Contains some IRE_IS_CONDEMNED */
1271 #define IRB_MARK_DYNAMIC 0x0002 /* Dynamically allocated */
1272 /* Once IPv6 uses radix then IRB_MARK_DYNAMIC will be always be set */
1273 uint_t irb_ire_cnt; /* Num of active IRE in this bucket */
1274 int irb_nire; /* Num of ftable ire's that ref irb */
1275 ip_stack_t *irb_ipst; /* Does not have a netstack_hold */
1276 } irb_t;
1277
1278 /*
1279 * This is the structure used to store the multicast physical addresses
1280 * that an interface has joined.
1281 * The refcnt keeps track of the number of multicast IP addresses mapping
1282 * to a physical multicast address.
1283 */
1284 typedef struct multiphysaddr_s {
1285 struct multiphysaddr_s *mpa_next;
1286 char mpa_addr[IP_MAX_HW_LEN];
1287 int mpa_refcnt;
1288 } multiphysaddr_t;
1289
1290 #define IRB2RT(irb) (rt_t *)((caddr_t)(irb) - offsetof(rt_t, rt_irb))
1291
1292 /* Forward declarations */
1293 struct dce_s;
1294 typedef struct dce_s dce_t;
1295 struct ire_s;
1296 typedef struct ire_s ire_t;
1297 struct ncec_s;
1298 typedef struct ncec_s ncec_t;
1299 struct nce_s;
1300 typedef struct nce_s nce_t;
1301 struct ip_recv_attr_s;
1302 typedef struct ip_recv_attr_s ip_recv_attr_t;
1303 struct ip_xmit_attr_s;
1304 typedef struct ip_xmit_attr_s ip_xmit_attr_t;
1305
1306 struct tsol_ire_gw_secattr_s;
1307 typedef struct tsol_ire_gw_secattr_s tsol_ire_gw_secattr_t;
1308
1309 /*
1310 * This is a structure for a one-element route cache that is passed
1311 * by reference between ip_input and ill_inputfn.
1312 */
1313 typedef struct {
1314 ire_t *rtc_ire;
1315 ipaddr_t rtc_ipaddr;
1316 in6_addr_t rtc_ip6addr;
1317 } rtc_t;
1318
1319 /*
1320 * Note: Temporarily use 64 bits, and will probably go back to 32 bits after
1321 * more cleanup work is done.
1322 */
1323 typedef uint64_t iaflags_t;
1324
1325 /* The ill input function pointer type */
1326 typedef void (*pfillinput_t)(mblk_t *, void *, void *, ip_recv_attr_t *,
1327 rtc_t *);
1328
1329 /* The ire receive function pointer type */
1330 typedef void (*pfirerecv_t)(ire_t *, mblk_t *, void *, ip_recv_attr_t *);
1331
1332 /* The ire send and postfrag function pointer types */
1333 typedef int (*pfiresend_t)(ire_t *, mblk_t *, void *,
1334 ip_xmit_attr_t *, uint32_t *);
1335 typedef int (*pfirepostfrag_t)(mblk_t *, nce_t *, iaflags_t, uint_t, uint32_t,
1336 zoneid_t, zoneid_t, uintptr_t *);
1337
1338
1339 #define IP_V4_G_HEAD 0
1340 #define IP_V6_G_HEAD 1
1341
1342 #define MAX_G_HEADS 2
1343
1344 /*
1345 * unpadded ill_if structure
1346 */
1347 struct _ill_if_s_ {
1348 union ill_if_u *illif_next;
1349 union ill_if_u *illif_prev;
1350 avl_tree_t illif_avl_by_ppa; /* AVL tree sorted on ppa */
1351 vmem_t *illif_ppa_arena; /* ppa index space */
1352 uint16_t illif_mcast_v1; /* hints for */
1353 uint16_t illif_mcast_v2; /* [igmp|mld]_slowtimo */
1354 int illif_name_len; /* name length */
1355 char illif_name[LIFNAMSIZ]; /* name of interface type */
1356 };
1357
1358 /* cache aligned ill_if structure */
1359 typedef union ill_if_u {
1360 struct _ill_if_s_ ill_if_s;
1361 char illif_filler[CACHE_ALIGN(_ill_if_s_)];
1362 } ill_if_t;
1363
1364 #define illif_next ill_if_s.illif_next
1365 #define illif_prev ill_if_s.illif_prev
1366 #define illif_avl_by_ppa ill_if_s.illif_avl_by_ppa
1367 #define illif_ppa_arena ill_if_s.illif_ppa_arena
1368 #define illif_mcast_v1 ill_if_s.illif_mcast_v1
1369 #define illif_mcast_v2 ill_if_s.illif_mcast_v2
1370 #define illif_name ill_if_s.illif_name
1371 #define illif_name_len ill_if_s.illif_name_len
1372
1373 typedef struct ill_walk_context_s {
1374 int ctx_current_list; /* current list being searched */
1375 int ctx_last_list; /* last list to search */
1376 } ill_walk_context_t;
1377
1378 /*
1379 * ill_g_heads structure, one for IPV4 and one for IPV6
1380 */
1381 struct _ill_g_head_s_ {
1382 ill_if_t *ill_g_list_head;
1383 ill_if_t *ill_g_list_tail;
1384 };
1385
1386 typedef union ill_g_head_u {
1387 struct _ill_g_head_s_ ill_g_head_s;
1388 char ill_g_head_filler[CACHE_ALIGN(_ill_g_head_s_)];
1389 } ill_g_head_t;
1390
1391 #define ill_g_list_head ill_g_head_s.ill_g_list_head
1392 #define ill_g_list_tail ill_g_head_s.ill_g_list_tail
1393
1394 #define IP_V4_ILL_G_LIST(ipst) \
1395 (ipst)->ips_ill_g_heads[IP_V4_G_HEAD].ill_g_list_head
1396 #define IP_V6_ILL_G_LIST(ipst) \
1397 (ipst)->ips_ill_g_heads[IP_V6_G_HEAD].ill_g_list_head
1398 #define IP_VX_ILL_G_LIST(i, ipst) \
1399 (ipst)->ips_ill_g_heads[i].ill_g_list_head
1400
1401 #define ILL_START_WALK_V4(ctx_ptr, ipst) \
1402 ill_first(IP_V4_G_HEAD, IP_V4_G_HEAD, ctx_ptr, ipst)
1403 #define ILL_START_WALK_V6(ctx_ptr, ipst) \
1404 ill_first(IP_V6_G_HEAD, IP_V6_G_HEAD, ctx_ptr, ipst)
1405 #define ILL_START_WALK_ALL(ctx_ptr, ipst) \
1406 ill_first(MAX_G_HEADS, MAX_G_HEADS, ctx_ptr, ipst)
1407
1408 /*
1409 * Capabilities, possible flags for ill_capabilities.
1410 */
1411 #define ILL_CAPAB_LSO 0x04 /* Large Send Offload */
1412 #define ILL_CAPAB_HCKSUM 0x08 /* Hardware checksumming */
1413 #define ILL_CAPAB_ZEROCOPY 0x10 /* Zero-copy */
1414 #define ILL_CAPAB_DLD 0x20 /* DLD capabilities */
1415 #define ILL_CAPAB_DLD_POLL 0x40 /* Polling */
1416 #define ILL_CAPAB_DLD_DIRECT 0x80 /* Direct function call */
1417
1418 /*
1419 * Per-ill Hardware Checksumming capbilities.
1420 */
1421 typedef struct ill_hcksum_capab_s ill_hcksum_capab_t;
1422
1423 /*
1424 * Per-ill Zero-copy capabilities.
1425 */
1426 typedef struct ill_zerocopy_capab_s ill_zerocopy_capab_t;
1427
1428 /*
1429 * DLD capbilities.
1430 */
1431 typedef struct ill_dld_capab_s ill_dld_capab_t;
1432
1433 /*
1434 * Per-ill polling resource map.
1435 */
1436 typedef struct ill_rx_ring ill_rx_ring_t;
1437
1438 /*
1439 * Per-ill Large Send Offload capabilities.
1440 */
1441 typedef struct ill_lso_capab_s ill_lso_capab_t;
1442
1443 /* The following are ill_state_flags */
1444 #define ILL_LL_SUBNET_PENDING 0x01 /* Waiting for DL_INFO_ACK from drv */
1445 #define ILL_CONDEMNED 0x02 /* No more new ref's to the ILL */
1446 #define ILL_DL_UNBIND_IN_PROGRESS 0x04 /* UNBIND_REQ is sent */
1447 /*
1448 * ILL_DOWN_IN_PROGRESS is set to ensure the following:
1449 * - no packets are sent to the driver after the DL_UNBIND_REQ is sent,
1450 * - no longstanding references will be acquired on objects that are being
1451 * brought down.
1452 */
1453 #define ILL_DOWN_IN_PROGRESS 0x08
1454
1455 /* Is this an ILL whose source address is used by other ILL's ? */
1456 #define IS_USESRC_ILL(ill) \
1457 (((ill)->ill_usesrc_ifindex == 0) && \
1458 ((ill)->ill_usesrc_grp_next != NULL))
1459
1460 /* Is this a client/consumer of the usesrc ILL ? */
1461 #define IS_USESRC_CLI_ILL(ill) \
1462 (((ill)->ill_usesrc_ifindex != 0) && \
1463 ((ill)->ill_usesrc_grp_next != NULL))
1464
1465 /* Is this an virtual network interface (vni) ILL ? */
1466 #define IS_VNI(ill) \
1467 (((ill)->ill_phyint->phyint_flags & (PHYI_LOOPBACK|PHYI_VIRTUAL)) == \
1468 PHYI_VIRTUAL)
1469
1470 /* Is this a loopback ILL? */
1471 #define IS_LOOPBACK(ill) \
1472 ((ill)->ill_phyint->phyint_flags & PHYI_LOOPBACK)
1473
1474 /* Is this an IPMP meta-interface ILL? */
1475 #define IS_IPMP(ill) \
1476 ((ill)->ill_phyint->phyint_flags & PHYI_IPMP)
1477
1478 /* Is this ILL under an IPMP meta-interface? (aka "in a group?") */
1479 #define IS_UNDER_IPMP(ill) \
1480 ((ill)->ill_grp != NULL && !IS_IPMP(ill))
1481
1482 /* Is ill1 in the same illgrp as ill2? */
1483 #define IS_IN_SAME_ILLGRP(ill1, ill2) \
1484 ((ill1)->ill_grp != NULL && ((ill1)->ill_grp == (ill2)->ill_grp))
1485
1486 /* Is ill1 on the same LAN as ill2? */
1487 #define IS_ON_SAME_LAN(ill1, ill2) \
1488 ((ill1) == (ill2) || IS_IN_SAME_ILLGRP(ill1, ill2))
1489
1490 #define ILL_OTHER(ill) \
1491 ((ill)->ill_isv6 ? (ill)->ill_phyint->phyint_illv4 : \
1492 (ill)->ill_phyint->phyint_illv6)
1493
1494 /*
1495 * IPMP group ILL state structure -- up to two per IPMP group (V4 and V6).
1496 * Created when the V4 and/or V6 IPMP meta-interface is I_PLINK'd. It is
1497 * guaranteed to persist while there are interfaces of that type in the group.
1498 * In general, most fields are accessed outside of the IPSQ (e.g., in the
1499 * datapath), and thus use locks in addition to the IPSQ for protection.
1500 *
1501 * synchronization: read write
1502 *
1503 * ig_if ipsq or ill_g_lock ipsq and ill_g_lock
1504 * ig_actif ipsq or ipmp_lock ipsq and ipmp_lock
1505 * ig_nactif ipsq or ipmp_lock ipsq and ipmp_lock
1506 * ig_next_ill ipsq or ipmp_lock ipsq and ipmp_lock
1507 * ig_ipmp_ill write once write once
1508 * ig_cast_ill ipsq or ipmp_lock ipsq and ipmp_lock
1509 * ig_arpent ipsq ipsq
1510 * ig_mtu ipsq ipsq
1511 * ig_mc_mtu ipsq ipsq
1512 */
1513 typedef struct ipmp_illgrp_s {
1514 list_t ig_if; /* list of all interfaces */
1515 list_t ig_actif; /* list of active interfaces */
1516 uint_t ig_nactif; /* number of active interfaces */
1517 struct ill_s *ig_next_ill; /* next active interface to use */
1518 struct ill_s *ig_ipmp_ill; /* backpointer to IPMP meta-interface */
1519 struct ill_s *ig_cast_ill; /* nominated ill for multi/broadcast */
1520 list_t ig_arpent; /* list of ARP entries */
1521 uint_t ig_mtu; /* ig_ipmp_ill->ill_mtu */
1522 uint_t ig_mc_mtu; /* ig_ipmp_ill->ill_mc_mtu */
1523 } ipmp_illgrp_t;
1524
1525 /*
1526 * IPMP group state structure -- one per IPMP group. Created when the
1527 * IPMP meta-interface is plumbed; it is guaranteed to persist while there
1528 * are interfaces in it.
1529 *
1530 * ipmp_grp_t synchronization: read write
1531 *
1532 * gr_name ipmp_lock ipmp_lock
1533 * gr_ifname write once write once
1534 * gr_mactype ipmp_lock ipmp_lock
1535 * gr_phyint write once write once
1536 * gr_nif ipmp_lock ipmp_lock
1537 * gr_nactif ipsq ipsq
1538 * gr_v4 ipmp_lock ipmp_lock
1539 * gr_v6 ipmp_lock ipmp_lock
1540 * gr_nv4 ipmp_lock ipmp_lock
1541 * gr_nv6 ipmp_lock ipmp_lock
1542 * gr_pendv4 ipmp_lock ipmp_lock
1543 * gr_pendv6 ipmp_lock ipmp_lock
1544 * gr_linkdownmp ipsq ipsq
1545 * gr_ksp ipmp_lock ipmp_lock
1546 * gr_kstats0 atomic atomic
1547 */
1548 typedef struct ipmp_grp_s {
1549 char gr_name[LIFGRNAMSIZ]; /* group name */
1550 char gr_ifname[LIFNAMSIZ]; /* interface name */
1551 t_uscalar_t gr_mactype; /* DLPI mactype of group */
1552 phyint_t *gr_phyint; /* IPMP group phyint */
1553 uint_t gr_nif; /* number of interfaces in group */
1554 uint_t gr_nactif; /* number of active interfaces */
1555 ipmp_illgrp_t *gr_v4; /* V4 group information */
1556 ipmp_illgrp_t *gr_v6; /* V6 group information */
1557 uint_t gr_nv4; /* number of ills in V4 group */
1558 uint_t gr_nv6; /* number of ills in V6 group */
1559 uint_t gr_pendv4; /* number of pending ills in V4 group */
1560 uint_t gr_pendv6; /* number of pending ills in V6 group */
1561 mblk_t *gr_linkdownmp; /* message used to bring link down */
1562 kstat_t *gr_ksp; /* group kstat pointer */
1563 uint64_t gr_kstats0[IPMP_KSTAT_MAX]; /* baseline group kstats */
1564 } ipmp_grp_t;
1565
1566 /*
1567 * IPMP ARP entry -- one per SIOCS*ARP entry tied to the group. Used to keep
1568 * ARP up-to-date as the active set of interfaces in the group changes.
1569 */
1570 typedef struct ipmp_arpent_s {
1571 ipaddr_t ia_ipaddr; /* IP address for this entry */
1572 boolean_t ia_proxyarp; /* proxy ARP entry? */
1573 boolean_t ia_notified; /* ARP notified about this entry? */
1574 list_node_t ia_node; /* next ARP entry in list */
1575 uint16_t ia_flags; /* nce_flags for the address */
1576 size_t ia_lladdr_len;
1577 uchar_t *ia_lladdr;
1578 } ipmp_arpent_t;
1579
1580 struct arl_s;
1581
1582 /*
1583 * Per-ill capabilities.
1584 */
1585 struct ill_hcksum_capab_s {
1586 uint_t ill_hcksum_version; /* interface version */
1587 uint_t ill_hcksum_txflags; /* capabilities on transmit */
1588 };
1589
1590 struct ill_zerocopy_capab_s {
1591 uint_t ill_zerocopy_version; /* interface version */
1592 uint_t ill_zerocopy_flags; /* capabilities */
1593 };
1594
1595 struct ill_lso_capab_s {
1596 uint_t ill_lso_flags; /* capabilities */
1597 uint_t ill_lso_max; /* maximum size of payload */
1598 };
1599
1600 /*
1601 * IP Lower level Structure.
1602 * Instance data structure in ip_open when there is a device below us.
1603 */
1604 typedef struct ill_s {
1605 pfillinput_t ill_inputfn; /* Fast input function selector */
1606 ill_if_t *ill_ifptr; /* pointer to interface type */
1607 queue_t *ill_rq; /* Read queue. */
1608 queue_t *ill_wq; /* Write queue. */
1609
1610 int ill_error; /* Error value sent up by device. */
1611
1612 ipif_t *ill_ipif; /* Interface chain for this ILL. */
1613
1614 uint_t ill_ipif_up_count; /* Number of IPIFs currently up. */
1615 uint_t ill_max_frag; /* Max IDU from DLPI. */
1616 uint_t ill_current_frag; /* Current IDU from DLPI. */
1617 uint_t ill_mtu; /* User-specified MTU; SIOCSLIFMTU */
1618 uint_t ill_mc_mtu; /* MTU for multi/broadcast */
1619 uint_t ill_metric; /* BSD if metric, for compatibility. */
1620 char *ill_name; /* Our name. */
1621 uint_t ill_ipif_dup_count; /* Number of duplicate addresses. */
1622 uint_t ill_name_length; /* Name length, incl. terminator. */
1623 uint_t ill_net_type; /* IRE_IF_RESOLVER/IRE_IF_NORESOLVER. */
1624 /*
1625 * Physical Point of Attachment num. If DLPI style 1 provider
1626 * then this is derived from the devname.
1627 */
1628 uint_t ill_ppa;
1629 t_uscalar_t ill_sap;
1630 t_scalar_t ill_sap_length; /* Including sign (for position) */
1631 uint_t ill_phys_addr_length; /* Excluding the sap. */
1632 uint_t ill_bcast_addr_length; /* Only set when the DL provider */
1633 /* supports broadcast. */
1634 t_uscalar_t ill_mactype;
1635 uint8_t *ill_frag_ptr; /* Reassembly state. */
1636 timeout_id_t ill_frag_timer_id; /* timeout id for the frag timer */
1637 ipfb_t *ill_frag_hash_tbl; /* Fragment hash list head. */
1638
1639 krwlock_t ill_mcast_lock; /* Protects multicast state */
1640 kmutex_t ill_mcast_serializer; /* Serialize across ilg and ilm state */
1641 ilm_t *ill_ilm; /* Multicast membership for ill */
1642 uint_t ill_global_timer; /* for IGMPv3/MLDv2 general queries */
1643 int ill_mcast_type; /* type of router which is querier */
1644 /* on this interface */
1645 uint16_t ill_mcast_v1_time; /* # slow timeouts since last v1 qry */
1646 uint16_t ill_mcast_v2_time; /* # slow timeouts since last v2 qry */
1647 uint8_t ill_mcast_v1_tset; /* 1 => timer is set; 0 => not set */
1648 uint8_t ill_mcast_v2_tset; /* 1 => timer is set; 0 => not set */
1649
1650 uint8_t ill_mcast_rv; /* IGMPv3/MLDv2 robustness variable */
1651 int ill_mcast_qi; /* IGMPv3/MLDv2 query interval var */
1652
1653 /*
1654 * All non-NULL cells between 'ill_first_mp_to_free' and
1655 * 'ill_last_mp_to_free' are freed in ill_delete.
1656 */
1657 #define ill_first_mp_to_free ill_bcast_mp
1658 mblk_t *ill_bcast_mp; /* DLPI header for broadcasts. */
1659 mblk_t *ill_unbind_mp; /* unbind mp from ill_dl_up() */
1660 mblk_t *ill_promiscoff_mp; /* for ill_leave_allmulti() */
1661 mblk_t *ill_dlpi_deferred; /* b_next chain of control messages */
1662 mblk_t *ill_dest_addr_mp; /* mblk which holds ill_dest_addr */
1663 mblk_t *ill_replumb_mp; /* replumb mp from ill_replumb() */
1664 mblk_t *ill_phys_addr_mp; /* mblk which holds ill_phys_addr */
1665 mblk_t *ill_mcast_deferred; /* b_next chain of IGMP/MLD packets */
1666 #define ill_last_mp_to_free ill_mcast_deferred
1667
1668 cred_t *ill_credp; /* opener's credentials */
1669 uint8_t *ill_phys_addr; /* ill_phys_addr_mp->b_rptr + off */
1670 uint8_t *ill_dest_addr; /* ill_dest_addr_mp->b_rptr + off */
1671
1672 uint_t ill_state_flags; /* see ILL_* flags above */
1673
1674 /* Following bit fields protected by ipsq_t */
1675 uint_t
1676 ill_needs_attach : 1,
1677 ill_reserved : 1,
1678 ill_isv6 : 1,
1679 ill_dlpi_style_set : 1,
1680
1681 ill_ifname_pending : 1,
1682 ill_logical_down : 1,
1683 ill_dl_up : 1,
1684 ill_up_ipifs : 1,
1685
1686 ill_note_link : 1, /* supports link-up notification */
1687 ill_capab_reneg : 1, /* capability renegotiation to be done */
1688 ill_dld_capab_inprog : 1, /* direct dld capab call in prog */
1689 ill_need_recover_multicast : 1,
1690
1691 ill_replumbing : 1,
1692 ill_arl_dlpi_pending : 1,
1693 ill_grp_pending : 1,
1694
1695 ill_pad_to_bit_31 : 17;
1696
1697 /* Following bit fields protected by ill_lock */
1698 uint_t
1699 ill_fragtimer_executing : 1,
1700 ill_fragtimer_needrestart : 1,
1701 ill_manual_token : 1, /* system won't override ill_token */
1702 /*
1703 * ill_manual_linklocal : system will not change the
1704 * linklocal whenever ill_token changes.
1705 */
1706 ill_manual_linklocal : 1,
1707
1708 ill_manual_dst_linklocal : 1, /* same for pt-pt dst linklocal */
1709
1710 ill_pad_bit_31 : 27;
1711
1712 /*
1713 * Used in SIOCSIFMUXID and SIOCGIFMUXID for 'ifconfig unplumb'.
1714 */
1715 int ill_muxid; /* muxid returned from plink */
1716
1717 /* Used for IP frag reassembly throttling on a per ILL basis. */
1718 uint_t ill_ipf_gen; /* Generation of next fragment queue */
1719 uint_t ill_frag_count; /* Count of all reassembly mblk bytes */
1720 uint_t ill_frag_free_num_pkts; /* num of fragmented packets to free */
1721 clock_t ill_last_frag_clean_time; /* time when frag's were pruned */
1722 int ill_type; /* From <net/if_types.h> */
1723 uint_t ill_dlpi_multicast_state; /* See below IDS_* */
1724 uint_t ill_dlpi_fastpath_state; /* See below IDS_* */
1725
1726 /*
1727 * Capabilities related fields.
1728 */
1729 uint_t ill_dlpi_capab_state; /* State of capability query, IDCS_* */
1730 uint_t ill_capab_pending_cnt;
1731 uint64_t ill_capabilities; /* Enabled capabilities, ILL_CAPAB_* */
1732 ill_hcksum_capab_t *ill_hcksum_capab; /* H/W cksumming capabilities */
1733 ill_zerocopy_capab_t *ill_zerocopy_capab; /* Zero-copy capabilities */
1734 ill_dld_capab_t *ill_dld_capab; /* DLD capabilities */
1735 ill_lso_capab_t *ill_lso_capab; /* Large Segment Offload capabilities */
1736 mblk_t *ill_capab_reset_mp; /* Preallocated mblk for capab reset */
1737
1738 uint8_t ill_max_hops; /* Maximum hops for any logical interface */
1739 uint_t ill_user_mtu; /* User-specified MTU via SIOCSLIFLNKINFO */
1740 uint32_t ill_reachable_time; /* Value for ND algorithm in msec */
1741 uint32_t ill_reachable_retrans_time; /* Value for ND algorithm msec */
1742 uint_t ill_max_buf; /* Max # of req to buffer for ND */
1743 in6_addr_t ill_token; /* IPv6 interface id */
1744 in6_addr_t ill_dest_token; /* Destination IPv6 interface id */
1745 uint_t ill_token_length;
1746 uint32_t ill_xmit_count; /* ndp max multicast xmits */
1747 mib2_ipIfStatsEntry_t *ill_ip_mib; /* ver indep. interface mib */
1748 mib2_ipv6IfIcmpEntry_t *ill_icmp6_mib; /* Per interface mib */
1749
1750 phyint_t *ill_phyint;
1751 uint64_t ill_flags;
1752
1753 kmutex_t ill_lock; /* Please see table below */
1754 /*
1755 * The ill_nd_lla* fields handle the link layer address option
1756 * from neighbor discovery. This is used for external IPv6
1757 * address resolution.
1758 */
1759 mblk_t *ill_nd_lla_mp; /* mblk which holds ill_nd_lla */
1760 uint8_t *ill_nd_lla; /* Link Layer Address */
1761 uint_t ill_nd_lla_len; /* Link Layer Address length */
1762 /*
1763 * We have 4 phys_addr_req's sent down. This field keeps track
1764 * of which one is pending.
1765 */
1766 t_uscalar_t ill_phys_addr_pend; /* which dl_phys_addr_req pending */
1767 /*
1768 * Used to save errors that occur during plumbing
1769 */
1770 uint_t ill_ifname_pending_err;
1771 avl_node_t ill_avl_byppa; /* avl node based on ppa */
1772 list_t ill_nce; /* pointer to nce_s list */
1773 uint_t ill_refcnt; /* active refcnt by threads */
1774 uint_t ill_ire_cnt; /* ires associated with this ill */
1775 kcondvar_t ill_cv;
1776 uint_t ill_ncec_cnt; /* ncecs associated with this ill */
1777 uint_t ill_nce_cnt; /* nces associated with this ill */
1778 uint_t ill_waiters; /* threads waiting in ipsq_enter */
1779 /*
1780 * Contains the upper read queue pointer of the module immediately
1781 * beneath IP. This field allows IP to validate sub-capability
1782 * acknowledgments coming up from downstream.
1783 */
1784 queue_t *ill_lmod_rq; /* read queue pointer of module below */
1785 uint_t ill_lmod_cnt; /* number of modules beneath IP */
1786 ip_m_t *ill_media; /* media specific params/functions */
1787 t_uscalar_t ill_dlpi_pending; /* Last DLPI primitive issued */
1788 uint_t ill_usesrc_ifindex; /* use src addr from this ILL */
1789 struct ill_s *ill_usesrc_grp_next; /* Next ILL in the usesrc group */
1790 boolean_t ill_trace_disable; /* True when alloc fails */
1791 zoneid_t ill_zoneid;
1792 ip_stack_t *ill_ipst; /* Corresponds to a netstack_hold */
1793 uint32_t ill_dhcpinit; /* IP_DHCPINIT_IFs for ill */
1794 void *ill_flownotify_mh; /* Tx flow ctl, mac cb handle */
1795 uint_t ill_ilm_cnt; /* ilms referencing this ill */
1796 uint_t ill_ipallmulti_cnt; /* ip_join_allmulti() calls */
1797 ilm_t *ill_ipallmulti_ilm;
1798
1799 mblk_t *ill_saved_ire_mp; /* Allocated for each extra IRE */
1800 /* with ire_ill set so they can */
1801 /* survive the ill going down and up. */
1802 kmutex_t ill_saved_ire_lock; /* Protects ill_saved_ire_mp, cnt */
1803 uint_t ill_saved_ire_cnt; /* # entries */
1804 struct arl_ill_common_s *ill_common;
1805 ire_t *ill_ire_multicast; /* IRE_MULTICAST for ill */
1806 clock_t ill_defend_start; /* start of 1 hour period */
1807 uint_t ill_defend_count; /* # of announce/defends per ill */
1808 /*
1809 * IPMP fields.
1810 */
1811 ipmp_illgrp_t *ill_grp; /* IPMP group information */
1812 list_node_t ill_actnode; /* next active ill in group */
1813 list_node_t ill_grpnode; /* next ill in group */
1814 ipif_t *ill_src_ipif; /* source address selection rotor */
1815 ipif_t *ill_move_ipif; /* ipif awaiting move to new ill */
1816 boolean_t ill_nom_cast; /* nominated for mcast/bcast */
1817 uint_t ill_bound_cnt; /* # of data addresses bound to ill */
1818 ipif_t *ill_bound_ipif; /* ipif chain bound to ill */
1819 timeout_id_t ill_refresh_tid; /* ill refresh retry timeout id */
1820
1821 uint32_t ill_mrouter_cnt; /* mrouter allmulti joins */
1822 uint32_t ill_allowed_ips_cnt;
1823 in6_addr_t *ill_allowed_ips;
1824
1825 /* list of multicast physical addresses joined on this ill */
1826 multiphysaddr_t *ill_mphysaddr_list;
1827 } ill_t;
1828
1829 /*
1830 * ILL_FREE_OK() means that there are no incoming pointer references
1831 * to the ill.
1832 */
1833 #define ILL_FREE_OK(ill) \
1834 ((ill)->ill_ire_cnt == 0 && (ill)->ill_ilm_cnt == 0 && \
1835 (ill)->ill_ncec_cnt == 0 && (ill)->ill_nce_cnt == 0)
1836
1837 /*
1838 * An ipif/ill can be marked down only when the ire and ncec references
1839 * to that ipif/ill goes to zero. ILL_DOWN_OK() is a necessary condition
1840 * quiescence checks. See comments above IPIF_DOWN_OK for details
1841 * on why ires and nces are selectively considered for this macro.
1842 */
1843 #define ILL_DOWN_OK(ill) \
1844 (ill->ill_ire_cnt == 0 && ill->ill_ncec_cnt == 0 && \
1845 ill->ill_nce_cnt == 0)
1846
1847 /*
1848 * The following table lists the protection levels of the various members
1849 * of the ill_t. Same notation as that used for ipif_t above is used.
1850 *
1851 * Write Read
1852 *
1853 * ill_ifptr ill_g_lock + s Write once
1854 * ill_rq ipsq Write once
1855 * ill_wq ipsq Write once
1856 *
1857 * ill_error ipsq None
1858 * ill_ipif ill_g_lock + ipsq ill_g_lock OR ipsq
1859 * ill_ipif_up_count ill_lock + ipsq ill_lock OR ipsq
1860 * ill_max_frag ill_lock ill_lock
1861 * ill_current_frag ill_lock ill_lock
1862 *
1863 * ill_name ill_g_lock + ipsq Write once
1864 * ill_name_length ill_g_lock + ipsq Write once
1865 * ill_ndd_name ipsq Write once
1866 * ill_net_type ipsq Write once
1867 * ill_ppa ill_g_lock + ipsq Write once
1868 * ill_sap ipsq + down ill Write once
1869 * ill_sap_length ipsq + down ill Write once
1870 * ill_phys_addr_length ipsq + down ill Write once
1871 *
1872 * ill_bcast_addr_length ipsq ipsq
1873 * ill_mactype ipsq ipsq
1874 * ill_frag_ptr ipsq ipsq
1875 *
1876 * ill_frag_timer_id ill_lock ill_lock
1877 * ill_frag_hash_tbl ipsq up ill
1878 * ill_ilm ill_mcast_lock(WRITER) ill_mcast_lock(READER)
1879 * ill_global_timer ill_mcast_lock(WRITER) ill_mcast_lock(READER)
1880 * ill_mcast_type ill_mcast_lock(WRITER) ill_mcast_lock(READER)
1881 * ill_mcast_v1_time ill_mcast_lock(WRITER) ill_mcast_lock(READER)
1882 * ill_mcast_v2_time ill_mcast_lock(WRITER) ill_mcast_lock(READER)
1883 * ill_mcast_v1_tset ill_mcast_lock(WRITER) ill_mcast_lock(READER)
1884 * ill_mcast_v2_tset ill_mcast_lock(WRITER) ill_mcast_lock(READER)
1885 * ill_mcast_rv ill_mcast_lock(WRITER) ill_mcast_lock(READER)
1886 * ill_mcast_qi ill_mcast_lock(WRITER) ill_mcast_lock(READER)
1887 *
1888 * ill_down_mp ipsq ipsq
1889 * ill_dlpi_deferred ill_lock ill_lock
1890 * ill_dlpi_pending ipsq + ill_lock ipsq or ill_lock or
1891 * absence of ipsq writer.
1892 * ill_phys_addr_mp ipsq + down ill only when ill is up
1893 * ill_mcast_deferred ill_lock ill_lock
1894 * ill_phys_addr ipsq + down ill only when ill is up
1895 * ill_dest_addr_mp ipsq + down ill only when ill is up
1896 * ill_dest_addr ipsq + down ill only when ill is up
1897 *
1898 * ill_state_flags ill_lock ill_lock
1899 * exclusive bit flags ipsq_t ipsq_t
1900 * shared bit flags ill_lock ill_lock
1901 *
1902 * ill_muxid ipsq Not atomic
1903 *
1904 * ill_ipf_gen Not atomic
1905 * ill_frag_count atomics atomics
1906 * ill_type ipsq + down ill only when ill is up
1907 * ill_dlpi_multicast_state ill_lock ill_lock
1908 * ill_dlpi_fastpath_state ill_lock ill_lock
1909 * ill_dlpi_capab_state ipsq ipsq
1910 * ill_max_hops ipsq Not atomic
1911 *
1912 * ill_mtu ill_lock None
1913 * ill_mc_mtu ill_lock None
1914 *
1915 * ill_user_mtu ipsq + ill_lock ill_lock
1916 * ill_reachable_time ipsq + ill_lock ill_lock
1917 * ill_reachable_retrans_time ipsq + ill_lock ill_lock
1918 * ill_max_buf ipsq + ill_lock ill_lock
1919 *
1920 * Next 2 fields need ill_lock because of the get ioctls. They should not
1921 * report partially updated results without executing in the ipsq.
1922 * ill_token ipsq + ill_lock ill_lock
1923 * ill_token_length ipsq + ill_lock ill_lock
1924 * ill_dest_token ipsq + down ill only when ill is up
1925 * ill_xmit_count ipsq + down ill write once
1926 * ill_ip6_mib ipsq + down ill only when ill is up
1927 * ill_icmp6_mib ipsq + down ill only when ill is up
1928 *
1929 * ill_phyint ipsq, ill_g_lock, ill_lock Any of them
1930 * ill_flags ill_lock ill_lock
1931 * ill_nd_lla_mp ipsq + down ill only when ill is up
1932 * ill_nd_lla ipsq + down ill only when ill is up
1933 * ill_nd_lla_len ipsq + down ill only when ill is up
1934 * ill_phys_addr_pend ipsq + down ill only when ill is up
1935 * ill_ifname_pending_err ipsq ipsq
1936 * ill_avl_byppa ipsq, ill_g_lock write once
1937 *
1938 * ill_fastpath_list ill_lock ill_lock
1939 * ill_refcnt ill_lock ill_lock
1940 * ill_ire_cnt ill_lock ill_lock
1941 * ill_cv ill_lock ill_lock
1942 * ill_ncec_cnt ill_lock ill_lock
1943 * ill_nce_cnt ill_lock ill_lock
1944 * ill_ilm_cnt ill_lock ill_lock
1945 * ill_src_ipif ill_g_lock ill_g_lock
1946 * ill_trace ill_lock ill_lock
1947 * ill_usesrc_grp_next ill_g_usesrc_lock ill_g_usesrc_lock
1948 * ill_dhcpinit atomics atomics
1949 * ill_flownotify_mh write once write once
1950 * ill_capab_pending_cnt ipsq ipsq
1951 * ill_ipallmulti_cnt ill_lock ill_lock
1952 * ill_ipallmulti_ilm ill_lock ill_lock
1953 * ill_saved_ire_mp ill_saved_ire_lock ill_saved_ire_lock
1954 * ill_saved_ire_cnt ill_saved_ire_lock ill_saved_ire_lock
1955 * ill_arl ??? ???
1956 * ill_ire_multicast ipsq + quiescent none
1957 * ill_bound_ipif ipsq ipsq
1958 * ill_actnode ipsq + ipmp_lock ipsq OR ipmp_lock
1959 * ill_grpnode ipsq + ill_g_lock ipsq OR ill_g_lock
1960 * ill_src_ipif ill_g_lock ill_g_lock
1961 * ill_move_ipif ipsq ipsq
1962 * ill_nom_cast ipsq ipsq OR advisory
1963 * ill_refresh_tid ill_lock ill_lock
1964 * ill_grp (for IPMP ill) write once write once
1965 * ill_grp (for underlying ill) ipsq + ill_g_lock ipsq OR ill_g_lock
1966 * ill_grp_pending ill_mcast_serializer ill_mcast_serializer
1967 * ill_mrouter_cnt atomics atomics
1968 * ill_mphysaddr_list ill_lock ill_lock
1969 *
1970 * NOTE: It's OK to make heuristic decisions on an underlying interface
1971 * by using IS_UNDER_IPMP() or comparing ill_grp's raw pointer value.
1972 */
1973
1974 /*
1975 * For ioctl restart mechanism see ip_reprocess_ioctl()
1976 */
1977 struct ip_ioctl_cmd_s;
1978
1979 typedef int (*ifunc_t)(ipif_t *, struct sockaddr_in *, queue_t *, mblk_t *,
1980 struct ip_ioctl_cmd_s *, void *);
1981
1982 typedef struct ip_ioctl_cmd_s {
1983 int ipi_cmd;
1984 size_t ipi_copyin_size;
1985 uint_t ipi_flags;
1986 uint_t ipi_cmd_type;
1987 ifunc_t ipi_func;
1988 ifunc_t ipi_func_restart;
1989 } ip_ioctl_cmd_t;
1990
1991 /*
1992 * ipi_cmd_type:
1993 *
1994 * IF_CMD 1 old style ifreq cmd
1995 * LIF_CMD 2 new style lifreq cmd
1996 * ARP_CMD 3 arpreq cmd
1997 * XARP_CMD 4 xarpreq cmd
1998 * MSFILT_CMD 5 multicast source filter cmd
1999 * MISC_CMD 6 misc cmd (not a more specific one above)
2000 */
2001
2002 enum { IF_CMD = 1, LIF_CMD, ARP_CMD, XARP_CMD, MSFILT_CMD, MISC_CMD };
2003
2004 #define IPI_DONTCARE 0 /* For ioctl encoded values that don't matter */
2005
2006 /* Flag values in ipi_flags */
2007 #define IPI_PRIV 0x1 /* Root only command */
2008 #define IPI_MODOK 0x2 /* Permitted on mod instance of IP */
2009 #define IPI_WR 0x4 /* Need to grab writer access */
2010 #define IPI_GET_CMD 0x8 /* branch to mi_copyout on success */
2011 /* unused 0x10 */
2012 #define IPI_NULL_BCONT 0x20 /* ioctl has not data and hence no b_cont */
2013
2014 extern ip_ioctl_cmd_t ip_ndx_ioctl_table[];
2015 extern ip_ioctl_cmd_t ip_misc_ioctl_table[];
2016 extern int ip_ndx_ioctl_count;
2017 extern int ip_misc_ioctl_count;
2018
2019 /* Passed down by ARP to IP during I_PLINK/I_PUNLINK */
2020 typedef struct ipmx_s {
2021 char ipmx_name[LIFNAMSIZ]; /* if name */
2022 uint_t
2023 ipmx_arpdev_stream : 1, /* This is the arp stream */
2024 ipmx_notused : 31;
2025 } ipmx_t;
2026
2027 /*
2028 * State for detecting if a driver supports certain features.
2029 * Support for DL_ENABMULTI_REQ uses ill_dlpi_multicast_state.
2030 * Support for DLPI M_DATA fastpath uses ill_dlpi_fastpath_state.
2031 */
2032 #define IDS_UNKNOWN 0 /* No DLPI request sent */
2033 #define IDS_INPROGRESS 1 /* DLPI request sent */
2034 #define IDS_OK 2 /* DLPI request completed successfully */
2035 #define IDS_FAILED 3 /* DLPI request failed */
2036
2037 /* Support for DL_CAPABILITY_REQ uses ill_dlpi_capab_state. */
2038 enum {
2039 IDCS_UNKNOWN,
2040 IDCS_PROBE_SENT,
2041 IDCS_OK,
2042 IDCS_RESET_SENT,
2043 IDCS_RENEG,
2044 IDCS_FAILED
2045 };
2046
2047 /* Extended NDP Management Structure */
2048 typedef struct ipndp_s {
2049 ndgetf_t ip_ndp_getf;
2050 ndsetf_t ip_ndp_setf;
2051 caddr_t ip_ndp_data;
2052 char *ip_ndp_name;
2053 } ipndp_t;
2054
2055 /* IXA Notification types */
2056 typedef enum {
2057 IXAN_LSO, /* LSO capability change */
2058 IXAN_PMTU, /* PMTU change */
2059 IXAN_ZCOPY /* ZEROCOPY capability change */
2060 } ixa_notify_type_t;
2061
2062 typedef uint_t ixa_notify_arg_t;
2063
2064 typedef void (*ixa_notify_t)(void *, ip_xmit_attr_t *ixa, ixa_notify_type_t,
2065 ixa_notify_arg_t);
2066
2067 /*
2068 * Attribute flags that are common to the transmit and receive attributes
2069 */
2070 #define IAF_IS_IPV4 0x80000000 /* ipsec_*_v4 */
2071 #define IAF_TRUSTED_ICMP 0x40000000 /* ipsec_*_icmp_loopback */
2072 #define IAF_NO_LOOP_ZONEID_SET 0x20000000 /* Zone that shouldn't have */
2073 /* a copy */
2074 #define IAF_LOOPBACK_COPY 0x10000000 /* For multi and broadcast */
2075
2076 #define IAF_MASK 0xf0000000 /* Flags that are common */
2077
2078 /*
2079 * Transmit side attributes used between the transport protocols and IP as
2080 * well as inside IP. It is also used to cache information in the conn_t i.e.
2081 * replaces conn_ire and the IPsec caching in the conn_t.
2082 */
2083 struct ip_xmit_attr_s {
2084 iaflags_t ixa_flags; /* IXAF_*. See below */
2085
2086 uint32_t ixa_free_flags; /* IXA_FREE_*. See below */
2087 uint32_t ixa_refcnt; /* Using atomics */
2088
2089 /*
2090 * Always initialized independently of ixa_flags settings.
2091 * Used by ip_xmit so we keep them up front for cache locality.
2092 */
2093 uint32_t ixa_xmit_hint; /* For ECMP and GLD TX ring fanout */
2094 uint_t ixa_pktlen; /* Always set. For frag and stats */
2095 zoneid_t ixa_zoneid; /* Assumed always set */
2096
2097 /* Always set for conn_ip_output(); might be stale */
2098 /*
2099 * Since TCP keeps the conn_t around past the process going away
2100 * we need to use the "notr" (e.g, ire_refhold_notr) for ixa_ire,
2101 * ixa_nce, and ixa_dce.
2102 */
2103 ire_t *ixa_ire; /* Forwarding table entry */
2104 uint_t ixa_ire_generation;
2105 nce_t *ixa_nce; /* Neighbor cache entry */
2106 dce_t *ixa_dce; /* Destination cache entry */
2107 uint_t ixa_dce_generation;
2108 uint_t ixa_src_generation; /* If IXAF_VERIFY_SOURCE */
2109
2110 uint32_t ixa_src_preferences; /* prefs for src addr select */
2111 uint32_t ixa_pmtu; /* IXAF_VERIFY_PMTU */
2112
2113 /* Set by ULP if IXAF_VERIFY_PMTU; otherwise set by IP */
2114 uint32_t ixa_fragsize;
2115
2116 int8_t ixa_use_min_mtu; /* IXAF_USE_MIN_MTU values */
2117
2118 pfirepostfrag_t ixa_postfragfn; /* Set internally in IP */
2119
2120 in6_addr_t ixa_nexthop_v6; /* IXAF_NEXTHOP_SET */
2121 #define ixa_nexthop_v4 V4_PART_OF_V6(ixa_nexthop_v6)
2122
2123 zoneid_t ixa_no_loop_zoneid; /* IXAF_NO_LOOP_ZONEID_SET */
2124
2125 uint_t ixa_scopeid; /* For IPv6 link-locals */
2126
2127 uint_t ixa_broadcast_ttl; /* IXAF_BROACAST_TTL_SET */
2128
2129 uint_t ixa_multicast_ttl; /* Assumed set for multicast */
2130 uint_t ixa_multicast_ifindex; /* Assumed set for multicast */
2131 ipaddr_t ixa_multicast_ifaddr; /* Assumed set for multicast */
2132
2133 int ixa_raw_cksum_offset; /* If IXAF_SET_RAW_CKSUM */
2134
2135 uint32_t ixa_ident; /* For IPv6 fragment header */
2136
2137 uint64_t ixa_conn_id; /* Used by DTrace */
2138 /*
2139 * Cached LSO information.
2140 */
2141 ill_lso_capab_t ixa_lso_capab; /* Valid when IXAF_LSO_CAPAB */
2142
2143 uint64_t ixa_ipsec_policy_gen; /* Generation from iph_gen */
2144 /*
2145 * The following IPsec fields are only initialized when
2146 * IXAF_IPSEC_SECURE is set. Otherwise they contain garbage.
2147 */
2148 ipsec_latch_t *ixa_ipsec_latch; /* Just the ids */
2149 struct ipsa_s *ixa_ipsec_ah_sa; /* Hard reference SA for AH */
2150 struct ipsa_s *ixa_ipsec_esp_sa; /* Hard reference SA for ESP */
2151 struct ipsec_policy_s *ixa_ipsec_policy; /* why are we here? */
2152 struct ipsec_action_s *ixa_ipsec_action; /* For reflected packets */
2153 ipsa_ref_t ixa_ipsec_ref[2]; /* Soft reference to SA */
2154 /* 0: ESP, 1: AH */
2155
2156 /*
2157 * The selectors here are potentially different than the SPD rule's
2158 * selectors, and we need to have both available for IKEv2.
2159 *
2160 * NOTE: "Source" and "Dest" are w.r.t. outbound datagrams. Ports can
2161 * be zero, and the protocol number is needed to make the ports
2162 * significant.
2163 */
2164 uint16_t ixa_ipsec_src_port; /* Source port number of d-gram. */
2165 uint16_t ixa_ipsec_dst_port; /* Destination port number of d-gram. */
2166 uint8_t ixa_ipsec_icmp_type; /* ICMP type of d-gram */
2167 uint8_t ixa_ipsec_icmp_code; /* ICMP code of d-gram */
2168
2169 sa_family_t ixa_ipsec_inaf; /* Inner address family */
2170 #define IXA_MAX_ADDRLEN 4 /* Max addr len. (in 32-bit words) */
2171 uint32_t ixa_ipsec_insrc[IXA_MAX_ADDRLEN]; /* Inner src address */
2172 uint32_t ixa_ipsec_indst[IXA_MAX_ADDRLEN]; /* Inner dest address */
2173 uint8_t ixa_ipsec_insrcpfx; /* Inner source prefix */
2174 uint8_t ixa_ipsec_indstpfx; /* Inner destination prefix */
2175
2176 uint8_t ixa_ipsec_proto; /* IP protocol number for d-gram. */
2177
2178 /* Always initialized independently of ixa_flags settings */
2179 uint_t ixa_ifindex; /* Assumed always set */
2180 uint16_t ixa_ip_hdr_length; /* Points to ULP header */
2181 uint8_t ixa_protocol; /* Protocol number for ULP cksum */
2182 ts_label_t *ixa_tsl; /* Always set. NULL if not TX */
2183 ip_stack_t *ixa_ipst; /* Always set */
2184 uint32_t ixa_extra_ident; /* Set if LSO */
2185 cred_t *ixa_cred; /* For getpeerucred */
2186 pid_t ixa_cpid; /* For getpeerucred */
2187
2188 #ifdef DEBUG
2189 kthread_t *ixa_curthread; /* For serialization assert */
2190 #endif
2191 squeue_t *ixa_sqp; /* Set from conn_sqp as a hint */
2192 uintptr_t ixa_cookie; /* cookie to use for tx flow control */
2193
2194 /*
2195 * Must be set by ULP if any of IXAF_VERIFY_LSO, IXAF_VERIFY_PMTU,
2196 * or IXAF_VERIFY_ZCOPY is set.
2197 */
2198 ixa_notify_t ixa_notify; /* Registered upcall notify function */
2199 void *ixa_notify_cookie; /* ULP cookie for ixa_notify */
2200
2201 uint_t ixa_tcpcleanup; /* Used by conn_ixa_cleanup */
2202 };
2203
2204 /*
2205 * Flags to indicate which transmit attributes are set.
2206 * Split into "xxx_SET" ones which indicate that the "xxx" field it set, and
2207 * single flags.
2208 */
2209 #define IXAF_REACH_CONF 0x00000001 /* Reachability confirmation */
2210 #define IXAF_BROADCAST_TTL_SET 0x00000002 /* ixa_broadcast_ttl valid */
2211 #define IXAF_SET_SOURCE 0x00000004 /* Replace if broadcast */
2212 #define IXAF_USE_MIN_MTU 0x00000008 /* IPV6_USE_MIN_MTU */
2213
2214 #define IXAF_DONTFRAG 0x00000010 /* IP*_DONTFRAG */
2215 #define IXAF_VERIFY_PMTU 0x00000020 /* ixa_pmtu/ixa_fragsize set */
2216 #define IXAF_PMTU_DISCOVERY 0x00000040 /* Create/use PMTU state */
2217 #define IXAF_MULTICAST_LOOP 0x00000080 /* IP_MULTICAST_LOOP */
2218
2219 #define IXAF_IPSEC_SECURE 0x00000100 /* Need IPsec processing */
2220 #define IXAF_UCRED_TSL 0x00000200 /* ixa_tsl from SCM_UCRED */
2221 #define IXAF_DONTROUTE 0x00000400 /* SO_DONTROUTE */
2222 #define IXAF_NO_IPSEC 0x00000800 /* Ignore policy */
2223
2224 #define IXAF_PMTU_TOO_SMALL 0x00001000 /* PMTU too small */
2225 #define IXAF_SET_ULP_CKSUM 0x00002000 /* Calculate ULP checksum */
2226 #define IXAF_VERIFY_SOURCE 0x00004000 /* Check that source is ok */
2227 #define IXAF_NEXTHOP_SET 0x00008000 /* ixa_nexthop set */
2228
2229 #define IXAF_PMTU_IPV4_DF 0x00010000 /* Set IPv4 DF */
2230 #define IXAF_NO_DEV_FLOW_CTL 0x00020000 /* Protocol needs no flow ctl */
2231 #define IXAF_NO_TTL_CHANGE 0x00040000 /* Internal to IP */
2232 #define IXAF_IPV6_ADD_FRAGHDR 0x00080000 /* Add fragment header */
2233
2234 #define IXAF_IPSEC_TUNNEL 0x00100000 /* Tunnel mode */
2235 #define IXAF_NO_PFHOOK 0x00200000 /* Skip xmit pfhook */
2236 #define IXAF_NO_TRACE 0x00400000 /* When back from ARP/ND */
2237 #define IXAF_SCOPEID_SET 0x00800000 /* ixa_scopeid set */
2238
2239 #define IXAF_MULTIRT_MULTICAST 0x01000000 /* MULTIRT for multicast */
2240 #define IXAF_NO_HW_CKSUM 0x02000000 /* Force software cksum */
2241 #define IXAF_SET_RAW_CKSUM 0x04000000 /* Use ixa_raw_cksum_offset */
2242 #define IXAF_IPSEC_GLOBAL_POLICY 0x08000000 /* Policy came from global */
2243
2244 /* Note the following uses bits 0x10000000 through 0x80000000 */
2245 #define IXAF_IS_IPV4 IAF_IS_IPV4
2246 #define IXAF_TRUSTED_ICMP IAF_TRUSTED_ICMP
2247 #define IXAF_NO_LOOP_ZONEID_SET IAF_NO_LOOP_ZONEID_SET
2248 #define IXAF_LOOPBACK_COPY IAF_LOOPBACK_COPY
2249
2250 /* Note: use the upper 32 bits */
2251 #define IXAF_VERIFY_LSO 0x100000000 /* Check LSO capability */
2252 #define IXAF_LSO_CAPAB 0x200000000 /* Capable of LSO */
2253 #define IXAF_VERIFY_ZCOPY 0x400000000 /* Check Zero Copy capability */
2254 #define IXAF_ZCOPY_CAPAB 0x800000000 /* Capable of ZEROCOPY */
2255
2256 /*
2257 * The normal flags for sending packets e.g., icmp errors
2258 */
2259 #define IXAF_BASIC_SIMPLE_V4 \
2260 (IXAF_SET_ULP_CKSUM | IXAF_IS_IPV4 | IXAF_VERIFY_SOURCE)
2261 #define IXAF_BASIC_SIMPLE_V6 (IXAF_SET_ULP_CKSUM | IXAF_VERIFY_SOURCE)
2262
2263 /*
2264 * Normally these fields do not have a hold. But in some cases they do, for
2265 * instance when we've gone through ip_*_attr_to/from_mblk.
2266 * We use ixa_free_flags to indicate that they have a hold and need to be
2267 * released on cleanup.
2268 */
2269 #define IXA_FREE_CRED 0x00000001 /* ixa_cred needs to be rele */
2270 #define IXA_FREE_TSL 0x00000002 /* ixa_tsl needs to be rele */
2271
2272 /*
2273 * Trivial state machine used to synchronize IXA cleanup for TCP connections.
2274 * See conn_ixa_cleanup().
2275 */
2276 #define IXATC_IDLE 0x00000000
2277 #define IXATC_INPROGRESS 0x00000001
2278 #define IXATC_COMPLETE 0x00000002
2279
2280 /*
2281 * Simplistic way to set the ixa_xmit_hint for locally generated traffic
2282 * and forwarded traffic. The shift amount are based on the size of the
2283 * structs to discard the low order bits which don't have much if any variation
2284 * (coloring in kmem_cache_alloc might provide some variation).
2285 *
2286 * Basing the locally generated hint on the address of the conn_t means that
2287 * the packets from the same socket/connection do not get reordered.
2288 * Basing the hint for forwarded traffic on the ill_ring_t means that
2289 * packets from the same NIC+ring are likely to use the same outbound ring
2290 * hence we get low contention on the ring in the transmitting driver.
2291 */
2292 #define CONN_TO_XMIT_HINT(connp) ((uint32_t)(((uintptr_t)connp) >> 11))
2293 #define ILL_RING_TO_XMIT_HINT(ring) ((uint32_t)(((uintptr_t)ring) >> 7))
2294
2295 /*
2296 * IP set Destination Flags used by function ip_set_destination,
2297 * ip_attr_connect, and conn_connect.
2298 */
2299 #define IPDF_ALLOW_MCBC 0x1 /* Allow multi/broadcast */
2300 #define IPDF_VERIFY_DST 0x2 /* Verify destination addr */
2301 #define IPDF_SELECT_SRC 0x4 /* Select source address */
2302 #define IPDF_LSO 0x8 /* Try LSO */
2303 #define IPDF_IPSEC 0x10 /* Set IPsec policy */
2304 #define IPDF_ZONE_IS_GLOBAL 0x20 /* From conn_zone_is_global */
2305 #define IPDF_ZCOPY 0x40 /* Try ZEROCOPY */
2306 #define IPDF_UNIQUE_DCE 0x80 /* Get a per-destination DCE */
2307
2308 /*
2309 * Receive side attributes used between the transport protocols and IP as
2310 * well as inside IP.
2311 */
2312 struct ip_recv_attr_s {
2313 iaflags_t ira_flags; /* See below */
2314
2315 uint32_t ira_free_flags; /* IRA_FREE_*. See below */
2316
2317 /*
2318 * This is a hint for TCP SYN packets.
2319 * Always initialized independently of ira_flags settings
2320 */
2321 squeue_t *ira_sqp;
2322 ill_rx_ring_t *ira_ring; /* Internal to IP */
2323
2324 /* For ip_accept_tcp when IRAF_TARGET_SQP is set */
2325 squeue_t *ira_target_sqp;
2326 mblk_t *ira_target_sqp_mp;
2327
2328 /* Always initialized independently of ira_flags settings */
2329 uint32_t ira_xmit_hint; /* For ECMP and GLD TX ring fanout */
2330 zoneid_t ira_zoneid; /* ALL_ZONES unless local delivery */
2331 uint_t ira_pktlen; /* Always set. For frag and stats */
2332 uint16_t ira_ip_hdr_length; /* Points to ULP header */
2333 uint8_t ira_protocol; /* Protocol number for ULP cksum */
2334 uint_t ira_rifindex; /* Received ifindex */
2335 uint_t ira_ruifindex; /* Received upper ifindex */
2336 ts_label_t *ira_tsl; /* Always set. NULL if not TX */
2337 /*
2338 * ira_rill and ira_ill is set inside IP, but not when conn_recv is
2339 * called; ULPs should use ira_ruifindex instead.
2340 */
2341 ill_t *ira_rill; /* ill where packet came */
2342 ill_t *ira_ill; /* ill where IP address hosted */
2343 cred_t *ira_cred; /* For getpeerucred */
2344 pid_t ira_cpid; /* For getpeerucred */
2345
2346 /* Used when IRAF_VERIFIED_SRC is set; this source was ok */
2347 ipaddr_t ira_verified_src;
2348
2349 /*
2350 * The following IPsec fields are only initialized when
2351 * IRAF_IPSEC_SECURE is set. Otherwise they contain garbage.
2352 */
2353 struct ipsec_action_s *ira_ipsec_action; /* how we made it in.. */
2354 struct ipsa_s *ira_ipsec_ah_sa; /* SA for AH */
2355 struct ipsa_s *ira_ipsec_esp_sa; /* SA for ESP */
2356
2357 ipaddr_t ira_mroute_tunnel; /* IRAF_MROUTE_TUNNEL_SET */
2358
2359 zoneid_t ira_no_loop_zoneid; /* IRAF_NO_LOOP_ZONEID_SET */
2360
2361 uint32_t ira_esp_udp_ports; /* IRAF_ESP_UDP_PORTS */
2362
2363 /*
2364 * For IP_RECVSLLA and ip_ndp_conflict/find_solicitation.
2365 * Same size as max for sockaddr_dl
2366 */
2367 #define IRA_L2SRC_SIZE 244
2368 uint8_t ira_l2src[IRA_L2SRC_SIZE]; /* If IRAF_L2SRC_SET */
2369
2370 /*
2371 * Local handle that we use to do lazy setting of ira_l2src.
2372 * We defer setting l2src until needed but we do before any
2373 * ip_input pullupmsg or copymsg.
2374 */
2375 struct mac_header_info_s *ira_mhip; /* Could be NULL */
2376 };
2377
2378 /*
2379 * Flags to indicate which receive attributes are set.
2380 */
2381 #define IRAF_SYSTEM_LABELED 0x00000001 /* is_system_labeled() */
2382 #define IRAF_IPV4_OPTIONS 0x00000002 /* Performance */
2383 #define IRAF_MULTICAST 0x00000004 /* Was multicast at L3 */
2384 #define IRAF_BROADCAST 0x00000008 /* Was broadcast at L3 */
2385 #define IRAF_MULTIBROADCAST (IRAF_MULTICAST|IRAF_BROADCAST)
2386
2387 #define IRAF_LOOPBACK 0x00000010 /* Looped back by IP */
2388 #define IRAF_VERIFY_IP_CKSUM 0x00000020 /* Need to verify IP */
2389 #define IRAF_VERIFY_ULP_CKSUM 0x00000040 /* Need to verify TCP,UDP,etc */
2390 #define IRAF_SCTP_CSUM_ERR 0x00000080 /* sctp pkt has failed chksum */
2391
2392 #define IRAF_IPSEC_SECURE 0x00000100 /* Passed AH and/or ESP */
2393 #define IRAF_DHCP_UNICAST 0x00000200
2394 #define IRAF_IPSEC_DECAPS 0x00000400 /* Was packet decapsulated */
2395 /* from a matching inner packet? */
2396 #define IRAF_TARGET_SQP 0x00000800 /* ira_target_sqp is set */
2397 #define IRAF_VERIFIED_SRC 0x00001000 /* ira_verified_src set */
2398 #define IRAF_RSVP 0x00002000 /* RSVP packet for rsvpd */
2399 #define IRAF_MROUTE_TUNNEL_SET 0x00004000 /* From ip_mroute_decap */
2400 #define IRAF_PIM_REGISTER 0x00008000 /* From register_mforward */
2401
2402 #define IRAF_TX_MAC_EXEMPTABLE 0x00010000 /* Allow MAC_EXEMPT readdown */
2403 #define IRAF_TX_SHARED_ADDR 0x00020000 /* Arrived on ALL_ZONES addr */
2404 #define IRAF_ESP_UDP_PORTS 0x00040000 /* NAT-traversal packet */
2405 #define IRAF_NO_HW_CKSUM 0x00080000 /* Force software cksum */
2406
2407 #define IRAF_ICMP_ERROR 0x00100000 /* Send to conn_recvicmp */
2408 #define IRAF_ROUTER_ALERT 0x00200000 /* IPv6 router alert */
2409 #define IRAF_L2SRC_SET 0x00400000 /* ira_l2src has been set */
2410 #define IRAF_L2SRC_LOOPBACK 0x00800000 /* Came from us */
2411
2412 #define IRAF_L2DST_MULTICAST 0x01000000 /* Multicast at L2 */
2413 #define IRAF_L2DST_BROADCAST 0x02000000 /* Broadcast at L2 */
2414 /* Unused 0x04000000 */
2415 /* Unused 0x08000000 */
2416
2417 /* Below starts with 0x10000000 */
2418 #define IRAF_IS_IPV4 IAF_IS_IPV4
2419 #define IRAF_TRUSTED_ICMP IAF_TRUSTED_ICMP
2420 #define IRAF_NO_LOOP_ZONEID_SET IAF_NO_LOOP_ZONEID_SET
2421 #define IRAF_LOOPBACK_COPY IAF_LOOPBACK_COPY
2422
2423 /*
2424 * Normally these fields do not have a hold. But in some cases they do, for
2425 * instance when we've gone through ip_*_attr_to/from_mblk.
2426 * We use ira_free_flags to indicate that they have a hold and need to be
2427 * released on cleanup.
2428 */
2429 #define IRA_FREE_CRED 0x00000001 /* ira_cred needs to be rele */
2430 #define IRA_FREE_TSL 0x00000002 /* ira_tsl needs to be rele */
2431
2432 /*
2433 * Optional destination cache entry for path MTU information,
2434 * and ULP metrics.
2435 */
2436 struct dce_s {
2437 uint_t dce_generation; /* Changed since cached? */
2438 uint_t dce_flags; /* See below */
2439 uint_t dce_ipversion; /* IPv4/IPv6 version */
2440 uint32_t dce_pmtu; /* Path MTU if DCEF_PMTU */
2441 uint32_t dce_ident; /* Per destination IP ident. */
2442 iulp_t dce_uinfo; /* Metrics if DCEF_UINFO */
2443
2444 struct dce_s *dce_next;
2445 struct dce_s **dce_ptpn;
2446 struct dcb_s *dce_bucket;
2447
2448 union {
2449 in6_addr_t dceu_v6addr;
2450 ipaddr_t dceu_v4addr;
2451 } dce_u;
2452 #define dce_v4addr dce_u.dceu_v4addr
2453 #define dce_v6addr dce_u.dceu_v6addr
2454 /* Note that for IPv6+IPMP we use the ifindex for the upper interface */
2455 uint_t dce_ifindex; /* For IPv6 link-locals */
2456
2457 kmutex_t dce_lock;
2458 uint_t dce_refcnt;
2459 uint64_t dce_last_change_time; /* Path MTU. In seconds */
2460
2461 ip_stack_t *dce_ipst; /* Does not have a netstack_hold */
2462 };
2463
2464 /*
2465 * Values for dce_generation.
2466 *
2467 * If a DCE has DCE_GENERATION_CONDEMNED, the last dce_refrele should delete
2468 * it.
2469 *
2470 * DCE_GENERATION_VERIFY is never stored in dce_generation but it is
2471 * stored in places that cache DCE (such as ixa_dce_generation).
2472 * It is used as a signal that the cache is stale and needs to be reverified.
2473 */
2474 #define DCE_GENERATION_CONDEMNED 0
2475 #define DCE_GENERATION_VERIFY 1
2476 #define DCE_GENERATION_INITIAL 2
2477 #define DCE_IS_CONDEMNED(dce) \
2478 ((dce)->dce_generation == DCE_GENERATION_CONDEMNED)
2479
2480
2481 /*
2482 * Values for ips_src_generation.
2483 *
2484 * SRC_GENERATION_VERIFY is never stored in ips_src_generation but it is
2485 * stored in places that cache IREs (ixa_src_generation). It is used as a
2486 * signal that the cache is stale and needs to be reverified.
2487 */
2488 #define SRC_GENERATION_VERIFY 0
2489 #define SRC_GENERATION_INITIAL 1
2490
2491 /*
2492 * The kernel stores security attributes of all gateways in a database made
2493 * up of one or more tsol_gcdb_t elements. Each tsol_gcdb_t contains the
2494 * security-related credentials of the gateway. More than one gateways may
2495 * share entries in the database.
2496 *
2497 * The tsol_gc_t structure represents the gateway to credential association,
2498 * and refers to an entry in the database. One or more tsol_gc_t entities are
2499 * grouped together to form one or more tsol_gcgrp_t, each representing the
2500 * list of security attributes specific to the gateway. A gateway may be
2501 * associated with at most one credentials group.
2502 */
2503 struct tsol_gcgrp_s;
2504
2505 extern uchar_t ip6opt_ls; /* TX IPv6 enabler */
2506
2507 /*
2508 * Gateway security credential record.
2509 */
2510 typedef struct tsol_gcdb_s {
2511 uint_t gcdb_refcnt; /* reference count */
2512 struct rtsa_s gcdb_attr; /* security attributes */
2513 #define gcdb_mask gcdb_attr.rtsa_mask
2514 #define gcdb_doi gcdb_attr.rtsa_doi
2515 #define gcdb_slrange gcdb_attr.rtsa_slrange
2516 } tsol_gcdb_t;
2517
2518 /*
2519 * Gateway to credential association.
2520 */
2521 typedef struct tsol_gc_s {
2522 uint_t gc_refcnt; /* reference count */
2523 struct tsol_gcgrp_s *gc_grp; /* pointer to group */
2524 struct tsol_gc_s *gc_prev; /* previous in list */
2525 struct tsol_gc_s *gc_next; /* next in list */
2526 tsol_gcdb_t *gc_db; /* pointer to actual credentials */
2527 } tsol_gc_t;
2528
2529 /*
2530 * Gateway credentials group address.
2531 */
2532 typedef struct tsol_gcgrp_addr_s {
2533 int ga_af; /* address family */
2534 in6_addr_t ga_addr; /* IPv4 mapped or IPv6 address */
2535 } tsol_gcgrp_addr_t;
2536
2537 /*
2538 * Gateway credentials group.
2539 */
2540 typedef struct tsol_gcgrp_s {
2541 uint_t gcgrp_refcnt; /* reference count */
2542 krwlock_t gcgrp_rwlock; /* lock to protect following */
2543 uint_t gcgrp_count; /* number of credentials */
2544 tsol_gc_t *gcgrp_head; /* first credential in list */
2545 tsol_gc_t *gcgrp_tail; /* last credential in list */
2546 tsol_gcgrp_addr_t gcgrp_addr; /* next-hop gateway address */
2547 } tsol_gcgrp_t;
2548
2549 extern kmutex_t gcgrp_lock;
2550
2551 #define GC_REFRELE(p) { \
2552 ASSERT((p)->gc_grp != NULL); \
2553 rw_enter(&(p)->gc_grp->gcgrp_rwlock, RW_WRITER); \
2554 ASSERT((p)->gc_refcnt > 0); \
2555 if (--((p)->gc_refcnt) == 0) \
2556 gc_inactive(p); \
2557 else \
2558 rw_exit(&(p)->gc_grp->gcgrp_rwlock); \
2559 }
2560
2561 #define GCGRP_REFHOLD(p) { \
2562 mutex_enter(&gcgrp_lock); \
2563 ++((p)->gcgrp_refcnt); \
2564 ASSERT((p)->gcgrp_refcnt != 0); \
2565 mutex_exit(&gcgrp_lock); \
2566 }
2567
2568 #define GCGRP_REFRELE(p) { \
2569 mutex_enter(&gcgrp_lock); \
2570 ASSERT((p)->gcgrp_refcnt > 0); \
2571 if (--((p)->gcgrp_refcnt) == 0) \
2572 gcgrp_inactive(p); \
2573 ASSERT(MUTEX_HELD(&gcgrp_lock)); \
2574 mutex_exit(&gcgrp_lock); \
2575 }
2576
2577 /*
2578 * IRE gateway security attributes structure, pointed to by tsol_ire_gw_secattr
2579 */
2580 struct tsol_tnrhc;
2581
2582 struct tsol_ire_gw_secattr_s {
2583 kmutex_t igsa_lock; /* lock to protect following */
2584 struct tsol_tnrhc *igsa_rhc; /* host entry for gateway */
2585 tsol_gc_t *igsa_gc; /* for prefix IREs */
2586 };
2587
2588 void irb_refrele_ftable(irb_t *);
2589
2590 extern struct kmem_cache *rt_entry_cache;
2591
2592 typedef struct ire4 {
2593 ipaddr_t ire4_mask; /* Mask for matching this IRE. */
2594 ipaddr_t ire4_addr; /* Address this IRE represents. */
2595 ipaddr_t ire4_gateway_addr; /* Gateway including for IRE_ONLINK */
2596 ipaddr_t ire4_setsrc_addr; /* RTF_SETSRC */
2597 } ire4_t;
2598
2599 typedef struct ire6 {
2600 in6_addr_t ire6_mask; /* Mask for matching this IRE. */
2601 in6_addr_t ire6_addr; /* Address this IRE represents. */
2602 in6_addr_t ire6_gateway_addr; /* Gateway including for IRE_ONLINK */
2603 in6_addr_t ire6_setsrc_addr; /* RTF_SETSRC */
2604 } ire6_t;
2605
2606 typedef union ire_addr {
2607 ire6_t ire6_u;
2608 ire4_t ire4_u;
2609 } ire_addr_u_t;
2610
2611 /*
2612 * Internet Routing Entry
2613 * When we have multiple identical IREs we logically add them by manipulating
2614 * ire_identical_ref and ire_delete first decrements
2615 * that and when it reaches 1 we know it is the last IRE.
2616 * "identical" is defined as being the same for:
2617 * ire_addr, ire_netmask, ire_gateway, ire_ill, ire_zoneid, and ire_type
2618 * For instance, multiple IRE_BROADCASTs for the same subnet number are
2619 * viewed as identical, and so are the IRE_INTERFACEs when there are
2620 * multiple logical interfaces (on the same ill) with the same subnet prefix.
2621 */
2622 struct ire_s {
2623 struct ire_s *ire_next; /* The hash chain must be first. */
2624 struct ire_s **ire_ptpn; /* Pointer to previous next. */
2625 uint32_t ire_refcnt; /* Number of references */
2626 ill_t *ire_ill;
2627 uint32_t ire_identical_ref; /* IRE_INTERFACE, IRE_BROADCAST */
2628 uchar_t ire_ipversion; /* IPv4/IPv6 version */
2629 ushort_t ire_type; /* Type of IRE */
2630 uint_t ire_generation; /* Generation including CONDEMNED */
2631 uint_t ire_ib_pkt_count; /* Inbound packets for ire_addr */
2632 uint_t ire_ob_pkt_count; /* Outbound packets to ire_addr */
2633 time_t ire_create_time; /* Time (in secs) IRE was created. */
2634 uint32_t ire_flags; /* flags related to route (RTF_*) */
2635 /*
2636 * ire_testhidden is TRUE for INTERFACE IREs of IS_UNDER_IPMP(ill)
2637 * interfaces
2638 */
2639 boolean_t ire_testhidden;
2640 pfirerecv_t ire_recvfn; /* Receive side handling */
2641 pfiresend_t ire_sendfn; /* Send side handling */
2642 pfirepostfrag_t ire_postfragfn; /* Bottom end of send handling */
2643
2644 uint_t ire_masklen; /* # bits in ire_mask{,_v6} */
2645 ire_addr_u_t ire_u; /* IPv4/IPv6 address info. */
2646
2647 irb_t *ire_bucket; /* Hash bucket when ire_ptphn is set */
2648 kmutex_t ire_lock;
2649 clock_t ire_last_used_time; /* For IRE_LOCAL reception */
2650 tsol_ire_gw_secattr_t *ire_gw_secattr; /* gateway security attributes */
2651 zoneid_t ire_zoneid;
2652
2653 /*
2654 * Cached information of where to send packets that match this route.
2655 * The ire_dep_* information is used to determine when ire_nce_cache
2656 * needs to be updated.
2657 * ire_nce_cache is the fastpath for the Neighbor Cache Entry
2658 * for IPv6; arp info for IPv4
2659 * Since this is a cache setup and torn down independently of
2660 * applications we need to use nce_ref{rele,hold}_notr for it.
2661 */
2662 nce_t *ire_nce_cache;
2663
2664 /*
2665 * Quick check whether the ire_type and ire_masklen indicates
2666 * that the IRE can have ire_nce_cache set i.e., whether it is
2667 * IRE_ONLINK and for a single destination.
2668 */
2669 boolean_t ire_nce_capable;
2670
2671 /*
2672 * Dependency tracking so we can safely cache IRE and NCE pointers
2673 * in offlink and onlink IREs.
2674 * These are locked under the ips_ire_dep_lock rwlock. Write held
2675 * when modifying the linkage.
2676 * ire_dep_parent (Also chain towards IRE for nexthop)
2677 * ire_dep_parent_generation: ire_generation of ire_dep_parent
2678 * ire_dep_children (From parent to first child)
2679 * ire_dep_sib_next (linked list of siblings)
2680 * ire_dep_sib_ptpn (linked list of siblings)
2681 *
2682 * The parent has a ire_refhold on each child, and each child has
2683 * an ire_refhold on its parent.
2684 * Since ire_dep_parent is a cache setup and torn down independently of
2685 * applications we need to use ire_ref{rele,hold}_notr for it.
2686 */
2687 ire_t *ire_dep_parent;
2688 ire_t *ire_dep_children;
2689 ire_t *ire_dep_sib_next;
2690 ire_t **ire_dep_sib_ptpn; /* Pointer to previous next */
2691 uint_t ire_dep_parent_generation;
2692
2693 uint_t ire_badcnt; /* Number of times ND_UNREACHABLE */
2694 uint64_t ire_last_badcnt; /* In seconds */
2695
2696 /* ire_defense* and ire_last_used_time are only used on IRE_LOCALs */
2697 uint_t ire_defense_count; /* number of ARP conflicts */
2698 uint_t ire_defense_time; /* last time defended (secs) */
2699
2700 boolean_t ire_trace_disable; /* True when alloc fails */
2701 ip_stack_t *ire_ipst; /* Does not have a netstack_hold */
2702 iulp_t ire_metrics;
2703 /*
2704 * default and prefix routes that are added without explicitly
2705 * specifying the interface are termed "unbound" routes, and will
2706 * have ire_unbound set to true.
2707 */
2708 boolean_t ire_unbound;
2709 };
2710
2711 /* IPv4 compatibility macros */
2712 #define ire_mask ire_u.ire4_u.ire4_mask
2713 #define ire_addr ire_u.ire4_u.ire4_addr
2714 #define ire_gateway_addr ire_u.ire4_u.ire4_gateway_addr
2715 #define ire_setsrc_addr ire_u.ire4_u.ire4_setsrc_addr
2716
2717 #define ire_mask_v6 ire_u.ire6_u.ire6_mask
2718 #define ire_addr_v6 ire_u.ire6_u.ire6_addr
2719 #define ire_gateway_addr_v6 ire_u.ire6_u.ire6_gateway_addr
2720 #define ire_setsrc_addr_v6 ire_u.ire6_u.ire6_setsrc_addr
2721
2722 /*
2723 * Values for ire_generation.
2724 *
2725 * If an IRE is marked with IRE_IS_CONDEMNED, the last walker of
2726 * the bucket should delete this IRE from this bucket.
2727 *
2728 * IRE_GENERATION_VERIFY is never stored in ire_generation but it is
2729 * stored in places that cache IREs (such as ixa_ire_generation and
2730 * ire_dep_parent_generation). It is used as a signal that the cache is
2731 * stale and needs to be reverified.
2732 */
2733 #define IRE_GENERATION_CONDEMNED 0
2734 #define IRE_GENERATION_VERIFY 1
2735 #define IRE_GENERATION_INITIAL 2
2736 #define IRE_IS_CONDEMNED(ire) \
2737 ((ire)->ire_generation == IRE_GENERATION_CONDEMNED)
2738
2739 /* Convenient typedefs for sockaddrs */
2740 typedef struct sockaddr_in sin_t;
2741 typedef struct sockaddr_in6 sin6_t;
2742
2743 /* Name/Value Descriptor. */
2744 typedef struct nv_s {
2745 uint64_t nv_value;
2746 char *nv_name;
2747 } nv_t;
2748
2749 #define ILL_FRAG_HASH(s, i) \
2750 ((ntohl(s) ^ ((i) ^ ((i) >> 8))) % ILL_FRAG_HASH_TBL_COUNT)
2751
2752 /*
2753 * The MAX number of allowed fragmented packets per hash bucket
2754 * calculation is based on the most common mtu size of 1500. This limit
2755 * will work well for other mtu sizes as well.
2756 */
2757 #define COMMON_IP_MTU 1500
2758 #define MAX_FRAG_MIN 10
2759 #define MAX_FRAG_PKTS(ipst) \
2760 MAX(MAX_FRAG_MIN, (2 * (ipst->ips_ip_reass_queue_bytes / \
2761 (COMMON_IP_MTU * ILL_FRAG_HASH_TBL_COUNT))))
2762
2763 /*
2764 * Maximum dups allowed per packet.
2765 */
2766 extern uint_t ip_max_frag_dups;
2767
2768 /*
2769 * Per-packet information for received packets and transmitted.
2770 * Used by the transport protocols when converting between the packet
2771 * and ancillary data and socket options.
2772 *
2773 * Note: This private data structure and related IPPF_* constant
2774 * definitions are exposed to enable compilation of some debugging tools
2775 * like lsof which use struct tcp_t in <inet/tcp.h>. This is intended to be
2776 * a temporary hack and long term alternate interfaces should be defined
2777 * to support the needs of such tools and private definitions moved to
2778 * private headers.
2779 */
2780 struct ip_pkt_s {
2781 uint_t ipp_fields; /* Which fields are valid */
2782 in6_addr_t ipp_addr; /* pktinfo src/dst addr */
2783 #define ipp_addr_v4 V4_PART_OF_V6(ipp_addr)
2784 uint_t ipp_unicast_hops; /* IPV6_UNICAST_HOPS, IP_TTL */
2785 uint_t ipp_hoplimit; /* IPV6_HOPLIMIT */
2786 uint_t ipp_hopoptslen;
2787 uint_t ipp_rthdrdstoptslen;
2788 uint_t ipp_rthdrlen;
2789 uint_t ipp_dstoptslen;
2790 uint_t ipp_fraghdrlen;
2791 ip6_hbh_t *ipp_hopopts;
2792 ip6_dest_t *ipp_rthdrdstopts;
2793 ip6_rthdr_t *ipp_rthdr;
2794 ip6_dest_t *ipp_dstopts;
2795 ip6_frag_t *ipp_fraghdr;
2796 uint8_t ipp_tclass; /* IPV6_TCLASS */
2797 uint8_t ipp_type_of_service; /* IP_TOS */
2798 uint_t ipp_ipv4_options_len; /* Len of IPv4 options */
2799 uint8_t *ipp_ipv4_options; /* Ptr to IPv4 options */
2800 uint_t ipp_label_len_v4; /* Len of TX label for IPv4 */
2801 uint8_t *ipp_label_v4; /* TX label for IPv4 */
2802 uint_t ipp_label_len_v6; /* Len of TX label for IPv6 */
2803 uint8_t *ipp_label_v6; /* TX label for IPv6 */
2804 };
2805 typedef struct ip_pkt_s ip_pkt_t;
2806
2807 extern void ip_pkt_free(ip_pkt_t *); /* free storage inside ip_pkt_t */
2808 extern ipaddr_t ip_pkt_source_route_v4(const ip_pkt_t *);
2809 extern in6_addr_t *ip_pkt_source_route_v6(const ip_pkt_t *);
2810 extern int ip_pkt_copy(ip_pkt_t *, ip_pkt_t *, int);
2811 extern void ip_pkt_source_route_reverse_v4(ip_pkt_t *);
2812
2813 /* ipp_fields values */
2814 #define IPPF_ADDR 0x0001 /* Part of in6_pktinfo: src/dst addr */
2815 #define IPPF_HOPLIMIT 0x0002 /* Overrides unicast and multicast */
2816 #define IPPF_TCLASS 0x0004 /* Overrides class in sin6_flowinfo */
2817
2818 #define IPPF_HOPOPTS 0x0010 /* ipp_hopopts set */
2819 #define IPPF_RTHDR 0x0020 /* ipp_rthdr set */
2820 #define IPPF_RTHDRDSTOPTS 0x0040 /* ipp_rthdrdstopts set */
2821 #define IPPF_DSTOPTS 0x0080 /* ipp_dstopts set */
2822
2823 #define IPPF_IPV4_OPTIONS 0x0100 /* ipp_ipv4_options set */
2824 #define IPPF_LABEL_V4 0x0200 /* ipp_label_v4 set */
2825 #define IPPF_LABEL_V6 0x0400 /* ipp_label_v6 set */
2826
2827 #define IPPF_FRAGHDR 0x0800 /* Used for IPsec receive side */
2828
2829 /*
2830 * Data structure which is passed to conn_opt_get/set.
2831 * The conn_t is included even though it can be inferred from queue_t.
2832 * setsockopt and getsockopt use conn_ixa and conn_xmit_ipp. However,
2833 * when handling ancillary data we use separate ixa and ipps.
2834 */
2835 typedef struct conn_opt_arg_s {
2836 conn_t *coa_connp;
2837 ip_xmit_attr_t *coa_ixa;
2838 ip_pkt_t *coa_ipp;
2839 boolean_t coa_ancillary; /* Ancillary data and not setsockopt */
2840 uint_t coa_changed; /* See below */
2841 } conn_opt_arg_t;
2842
2843 /*
2844 * Flags for what changed.
2845 * If we want to be more efficient in the future we can have more fine
2846 * grained flags e.g., a flag for just IP_TOS changing.
2847 * For now we either call ip_set_destination (for "route changed")
2848 * and/or conn_build_hdr_template/conn_prepend_hdr (for "header changed").
2849 */
2850 #define COA_HEADER_CHANGED 0x0001
2851 #define COA_ROUTE_CHANGED 0x0002
2852 #define COA_RCVBUF_CHANGED 0x0004 /* SO_RCVBUF */
2853 #define COA_SNDBUF_CHANGED 0x0008 /* SO_SNDBUF */
2854 #define COA_WROFF_CHANGED 0x0010 /* Header size changed */
2855 #define COA_ICMP_BIND_NEEDED 0x0020
2856 #define COA_OOBINLINE_CHANGED 0x0040
2857
2858 #define TCP_PORTS_OFFSET 0
2859 #define UDP_PORTS_OFFSET 0
2860
2861 /*
2862 * lookups return the ill/ipif only if the flags are clear OR Iam writer.
2863 * ill / ipif lookup functions increment the refcnt on the ill / ipif only
2864 * after calling these macros. This ensures that the refcnt on the ipif or
2865 * ill will eventually drop down to zero.
2866 */
2867 #define ILL_LOOKUP_FAILED 1 /* Used as error code */
2868 #define IPIF_LOOKUP_FAILED 2 /* Used as error code */
2869
2870 #define ILL_CAN_LOOKUP(ill) \
2871 (!((ill)->ill_state_flags & ILL_CONDEMNED) || \
2872 IAM_WRITER_ILL(ill))
2873
2874 #define ILL_IS_CONDEMNED(ill) \
2875 ((ill)->ill_state_flags & ILL_CONDEMNED)
2876
2877 #define IPIF_CAN_LOOKUP(ipif) \
2878 (!((ipif)->ipif_state_flags & IPIF_CONDEMNED) || \
2879 IAM_WRITER_IPIF(ipif))
2880
2881 #define IPIF_IS_CONDEMNED(ipif) \
2882 ((ipif)->ipif_state_flags & IPIF_CONDEMNED)
2883
2884 #define IPIF_IS_CHANGING(ipif) \
2885 ((ipif)->ipif_state_flags & IPIF_CHANGING)
2886
2887 /* Macros used to assert that this thread is a writer */
2888 #define IAM_WRITER_IPSQ(ipsq) ((ipsq)->ipsq_xop->ipx_writer == curthread)
2889 #define IAM_WRITER_ILL(ill) IAM_WRITER_IPSQ((ill)->ill_phyint->phyint_ipsq)
2890 #define IAM_WRITER_IPIF(ipif) IAM_WRITER_ILL((ipif)->ipif_ill)
2891
2892 /*
2893 * Grab ill locks in the proper order. The order is highest addressed
2894 * ill is locked first.
2895 */
2896 #define GRAB_ILL_LOCKS(ill_1, ill_2) \
2897 { \
2898 if ((ill_1) > (ill_2)) { \
2899 if (ill_1 != NULL) \
2900 mutex_enter(&(ill_1)->ill_lock); \
2901 if (ill_2 != NULL) \
2902 mutex_enter(&(ill_2)->ill_lock); \
2903 } else { \
2904 if (ill_2 != NULL) \
2905 mutex_enter(&(ill_2)->ill_lock); \
2906 if (ill_1 != NULL && ill_1 != ill_2) \
2907 mutex_enter(&(ill_1)->ill_lock); \
2908 } \
2909 }
2910
2911 #define RELEASE_ILL_LOCKS(ill_1, ill_2) \
2912 { \
2913 if (ill_1 != NULL) \
2914 mutex_exit(&(ill_1)->ill_lock); \
2915 if (ill_2 != NULL && ill_2 != ill_1) \
2916 mutex_exit(&(ill_2)->ill_lock); \
2917 }
2918
2919 /* Get the other protocol instance ill */
2920 #define ILL_OTHER(ill) \
2921 ((ill)->ill_isv6 ? (ill)->ill_phyint->phyint_illv4 : \
2922 (ill)->ill_phyint->phyint_illv6)
2923
2924 /* ioctl command info: Ioctl properties extracted and stored in here */
2925 typedef struct cmd_info_s
2926 {
2927 ipif_t *ci_ipif; /* ipif associated with [l]ifreq ioctl's */
2928 sin_t *ci_sin; /* the sin struct passed down */
2929 sin6_t *ci_sin6; /* the sin6_t struct passed down */
2930 struct lifreq *ci_lifr; /* the lifreq struct passed down */
2931 } cmd_info_t;
2932
2933 extern struct kmem_cache *ire_cache;
2934
2935 extern ipaddr_t ip_g_all_ones;
2936
2937 extern uint_t ip_loopback_mtu; /* /etc/system */
2938 extern uint_t ip_loopback_mtuplus;
2939 extern uint_t ip_loopback_mtu_v6plus;
2940
2941 extern vmem_t *ip_minor_arena_sa;
2942 extern vmem_t *ip_minor_arena_la;
2943
2944 /*
2945 * ip_g_forward controls IP forwarding. It takes two values:
2946 * 0: IP_FORWARD_NEVER Don't forward packets ever.
2947 * 1: IP_FORWARD_ALWAYS Forward packets for elsewhere.
2948 *
2949 * RFC1122 says there must be a configuration switch to control forwarding,
2950 * but that the default MUST be to not forward packets ever. Implicit
2951 * control based on configuration of multiple interfaces MUST NOT be
2952 * implemented (Section 3.1). SunOS 4.1 did provide the "automatic" capability
2953 * and, in fact, it was the default. That capability is now provided in the
2954 * /etc/rc2.d/S69inet script.
2955 */
2956
2957 #define ips_ip_respond_to_address_mask_broadcast \
2958 ips_propinfo_tbl[0].prop_cur_bval
2959 #define ips_ip_g_resp_to_echo_bcast ips_propinfo_tbl[1].prop_cur_bval
2960 #define ips_ip_g_resp_to_echo_mcast ips_propinfo_tbl[2].prop_cur_bval
2961 #define ips_ip_g_resp_to_timestamp ips_propinfo_tbl[3].prop_cur_bval
2962 #define ips_ip_g_resp_to_timestamp_bcast ips_propinfo_tbl[4].prop_cur_bval
2963 #define ips_ip_g_send_redirects ips_propinfo_tbl[5].prop_cur_bval
2964 #define ips_ip_g_forward_directed_bcast ips_propinfo_tbl[6].prop_cur_bval
2965 #define ips_ip_mrtdebug ips_propinfo_tbl[7].prop_cur_uval
2966 #define ips_ip_ire_reclaim_fraction ips_propinfo_tbl[8].prop_cur_uval
2967 #define ips_ip_nce_reclaim_fraction ips_propinfo_tbl[9].prop_cur_uval
2968 #define ips_ip_dce_reclaim_fraction ips_propinfo_tbl[10].prop_cur_uval
2969 #define ips_ip_def_ttl ips_propinfo_tbl[11].prop_cur_uval
2970 #define ips_ip_forward_src_routed ips_propinfo_tbl[12].prop_cur_bval
2971 #define ips_ip_wroff_extra ips_propinfo_tbl[13].prop_cur_uval
2972 #define ips_ip_pathmtu_interval ips_propinfo_tbl[14].prop_cur_uval
2973 #define ips_ip_icmp_return ips_propinfo_tbl[15].prop_cur_uval
2974 #define ips_ip_path_mtu_discovery ips_propinfo_tbl[16].prop_cur_bval
2975 #define ips_ip_pmtu_min ips_propinfo_tbl[17].prop_cur_uval
2976 #define ips_ip_ignore_redirect ips_propinfo_tbl[18].prop_cur_bval
2977 #define ips_ip_arp_icmp_error ips_propinfo_tbl[19].prop_cur_bval
2978 #define ips_ip_broadcast_ttl ips_propinfo_tbl[20].prop_cur_uval
2979 #define ips_ip_icmp_err_interval ips_propinfo_tbl[21].prop_cur_uval
2980 #define ips_ip_icmp_err_burst ips_propinfo_tbl[22].prop_cur_uval
2981 #define ips_ip_reass_queue_bytes ips_propinfo_tbl[23].prop_cur_uval
2982 #define ips_ip_strict_dst_multihoming ips_propinfo_tbl[24].prop_cur_uval
2983 #define ips_ip_addrs_per_if ips_propinfo_tbl[25].prop_cur_uval
2984 #define ips_ipsec_override_persocket_policy ips_propinfo_tbl[26].prop_cur_bval
2985 #define ips_icmp_accept_clear_messages ips_propinfo_tbl[27].prop_cur_bval
2986 #define ips_igmp_accept_clear_messages ips_propinfo_tbl[28].prop_cur_bval
2987
2988 /* IPv6 configuration knobs */
2989 #define ips_delay_first_probe_time ips_propinfo_tbl[29].prop_cur_uval
2990 #define ips_max_unicast_solicit ips_propinfo_tbl[30].prop_cur_uval
2991 #define ips_ipv6_def_hops ips_propinfo_tbl[31].prop_cur_uval
2992 #define ips_ipv6_icmp_return ips_propinfo_tbl[32].prop_cur_uval
2993 #define ips_ipv6_forward_src_routed ips_propinfo_tbl[33].prop_cur_bval
2994 #define ips_ipv6_resp_echo_mcast ips_propinfo_tbl[34].prop_cur_bval
2995 #define ips_ipv6_send_redirects ips_propinfo_tbl[35].prop_cur_bval
2996 #define ips_ipv6_ignore_redirect ips_propinfo_tbl[36].prop_cur_bval
2997 #define ips_ipv6_strict_dst_multihoming ips_propinfo_tbl[37].prop_cur_uval
2998 #define ips_src_check ips_propinfo_tbl[38].prop_cur_uval
2999 #define ips_ipsec_policy_log_interval ips_propinfo_tbl[39].prop_cur_uval
3000 #define ips_pim_accept_clear_messages ips_propinfo_tbl[40].prop_cur_bval
3001 #define ips_ip_ndp_unsolicit_interval ips_propinfo_tbl[41].prop_cur_uval
3002 #define ips_ip_ndp_unsolicit_count ips_propinfo_tbl[42].prop_cur_uval
3003 #define ips_ipv6_ignore_home_address_opt ips_propinfo_tbl[43].prop_cur_bval
3004
3005 /* Misc IP configuration knobs */
3006 #define ips_ip_policy_mask ips_propinfo_tbl[44].prop_cur_uval
3007 #define ips_ip_ecmp_behavior ips_propinfo_tbl[45].prop_cur_uval
3008 #define ips_ip_multirt_ttl ips_propinfo_tbl[46].prop_cur_uval
3009 #define ips_ip_ire_badcnt_lifetime ips_propinfo_tbl[47].prop_cur_uval
3010 #define ips_ip_max_temp_idle ips_propinfo_tbl[48].prop_cur_uval
3011 #define ips_ip_max_temp_defend ips_propinfo_tbl[49].prop_cur_uval
3012 #define ips_ip_max_defend ips_propinfo_tbl[50].prop_cur_uval
3013 #define ips_ip_defend_interval ips_propinfo_tbl[51].prop_cur_uval
3014 #define ips_ip_dup_recovery ips_propinfo_tbl[52].prop_cur_uval
3015 #define ips_ip_restrict_interzone_loopback ips_propinfo_tbl[53].prop_cur_bval
3016 #define ips_ip_lso_outbound ips_propinfo_tbl[54].prop_cur_bval
3017 #define ips_igmp_max_version ips_propinfo_tbl[55].prop_cur_uval
3018 #define ips_mld_max_version ips_propinfo_tbl[56].prop_cur_uval
3019 #define ips_ip_forwarding ips_propinfo_tbl[57].prop_cur_bval
3020 #define ips_ipv6_forwarding ips_propinfo_tbl[58].prop_cur_bval
3021 #define ips_ip_reassembly_timeout ips_propinfo_tbl[59].prop_cur_uval
3022 #define ips_ipv6_reassembly_timeout ips_propinfo_tbl[60].prop_cur_uval
3023 #define ips_ip_cgtp_filter ips_propinfo_tbl[61].prop_cur_bval
3024 #define ips_arp_probe_delay ips_propinfo_tbl[62].prop_cur_uval
3025 #define ips_arp_fastprobe_delay ips_propinfo_tbl[63].prop_cur_uval
3026 #define ips_arp_probe_interval ips_propinfo_tbl[64].prop_cur_uval
3027 #define ips_arp_fastprobe_interval ips_propinfo_tbl[65].prop_cur_uval
3028 #define ips_arp_probe_count ips_propinfo_tbl[66].prop_cur_uval
3029 #define ips_arp_fastprobe_count ips_propinfo_tbl[67].prop_cur_uval
3030 #define ips_ipv4_dad_announce_interval ips_propinfo_tbl[68].prop_cur_uval
3031 #define ips_ipv6_dad_announce_interval ips_propinfo_tbl[69].prop_cur_uval
3032 #define ips_arp_defend_interval ips_propinfo_tbl[70].prop_cur_uval
3033 #define ips_arp_defend_rate ips_propinfo_tbl[71].prop_cur_uval
3034 #define ips_ndp_defend_interval ips_propinfo_tbl[72].prop_cur_uval
3035 #define ips_ndp_defend_rate ips_propinfo_tbl[73].prop_cur_uval
3036 #define ips_arp_defend_period ips_propinfo_tbl[74].prop_cur_uval
3037 #define ips_ndp_defend_period ips_propinfo_tbl[75].prop_cur_uval
3038 #define ips_ipv4_icmp_return_pmtu ips_propinfo_tbl[76].prop_cur_bval
3039 #define ips_ipv6_icmp_return_pmtu ips_propinfo_tbl[77].prop_cur_bval
3040 #define ips_ip_arp_publish_count ips_propinfo_tbl[78].prop_cur_uval
3041 #define ips_ip_arp_publish_interval ips_propinfo_tbl[79].prop_cur_uval
3042 #define ips_ip_strict_src_multihoming ips_propinfo_tbl[80].prop_cur_uval
3043 #define ips_ipv6_strict_src_multihoming ips_propinfo_tbl[81].prop_cur_uval
3044 #define ips_ipv6_drop_inbound_icmpv6 ips_propinfo_tbl[82].prop_cur_bval
3045 #define ips_ip_dce_reclaim_threshold ips_propinfo_tbl[83].prop_cur_uval
3046
3047 extern int dohwcksum; /* use h/w cksum if supported by the h/w */
3048 #ifdef ZC_TEST
3049 extern int noswcksum;
3050 #endif
3051
3052 extern char ipif_loopback_name[];
3053
3054 extern nv_t *ire_nv_tbl;
3055
3056 extern struct module_info ip_mod_info;
3057
3058 #define HOOKS4_INTERESTED_PHYSICAL_IN(ipst) \
3059 ((ipst)->ips_ip4_physical_in_event.he_interested)
3060 #define HOOKS6_INTERESTED_PHYSICAL_IN(ipst) \
3061 ((ipst)->ips_ip6_physical_in_event.he_interested)
3062 #define HOOKS4_INTERESTED_PHYSICAL_OUT(ipst) \
3063 ((ipst)->ips_ip4_physical_out_event.he_interested)
3064 #define HOOKS6_INTERESTED_PHYSICAL_OUT(ipst) \
3065 ((ipst)->ips_ip6_physical_out_event.he_interested)
3066 #define HOOKS4_INTERESTED_FORWARDING(ipst) \
3067 ((ipst)->ips_ip4_forwarding_event.he_interested)
3068 #define HOOKS6_INTERESTED_FORWARDING(ipst) \
3069 ((ipst)->ips_ip6_forwarding_event.he_interested)
3070 #define HOOKS4_INTERESTED_LOOPBACK_IN(ipst) \
3071 ((ipst)->ips_ip4_loopback_in_event.he_interested)
3072 #define HOOKS6_INTERESTED_LOOPBACK_IN(ipst) \
3073 ((ipst)->ips_ip6_loopback_in_event.he_interested)
3074 #define HOOKS4_INTERESTED_LOOPBACK_OUT(ipst) \
3075 ((ipst)->ips_ip4_loopback_out_event.he_interested)
3076 #define HOOKS6_INTERESTED_LOOPBACK_OUT(ipst) \
3077 ((ipst)->ips_ip6_loopback_out_event.he_interested)
3078 /*
3079 * Hooks marcos used inside of ip
3080 * The callers use the above INTERESTED macros first, hence
3081 * the he_interested check is superflous.
3082 */
3083 #define FW_HOOKS(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst, _err) \
3084 if ((_hook).he_interested) { \
3085 hook_pkt_event_t info; \
3086 \
3087 _NOTE(CONSTCOND) \
3088 ASSERT((_ilp != NULL) || (_olp != NULL)); \
3089 \
3090 FW_SET_ILL_INDEX(info.hpe_ifp, (ill_t *)_ilp); \
3091 FW_SET_ILL_INDEX(info.hpe_ofp, (ill_t *)_olp); \
3092 info.hpe_protocol = ipst->ips_ipv4_net_data; \
3093 info.hpe_hdr = _iph; \
3094 info.hpe_mp = &(_fm); \
3095 info.hpe_mb = _m; \
3096 info.hpe_flags = _llm; \
3097 _err = hook_run(ipst->ips_ipv4_net_data->netd_hooks, \
3098 _event, (hook_data_t)&info); \
3099 if (_err != 0) { \
3100 ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
3101 (_hook).he_name, (void *)_fm, (void *)_m)); \
3102 if (_fm != NULL) { \
3103 freemsg(_fm); \
3104 _fm = NULL; \
3105 } \
3106 _iph = NULL; \
3107 _m = NULL; \
3108 } else { \
3109 _iph = info.hpe_hdr; \
3110 _m = info.hpe_mb; \
3111 } \
3112 }
3113
3114 #define FW_HOOKS6(_hook, _event, _ilp, _olp, _iph, _fm, _m, _llm, ipst, _err) \
3115 if ((_hook).he_interested) { \
3116 hook_pkt_event_t info; \
3117 \
3118 _NOTE(CONSTCOND) \
3119 ASSERT((_ilp != NULL) || (_olp != NULL)); \
3120 \
3121 FW_SET_ILL_INDEX(info.hpe_ifp, (ill_t *)_ilp); \
3122 FW_SET_ILL_INDEX(info.hpe_ofp, (ill_t *)_olp); \
3123 info.hpe_protocol = ipst->ips_ipv6_net_data; \
3124 info.hpe_hdr = _iph; \
3125 info.hpe_mp = &(_fm); \
3126 info.hpe_mb = _m; \
3127 info.hpe_flags = _llm; \
3128 _err = hook_run(ipst->ips_ipv6_net_data->netd_hooks, \
3129 _event, (hook_data_t)&info); \
3130 if (_err != 0) { \
3131 ip2dbg(("%s hook dropped mblk chain %p hdr %p\n",\
3132 (_hook).he_name, (void *)_fm, (void *)_m)); \
3133 if (_fm != NULL) { \
3134 freemsg(_fm); \
3135 _fm = NULL; \
3136 } \
3137 _iph = NULL; \
3138 _m = NULL; \
3139 } else { \
3140 _iph = info.hpe_hdr; \
3141 _m = info.hpe_mb; \
3142 } \
3143 }
3144
3145 #define FW_SET_ILL_INDEX(fp, ill) \
3146 _NOTE(CONSTCOND) \
3147 if ((ill) == NULL || (ill)->ill_phyint == NULL) { \
3148 (fp) = 0; \
3149 _NOTE(CONSTCOND) \
3150 } else if (IS_UNDER_IPMP(ill)) { \
3151 (fp) = ipmp_ill_get_ipmp_ifindex(ill); \
3152 } else { \
3153 (fp) = (ill)->ill_phyint->phyint_ifindex; \
3154 }
3155
3156 /*
3157 * Network byte order macros
3158 */
3159 #ifdef _BIG_ENDIAN
3160 #define N_IN_CLASSA_NET IN_CLASSA_NET
3161 #define N_IN_CLASSD_NET IN_CLASSD_NET
3162 #define N_INADDR_UNSPEC_GROUP INADDR_UNSPEC_GROUP
3163 #define N_IN_LOOPBACK_NET (ipaddr_t)0x7f000000U
3164 #else /* _BIG_ENDIAN */
3165 #define N_IN_CLASSA_NET (ipaddr_t)0x000000ffU
3166 #define N_IN_CLASSD_NET (ipaddr_t)0x000000f0U
3167 #define N_INADDR_UNSPEC_GROUP (ipaddr_t)0x000000e0U
3168 #define N_IN_LOOPBACK_NET (ipaddr_t)0x0000007fU
3169 #endif /* _BIG_ENDIAN */
3170 #define CLASSD(addr) (((addr) & N_IN_CLASSD_NET) == N_INADDR_UNSPEC_GROUP)
3171 #define CLASSE(addr) (((addr) & N_IN_CLASSD_NET) == N_IN_CLASSD_NET)
3172 #define IP_LOOPBACK_ADDR(addr) \
3173 (((addr) & N_IN_CLASSA_NET == N_IN_LOOPBACK_NET))
3174
3175 extern int ip_debug;
3176 extern uint_t ip_thread_data;
3177 extern krwlock_t ip_thread_rwlock;
3178 extern list_t ip_thread_list;
3179
3180 #ifdef IP_DEBUG
3181 #include <sys/debug.h>
3182 #include <sys/promif.h>
3183
3184 #define ip0dbg(a) printf a
3185 #define ip1dbg(a) if (ip_debug > 2) printf a
3186 #define ip2dbg(a) if (ip_debug > 3) printf a
3187 #define ip3dbg(a) if (ip_debug > 4) printf a
3188 #else
3189 #define ip0dbg(a) /* */
3190 #define ip1dbg(a) /* */
3191 #define ip2dbg(a) /* */
3192 #define ip3dbg(a) /* */
3193 #endif /* IP_DEBUG */
3194
3195 /* Default MAC-layer address string length for mac_colon_addr */
3196 #define MAC_STR_LEN 128
3197
3198 struct mac_header_info_s;
3199
3200 extern void ill_frag_timer(void *);
3201 extern ill_t *ill_first(int, int, ill_walk_context_t *, ip_stack_t *);
3202 extern ill_t *ill_next(ill_walk_context_t *, ill_t *);
3203 extern void ill_frag_timer_start(ill_t *);
3204 extern void ill_nic_event_dispatch(ill_t *, lif_if_t, nic_event_t,
3205 nic_event_data_t, size_t);
3206 extern mblk_t *ip_carve_mp(mblk_t **, ssize_t);
3207 extern mblk_t *ip_dlpi_alloc(size_t, t_uscalar_t);
3208 extern mblk_t *ip_dlnotify_alloc(uint_t, uint_t);
3209 extern mblk_t *ip_dlnotify_alloc2(uint_t, uint_t, uint_t);
3210 extern char *ip_dot_addr(ipaddr_t, char *);
3211 extern const char *mac_colon_addr(const uint8_t *, size_t, char *, size_t);
3212 extern void ip_lwput(queue_t *, mblk_t *);
3213 extern boolean_t icmp_err_rate_limit(ip_stack_t *);
3214 extern void icmp_frag_needed(mblk_t *, int, ip_recv_attr_t *);
3215 extern mblk_t *icmp_inbound_v4(mblk_t *, ip_recv_attr_t *);
3216 extern void icmp_time_exceeded(mblk_t *, uint8_t, ip_recv_attr_t *);
3217 extern void icmp_unreachable(mblk_t *, uint8_t, ip_recv_attr_t *);
3218 extern boolean_t ip_ipsec_policy_inherit(conn_t *, conn_t *, ip_recv_attr_t *);
3219 extern void *ip_pullup(mblk_t *, ssize_t, ip_recv_attr_t *);
3220 extern void ip_setl2src(mblk_t *, ip_recv_attr_t *, ill_t *);
3221 extern mblk_t *ip_check_and_align_header(mblk_t *, uint_t, ip_recv_attr_t *);
3222 extern mblk_t *ip_check_length(mblk_t *, uchar_t *, ssize_t, uint_t, uint_t,
3223 ip_recv_attr_t *);
3224 extern mblk_t *ip_check_optlen(mblk_t *, ipha_t *, uint_t, uint_t,
3225 ip_recv_attr_t *);
3226 extern mblk_t *ip_fix_dbref(mblk_t *, ip_recv_attr_t *);
3227 extern uint_t ip_cksum(mblk_t *, int, uint32_t);
3228 extern int ip_close(queue_t *, int);
3229 extern uint16_t ip_csum_hdr(ipha_t *);
3230 extern void ip_forward_xmit_v4(nce_t *, ill_t *, mblk_t *, ipha_t *,
3231 ip_recv_attr_t *, uint32_t, uint32_t);
3232 extern boolean_t ip_forward_options(mblk_t *, ipha_t *, ill_t *,
3233 ip_recv_attr_t *);
3234 extern int ip_fragment_v4(mblk_t *, nce_t *, iaflags_t, uint_t, uint32_t,
3235 uint32_t, zoneid_t, zoneid_t, pfirepostfrag_t postfragfn,
3236 uintptr_t *cookie);
3237 extern void ip_proto_not_sup(mblk_t *, ip_recv_attr_t *);
3238 extern void ip_ire_g_fini(void);
3239 extern void ip_ire_g_init(void);
3240 extern void ip_ire_fini(ip_stack_t *);
3241 extern void ip_ire_init(ip_stack_t *);
3242 extern void ip_mdata_to_mhi(ill_t *, mblk_t *, struct mac_header_info_s *);
3243 extern int ip_openv4(queue_t *q, dev_t *devp, int flag, int sflag,
3244 cred_t *credp);
3245 extern int ip_openv6(queue_t *q, dev_t *devp, int flag, int sflag,
3246 cred_t *credp);
3247 extern int ip_reassemble(mblk_t *, ipf_t *, uint_t, boolean_t, ill_t *,
3248 size_t);
3249 extern void ip_rput(queue_t *, mblk_t *);
3250 extern void ip_input(ill_t *, ill_rx_ring_t *, mblk_t *,
3251 struct mac_header_info_s *);
3252 extern void ip_input_v6(ill_t *, ill_rx_ring_t *, mblk_t *,
3253 struct mac_header_info_s *);
3254 extern mblk_t *ip_input_common_v4(ill_t *, ill_rx_ring_t *, mblk_t *,
3255 struct mac_header_info_s *, squeue_t *, mblk_t **, uint_t *);
3256 extern mblk_t *ip_input_common_v6(ill_t *, ill_rx_ring_t *, mblk_t *,
3257 struct mac_header_info_s *, squeue_t *, mblk_t **, uint_t *);
3258 extern void ill_input_full_v4(mblk_t *, void *, void *,
3259 ip_recv_attr_t *, rtc_t *);
3260 extern void ill_input_short_v4(mblk_t *, void *, void *,
3261 ip_recv_attr_t *, rtc_t *);
3262 extern void ill_input_full_v6(mblk_t *, void *, void *,
3263 ip_recv_attr_t *, rtc_t *);
3264 extern void ill_input_short_v6(mblk_t *, void *, void *,
3265 ip_recv_attr_t *, rtc_t *);
3266 extern ipaddr_t ip_input_options(ipha_t *, ipaddr_t, mblk_t *,
3267 ip_recv_attr_t *, int *);
3268 extern boolean_t ip_input_local_options(mblk_t *, ipha_t *, ip_recv_attr_t *);
3269 extern mblk_t *ip_input_fragment(mblk_t *, ipha_t *, ip_recv_attr_t *);
3270 extern mblk_t *ip_input_fragment_v6(mblk_t *, ip6_t *, ip6_frag_t *, uint_t,
3271 ip_recv_attr_t *);
3272 extern void ip_input_post_ipsec(mblk_t *, ip_recv_attr_t *);
3273 extern void ip_fanout_v4(mblk_t *, ipha_t *, ip_recv_attr_t *);
3274 extern void ip_fanout_v6(mblk_t *, ip6_t *, ip_recv_attr_t *);
3275 extern void ip_fanout_proto_conn(conn_t *, mblk_t *, ipha_t *, ip6_t *,
3276 ip_recv_attr_t *);
3277 extern void ip_fanout_proto_v4(mblk_t *, ipha_t *, ip_recv_attr_t *);
3278 extern void ip_fanout_send_icmp_v4(mblk_t *, uint_t, uint_t,
3279 ip_recv_attr_t *);
3280 extern void ip_fanout_udp_conn(conn_t *, mblk_t *, ipha_t *, ip6_t *,
3281 ip_recv_attr_t *);
3282 extern void ip_fanout_udp_multi_v4(mblk_t *, ipha_t *, uint16_t, uint16_t,
3283 ip_recv_attr_t *);
3284 extern mblk_t *zero_spi_check(mblk_t *, ip_recv_attr_t *);
3285 extern void ip_build_hdrs_v4(uchar_t *, uint_t, const ip_pkt_t *, uint8_t);
3286 extern int ip_find_hdr_v4(ipha_t *, ip_pkt_t *, boolean_t);
3287 extern int ip_total_hdrs_len_v4(const ip_pkt_t *);
3288
3289 extern mblk_t *ip_accept_tcp(ill_t *, ill_rx_ring_t *, squeue_t *,
3290 mblk_t *, mblk_t **, uint_t *cnt);
3291 extern void ip_rput_dlpi(ill_t *, mblk_t *);
3292 extern void ip_rput_notdata(ill_t *, mblk_t *);
3293
3294 extern void ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *,
3295 mib2_ipIfStatsEntry_t *);
3296 extern void ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *,
3297 mib2_ipv6IfIcmpEntry_t *);
3298 extern void ip_rput_other(ipsq_t *, queue_t *, mblk_t *, void *);
3299 extern ire_t *ip_check_multihome(void *, ire_t *, ill_t *);
3300 extern void ip_send_potential_redirect_v4(mblk_t *, ipha_t *, ire_t *,
3301 ip_recv_attr_t *);
3302 extern int ip_set_destination_v4(ipaddr_t *, ipaddr_t, ipaddr_t,
3303 ip_xmit_attr_t *, iulp_t *, uint32_t, uint_t);
3304 extern int ip_set_destination_v6(in6_addr_t *, const in6_addr_t *,
3305 const in6_addr_t *, ip_xmit_attr_t *, iulp_t *, uint32_t, uint_t);
3306
3307 extern int ip_output_simple(mblk_t *, ip_xmit_attr_t *);
3308 extern int ip_output_simple_v4(mblk_t *, ip_xmit_attr_t *);
3309 extern int ip_output_simple_v6(mblk_t *, ip_xmit_attr_t *);
3310 extern int ip_output_options(mblk_t *, ipha_t *, ip_xmit_attr_t *,
3311 ill_t *);
3312 extern void ip_output_local_options(ipha_t *, ip_stack_t *);
3313
3314 extern ip_xmit_attr_t *conn_get_ixa(conn_t *, boolean_t);
3315 extern ip_xmit_attr_t *conn_get_ixa_tryhard(conn_t *, boolean_t);
3316 extern ip_xmit_attr_t *conn_replace_ixa(conn_t *, ip_xmit_attr_t *);
3317 extern ip_xmit_attr_t *conn_get_ixa_exclusive(conn_t *);
3318 extern ip_xmit_attr_t *ip_xmit_attr_duplicate(ip_xmit_attr_t *);
3319 extern void ip_xmit_attr_replace_tsl(ip_xmit_attr_t *, ts_label_t *);
3320 extern void ip_xmit_attr_restore_tsl(ip_xmit_attr_t *, cred_t *);
3321 boolean_t ip_recv_attr_replace_label(ip_recv_attr_t *, ts_label_t *);
3322 extern void ixa_inactive(ip_xmit_attr_t *);
3323 extern void ixa_refrele(ip_xmit_attr_t *);
3324 extern boolean_t ixa_check_drain_insert(conn_t *, ip_xmit_attr_t *);
3325 extern void ixa_cleanup(ip_xmit_attr_t *);
3326 extern void ira_cleanup(ip_recv_attr_t *, boolean_t);
3327 extern void ixa_safe_copy(ip_xmit_attr_t *, ip_xmit_attr_t *);
3328
3329 extern int conn_ip_output(mblk_t *, ip_xmit_attr_t *);
3330 extern boolean_t ip_output_verify_local(ip_xmit_attr_t *);
3331 extern mblk_t *ip_output_process_local(mblk_t *, ip_xmit_attr_t *, boolean_t,
3332 boolean_t, conn_t *);
3333
3334 extern int conn_opt_get(conn_opt_arg_t *, t_scalar_t, t_scalar_t,
3335 uchar_t *);
3336 extern int conn_opt_set(conn_opt_arg_t *, t_scalar_t, t_scalar_t, uint_t,
3337 uchar_t *, boolean_t, cred_t *);
3338 extern boolean_t conn_same_as_last_v4(conn_t *, sin_t *);
3339 extern boolean_t conn_same_as_last_v6(conn_t *, sin6_t *);
3340 extern int conn_update_label(const conn_t *, const ip_xmit_attr_t *,
3341 const in6_addr_t *, ip_pkt_t *);
3342
3343 extern int ip_opt_set_multicast_group(conn_t *, t_scalar_t,
3344 uchar_t *, boolean_t, boolean_t);
3345 extern int ip_opt_set_multicast_sources(conn_t *, t_scalar_t,
3346 uchar_t *, boolean_t, boolean_t);
3347 extern int conn_getsockname(conn_t *, struct sockaddr *, uint_t *);
3348 extern int conn_getpeername(conn_t *, struct sockaddr *, uint_t *);
3349
3350 extern int conn_build_hdr_template(conn_t *, uint_t, uint_t,
3351 const in6_addr_t *, const in6_addr_t *, uint32_t);
3352 extern mblk_t *conn_prepend_hdr(ip_xmit_attr_t *, const ip_pkt_t *,
3353 const in6_addr_t *, const in6_addr_t *, uint8_t, uint32_t, uint_t,
3354 mblk_t *, uint_t, uint_t, uint32_t *, int *);
3355 extern void ip_attr_newdst(ip_xmit_attr_t *);
3356 extern void ip_attr_nexthop(const ip_pkt_t *, const ip_xmit_attr_t *,
3357 const in6_addr_t *, in6_addr_t *);
3358 extern int conn_connect(conn_t *, iulp_t *, uint32_t);
3359 extern int ip_attr_connect(const conn_t *, ip_xmit_attr_t *,
3360 const in6_addr_t *, const in6_addr_t *, const in6_addr_t *, in_port_t,
3361 in6_addr_t *, iulp_t *, uint32_t);
3362 extern int conn_inherit_parent(conn_t *, conn_t *);
3363
3364 extern void conn_ixa_cleanup(conn_t *connp, void *arg);
3365
3366 extern boolean_t conn_wantpacket(conn_t *, ip_recv_attr_t *, ipha_t *);
3367 extern uint_t ip_type_v4(ipaddr_t, ip_stack_t *);
3368 extern uint_t ip_type_v6(const in6_addr_t *, ip_stack_t *);
3369
3370 extern void ip_wput_nondata(queue_t *, mblk_t *);
3371 extern void ip_wsrv(queue_t *);
3372 extern char *ip_nv_lookup(nv_t *, int);
3373 extern boolean_t ip_local_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3374 extern boolean_t ip_remote_addr_ok_v6(const in6_addr_t *, const in6_addr_t *);
3375 extern ipaddr_t ip_massage_options(ipha_t *, netstack_t *);
3376 extern ipaddr_t ip_net_mask(ipaddr_t);
3377 extern void arp_bringup_done(ill_t *, int);
3378 extern void arp_replumb_done(ill_t *, int);
3379
3380 extern struct qinit iprinitv6;
3381
3382 extern void ipmp_init(ip_stack_t *);
3383 extern void ipmp_destroy(ip_stack_t *);
3384 extern ipmp_grp_t *ipmp_grp_create(const char *, phyint_t *);
3385 extern void ipmp_grp_destroy(ipmp_grp_t *);
3386 extern void ipmp_grp_info(const ipmp_grp_t *, lifgroupinfo_t *);
3387 extern int ipmp_grp_rename(ipmp_grp_t *, const char *);
3388 extern ipmp_grp_t *ipmp_grp_lookup(const char *, ip_stack_t *);
3389 extern int ipmp_grp_vet_phyint(ipmp_grp_t *, phyint_t *);
3390 extern ipmp_illgrp_t *ipmp_illgrp_create(ill_t *);
3391 extern void ipmp_illgrp_destroy(ipmp_illgrp_t *);
3392 extern ill_t *ipmp_illgrp_add_ipif(ipmp_illgrp_t *, ipif_t *);
3393 extern void ipmp_illgrp_del_ipif(ipmp_illgrp_t *, ipif_t *);
3394 extern ill_t *ipmp_illgrp_next_ill(ipmp_illgrp_t *);
3395 extern ill_t *ipmp_illgrp_hold_next_ill(ipmp_illgrp_t *);
3396 extern ill_t *ipmp_illgrp_hold_cast_ill(ipmp_illgrp_t *);
3397 extern ill_t *ipmp_illgrp_ipmp_ill(ipmp_illgrp_t *);
3398 extern void ipmp_illgrp_refresh_mtu(ipmp_illgrp_t *);
3399 extern ipmp_arpent_t *ipmp_illgrp_create_arpent(ipmp_illgrp_t *,
3400 boolean_t, ipaddr_t, uchar_t *, size_t, uint16_t);
3401 extern void ipmp_illgrp_destroy_arpent(ipmp_illgrp_t *, ipmp_arpent_t *);
3402 extern ipmp_arpent_t *ipmp_illgrp_lookup_arpent(ipmp_illgrp_t *, ipaddr_t *);
3403 extern void ipmp_illgrp_refresh_arpent(ipmp_illgrp_t *);
3404 extern void ipmp_illgrp_mark_arpent(ipmp_illgrp_t *, ipmp_arpent_t *);
3405 extern ill_t *ipmp_illgrp_find_ill(ipmp_illgrp_t *, uchar_t *, uint_t);
3406 extern void ipmp_illgrp_link_grp(ipmp_illgrp_t *, ipmp_grp_t *);
3407 extern int ipmp_illgrp_unlink_grp(ipmp_illgrp_t *);
3408 extern uint_t ipmp_ill_get_ipmp_ifindex(const ill_t *);
3409 extern void ipmp_ill_join_illgrp(ill_t *, ipmp_illgrp_t *);
3410 extern void ipmp_ill_leave_illgrp(ill_t *);
3411 extern ill_t *ipmp_ill_hold_ipmp_ill(ill_t *);
3412 extern ill_t *ipmp_ill_hold_xmit_ill(ill_t *, boolean_t);
3413 extern boolean_t ipmp_ill_is_active(ill_t *);
3414 extern void ipmp_ill_refresh_active(ill_t *);
3415 extern void ipmp_phyint_join_grp(phyint_t *, ipmp_grp_t *);
3416 extern void ipmp_phyint_leave_grp(phyint_t *);
3417 extern void ipmp_phyint_refresh_active(phyint_t *);
3418 extern ill_t *ipmp_ipif_bound_ill(const ipif_t *);
3419 extern ill_t *ipmp_ipif_hold_bound_ill(const ipif_t *);
3420 extern boolean_t ipmp_ipif_is_dataaddr(const ipif_t *);
3421 extern boolean_t ipmp_ipif_is_stubaddr(const ipif_t *);
3422 extern boolean_t ipmp_packet_is_probe(mblk_t *, ill_t *);
3423 extern void ipmp_ncec_delete_nce(ncec_t *);
3424 extern void ipmp_ncec_refresh_nce(ncec_t *);
3425
3426 extern void conn_drain_insert(conn_t *, idl_tx_list_t *);
3427 extern void conn_setqfull(conn_t *, boolean_t *);
3428 extern void conn_clrqfull(conn_t *, boolean_t *);
3429 extern int conn_ipsec_length(conn_t *);
3430 extern ipaddr_t ip_get_dst(ipha_t *);
3431 extern uint_t ip_get_pmtu(ip_xmit_attr_t *);
3432 extern uint_t ip_get_base_mtu(ill_t *, ire_t *);
3433 extern mblk_t *ip_output_attach_policy(mblk_t *, ipha_t *, ip6_t *,
3434 const conn_t *, ip_xmit_attr_t *);
3435 extern int ipsec_out_extra_length(ip_xmit_attr_t *);
3436 extern int ipsec_out_process(mblk_t *, ip_xmit_attr_t *);
3437 extern int ip_output_post_ipsec(mblk_t *, ip_xmit_attr_t *);
3438 extern void ipsec_out_to_in(ip_xmit_attr_t *, ill_t *ill,
3439 ip_recv_attr_t *);
3440
3441 extern void ire_cleanup(ire_t *);
3442 extern void ire_inactive(ire_t *);
3443 extern boolean_t irb_inactive(irb_t *);
3444 extern ire_t *ire_unlink(irb_t *);
3445
3446 #ifdef DEBUG
3447 extern boolean_t th_trace_ref(const void *, ip_stack_t *);
3448 extern void th_trace_unref(const void *);
3449 extern void th_trace_cleanup(const void *, boolean_t);
3450 extern void ire_trace_ref(ire_t *);
3451 extern void ire_untrace_ref(ire_t *);
3452 #endif
3453
3454 extern int ip_srcid_insert(const in6_addr_t *, zoneid_t, ip_stack_t *);
3455 extern int ip_srcid_remove(const in6_addr_t *, zoneid_t, ip_stack_t *);
3456 extern void ip_srcid_find_id(uint_t, in6_addr_t *, zoneid_t, netstack_t *);
3457 extern uint_t ip_srcid_find_addr(const in6_addr_t *, zoneid_t, netstack_t *);
3458
3459 extern uint8_t ipoptp_next(ipoptp_t *);
3460 extern uint8_t ipoptp_first(ipoptp_t *, ipha_t *);
3461 extern int ip_opt_get_user(conn_t *, uchar_t *);
3462 extern int ipsec_req_from_conn(conn_t *, ipsec_req_t *, int);
3463 extern int ip_snmp_get(queue_t *q, mblk_t *mctl, int level, boolean_t);
3464 extern int ip_snmp_set(queue_t *q, int, int, uchar_t *, int);
3465 extern void ip_process_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3466 extern void ip_quiesce_conn(conn_t *);
3467 extern void ip_reprocess_ioctl(ipsq_t *, queue_t *, mblk_t *, void *);
3468 extern void ip_ioctl_finish(queue_t *, mblk_t *, int, int, ipsq_t *);
3469
3470 extern boolean_t ip_cmpbuf(const void *, uint_t, boolean_t, const void *,
3471 uint_t);
3472 extern boolean_t ip_allocbuf(void **, uint_t *, boolean_t, const void *,
3473 uint_t);
3474 extern void ip_savebuf(void **, uint_t *, boolean_t, const void *, uint_t);
3475
3476 extern boolean_t ipsq_pending_mp_cleanup(ill_t *, conn_t *);
3477 extern void conn_ioctl_cleanup(conn_t *);
3478
3479 extern void ip_unbind(conn_t *);
3480
3481 extern void tnet_init(void);
3482 extern void tnet_fini(void);
3483
3484 /*
3485 * Hook functions to enable cluster networking
3486 * On non-clustered systems these vectors must always be NULL.
3487 */
3488 extern int (*cl_inet_isclusterwide)(netstackid_t stack_id, uint8_t protocol,
3489 sa_family_t addr_family, uint8_t *laddrp, void *args);
3490 extern uint32_t (*cl_inet_ipident)(netstackid_t stack_id, uint8_t protocol,
3491 sa_family_t addr_family, uint8_t *laddrp, uint8_t *faddrp,
3492 void *args);
3493 extern int (*cl_inet_connect2)(netstackid_t stack_id, uint8_t protocol,
3494 boolean_t is_outgoing, sa_family_t addr_family, uint8_t *laddrp,
3495 in_port_t lport, uint8_t *faddrp, in_port_t fport, void *args);
3496 extern void (*cl_inet_getspi)(netstackid_t, uint8_t, uint8_t *, size_t,
3497 void *);
3498 extern void (*cl_inet_getspi)(netstackid_t stack_id, uint8_t protocol,
3499 uint8_t *ptr, size_t len, void *args);
3500 extern int (*cl_inet_checkspi)(netstackid_t stack_id, uint8_t protocol,
3501 uint32_t spi, void *args);
3502 extern void (*cl_inet_deletespi)(netstackid_t stack_id, uint8_t protocol,
3503 uint32_t spi, void *args);
3504 extern void (*cl_inet_idlesa)(netstackid_t, uint8_t, uint32_t,
3505 sa_family_t, in6_addr_t, in6_addr_t, void *);
3506
3507
3508 /* Hooks for CGTP (multirt routes) filtering module */
3509 #define CGTP_FILTER_REV_1 1
3510 #define CGTP_FILTER_REV_2 2
3511 #define CGTP_FILTER_REV_3 3
3512 #define CGTP_FILTER_REV CGTP_FILTER_REV_3
3513
3514 /* cfo_filter and cfo_filter_v6 hooks return values */
3515 #define CGTP_IP_PKT_NOT_CGTP 0
3516 #define CGTP_IP_PKT_PREMIUM 1
3517 #define CGTP_IP_PKT_DUPLICATE 2
3518
3519 /* Version 3 of the filter interface */
3520 typedef struct cgtp_filter_ops {
3521 int cfo_filter_rev; /* CGTP_FILTER_REV_3 */
3522 int (*cfo_change_state)(netstackid_t, int);
3523 int (*cfo_add_dest_v4)(netstackid_t, ipaddr_t, ipaddr_t,
3524 ipaddr_t, ipaddr_t);
3525 int (*cfo_del_dest_v4)(netstackid_t, ipaddr_t, ipaddr_t);
3526 int (*cfo_add_dest_v6)(netstackid_t, in6_addr_t *, in6_addr_t *,
3527 in6_addr_t *, in6_addr_t *);
3528 int (*cfo_del_dest_v6)(netstackid_t, in6_addr_t *, in6_addr_t *);
3529 int (*cfo_filter)(netstackid_t, uint_t, mblk_t *);
3530 int (*cfo_filter_v6)(netstackid_t, uint_t, ip6_t *,
3531 ip6_frag_t *);
3532 } cgtp_filter_ops_t;
3533
3534 #define CGTP_MCAST_SUCCESS 1
3535
3536 /*
3537 * The separate CGTP module needs this global symbol so that it
3538 * can check the version and determine whether to use the old or the new
3539 * version of the filtering interface.
3540 */
3541 extern int ip_cgtp_filter_rev;
3542
3543 extern int ip_cgtp_filter_supported(void);
3544 extern int ip_cgtp_filter_register(netstackid_t, cgtp_filter_ops_t *);
3545 extern int ip_cgtp_filter_unregister(netstackid_t);
3546 extern int ip_cgtp_filter_is_registered(netstackid_t);
3547
3548 /*
3549 * rr_ring_state cycles in the order shown below from RR_FREE through
3550 * RR_FREE_IN_PROG and back to RR_FREE.
3551 */
3552 typedef enum {
3553 RR_FREE, /* Free slot */
3554 RR_SQUEUE_UNBOUND, /* Ring's squeue is unbound */
3555 RR_SQUEUE_BIND_INPROG, /* Ring's squeue bind in progress */
3556 RR_SQUEUE_BOUND, /* Ring's squeue bound to cpu */
3557 RR_FREE_INPROG /* Ring is being freed */
3558 } ip_ring_state_t;
3559
3560 #define ILL_MAX_RINGS 256 /* Max num of rx rings we can manage */
3561 #define ILL_POLLING 0x01 /* Polling in use */
3562
3563 /*
3564 * These functions pointer types are exported by the mac/dls layer.
3565 * we need to duplicate the definitions here because we cannot
3566 * include mac/dls header files here.
3567 */
3568 typedef boolean_t (*ip_mac_intr_disable_t)(void *);
3569 typedef void (*ip_mac_intr_enable_t)(void *);
3570 typedef ip_mac_tx_cookie_t (*ip_dld_tx_t)(void *, mblk_t *,
3571 uint64_t, uint16_t);
3572 typedef void (*ip_flow_enable_t)(void *, ip_mac_tx_cookie_t);
3573 typedef void *(*ip_dld_callb_t)(void *,
3574 ip_flow_enable_t, void *);
3575 typedef boolean_t (*ip_dld_fctl_t)(void *, ip_mac_tx_cookie_t);
3576 typedef int (*ip_capab_func_t)(void *, uint_t,
3577 void *, uint_t);
3578
3579 /*
3580 * POLLING README
3581 * sq_get_pkts() is called to pick packets from softring in poll mode. It
3582 * calls rr_rx to get the chain and process it with rr_ip_accept.
3583 * rr_rx = mac_soft_ring_poll() to pick packets
3584 * rr_ip_accept = ip_accept_tcp() to process packets
3585 */
3586
3587 /*
3588 * XXX: With protocol, service specific squeues, they will have
3589 * specific acceptor functions.
3590 */
3591 typedef mblk_t *(*ip_mac_rx_t)(void *, size_t);
3592 typedef mblk_t *(*ip_accept_t)(ill_t *, ill_rx_ring_t *,
3593 squeue_t *, mblk_t *, mblk_t **, uint_t *);
3594
3595 /*
3596 * rr_intr_enable, rr_intr_disable, rr_rx_handle, rr_rx:
3597 * May be accessed while in the squeue AND after checking that SQS_POLL_CAPAB
3598 * is set.
3599 *
3600 * rr_ring_state: Protected by ill_lock.
3601 */
3602 struct ill_rx_ring {
3603 ip_mac_intr_disable_t rr_intr_disable; /* Interrupt disabling func */
3604 ip_mac_intr_enable_t rr_intr_enable; /* Interrupt enabling func */
3605 void *rr_intr_handle; /* Handle interrupt funcs */
3606 ip_mac_rx_t rr_rx; /* Driver receive function */
3607 ip_accept_t rr_ip_accept; /* IP accept function */
3608 void *rr_rx_handle; /* Handle for Rx ring */
3609 squeue_t *rr_sqp; /* Squeue the ring is bound to */
3610 ill_t *rr_ill; /* back pointer to ill */
3611 ip_ring_state_t rr_ring_state; /* State of this ring */
3612 };
3613
3614 /*
3615 * IP - DLD direct function call capability
3616 * Suffixes, df - dld function, dh - dld handle,
3617 * cf - client (IP) function, ch - client handle
3618 */
3619 typedef struct ill_dld_direct_s { /* DLD provided driver Tx */
3620 ip_dld_tx_t idd_tx_df; /* str_mdata_fastpath_put */
3621 void *idd_tx_dh; /* dld_str_t *dsp */
3622 ip_dld_callb_t idd_tx_cb_df; /* mac_tx_srs_notify */
3623 void *idd_tx_cb_dh; /* mac_client_handle_t *mch */
3624 ip_dld_fctl_t idd_tx_fctl_df; /* mac_tx_is_flow_blocked */
3625 void *idd_tx_fctl_dh; /* mac_client_handle */
3626 } ill_dld_direct_t;
3627
3628 /* IP - DLD polling capability */
3629 typedef struct ill_dld_poll_s {
3630 ill_rx_ring_t idp_ring_tbl[ILL_MAX_RINGS];
3631 } ill_dld_poll_t;
3632
3633 /* Describes ill->ill_dld_capab */
3634 struct ill_dld_capab_s {
3635 ip_capab_func_t idc_capab_df; /* dld_capab_func */
3636 void *idc_capab_dh; /* dld_str_t *dsp */
3637 ill_dld_direct_t idc_direct;
3638 ill_dld_poll_t idc_poll;
3639 };
3640
3641 /*
3642 * IP squeues exports
3643 */
3644 extern boolean_t ip_squeue_fanout;
3645
3646 #define IP_SQUEUE_GET(hint) ip_squeue_random(hint)
3647
3648 extern void ip_squeue_init(void (*)(squeue_t *));
3649 extern squeue_t *ip_squeue_random(uint_t);
3650 extern squeue_t *ip_squeue_get(ill_rx_ring_t *);
3651 extern squeue_t *ip_squeue_getfree(pri_t);
3652 extern int ip_squeue_cpu_move(squeue_t *, processorid_t);
3653 extern void *ip_squeue_add_ring(ill_t *, void *);
3654 extern void ip_squeue_bind_ring(ill_t *, ill_rx_ring_t *, processorid_t);
3655 extern void ip_squeue_clean_ring(ill_t *, ill_rx_ring_t *);
3656 extern void ip_squeue_quiesce_ring(ill_t *, ill_rx_ring_t *);
3657 extern void ip_squeue_restart_ring(ill_t *, ill_rx_ring_t *);
3658 extern void ip_squeue_clean_all(ill_t *);
3659 extern boolean_t ip_source_routed(ipha_t *, ip_stack_t *);
3660
3661 extern void tcp_wput(queue_t *, mblk_t *);
3662
3663 extern int ip_fill_mtuinfo(conn_t *, ip_xmit_attr_t *,
3664 struct ip6_mtuinfo *);
3665 extern hook_t *ipobs_register_hook(netstack_t *, pfv_t);
3666 extern void ipobs_unregister_hook(netstack_t *, hook_t *);
3667 extern void ipobs_hook(mblk_t *, int, zoneid_t, zoneid_t, const ill_t *,
3668 ip_stack_t *);
3669 typedef void (*ipsq_func_t)(ipsq_t *, queue_t *, mblk_t *, void *);
3670
3671 extern void dce_g_init(void);
3672 extern void dce_g_destroy(void);
3673 extern void dce_stack_init(ip_stack_t *);
3674 extern void dce_stack_destroy(ip_stack_t *);
3675 extern void dce_cleanup(uint_t, ip_stack_t *);
3676 extern dce_t *dce_get_default(ip_stack_t *);
3677 extern dce_t *dce_lookup_pkt(mblk_t *, ip_xmit_attr_t *, uint_t *);
3678 extern dce_t *dce_lookup_v4(ipaddr_t, ip_stack_t *, uint_t *);
3679 extern dce_t *dce_lookup_v6(const in6_addr_t *, uint_t, ip_stack_t *,
3680 uint_t *);
3681 extern dce_t *dce_lookup_and_add_v4(ipaddr_t, ip_stack_t *);
3682 extern dce_t *dce_lookup_and_add_v6(const in6_addr_t *, uint_t,
3683 ip_stack_t *);
3684 extern int dce_update_uinfo_v4(ipaddr_t, iulp_t *, ip_stack_t *);
3685 extern int dce_update_uinfo_v6(const in6_addr_t *, uint_t, iulp_t *,
3686 ip_stack_t *);
3687 extern int dce_update_uinfo(const in6_addr_t *, uint_t, iulp_t *,
3688 ip_stack_t *);
3689 extern void dce_increment_generation(dce_t *);
3690 extern void dce_increment_all_generations(boolean_t, ip_stack_t *);
3691 extern void dce_refrele(dce_t *);
3692 extern void dce_refhold(dce_t *);
3693 extern void dce_refrele_notr(dce_t *);
3694 extern void dce_refhold_notr(dce_t *);
3695 mblk_t *ip_snmp_get_mib2_ip_dce(queue_t *, mblk_t *, ip_stack_t *ipst);
3696
3697 extern ip_laddr_t ip_laddr_verify_v4(ipaddr_t, zoneid_t,
3698 ip_stack_t *, boolean_t);
3699 extern ip_laddr_t ip_laddr_verify_v6(const in6_addr_t *, zoneid_t,
3700 ip_stack_t *, boolean_t, uint_t);
3701 extern int ip_laddr_fanout_insert(conn_t *);
3702
3703 extern boolean_t ip_verify_src(mblk_t *, ip_xmit_attr_t *, uint_t *);
3704 extern int ip_verify_ire(mblk_t *, ip_xmit_attr_t *);
3705
3706 extern mblk_t *ip_xmit_attr_to_mblk(ip_xmit_attr_t *);
3707 extern boolean_t ip_xmit_attr_from_mblk(mblk_t *, ip_xmit_attr_t *);
3708 extern mblk_t *ip_xmit_attr_free_mblk(mblk_t *);
3709 extern mblk_t *ip_recv_attr_to_mblk(ip_recv_attr_t *);
3710 extern boolean_t ip_recv_attr_from_mblk(mblk_t *, ip_recv_attr_t *);
3711 extern mblk_t *ip_recv_attr_free_mblk(mblk_t *);
3712 extern boolean_t ip_recv_attr_is_mblk(mblk_t *);
3713
3714 #ifdef __PRAGMA_REDEFINE_EXTNAME
3715 #pragma redefine_extname inet_pton _inet_pton
3716 #else /* __PRAGMA_REDEFINE_EXTNAME */
3717 #define inet_pton _inet_pton
3718 #endif /* __PRAGMA_REDEFINE_EXTNAME */
3719
3720 extern char *inet_ntop(int, const void *, char *, int);
3721 extern int inet_pton(int, char *, void *);
3722
3723 /*
3724 * Squeue tags. Tags only need to be unique when the callback function is the
3725 * same to distinguish between different calls, but we use unique tags for
3726 * convenience anyway.
3727 */
3728 #define SQTAG_IP_INPUT 1
3729 #define SQTAG_TCP_INPUT_ICMP_ERR 2
3730 #define SQTAG_TCP6_INPUT_ICMP_ERR 3
3731 #define SQTAG_IP_TCP_INPUT 4
3732 #define SQTAG_IP6_TCP_INPUT 5
3733 #define SQTAG_IP_TCP_CLOSE 6
3734 #define SQTAG_TCP_OUTPUT 7
3735 #define SQTAG_TCP_TIMER 8
3736 #define SQTAG_TCP_TIMEWAIT 9
3737 #define SQTAG_TCP_ACCEPT_FINISH 10
3738 #define SQTAG_TCP_ACCEPT_FINISH_Q0 11
3739 #define SQTAG_TCP_ACCEPT_PENDING 12
3740 #define SQTAG_TCP_LISTEN_DISCON 13
3741 #define SQTAG_TCP_CONN_REQ_1 14
3742 #define SQTAG_TCP_EAGER_BLOWOFF 15
3743 #define SQTAG_TCP_EAGER_CLEANUP 16
3744 #define SQTAG_TCP_EAGER_CLEANUP_Q0 17
3745 #define SQTAG_TCP_CONN_IND 18
3746 #define SQTAG_TCP_RSRV 19
3747 #define SQTAG_TCP_ABORT_BUCKET 20
3748 #define SQTAG_TCP_REINPUT 21
3749 #define SQTAG_TCP_REINPUT_EAGER 22
3750 #define SQTAG_TCP_INPUT_MCTL 23
3751 #define SQTAG_TCP_RPUTOTHER 24
3752 #define SQTAG_IP_PROTO_AGAIN 25
3753 #define SQTAG_IP_FANOUT_TCP 26
3754 #define SQTAG_IPSQ_CLEAN_RING 27
3755 #define SQTAG_TCP_WPUT_OTHER 28
3756 #define SQTAG_TCP_CONN_REQ_UNBOUND 29
3757 #define SQTAG_TCP_SEND_PENDING 30
3758 #define SQTAG_BIND_RETRY 31
3759 #define SQTAG_UDP_FANOUT 32
3760 #define SQTAG_UDP_INPUT 33
3761 #define SQTAG_UDP_WPUT 34
3762 #define SQTAG_UDP_OUTPUT 35
3763 #define SQTAG_TCP_KSSL_INPUT 36
3764 #define SQTAG_TCP_DROP_Q0 37
3765 #define SQTAG_TCP_CONN_REQ_2 38
3766 #define SQTAG_IP_INPUT_RX_RING 39
3767 #define SQTAG_SQUEUE_CHANGE 40
3768 #define SQTAG_CONNECT_FINISH 41
3769 #define SQTAG_SYNCHRONOUS_OP 42
3770 #define SQTAG_TCP_SHUTDOWN_OUTPUT 43
3771 #define SQTAG_TCP_IXA_CLEANUP 44
3772 #define SQTAG_TCP_SEND_SYNACK 45
3773
3774 extern sin_t sin_null; /* Zero address for quick clears */
3775 extern sin6_t sin6_null; /* Zero address for quick clears */
3776
3777 #endif /* _KERNEL */
3778
3779 #ifdef __cplusplus
3780 }
3781 #endif
3782
3783 #endif /* _INET_IP_H */