Print this page
11547 Want connstat(1M) command to display per-connection TCP statistics
Portions contributed by: Cody Peter Mello <cody.mello@joyent.com>
Portions contributed by: Ahmed G <ahmedg@delphix.com>
Reviewed by: Jason King <jason.king@joyent.com>
Reviewed by: Robert Mustacchi <rm@joyent.com>
Reviewed by: Dan McDonald <danmcd@joyent.com>
| Split |
Close |
| Expand all |
| Collapse all |
--- old/usr/src/uts/common/inet/ip/ip.c
+++ new/usr/src/uts/common/inet/ip/ip.c
1 1 /*
2 2 * CDDL HEADER START
3 3 *
4 4 * The contents of this file are subject to the terms of the
5 5 * Common Development and Distribution License (the "License").
6 6 * You may not use this file except in compliance with the License.
7 7 *
8 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 9 * or http://www.opensolaris.org/os/licensing.
10 10 * See the License for the specific language governing permissions
11 11 * and limitations under the License.
12 12 *
13 13 * When distributing Covered Code, include this CDDL HEADER in each
14 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 15 * If applicable, add the following below this CDDL HEADER, with the
16 16 * fields enclosed by brackets "[]" replaced with your own identifying
17 17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 18 *
19 19 * CDDL HEADER END
20 20 */
21 21
22 22 /*
23 23 * Copyright (c) 1991, 2010, Oracle and/or its affiliates. All rights reserved.
24 24 * Copyright (c) 1990 Mentat Inc.
25 25 * Copyright (c) 2017 OmniTI Computer Consulting, Inc. All rights reserved.
26 26 * Copyright (c) 2016 by Delphix. All rights reserved.
27 27 * Copyright (c) 2019 Joyent, Inc. All rights reserved.
28 28 */
29 29
30 30 #include <sys/types.h>
31 31 #include <sys/stream.h>
32 32 #include <sys/dlpi.h>
33 33 #include <sys/stropts.h>
34 34 #include <sys/sysmacros.h>
35 35 #include <sys/strsubr.h>
36 36 #include <sys/strlog.h>
37 37 #include <sys/strsun.h>
38 38 #include <sys/zone.h>
39 39 #define _SUN_TPI_VERSION 2
40 40 #include <sys/tihdr.h>
41 41 #include <sys/xti_inet.h>
42 42 #include <sys/ddi.h>
43 43 #include <sys/suntpi.h>
44 44 #include <sys/cmn_err.h>
45 45 #include <sys/debug.h>
46 46 #include <sys/kobj.h>
47 47 #include <sys/modctl.h>
48 48 #include <sys/atomic.h>
49 49 #include <sys/policy.h>
50 50 #include <sys/priv.h>
51 51 #include <sys/taskq.h>
52 52
53 53 #include <sys/systm.h>
54 54 #include <sys/param.h>
55 55 #include <sys/kmem.h>
56 56 #include <sys/sdt.h>
57 57 #include <sys/socket.h>
58 58 #include <sys/vtrace.h>
59 59 #include <sys/isa_defs.h>
60 60 #include <sys/mac.h>
61 61 #include <net/if.h>
62 62 #include <net/if_arp.h>
63 63 #include <net/route.h>
64 64 #include <sys/sockio.h>
65 65 #include <netinet/in.h>
66 66 #include <net/if_dl.h>
67 67
68 68 #include <inet/common.h>
69 69 #include <inet/mi.h>
70 70 #include <inet/mib2.h>
71 71 #include <inet/nd.h>
72 72 #include <inet/arp.h>
73 73 #include <inet/snmpcom.h>
74 74 #include <inet/optcom.h>
75 75 #include <inet/kstatcom.h>
76 76
77 77 #include <netinet/igmp_var.h>
78 78 #include <netinet/ip6.h>
79 79 #include <netinet/icmp6.h>
80 80 #include <netinet/sctp.h>
81 81
82 82 #include <inet/ip.h>
83 83 #include <inet/ip_impl.h>
84 84 #include <inet/ip6.h>
85 85 #include <inet/ip6_asp.h>
86 86 #include <inet/tcp.h>
87 87 #include <inet/tcp_impl.h>
88 88 #include <inet/ip_multi.h>
89 89 #include <inet/ip_if.h>
90 90 #include <inet/ip_ire.h>
91 91 #include <inet/ip_ftable.h>
92 92 #include <inet/ip_rts.h>
93 93 #include <inet/ip_ndp.h>
94 94 #include <inet/ip_listutils.h>
95 95 #include <netinet/igmp.h>
96 96 #include <netinet/ip_mroute.h>
97 97 #include <inet/ipp_common.h>
98 98
99 99 #include <net/pfkeyv2.h>
100 100 #include <inet/sadb.h>
101 101 #include <inet/ipsec_impl.h>
102 102 #include <inet/iptun/iptun_impl.h>
103 103 #include <inet/ipdrop.h>
104 104 #include <inet/ip_netinfo.h>
105 105 #include <inet/ilb_ip.h>
106 106
107 107 #include <sys/ethernet.h>
108 108 #include <net/if_types.h>
109 109 #include <sys/cpuvar.h>
110 110
111 111 #include <ipp/ipp.h>
112 112 #include <ipp/ipp_impl.h>
113 113 #include <ipp/ipgpc/ipgpc.h>
114 114
115 115 #include <sys/pattr.h>
116 116 #include <inet/ipclassifier.h>
117 117 #include <inet/sctp_ip.h>
118 118 #include <inet/sctp/sctp_impl.h>
119 119 #include <inet/udp_impl.h>
120 120 #include <inet/rawip_impl.h>
121 121 #include <inet/rts_impl.h>
122 122
123 123 #include <sys/tsol/label.h>
124 124 #include <sys/tsol/tnet.h>
125 125
126 126 #include <sys/squeue_impl.h>
127 127 #include <inet/ip_arp.h>
128 128
129 129 #include <sys/clock_impl.h> /* For LBOLT_FASTPATH{,64} */
130 130
131 131 /*
132 132 * Values for squeue switch:
133 133 * IP_SQUEUE_ENTER_NODRAIN: SQ_NODRAIN
134 134 * IP_SQUEUE_ENTER: SQ_PROCESS
135 135 * IP_SQUEUE_FILL: SQ_FILL
136 136 */
137 137 int ip_squeue_enter = IP_SQUEUE_ENTER; /* Setable in /etc/system */
138 138
139 139 int ip_squeue_flag;
140 140
141 141 /*
142 142 * Setable in /etc/system
143 143 */
144 144 int ip_poll_normal_ms = 100;
145 145 int ip_poll_normal_ticks = 0;
146 146 int ip_modclose_ackwait_ms = 3000;
147 147
148 148 /*
149 149 * It would be nice to have these present only in DEBUG systems, but the
150 150 * current design of the global symbol checking logic requires them to be
151 151 * unconditionally present.
152 152 */
153 153 uint_t ip_thread_data; /* TSD key for debug support */
154 154 krwlock_t ip_thread_rwlock;
155 155 list_t ip_thread_list;
156 156
157 157 /*
158 158 * Structure to represent a linked list of msgblks. Used by ip_snmp_ functions.
159 159 */
160 160
161 161 struct listptr_s {
162 162 mblk_t *lp_head; /* pointer to the head of the list */
163 163 mblk_t *lp_tail; /* pointer to the tail of the list */
164 164 };
165 165
166 166 typedef struct listptr_s listptr_t;
167 167
168 168 /*
169 169 * This is used by ip_snmp_get_mib2_ip_route_media and
170 170 * ip_snmp_get_mib2_ip6_route_media to carry the lists of return data.
171 171 */
172 172 typedef struct iproutedata_s {
173 173 uint_t ird_idx;
174 174 uint_t ird_flags; /* see below */
175 175 listptr_t ird_route; /* ipRouteEntryTable */
176 176 listptr_t ird_netmedia; /* ipNetToMediaEntryTable */
177 177 listptr_t ird_attrs; /* ipRouteAttributeTable */
178 178 } iproutedata_t;
179 179
180 180 /* Include ire_testhidden and IRE_IF_CLONE routes */
181 181 #define IRD_REPORT_ALL 0x01
182 182
183 183 /*
184 184 * Cluster specific hooks. These should be NULL when booted as a non-cluster
185 185 */
186 186
187 187 /*
188 188 * Hook functions to enable cluster networking
189 189 * On non-clustered systems these vectors must always be NULL.
190 190 *
191 191 * Hook function to Check ip specified ip address is a shared ip address
192 192 * in the cluster
193 193 *
194 194 */
195 195 int (*cl_inet_isclusterwide)(netstackid_t stack_id, uint8_t protocol,
196 196 sa_family_t addr_family, uint8_t *laddrp, void *args) = NULL;
197 197
198 198 /*
199 199 * Hook function to generate cluster wide ip fragment identifier
200 200 */
201 201 uint32_t (*cl_inet_ipident)(netstackid_t stack_id, uint8_t protocol,
202 202 sa_family_t addr_family, uint8_t *laddrp, uint8_t *faddrp,
203 203 void *args) = NULL;
204 204
205 205 /*
206 206 * Hook function to generate cluster wide SPI.
207 207 */
208 208 void (*cl_inet_getspi)(netstackid_t, uint8_t, uint8_t *, size_t,
209 209 void *) = NULL;
210 210
211 211 /*
212 212 * Hook function to verify if the SPI is already utlized.
213 213 */
214 214
215 215 int (*cl_inet_checkspi)(netstackid_t, uint8_t, uint32_t, void *) = NULL;
216 216
217 217 /*
218 218 * Hook function to delete the SPI from the cluster wide repository.
219 219 */
220 220
221 221 void (*cl_inet_deletespi)(netstackid_t, uint8_t, uint32_t, void *) = NULL;
222 222
223 223 /*
224 224 * Hook function to inform the cluster when packet received on an IDLE SA
225 225 */
226 226
227 227 void (*cl_inet_idlesa)(netstackid_t, uint8_t, uint32_t, sa_family_t,
228 228 in6_addr_t, in6_addr_t, void *) = NULL;
229 229
230 230 /*
231 231 * Synchronization notes:
232 232 *
233 233 * IP is a fully D_MP STREAMS module/driver. Thus it does not depend on any
234 234 * MT level protection given by STREAMS. IP uses a combination of its own
235 235 * internal serialization mechanism and standard Solaris locking techniques.
236 236 * The internal serialization is per phyint. This is used to serialize
237 237 * plumbing operations, IPMP operations, most set ioctls, etc.
238 238 *
239 239 * Plumbing is a long sequence of operations involving message
240 240 * exchanges between IP, ARP and device drivers. Many set ioctls are typically
241 241 * involved in plumbing operations. A natural model is to serialize these
242 242 * ioctls one per ill. For example plumbing of hme0 and qfe0 can go on in
243 243 * parallel without any interference. But various set ioctls on hme0 are best
244 244 * serialized, along with IPMP operations and processing of DLPI control
245 245 * messages received from drivers on a per phyint basis. This serialization is
246 246 * provided by the ipsq_t and primitives operating on this. Details can
247 247 * be found in ip_if.c above the core primitives operating on ipsq_t.
248 248 *
249 249 * Lookups of an ipif or ill by a thread return a refheld ipif / ill.
250 250 * Simiarly lookup of an ire by a thread also returns a refheld ire.
251 251 * In addition ipif's and ill's referenced by the ire are also indirectly
252 252 * refheld. Thus no ipif or ill can vanish as long as an ipif is refheld
253 253 * directly or indirectly. For example an SIOCSLIFADDR ioctl that changes the
254 254 * address of an ipif has to go through the ipsq_t. This ensures that only
255 255 * one such exclusive operation proceeds at any time on the ipif. It then
256 256 * waits for all refcnts
257 257 * associated with this ipif to come down to zero. The address is changed
258 258 * only after the ipif has been quiesced. Then the ipif is brought up again.
259 259 * More details are described above the comment in ip_sioctl_flags.
260 260 *
261 261 * Packet processing is based mostly on IREs and are fully multi-threaded
262 262 * using standard Solaris MT techniques.
263 263 *
264 264 * There are explicit locks in IP to handle:
265 265 * - The ip_g_head list maintained by mi_open_link() and friends.
266 266 *
267 267 * - The reassembly data structures (one lock per hash bucket)
268 268 *
269 269 * - conn_lock is meant to protect conn_t fields. The fields actually
270 270 * protected by conn_lock are documented in the conn_t definition.
271 271 *
272 272 * - ire_lock to protect some of the fields of the ire, IRE tables
273 273 * (one lock per hash bucket). Refer to ip_ire.c for details.
274 274 *
275 275 * - ndp_g_lock and ncec_lock for protecting NCEs.
276 276 *
277 277 * - ill_lock protects fields of the ill and ipif. Details in ip.h
278 278 *
279 279 * - ill_g_lock: This is a global reader/writer lock. Protects the following
280 280 * * The AVL tree based global multi list of all ills.
281 281 * * The linked list of all ipifs of an ill
282 282 * * The <ipsq-xop> mapping
283 283 * * <ill-phyint> association
284 284 * Insertion/deletion of an ill in the system, insertion/deletion of an ipif
285 285 * into an ill, changing the <ipsq-xop> mapping of an ill, changing the
286 286 * <ill-phyint> assoc of an ill will all have to hold the ill_g_lock as
287 287 * writer for the actual duration of the insertion/deletion/change.
288 288 *
289 289 * - ill_lock: This is a per ill mutex.
290 290 * It protects some members of the ill_t struct; see ip.h for details.
291 291 * It also protects the <ill-phyint> assoc.
292 292 * It also protects the list of ipifs hanging off the ill.
293 293 *
294 294 * - ipsq_lock: This is a per ipsq_t mutex lock.
295 295 * This protects some members of the ipsq_t struct; see ip.h for details.
296 296 * It also protects the <ipsq-ipxop> mapping
297 297 *
298 298 * - ipx_lock: This is a per ipxop_t mutex lock.
299 299 * This protects some members of the ipxop_t struct; see ip.h for details.
300 300 *
301 301 * - phyint_lock: This is a per phyint mutex lock. Protects just the
302 302 * phyint_flags
303 303 *
304 304 * - ip_addr_avail_lock: This is used to ensure the uniqueness of IP addresses.
305 305 * This lock is held in ipif_up_done and the ipif is marked IPIF_UP and the
306 306 * uniqueness check also done atomically.
307 307 *
308 308 * - ill_g_usesrc_lock: This readers/writer lock protects the usesrc
309 309 * group list linked by ill_usesrc_grp_next. It also protects the
310 310 * ill_usesrc_ifindex field. It is taken as a writer when a member of the
311 311 * group is being added or deleted. This lock is taken as a reader when
312 312 * walking the list/group(eg: to get the number of members in a usesrc group).
313 313 * Note, it is only necessary to take this lock if the ill_usesrc_grp_next
314 314 * field is changing state i.e from NULL to non-NULL or vice-versa. For
315 315 * example, it is not necessary to take this lock in the initial portion
316 316 * of ip_sioctl_slifusesrc or at all in ip_sioctl_flags since these
317 317 * operations are executed exclusively and that ensures that the "usesrc
318 318 * group state" cannot change. The "usesrc group state" change can happen
319 319 * only in the latter part of ip_sioctl_slifusesrc and in ill_delete.
320 320 *
321 321 * Changing <ill-phyint>, <ipsq-xop> assocications:
322 322 *
323 323 * To change the <ill-phyint> association, the ill_g_lock must be held
324 324 * as writer, and the ill_locks of both the v4 and v6 instance of the ill
325 325 * must be held.
326 326 *
327 327 * To change the <ipsq-xop> association, the ill_g_lock must be held as
328 328 * writer, the ipsq_lock must be held, and one must be writer on the ipsq.
329 329 * This is only done when ills are added or removed from IPMP groups.
330 330 *
331 331 * To add or delete an ipif from the list of ipifs hanging off the ill,
332 332 * ill_g_lock (writer) and ill_lock must be held and the thread must be
333 333 * a writer on the associated ipsq.
334 334 *
335 335 * To add or delete an ill to the system, the ill_g_lock must be held as
336 336 * writer and the thread must be a writer on the associated ipsq.
337 337 *
338 338 * To add or delete an ilm to an ill, the ill_lock must be held and the thread
339 339 * must be a writer on the associated ipsq.
340 340 *
341 341 * Lock hierarchy
342 342 *
343 343 * Some lock hierarchy scenarios are listed below.
344 344 *
345 345 * ill_g_lock -> conn_lock -> ill_lock -> ipsq_lock -> ipx_lock
346 346 * ill_g_lock -> ill_lock(s) -> phyint_lock
347 347 * ill_g_lock -> ndp_g_lock -> ill_lock -> ncec_lock
348 348 * ill_g_lock -> ip_addr_avail_lock
349 349 * conn_lock -> irb_lock -> ill_lock -> ire_lock
350 350 * ill_g_lock -> ip_g_nd_lock
351 351 * ill_g_lock -> ips_ipmp_lock -> ill_lock -> nce_lock
352 352 * ill_g_lock -> ndp_g_lock -> ill_lock -> ncec_lock -> nce_lock
353 353 * arl_lock -> ill_lock
354 354 * ips_ire_dep_lock -> irb_lock
355 355 *
356 356 * When more than 1 ill lock is needed to be held, all ill lock addresses
357 357 * are sorted on address and locked starting from highest addressed lock
358 358 * downward.
359 359 *
360 360 * Multicast scenarios
361 361 * ips_ill_g_lock -> ill_mcast_lock
362 362 * conn_ilg_lock -> ips_ill_g_lock -> ill_lock
363 363 * ill_mcast_serializer -> ill_mcast_lock -> ips_ipmp_lock -> ill_lock
364 364 * ill_mcast_serializer -> ill_mcast_lock -> connf_lock -> conn_lock
365 365 * ill_mcast_serializer -> ill_mcast_lock -> conn_ilg_lock
366 366 * ill_mcast_serializer -> ill_mcast_lock -> ips_igmp_timer_lock
367 367 *
368 368 * IPsec scenarios
369 369 *
370 370 * ipsa_lock -> ill_g_lock -> ill_lock
371 371 * ill_g_usesrc_lock -> ill_g_lock -> ill_lock
372 372 *
373 373 * Trusted Solaris scenarios
374 374 *
375 375 * igsa_lock -> gcgrp_rwlock -> gcgrp_lock
376 376 * igsa_lock -> gcdb_lock
377 377 * gcgrp_rwlock -> ire_lock
378 378 * gcgrp_rwlock -> gcdb_lock
379 379 *
380 380 * squeue(sq_lock), flow related (ft_lock, fe_lock) locking
381 381 *
382 382 * cpu_lock --> ill_lock --> sqset_lock --> sq_lock
383 383 * sq_lock -> conn_lock -> QLOCK(q)
384 384 * ill_lock -> ft_lock -> fe_lock
385 385 *
386 386 * Routing/forwarding table locking notes:
387 387 *
388 388 * Lock acquisition order: Radix tree lock, irb_lock.
389 389 * Requirements:
390 390 * i. Walker must not hold any locks during the walker callback.
391 391 * ii Walker must not see a truncated tree during the walk because of any node
392 392 * deletion.
393 393 * iii Existing code assumes ire_bucket is valid if it is non-null and is used
394 394 * in many places in the code to walk the irb list. Thus even if all the
395 395 * ires in a bucket have been deleted, we still can't free the radix node
396 396 * until the ires have actually been inactive'd (freed).
397 397 *
398 398 * Tree traversal - Need to hold the global tree lock in read mode.
399 399 * Before dropping the global tree lock, need to either increment the ire_refcnt
400 400 * to ensure that the radix node can't be deleted.
401 401 *
402 402 * Tree add - Need to hold the global tree lock in write mode to add a
403 403 * radix node. To prevent the node from being deleted, increment the
404 404 * irb_refcnt, after the node is added to the tree. The ire itself is
405 405 * added later while holding the irb_lock, but not the tree lock.
406 406 *
407 407 * Tree delete - Need to hold the global tree lock and irb_lock in write mode.
408 408 * All associated ires must be inactive (i.e. freed), and irb_refcnt
409 409 * must be zero.
410 410 *
411 411 * Walker - Increment irb_refcnt before calling the walker callback. Hold the
412 412 * global tree lock (read mode) for traversal.
413 413 *
414 414 * IRE dependencies - In some cases we hold ips_ire_dep_lock across ire_refrele
415 415 * hence we will acquire irb_lock while holding ips_ire_dep_lock.
416 416 *
417 417 * IPsec notes :
418 418 *
419 419 * IP interacts with the IPsec code (AH/ESP) by storing IPsec attributes
420 420 * in the ip_xmit_attr_t ip_recv_attr_t. For outbound datagrams, the
421 421 * ip_xmit_attr_t has the
422 422 * information used by the IPsec code for applying the right level of
423 423 * protection. The information initialized by IP in the ip_xmit_attr_t
424 424 * is determined by the per-socket policy or global policy in the system.
425 425 * For inbound datagrams, the ip_recv_attr_t
426 426 * starts out with nothing in it. It gets filled
427 427 * with the right information if it goes through the AH/ESP code, which
428 428 * happens if the incoming packet is secure. The information initialized
429 429 * by AH/ESP, is later used by IP (during fanouts to ULP) to see whether
430 430 * the policy requirements needed by per-socket policy or global policy
431 431 * is met or not.
432 432 *
433 433 * For fully connected sockets i.e dst, src [addr, port] is known,
434 434 * conn_policy_cached is set indicating that policy has been cached.
435 435 * conn_in_enforce_policy may or may not be set depending on whether
436 436 * there is a global policy match or per-socket policy match.
437 437 * Policy inheriting happpens in ip_policy_set once the destination is known.
438 438 * Once the right policy is set on the conn_t, policy cannot change for
439 439 * this socket. This makes life simpler for TCP (UDP ?) where
440 440 * re-transmissions go out with the same policy. For symmetry, policy
441 441 * is cached for fully connected UDP sockets also. Thus if policy is cached,
442 442 * it also implies that policy is latched i.e policy cannot change
443 443 * on these sockets. As we have the right policy on the conn, we don't
444 444 * have to lookup global policy for every outbound and inbound datagram
445 445 * and thus serving as an optimization. Note that a global policy change
446 446 * does not affect fully connected sockets if they have policy. If fully
447 447 * connected sockets did not have any policy associated with it, global
448 448 * policy change may affect them.
449 449 *
450 450 * IP Flow control notes:
451 451 * ---------------------
452 452 * Non-TCP streams are flow controlled by IP. The way this is accomplished
453 453 * differs when ILL_CAPAB_DLD_DIRECT is enabled for that IP instance. When
454 454 * ILL_DIRECT_CAPABLE(ill) is TRUE, IP can do direct function calls into
455 455 * GLDv3. Otherwise packets are sent down to lower layers using STREAMS
456 456 * functions.
457 457 *
458 458 * Per Tx ring udp flow control:
459 459 * This is applicable only when ILL_CAPAB_DLD_DIRECT capability is set in
460 460 * the ill (i.e. ILL_DIRECT_CAPABLE(ill) is true).
461 461 *
462 462 * The underlying link can expose multiple Tx rings to the GLDv3 mac layer.
463 463 * To achieve best performance, outgoing traffic need to be fanned out among
464 464 * these Tx ring. mac_tx() is called (via str_mdata_fastpath_put()) to send
465 465 * traffic out of the NIC and it takes a fanout hint. UDP connections pass
466 466 * the address of connp as fanout hint to mac_tx(). Under flow controlled
467 467 * condition, mac_tx() returns a non-NULL cookie (ip_mac_tx_cookie_t). This
468 468 * cookie points to a specific Tx ring that is blocked. The cookie is used to
469 469 * hash into an idl_tx_list[] entry in idl_tx_list[] array. Each idl_tx_list_t
470 470 * point to drain_lists (idl_t's). These drain list will store the blocked UDP
471 471 * connp's. The drain list is not a single list but a configurable number of
472 472 * lists.
473 473 *
474 474 * The diagram below shows idl_tx_list_t's and their drain_lists. ip_stack_t
475 475 * has an array of idl_tx_list_t. The size of the array is TX_FANOUT_SIZE
476 476 * which is equal to 128. This array in turn contains a pointer to idl_t[],
477 477 * the ip drain list. The idl_t[] array size is MIN(max_ncpus, 8). The drain
478 478 * list will point to the list of connp's that are flow controlled.
479 479 *
480 480 * --------------- ------- ------- -------
481 481 * |->|drain_list[0]|-->|connp|-->|connp|-->|connp|-->
482 482 * | --------------- ------- ------- -------
483 483 * | --------------- ------- ------- -------
484 484 * |->|drain_list[1]|-->|connp|-->|connp|-->|connp|-->
485 485 * ---------------- | --------------- ------- ------- -------
486 486 * |idl_tx_list[0]|->| --------------- ------- ------- -------
487 487 * ---------------- |->|drain_list[2]|-->|connp|-->|connp|-->|connp|-->
488 488 * | --------------- ------- ------- -------
489 489 * . . . . .
490 490 * | --------------- ------- ------- -------
491 491 * |->|drain_list[n]|-->|connp|-->|connp|-->|connp|-->
492 492 * --------------- ------- ------- -------
493 493 * --------------- ------- ------- -------
494 494 * |->|drain_list[0]|-->|connp|-->|connp|-->|connp|-->
495 495 * | --------------- ------- ------- -------
496 496 * | --------------- ------- ------- -------
497 497 * ---------------- |->|drain_list[1]|-->|connp|-->|connp|-->|connp|-->
498 498 * |idl_tx_list[1]|->| --------------- ------- ------- -------
499 499 * ---------------- | . . . .
500 500 * | --------------- ------- ------- -------
501 501 * |->|drain_list[n]|-->|connp|-->|connp|-->|connp|-->
502 502 * --------------- ------- ------- -------
503 503 * .....
504 504 * ----------------
505 505 * |idl_tx_list[n]|-> ...
506 506 * ----------------
507 507 *
508 508 * When mac_tx() returns a cookie, the cookie is hashed into an index into
509 509 * ips_idl_tx_list[], and conn_drain_insert() is called with the idl_tx_list
510 510 * to insert the conn onto. conn_drain_insert() asserts flow control for the
511 511 * sockets via su_txq_full() (non-STREAMS) or QFULL on conn_wq (STREAMS).
512 512 * Further, conn_blocked is set to indicate that the conn is blocked.
513 513 *
514 514 * GLDv3 calls ill_flow_enable() when flow control is relieved. The cookie
515 515 * passed in the call to ill_flow_enable() identifies the blocked Tx ring and
516 516 * is again hashed to locate the appropriate idl_tx_list, which is then
517 517 * drained via conn_walk_drain(). conn_walk_drain() goes through each conn in
518 518 * the drain list and calls conn_drain_remove() to clear flow control (via
519 519 * calling su_txq_full() or clearing QFULL), and remove the conn from the
520 520 * drain list.
521 521 *
522 522 * Note that the drain list is not a single list but a (configurable) array of
523 523 * lists (8 elements by default). Synchronization between drain insertion and
524 524 * flow control wakeup is handled by using idl_txl->txl_lock, and only
525 525 * conn_drain_insert() and conn_drain_remove() manipulate the drain list.
526 526 *
527 527 * Flow control via STREAMS is used when ILL_DIRECT_CAPABLE() returns FALSE.
528 528 * On the send side, if the packet cannot be sent down to the driver by IP
529 529 * (canput() fails), ip_xmit() drops the packet and returns EWOULDBLOCK to the
530 530 * caller, who may then invoke ixa_check_drain_insert() to insert the conn on
531 531 * the 0'th drain list. When ip_wsrv() runs on the ill_wq because flow
532 532 * control has been relieved, the blocked conns in the 0'th drain list are
533 533 * drained as in the non-STREAMS case.
534 534 *
535 535 * In both the STREAMS and non-STREAMS cases, the sockfs upcall to set QFULL
536 536 * is done when the conn is inserted into the drain list (conn_drain_insert())
537 537 * and cleared when the conn is removed from the it (conn_drain_remove()).
538 538 *
539 539 * IPQOS notes:
540 540 *
541 541 * IPQoS Policies are applied to packets using IPPF (IP Policy framework)
542 542 * and IPQoS modules. IPPF includes hooks in IP at different control points
543 543 * (callout positions) which direct packets to IPQoS modules for policy
544 544 * processing. Policies, if present, are global.
545 545 *
546 546 * The callout positions are located in the following paths:
547 547 * o local_in (packets destined for this host)
548 548 * o local_out (packets orginating from this host )
549 549 * o fwd_in (packets forwarded by this m/c - inbound)
550 550 * o fwd_out (packets forwarded by this m/c - outbound)
551 551 * Hooks at these callout points can be enabled/disabled using the ndd variable
552 552 * ip_policy_mask (a bit mask with the 4 LSB indicating the callout positions).
553 553 * By default all the callout positions are enabled.
554 554 *
555 555 * Outbound (local_out)
556 556 * Hooks are placed in ire_send_wire_v4 and ire_send_wire_v6.
557 557 *
558 558 * Inbound (local_in)
559 559 * Hooks are placed in ip_fanout_v4 and ip_fanout_v6.
560 560 *
561 561 * Forwarding (in and out)
562 562 * Hooks are placed in ire_recv_forward_v4/v6.
563 563 *
564 564 * IP Policy Framework processing (IPPF processing)
565 565 * Policy processing for a packet is initiated by ip_process, which ascertains
566 566 * that the classifier (ipgpc) is loaded and configured, failing which the
567 567 * packet resumes normal processing in IP. If the clasifier is present, the
568 568 * packet is acted upon by one or more IPQoS modules (action instances), per
569 569 * filters configured in ipgpc and resumes normal IP processing thereafter.
570 570 * An action instance can drop a packet in course of its processing.
571 571 *
572 572 * Zones notes:
573 573 *
574 574 * The partitioning rules for networking are as follows:
575 575 * 1) Packets coming from a zone must have a source address belonging to that
576 576 * zone.
577 577 * 2) Packets coming from a zone can only be sent on a physical interface on
578 578 * which the zone has an IP address.
579 579 * 3) Between two zones on the same machine, packet delivery is only allowed if
580 580 * there's a matching route for the destination and zone in the forwarding
581 581 * table.
582 582 * 4) The TCP and UDP port spaces are per-zone; that is, two processes in
583 583 * different zones can bind to the same port with the wildcard address
584 584 * (INADDR_ANY).
585 585 *
586 586 * The granularity of interface partitioning is at the logical interface level.
587 587 * Therefore, every zone has its own IP addresses, and incoming packets can be
588 588 * attributed to a zone unambiguously. A logical interface is placed into a zone
589 589 * using the SIOCSLIFZONE ioctl; this sets the ipif_zoneid field in the ipif_t
590 590 * structure. Rule (1) is implemented by modifying the source address selection
591 591 * algorithm so that the list of eligible addresses is filtered based on the
592 592 * sending process zone.
593 593 *
594 594 * The Internet Routing Entries (IREs) are either exclusive to a zone or shared
595 595 * across all zones, depending on their type. Here is the break-up:
596 596 *
597 597 * IRE type Shared/exclusive
598 598 * -------- ----------------
599 599 * IRE_BROADCAST Exclusive
600 600 * IRE_DEFAULT (default routes) Shared (*)
601 601 * IRE_LOCAL Exclusive (x)
602 602 * IRE_LOOPBACK Exclusive
603 603 * IRE_PREFIX (net routes) Shared (*)
604 604 * IRE_IF_NORESOLVER (interface routes) Exclusive
605 605 * IRE_IF_RESOLVER (interface routes) Exclusive
606 606 * IRE_IF_CLONE (interface routes) Exclusive
607 607 * IRE_HOST (host routes) Shared (*)
608 608 *
609 609 * (*) A zone can only use a default or off-subnet route if the gateway is
610 610 * directly reachable from the zone, that is, if the gateway's address matches
611 611 * one of the zone's logical interfaces.
612 612 *
613 613 * (x) IRE_LOCAL are handled a bit differently.
614 614 * When ip_restrict_interzone_loopback is set (the default),
615 615 * ire_route_recursive restricts loopback using an IRE_LOCAL
616 616 * between zone to the case when L2 would have conceptually looped the packet
617 617 * back, i.e. the loopback which is required since neither Ethernet drivers
618 618 * nor Ethernet hardware loops them back. This is the case when the normal
619 619 * routes (ignoring IREs with different zoneids) would send out the packet on
620 620 * the same ill as the ill with which is IRE_LOCAL is associated.
621 621 *
622 622 * Multiple zones can share a common broadcast address; typically all zones
623 623 * share the 255.255.255.255 address. Incoming as well as locally originated
624 624 * broadcast packets must be dispatched to all the zones on the broadcast
625 625 * network. For directed broadcasts (e.g. 10.16.72.255) this is not trivial
626 626 * since some zones may not be on the 10.16.72/24 network. To handle this, each
627 627 * zone has its own set of IRE_BROADCAST entries; then, broadcast packets are
628 628 * sent to every zone that has an IRE_BROADCAST entry for the destination
629 629 * address on the input ill, see ip_input_broadcast().
630 630 *
631 631 * Applications in different zones can join the same multicast group address.
632 632 * The same logic applies for multicast as for broadcast. ip_input_multicast
633 633 * dispatches packets to all zones that have members on the physical interface.
634 634 */
635 635
636 636 /*
637 637 * Squeue Fanout flags:
638 638 * 0: No fanout.
639 639 * 1: Fanout across all squeues
640 640 */
641 641 boolean_t ip_squeue_fanout = 0;
642 642
643 643 /*
644 644 * Maximum dups allowed per packet.
645 645 */
646 646 uint_t ip_max_frag_dups = 10;
647 647
648 648 static int ip_open(queue_t *q, dev_t *devp, int flag, int sflag,
649 649 cred_t *credp, boolean_t isv6);
650 650 static mblk_t *ip_xmit_attach_llhdr(mblk_t *, nce_t *);
651 651
652 652 static boolean_t icmp_inbound_verify_v4(mblk_t *, icmph_t *, ip_recv_attr_t *);
653 653 static void icmp_inbound_too_big_v4(icmph_t *, ip_recv_attr_t *);
654 654 static void icmp_inbound_error_fanout_v4(mblk_t *, icmph_t *,
655 655 ip_recv_attr_t *);
656 656 static void icmp_options_update(ipha_t *);
657 657 static void icmp_param_problem(mblk_t *, uint8_t, ip_recv_attr_t *);
658 658 static void icmp_pkt(mblk_t *, void *, size_t, ip_recv_attr_t *);
659 659 static mblk_t *icmp_pkt_err_ok(mblk_t *, ip_recv_attr_t *);
660 660 static void icmp_redirect_v4(mblk_t *mp, ipha_t *, icmph_t *,
661 661 ip_recv_attr_t *);
662 662 static void icmp_send_redirect(mblk_t *, ipaddr_t, ip_recv_attr_t *);
663 663 static void icmp_send_reply_v4(mblk_t *, ipha_t *, icmph_t *,
664 664 ip_recv_attr_t *);
665 665
666 666 mblk_t *ip_dlpi_alloc(size_t, t_uscalar_t);
667 667 char *ip_dot_addr(ipaddr_t, char *);
668 668 mblk_t *ip_carve_mp(mblk_t **, ssize_t);
669 669 static char *ip_dot_saddr(uchar_t *, char *);
670 670 static int ip_lrput(queue_t *, mblk_t *);
671 671 ipaddr_t ip_net_mask(ipaddr_t);
672 672 char *ip_nv_lookup(nv_t *, int);
673 673 int ip_rput(queue_t *, mblk_t *);
674 674 static void ip_rput_dlpi_writer(ipsq_t *dummy_sq, queue_t *q, mblk_t *mp,
675 675 void *dummy_arg);
676 676 int ip_snmp_get(queue_t *, mblk_t *, int, boolean_t);
677 677 static mblk_t *ip_snmp_get_mib2_ip(queue_t *, mblk_t *,
678 678 mib2_ipIfStatsEntry_t *, ip_stack_t *, boolean_t);
679 679 static mblk_t *ip_snmp_get_mib2_ip_traffic_stats(queue_t *, mblk_t *,
680 680 ip_stack_t *, boolean_t);
681 681 static mblk_t *ip_snmp_get_mib2_ip6(queue_t *, mblk_t *, ip_stack_t *,
682 682 boolean_t);
683 683 static mblk_t *ip_snmp_get_mib2_icmp(queue_t *, mblk_t *, ip_stack_t *ipst);
684 684 static mblk_t *ip_snmp_get_mib2_icmp6(queue_t *, mblk_t *, ip_stack_t *ipst);
685 685 static mblk_t *ip_snmp_get_mib2_igmp(queue_t *, mblk_t *, ip_stack_t *ipst);
686 686 static mblk_t *ip_snmp_get_mib2_multi(queue_t *, mblk_t *, ip_stack_t *ipst);
687 687 static mblk_t *ip_snmp_get_mib2_ip_addr(queue_t *, mblk_t *,
688 688 ip_stack_t *ipst, boolean_t);
689 689 static mblk_t *ip_snmp_get_mib2_ip6_addr(queue_t *, mblk_t *,
690 690 ip_stack_t *ipst, boolean_t);
691 691 static mblk_t *ip_snmp_get_mib2_ip_group_src(queue_t *, mblk_t *,
692 692 ip_stack_t *ipst);
693 693 static mblk_t *ip_snmp_get_mib2_ip6_group_src(queue_t *, mblk_t *,
694 694 ip_stack_t *ipst);
695 695 static mblk_t *ip_snmp_get_mib2_ip_group_mem(queue_t *, mblk_t *,
696 696 ip_stack_t *ipst);
697 697 static mblk_t *ip_snmp_get_mib2_ip6_group_mem(queue_t *, mblk_t *,
698 698 ip_stack_t *ipst);
699 699 static mblk_t *ip_snmp_get_mib2_virt_multi(queue_t *, mblk_t *,
700 700 ip_stack_t *ipst);
701 701 static mblk_t *ip_snmp_get_mib2_multi_rtable(queue_t *, mblk_t *,
702 702 ip_stack_t *ipst);
703 703 static mblk_t *ip_snmp_get_mib2_ip_route_media(queue_t *, mblk_t *, int,
704 704 ip_stack_t *ipst);
705 705 static mblk_t *ip_snmp_get_mib2_ip6_route_media(queue_t *, mblk_t *, int,
706 706 ip_stack_t *ipst);
707 707 static void ip_snmp_get2_v4(ire_t *, iproutedata_t *);
708 708 static void ip_snmp_get2_v6_route(ire_t *, iproutedata_t *);
709 709 static void ip_snmp_get2_v4_media(ncec_t *, void *);
710 710 static void ip_snmp_get2_v6_media(ncec_t *, void *);
711 711 int ip_snmp_set(queue_t *, int, int, uchar_t *, int);
712 712
713 713 static mblk_t *ip_fragment_copyhdr(uchar_t *, int, int, ip_stack_t *,
714 714 mblk_t *);
715 715
716 716 static void conn_drain_init(ip_stack_t *);
717 717 static void conn_drain_fini(ip_stack_t *);
718 718 static void conn_drain(conn_t *connp, boolean_t closing);
719 719
720 720 static void conn_walk_drain(ip_stack_t *, idl_tx_list_t *);
721 721 static void conn_walk_sctp(pfv_t, void *, zoneid_t, netstack_t *);
722 722
723 723 static void *ip_stack_init(netstackid_t stackid, netstack_t *ns);
724 724 static void ip_stack_shutdown(netstackid_t stackid, void *arg);
725 725 static void ip_stack_fini(netstackid_t stackid, void *arg);
726 726
727 727 static int ip_multirt_apply_membership(int (*fn)(conn_t *, boolean_t,
728 728 const in6_addr_t *, ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *),
729 729 ire_t *, conn_t *, boolean_t, const in6_addr_t *, mcast_record_t,
730 730 const in6_addr_t *);
731 731
732 732 static int ip_squeue_switch(int);
733 733
734 734 static void *ip_kstat_init(netstackid_t, ip_stack_t *);
735 735 static void ip_kstat_fini(netstackid_t, kstat_t *);
736 736 static int ip_kstat_update(kstat_t *kp, int rw);
737 737 static void *icmp_kstat_init(netstackid_t);
738 738 static void icmp_kstat_fini(netstackid_t, kstat_t *);
739 739 static int icmp_kstat_update(kstat_t *kp, int rw);
740 740 static void *ip_kstat2_init(netstackid_t, ip_stat_t *);
741 741 static void ip_kstat2_fini(netstackid_t, kstat_t *);
742 742
743 743 static void ipobs_init(ip_stack_t *);
744 744 static void ipobs_fini(ip_stack_t *);
745 745
746 746 static int ip_tp_cpu_update(cpu_setup_t, int, void *);
747 747
748 748 ipaddr_t ip_g_all_ones = IP_HOST_MASK;
749 749
750 750 static long ip_rput_pullups;
751 751 int dohwcksum = 1; /* use h/w cksum if supported by the hardware */
752 752
753 753 vmem_t *ip_minor_arena_sa; /* for minor nos. from INET_MIN_DEV+2 thru 2^^18-1 */
754 754 vmem_t *ip_minor_arena_la; /* for minor nos. from 2^^18 thru 2^^32-1 */
755 755
756 756 int ip_debug;
757 757
758 758 /*
759 759 * Multirouting/CGTP stuff
760 760 */
761 761 int ip_cgtp_filter_rev = CGTP_FILTER_REV; /* CGTP hooks version */
762 762
763 763 /*
764 764 * IP tunables related declarations. Definitions are in ip_tunables.c
765 765 */
766 766 extern mod_prop_info_t ip_propinfo_tbl[];
767 767 extern int ip_propinfo_count;
768 768
769 769 /*
770 770 * Table of IP ioctls encoding the various properties of the ioctl and
771 771 * indexed based on the last byte of the ioctl command. Occasionally there
772 772 * is a clash, and there is more than 1 ioctl with the same last byte.
773 773 * In such a case 1 ioctl is encoded in the ndx table and the remaining
774 774 * ioctls are encoded in the misc table. An entry in the ndx table is
775 775 * retrieved by indexing on the last byte of the ioctl command and comparing
776 776 * the ioctl command with the value in the ndx table. In the event of a
777 777 * mismatch the misc table is then searched sequentially for the desired
778 778 * ioctl command.
779 779 *
780 780 * Entry: <command> <copyin_size> <flags> <cmd_type> <function> <restart_func>
781 781 */
782 782 ip_ioctl_cmd_t ip_ndx_ioctl_table[] = {
783 783 /* 000 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
784 784 /* 001 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
785 785 /* 002 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
786 786 /* 003 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
787 787 /* 004 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
788 788 /* 005 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
789 789 /* 006 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
790 790 /* 007 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
791 791 /* 008 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
792 792 /* 009 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
793 793
794 794 /* 010 */ { SIOCADDRT, sizeof (struct rtentry), IPI_PRIV,
795 795 MISC_CMD, ip_siocaddrt, NULL },
796 796 /* 011 */ { SIOCDELRT, sizeof (struct rtentry), IPI_PRIV,
797 797 MISC_CMD, ip_siocdelrt, NULL },
798 798
799 799 /* 012 */ { SIOCSIFADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
800 800 IF_CMD, ip_sioctl_addr, ip_sioctl_addr_restart },
801 801 /* 013 */ { SIOCGIFADDR, sizeof (struct ifreq), IPI_GET_CMD,
802 802 IF_CMD, ip_sioctl_get_addr, NULL },
803 803
804 804 /* 014 */ { SIOCSIFDSTADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
805 805 IF_CMD, ip_sioctl_dstaddr, ip_sioctl_dstaddr_restart },
806 806 /* 015 */ { SIOCGIFDSTADDR, sizeof (struct ifreq),
807 807 IPI_GET_CMD, IF_CMD, ip_sioctl_get_dstaddr, NULL },
808 808
809 809 /* 016 */ { SIOCSIFFLAGS, sizeof (struct ifreq),
810 810 IPI_PRIV | IPI_WR,
811 811 IF_CMD, ip_sioctl_flags, ip_sioctl_flags_restart },
812 812 /* 017 */ { SIOCGIFFLAGS, sizeof (struct ifreq),
813 813 IPI_MODOK | IPI_GET_CMD,
814 814 IF_CMD, ip_sioctl_get_flags, NULL },
815 815
816 816 /* 018 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
817 817 /* 019 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
818 818
819 819 /* copyin size cannot be coded for SIOCGIFCONF */
820 820 /* 020 */ { O_SIOCGIFCONF, 0, IPI_GET_CMD,
821 821 MISC_CMD, ip_sioctl_get_ifconf, NULL },
822 822
823 823 /* 021 */ { SIOCSIFMTU, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
824 824 IF_CMD, ip_sioctl_mtu, NULL },
825 825 /* 022 */ { SIOCGIFMTU, sizeof (struct ifreq), IPI_GET_CMD,
826 826 IF_CMD, ip_sioctl_get_mtu, NULL },
827 827 /* 023 */ { SIOCGIFBRDADDR, sizeof (struct ifreq),
828 828 IPI_GET_CMD, IF_CMD, ip_sioctl_get_brdaddr, NULL },
829 829 /* 024 */ { SIOCSIFBRDADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
830 830 IF_CMD, ip_sioctl_brdaddr, NULL },
831 831 /* 025 */ { SIOCGIFNETMASK, sizeof (struct ifreq),
832 832 IPI_GET_CMD, IF_CMD, ip_sioctl_get_netmask, NULL },
833 833 /* 026 */ { SIOCSIFNETMASK, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
834 834 IF_CMD, ip_sioctl_netmask, ip_sioctl_netmask_restart },
835 835 /* 027 */ { SIOCGIFMETRIC, sizeof (struct ifreq),
836 836 IPI_GET_CMD, IF_CMD, ip_sioctl_get_metric, NULL },
837 837 /* 028 */ { SIOCSIFMETRIC, sizeof (struct ifreq), IPI_PRIV,
838 838 IF_CMD, ip_sioctl_metric, NULL },
839 839 /* 029 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
840 840
841 841 /* See 166-168 below for extended SIOC*XARP ioctls */
842 842 /* 030 */ { SIOCSARP, sizeof (struct arpreq), IPI_PRIV | IPI_WR,
843 843 ARP_CMD, ip_sioctl_arp, NULL },
844 844 /* 031 */ { SIOCGARP, sizeof (struct arpreq), IPI_GET_CMD,
845 845 ARP_CMD, ip_sioctl_arp, NULL },
846 846 /* 032 */ { SIOCDARP, sizeof (struct arpreq), IPI_PRIV | IPI_WR,
847 847 ARP_CMD, ip_sioctl_arp, NULL },
848 848
849 849 /* 033 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
850 850 /* 034 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
851 851 /* 035 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
852 852 /* 036 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
853 853 /* 037 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
854 854 /* 038 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
855 855 /* 039 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
856 856 /* 040 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
857 857 /* 041 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
858 858 /* 042 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
859 859 /* 043 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
860 860 /* 044 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
861 861 /* 045 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
862 862 /* 046 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
863 863 /* 047 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
864 864 /* 048 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
865 865 /* 049 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
866 866 /* 050 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
867 867 /* 051 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
868 868 /* 052 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
869 869 /* 053 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
870 870
871 871 /* 054 */ { IF_UNITSEL, sizeof (int), IPI_PRIV | IPI_WR | IPI_MODOK,
872 872 MISC_CMD, if_unitsel, if_unitsel_restart },
873 873
874 874 /* 055 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
875 875 /* 056 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
876 876 /* 057 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
877 877 /* 058 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
878 878 /* 059 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
879 879 /* 060 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
880 880 /* 061 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
881 881 /* 062 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
882 882 /* 063 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
883 883 /* 064 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
884 884 /* 065 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
885 885 /* 066 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
886 886 /* 067 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
887 887 /* 068 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
888 888 /* 069 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
889 889 /* 070 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
890 890 /* 071 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
891 891 /* 072 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
892 892
893 893 /* 073 */ { SIOCSIFNAME, sizeof (struct ifreq),
894 894 IPI_PRIV | IPI_WR | IPI_MODOK,
895 895 IF_CMD, ip_sioctl_sifname, NULL },
896 896
897 897 /* 074 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
898 898 /* 075 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
899 899 /* 076 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
900 900 /* 077 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
901 901 /* 078 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
902 902 /* 079 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
903 903 /* 080 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
904 904 /* 081 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
905 905 /* 082 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
906 906 /* 083 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
907 907 /* 084 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
908 908 /* 085 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
909 909 /* 086 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
910 910
911 911 /* 087 */ { SIOCGIFNUM, sizeof (int), IPI_GET_CMD,
912 912 MISC_CMD, ip_sioctl_get_ifnum, NULL },
913 913 /* 088 */ { SIOCGIFMUXID, sizeof (struct ifreq), IPI_GET_CMD,
914 914 IF_CMD, ip_sioctl_get_muxid, NULL },
915 915 /* 089 */ { SIOCSIFMUXID, sizeof (struct ifreq),
916 916 IPI_PRIV | IPI_WR, IF_CMD, ip_sioctl_muxid, NULL },
917 917
918 918 /* Both if and lif variants share same func */
919 919 /* 090 */ { SIOCGIFINDEX, sizeof (struct ifreq), IPI_GET_CMD,
920 920 IF_CMD, ip_sioctl_get_lifindex, NULL },
921 921 /* Both if and lif variants share same func */
922 922 /* 091 */ { SIOCSIFINDEX, sizeof (struct ifreq),
923 923 IPI_PRIV | IPI_WR, IF_CMD, ip_sioctl_slifindex, NULL },
924 924
925 925 /* copyin size cannot be coded for SIOCGIFCONF */
926 926 /* 092 */ { SIOCGIFCONF, 0, IPI_GET_CMD,
927 927 MISC_CMD, ip_sioctl_get_ifconf, NULL },
928 928 /* 093 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
929 929 /* 094 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
930 930 /* 095 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
931 931 /* 096 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
932 932 /* 097 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
933 933 /* 098 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
934 934 /* 099 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
935 935 /* 100 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
936 936 /* 101 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
937 937 /* 102 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
938 938 /* 103 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
939 939 /* 104 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
940 940 /* 105 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
941 941 /* 106 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
942 942 /* 107 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
943 943 /* 108 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
944 944 /* 109 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
945 945
946 946 /* 110 */ { SIOCLIFREMOVEIF, sizeof (struct lifreq),
947 947 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_removeif,
948 948 ip_sioctl_removeif_restart },
949 949 /* 111 */ { SIOCLIFADDIF, sizeof (struct lifreq),
950 950 IPI_GET_CMD | IPI_PRIV | IPI_WR,
951 951 LIF_CMD, ip_sioctl_addif, NULL },
952 952 #define SIOCLIFADDR_NDX 112
953 953 /* 112 */ { SIOCSLIFADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
954 954 LIF_CMD, ip_sioctl_addr, ip_sioctl_addr_restart },
955 955 /* 113 */ { SIOCGLIFADDR, sizeof (struct lifreq),
956 956 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_addr, NULL },
957 957 /* 114 */ { SIOCSLIFDSTADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
958 958 LIF_CMD, ip_sioctl_dstaddr, ip_sioctl_dstaddr_restart },
959 959 /* 115 */ { SIOCGLIFDSTADDR, sizeof (struct lifreq),
960 960 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_dstaddr, NULL },
961 961 /* 116 */ { SIOCSLIFFLAGS, sizeof (struct lifreq),
962 962 IPI_PRIV | IPI_WR,
963 963 LIF_CMD, ip_sioctl_flags, ip_sioctl_flags_restart },
964 964 /* 117 */ { SIOCGLIFFLAGS, sizeof (struct lifreq),
965 965 IPI_GET_CMD | IPI_MODOK,
966 966 LIF_CMD, ip_sioctl_get_flags, NULL },
967 967
968 968 /* 118 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
969 969 /* 119 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
970 970
971 971 /* 120 */ { O_SIOCGLIFCONF, 0, IPI_GET_CMD, MISC_CMD,
972 972 ip_sioctl_get_lifconf, NULL },
973 973 /* 121 */ { SIOCSLIFMTU, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
974 974 LIF_CMD, ip_sioctl_mtu, NULL },
975 975 /* 122 */ { SIOCGLIFMTU, sizeof (struct lifreq), IPI_GET_CMD,
976 976 LIF_CMD, ip_sioctl_get_mtu, NULL },
977 977 /* 123 */ { SIOCGLIFBRDADDR, sizeof (struct lifreq),
978 978 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_brdaddr, NULL },
979 979 /* 124 */ { SIOCSLIFBRDADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
980 980 LIF_CMD, ip_sioctl_brdaddr, NULL },
981 981 /* 125 */ { SIOCGLIFNETMASK, sizeof (struct lifreq),
982 982 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_netmask, NULL },
983 983 /* 126 */ { SIOCSLIFNETMASK, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
984 984 LIF_CMD, ip_sioctl_netmask, ip_sioctl_netmask_restart },
985 985 /* 127 */ { SIOCGLIFMETRIC, sizeof (struct lifreq),
986 986 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_metric, NULL },
987 987 /* 128 */ { SIOCSLIFMETRIC, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
988 988 LIF_CMD, ip_sioctl_metric, NULL },
989 989 /* 129 */ { SIOCSLIFNAME, sizeof (struct lifreq),
990 990 IPI_PRIV | IPI_WR | IPI_MODOK,
991 991 LIF_CMD, ip_sioctl_slifname,
992 992 ip_sioctl_slifname_restart },
993 993
994 994 /* 130 */ { SIOCGLIFNUM, sizeof (struct lifnum), IPI_GET_CMD,
995 995 MISC_CMD, ip_sioctl_get_lifnum, NULL },
996 996 /* 131 */ { SIOCGLIFMUXID, sizeof (struct lifreq),
997 997 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_muxid, NULL },
998 998 /* 132 */ { SIOCSLIFMUXID, sizeof (struct lifreq),
999 999 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_muxid, NULL },
1000 1000 /* 133 */ { SIOCGLIFINDEX, sizeof (struct lifreq),
1001 1001 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lifindex, 0 },
1002 1002 /* 134 */ { SIOCSLIFINDEX, sizeof (struct lifreq),
1003 1003 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_slifindex, 0 },
1004 1004 /* 135 */ { SIOCSLIFTOKEN, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1005 1005 LIF_CMD, ip_sioctl_token, NULL },
1006 1006 /* 136 */ { SIOCGLIFTOKEN, sizeof (struct lifreq),
1007 1007 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_token, NULL },
1008 1008 /* 137 */ { SIOCSLIFSUBNET, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1009 1009 LIF_CMD, ip_sioctl_subnet, ip_sioctl_subnet_restart },
1010 1010 /* 138 */ { SIOCGLIFSUBNET, sizeof (struct lifreq),
1011 1011 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_subnet, NULL },
1012 1012 /* 139 */ { SIOCSLIFLNKINFO, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1013 1013 LIF_CMD, ip_sioctl_lnkinfo, NULL },
1014 1014
1015 1015 /* 140 */ { SIOCGLIFLNKINFO, sizeof (struct lifreq),
1016 1016 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lnkinfo, NULL },
1017 1017 /* 141 */ { SIOCLIFDELND, sizeof (struct lifreq), IPI_PRIV,
1018 1018 LIF_CMD, ip_siocdelndp_v6, NULL },
1019 1019 /* 142 */ { SIOCLIFGETND, sizeof (struct lifreq), IPI_GET_CMD,
1020 1020 LIF_CMD, ip_siocqueryndp_v6, NULL },
1021 1021 /* 143 */ { SIOCLIFSETND, sizeof (struct lifreq), IPI_PRIV,
1022 1022 LIF_CMD, ip_siocsetndp_v6, NULL },
1023 1023 /* 144 */ { SIOCTMYADDR, sizeof (struct sioc_addrreq), IPI_GET_CMD,
1024 1024 MISC_CMD, ip_sioctl_tmyaddr, NULL },
1025 1025 /* 145 */ { SIOCTONLINK, sizeof (struct sioc_addrreq), IPI_GET_CMD,
1026 1026 MISC_CMD, ip_sioctl_tonlink, NULL },
1027 1027 /* 146 */ { SIOCTMYSITE, sizeof (struct sioc_addrreq), 0,
1028 1028 MISC_CMD, ip_sioctl_tmysite, NULL },
1029 1029 /* 147 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1030 1030 /* 148 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1031 1031
1032 1032 /* Old *IPSECONFIG ioctls are now deprecated, now see spdsock.c */
1033 1033 /* 149 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1034 1034 /* 150 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1035 1035 /* 151 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1036 1036 /* 152 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1037 1037
1038 1038 /* 153 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1039 1039
1040 1040 /* 154 */ { SIOCGLIFBINDING, sizeof (struct lifreq), IPI_GET_CMD,
1041 1041 LIF_CMD, ip_sioctl_get_binding, NULL },
1042 1042 /* 155 */ { SIOCSLIFGROUPNAME, sizeof (struct lifreq),
1043 1043 IPI_PRIV | IPI_WR,
1044 1044 LIF_CMD, ip_sioctl_groupname, ip_sioctl_groupname },
1045 1045 /* 156 */ { SIOCGLIFGROUPNAME, sizeof (struct lifreq),
1046 1046 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_groupname, NULL },
1047 1047 /* 157 */ { SIOCGLIFGROUPINFO, sizeof (lifgroupinfo_t),
1048 1048 IPI_GET_CMD, MISC_CMD, ip_sioctl_groupinfo, NULL },
1049 1049
1050 1050 /* Leave 158-160 unused; used to be SIOC*IFARP ioctls */
1051 1051 /* 158 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1052 1052 /* 159 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1053 1053 /* 160 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1054 1054
1055 1055 /* 161 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1056 1056
1057 1057 /* These are handled in ip_sioctl_copyin_setup itself */
1058 1058 /* 162 */ { SIOCGIP6ADDRPOLICY, 0, IPI_NULL_BCONT,
1059 1059 MISC_CMD, NULL, NULL },
1060 1060 /* 163 */ { SIOCSIP6ADDRPOLICY, 0, IPI_PRIV | IPI_NULL_BCONT,
1061 1061 MISC_CMD, NULL, NULL },
1062 1062 /* 164 */ { SIOCGDSTINFO, 0, IPI_GET_CMD, MISC_CMD, NULL, NULL },
1063 1063
1064 1064 /* 165 */ { SIOCGLIFCONF, 0, IPI_GET_CMD, MISC_CMD,
1065 1065 ip_sioctl_get_lifconf, NULL },
1066 1066
1067 1067 /* 166 */ { SIOCSXARP, sizeof (struct xarpreq), IPI_PRIV | IPI_WR,
1068 1068 XARP_CMD, ip_sioctl_arp, NULL },
1069 1069 /* 167 */ { SIOCGXARP, sizeof (struct xarpreq), IPI_GET_CMD,
1070 1070 XARP_CMD, ip_sioctl_arp, NULL },
1071 1071 /* 168 */ { SIOCDXARP, sizeof (struct xarpreq), IPI_PRIV | IPI_WR,
1072 1072 XARP_CMD, ip_sioctl_arp, NULL },
1073 1073
1074 1074 /* SIOCPOPSOCKFS is not handled by IP */
1075 1075 /* 169 */ { IPI_DONTCARE /* SIOCPOPSOCKFS */, 0, 0, 0, NULL, NULL },
1076 1076
1077 1077 /* 170 */ { SIOCGLIFZONE, sizeof (struct lifreq),
1078 1078 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lifzone, NULL },
1079 1079 /* 171 */ { SIOCSLIFZONE, sizeof (struct lifreq),
1080 1080 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_slifzone,
1081 1081 ip_sioctl_slifzone_restart },
1082 1082 /* 172-174 are SCTP ioctls and not handled by IP */
1083 1083 /* 172 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1084 1084 /* 173 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1085 1085 /* 174 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1086 1086 /* 175 */ { SIOCGLIFUSESRC, sizeof (struct lifreq),
1087 1087 IPI_GET_CMD, LIF_CMD,
1088 1088 ip_sioctl_get_lifusesrc, 0 },
1089 1089 /* 176 */ { SIOCSLIFUSESRC, sizeof (struct lifreq),
1090 1090 IPI_PRIV | IPI_WR,
1091 1091 LIF_CMD, ip_sioctl_slifusesrc,
1092 1092 NULL },
1093 1093 /* 177 */ { SIOCGLIFSRCOF, 0, IPI_GET_CMD, MISC_CMD,
1094 1094 ip_sioctl_get_lifsrcof, NULL },
1095 1095 /* 178 */ { SIOCGMSFILTER, sizeof (struct group_filter), IPI_GET_CMD,
1096 1096 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1097 1097 /* 179 */ { SIOCSMSFILTER, sizeof (struct group_filter), 0,
1098 1098 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1099 1099 /* 180 */ { SIOCGIPMSFILTER, sizeof (struct ip_msfilter), IPI_GET_CMD,
1100 1100 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1101 1101 /* 181 */ { SIOCSIPMSFILTER, sizeof (struct ip_msfilter), 0,
1102 1102 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1103 1103 /* 182 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1104 1104 /* SIOCSENABLESDP is handled by SDP */
1105 1105 /* 183 */ { IPI_DONTCARE /* SIOCSENABLESDP */, 0, 0, 0, NULL, NULL },
1106 1106 /* 184 */ { IPI_DONTCARE /* SIOCSQPTR */, 0, 0, 0, NULL, NULL },
1107 1107 /* 185 */ { SIOCGIFHWADDR, sizeof (struct ifreq), IPI_GET_CMD,
1108 1108 IF_CMD, ip_sioctl_get_ifhwaddr, NULL },
1109 1109 /* 186 */ { IPI_DONTCARE /* SIOCGSTAMP */, 0, 0, 0, NULL, NULL },
1110 1110 /* 187 */ { SIOCILB, 0, IPI_PRIV | IPI_GET_CMD, MISC_CMD,
1111 1111 ip_sioctl_ilb_cmd, NULL },
1112 1112 /* 188 */ { SIOCGETPROP, 0, IPI_GET_CMD, 0, NULL, NULL },
1113 1113 /* 189 */ { SIOCSETPROP, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL},
1114 1114 /* 190 */ { SIOCGLIFDADSTATE, sizeof (struct lifreq),
1115 1115 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_dadstate, NULL },
1116 1116 /* 191 */ { SIOCSLIFPREFIX, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1117 1117 LIF_CMD, ip_sioctl_prefix, ip_sioctl_prefix_restart },
1118 1118 /* 192 */ { SIOCGLIFHWADDR, sizeof (struct lifreq), IPI_GET_CMD,
1119 1119 LIF_CMD, ip_sioctl_get_lifhwaddr, NULL }
1120 1120 };
1121 1121
1122 1122 int ip_ndx_ioctl_count = sizeof (ip_ndx_ioctl_table) / sizeof (ip_ioctl_cmd_t);
1123 1123
1124 1124 ip_ioctl_cmd_t ip_misc_ioctl_table[] = {
1125 1125 { I_LINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1126 1126 { I_UNLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1127 1127 { I_PLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1128 1128 { I_PUNLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1129 1129 { ND_GET, 0, 0, 0, NULL, NULL },
1130 1130 { ND_SET, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1131 1131 { IP_IOCTL, 0, 0, 0, NULL, NULL },
1132 1132 { SIOCGETVIFCNT, sizeof (struct sioc_vif_req), IPI_GET_CMD,
1133 1133 MISC_CMD, mrt_ioctl},
1134 1134 { SIOCGETSGCNT, sizeof (struct sioc_sg_req), IPI_GET_CMD,
1135 1135 MISC_CMD, mrt_ioctl},
1136 1136 { SIOCGETLSGCNT, sizeof (struct sioc_lsg_req), IPI_GET_CMD,
1137 1137 MISC_CMD, mrt_ioctl}
1138 1138 };
1139 1139
1140 1140 int ip_misc_ioctl_count =
1141 1141 sizeof (ip_misc_ioctl_table) / sizeof (ip_ioctl_cmd_t);
1142 1142
1143 1143 int conn_drain_nthreads; /* Number of drainers reqd. */
1144 1144 /* Settable in /etc/system */
1145 1145 /* Defined in ip_ire.c */
1146 1146 extern uint32_t ip_ire_max_bucket_cnt, ip6_ire_max_bucket_cnt;
1147 1147 extern uint32_t ip_ire_min_bucket_cnt, ip6_ire_min_bucket_cnt;
1148 1148 extern uint32_t ip_ire_mem_ratio, ip_ire_cpu_ratio;
1149 1149
1150 1150 static nv_t ire_nv_arr[] = {
1151 1151 { IRE_BROADCAST, "BROADCAST" },
1152 1152 { IRE_LOCAL, "LOCAL" },
1153 1153 { IRE_LOOPBACK, "LOOPBACK" },
1154 1154 { IRE_DEFAULT, "DEFAULT" },
1155 1155 { IRE_PREFIX, "PREFIX" },
1156 1156 { IRE_IF_NORESOLVER, "IF_NORESOL" },
1157 1157 { IRE_IF_RESOLVER, "IF_RESOLV" },
1158 1158 { IRE_IF_CLONE, "IF_CLONE" },
1159 1159 { IRE_HOST, "HOST" },
1160 1160 { IRE_MULTICAST, "MULTICAST" },
1161 1161 { IRE_NOROUTE, "NOROUTE" },
1162 1162 { 0 }
1163 1163 };
1164 1164
1165 1165 nv_t *ire_nv_tbl = ire_nv_arr;
1166 1166
1167 1167 /* Simple ICMP IP Header Template */
1168 1168 static ipha_t icmp_ipha = {
1169 1169 IP_SIMPLE_HDR_VERSION, 0, 0, 0, 0, 0, IPPROTO_ICMP
1170 1170 };
1171 1171
1172 1172 struct module_info ip_mod_info = {
1173 1173 IP_MOD_ID, IP_MOD_NAME, IP_MOD_MINPSZ, IP_MOD_MAXPSZ, IP_MOD_HIWAT,
1174 1174 IP_MOD_LOWAT
1175 1175 };
1176 1176
1177 1177 /*
1178 1178 * Duplicate static symbols within a module confuses mdb; so we avoid the
1179 1179 * problem by making the symbols here distinct from those in udp.c.
1180 1180 */
1181 1181
1182 1182 /*
1183 1183 * Entry points for IP as a device and as a module.
1184 1184 * We have separate open functions for the /dev/ip and /dev/ip6 devices.
1185 1185 */
1186 1186 static struct qinit iprinitv4 = {
1187 1187 ip_rput, NULL, ip_openv4, ip_close, NULL, &ip_mod_info
1188 1188 };
1189 1189
1190 1190 struct qinit iprinitv6 = {
1191 1191 ip_rput_v6, NULL, ip_openv6, ip_close, NULL, &ip_mod_info
1192 1192 };
1193 1193
1194 1194 static struct qinit ipwinit = {
1195 1195 ip_wput_nondata, ip_wsrv, NULL, NULL, NULL, &ip_mod_info
1196 1196 };
1197 1197
1198 1198 static struct qinit iplrinit = {
1199 1199 ip_lrput, NULL, ip_openv4, ip_close, NULL, &ip_mod_info
1200 1200 };
1201 1201
1202 1202 static struct qinit iplwinit = {
1203 1203 ip_lwput, NULL, NULL, NULL, NULL, &ip_mod_info
1204 1204 };
1205 1205
1206 1206 /* For AF_INET aka /dev/ip */
1207 1207 struct streamtab ipinfov4 = {
1208 1208 &iprinitv4, &ipwinit, &iplrinit, &iplwinit
1209 1209 };
1210 1210
1211 1211 /* For AF_INET6 aka /dev/ip6 */
1212 1212 struct streamtab ipinfov6 = {
1213 1213 &iprinitv6, &ipwinit, &iplrinit, &iplwinit
1214 1214 };
1215 1215
1216 1216 #ifdef DEBUG
1217 1217 boolean_t skip_sctp_cksum = B_FALSE;
1218 1218 #endif
1219 1219
1220 1220 /*
1221 1221 * Generate an ICMP fragmentation needed message.
1222 1222 * When called from ip_output side a minimal ip_recv_attr_t needs to be
1223 1223 * constructed by the caller.
1224 1224 */
1225 1225 void
1226 1226 icmp_frag_needed(mblk_t *mp, int mtu, ip_recv_attr_t *ira)
1227 1227 {
1228 1228 icmph_t icmph;
1229 1229 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
1230 1230
1231 1231 mp = icmp_pkt_err_ok(mp, ira);
1232 1232 if (mp == NULL)
1233 1233 return;
1234 1234
1235 1235 bzero(&icmph, sizeof (icmph_t));
1236 1236 icmph.icmph_type = ICMP_DEST_UNREACHABLE;
1237 1237 icmph.icmph_code = ICMP_FRAGMENTATION_NEEDED;
1238 1238 icmph.icmph_du_mtu = htons((uint16_t)mtu);
1239 1239 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutFragNeeded);
1240 1240 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDestUnreachs);
1241 1241
1242 1242 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
1243 1243 }
1244 1244
1245 1245 /*
1246 1246 * icmp_inbound_v4 deals with ICMP messages that are handled by IP.
1247 1247 * If the ICMP message is consumed by IP, i.e., it should not be delivered
1248 1248 * to any IPPROTO_ICMP raw sockets, then it returns NULL.
1249 1249 * Likewise, if the ICMP error is misformed (too short, etc), then it
1250 1250 * returns NULL. The caller uses this to determine whether or not to send
1251 1251 * to raw sockets.
1252 1252 *
1253 1253 * All error messages are passed to the matching transport stream.
1254 1254 *
1255 1255 * The following cases are handled by icmp_inbound:
1256 1256 * 1) It needs to send a reply back and possibly delivering it
1257 1257 * to the "interested" upper clients.
1258 1258 * 2) Return the mblk so that the caller can pass it to the RAW socket clients.
1259 1259 * 3) It needs to change some values in IP only.
1260 1260 * 4) It needs to change some values in IP and upper layers e.g TCP
1261 1261 * by delivering an error to the upper layers.
1262 1262 *
1263 1263 * We handle the above three cases in the context of IPsec in the
1264 1264 * following way :
1265 1265 *
1266 1266 * 1) Send the reply back in the same way as the request came in.
1267 1267 * If it came in encrypted, it goes out encrypted. If it came in
1268 1268 * clear, it goes out in clear. Thus, this will prevent chosen
1269 1269 * plain text attack.
1270 1270 * 2) The client may or may not expect things to come in secure.
1271 1271 * If it comes in secure, the policy constraints are checked
1272 1272 * before delivering it to the upper layers. If it comes in
1273 1273 * clear, ipsec_inbound_accept_clear will decide whether to
1274 1274 * accept this in clear or not. In both the cases, if the returned
1275 1275 * message (IP header + 8 bytes) that caused the icmp message has
1276 1276 * AH/ESP headers, it is sent up to AH/ESP for validation before
1277 1277 * sending up. If there are only 8 bytes of returned message, then
1278 1278 * upper client will not be notified.
1279 1279 * 3) Check with global policy to see whether it matches the constaints.
1280 1280 * But this will be done only if icmp_accept_messages_in_clear is
1281 1281 * zero.
1282 1282 * 4) If we need to change both in IP and ULP, then the decision taken
1283 1283 * while affecting the values in IP and while delivering up to TCP
1284 1284 * should be the same.
1285 1285 *
1286 1286 * There are two cases.
1287 1287 *
1288 1288 * a) If we reject data at the IP layer (ipsec_check_global_policy()
1289 1289 * failed), we will not deliver it to the ULP, even though they
1290 1290 * are *willing* to accept in *clear*. This is fine as our global
1291 1291 * disposition to icmp messages asks us reject the datagram.
1292 1292 *
1293 1293 * b) If we accept data at the IP layer (ipsec_check_global_policy()
1294 1294 * succeeded or icmp_accept_messages_in_clear is 1), and not able
1295 1295 * to deliver it to ULP (policy failed), it can lead to
1296 1296 * consistency problems. The cases known at this time are
1297 1297 * ICMP_DESTINATION_UNREACHABLE messages with following code
1298 1298 * values :
1299 1299 *
1300 1300 * - ICMP_FRAGMENTATION_NEEDED : IP adapts to the new value
1301 1301 * and Upper layer rejects. Then the communication will
1302 1302 * come to a stop. This is solved by making similar decisions
1303 1303 * at both levels. Currently, when we are unable to deliver
1304 1304 * to the Upper Layer (due to policy failures) while IP has
1305 1305 * adjusted dce_pmtu, the next outbound datagram would
1306 1306 * generate a local ICMP_FRAGMENTATION_NEEDED message - which
1307 1307 * will be with the right level of protection. Thus the right
1308 1308 * value will be communicated even if we are not able to
1309 1309 * communicate when we get from the wire initially. But this
1310 1310 * assumes there would be at least one outbound datagram after
1311 1311 * IP has adjusted its dce_pmtu value. To make things
1312 1312 * simpler, we accept in clear after the validation of
1313 1313 * AH/ESP headers.
1314 1314 *
1315 1315 * - Other ICMP ERRORS : We may not be able to deliver it to the
1316 1316 * upper layer depending on the level of protection the upper
1317 1317 * layer expects and the disposition in ipsec_inbound_accept_clear().
1318 1318 * ipsec_inbound_accept_clear() decides whether a given ICMP error
1319 1319 * should be accepted in clear when the Upper layer expects secure.
1320 1320 * Thus the communication may get aborted by some bad ICMP
1321 1321 * packets.
1322 1322 */
1323 1323 mblk_t *
1324 1324 icmp_inbound_v4(mblk_t *mp, ip_recv_attr_t *ira)
1325 1325 {
1326 1326 icmph_t *icmph;
1327 1327 ipha_t *ipha; /* Outer header */
1328 1328 int ip_hdr_length; /* Outer header length */
1329 1329 boolean_t interested;
1330 1330 ipif_t *ipif;
1331 1331 uint32_t ts;
1332 1332 uint32_t *tsp;
1333 1333 timestruc_t now;
1334 1334 ill_t *ill = ira->ira_ill;
1335 1335 ip_stack_t *ipst = ill->ill_ipst;
1336 1336 zoneid_t zoneid = ira->ira_zoneid;
1337 1337 int len_needed;
1338 1338 mblk_t *mp_ret = NULL;
1339 1339
1340 1340 ipha = (ipha_t *)mp->b_rptr;
1341 1341
1342 1342 BUMP_MIB(&ipst->ips_icmp_mib, icmpInMsgs);
1343 1343
1344 1344 ip_hdr_length = ira->ira_ip_hdr_length;
1345 1345 if ((mp->b_wptr - mp->b_rptr) < (ip_hdr_length + ICMPH_SIZE)) {
1346 1346 if (ira->ira_pktlen < (ip_hdr_length + ICMPH_SIZE)) {
1347 1347 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
1348 1348 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
1349 1349 freemsg(mp);
1350 1350 return (NULL);
1351 1351 }
1352 1352 /* Last chance to get real. */
1353 1353 ipha = ip_pullup(mp, ip_hdr_length + ICMPH_SIZE, ira);
1354 1354 if (ipha == NULL) {
1355 1355 BUMP_MIB(&ipst->ips_icmp_mib, icmpInErrors);
1356 1356 freemsg(mp);
1357 1357 return (NULL);
1358 1358 }
1359 1359 }
1360 1360
1361 1361 /* The IP header will always be a multiple of four bytes */
1362 1362 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1363 1363 ip2dbg(("icmp_inbound_v4: type %d code %d\n", icmph->icmph_type,
1364 1364 icmph->icmph_code));
1365 1365
1366 1366 /*
1367 1367 * We will set "interested" to "true" if we should pass a copy to
1368 1368 * the transport or if we handle the packet locally.
1369 1369 */
1370 1370 interested = B_FALSE;
1371 1371 switch (icmph->icmph_type) {
1372 1372 case ICMP_ECHO_REPLY:
1373 1373 BUMP_MIB(&ipst->ips_icmp_mib, icmpInEchoReps);
1374 1374 break;
1375 1375 case ICMP_DEST_UNREACHABLE:
1376 1376 if (icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED)
1377 1377 BUMP_MIB(&ipst->ips_icmp_mib, icmpInFragNeeded);
1378 1378 interested = B_TRUE; /* Pass up to transport */
1379 1379 BUMP_MIB(&ipst->ips_icmp_mib, icmpInDestUnreachs);
1380 1380 break;
1381 1381 case ICMP_SOURCE_QUENCH:
1382 1382 interested = B_TRUE; /* Pass up to transport */
1383 1383 BUMP_MIB(&ipst->ips_icmp_mib, icmpInSrcQuenchs);
1384 1384 break;
1385 1385 case ICMP_REDIRECT:
1386 1386 if (!ipst->ips_ip_ignore_redirect)
1387 1387 interested = B_TRUE;
1388 1388 BUMP_MIB(&ipst->ips_icmp_mib, icmpInRedirects);
1389 1389 break;
1390 1390 case ICMP_ECHO_REQUEST:
1391 1391 /*
1392 1392 * Whether to respond to echo requests that come in as IP
1393 1393 * broadcasts or as IP multicast is subject to debate
1394 1394 * (what isn't?). We aim to please, you pick it.
1395 1395 * Default is do it.
1396 1396 */
1397 1397 if (ira->ira_flags & IRAF_MULTICAST) {
1398 1398 /* multicast: respond based on tunable */
1399 1399 interested = ipst->ips_ip_g_resp_to_echo_mcast;
1400 1400 } else if (ira->ira_flags & IRAF_BROADCAST) {
1401 1401 /* broadcast: respond based on tunable */
1402 1402 interested = ipst->ips_ip_g_resp_to_echo_bcast;
1403 1403 } else {
1404 1404 /* unicast: always respond */
1405 1405 interested = B_TRUE;
1406 1406 }
1407 1407 BUMP_MIB(&ipst->ips_icmp_mib, icmpInEchos);
1408 1408 if (!interested) {
1409 1409 /* We never pass these to RAW sockets */
1410 1410 freemsg(mp);
1411 1411 return (NULL);
1412 1412 }
1413 1413
1414 1414 /* Check db_ref to make sure we can modify the packet. */
1415 1415 if (mp->b_datap->db_ref > 1) {
1416 1416 mblk_t *mp1;
1417 1417
1418 1418 mp1 = copymsg(mp);
1419 1419 freemsg(mp);
1420 1420 if (!mp1) {
1421 1421 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1422 1422 return (NULL);
1423 1423 }
1424 1424 mp = mp1;
1425 1425 ipha = (ipha_t *)mp->b_rptr;
1426 1426 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1427 1427 }
1428 1428 icmph->icmph_type = ICMP_ECHO_REPLY;
1429 1429 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutEchoReps);
1430 1430 icmp_send_reply_v4(mp, ipha, icmph, ira);
1431 1431 return (NULL);
1432 1432
1433 1433 case ICMP_ROUTER_ADVERTISEMENT:
1434 1434 case ICMP_ROUTER_SOLICITATION:
1435 1435 break;
1436 1436 case ICMP_TIME_EXCEEDED:
1437 1437 interested = B_TRUE; /* Pass up to transport */
1438 1438 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimeExcds);
1439 1439 break;
1440 1440 case ICMP_PARAM_PROBLEM:
1441 1441 interested = B_TRUE; /* Pass up to transport */
1442 1442 BUMP_MIB(&ipst->ips_icmp_mib, icmpInParmProbs);
1443 1443 break;
1444 1444 case ICMP_TIME_STAMP_REQUEST:
1445 1445 /* Response to Time Stamp Requests is local policy. */
1446 1446 if (ipst->ips_ip_g_resp_to_timestamp) {
1447 1447 if (ira->ira_flags & IRAF_MULTIBROADCAST)
1448 1448 interested =
1449 1449 ipst->ips_ip_g_resp_to_timestamp_bcast;
1450 1450 else
1451 1451 interested = B_TRUE;
1452 1452 }
1453 1453 if (!interested) {
1454 1454 /* We never pass these to RAW sockets */
1455 1455 freemsg(mp);
1456 1456 return (NULL);
1457 1457 }
1458 1458
1459 1459 /* Make sure we have enough of the packet */
1460 1460 len_needed = ip_hdr_length + ICMPH_SIZE +
1461 1461 3 * sizeof (uint32_t);
1462 1462
1463 1463 if (mp->b_wptr - mp->b_rptr < len_needed) {
1464 1464 ipha = ip_pullup(mp, len_needed, ira);
1465 1465 if (ipha == NULL) {
1466 1466 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1467 1467 ip_drop_input("ipIfStatsInDiscards - ip_pullup",
1468 1468 mp, ill);
1469 1469 freemsg(mp);
1470 1470 return (NULL);
1471 1471 }
1472 1472 /* Refresh following the pullup. */
1473 1473 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1474 1474 }
1475 1475 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimestamps);
1476 1476 /* Check db_ref to make sure we can modify the packet. */
1477 1477 if (mp->b_datap->db_ref > 1) {
1478 1478 mblk_t *mp1;
1479 1479
1480 1480 mp1 = copymsg(mp);
1481 1481 freemsg(mp);
1482 1482 if (!mp1) {
1483 1483 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1484 1484 return (NULL);
1485 1485 }
1486 1486 mp = mp1;
1487 1487 ipha = (ipha_t *)mp->b_rptr;
1488 1488 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1489 1489 }
1490 1490 icmph->icmph_type = ICMP_TIME_STAMP_REPLY;
1491 1491 tsp = (uint32_t *)&icmph[1];
1492 1492 tsp++; /* Skip past 'originate time' */
1493 1493 /* Compute # of milliseconds since midnight */
1494 1494 gethrestime(&now);
1495 1495 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
1496 1496 NSEC2MSEC(now.tv_nsec);
1497 1497 *tsp++ = htonl(ts); /* Lay in 'receive time' */
1498 1498 *tsp++ = htonl(ts); /* Lay in 'send time' */
1499 1499 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutTimestampReps);
1500 1500 icmp_send_reply_v4(mp, ipha, icmph, ira);
1501 1501 return (NULL);
1502 1502
1503 1503 case ICMP_TIME_STAMP_REPLY:
1504 1504 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimestampReps);
1505 1505 break;
1506 1506 case ICMP_INFO_REQUEST:
1507 1507 /* Per RFC 1122 3.2.2.7, ignore this. */
1508 1508 case ICMP_INFO_REPLY:
1509 1509 break;
1510 1510 case ICMP_ADDRESS_MASK_REQUEST:
1511 1511 if (ira->ira_flags & IRAF_MULTIBROADCAST) {
1512 1512 interested =
1513 1513 ipst->ips_ip_respond_to_address_mask_broadcast;
1514 1514 } else {
1515 1515 interested = B_TRUE;
1516 1516 }
1517 1517 if (!interested) {
1518 1518 /* We never pass these to RAW sockets */
1519 1519 freemsg(mp);
1520 1520 return (NULL);
1521 1521 }
1522 1522 len_needed = ip_hdr_length + ICMPH_SIZE + IP_ADDR_LEN;
1523 1523 if (mp->b_wptr - mp->b_rptr < len_needed) {
1524 1524 ipha = ip_pullup(mp, len_needed, ira);
1525 1525 if (ipha == NULL) {
1526 1526 BUMP_MIB(ill->ill_ip_mib,
1527 1527 ipIfStatsInTruncatedPkts);
1528 1528 ip_drop_input("ipIfStatsInTruncatedPkts", mp,
1529 1529 ill);
1530 1530 freemsg(mp);
1531 1531 return (NULL);
1532 1532 }
1533 1533 /* Refresh following the pullup. */
1534 1534 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1535 1535 }
1536 1536 BUMP_MIB(&ipst->ips_icmp_mib, icmpInAddrMasks);
1537 1537 /* Check db_ref to make sure we can modify the packet. */
1538 1538 if (mp->b_datap->db_ref > 1) {
1539 1539 mblk_t *mp1;
1540 1540
1541 1541 mp1 = copymsg(mp);
1542 1542 freemsg(mp);
1543 1543 if (!mp1) {
1544 1544 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1545 1545 return (NULL);
1546 1546 }
1547 1547 mp = mp1;
1548 1548 ipha = (ipha_t *)mp->b_rptr;
1549 1549 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1550 1550 }
1551 1551 /*
1552 1552 * Need the ipif with the mask be the same as the source
1553 1553 * address of the mask reply. For unicast we have a specific
1554 1554 * ipif. For multicast/broadcast we only handle onlink
1555 1555 * senders, and use the source address to pick an ipif.
1556 1556 */
1557 1557 ipif = ipif_lookup_addr(ipha->ipha_dst, ill, zoneid, ipst);
1558 1558 if (ipif == NULL) {
1559 1559 /* Broadcast or multicast */
1560 1560 ipif = ipif_lookup_remote(ill, ipha->ipha_src, zoneid);
1561 1561 if (ipif == NULL) {
1562 1562 freemsg(mp);
1563 1563 return (NULL);
1564 1564 }
1565 1565 }
1566 1566 icmph->icmph_type = ICMP_ADDRESS_MASK_REPLY;
1567 1567 bcopy(&ipif->ipif_net_mask, &icmph[1], IP_ADDR_LEN);
1568 1568 ipif_refrele(ipif);
1569 1569 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutAddrMaskReps);
1570 1570 icmp_send_reply_v4(mp, ipha, icmph, ira);
1571 1571 return (NULL);
1572 1572
1573 1573 case ICMP_ADDRESS_MASK_REPLY:
1574 1574 BUMP_MIB(&ipst->ips_icmp_mib, icmpInAddrMaskReps);
1575 1575 break;
1576 1576 default:
1577 1577 interested = B_TRUE; /* Pass up to transport */
1578 1578 BUMP_MIB(&ipst->ips_icmp_mib, icmpInUnknowns);
1579 1579 break;
1580 1580 }
1581 1581 /*
1582 1582 * See if there is an ICMP client to avoid an extra copymsg/freemsg
1583 1583 * if there isn't one.
1584 1584 */
1585 1585 if (ipst->ips_ipcl_proto_fanout_v4[IPPROTO_ICMP].connf_head != NULL) {
1586 1586 /* If there is an ICMP client and we want one too, copy it. */
1587 1587
1588 1588 if (!interested) {
1589 1589 /* Caller will deliver to RAW sockets */
1590 1590 return (mp);
1591 1591 }
1592 1592 mp_ret = copymsg(mp);
1593 1593 if (mp_ret == NULL) {
1594 1594 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1595 1595 ip_drop_input("ipIfStatsInDiscards - copymsg", mp, ill);
1596 1596 }
1597 1597 } else if (!interested) {
1598 1598 /* Neither we nor raw sockets are interested. Drop packet now */
1599 1599 freemsg(mp);
1600 1600 return (NULL);
1601 1601 }
1602 1602
1603 1603 /*
1604 1604 * ICMP error or redirect packet. Make sure we have enough of
1605 1605 * the header and that db_ref == 1 since we might end up modifying
1606 1606 * the packet.
1607 1607 */
1608 1608 if (mp->b_cont != NULL) {
1609 1609 if (ip_pullup(mp, -1, ira) == NULL) {
1610 1610 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1611 1611 ip_drop_input("ipIfStatsInDiscards - ip_pullup",
1612 1612 mp, ill);
1613 1613 freemsg(mp);
1614 1614 return (mp_ret);
1615 1615 }
1616 1616 }
1617 1617
1618 1618 if (mp->b_datap->db_ref > 1) {
1619 1619 mblk_t *mp1;
1620 1620
1621 1621 mp1 = copymsg(mp);
1622 1622 if (mp1 == NULL) {
1623 1623 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1624 1624 ip_drop_input("ipIfStatsInDiscards - copymsg", mp, ill);
1625 1625 freemsg(mp);
1626 1626 return (mp_ret);
1627 1627 }
1628 1628 freemsg(mp);
1629 1629 mp = mp1;
1630 1630 }
1631 1631
1632 1632 /*
1633 1633 * In case mp has changed, verify the message before any further
1634 1634 * processes.
1635 1635 */
1636 1636 ipha = (ipha_t *)mp->b_rptr;
1637 1637 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1638 1638 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
1639 1639 freemsg(mp);
1640 1640 return (mp_ret);
1641 1641 }
1642 1642
1643 1643 switch (icmph->icmph_type) {
1644 1644 case ICMP_REDIRECT:
1645 1645 icmp_redirect_v4(mp, ipha, icmph, ira);
1646 1646 break;
1647 1647 case ICMP_DEST_UNREACHABLE:
1648 1648 if (icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED) {
1649 1649 /* Update DCE and adjust MTU is icmp header if needed */
1650 1650 icmp_inbound_too_big_v4(icmph, ira);
1651 1651 }
1652 1652 /* FALLTHROUGH */
1653 1653 default:
1654 1654 icmp_inbound_error_fanout_v4(mp, icmph, ira);
1655 1655 break;
1656 1656 }
1657 1657 return (mp_ret);
1658 1658 }
1659 1659
1660 1660 /*
1661 1661 * Send an ICMP echo, timestamp or address mask reply.
1662 1662 * The caller has already updated the payload part of the packet.
1663 1663 * We handle the ICMP checksum, IP source address selection and feed
1664 1664 * the packet into ip_output_simple.
1665 1665 */
1666 1666 static void
1667 1667 icmp_send_reply_v4(mblk_t *mp, ipha_t *ipha, icmph_t *icmph,
1668 1668 ip_recv_attr_t *ira)
1669 1669 {
1670 1670 uint_t ip_hdr_length = ira->ira_ip_hdr_length;
1671 1671 ill_t *ill = ira->ira_ill;
1672 1672 ip_stack_t *ipst = ill->ill_ipst;
1673 1673 ip_xmit_attr_t ixas;
1674 1674
1675 1675 /* Send out an ICMP packet */
1676 1676 icmph->icmph_checksum = 0;
1677 1677 icmph->icmph_checksum = IP_CSUM(mp, ip_hdr_length, 0);
1678 1678 /* Reset time to live. */
1679 1679 ipha->ipha_ttl = ipst->ips_ip_def_ttl;
1680 1680 {
1681 1681 /* Swap source and destination addresses */
1682 1682 ipaddr_t tmp;
1683 1683
1684 1684 tmp = ipha->ipha_src;
1685 1685 ipha->ipha_src = ipha->ipha_dst;
1686 1686 ipha->ipha_dst = tmp;
1687 1687 }
1688 1688 ipha->ipha_ident = 0;
1689 1689 if (!IS_SIMPLE_IPH(ipha))
1690 1690 icmp_options_update(ipha);
1691 1691
1692 1692 bzero(&ixas, sizeof (ixas));
1693 1693 ixas.ixa_flags = IXAF_BASIC_SIMPLE_V4;
1694 1694 ixas.ixa_zoneid = ira->ira_zoneid;
1695 1695 ixas.ixa_cred = kcred;
1696 1696 ixas.ixa_cpid = NOPID;
1697 1697 ixas.ixa_tsl = ira->ira_tsl; /* Behave as a multi-level responder */
1698 1698 ixas.ixa_ifindex = 0;
1699 1699 ixas.ixa_ipst = ipst;
1700 1700 ixas.ixa_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1701 1701
1702 1702 if (!(ira->ira_flags & IRAF_IPSEC_SECURE)) {
1703 1703 /*
1704 1704 * This packet should go out the same way as it
1705 1705 * came in i.e in clear, independent of the IPsec policy
1706 1706 * for transmitting packets.
1707 1707 */
1708 1708 ixas.ixa_flags |= IXAF_NO_IPSEC;
1709 1709 } else {
1710 1710 if (!ipsec_in_to_out(ira, &ixas, mp, ipha, NULL)) {
1711 1711 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1712 1712 /* Note: mp already consumed and ip_drop_packet done */
1713 1713 return;
1714 1714 }
1715 1715 }
1716 1716 if (ira->ira_flags & IRAF_MULTIBROADCAST) {
1717 1717 /*
1718 1718 * Not one or our addresses (IRE_LOCALs), thus we let
1719 1719 * ip_output_simple pick the source.
1720 1720 */
1721 1721 ipha->ipha_src = INADDR_ANY;
1722 1722 ixas.ixa_flags |= IXAF_SET_SOURCE;
1723 1723 }
1724 1724 /* Should we send with DF and use dce_pmtu? */
1725 1725 if (ipst->ips_ipv4_icmp_return_pmtu) {
1726 1726 ixas.ixa_flags |= IXAF_PMTU_DISCOVERY;
1727 1727 ipha->ipha_fragment_offset_and_flags |= IPH_DF_HTONS;
1728 1728 }
1729 1729
1730 1730 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutMsgs);
1731 1731
1732 1732 (void) ip_output_simple(mp, &ixas);
1733 1733 ixa_cleanup(&ixas);
1734 1734 }
1735 1735
1736 1736 /*
1737 1737 * Verify the ICMP messages for either for ICMP error or redirect packet.
1738 1738 * The caller should have fully pulled up the message. If it's a redirect
1739 1739 * packet, only basic checks on IP header will be done; otherwise, verify
1740 1740 * the packet by looking at the included ULP header.
1741 1741 *
1742 1742 * Called before icmp_inbound_error_fanout_v4 is called.
1743 1743 */
1744 1744 static boolean_t
1745 1745 icmp_inbound_verify_v4(mblk_t *mp, icmph_t *icmph, ip_recv_attr_t *ira)
1746 1746 {
1747 1747 ill_t *ill = ira->ira_ill;
1748 1748 int hdr_length;
1749 1749 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
1750 1750 conn_t *connp;
1751 1751 ipha_t *ipha; /* Inner IP header */
1752 1752
1753 1753 ipha = (ipha_t *)&icmph[1];
1754 1754 if ((uchar_t *)ipha + IP_SIMPLE_HDR_LENGTH > mp->b_wptr)
1755 1755 goto truncated;
1756 1756
1757 1757 hdr_length = IPH_HDR_LENGTH(ipha);
1758 1758
1759 1759 if ((IPH_HDR_VERSION(ipha) != IPV4_VERSION))
1760 1760 goto discard_pkt;
1761 1761
1762 1762 if (hdr_length < sizeof (ipha_t))
1763 1763 goto truncated;
1764 1764
1765 1765 if ((uchar_t *)ipha + hdr_length > mp->b_wptr)
1766 1766 goto truncated;
1767 1767
1768 1768 /*
1769 1769 * Stop here for ICMP_REDIRECT.
1770 1770 */
1771 1771 if (icmph->icmph_type == ICMP_REDIRECT)
1772 1772 return (B_TRUE);
1773 1773
1774 1774 /*
1775 1775 * ICMP errors only.
1776 1776 */
1777 1777 switch (ipha->ipha_protocol) {
1778 1778 case IPPROTO_UDP:
1779 1779 /*
1780 1780 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1781 1781 * transport header.
1782 1782 */
1783 1783 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1784 1784 mp->b_wptr)
1785 1785 goto truncated;
1786 1786 break;
1787 1787 case IPPROTO_TCP: {
1788 1788 tcpha_t *tcpha;
1789 1789
1790 1790 /*
1791 1791 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1792 1792 * transport header.
1793 1793 */
1794 1794 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1795 1795 mp->b_wptr)
1796 1796 goto truncated;
1797 1797
1798 1798 tcpha = (tcpha_t *)((uchar_t *)ipha + hdr_length);
1799 1799 connp = ipcl_tcp_lookup_reversed_ipv4(ipha, tcpha, TCPS_LISTEN,
1800 1800 ipst);
1801 1801 if (connp == NULL)
1802 1802 goto discard_pkt;
1803 1803
1804 1804 if ((connp->conn_verifyicmp != NULL) &&
1805 1805 !connp->conn_verifyicmp(connp, tcpha, icmph, NULL, ira)) {
1806 1806 CONN_DEC_REF(connp);
1807 1807 goto discard_pkt;
1808 1808 }
1809 1809 CONN_DEC_REF(connp);
1810 1810 break;
1811 1811 }
1812 1812 case IPPROTO_SCTP:
1813 1813 /*
1814 1814 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1815 1815 * transport header.
1816 1816 */
1817 1817 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1818 1818 mp->b_wptr)
1819 1819 goto truncated;
1820 1820 break;
1821 1821 case IPPROTO_ESP:
1822 1822 case IPPROTO_AH:
1823 1823 break;
1824 1824 case IPPROTO_ENCAP:
1825 1825 if ((uchar_t *)ipha + hdr_length + sizeof (ipha_t) >
1826 1826 mp->b_wptr)
1827 1827 goto truncated;
1828 1828 break;
1829 1829 default:
1830 1830 break;
1831 1831 }
1832 1832
1833 1833 return (B_TRUE);
1834 1834
1835 1835 discard_pkt:
1836 1836 /* Bogus ICMP error. */
1837 1837 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1838 1838 return (B_FALSE);
1839 1839
1840 1840 truncated:
1841 1841 /* We pulled up everthing already. Must be truncated */
1842 1842 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
1843 1843 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
1844 1844 return (B_FALSE);
1845 1845 }
1846 1846
1847 1847 /* Table from RFC 1191 */
1848 1848 static int icmp_frag_size_table[] =
1849 1849 { 32000, 17914, 8166, 4352, 2002, 1496, 1006, 508, 296, 68 };
1850 1850
1851 1851 /*
1852 1852 * Process received ICMP Packet too big.
1853 1853 * Just handles the DCE create/update, including using the above table of
1854 1854 * PMTU guesses. The caller is responsible for validating the packet before
1855 1855 * passing it in and also to fanout the ICMP error to any matching transport
1856 1856 * conns. Assumes the message has been fully pulled up and verified.
1857 1857 *
1858 1858 * Before getting here, the caller has called icmp_inbound_verify_v4()
1859 1859 * that should have verified with ULP to prevent undoing the changes we're
1860 1860 * going to make to DCE. For example, TCP might have verified that the packet
1861 1861 * which generated error is in the send window.
1862 1862 *
1863 1863 * In some cases modified this MTU in the ICMP header packet; the caller
1864 1864 * should pass to the matching ULP after this returns.
1865 1865 */
1866 1866 static void
1867 1867 icmp_inbound_too_big_v4(icmph_t *icmph, ip_recv_attr_t *ira)
1868 1868 {
1869 1869 dce_t *dce;
1870 1870 int old_mtu;
1871 1871 int mtu, orig_mtu;
1872 1872 ipaddr_t dst;
1873 1873 boolean_t disable_pmtud;
1874 1874 ill_t *ill = ira->ira_ill;
1875 1875 ip_stack_t *ipst = ill->ill_ipst;
1876 1876 uint_t hdr_length;
1877 1877 ipha_t *ipha;
1878 1878
1879 1879 /* Caller already pulled up everything. */
1880 1880 ipha = (ipha_t *)&icmph[1];
1881 1881 ASSERT(icmph->icmph_type == ICMP_DEST_UNREACHABLE &&
1882 1882 icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED);
1883 1883 ASSERT(ill != NULL);
1884 1884
1885 1885 hdr_length = IPH_HDR_LENGTH(ipha);
1886 1886
1887 1887 /*
1888 1888 * We handle path MTU for source routed packets since the DCE
1889 1889 * is looked up using the final destination.
1890 1890 */
1891 1891 dst = ip_get_dst(ipha);
1892 1892
1893 1893 dce = dce_lookup_and_add_v4(dst, ipst);
1894 1894 if (dce == NULL) {
1895 1895 /* Couldn't add a unique one - ENOMEM */
1896 1896 ip1dbg(("icmp_inbound_too_big_v4: no dce for 0x%x\n",
1897 1897 ntohl(dst)));
1898 1898 return;
1899 1899 }
1900 1900
1901 1901 /* Check for MTU discovery advice as described in RFC 1191 */
1902 1902 mtu = ntohs(icmph->icmph_du_mtu);
1903 1903 orig_mtu = mtu;
1904 1904 disable_pmtud = B_FALSE;
1905 1905
1906 1906 mutex_enter(&dce->dce_lock);
1907 1907 if (dce->dce_flags & DCEF_PMTU)
1908 1908 old_mtu = dce->dce_pmtu;
1909 1909 else
1910 1910 old_mtu = ill->ill_mtu;
1911 1911
1912 1912 if (icmph->icmph_du_zero != 0 || mtu < ipst->ips_ip_pmtu_min) {
1913 1913 uint32_t length;
1914 1914 int i;
1915 1915
1916 1916 /*
1917 1917 * Use the table from RFC 1191 to figure out
1918 1918 * the next "plateau" based on the length in
1919 1919 * the original IP packet.
1920 1920 */
1921 1921 length = ntohs(ipha->ipha_length);
1922 1922 DTRACE_PROBE2(ip4__pmtu__guess, dce_t *, dce,
1923 1923 uint32_t, length);
1924 1924 if (old_mtu <= length &&
1925 1925 old_mtu >= length - hdr_length) {
1926 1926 /*
1927 1927 * Handle broken BSD 4.2 systems that
1928 1928 * return the wrong ipha_length in ICMP
1929 1929 * errors.
1930 1930 */
1931 1931 ip1dbg(("Wrong mtu: sent %d, dce %d\n",
1932 1932 length, old_mtu));
1933 1933 length -= hdr_length;
1934 1934 }
1935 1935 for (i = 0; i < A_CNT(icmp_frag_size_table); i++) {
1936 1936 if (length > icmp_frag_size_table[i])
1937 1937 break;
1938 1938 }
1939 1939 if (i == A_CNT(icmp_frag_size_table)) {
1940 1940 /* Smaller than IP_MIN_MTU! */
1941 1941 ip1dbg(("Too big for packet size %d\n",
1942 1942 length));
1943 1943 disable_pmtud = B_TRUE;
1944 1944 mtu = ipst->ips_ip_pmtu_min;
1945 1945 } else {
1946 1946 mtu = icmp_frag_size_table[i];
1947 1947 ip1dbg(("Calculated mtu %d, packet size %d, "
1948 1948 "before %d\n", mtu, length, old_mtu));
1949 1949 if (mtu < ipst->ips_ip_pmtu_min) {
1950 1950 mtu = ipst->ips_ip_pmtu_min;
1951 1951 disable_pmtud = B_TRUE;
1952 1952 }
1953 1953 }
1954 1954 }
1955 1955 if (disable_pmtud)
1956 1956 dce->dce_flags |= DCEF_TOO_SMALL_PMTU;
1957 1957 else
1958 1958 dce->dce_flags &= ~DCEF_TOO_SMALL_PMTU;
1959 1959
1960 1960 dce->dce_pmtu = MIN(old_mtu, mtu);
1961 1961 /* Prepare to send the new max frag size for the ULP. */
1962 1962 icmph->icmph_du_zero = 0;
1963 1963 icmph->icmph_du_mtu = htons((uint16_t)dce->dce_pmtu);
1964 1964 DTRACE_PROBE4(ip4__pmtu__change, icmph_t *, icmph, dce_t *,
1965 1965 dce, int, orig_mtu, int, mtu);
1966 1966
1967 1967 /* We now have a PMTU for sure */
1968 1968 dce->dce_flags |= DCEF_PMTU;
1969 1969 dce->dce_last_change_time = TICK_TO_SEC(ddi_get_lbolt64());
1970 1970 mutex_exit(&dce->dce_lock);
1971 1971 /*
1972 1972 * After dropping the lock the new value is visible to everyone.
1973 1973 * Then we bump the generation number so any cached values reinspect
1974 1974 * the dce_t.
1975 1975 */
1976 1976 dce_increment_generation(dce);
1977 1977 dce_refrele(dce);
1978 1978 }
1979 1979
1980 1980 /*
1981 1981 * If the packet in error is Self-Encapsulated, icmp_inbound_error_fanout_v4
1982 1982 * calls this function.
1983 1983 */
1984 1984 static mblk_t *
1985 1985 icmp_inbound_self_encap_error_v4(mblk_t *mp, ipha_t *ipha, ipha_t *in_ipha)
1986 1986 {
1987 1987 int length;
1988 1988
1989 1989 ASSERT(mp->b_datap->db_type == M_DATA);
1990 1990
1991 1991 /* icmp_inbound_v4 has already pulled up the whole error packet */
1992 1992 ASSERT(mp->b_cont == NULL);
1993 1993
1994 1994 /*
1995 1995 * The length that we want to overlay is the inner header
1996 1996 * and what follows it.
1997 1997 */
1998 1998 length = msgdsize(mp) - ((uchar_t *)in_ipha - mp->b_rptr);
1999 1999
2000 2000 /*
2001 2001 * Overlay the inner header and whatever follows it over the
2002 2002 * outer header.
2003 2003 */
2004 2004 bcopy((uchar_t *)in_ipha, (uchar_t *)ipha, length);
2005 2005
2006 2006 /* Adjust for what we removed */
2007 2007 mp->b_wptr -= (uchar_t *)in_ipha - (uchar_t *)ipha;
2008 2008 return (mp);
2009 2009 }
2010 2010
2011 2011 /*
2012 2012 * Try to pass the ICMP message upstream in case the ULP cares.
2013 2013 *
2014 2014 * If the packet that caused the ICMP error is secure, we send
2015 2015 * it to AH/ESP to make sure that the attached packet has a
2016 2016 * valid association. ipha in the code below points to the
2017 2017 * IP header of the packet that caused the error.
2018 2018 *
2019 2019 * For IPsec cases, we let the next-layer-up (which has access to
2020 2020 * cached policy on the conn_t, or can query the SPD directly)
2021 2021 * subtract out any IPsec overhead if they must. We therefore make no
2022 2022 * adjustments here for IPsec overhead.
2023 2023 *
2024 2024 * IFN could have been generated locally or by some router.
2025 2025 *
2026 2026 * LOCAL : ire_send_wire (before calling ipsec_out_process) can call
2027 2027 * icmp_frag_needed/icmp_pkt2big_v6 to generated a local IFN.
2028 2028 * This happens because IP adjusted its value of MTU on an
2029 2029 * earlier IFN message and could not tell the upper layer,
2030 2030 * the new adjusted value of MTU e.g. Packet was encrypted
2031 2031 * or there was not enough information to fanout to upper
2032 2032 * layers. Thus on the next outbound datagram, ire_send_wire
2033 2033 * generates the IFN, where IPsec processing has *not* been
2034 2034 * done.
2035 2035 *
2036 2036 * Note that we retain ixa_fragsize across IPsec thus once
2037 2037 * we have picking ixa_fragsize and entered ipsec_out_process we do
2038 2038 * no change the fragsize even if the path MTU changes before
2039 2039 * we reach ip_output_post_ipsec.
2040 2040 *
2041 2041 * In the local case, IRAF_LOOPBACK will be set indicating
2042 2042 * that IFN was generated locally.
2043 2043 *
2044 2044 * ROUTER : IFN could be secure or non-secure.
2045 2045 *
2046 2046 * * SECURE : We use the IPSEC_IN to fanout to AH/ESP if the
2047 2047 * packet in error has AH/ESP headers to validate the AH/ESP
2048 2048 * headers. AH/ESP will verify whether there is a valid SA or
2049 2049 * not and send it back. We will fanout again if we have more
2050 2050 * data in the packet.
2051 2051 *
2052 2052 * If the packet in error does not have AH/ESP, we handle it
2053 2053 * like any other case.
2054 2054 *
2055 2055 * * NON_SECURE : If the packet in error has AH/ESP headers, we send it
2056 2056 * up to AH/ESP for validation. AH/ESP will verify whether there is a
2057 2057 * valid SA or not and send it back. We will fanout again if
2058 2058 * we have more data in the packet.
2059 2059 *
2060 2060 * If the packet in error does not have AH/ESP, we handle it
2061 2061 * like any other case.
2062 2062 *
2063 2063 * The caller must have called icmp_inbound_verify_v4.
2064 2064 */
2065 2065 static void
2066 2066 icmp_inbound_error_fanout_v4(mblk_t *mp, icmph_t *icmph, ip_recv_attr_t *ira)
2067 2067 {
2068 2068 uint16_t *up; /* Pointer to ports in ULP header */
2069 2069 uint32_t ports; /* reversed ports for fanout */
2070 2070 ipha_t ripha; /* With reversed addresses */
2071 2071 ipha_t *ipha; /* Inner IP header */
2072 2072 uint_t hdr_length; /* Inner IP header length */
2073 2073 tcpha_t *tcpha;
2074 2074 conn_t *connp;
2075 2075 ill_t *ill = ira->ira_ill;
2076 2076 ip_stack_t *ipst = ill->ill_ipst;
2077 2077 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
2078 2078 ill_t *rill = ira->ira_rill;
2079 2079
2080 2080 /* Caller already pulled up everything. */
2081 2081 ipha = (ipha_t *)&icmph[1];
2082 2082 ASSERT((uchar_t *)&ipha[1] <= mp->b_wptr);
2083 2083 ASSERT(mp->b_cont == NULL);
2084 2084
2085 2085 hdr_length = IPH_HDR_LENGTH(ipha);
2086 2086 ira->ira_protocol = ipha->ipha_protocol;
2087 2087
2088 2088 /*
2089 2089 * We need a separate IP header with the source and destination
2090 2090 * addresses reversed to do fanout/classification because the ipha in
2091 2091 * the ICMP error is in the form we sent it out.
2092 2092 */
2093 2093 ripha.ipha_src = ipha->ipha_dst;
2094 2094 ripha.ipha_dst = ipha->ipha_src;
2095 2095 ripha.ipha_protocol = ipha->ipha_protocol;
2096 2096 ripha.ipha_version_and_hdr_length = ipha->ipha_version_and_hdr_length;
2097 2097
2098 2098 ip2dbg(("icmp_inbound_error_v4: proto %d %x to %x: %d/%d\n",
2099 2099 ripha.ipha_protocol, ntohl(ipha->ipha_src),
2100 2100 ntohl(ipha->ipha_dst),
2101 2101 icmph->icmph_type, icmph->icmph_code));
2102 2102
2103 2103 switch (ipha->ipha_protocol) {
2104 2104 case IPPROTO_UDP:
2105 2105 up = (uint16_t *)((uchar_t *)ipha + hdr_length);
2106 2106
2107 2107 /* Attempt to find a client stream based on port. */
2108 2108 ip2dbg(("icmp_inbound_error_v4: UDP ports %d to %d\n",
2109 2109 ntohs(up[0]), ntohs(up[1])));
2110 2110
2111 2111 /* Note that we send error to all matches. */
2112 2112 ira->ira_flags |= IRAF_ICMP_ERROR;
2113 2113 ip_fanout_udp_multi_v4(mp, &ripha, up[0], up[1], ira);
2114 2114 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2115 2115 return;
2116 2116
2117 2117 case IPPROTO_TCP:
2118 2118 /*
2119 2119 * Find a TCP client stream for this packet.
2120 2120 * Note that we do a reverse lookup since the header is
2121 2121 * in the form we sent it out.
2122 2122 */
2123 2123 tcpha = (tcpha_t *)((uchar_t *)ipha + hdr_length);
2124 2124 connp = ipcl_tcp_lookup_reversed_ipv4(ipha, tcpha, TCPS_LISTEN,
2125 2125 ipst);
2126 2126 if (connp == NULL)
2127 2127 goto discard_pkt;
2128 2128
2129 2129 if (CONN_INBOUND_POLICY_PRESENT(connp, ipss) ||
2130 2130 (ira->ira_flags & IRAF_IPSEC_SECURE)) {
2131 2131 mp = ipsec_check_inbound_policy(mp, connp,
2132 2132 ipha, NULL, ira);
2133 2133 if (mp == NULL) {
2134 2134 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2135 2135 /* Note that mp is NULL */
2136 2136 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2137 2137 CONN_DEC_REF(connp);
2138 2138 return;
2139 2139 }
2140 2140 }
2141 2141
2142 2142 ira->ira_flags |= IRAF_ICMP_ERROR;
2143 2143 ira->ira_ill = ira->ira_rill = NULL;
2144 2144 if (IPCL_IS_TCP(connp)) {
2145 2145 SQUEUE_ENTER_ONE(connp->conn_sqp, mp,
2146 2146 connp->conn_recvicmp, connp, ira, SQ_FILL,
2147 2147 SQTAG_TCP_INPUT_ICMP_ERR);
2148 2148 } else {
2149 2149 /* Not TCP; must be SOCK_RAW, IPPROTO_TCP */
2150 2150 (connp->conn_recv)(connp, mp, NULL, ira);
2151 2151 CONN_DEC_REF(connp);
2152 2152 }
2153 2153 ira->ira_ill = ill;
2154 2154 ira->ira_rill = rill;
2155 2155 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2156 2156 return;
2157 2157
2158 2158 case IPPROTO_SCTP:
2159 2159 up = (uint16_t *)((uchar_t *)ipha + hdr_length);
2160 2160 /* Find a SCTP client stream for this packet. */
2161 2161 ((uint16_t *)&ports)[0] = up[1];
2162 2162 ((uint16_t *)&ports)[1] = up[0];
2163 2163
2164 2164 ira->ira_flags |= IRAF_ICMP_ERROR;
2165 2165 ip_fanout_sctp(mp, &ripha, NULL, ports, ira);
2166 2166 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2167 2167 return;
2168 2168
2169 2169 case IPPROTO_ESP:
2170 2170 case IPPROTO_AH:
2171 2171 if (!ipsec_loaded(ipss)) {
2172 2172 ip_proto_not_sup(mp, ira);
2173 2173 return;
2174 2174 }
2175 2175
2176 2176 if (ipha->ipha_protocol == IPPROTO_ESP)
2177 2177 mp = ipsecesp_icmp_error(mp, ira);
2178 2178 else
2179 2179 mp = ipsecah_icmp_error(mp, ira);
2180 2180 if (mp == NULL)
2181 2181 return;
2182 2182
2183 2183 /* Just in case ipsec didn't preserve the NULL b_cont */
2184 2184 if (mp->b_cont != NULL) {
2185 2185 if (!pullupmsg(mp, -1))
2186 2186 goto discard_pkt;
2187 2187 }
2188 2188
2189 2189 /*
2190 2190 * Note that ira_pktlen and ira_ip_hdr_length are no longer
2191 2191 * correct, but we don't use them any more here.
2192 2192 *
2193 2193 * If succesful, the mp has been modified to not include
2194 2194 * the ESP/AH header so we can fanout to the ULP's icmp
2195 2195 * error handler.
2196 2196 */
2197 2197 if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH)
2198 2198 goto truncated;
2199 2199
2200 2200 /* Verify the modified message before any further processes. */
2201 2201 ipha = (ipha_t *)mp->b_rptr;
2202 2202 hdr_length = IPH_HDR_LENGTH(ipha);
2203 2203 icmph = (icmph_t *)&mp->b_rptr[hdr_length];
2204 2204 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
2205 2205 freemsg(mp);
2206 2206 return;
2207 2207 }
2208 2208
2209 2209 icmp_inbound_error_fanout_v4(mp, icmph, ira);
2210 2210 return;
2211 2211
2212 2212 case IPPROTO_ENCAP: {
2213 2213 /* Look for self-encapsulated packets that caused an error */
2214 2214 ipha_t *in_ipha;
2215 2215
2216 2216 /*
2217 2217 * Caller has verified that length has to be
2218 2218 * at least the size of IP header.
2219 2219 */
2220 2220 ASSERT(hdr_length >= sizeof (ipha_t));
2221 2221 /*
2222 2222 * Check the sanity of the inner IP header like
2223 2223 * we did for the outer header.
2224 2224 */
2225 2225 in_ipha = (ipha_t *)((uchar_t *)ipha + hdr_length);
2226 2226 if ((IPH_HDR_VERSION(in_ipha) != IPV4_VERSION)) {
2227 2227 goto discard_pkt;
2228 2228 }
2229 2229 if (IPH_HDR_LENGTH(in_ipha) < sizeof (ipha_t)) {
2230 2230 goto discard_pkt;
2231 2231 }
2232 2232 /* Check for Self-encapsulated tunnels */
2233 2233 if (in_ipha->ipha_src == ipha->ipha_src &&
2234 2234 in_ipha->ipha_dst == ipha->ipha_dst) {
2235 2235
2236 2236 mp = icmp_inbound_self_encap_error_v4(mp, ipha,
2237 2237 in_ipha);
2238 2238 if (mp == NULL)
2239 2239 goto discard_pkt;
2240 2240
2241 2241 /*
2242 2242 * Just in case self_encap didn't preserve the NULL
2243 2243 * b_cont
2244 2244 */
2245 2245 if (mp->b_cont != NULL) {
2246 2246 if (!pullupmsg(mp, -1))
2247 2247 goto discard_pkt;
2248 2248 }
2249 2249 /*
2250 2250 * Note that ira_pktlen and ira_ip_hdr_length are no
2251 2251 * longer correct, but we don't use them any more here.
2252 2252 */
2253 2253 if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH)
2254 2254 goto truncated;
2255 2255
2256 2256 /*
2257 2257 * Verify the modified message before any further
2258 2258 * processes.
2259 2259 */
2260 2260 ipha = (ipha_t *)mp->b_rptr;
2261 2261 hdr_length = IPH_HDR_LENGTH(ipha);
2262 2262 icmph = (icmph_t *)&mp->b_rptr[hdr_length];
2263 2263 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
2264 2264 freemsg(mp);
2265 2265 return;
2266 2266 }
2267 2267
2268 2268 /*
2269 2269 * The packet in error is self-encapsualted.
2270 2270 * And we are finding it further encapsulated
2271 2271 * which we could not have possibly generated.
2272 2272 */
2273 2273 if (ipha->ipha_protocol == IPPROTO_ENCAP) {
2274 2274 goto discard_pkt;
2275 2275 }
2276 2276 icmp_inbound_error_fanout_v4(mp, icmph, ira);
2277 2277 return;
2278 2278 }
2279 2279 /* No self-encapsulated */
2280 2280 }
2281 2281 /* FALLTHROUGH */
2282 2282 case IPPROTO_IPV6:
2283 2283 if ((connp = ipcl_iptun_classify_v4(&ripha.ipha_src,
2284 2284 &ripha.ipha_dst, ipst)) != NULL) {
2285 2285 ira->ira_flags |= IRAF_ICMP_ERROR;
2286 2286 connp->conn_recvicmp(connp, mp, NULL, ira);
2287 2287 CONN_DEC_REF(connp);
2288 2288 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2289 2289 return;
2290 2290 }
2291 2291 /*
2292 2292 * No IP tunnel is interested, fallthrough and see
2293 2293 * if a raw socket will want it.
2294 2294 */
2295 2295 /* FALLTHROUGH */
2296 2296 default:
2297 2297 ira->ira_flags |= IRAF_ICMP_ERROR;
2298 2298 ip_fanout_proto_v4(mp, &ripha, ira);
2299 2299 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2300 2300 return;
2301 2301 }
2302 2302 /* NOTREACHED */
2303 2303 discard_pkt:
2304 2304 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2305 2305 ip1dbg(("icmp_inbound_error_fanout_v4: drop pkt\n"));
2306 2306 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2307 2307 freemsg(mp);
2308 2308 return;
2309 2309
2310 2310 truncated:
2311 2311 /* We pulled up everthing already. Must be truncated */
2312 2312 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
2313 2313 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
2314 2314 freemsg(mp);
2315 2315 }
2316 2316
2317 2317 /*
2318 2318 * Common IP options parser.
2319 2319 *
2320 2320 * Setup routine: fill in *optp with options-parsing state, then
2321 2321 * tail-call ipoptp_next to return the first option.
2322 2322 */
2323 2323 uint8_t
2324 2324 ipoptp_first(ipoptp_t *optp, ipha_t *ipha)
2325 2325 {
2326 2326 uint32_t totallen; /* total length of all options */
2327 2327
2328 2328 totallen = ipha->ipha_version_and_hdr_length -
2329 2329 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
2330 2330 totallen <<= 2;
2331 2331 optp->ipoptp_next = (uint8_t *)(&ipha[1]);
2332 2332 optp->ipoptp_end = optp->ipoptp_next + totallen;
2333 2333 optp->ipoptp_flags = 0;
2334 2334 return (ipoptp_next(optp));
2335 2335 }
2336 2336
2337 2337 /* Like above but without an ipha_t */
2338 2338 uint8_t
2339 2339 ipoptp_first2(ipoptp_t *optp, uint32_t totallen, uint8_t *opt)
2340 2340 {
2341 2341 optp->ipoptp_next = opt;
2342 2342 optp->ipoptp_end = optp->ipoptp_next + totallen;
2343 2343 optp->ipoptp_flags = 0;
2344 2344 return (ipoptp_next(optp));
2345 2345 }
2346 2346
2347 2347 /*
2348 2348 * Common IP options parser: extract next option.
2349 2349 */
2350 2350 uint8_t
2351 2351 ipoptp_next(ipoptp_t *optp)
2352 2352 {
2353 2353 uint8_t *end = optp->ipoptp_end;
2354 2354 uint8_t *cur = optp->ipoptp_next;
2355 2355 uint8_t opt, len, pointer;
2356 2356
2357 2357 /*
2358 2358 * If cur > end already, then the ipoptp_end or ipoptp_next pointer
2359 2359 * has been corrupted.
2360 2360 */
2361 2361 ASSERT(cur <= end);
2362 2362
2363 2363 if (cur == end)
2364 2364 return (IPOPT_EOL);
2365 2365
2366 2366 opt = cur[IPOPT_OPTVAL];
2367 2367
2368 2368 /*
2369 2369 * Skip any NOP options.
2370 2370 */
2371 2371 while (opt == IPOPT_NOP) {
2372 2372 cur++;
2373 2373 if (cur == end)
2374 2374 return (IPOPT_EOL);
2375 2375 opt = cur[IPOPT_OPTVAL];
2376 2376 }
2377 2377
2378 2378 if (opt == IPOPT_EOL)
2379 2379 return (IPOPT_EOL);
2380 2380
2381 2381 /*
2382 2382 * Option requiring a length.
2383 2383 */
2384 2384 if ((cur + 1) >= end) {
2385 2385 optp->ipoptp_flags |= IPOPTP_ERROR;
2386 2386 return (IPOPT_EOL);
2387 2387 }
2388 2388 len = cur[IPOPT_OLEN];
2389 2389 if (len < 2) {
2390 2390 optp->ipoptp_flags |= IPOPTP_ERROR;
2391 2391 return (IPOPT_EOL);
2392 2392 }
2393 2393 optp->ipoptp_cur = cur;
2394 2394 optp->ipoptp_len = len;
2395 2395 optp->ipoptp_next = cur + len;
2396 2396 if (cur + len > end) {
2397 2397 optp->ipoptp_flags |= IPOPTP_ERROR;
2398 2398 return (IPOPT_EOL);
2399 2399 }
2400 2400
2401 2401 /*
2402 2402 * For the options which require a pointer field, make sure
2403 2403 * its there, and make sure it points to either something
2404 2404 * inside this option, or the end of the option.
2405 2405 */
2406 2406 switch (opt) {
2407 2407 case IPOPT_RR:
2408 2408 case IPOPT_TS:
2409 2409 case IPOPT_LSRR:
2410 2410 case IPOPT_SSRR:
2411 2411 if (len <= IPOPT_OFFSET) {
2412 2412 optp->ipoptp_flags |= IPOPTP_ERROR;
2413 2413 return (opt);
2414 2414 }
2415 2415 pointer = cur[IPOPT_OFFSET];
2416 2416 if (pointer - 1 > len) {
2417 2417 optp->ipoptp_flags |= IPOPTP_ERROR;
2418 2418 return (opt);
2419 2419 }
2420 2420 break;
2421 2421 }
2422 2422
2423 2423 /*
2424 2424 * Sanity check the pointer field based on the type of the
2425 2425 * option.
2426 2426 */
2427 2427 switch (opt) {
2428 2428 case IPOPT_RR:
2429 2429 case IPOPT_SSRR:
2430 2430 case IPOPT_LSRR:
2431 2431 if (pointer < IPOPT_MINOFF_SR)
2432 2432 optp->ipoptp_flags |= IPOPTP_ERROR;
2433 2433 break;
2434 2434 case IPOPT_TS:
2435 2435 if (pointer < IPOPT_MINOFF_IT)
2436 2436 optp->ipoptp_flags |= IPOPTP_ERROR;
2437 2437 /*
2438 2438 * Note that the Internet Timestamp option also
2439 2439 * contains two four bit fields (the Overflow field,
2440 2440 * and the Flag field), which follow the pointer
2441 2441 * field. We don't need to check that these fields
2442 2442 * fall within the length of the option because this
2443 2443 * was implicitely done above. We've checked that the
2444 2444 * pointer value is at least IPOPT_MINOFF_IT, and that
2445 2445 * it falls within the option. Since IPOPT_MINOFF_IT >
2446 2446 * IPOPT_POS_OV_FLG, we don't need the explicit check.
2447 2447 */
2448 2448 ASSERT(len > IPOPT_POS_OV_FLG);
2449 2449 break;
2450 2450 }
2451 2451
2452 2452 return (opt);
2453 2453 }
2454 2454
2455 2455 /*
2456 2456 * Use the outgoing IP header to create an IP_OPTIONS option the way
2457 2457 * it was passed down from the application.
2458 2458 *
2459 2459 * This is compatible with BSD in that it returns
2460 2460 * the reverse source route with the final destination
2461 2461 * as the last entry. The first 4 bytes of the option
2462 2462 * will contain the final destination.
2463 2463 */
2464 2464 int
2465 2465 ip_opt_get_user(conn_t *connp, uchar_t *buf)
2466 2466 {
2467 2467 ipoptp_t opts;
2468 2468 uchar_t *opt;
2469 2469 uint8_t optval;
2470 2470 uint8_t optlen;
2471 2471 uint32_t len = 0;
2472 2472 uchar_t *buf1 = buf;
2473 2473 uint32_t totallen;
2474 2474 ipaddr_t dst;
2475 2475 ip_pkt_t *ipp = &connp->conn_xmit_ipp;
2476 2476
2477 2477 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
2478 2478 return (0);
2479 2479
2480 2480 totallen = ipp->ipp_ipv4_options_len;
2481 2481 if (totallen & 0x3)
2482 2482 return (0);
2483 2483
2484 2484 buf += IP_ADDR_LEN; /* Leave room for final destination */
2485 2485 len += IP_ADDR_LEN;
2486 2486 bzero(buf1, IP_ADDR_LEN);
2487 2487
2488 2488 dst = connp->conn_faddr_v4;
2489 2489
2490 2490 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
2491 2491 optval != IPOPT_EOL;
2492 2492 optval = ipoptp_next(&opts)) {
2493 2493 int off;
2494 2494
2495 2495 opt = opts.ipoptp_cur;
2496 2496 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
2497 2497 break;
2498 2498 }
2499 2499 optlen = opts.ipoptp_len;
2500 2500
2501 2501 switch (optval) {
2502 2502 case IPOPT_SSRR:
2503 2503 case IPOPT_LSRR:
2504 2504
2505 2505 /*
2506 2506 * Insert destination as the first entry in the source
2507 2507 * route and move down the entries on step.
2508 2508 * The last entry gets placed at buf1.
2509 2509 */
2510 2510 buf[IPOPT_OPTVAL] = optval;
2511 2511 buf[IPOPT_OLEN] = optlen;
2512 2512 buf[IPOPT_OFFSET] = optlen;
2513 2513
2514 2514 off = optlen - IP_ADDR_LEN;
2515 2515 if (off < 0) {
2516 2516 /* No entries in source route */
2517 2517 break;
2518 2518 }
2519 2519 /* Last entry in source route if not already set */
2520 2520 if (dst == INADDR_ANY)
2521 2521 bcopy(opt + off, buf1, IP_ADDR_LEN);
2522 2522 off -= IP_ADDR_LEN;
2523 2523
2524 2524 while (off > 0) {
2525 2525 bcopy(opt + off,
2526 2526 buf + off + IP_ADDR_LEN,
2527 2527 IP_ADDR_LEN);
2528 2528 off -= IP_ADDR_LEN;
2529 2529 }
2530 2530 /* ipha_dst into first slot */
2531 2531 bcopy(&dst, buf + off + IP_ADDR_LEN,
2532 2532 IP_ADDR_LEN);
2533 2533 buf += optlen;
2534 2534 len += optlen;
2535 2535 break;
2536 2536
2537 2537 default:
2538 2538 bcopy(opt, buf, optlen);
2539 2539 buf += optlen;
2540 2540 len += optlen;
2541 2541 break;
2542 2542 }
2543 2543 }
2544 2544 done:
2545 2545 /* Pad the resulting options */
2546 2546 while (len & 0x3) {
2547 2547 *buf++ = IPOPT_EOL;
2548 2548 len++;
2549 2549 }
2550 2550 return (len);
2551 2551 }
2552 2552
2553 2553 /*
2554 2554 * Update any record route or timestamp options to include this host.
2555 2555 * Reverse any source route option.
2556 2556 * This routine assumes that the options are well formed i.e. that they
2557 2557 * have already been checked.
2558 2558 */
2559 2559 static void
2560 2560 icmp_options_update(ipha_t *ipha)
2561 2561 {
2562 2562 ipoptp_t opts;
2563 2563 uchar_t *opt;
2564 2564 uint8_t optval;
2565 2565 ipaddr_t src; /* Our local address */
2566 2566 ipaddr_t dst;
2567 2567
2568 2568 ip2dbg(("icmp_options_update\n"));
2569 2569 src = ipha->ipha_src;
2570 2570 dst = ipha->ipha_dst;
2571 2571
2572 2572 for (optval = ipoptp_first(&opts, ipha);
2573 2573 optval != IPOPT_EOL;
2574 2574 optval = ipoptp_next(&opts)) {
2575 2575 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
2576 2576 opt = opts.ipoptp_cur;
2577 2577 ip2dbg(("icmp_options_update: opt %d, len %d\n",
2578 2578 optval, opts.ipoptp_len));
2579 2579 switch (optval) {
2580 2580 int off1, off2;
2581 2581 case IPOPT_SSRR:
2582 2582 case IPOPT_LSRR:
2583 2583 /*
2584 2584 * Reverse the source route. The first entry
2585 2585 * should be the next to last one in the current
2586 2586 * source route (the last entry is our address).
2587 2587 * The last entry should be the final destination.
2588 2588 */
2589 2589 off1 = IPOPT_MINOFF_SR - 1;
2590 2590 off2 = opt[IPOPT_OFFSET] - IP_ADDR_LEN - 1;
2591 2591 if (off2 < 0) {
2592 2592 /* No entries in source route */
2593 2593 ip1dbg((
2594 2594 "icmp_options_update: bad src route\n"));
2595 2595 break;
2596 2596 }
2597 2597 bcopy((char *)opt + off2, &dst, IP_ADDR_LEN);
2598 2598 bcopy(&ipha->ipha_dst, (char *)opt + off2, IP_ADDR_LEN);
2599 2599 bcopy(&dst, &ipha->ipha_dst, IP_ADDR_LEN);
2600 2600 off2 -= IP_ADDR_LEN;
2601 2601
2602 2602 while (off1 < off2) {
2603 2603 bcopy((char *)opt + off1, &src, IP_ADDR_LEN);
2604 2604 bcopy((char *)opt + off2, (char *)opt + off1,
2605 2605 IP_ADDR_LEN);
2606 2606 bcopy(&src, (char *)opt + off2, IP_ADDR_LEN);
2607 2607 off1 += IP_ADDR_LEN;
2608 2608 off2 -= IP_ADDR_LEN;
2609 2609 }
2610 2610 opt[IPOPT_OFFSET] = IPOPT_MINOFF_SR;
2611 2611 break;
2612 2612 }
2613 2613 }
2614 2614 }
2615 2615
2616 2616 /*
2617 2617 * Process received ICMP Redirect messages.
2618 2618 * Assumes the caller has verified that the headers are in the pulled up mblk.
2619 2619 * Consumes mp.
2620 2620 */
2621 2621 static void
2622 2622 icmp_redirect_v4(mblk_t *mp, ipha_t *ipha, icmph_t *icmph, ip_recv_attr_t *ira)
2623 2623 {
2624 2624 ire_t *ire, *nire;
2625 2625 ire_t *prev_ire;
2626 2626 ipaddr_t src, dst, gateway;
2627 2627 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2628 2628 ipha_t *inner_ipha; /* Inner IP header */
2629 2629
2630 2630 /* Caller already pulled up everything. */
2631 2631 inner_ipha = (ipha_t *)&icmph[1];
2632 2632 src = ipha->ipha_src;
2633 2633 dst = inner_ipha->ipha_dst;
2634 2634 gateway = icmph->icmph_rd_gateway;
2635 2635 /* Make sure the new gateway is reachable somehow. */
2636 2636 ire = ire_ftable_lookup_v4(gateway, 0, 0, IRE_ONLINK, NULL,
2637 2637 ALL_ZONES, NULL, MATCH_IRE_TYPE, 0, ipst, NULL);
2638 2638 /*
2639 2639 * Make sure we had a route for the dest in question and that
2640 2640 * that route was pointing to the old gateway (the source of the
2641 2641 * redirect packet.)
2642 2642 * We do longest match and then compare ire_gateway_addr below.
2643 2643 */
2644 2644 prev_ire = ire_ftable_lookup_v4(dst, 0, 0, 0, NULL, ALL_ZONES,
2645 2645 NULL, MATCH_IRE_DSTONLY, 0, ipst, NULL);
2646 2646 /*
2647 2647 * Check that
2648 2648 * the redirect was not from ourselves
2649 2649 * the new gateway and the old gateway are directly reachable
2650 2650 */
2651 2651 if (prev_ire == NULL || ire == NULL ||
2652 2652 (prev_ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK)) ||
2653 2653 (prev_ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) ||
2654 2654 !(ire->ire_type & IRE_IF_ALL) ||
2655 2655 prev_ire->ire_gateway_addr != src) {
2656 2656 BUMP_MIB(&ipst->ips_icmp_mib, icmpInBadRedirects);
2657 2657 ip_drop_input("icmpInBadRedirects - ire", mp, ira->ira_ill);
2658 2658 freemsg(mp);
2659 2659 if (ire != NULL)
2660 2660 ire_refrele(ire);
2661 2661 if (prev_ire != NULL)
2662 2662 ire_refrele(prev_ire);
2663 2663 return;
2664 2664 }
2665 2665
2666 2666 ire_refrele(prev_ire);
2667 2667 ire_refrele(ire);
2668 2668
2669 2669 /*
2670 2670 * TODO: more precise handling for cases 0, 2, 3, the latter two
2671 2671 * require TOS routing
2672 2672 */
2673 2673 switch (icmph->icmph_code) {
2674 2674 case 0:
2675 2675 case 1:
2676 2676 /* TODO: TOS specificity for cases 2 and 3 */
2677 2677 case 2:
2678 2678 case 3:
2679 2679 break;
2680 2680 default:
2681 2681 BUMP_MIB(&ipst->ips_icmp_mib, icmpInBadRedirects);
2682 2682 ip_drop_input("icmpInBadRedirects - code", mp, ira->ira_ill);
2683 2683 freemsg(mp);
2684 2684 return;
2685 2685 }
2686 2686 /*
2687 2687 * Create a Route Association. This will allow us to remember that
2688 2688 * someone we believe told us to use the particular gateway.
2689 2689 */
2690 2690 ire = ire_create(
2691 2691 (uchar_t *)&dst, /* dest addr */
2692 2692 (uchar_t *)&ip_g_all_ones, /* mask */
2693 2693 (uchar_t *)&gateway, /* gateway addr */
2694 2694 IRE_HOST,
2695 2695 NULL, /* ill */
2696 2696 ALL_ZONES,
2697 2697 (RTF_DYNAMIC | RTF_GATEWAY | RTF_HOST),
2698 2698 NULL, /* tsol_gc_t */
2699 2699 ipst);
2700 2700
2701 2701 if (ire == NULL) {
2702 2702 freemsg(mp);
2703 2703 return;
2704 2704 }
2705 2705 nire = ire_add(ire);
2706 2706 /* Check if it was a duplicate entry */
2707 2707 if (nire != NULL && nire != ire) {
2708 2708 ASSERT(nire->ire_identical_ref > 1);
2709 2709 ire_delete(nire);
2710 2710 ire_refrele(nire);
2711 2711 nire = NULL;
2712 2712 }
2713 2713 ire = nire;
2714 2714 if (ire != NULL) {
2715 2715 ire_refrele(ire); /* Held in ire_add */
2716 2716
2717 2717 /* tell routing sockets that we received a redirect */
2718 2718 ip_rts_change(RTM_REDIRECT, dst, gateway, IP_HOST_MASK, 0, src,
2719 2719 (RTF_DYNAMIC | RTF_GATEWAY | RTF_HOST), 0,
2720 2720 (RTA_DST | RTA_GATEWAY | RTA_NETMASK | RTA_AUTHOR), ipst);
2721 2721 }
2722 2722
2723 2723 /*
2724 2724 * Delete any existing IRE_HOST type redirect ires for this destination.
2725 2725 * This together with the added IRE has the effect of
2726 2726 * modifying an existing redirect.
2727 2727 */
2728 2728 prev_ire = ire_ftable_lookup_v4(dst, 0, src, IRE_HOST, NULL,
2729 2729 ALL_ZONES, NULL, (MATCH_IRE_GW | MATCH_IRE_TYPE), 0, ipst, NULL);
2730 2730 if (prev_ire != NULL) {
2731 2731 if (prev_ire ->ire_flags & RTF_DYNAMIC)
2732 2732 ire_delete(prev_ire);
2733 2733 ire_refrele(prev_ire);
2734 2734 }
2735 2735
2736 2736 freemsg(mp);
2737 2737 }
2738 2738
2739 2739 /*
2740 2740 * Generate an ICMP parameter problem message.
2741 2741 * When called from ip_output side a minimal ip_recv_attr_t needs to be
2742 2742 * constructed by the caller.
2743 2743 */
2744 2744 static void
2745 2745 icmp_param_problem(mblk_t *mp, uint8_t ptr, ip_recv_attr_t *ira)
2746 2746 {
2747 2747 icmph_t icmph;
2748 2748 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2749 2749
2750 2750 mp = icmp_pkt_err_ok(mp, ira);
2751 2751 if (mp == NULL)
2752 2752 return;
2753 2753
2754 2754 bzero(&icmph, sizeof (icmph_t));
2755 2755 icmph.icmph_type = ICMP_PARAM_PROBLEM;
2756 2756 icmph.icmph_pp_ptr = ptr;
2757 2757 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutParmProbs);
2758 2758 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
2759 2759 }
2760 2760
2761 2761 /*
2762 2762 * Build and ship an IPv4 ICMP message using the packet data in mp, and
2763 2763 * the ICMP header pointed to by "stuff". (May be called as writer.)
2764 2764 * Note: assumes that icmp_pkt_err_ok has been called to verify that
2765 2765 * an icmp error packet can be sent.
2766 2766 * Assigns an appropriate source address to the packet. If ipha_dst is
2767 2767 * one of our addresses use it for source. Otherwise let ip_output_simple
2768 2768 * pick the source address.
2769 2769 */
2770 2770 static void
2771 2771 icmp_pkt(mblk_t *mp, void *stuff, size_t len, ip_recv_attr_t *ira)
2772 2772 {
2773 2773 ipaddr_t dst;
2774 2774 icmph_t *icmph;
2775 2775 ipha_t *ipha;
2776 2776 uint_t len_needed;
2777 2777 size_t msg_len;
2778 2778 mblk_t *mp1;
2779 2779 ipaddr_t src;
2780 2780 ire_t *ire;
2781 2781 ip_xmit_attr_t ixas;
2782 2782 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2783 2783
2784 2784 ipha = (ipha_t *)mp->b_rptr;
2785 2785
2786 2786 bzero(&ixas, sizeof (ixas));
2787 2787 ixas.ixa_flags = IXAF_BASIC_SIMPLE_V4;
2788 2788 ixas.ixa_zoneid = ira->ira_zoneid;
2789 2789 ixas.ixa_ifindex = 0;
2790 2790 ixas.ixa_ipst = ipst;
2791 2791 ixas.ixa_cred = kcred;
2792 2792 ixas.ixa_cpid = NOPID;
2793 2793 ixas.ixa_tsl = ira->ira_tsl; /* Behave as a multi-level responder */
2794 2794 ixas.ixa_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
2795 2795
2796 2796 if (ira->ira_flags & IRAF_IPSEC_SECURE) {
2797 2797 /*
2798 2798 * Apply IPsec based on how IPsec was applied to
2799 2799 * the packet that had the error.
2800 2800 *
2801 2801 * If it was an outbound packet that caused the ICMP
2802 2802 * error, then the caller will have setup the IRA
2803 2803 * appropriately.
2804 2804 */
2805 2805 if (!ipsec_in_to_out(ira, &ixas, mp, ipha, NULL)) {
2806 2806 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
2807 2807 /* Note: mp already consumed and ip_drop_packet done */
2808 2808 return;
2809 2809 }
2810 2810 } else {
2811 2811 /*
2812 2812 * This is in clear. The icmp message we are building
2813 2813 * here should go out in clear, independent of our policy.
2814 2814 */
2815 2815 ixas.ixa_flags |= IXAF_NO_IPSEC;
2816 2816 }
2817 2817
2818 2818 /* Remember our eventual destination */
2819 2819 dst = ipha->ipha_src;
2820 2820
2821 2821 /*
2822 2822 * If the packet was for one of our unicast addresses, make
2823 2823 * sure we respond with that as the source. Otherwise
2824 2824 * have ip_output_simple pick the source address.
2825 2825 */
2826 2826 ire = ire_ftable_lookup_v4(ipha->ipha_dst, 0, 0,
2827 2827 (IRE_LOCAL|IRE_LOOPBACK), NULL, ira->ira_zoneid, NULL,
2828 2828 MATCH_IRE_TYPE|MATCH_IRE_ZONEONLY, 0, ipst, NULL);
2829 2829 if (ire != NULL) {
2830 2830 ire_refrele(ire);
2831 2831 src = ipha->ipha_dst;
2832 2832 } else {
2833 2833 src = INADDR_ANY;
2834 2834 ixas.ixa_flags |= IXAF_SET_SOURCE;
2835 2835 }
2836 2836
2837 2837 /*
2838 2838 * Check if we can send back more then 8 bytes in addition to
2839 2839 * the IP header. We try to send 64 bytes of data and the internal
2840 2840 * header in the special cases of ipv4 encapsulated ipv4 or ipv6.
2841 2841 */
2842 2842 len_needed = IPH_HDR_LENGTH(ipha);
2843 2843 if (ipha->ipha_protocol == IPPROTO_ENCAP ||
2844 2844 ipha->ipha_protocol == IPPROTO_IPV6) {
2845 2845 if (!pullupmsg(mp, -1)) {
2846 2846 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
2847 2847 ip_drop_output("ipIfStatsOutDiscards", mp, NULL);
2848 2848 freemsg(mp);
2849 2849 return;
2850 2850 }
2851 2851 ipha = (ipha_t *)mp->b_rptr;
2852 2852
2853 2853 if (ipha->ipha_protocol == IPPROTO_ENCAP) {
2854 2854 len_needed += IPH_HDR_LENGTH(((uchar_t *)ipha +
2855 2855 len_needed));
2856 2856 } else {
2857 2857 ip6_t *ip6h = (ip6_t *)((uchar_t *)ipha + len_needed);
2858 2858
2859 2859 ASSERT(ipha->ipha_protocol == IPPROTO_IPV6);
2860 2860 len_needed += ip_hdr_length_v6(mp, ip6h);
2861 2861 }
2862 2862 }
2863 2863 len_needed += ipst->ips_ip_icmp_return;
2864 2864 msg_len = msgdsize(mp);
2865 2865 if (msg_len > len_needed) {
2866 2866 (void) adjmsg(mp, len_needed - msg_len);
2867 2867 msg_len = len_needed;
2868 2868 }
2869 2869 mp1 = allocb(sizeof (icmp_ipha) + len, BPRI_MED);
2870 2870 if (mp1 == NULL) {
2871 2871 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutErrors);
2872 2872 freemsg(mp);
2873 2873 return;
2874 2874 }
2875 2875 mp1->b_cont = mp;
2876 2876 mp = mp1;
2877 2877
2878 2878 /*
2879 2879 * Set IXAF_TRUSTED_ICMP so we can let the ICMP messages this
2880 2880 * node generates be accepted in peace by all on-host destinations.
2881 2881 * If we do NOT assume that all on-host destinations trust
2882 2882 * self-generated ICMP messages, then rework here, ip6.c, and spd.c.
2883 2883 * (Look for IXAF_TRUSTED_ICMP).
2884 2884 */
2885 2885 ixas.ixa_flags |= IXAF_TRUSTED_ICMP;
2886 2886
2887 2887 ipha = (ipha_t *)mp->b_rptr;
2888 2888 mp1->b_wptr = (uchar_t *)ipha + (sizeof (icmp_ipha) + len);
2889 2889 *ipha = icmp_ipha;
2890 2890 ipha->ipha_src = src;
2891 2891 ipha->ipha_dst = dst;
2892 2892 ipha->ipha_ttl = ipst->ips_ip_def_ttl;
2893 2893 msg_len += sizeof (icmp_ipha) + len;
2894 2894 if (msg_len > IP_MAXPACKET) {
2895 2895 (void) adjmsg(mp, IP_MAXPACKET - msg_len);
2896 2896 msg_len = IP_MAXPACKET;
2897 2897 }
2898 2898 ipha->ipha_length = htons((uint16_t)msg_len);
2899 2899 icmph = (icmph_t *)&ipha[1];
2900 2900 bcopy(stuff, icmph, len);
2901 2901 icmph->icmph_checksum = 0;
2902 2902 icmph->icmph_checksum = IP_CSUM(mp, (int32_t)sizeof (ipha_t), 0);
2903 2903 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutMsgs);
2904 2904
2905 2905 (void) ip_output_simple(mp, &ixas);
2906 2906 ixa_cleanup(&ixas);
2907 2907 }
2908 2908
2909 2909 /*
2910 2910 * Determine if an ICMP error packet can be sent given the rate limit.
2911 2911 * The limit consists of an average frequency (icmp_pkt_err_interval measured
2912 2912 * in milliseconds) and a burst size. Burst size number of packets can
2913 2913 * be sent arbitrarely closely spaced.
2914 2914 * The state is tracked using two variables to implement an approximate
2915 2915 * token bucket filter:
2916 2916 * icmp_pkt_err_last - lbolt value when the last burst started
2917 2917 * icmp_pkt_err_sent - number of packets sent in current burst
2918 2918 */
2919 2919 boolean_t
2920 2920 icmp_err_rate_limit(ip_stack_t *ipst)
2921 2921 {
2922 2922 clock_t now = TICK_TO_MSEC(ddi_get_lbolt());
2923 2923 uint_t refilled; /* Number of packets refilled in tbf since last */
2924 2924 /* Guard against changes by loading into local variable */
2925 2925 uint_t err_interval = ipst->ips_ip_icmp_err_interval;
2926 2926
2927 2927 if (err_interval == 0)
2928 2928 return (B_FALSE);
2929 2929
2930 2930 if (ipst->ips_icmp_pkt_err_last > now) {
2931 2931 /* 100HZ lbolt in ms for 32bit arch wraps every 49.7 days */
2932 2932 ipst->ips_icmp_pkt_err_last = 0;
2933 2933 ipst->ips_icmp_pkt_err_sent = 0;
2934 2934 }
2935 2935 /*
2936 2936 * If we are in a burst update the token bucket filter.
2937 2937 * Update the "last" time to be close to "now" but make sure
2938 2938 * we don't loose precision.
2939 2939 */
2940 2940 if (ipst->ips_icmp_pkt_err_sent != 0) {
2941 2941 refilled = (now - ipst->ips_icmp_pkt_err_last)/err_interval;
2942 2942 if (refilled > ipst->ips_icmp_pkt_err_sent) {
2943 2943 ipst->ips_icmp_pkt_err_sent = 0;
2944 2944 } else {
2945 2945 ipst->ips_icmp_pkt_err_sent -= refilled;
2946 2946 ipst->ips_icmp_pkt_err_last += refilled * err_interval;
2947 2947 }
2948 2948 }
2949 2949 if (ipst->ips_icmp_pkt_err_sent == 0) {
2950 2950 /* Start of new burst */
2951 2951 ipst->ips_icmp_pkt_err_last = now;
2952 2952 }
2953 2953 if (ipst->ips_icmp_pkt_err_sent < ipst->ips_ip_icmp_err_burst) {
2954 2954 ipst->ips_icmp_pkt_err_sent++;
2955 2955 ip1dbg(("icmp_err_rate_limit: %d sent in burst\n",
2956 2956 ipst->ips_icmp_pkt_err_sent));
2957 2957 return (B_FALSE);
2958 2958 }
2959 2959 ip1dbg(("icmp_err_rate_limit: dropped\n"));
2960 2960 return (B_TRUE);
2961 2961 }
2962 2962
2963 2963 /*
2964 2964 * Check if it is ok to send an IPv4 ICMP error packet in
2965 2965 * response to the IPv4 packet in mp.
2966 2966 * Free the message and return null if no
2967 2967 * ICMP error packet should be sent.
2968 2968 */
2969 2969 static mblk_t *
2970 2970 icmp_pkt_err_ok(mblk_t *mp, ip_recv_attr_t *ira)
2971 2971 {
2972 2972 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2973 2973 icmph_t *icmph;
2974 2974 ipha_t *ipha;
2975 2975 uint_t len_needed;
2976 2976
2977 2977 if (!mp)
2978 2978 return (NULL);
2979 2979 ipha = (ipha_t *)mp->b_rptr;
2980 2980 if (ip_csum_hdr(ipha)) {
2981 2981 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsInCksumErrs);
2982 2982 ip_drop_input("ipIfStatsInCksumErrs", mp, NULL);
2983 2983 freemsg(mp);
2984 2984 return (NULL);
2985 2985 }
2986 2986 if (ip_type_v4(ipha->ipha_dst, ipst) == IRE_BROADCAST ||
2987 2987 ip_type_v4(ipha->ipha_src, ipst) == IRE_BROADCAST ||
2988 2988 CLASSD(ipha->ipha_dst) ||
2989 2989 CLASSD(ipha->ipha_src) ||
2990 2990 (ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_OFFSET)) {
2991 2991 /* Note: only errors to the fragment with offset 0 */
2992 2992 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
2993 2993 freemsg(mp);
2994 2994 return (NULL);
2995 2995 }
2996 2996 if (ipha->ipha_protocol == IPPROTO_ICMP) {
2997 2997 /*
2998 2998 * Check the ICMP type. RFC 1122 sez: don't send ICMP
2999 2999 * errors in response to any ICMP errors.
3000 3000 */
3001 3001 len_needed = IPH_HDR_LENGTH(ipha) + ICMPH_SIZE;
3002 3002 if (mp->b_wptr - mp->b_rptr < len_needed) {
3003 3003 if (!pullupmsg(mp, len_needed)) {
3004 3004 BUMP_MIB(&ipst->ips_icmp_mib, icmpInErrors);
3005 3005 freemsg(mp);
3006 3006 return (NULL);
3007 3007 }
3008 3008 ipha = (ipha_t *)mp->b_rptr;
3009 3009 }
3010 3010 icmph = (icmph_t *)
3011 3011 (&((char *)ipha)[IPH_HDR_LENGTH(ipha)]);
3012 3012 switch (icmph->icmph_type) {
3013 3013 case ICMP_DEST_UNREACHABLE:
3014 3014 case ICMP_SOURCE_QUENCH:
3015 3015 case ICMP_TIME_EXCEEDED:
3016 3016 case ICMP_PARAM_PROBLEM:
3017 3017 case ICMP_REDIRECT:
3018 3018 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
3019 3019 freemsg(mp);
3020 3020 return (NULL);
3021 3021 default:
3022 3022 break;
3023 3023 }
3024 3024 }
3025 3025 /*
3026 3026 * If this is a labeled system, then check to see if we're allowed to
3027 3027 * send a response to this particular sender. If not, then just drop.
3028 3028 */
3029 3029 if (is_system_labeled() && !tsol_can_reply_error(mp, ira)) {
3030 3030 ip2dbg(("icmp_pkt_err_ok: can't respond to packet\n"));
3031 3031 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
3032 3032 freemsg(mp);
3033 3033 return (NULL);
3034 3034 }
3035 3035 if (icmp_err_rate_limit(ipst)) {
3036 3036 /*
3037 3037 * Only send ICMP error packets every so often.
3038 3038 * This should be done on a per port/source basis,
3039 3039 * but for now this will suffice.
3040 3040 */
3041 3041 freemsg(mp);
3042 3042 return (NULL);
3043 3043 }
3044 3044 return (mp);
3045 3045 }
3046 3046
3047 3047 /*
3048 3048 * Called when a packet was sent out the same link that it arrived on.
3049 3049 * Check if it is ok to send a redirect and then send it.
3050 3050 */
3051 3051 void
3052 3052 ip_send_potential_redirect_v4(mblk_t *mp, ipha_t *ipha, ire_t *ire,
3053 3053 ip_recv_attr_t *ira)
3054 3054 {
3055 3055 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3056 3056 ipaddr_t src, nhop;
3057 3057 mblk_t *mp1;
3058 3058 ire_t *nhop_ire;
3059 3059
3060 3060 /*
3061 3061 * Check the source address to see if it originated
3062 3062 * on the same logical subnet it is going back out on.
3063 3063 * If so, we should be able to send it a redirect.
3064 3064 * Avoid sending a redirect if the destination
3065 3065 * is directly connected (i.e., we matched an IRE_ONLINK),
3066 3066 * or if the packet was source routed out this interface.
3067 3067 *
3068 3068 * We avoid sending a redirect if the
3069 3069 * destination is directly connected
3070 3070 * because it is possible that multiple
3071 3071 * IP subnets may have been configured on
3072 3072 * the link, and the source may not
3073 3073 * be on the same subnet as ip destination,
3074 3074 * even though they are on the same
3075 3075 * physical link.
3076 3076 */
3077 3077 if ((ire->ire_type & IRE_ONLINK) ||
3078 3078 ip_source_routed(ipha, ipst))
3079 3079 return;
3080 3080
3081 3081 nhop_ire = ire_nexthop(ire);
3082 3082 if (nhop_ire == NULL)
3083 3083 return;
3084 3084
3085 3085 nhop = nhop_ire->ire_addr;
3086 3086
3087 3087 if (nhop_ire->ire_type & IRE_IF_CLONE) {
3088 3088 ire_t *ire2;
3089 3089
3090 3090 /* Follow ire_dep_parent to find non-clone IRE_INTERFACE */
3091 3091 mutex_enter(&nhop_ire->ire_lock);
3092 3092 ire2 = nhop_ire->ire_dep_parent;
3093 3093 if (ire2 != NULL)
3094 3094 ire_refhold(ire2);
3095 3095 mutex_exit(&nhop_ire->ire_lock);
3096 3096 ire_refrele(nhop_ire);
3097 3097 nhop_ire = ire2;
3098 3098 }
3099 3099 if (nhop_ire == NULL)
3100 3100 return;
3101 3101
3102 3102 ASSERT(!(nhop_ire->ire_type & IRE_IF_CLONE));
3103 3103
3104 3104 src = ipha->ipha_src;
3105 3105
3106 3106 /*
3107 3107 * We look at the interface ire for the nexthop,
3108 3108 * to see if ipha_src is in the same subnet
3109 3109 * as the nexthop.
3110 3110 */
3111 3111 if ((src & nhop_ire->ire_mask) == (nhop & nhop_ire->ire_mask)) {
3112 3112 /*
3113 3113 * The source is directly connected.
3114 3114 */
3115 3115 mp1 = copymsg(mp);
3116 3116 if (mp1 != NULL) {
3117 3117 icmp_send_redirect(mp1, nhop, ira);
3118 3118 }
3119 3119 }
3120 3120 ire_refrele(nhop_ire);
3121 3121 }
3122 3122
3123 3123 /*
3124 3124 * Generate an ICMP redirect message.
3125 3125 */
3126 3126 static void
3127 3127 icmp_send_redirect(mblk_t *mp, ipaddr_t gateway, ip_recv_attr_t *ira)
3128 3128 {
3129 3129 icmph_t icmph;
3130 3130 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3131 3131
3132 3132 mp = icmp_pkt_err_ok(mp, ira);
3133 3133 if (mp == NULL)
3134 3134 return;
3135 3135
3136 3136 bzero(&icmph, sizeof (icmph_t));
3137 3137 icmph.icmph_type = ICMP_REDIRECT;
3138 3138 icmph.icmph_code = 1;
3139 3139 icmph.icmph_rd_gateway = gateway;
3140 3140 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutRedirects);
3141 3141 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3142 3142 }
3143 3143
3144 3144 /*
3145 3145 * Generate an ICMP time exceeded message.
3146 3146 */
3147 3147 void
3148 3148 icmp_time_exceeded(mblk_t *mp, uint8_t code, ip_recv_attr_t *ira)
3149 3149 {
3150 3150 icmph_t icmph;
3151 3151 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3152 3152
3153 3153 mp = icmp_pkt_err_ok(mp, ira);
3154 3154 if (mp == NULL)
3155 3155 return;
3156 3156
3157 3157 bzero(&icmph, sizeof (icmph_t));
3158 3158 icmph.icmph_type = ICMP_TIME_EXCEEDED;
3159 3159 icmph.icmph_code = code;
3160 3160 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutTimeExcds);
3161 3161 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3162 3162 }
3163 3163
3164 3164 /*
3165 3165 * Generate an ICMP unreachable message.
3166 3166 * When called from ip_output side a minimal ip_recv_attr_t needs to be
3167 3167 * constructed by the caller.
3168 3168 */
3169 3169 void
3170 3170 icmp_unreachable(mblk_t *mp, uint8_t code, ip_recv_attr_t *ira)
3171 3171 {
3172 3172 icmph_t icmph;
3173 3173 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3174 3174
3175 3175 mp = icmp_pkt_err_ok(mp, ira);
3176 3176 if (mp == NULL)
3177 3177 return;
3178 3178
3179 3179 bzero(&icmph, sizeof (icmph_t));
3180 3180 icmph.icmph_type = ICMP_DEST_UNREACHABLE;
3181 3181 icmph.icmph_code = code;
3182 3182 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDestUnreachs);
3183 3183 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3184 3184 }
3185 3185
3186 3186 /*
3187 3187 * Latch in the IPsec state for a stream based the policy in the listener
3188 3188 * and the actions in the ip_recv_attr_t.
3189 3189 * Called directly from TCP and SCTP.
3190 3190 */
3191 3191 boolean_t
3192 3192 ip_ipsec_policy_inherit(conn_t *connp, conn_t *lconnp, ip_recv_attr_t *ira)
3193 3193 {
3194 3194 ASSERT(lconnp->conn_policy != NULL);
3195 3195 ASSERT(connp->conn_policy == NULL);
3196 3196
3197 3197 IPPH_REFHOLD(lconnp->conn_policy);
3198 3198 connp->conn_policy = lconnp->conn_policy;
3199 3199
3200 3200 if (ira->ira_ipsec_action != NULL) {
3201 3201 if (connp->conn_latch == NULL) {
3202 3202 connp->conn_latch = iplatch_create();
3203 3203 if (connp->conn_latch == NULL)
3204 3204 return (B_FALSE);
3205 3205 }
3206 3206 ipsec_latch_inbound(connp, ira);
3207 3207 }
3208 3208 return (B_TRUE);
3209 3209 }
3210 3210
3211 3211 /*
3212 3212 * Verify whether or not the IP address is a valid local address.
3213 3213 * Could be a unicast, including one for a down interface.
3214 3214 * If allow_mcbc then a multicast or broadcast address is also
3215 3215 * acceptable.
3216 3216 *
3217 3217 * In the case of a broadcast/multicast address, however, the
3218 3218 * upper protocol is expected to reset the src address
3219 3219 * to zero when we return IPVL_MCAST/IPVL_BCAST so that
3220 3220 * no packets are emitted with broadcast/multicast address as
3221 3221 * source address (that violates hosts requirements RFC 1122)
3222 3222 * The addresses valid for bind are:
3223 3223 * (1) - INADDR_ANY (0)
3224 3224 * (2) - IP address of an UP interface
3225 3225 * (3) - IP address of a DOWN interface
3226 3226 * (4) - valid local IP broadcast addresses. In this case
3227 3227 * the conn will only receive packets destined to
3228 3228 * the specified broadcast address.
3229 3229 * (5) - a multicast address. In this case
3230 3230 * the conn will only receive packets destined to
3231 3231 * the specified multicast address. Note: the
3232 3232 * application still has to issue an
3233 3233 * IP_ADD_MEMBERSHIP socket option.
3234 3234 *
3235 3235 * In all the above cases, the bound address must be valid in the current zone.
3236 3236 * When the address is loopback, multicast or broadcast, there might be many
3237 3237 * matching IREs so bind has to look up based on the zone.
3238 3238 */
3239 3239 ip_laddr_t
3240 3240 ip_laddr_verify_v4(ipaddr_t src_addr, zoneid_t zoneid,
3241 3241 ip_stack_t *ipst, boolean_t allow_mcbc)
3242 3242 {
3243 3243 ire_t *src_ire;
3244 3244
3245 3245 ASSERT(src_addr != INADDR_ANY);
3246 3246
3247 3247 src_ire = ire_ftable_lookup_v4(src_addr, 0, 0, 0,
3248 3248 NULL, zoneid, NULL, MATCH_IRE_ZONEONLY, 0, ipst, NULL);
3249 3249
3250 3250 /*
3251 3251 * If an address other than in6addr_any is requested,
3252 3252 * we verify that it is a valid address for bind
3253 3253 * Note: Following code is in if-else-if form for
3254 3254 * readability compared to a condition check.
3255 3255 */
3256 3256 if (src_ire != NULL && (src_ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK))) {
3257 3257 /*
3258 3258 * (2) Bind to address of local UP interface
3259 3259 */
3260 3260 ire_refrele(src_ire);
3261 3261 return (IPVL_UNICAST_UP);
3262 3262 } else if (src_ire != NULL && src_ire->ire_type & IRE_BROADCAST) {
3263 3263 /*
3264 3264 * (4) Bind to broadcast address
3265 3265 */
3266 3266 ire_refrele(src_ire);
3267 3267 if (allow_mcbc)
3268 3268 return (IPVL_BCAST);
3269 3269 else
3270 3270 return (IPVL_BAD);
3271 3271 } else if (CLASSD(src_addr)) {
3272 3272 /* (5) bind to multicast address. */
3273 3273 if (src_ire != NULL)
3274 3274 ire_refrele(src_ire);
3275 3275
3276 3276 if (allow_mcbc)
3277 3277 return (IPVL_MCAST);
3278 3278 else
3279 3279 return (IPVL_BAD);
3280 3280 } else {
3281 3281 ipif_t *ipif;
3282 3282
3283 3283 /*
3284 3284 * (3) Bind to address of local DOWN interface?
3285 3285 * (ipif_lookup_addr() looks up all interfaces
3286 3286 * but we do not get here for UP interfaces
3287 3287 * - case (2) above)
3288 3288 */
3289 3289 if (src_ire != NULL)
3290 3290 ire_refrele(src_ire);
3291 3291
3292 3292 ipif = ipif_lookup_addr(src_addr, NULL, zoneid, ipst);
3293 3293 if (ipif == NULL)
3294 3294 return (IPVL_BAD);
3295 3295
3296 3296 /* Not a useful source? */
3297 3297 if (ipif->ipif_flags & (IPIF_NOLOCAL | IPIF_ANYCAST)) {
3298 3298 ipif_refrele(ipif);
3299 3299 return (IPVL_BAD);
3300 3300 }
3301 3301 ipif_refrele(ipif);
3302 3302 return (IPVL_UNICAST_DOWN);
3303 3303 }
3304 3304 }
3305 3305
3306 3306 /*
3307 3307 * Insert in the bind fanout for IPv4 and IPv6.
3308 3308 * The caller should already have used ip_laddr_verify_v*() before calling
3309 3309 * this.
3310 3310 */
3311 3311 int
3312 3312 ip_laddr_fanout_insert(conn_t *connp)
3313 3313 {
3314 3314 int error;
3315 3315
3316 3316 /*
3317 3317 * Allow setting new policies. For example, disconnects result
3318 3318 * in us being called. As we would have set conn_policy_cached
3319 3319 * to B_TRUE before, we should set it to B_FALSE, so that policy
3320 3320 * can change after the disconnect.
3321 3321 */
3322 3322 connp->conn_policy_cached = B_FALSE;
3323 3323
3324 3324 error = ipcl_bind_insert(connp);
3325 3325 if (error != 0) {
3326 3326 if (connp->conn_anon_port) {
3327 3327 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
3328 3328 connp->conn_mlp_type, connp->conn_proto,
3329 3329 ntohs(connp->conn_lport), B_FALSE);
3330 3330 }
3331 3331 connp->conn_mlp_type = mlptSingle;
3332 3332 }
3333 3333 return (error);
3334 3334 }
3335 3335
3336 3336 /*
3337 3337 * Verify that both the source and destination addresses are valid. If
3338 3338 * IPDF_VERIFY_DST is not set, then the destination address may be unreachable,
3339 3339 * i.e. have no route to it. Protocols like TCP want to verify destination
3340 3340 * reachability, while tunnels do not.
3341 3341 *
3342 3342 * Determine the route, the interface, and (optionally) the source address
3343 3343 * to use to reach a given destination.
3344 3344 * Note that we allow connect to broadcast and multicast addresses when
3345 3345 * IPDF_ALLOW_MCBC is set.
3346 3346 * first_hop and dst_addr are normally the same, but if source routing
3347 3347 * they will differ; in that case the first_hop is what we'll use for the
3348 3348 * routing lookup but the dce and label checks will be done on dst_addr,
3349 3349 *
3350 3350 * If uinfo is set, then we fill in the best available information
3351 3351 * we have for the destination. This is based on (in priority order) any
3352 3352 * metrics and path MTU stored in a dce_t, route metrics, and finally the
3353 3353 * ill_mtu/ill_mc_mtu.
3354 3354 *
3355 3355 * Tsol note: If we have a source route then dst_addr != firsthop. But we
3356 3356 * always do the label check on dst_addr.
3357 3357 */
3358 3358 int
3359 3359 ip_set_destination_v4(ipaddr_t *src_addrp, ipaddr_t dst_addr, ipaddr_t firsthop,
3360 3360 ip_xmit_attr_t *ixa, iulp_t *uinfo, uint32_t flags, uint_t mac_mode)
3361 3361 {
3362 3362 ire_t *ire = NULL;
3363 3363 int error = 0;
3364 3364 ipaddr_t setsrc; /* RTF_SETSRC */
3365 3365 zoneid_t zoneid = ixa->ixa_zoneid; /* Honors SO_ALLZONES */
3366 3366 ip_stack_t *ipst = ixa->ixa_ipst;
3367 3367 dce_t *dce;
3368 3368 uint_t pmtu;
3369 3369 uint_t generation;
3370 3370 nce_t *nce;
3371 3371 ill_t *ill = NULL;
3372 3372 boolean_t multirt = B_FALSE;
3373 3373
3374 3374 ASSERT(ixa->ixa_flags & IXAF_IS_IPV4);
3375 3375
3376 3376 /*
3377 3377 * We never send to zero; the ULPs map it to the loopback address.
3378 3378 * We can't allow it since we use zero to mean unitialized in some
3379 3379 * places.
3380 3380 */
3381 3381 ASSERT(dst_addr != INADDR_ANY);
3382 3382
3383 3383 if (is_system_labeled()) {
3384 3384 ts_label_t *tsl = NULL;
3385 3385
3386 3386 error = tsol_check_dest(ixa->ixa_tsl, &dst_addr, IPV4_VERSION,
3387 3387 mac_mode, (flags & IPDF_ZONE_IS_GLOBAL) != 0, &tsl);
3388 3388 if (error != 0)
3389 3389 return (error);
3390 3390 if (tsl != NULL) {
3391 3391 /* Update the label */
3392 3392 ip_xmit_attr_replace_tsl(ixa, tsl);
3393 3393 }
3394 3394 }
3395 3395
3396 3396 setsrc = INADDR_ANY;
3397 3397 /*
3398 3398 * Select a route; For IPMP interfaces, we would only select
3399 3399 * a "hidden" route (i.e., going through a specific under_ill)
3400 3400 * if ixa_ifindex has been specified.
3401 3401 */
3402 3402 ire = ip_select_route_v4(firsthop, *src_addrp, ixa,
3403 3403 &generation, &setsrc, &error, &multirt);
3404 3404 ASSERT(ire != NULL); /* IRE_NOROUTE if none found */
3405 3405 if (error != 0)
3406 3406 goto bad_addr;
3407 3407
3408 3408 /*
3409 3409 * ire can't be a broadcast or multicast unless IPDF_ALLOW_MCBC is set.
3410 3410 * If IPDF_VERIFY_DST is set, the destination must be reachable;
3411 3411 * Otherwise the destination needn't be reachable.
3412 3412 *
3413 3413 * If we match on a reject or black hole, then we've got a
3414 3414 * local failure. May as well fail out the connect() attempt,
3415 3415 * since it's never going to succeed.
3416 3416 */
3417 3417 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
3418 3418 /*
3419 3419 * If we're verifying destination reachability, we always want
3420 3420 * to complain here.
3421 3421 *
3422 3422 * If we're not verifying destination reachability but the
3423 3423 * destination has a route, we still want to fail on the
3424 3424 * temporary address and broadcast address tests.
3425 3425 *
3426 3426 * In both cases do we let the code continue so some reasonable
3427 3427 * information is returned to the caller. That enables the
3428 3428 * caller to use (and even cache) the IRE. conn_ip_ouput will
3429 3429 * use the generation mismatch path to check for the unreachable
3430 3430 * case thereby avoiding any specific check in the main path.
3431 3431 */
3432 3432 ASSERT(generation == IRE_GENERATION_VERIFY);
3433 3433 if (flags & IPDF_VERIFY_DST) {
3434 3434 /*
3435 3435 * Set errno but continue to set up ixa_ire to be
3436 3436 * the RTF_REJECT|RTF_BLACKHOLE IRE.
3437 3437 * That allows callers to use ip_output to get an
3438 3438 * ICMP error back.
3439 3439 */
3440 3440 if (!(ire->ire_type & IRE_HOST))
3441 3441 error = ENETUNREACH;
3442 3442 else
3443 3443 error = EHOSTUNREACH;
3444 3444 }
3445 3445 }
3446 3446
3447 3447 if ((ire->ire_type & (IRE_BROADCAST|IRE_MULTICAST)) &&
3448 3448 !(flags & IPDF_ALLOW_MCBC)) {
3449 3449 ire_refrele(ire);
3450 3450 ire = ire_reject(ipst, B_FALSE);
3451 3451 generation = IRE_GENERATION_VERIFY;
3452 3452 error = ENETUNREACH;
3453 3453 }
3454 3454
3455 3455 /* Cache things */
3456 3456 if (ixa->ixa_ire != NULL)
3457 3457 ire_refrele_notr(ixa->ixa_ire);
3458 3458 #ifdef DEBUG
3459 3459 ire_refhold_notr(ire);
3460 3460 ire_refrele(ire);
3461 3461 #endif
3462 3462 ixa->ixa_ire = ire;
3463 3463 ixa->ixa_ire_generation = generation;
3464 3464
3465 3465 /*
3466 3466 * Ensure that ixa_dce is always set any time that ixa_ire is set,
3467 3467 * since some callers will send a packet to conn_ip_output() even if
3468 3468 * there's an error.
3469 3469 */
3470 3470 if (flags & IPDF_UNIQUE_DCE) {
3471 3471 /* Fallback to the default dce if allocation fails */
3472 3472 dce = dce_lookup_and_add_v4(dst_addr, ipst);
3473 3473 if (dce != NULL)
3474 3474 generation = dce->dce_generation;
3475 3475 else
3476 3476 dce = dce_lookup_v4(dst_addr, ipst, &generation);
3477 3477 } else {
3478 3478 dce = dce_lookup_v4(dst_addr, ipst, &generation);
3479 3479 }
3480 3480 ASSERT(dce != NULL);
3481 3481 if (ixa->ixa_dce != NULL)
3482 3482 dce_refrele_notr(ixa->ixa_dce);
3483 3483 #ifdef DEBUG
3484 3484 dce_refhold_notr(dce);
3485 3485 dce_refrele(dce);
3486 3486 #endif
3487 3487 ixa->ixa_dce = dce;
3488 3488 ixa->ixa_dce_generation = generation;
3489 3489
3490 3490 /*
3491 3491 * For multicast with multirt we have a flag passed back from
3492 3492 * ire_lookup_multi_ill_v4 since we don't have an IRE for each
3493 3493 * possible multicast address.
3494 3494 * We also need a flag for multicast since we can't check
3495 3495 * whether RTF_MULTIRT is set in ixa_ire for multicast.
3496 3496 */
3497 3497 if (multirt) {
3498 3498 ixa->ixa_postfragfn = ip_postfrag_multirt_v4;
3499 3499 ixa->ixa_flags |= IXAF_MULTIRT_MULTICAST;
3500 3500 } else {
3501 3501 ixa->ixa_postfragfn = ire->ire_postfragfn;
3502 3502 ixa->ixa_flags &= ~IXAF_MULTIRT_MULTICAST;
3503 3503 }
3504 3504 if (!(ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))) {
3505 3505 /* Get an nce to cache. */
3506 3506 nce = ire_to_nce(ire, firsthop, NULL);
3507 3507 if (nce == NULL) {
3508 3508 /* Allocation failure? */
3509 3509 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3510 3510 } else {
3511 3511 if (ixa->ixa_nce != NULL)
3512 3512 nce_refrele(ixa->ixa_nce);
3513 3513 ixa->ixa_nce = nce;
3514 3514 }
3515 3515 }
3516 3516
3517 3517 /*
3518 3518 * If the source address is a loopback address, the
3519 3519 * destination had best be local or multicast.
3520 3520 * If we are sending to an IRE_LOCAL using a loopback source then
3521 3521 * it had better be the same zoneid.
3522 3522 */
3523 3523 if (*src_addrp == htonl(INADDR_LOOPBACK)) {
3524 3524 if ((ire->ire_type & IRE_LOCAL) && ire->ire_zoneid != zoneid) {
3525 3525 ire = NULL; /* Stored in ixa_ire */
3526 3526 error = EADDRNOTAVAIL;
3527 3527 goto bad_addr;
3528 3528 }
3529 3529 if (!(ire->ire_type & (IRE_LOOPBACK|IRE_LOCAL|IRE_MULTICAST))) {
3530 3530 ire = NULL; /* Stored in ixa_ire */
3531 3531 error = EADDRNOTAVAIL;
3532 3532 goto bad_addr;
3533 3533 }
3534 3534 }
3535 3535 if (ire->ire_type & IRE_BROADCAST) {
3536 3536 /*
3537 3537 * If the ULP didn't have a specified source, then we
3538 3538 * make sure we reselect the source when sending
3539 3539 * broadcasts out different interfaces.
3540 3540 */
3541 3541 if (flags & IPDF_SELECT_SRC)
3542 3542 ixa->ixa_flags |= IXAF_SET_SOURCE;
3543 3543 else
3544 3544 ixa->ixa_flags &= ~IXAF_SET_SOURCE;
3545 3545 }
3546 3546
3547 3547 /*
3548 3548 * Does the caller want us to pick a source address?
3549 3549 */
3550 3550 if (flags & IPDF_SELECT_SRC) {
3551 3551 ipaddr_t src_addr;
3552 3552
3553 3553 /*
3554 3554 * We use use ire_nexthop_ill to avoid the under ipmp
3555 3555 * interface for source address selection. Note that for ipmp
3556 3556 * probe packets, ixa_ifindex would have been specified, and
3557 3557 * the ip_select_route() invocation would have picked an ire
3558 3558 * will ire_ill pointing at an under interface.
3559 3559 */
3560 3560 ill = ire_nexthop_ill(ire);
3561 3561
3562 3562 /* If unreachable we have no ill but need some source */
3563 3563 if (ill == NULL) {
3564 3564 src_addr = htonl(INADDR_LOOPBACK);
3565 3565 /* Make sure we look for a better source address */
3566 3566 generation = SRC_GENERATION_VERIFY;
3567 3567 } else {
3568 3568 error = ip_select_source_v4(ill, setsrc, dst_addr,
3569 3569 ixa->ixa_multicast_ifaddr, zoneid,
3570 3570 ipst, &src_addr, &generation, NULL);
3571 3571 if (error != 0) {
3572 3572 ire = NULL; /* Stored in ixa_ire */
3573 3573 goto bad_addr;
3574 3574 }
3575 3575 }
3576 3576
3577 3577 /*
3578 3578 * We allow the source address to to down.
3579 3579 * However, we check that we don't use the loopback address
3580 3580 * as a source when sending out on the wire.
3581 3581 */
3582 3582 if ((src_addr == htonl(INADDR_LOOPBACK)) &&
3583 3583 !(ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK|IRE_MULTICAST)) &&
3584 3584 !(ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))) {
3585 3585 ire = NULL; /* Stored in ixa_ire */
3586 3586 error = EADDRNOTAVAIL;
3587 3587 goto bad_addr;
3588 3588 }
3589 3589
3590 3590 *src_addrp = src_addr;
3591 3591 ixa->ixa_src_generation = generation;
3592 3592 }
3593 3593
3594 3594 /*
3595 3595 * Make sure we don't leave an unreachable ixa_nce in place
3596 3596 * since ip_select_route is used when we unplumb i.e., remove
3597 3597 * references on ixa_ire, ixa_nce, and ixa_dce.
3598 3598 */
3599 3599 nce = ixa->ixa_nce;
3600 3600 if (nce != NULL && nce->nce_is_condemned) {
3601 3601 nce_refrele(nce);
3602 3602 ixa->ixa_nce = NULL;
3603 3603 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3604 3604 }
3605 3605
3606 3606 /*
3607 3607 * The caller has set IXAF_PMTU_DISCOVERY if path MTU is desired.
3608 3608 * However, we can't do it for IPv4 multicast or broadcast.
3609 3609 */
3610 3610 if (ire->ire_type & (IRE_BROADCAST|IRE_MULTICAST))
3611 3611 ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
3612 3612
3613 3613 /*
3614 3614 * Set initial value for fragmentation limit. Either conn_ip_output
3615 3615 * or ULP might updates it when there are routing changes.
3616 3616 * Handles a NULL ixa_ire->ire_ill or a NULL ixa_nce for RTF_REJECT.
3617 3617 */
3618 3618 pmtu = ip_get_pmtu(ixa);
3619 3619 ixa->ixa_fragsize = pmtu;
3620 3620 /* Make sure ixa_fragsize and ixa_pmtu remain identical */
3621 3621 if (ixa->ixa_flags & IXAF_VERIFY_PMTU)
3622 3622 ixa->ixa_pmtu = pmtu;
3623 3623
3624 3624 /*
3625 3625 * Extract information useful for some transports.
3626 3626 * First we look for DCE metrics. Then we take what we have in
3627 3627 * the metrics in the route, where the offlink is used if we have
3628 3628 * one.
3629 3629 */
3630 3630 if (uinfo != NULL) {
3631 3631 bzero(uinfo, sizeof (*uinfo));
3632 3632
3633 3633 if (dce->dce_flags & DCEF_UINFO)
3634 3634 *uinfo = dce->dce_uinfo;
3635 3635
3636 3636 rts_merge_metrics(uinfo, &ire->ire_metrics);
3637 3637
3638 3638 /* Allow ire_metrics to decrease the path MTU from above */
3639 3639 if (uinfo->iulp_mtu == 0 || uinfo->iulp_mtu > pmtu)
3640 3640 uinfo->iulp_mtu = pmtu;
3641 3641
3642 3642 uinfo->iulp_localnet = (ire->ire_type & IRE_ONLINK) != 0;
3643 3643 uinfo->iulp_loopback = (ire->ire_type & IRE_LOOPBACK) != 0;
3644 3644 uinfo->iulp_local = (ire->ire_type & IRE_LOCAL) != 0;
3645 3645 }
3646 3646
3647 3647 if (ill != NULL)
3648 3648 ill_refrele(ill);
3649 3649
3650 3650 return (error);
3651 3651
3652 3652 bad_addr:
3653 3653 if (ire != NULL)
3654 3654 ire_refrele(ire);
3655 3655
3656 3656 if (ill != NULL)
3657 3657 ill_refrele(ill);
3658 3658
3659 3659 /*
3660 3660 * Make sure we don't leave an unreachable ixa_nce in place
3661 3661 * since ip_select_route is used when we unplumb i.e., remove
3662 3662 * references on ixa_ire, ixa_nce, and ixa_dce.
3663 3663 */
3664 3664 nce = ixa->ixa_nce;
3665 3665 if (nce != NULL && nce->nce_is_condemned) {
3666 3666 nce_refrele(nce);
3667 3667 ixa->ixa_nce = NULL;
3668 3668 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3669 3669 }
3670 3670
3671 3671 return (error);
3672 3672 }
3673 3673
3674 3674
3675 3675 /*
3676 3676 * Get the base MTU for the case when path MTU discovery is not used.
3677 3677 * Takes the MTU of the IRE into account.
3678 3678 */
3679 3679 uint_t
3680 3680 ip_get_base_mtu(ill_t *ill, ire_t *ire)
3681 3681 {
3682 3682 uint_t mtu;
3683 3683 uint_t iremtu = ire->ire_metrics.iulp_mtu;
3684 3684
3685 3685 if (ire->ire_type & (IRE_MULTICAST|IRE_BROADCAST))
3686 3686 mtu = ill->ill_mc_mtu;
3687 3687 else
3688 3688 mtu = ill->ill_mtu;
3689 3689
3690 3690 if (iremtu != 0 && iremtu < mtu)
3691 3691 mtu = iremtu;
3692 3692
3693 3693 return (mtu);
3694 3694 }
3695 3695
3696 3696 /*
3697 3697 * Get the PMTU for the attributes. Handles both IPv4 and IPv6.
3698 3698 * Assumes that ixa_ire, dce, and nce have already been set up.
3699 3699 *
3700 3700 * The caller has set IXAF_PMTU_DISCOVERY if path MTU discovery is desired.
3701 3701 * We avoid path MTU discovery if it is disabled with ndd.
3702 3702 * Furtermore, if the path MTU is too small, then we don't set DF for IPv4.
3703 3703 *
3704 3704 * NOTE: We also used to turn it off for source routed packets. That
3705 3705 * is no longer required since the dce is per final destination.
3706 3706 */
3707 3707 uint_t
3708 3708 ip_get_pmtu(ip_xmit_attr_t *ixa)
3709 3709 {
3710 3710 ip_stack_t *ipst = ixa->ixa_ipst;
3711 3711 dce_t *dce;
3712 3712 nce_t *nce;
3713 3713 ire_t *ire;
3714 3714 uint_t pmtu;
3715 3715
3716 3716 ire = ixa->ixa_ire;
3717 3717 dce = ixa->ixa_dce;
3718 3718 nce = ixa->ixa_nce;
3719 3719
3720 3720 /*
3721 3721 * If path MTU discovery has been turned off by ndd, then we ignore
3722 3722 * any dce_pmtu and for IPv4 we will not set DF.
3723 3723 */
3724 3724 if (!ipst->ips_ip_path_mtu_discovery)
3725 3725 ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
3726 3726
3727 3727 pmtu = IP_MAXPACKET;
3728 3728 /*
3729 3729 * Decide whether whether IPv4 sets DF
3730 3730 * For IPv6 "no DF" means to use the 1280 mtu
3731 3731 */
3732 3732 if (ixa->ixa_flags & IXAF_PMTU_DISCOVERY) {
3733 3733 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3734 3734 } else {
3735 3735 ixa->ixa_flags &= ~IXAF_PMTU_IPV4_DF;
3736 3736 if (!(ixa->ixa_flags & IXAF_IS_IPV4))
3737 3737 pmtu = IPV6_MIN_MTU;
3738 3738 }
3739 3739
3740 3740 /* Check if the PMTU is to old before we use it */
3741 3741 if ((dce->dce_flags & DCEF_PMTU) &&
3742 3742 TICK_TO_SEC(ddi_get_lbolt64()) - dce->dce_last_change_time >
3743 3743 ipst->ips_ip_pathmtu_interval) {
3744 3744 /*
3745 3745 * Older than 20 minutes. Drop the path MTU information.
3746 3746 */
3747 3747 mutex_enter(&dce->dce_lock);
3748 3748 dce->dce_flags &= ~(DCEF_PMTU|DCEF_TOO_SMALL_PMTU);
3749 3749 dce->dce_last_change_time = TICK_TO_SEC(ddi_get_lbolt64());
3750 3750 mutex_exit(&dce->dce_lock);
3751 3751 dce_increment_generation(dce);
3752 3752 }
3753 3753
3754 3754 /* The metrics on the route can lower the path MTU */
3755 3755 if (ire->ire_metrics.iulp_mtu != 0 &&
3756 3756 ire->ire_metrics.iulp_mtu < pmtu)
3757 3757 pmtu = ire->ire_metrics.iulp_mtu;
3758 3758
3759 3759 /*
3760 3760 * If the path MTU is smaller than some minimum, we still use dce_pmtu
3761 3761 * above (would be 576 for IPv4 and 1280 for IPv6), but we clear
3762 3762 * IXAF_PMTU_IPV4_DF so that we avoid setting DF for IPv4.
3763 3763 */
3764 3764 if (ixa->ixa_flags & IXAF_PMTU_DISCOVERY) {
3765 3765 if (dce->dce_flags & DCEF_PMTU) {
3766 3766 if (dce->dce_pmtu < pmtu)
3767 3767 pmtu = dce->dce_pmtu;
3768 3768
3769 3769 if (dce->dce_flags & DCEF_TOO_SMALL_PMTU) {
3770 3770 ixa->ixa_flags |= IXAF_PMTU_TOO_SMALL;
3771 3771 ixa->ixa_flags &= ~IXAF_PMTU_IPV4_DF;
3772 3772 } else {
3773 3773 ixa->ixa_flags &= ~IXAF_PMTU_TOO_SMALL;
3774 3774 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3775 3775 }
3776 3776 } else {
3777 3777 ixa->ixa_flags &= ~IXAF_PMTU_TOO_SMALL;
3778 3778 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3779 3779 }
3780 3780 }
3781 3781
3782 3782 /*
3783 3783 * If we have an IRE_LOCAL we use the loopback mtu instead of
3784 3784 * the ill for going out the wire i.e., IRE_LOCAL gets the same
3785 3785 * mtu as IRE_LOOPBACK.
3786 3786 */
3787 3787 if (ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK)) {
3788 3788 uint_t loopback_mtu;
3789 3789
3790 3790 loopback_mtu = (ire->ire_ipversion == IPV6_VERSION) ?
3791 3791 ip_loopback_mtu_v6plus : ip_loopback_mtuplus;
3792 3792
3793 3793 if (loopback_mtu < pmtu)
3794 3794 pmtu = loopback_mtu;
3795 3795 } else if (nce != NULL) {
3796 3796 /*
3797 3797 * Make sure we don't exceed the interface MTU.
3798 3798 * In the case of RTF_REJECT or RTF_BLACKHOLE we might not have
3799 3799 * an ill. We'd use the above IP_MAXPACKET in that case just
3800 3800 * to tell the transport something larger than zero.
3801 3801 */
3802 3802 if (ire->ire_type & (IRE_MULTICAST|IRE_BROADCAST)) {
3803 3803 if (nce->nce_common->ncec_ill->ill_mc_mtu < pmtu)
3804 3804 pmtu = nce->nce_common->ncec_ill->ill_mc_mtu;
3805 3805 if (nce->nce_common->ncec_ill != nce->nce_ill &&
3806 3806 nce->nce_ill->ill_mc_mtu < pmtu) {
3807 3807 /*
3808 3808 * for interfaces in an IPMP group, the mtu of
3809 3809 * the nce_ill (under_ill) could be different
3810 3810 * from the mtu of the ncec_ill, so we take the
3811 3811 * min of the two.
3812 3812 */
3813 3813 pmtu = nce->nce_ill->ill_mc_mtu;
3814 3814 }
3815 3815 } else {
3816 3816 if (nce->nce_common->ncec_ill->ill_mtu < pmtu)
3817 3817 pmtu = nce->nce_common->ncec_ill->ill_mtu;
3818 3818 if (nce->nce_common->ncec_ill != nce->nce_ill &&
3819 3819 nce->nce_ill->ill_mtu < pmtu) {
3820 3820 /*
3821 3821 * for interfaces in an IPMP group, the mtu of
3822 3822 * the nce_ill (under_ill) could be different
3823 3823 * from the mtu of the ncec_ill, so we take the
3824 3824 * min of the two.
3825 3825 */
3826 3826 pmtu = nce->nce_ill->ill_mtu;
3827 3827 }
3828 3828 }
3829 3829 }
3830 3830
3831 3831 /*
3832 3832 * Handle the IPV6_USE_MIN_MTU socket option or ancillary data.
3833 3833 * Only applies to IPv6.
3834 3834 */
3835 3835 if (!(ixa->ixa_flags & IXAF_IS_IPV4)) {
3836 3836 if (ixa->ixa_flags & IXAF_USE_MIN_MTU) {
3837 3837 switch (ixa->ixa_use_min_mtu) {
3838 3838 case IPV6_USE_MIN_MTU_MULTICAST:
3839 3839 if (ire->ire_type & IRE_MULTICAST)
3840 3840 pmtu = IPV6_MIN_MTU;
3841 3841 break;
3842 3842 case IPV6_USE_MIN_MTU_ALWAYS:
3843 3843 pmtu = IPV6_MIN_MTU;
3844 3844 break;
3845 3845 case IPV6_USE_MIN_MTU_NEVER:
3846 3846 break;
3847 3847 }
3848 3848 } else {
3849 3849 /* Default is IPV6_USE_MIN_MTU_MULTICAST */
3850 3850 if (ire->ire_type & IRE_MULTICAST)
3851 3851 pmtu = IPV6_MIN_MTU;
3852 3852 }
3853 3853 }
3854 3854
3855 3855 /*
3856 3856 * For multirouted IPv6 packets, the IP layer will insert a 8-byte
3857 3857 * fragment header in every packet. We compensate for those cases by
3858 3858 * returning a smaller path MTU to the ULP.
3859 3859 *
3860 3860 * In the case of CGTP then ip_output will add a fragment header.
3861 3861 * Make sure there is room for it by telling a smaller number
3862 3862 * to the transport.
3863 3863 *
3864 3864 * When IXAF_IPV6_ADDR_FRAGHDR we subtract the frag hdr here
3865 3865 * so the ULPs consistently see a iulp_pmtu and ip_get_pmtu()
3866 3866 * which is the size of the packets it can send.
3867 3867 */
3868 3868 if (!(ixa->ixa_flags & IXAF_IS_IPV4)) {
3869 3869 if ((ire->ire_flags & RTF_MULTIRT) ||
3870 3870 (ixa->ixa_flags & IXAF_MULTIRT_MULTICAST)) {
3871 3871 pmtu -= sizeof (ip6_frag_t);
3872 3872 ixa->ixa_flags |= IXAF_IPV6_ADD_FRAGHDR;
3873 3873 }
3874 3874 }
3875 3875
3876 3876 return (pmtu);
3877 3877 }
3878 3878
3879 3879 /*
3880 3880 * Carve "len" bytes out of an mblk chain, consuming any we empty, and duping
3881 3881 * the final piece where we don't. Return a pointer to the first mblk in the
3882 3882 * result, and update the pointer to the next mblk to chew on. If anything
3883 3883 * goes wrong (i.e., dupb fails), we waste everything in sight and return a
3884 3884 * NULL pointer.
3885 3885 */
3886 3886 mblk_t *
3887 3887 ip_carve_mp(mblk_t **mpp, ssize_t len)
3888 3888 {
3889 3889 mblk_t *mp0;
3890 3890 mblk_t *mp1;
3891 3891 mblk_t *mp2;
3892 3892
3893 3893 if (!len || !mpp || !(mp0 = *mpp))
3894 3894 return (NULL);
3895 3895 /* If we aren't going to consume the first mblk, we need a dup. */
3896 3896 if (mp0->b_wptr - mp0->b_rptr > len) {
3897 3897 mp1 = dupb(mp0);
3898 3898 if (mp1) {
3899 3899 /* Partition the data between the two mblks. */
3900 3900 mp1->b_wptr = mp1->b_rptr + len;
3901 3901 mp0->b_rptr = mp1->b_wptr;
3902 3902 /*
3903 3903 * after adjustments if mblk not consumed is now
3904 3904 * unaligned, try to align it. If this fails free
3905 3905 * all messages and let upper layer recover.
3906 3906 */
3907 3907 if (!OK_32PTR(mp0->b_rptr)) {
3908 3908 if (!pullupmsg(mp0, -1)) {
3909 3909 freemsg(mp0);
3910 3910 freemsg(mp1);
3911 3911 *mpp = NULL;
3912 3912 return (NULL);
3913 3913 }
3914 3914 }
3915 3915 }
3916 3916 return (mp1);
3917 3917 }
3918 3918 /* Eat through as many mblks as we need to get len bytes. */
3919 3919 len -= mp0->b_wptr - mp0->b_rptr;
3920 3920 for (mp2 = mp1 = mp0; (mp2 = mp2->b_cont) != 0 && len; mp1 = mp2) {
3921 3921 if (mp2->b_wptr - mp2->b_rptr > len) {
3922 3922 /*
3923 3923 * We won't consume the entire last mblk. Like
3924 3924 * above, dup and partition it.
3925 3925 */
3926 3926 mp1->b_cont = dupb(mp2);
3927 3927 mp1 = mp1->b_cont;
3928 3928 if (!mp1) {
3929 3929 /*
3930 3930 * Trouble. Rather than go to a lot of
3931 3931 * trouble to clean up, we free the messages.
3932 3932 * This won't be any worse than losing it on
3933 3933 * the wire.
3934 3934 */
3935 3935 freemsg(mp0);
3936 3936 freemsg(mp2);
3937 3937 *mpp = NULL;
3938 3938 return (NULL);
3939 3939 }
3940 3940 mp1->b_wptr = mp1->b_rptr + len;
3941 3941 mp2->b_rptr = mp1->b_wptr;
3942 3942 /*
3943 3943 * after adjustments if mblk not consumed is now
3944 3944 * unaligned, try to align it. If this fails free
3945 3945 * all messages and let upper layer recover.
3946 3946 */
3947 3947 if (!OK_32PTR(mp2->b_rptr)) {
3948 3948 if (!pullupmsg(mp2, -1)) {
3949 3949 freemsg(mp0);
3950 3950 freemsg(mp2);
3951 3951 *mpp = NULL;
3952 3952 return (NULL);
3953 3953 }
3954 3954 }
3955 3955 *mpp = mp2;
3956 3956 return (mp0);
3957 3957 }
3958 3958 /* Decrement len by the amount we just got. */
3959 3959 len -= mp2->b_wptr - mp2->b_rptr;
3960 3960 }
3961 3961 /*
3962 3962 * len should be reduced to zero now. If not our caller has
3963 3963 * screwed up.
3964 3964 */
3965 3965 if (len) {
3966 3966 /* Shouldn't happen! */
3967 3967 freemsg(mp0);
3968 3968 *mpp = NULL;
3969 3969 return (NULL);
3970 3970 }
3971 3971 /*
3972 3972 * We consumed up to exactly the end of an mblk. Detach the part
3973 3973 * we are returning from the rest of the chain.
3974 3974 */
3975 3975 mp1->b_cont = NULL;
3976 3976 *mpp = mp2;
3977 3977 return (mp0);
3978 3978 }
3979 3979
3980 3980 /* The ill stream is being unplumbed. Called from ip_close */
3981 3981 int
3982 3982 ip_modclose(ill_t *ill)
3983 3983 {
3984 3984 boolean_t success;
3985 3985 ipsq_t *ipsq;
3986 3986 ipif_t *ipif;
3987 3987 queue_t *q = ill->ill_rq;
3988 3988 ip_stack_t *ipst = ill->ill_ipst;
3989 3989 int i;
3990 3990 arl_ill_common_t *ai = ill->ill_common;
3991 3991
3992 3992 /*
3993 3993 * The punlink prior to this may have initiated a capability
3994 3994 * negotiation. But ipsq_enter will block until that finishes or
3995 3995 * times out.
3996 3996 */
3997 3997 success = ipsq_enter(ill, B_FALSE, NEW_OP);
3998 3998
3999 3999 /*
4000 4000 * Open/close/push/pop is guaranteed to be single threaded
4001 4001 * per stream by STREAMS. FS guarantees that all references
4002 4002 * from top are gone before close is called. So there can't
4003 4003 * be another close thread that has set CONDEMNED on this ill.
4004 4004 * and cause ipsq_enter to return failure.
4005 4005 */
4006 4006 ASSERT(success);
4007 4007 ipsq = ill->ill_phyint->phyint_ipsq;
4008 4008
4009 4009 /*
4010 4010 * Mark it condemned. No new reference will be made to this ill.
4011 4011 * Lookup functions will return an error. Threads that try to
4012 4012 * increment the refcnt must check for ILL_CAN_LOOKUP. This ensures
4013 4013 * that the refcnt will drop down to zero.
4014 4014 */
4015 4015 mutex_enter(&ill->ill_lock);
4016 4016 ill->ill_state_flags |= ILL_CONDEMNED;
4017 4017 for (ipif = ill->ill_ipif; ipif != NULL;
4018 4018 ipif = ipif->ipif_next) {
4019 4019 ipif->ipif_state_flags |= IPIF_CONDEMNED;
4020 4020 }
4021 4021 /*
4022 4022 * Wake up anybody waiting to enter the ipsq. ipsq_enter
4023 4023 * returns error if ILL_CONDEMNED is set
4024 4024 */
4025 4025 cv_broadcast(&ill->ill_cv);
4026 4026 mutex_exit(&ill->ill_lock);
4027 4027
4028 4028 /*
4029 4029 * Send all the deferred DLPI messages downstream which came in
4030 4030 * during the small window right before ipsq_enter(). We do this
4031 4031 * without waiting for the ACKs because all the ACKs for M_PROTO
4032 4032 * messages are ignored in ip_rput() when ILL_CONDEMNED is set.
4033 4033 */
4034 4034 ill_dlpi_send_deferred(ill);
4035 4035
4036 4036 /*
4037 4037 * Shut down fragmentation reassembly.
4038 4038 * ill_frag_timer won't start a timer again.
4039 4039 * Now cancel any existing timer
4040 4040 */
4041 4041 (void) untimeout(ill->ill_frag_timer_id);
4042 4042 (void) ill_frag_timeout(ill, 0);
4043 4043
4044 4044 /*
4045 4045 * Call ill_delete to bring down the ipifs, ilms and ill on
4046 4046 * this ill. Then wait for the refcnts to drop to zero.
4047 4047 * ill_is_freeable checks whether the ill is really quiescent.
4048 4048 * Then make sure that threads that are waiting to enter the
4049 4049 * ipsq have seen the error returned by ipsq_enter and have
4050 4050 * gone away. Then we call ill_delete_tail which does the
4051 4051 * DL_UNBIND_REQ with the driver and then qprocsoff.
4052 4052 */
4053 4053 ill_delete(ill);
4054 4054 mutex_enter(&ill->ill_lock);
4055 4055 while (!ill_is_freeable(ill))
4056 4056 cv_wait(&ill->ill_cv, &ill->ill_lock);
4057 4057
4058 4058 while (ill->ill_waiters)
4059 4059 cv_wait(&ill->ill_cv, &ill->ill_lock);
4060 4060
4061 4061 mutex_exit(&ill->ill_lock);
4062 4062
4063 4063 /*
4064 4064 * ill_delete_tail drops reference on ill_ipst, but we need to keep
4065 4065 * it held until the end of the function since the cleanup
4066 4066 * below needs to be able to use the ip_stack_t.
4067 4067 */
4068 4068 netstack_hold(ipst->ips_netstack);
4069 4069
4070 4070 /* qprocsoff is done via ill_delete_tail */
4071 4071 ill_delete_tail(ill);
4072 4072 /*
4073 4073 * synchronously wait for arp stream to unbind. After this, we
4074 4074 * cannot get any data packets up from the driver.
4075 4075 */
4076 4076 arp_unbind_complete(ill);
4077 4077 ASSERT(ill->ill_ipst == NULL);
4078 4078
4079 4079 /*
4080 4080 * Walk through all conns and qenable those that have queued data.
4081 4081 * Close synchronization needs this to
4082 4082 * be done to ensure that all upper layers blocked
4083 4083 * due to flow control to the closing device
4084 4084 * get unblocked.
4085 4085 */
4086 4086 ip1dbg(("ip_wsrv: walking\n"));
4087 4087 for (i = 0; i < TX_FANOUT_SIZE; i++) {
4088 4088 conn_walk_drain(ipst, &ipst->ips_idl_tx_list[i]);
4089 4089 }
4090 4090
4091 4091 /*
4092 4092 * ai can be null if this is an IPv6 ill, or if the IPv4
4093 4093 * stream is being torn down before ARP was plumbed (e.g.,
4094 4094 * /sbin/ifconfig plumbing a stream twice, and encountering
4095 4095 * an error
4096 4096 */
4097 4097 if (ai != NULL) {
4098 4098 ASSERT(!ill->ill_isv6);
4099 4099 mutex_enter(&ai->ai_lock);
4100 4100 ai->ai_ill = NULL;
4101 4101 if (ai->ai_arl == NULL) {
4102 4102 mutex_destroy(&ai->ai_lock);
4103 4103 kmem_free(ai, sizeof (*ai));
4104 4104 } else {
4105 4105 cv_signal(&ai->ai_ill_unplumb_done);
4106 4106 mutex_exit(&ai->ai_lock);
4107 4107 }
4108 4108 }
4109 4109
4110 4110 mutex_enter(&ipst->ips_ip_mi_lock);
4111 4111 mi_close_unlink(&ipst->ips_ip_g_head, (IDP)ill);
4112 4112 mutex_exit(&ipst->ips_ip_mi_lock);
4113 4113
4114 4114 /*
4115 4115 * credp could be null if the open didn't succeed and ip_modopen
4116 4116 * itself calls ip_close.
4117 4117 */
4118 4118 if (ill->ill_credp != NULL)
4119 4119 crfree(ill->ill_credp);
4120 4120
4121 4121 mutex_destroy(&ill->ill_saved_ire_lock);
4122 4122 mutex_destroy(&ill->ill_lock);
4123 4123 rw_destroy(&ill->ill_mcast_lock);
4124 4124 mutex_destroy(&ill->ill_mcast_serializer);
4125 4125 list_destroy(&ill->ill_nce);
4126 4126
4127 4127 /*
4128 4128 * Now we are done with the module close pieces that
4129 4129 * need the netstack_t.
4130 4130 */
4131 4131 netstack_rele(ipst->ips_netstack);
4132 4132
4133 4133 mi_close_free((IDP)ill);
4134 4134 q->q_ptr = WR(q)->q_ptr = NULL;
4135 4135
4136 4136 ipsq_exit(ipsq);
4137 4137
4138 4138 return (0);
4139 4139 }
4140 4140
4141 4141 /*
4142 4142 * This is called as part of close() for IP, UDP, ICMP, and RTS
4143 4143 * in order to quiesce the conn.
4144 4144 */
4145 4145 void
4146 4146 ip_quiesce_conn(conn_t *connp)
4147 4147 {
4148 4148 boolean_t drain_cleanup_reqd = B_FALSE;
4149 4149 boolean_t conn_ioctl_cleanup_reqd = B_FALSE;
4150 4150 boolean_t ilg_cleanup_reqd = B_FALSE;
4151 4151 ip_stack_t *ipst;
4152 4152
4153 4153 ASSERT(!IPCL_IS_TCP(connp));
4154 4154 ipst = connp->conn_netstack->netstack_ip;
4155 4155
4156 4156 /*
4157 4157 * Mark the conn as closing, and this conn must not be
4158 4158 * inserted in future into any list. Eg. conn_drain_insert(),
4159 4159 * won't insert this conn into the conn_drain_list.
4160 4160 *
4161 4161 * conn_idl, and conn_ilg cannot get set henceforth.
4162 4162 */
4163 4163 mutex_enter(&connp->conn_lock);
4164 4164 ASSERT(!(connp->conn_state_flags & CONN_QUIESCED));
4165 4165 connp->conn_state_flags |= CONN_CLOSING;
4166 4166 if (connp->conn_idl != NULL)
4167 4167 drain_cleanup_reqd = B_TRUE;
4168 4168 if (connp->conn_oper_pending_ill != NULL)
4169 4169 conn_ioctl_cleanup_reqd = B_TRUE;
4170 4170 if (connp->conn_dhcpinit_ill != NULL) {
4171 4171 ASSERT(connp->conn_dhcpinit_ill->ill_dhcpinit != 0);
4172 4172 atomic_dec_32(&connp->conn_dhcpinit_ill->ill_dhcpinit);
4173 4173 ill_set_inputfn(connp->conn_dhcpinit_ill);
4174 4174 connp->conn_dhcpinit_ill = NULL;
4175 4175 }
4176 4176 if (connp->conn_ilg != NULL)
4177 4177 ilg_cleanup_reqd = B_TRUE;
4178 4178 mutex_exit(&connp->conn_lock);
4179 4179
4180 4180 if (conn_ioctl_cleanup_reqd)
4181 4181 conn_ioctl_cleanup(connp);
4182 4182
4183 4183 if (is_system_labeled() && connp->conn_anon_port) {
4184 4184 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
4185 4185 connp->conn_mlp_type, connp->conn_proto,
4186 4186 ntohs(connp->conn_lport), B_FALSE);
4187 4187 connp->conn_anon_port = 0;
4188 4188 }
4189 4189 connp->conn_mlp_type = mlptSingle;
4190 4190
4191 4191 /*
4192 4192 * Remove this conn from any fanout list it is on.
4193 4193 * and then wait for any threads currently operating
4194 4194 * on this endpoint to finish
4195 4195 */
4196 4196 ipcl_hash_remove(connp);
4197 4197
4198 4198 /*
4199 4199 * Remove this conn from the drain list, and do any other cleanup that
4200 4200 * may be required. (TCP conns are never flow controlled, and
4201 4201 * conn_idl will be NULL.)
4202 4202 */
4203 4203 if (drain_cleanup_reqd && connp->conn_idl != NULL) {
4204 4204 idl_t *idl = connp->conn_idl;
4205 4205
4206 4206 mutex_enter(&idl->idl_lock);
4207 4207 conn_drain(connp, B_TRUE);
4208 4208 mutex_exit(&idl->idl_lock);
4209 4209 }
4210 4210
4211 4211 if (connp == ipst->ips_ip_g_mrouter)
4212 4212 (void) ip_mrouter_done(ipst);
4213 4213
4214 4214 if (ilg_cleanup_reqd)
4215 4215 ilg_delete_all(connp);
4216 4216
4217 4217 /*
4218 4218 * Now conn refcnt can increase only thru CONN_INC_REF_LOCKED.
4219 4219 * callers from write side can't be there now because close
4220 4220 * is in progress. The only other caller is ipcl_walk
4221 4221 * which checks for the condemned flag.
4222 4222 */
4223 4223 mutex_enter(&connp->conn_lock);
4224 4224 connp->conn_state_flags |= CONN_CONDEMNED;
4225 4225 while (connp->conn_ref != 1)
4226 4226 cv_wait(&connp->conn_cv, &connp->conn_lock);
4227 4227 connp->conn_state_flags |= CONN_QUIESCED;
4228 4228 mutex_exit(&connp->conn_lock);
4229 4229 }
4230 4230
4231 4231 /* ARGSUSED */
4232 4232 int
4233 4233 ip_close(queue_t *q, int flags, cred_t *credp __unused)
4234 4234 {
4235 4235 conn_t *connp;
4236 4236
4237 4237 /*
4238 4238 * Call the appropriate delete routine depending on whether this is
4239 4239 * a module or device.
4240 4240 */
4241 4241 if (WR(q)->q_next != NULL) {
4242 4242 /* This is a module close */
4243 4243 return (ip_modclose((ill_t *)q->q_ptr));
4244 4244 }
4245 4245
4246 4246 connp = q->q_ptr;
4247 4247 ip_quiesce_conn(connp);
4248 4248
4249 4249 qprocsoff(q);
4250 4250
4251 4251 /*
4252 4252 * Now we are truly single threaded on this stream, and can
4253 4253 * delete the things hanging off the connp, and finally the connp.
4254 4254 * We removed this connp from the fanout list, it cannot be
4255 4255 * accessed thru the fanouts, and we already waited for the
4256 4256 * conn_ref to drop to 0. We are already in close, so
4257 4257 * there cannot be any other thread from the top. qprocsoff
4258 4258 * has completed, and service has completed or won't run in
4259 4259 * future.
4260 4260 */
4261 4261 ASSERT(connp->conn_ref == 1);
4262 4262
4263 4263 inet_minor_free(connp->conn_minor_arena, connp->conn_dev);
4264 4264
4265 4265 connp->conn_ref--;
4266 4266 ipcl_conn_destroy(connp);
4267 4267
4268 4268 q->q_ptr = WR(q)->q_ptr = NULL;
4269 4269 return (0);
4270 4270 }
4271 4271
4272 4272 /*
4273 4273 * Wapper around putnext() so that ip_rts_request can merely use
4274 4274 * conn_recv.
4275 4275 */
4276 4276 /*ARGSUSED2*/
4277 4277 static void
4278 4278 ip_conn_input(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira)
4279 4279 {
4280 4280 conn_t *connp = (conn_t *)arg1;
4281 4281
4282 4282 putnext(connp->conn_rq, mp);
4283 4283 }
4284 4284
4285 4285 /* Dummy in case ICMP error delivery is attempted to a /dev/ip instance */
4286 4286 /* ARGSUSED */
4287 4287 static void
4288 4288 ip_conn_input_icmp(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira)
4289 4289 {
4290 4290 freemsg(mp);
4291 4291 }
4292 4292
4293 4293 /*
4294 4294 * Called when the module is about to be unloaded
4295 4295 */
4296 4296 void
4297 4297 ip_ddi_destroy(void)
4298 4298 {
4299 4299 /* This needs to be called before destroying any transports. */
4300 4300 mutex_enter(&cpu_lock);
4301 4301 unregister_cpu_setup_func(ip_tp_cpu_update, NULL);
4302 4302 mutex_exit(&cpu_lock);
4303 4303
4304 4304 tnet_fini();
4305 4305
4306 4306 icmp_ddi_g_destroy();
4307 4307 rts_ddi_g_destroy();
4308 4308 udp_ddi_g_destroy();
4309 4309 sctp_ddi_g_destroy();
4310 4310 tcp_ddi_g_destroy();
4311 4311 ilb_ddi_g_destroy();
4312 4312 dce_g_destroy();
4313 4313 ipsec_policy_g_destroy();
4314 4314 ipcl_g_destroy();
4315 4315 ip_net_g_destroy();
4316 4316 ip_ire_g_fini();
4317 4317 inet_minor_destroy(ip_minor_arena_sa);
4318 4318 #if defined(_LP64)
4319 4319 inet_minor_destroy(ip_minor_arena_la);
4320 4320 #endif
4321 4321
4322 4322 #ifdef DEBUG
4323 4323 list_destroy(&ip_thread_list);
4324 4324 rw_destroy(&ip_thread_rwlock);
4325 4325 tsd_destroy(&ip_thread_data);
4326 4326 #endif
4327 4327
4328 4328 netstack_unregister(NS_IP);
4329 4329 }
4330 4330
4331 4331 /*
4332 4332 * First step in cleanup.
4333 4333 */
4334 4334 /* ARGSUSED */
4335 4335 static void
4336 4336 ip_stack_shutdown(netstackid_t stackid, void *arg)
4337 4337 {
4338 4338 ip_stack_t *ipst = (ip_stack_t *)arg;
4339 4339 kt_did_t ktid;
4340 4340
4341 4341 #ifdef NS_DEBUG
4342 4342 printf("ip_stack_shutdown(%p, stack %d)\n", (void *)ipst, stackid);
4343 4343 #endif
4344 4344
4345 4345 /*
4346 4346 * Perform cleanup for special interfaces (loopback and IPMP).
4347 4347 */
4348 4348 ip_interface_cleanup(ipst);
4349 4349
4350 4350 /*
4351 4351 * The *_hook_shutdown()s start the process of notifying any
4352 4352 * consumers that things are going away.... nothing is destroyed.
4353 4353 */
4354 4354 ipv4_hook_shutdown(ipst);
4355 4355 ipv6_hook_shutdown(ipst);
4356 4356 arp_hook_shutdown(ipst);
4357 4357
4358 4358 mutex_enter(&ipst->ips_capab_taskq_lock);
4359 4359 ktid = ipst->ips_capab_taskq_thread->t_did;
4360 4360 ipst->ips_capab_taskq_quit = B_TRUE;
4361 4361 cv_signal(&ipst->ips_capab_taskq_cv);
4362 4362 mutex_exit(&ipst->ips_capab_taskq_lock);
4363 4363
4364 4364 /*
4365 4365 * In rare occurrences, particularly on virtual hardware where CPUs can
4366 4366 * be de-scheduled, the thread that we just signaled will not run until
4367 4367 * after we have gotten through parts of ip_stack_fini. If that happens
4368 4368 * then we'll try to grab the ips_capab_taskq_lock as part of returning
4369 4369 * from cv_wait which no longer exists.
4370 4370 */
4371 4371 thread_join(ktid);
4372 4372 }
4373 4373
4374 4374 /*
4375 4375 * Free the IP stack instance.
4376 4376 */
4377 4377 static void
4378 4378 ip_stack_fini(netstackid_t stackid, void *arg)
4379 4379 {
4380 4380 ip_stack_t *ipst = (ip_stack_t *)arg;
4381 4381 int ret;
4382 4382
4383 4383 #ifdef NS_DEBUG
4384 4384 printf("ip_stack_fini(%p, stack %d)\n", (void *)ipst, stackid);
4385 4385 #endif
4386 4386 /*
4387 4387 * At this point, all of the notifications that the events and
4388 4388 * protocols are going away have been run, meaning that we can
4389 4389 * now set about starting to clean things up.
4390 4390 */
4391 4391 ipobs_fini(ipst);
4392 4392 ipv4_hook_destroy(ipst);
4393 4393 ipv6_hook_destroy(ipst);
4394 4394 arp_hook_destroy(ipst);
4395 4395 ip_net_destroy(ipst);
4396 4396
4397 4397 ipmp_destroy(ipst);
4398 4398
4399 4399 ip_kstat_fini(stackid, ipst->ips_ip_mibkp);
4400 4400 ipst->ips_ip_mibkp = NULL;
4401 4401 icmp_kstat_fini(stackid, ipst->ips_icmp_mibkp);
4402 4402 ipst->ips_icmp_mibkp = NULL;
4403 4403 ip_kstat2_fini(stackid, ipst->ips_ip_kstat);
4404 4404 ipst->ips_ip_kstat = NULL;
4405 4405 bzero(&ipst->ips_ip_statistics, sizeof (ipst->ips_ip_statistics));
4406 4406 ip6_kstat_fini(stackid, ipst->ips_ip6_kstat);
4407 4407 ipst->ips_ip6_kstat = NULL;
4408 4408 bzero(&ipst->ips_ip6_statistics, sizeof (ipst->ips_ip6_statistics));
4409 4409
4410 4410 kmem_free(ipst->ips_propinfo_tbl,
4411 4411 ip_propinfo_count * sizeof (mod_prop_info_t));
4412 4412 ipst->ips_propinfo_tbl = NULL;
4413 4413
4414 4414 dce_stack_destroy(ipst);
4415 4415 ip_mrouter_stack_destroy(ipst);
4416 4416
4417 4417 /*
4418 4418 * Quiesce all of our timers. Note we set the quiesce flags before we
4419 4419 * call untimeout. The slowtimers may actually kick off another instance
4420 4420 * of the non-slow timers.
4421 4421 */
4422 4422 mutex_enter(&ipst->ips_igmp_timer_lock);
4423 4423 ipst->ips_igmp_timer_quiesce = B_TRUE;
4424 4424 mutex_exit(&ipst->ips_igmp_timer_lock);
4425 4425
4426 4426 mutex_enter(&ipst->ips_mld_timer_lock);
4427 4427 ipst->ips_mld_timer_quiesce = B_TRUE;
4428 4428 mutex_exit(&ipst->ips_mld_timer_lock);
4429 4429
4430 4430 mutex_enter(&ipst->ips_igmp_slowtimeout_lock);
4431 4431 ipst->ips_igmp_slowtimeout_quiesce = B_TRUE;
4432 4432 mutex_exit(&ipst->ips_igmp_slowtimeout_lock);
4433 4433
4434 4434 mutex_enter(&ipst->ips_mld_slowtimeout_lock);
4435 4435 ipst->ips_mld_slowtimeout_quiesce = B_TRUE;
4436 4436 mutex_exit(&ipst->ips_mld_slowtimeout_lock);
4437 4437
4438 4438 ret = untimeout(ipst->ips_igmp_timeout_id);
4439 4439 if (ret == -1) {
4440 4440 ASSERT(ipst->ips_igmp_timeout_id == 0);
4441 4441 } else {
4442 4442 ASSERT(ipst->ips_igmp_timeout_id != 0);
4443 4443 ipst->ips_igmp_timeout_id = 0;
4444 4444 }
4445 4445 ret = untimeout(ipst->ips_igmp_slowtimeout_id);
4446 4446 if (ret == -1) {
4447 4447 ASSERT(ipst->ips_igmp_slowtimeout_id == 0);
4448 4448 } else {
4449 4449 ASSERT(ipst->ips_igmp_slowtimeout_id != 0);
4450 4450 ipst->ips_igmp_slowtimeout_id = 0;
4451 4451 }
4452 4452 ret = untimeout(ipst->ips_mld_timeout_id);
4453 4453 if (ret == -1) {
4454 4454 ASSERT(ipst->ips_mld_timeout_id == 0);
4455 4455 } else {
4456 4456 ASSERT(ipst->ips_mld_timeout_id != 0);
4457 4457 ipst->ips_mld_timeout_id = 0;
4458 4458 }
4459 4459 ret = untimeout(ipst->ips_mld_slowtimeout_id);
4460 4460 if (ret == -1) {
4461 4461 ASSERT(ipst->ips_mld_slowtimeout_id == 0);
4462 4462 } else {
4463 4463 ASSERT(ipst->ips_mld_slowtimeout_id != 0);
4464 4464 ipst->ips_mld_slowtimeout_id = 0;
4465 4465 }
4466 4466
4467 4467 ip_ire_fini(ipst);
4468 4468 ip6_asp_free(ipst);
4469 4469 conn_drain_fini(ipst);
4470 4470 ipcl_destroy(ipst);
4471 4471
4472 4472 mutex_destroy(&ipst->ips_ndp4->ndp_g_lock);
4473 4473 mutex_destroy(&ipst->ips_ndp6->ndp_g_lock);
4474 4474 kmem_free(ipst->ips_ndp4, sizeof (ndp_g_t));
4475 4475 ipst->ips_ndp4 = NULL;
4476 4476 kmem_free(ipst->ips_ndp6, sizeof (ndp_g_t));
4477 4477 ipst->ips_ndp6 = NULL;
4478 4478
4479 4479 if (ipst->ips_loopback_ksp != NULL) {
4480 4480 kstat_delete_netstack(ipst->ips_loopback_ksp, stackid);
4481 4481 ipst->ips_loopback_ksp = NULL;
4482 4482 }
4483 4483
4484 4484 mutex_destroy(&ipst->ips_capab_taskq_lock);
4485 4485 cv_destroy(&ipst->ips_capab_taskq_cv);
4486 4486
4487 4487 rw_destroy(&ipst->ips_srcid_lock);
4488 4488
4489 4489 mutex_destroy(&ipst->ips_ip_mi_lock);
4490 4490 rw_destroy(&ipst->ips_ill_g_usesrc_lock);
4491 4491
4492 4492 mutex_destroy(&ipst->ips_igmp_timer_lock);
4493 4493 mutex_destroy(&ipst->ips_mld_timer_lock);
4494 4494 mutex_destroy(&ipst->ips_igmp_slowtimeout_lock);
4495 4495 mutex_destroy(&ipst->ips_mld_slowtimeout_lock);
4496 4496 mutex_destroy(&ipst->ips_ip_addr_avail_lock);
4497 4497 rw_destroy(&ipst->ips_ill_g_lock);
4498 4498
4499 4499 kmem_free(ipst->ips_phyint_g_list, sizeof (phyint_list_t));
4500 4500 ipst->ips_phyint_g_list = NULL;
4501 4501 kmem_free(ipst->ips_ill_g_heads, sizeof (ill_g_head_t) * MAX_G_HEADS);
4502 4502 ipst->ips_ill_g_heads = NULL;
4503 4503
4504 4504 ldi_ident_release(ipst->ips_ldi_ident);
4505 4505 kmem_free(ipst, sizeof (*ipst));
4506 4506 }
4507 4507
4508 4508 /*
4509 4509 * This function is called from the TSD destructor, and is used to debug
4510 4510 * reference count issues in IP. See block comment in <inet/ip_if.h> for
4511 4511 * details.
4512 4512 */
4513 4513 static void
4514 4514 ip_thread_exit(void *phash)
4515 4515 {
4516 4516 th_hash_t *thh = phash;
4517 4517
4518 4518 rw_enter(&ip_thread_rwlock, RW_WRITER);
4519 4519 list_remove(&ip_thread_list, thh);
4520 4520 rw_exit(&ip_thread_rwlock);
4521 4521 mod_hash_destroy_hash(thh->thh_hash);
4522 4522 kmem_free(thh, sizeof (*thh));
4523 4523 }
4524 4524
4525 4525 /*
4526 4526 * Called when the IP kernel module is loaded into the kernel
4527 4527 */
4528 4528 void
4529 4529 ip_ddi_init(void)
4530 4530 {
4531 4531 ip_squeue_flag = ip_squeue_switch(ip_squeue_enter);
4532 4532
4533 4533 /*
4534 4534 * For IP and TCP the minor numbers should start from 2 since we have 4
4535 4535 * initial devices: ip, ip6, tcp, tcp6.
4536 4536 */
4537 4537 /*
4538 4538 * If this is a 64-bit kernel, then create two separate arenas -
4539 4539 * one for TLIs in the range of INET_MIN_DEV+2 through 2^^18-1, and the
4540 4540 * other for socket apps in the range 2^^18 through 2^^32-1.
4541 4541 */
4542 4542 ip_minor_arena_la = NULL;
4543 4543 ip_minor_arena_sa = NULL;
4544 4544 #if defined(_LP64)
4545 4545 if ((ip_minor_arena_sa = inet_minor_create("ip_minor_arena_sa",
4546 4546 INET_MIN_DEV + 2, MAXMIN32, KM_SLEEP)) == NULL) {
4547 4547 cmn_err(CE_PANIC,
4548 4548 "ip_ddi_init: ip_minor_arena_sa creation failed\n");
4549 4549 }
4550 4550 if ((ip_minor_arena_la = inet_minor_create("ip_minor_arena_la",
4551 4551 MAXMIN32 + 1, MAXMIN64, KM_SLEEP)) == NULL) {
4552 4552 cmn_err(CE_PANIC,
4553 4553 "ip_ddi_init: ip_minor_arena_la creation failed\n");
4554 4554 }
4555 4555 #else
4556 4556 if ((ip_minor_arena_sa = inet_minor_create("ip_minor_arena_sa",
4557 4557 INET_MIN_DEV + 2, MAXMIN, KM_SLEEP)) == NULL) {
4558 4558 cmn_err(CE_PANIC,
4559 4559 "ip_ddi_init: ip_minor_arena_sa creation failed\n");
4560 4560 }
4561 4561 #endif
4562 4562 ip_poll_normal_ticks = MSEC_TO_TICK_ROUNDUP(ip_poll_normal_ms);
4563 4563
4564 4564 ipcl_g_init();
4565 4565 ip_ire_g_init();
4566 4566 ip_net_g_init();
4567 4567
4568 4568 #ifdef DEBUG
4569 4569 tsd_create(&ip_thread_data, ip_thread_exit);
4570 4570 rw_init(&ip_thread_rwlock, NULL, RW_DEFAULT, NULL);
4571 4571 list_create(&ip_thread_list, sizeof (th_hash_t),
4572 4572 offsetof(th_hash_t, thh_link));
4573 4573 #endif
4574 4574 ipsec_policy_g_init();
4575 4575 tcp_ddi_g_init();
4576 4576 sctp_ddi_g_init();
4577 4577 dce_g_init();
4578 4578
4579 4579 /*
4580 4580 * We want to be informed each time a stack is created or
4581 4581 * destroyed in the kernel, so we can maintain the
4582 4582 * set of udp_stack_t's.
4583 4583 */
4584 4584 netstack_register(NS_IP, ip_stack_init, ip_stack_shutdown,
4585 4585 ip_stack_fini);
4586 4586
4587 4587 tnet_init();
4588 4588
4589 4589 udp_ddi_g_init();
4590 4590 rts_ddi_g_init();
4591 4591 icmp_ddi_g_init();
4592 4592 ilb_ddi_g_init();
4593 4593
4594 4594 /* This needs to be called after all transports are initialized. */
4595 4595 mutex_enter(&cpu_lock);
4596 4596 register_cpu_setup_func(ip_tp_cpu_update, NULL);
4597 4597 mutex_exit(&cpu_lock);
4598 4598 }
4599 4599
4600 4600 /*
4601 4601 * Initialize the IP stack instance.
4602 4602 */
4603 4603 static void *
4604 4604 ip_stack_init(netstackid_t stackid, netstack_t *ns)
4605 4605 {
4606 4606 ip_stack_t *ipst;
4607 4607 size_t arrsz;
4608 4608 major_t major;
4609 4609
4610 4610 #ifdef NS_DEBUG
4611 4611 printf("ip_stack_init(stack %d)\n", stackid);
4612 4612 #endif
4613 4613
4614 4614 ipst = (ip_stack_t *)kmem_zalloc(sizeof (*ipst), KM_SLEEP);
4615 4615 ipst->ips_netstack = ns;
4616 4616
4617 4617 ipst->ips_ill_g_heads = kmem_zalloc(sizeof (ill_g_head_t) * MAX_G_HEADS,
4618 4618 KM_SLEEP);
4619 4619 ipst->ips_phyint_g_list = kmem_zalloc(sizeof (phyint_list_t),
4620 4620 KM_SLEEP);
4621 4621 ipst->ips_ndp4 = kmem_zalloc(sizeof (ndp_g_t), KM_SLEEP);
4622 4622 ipst->ips_ndp6 = kmem_zalloc(sizeof (ndp_g_t), KM_SLEEP);
4623 4623 mutex_init(&ipst->ips_ndp4->ndp_g_lock, NULL, MUTEX_DEFAULT, NULL);
4624 4624 mutex_init(&ipst->ips_ndp6->ndp_g_lock, NULL, MUTEX_DEFAULT, NULL);
4625 4625
4626 4626 mutex_init(&ipst->ips_igmp_timer_lock, NULL, MUTEX_DEFAULT, NULL);
4627 4627 ipst->ips_igmp_deferred_next = INFINITY;
4628 4628 mutex_init(&ipst->ips_mld_timer_lock, NULL, MUTEX_DEFAULT, NULL);
4629 4629 ipst->ips_mld_deferred_next = INFINITY;
4630 4630 mutex_init(&ipst->ips_igmp_slowtimeout_lock, NULL, MUTEX_DEFAULT, NULL);
4631 4631 mutex_init(&ipst->ips_mld_slowtimeout_lock, NULL, MUTEX_DEFAULT, NULL);
4632 4632 mutex_init(&ipst->ips_ip_mi_lock, NULL, MUTEX_DEFAULT, NULL);
4633 4633 mutex_init(&ipst->ips_ip_addr_avail_lock, NULL, MUTEX_DEFAULT, NULL);
4634 4634 rw_init(&ipst->ips_ill_g_lock, NULL, RW_DEFAULT, NULL);
4635 4635 rw_init(&ipst->ips_ill_g_usesrc_lock, NULL, RW_DEFAULT, NULL);
4636 4636
4637 4637 ipcl_init(ipst);
4638 4638 ip_ire_init(ipst);
4639 4639 ip6_asp_init(ipst);
4640 4640 ipif_init(ipst);
4641 4641 conn_drain_init(ipst);
4642 4642 ip_mrouter_stack_init(ipst);
4643 4643 dce_stack_init(ipst);
4644 4644
4645 4645 ipst->ips_ip_multirt_log_interval = 1000;
4646 4646
4647 4647 ipst->ips_ill_index = 1;
4648 4648
4649 4649 ipst->ips_saved_ip_forwarding = -1;
4650 4650 ipst->ips_reg_vif_num = ALL_VIFS; /* Index to Register vif */
4651 4651
4652 4652 arrsz = ip_propinfo_count * sizeof (mod_prop_info_t);
4653 4653 ipst->ips_propinfo_tbl = (mod_prop_info_t *)kmem_alloc(arrsz, KM_SLEEP);
4654 4654 bcopy(ip_propinfo_tbl, ipst->ips_propinfo_tbl, arrsz);
4655 4655
4656 4656 ipst->ips_ip_mibkp = ip_kstat_init(stackid, ipst);
4657 4657 ipst->ips_icmp_mibkp = icmp_kstat_init(stackid);
4658 4658 ipst->ips_ip_kstat = ip_kstat2_init(stackid, &ipst->ips_ip_statistics);
4659 4659 ipst->ips_ip6_kstat =
4660 4660 ip6_kstat_init(stackid, &ipst->ips_ip6_statistics);
4661 4661
4662 4662 ipst->ips_ip_src_id = 1;
4663 4663 rw_init(&ipst->ips_srcid_lock, NULL, RW_DEFAULT, NULL);
4664 4664
4665 4665 ipst->ips_src_generation = SRC_GENERATION_INITIAL;
4666 4666
4667 4667 ip_net_init(ipst, ns);
4668 4668 ipv4_hook_init(ipst);
4669 4669 ipv6_hook_init(ipst);
4670 4670 arp_hook_init(ipst);
4671 4671 ipmp_init(ipst);
4672 4672 ipobs_init(ipst);
4673 4673
4674 4674 /*
4675 4675 * Create the taskq dispatcher thread and initialize related stuff.
4676 4676 */
4677 4677 mutex_init(&ipst->ips_capab_taskq_lock, NULL, MUTEX_DEFAULT, NULL);
4678 4678 cv_init(&ipst->ips_capab_taskq_cv, NULL, CV_DEFAULT, NULL);
4679 4679 ipst->ips_capab_taskq_thread = thread_create(NULL, 0,
4680 4680 ill_taskq_dispatch, ipst, 0, &p0, TS_RUN, minclsyspri);
4681 4681
4682 4682 major = mod_name_to_major(INET_NAME);
4683 4683 (void) ldi_ident_from_major(major, &ipst->ips_ldi_ident);
4684 4684 return (ipst);
4685 4685 }
4686 4686
4687 4687 /*
4688 4688 * Allocate and initialize a DLPI template of the specified length. (May be
4689 4689 * called as writer.)
4690 4690 */
4691 4691 mblk_t *
4692 4692 ip_dlpi_alloc(size_t len, t_uscalar_t prim)
4693 4693 {
4694 4694 mblk_t *mp;
4695 4695
4696 4696 mp = allocb(len, BPRI_MED);
4697 4697 if (!mp)
4698 4698 return (NULL);
4699 4699
4700 4700 /*
4701 4701 * DLPIv2 says that DL_INFO_REQ and DL_TOKEN_REQ (the latter
4702 4702 * of which we don't seem to use) are sent with M_PCPROTO, and
4703 4703 * that other DLPI are M_PROTO.
4704 4704 */
4705 4705 if (prim == DL_INFO_REQ) {
4706 4706 mp->b_datap->db_type = M_PCPROTO;
4707 4707 } else {
4708 4708 mp->b_datap->db_type = M_PROTO;
4709 4709 }
4710 4710
4711 4711 mp->b_wptr = mp->b_rptr + len;
4712 4712 bzero(mp->b_rptr, len);
4713 4713 ((dl_unitdata_req_t *)mp->b_rptr)->dl_primitive = prim;
4714 4714 return (mp);
4715 4715 }
4716 4716
4717 4717 /*
4718 4718 * Allocate and initialize a DLPI notification. (May be called as writer.)
4719 4719 */
4720 4720 mblk_t *
4721 4721 ip_dlnotify_alloc(uint_t notification, uint_t data)
4722 4722 {
4723 4723 dl_notify_ind_t *notifyp;
4724 4724 mblk_t *mp;
4725 4725
4726 4726 if ((mp = ip_dlpi_alloc(DL_NOTIFY_IND_SIZE, DL_NOTIFY_IND)) == NULL)
4727 4727 return (NULL);
4728 4728
4729 4729 notifyp = (dl_notify_ind_t *)mp->b_rptr;
4730 4730 notifyp->dl_notification = notification;
4731 4731 notifyp->dl_data = data;
4732 4732 return (mp);
4733 4733 }
4734 4734
4735 4735 mblk_t *
4736 4736 ip_dlnotify_alloc2(uint_t notification, uint_t data1, uint_t data2)
4737 4737 {
4738 4738 dl_notify_ind_t *notifyp;
4739 4739 mblk_t *mp;
4740 4740
4741 4741 if ((mp = ip_dlpi_alloc(DL_NOTIFY_IND_SIZE, DL_NOTIFY_IND)) == NULL)
4742 4742 return (NULL);
4743 4743
4744 4744 notifyp = (dl_notify_ind_t *)mp->b_rptr;
4745 4745 notifyp->dl_notification = notification;
4746 4746 notifyp->dl_data1 = data1;
4747 4747 notifyp->dl_data2 = data2;
4748 4748 return (mp);
4749 4749 }
4750 4750
4751 4751 /*
4752 4752 * Debug formatting routine. Returns a character string representation of the
4753 4753 * addr in buf, of the form xxx.xxx.xxx.xxx. This routine takes the address
4754 4754 * in the form of a ipaddr_t and calls ip_dot_saddr with a pointer.
4755 4755 *
4756 4756 * Once the ndd table-printing interfaces are removed, this can be changed to
4757 4757 * standard dotted-decimal form.
4758 4758 */
4759 4759 char *
4760 4760 ip_dot_addr(ipaddr_t addr, char *buf)
4761 4761 {
4762 4762 uint8_t *ap = (uint8_t *)&addr;
4763 4763
4764 4764 (void) mi_sprintf(buf, "%03d.%03d.%03d.%03d",
4765 4765 ap[0] & 0xFF, ap[1] & 0xFF, ap[2] & 0xFF, ap[3] & 0xFF);
4766 4766 return (buf);
4767 4767 }
4768 4768
4769 4769 /*
4770 4770 * Write the given MAC address as a printable string in the usual colon-
4771 4771 * separated format.
4772 4772 */
4773 4773 const char *
4774 4774 mac_colon_addr(const uint8_t *addr, size_t alen, char *buf, size_t buflen)
4775 4775 {
4776 4776 char *bp;
4777 4777
4778 4778 if (alen == 0 || buflen < 4)
4779 4779 return ("?");
4780 4780 bp = buf;
4781 4781 for (;;) {
4782 4782 /*
4783 4783 * If there are more MAC address bytes available, but we won't
4784 4784 * have any room to print them, then add "..." to the string
4785 4785 * instead. See below for the 'magic number' explanation.
4786 4786 */
4787 4787 if ((alen == 2 && buflen < 6) || (alen > 2 && buflen < 7)) {
4788 4788 (void) strcpy(bp, "...");
4789 4789 break;
4790 4790 }
4791 4791 (void) sprintf(bp, "%02x", *addr++);
4792 4792 bp += 2;
4793 4793 if (--alen == 0)
4794 4794 break;
4795 4795 *bp++ = ':';
4796 4796 buflen -= 3;
4797 4797 /*
4798 4798 * At this point, based on the first 'if' statement above,
4799 4799 * either alen == 1 and buflen >= 3, or alen > 1 and
4800 4800 * buflen >= 4. The first case leaves room for the final "xx"
4801 4801 * number and trailing NUL byte. The second leaves room for at
4802 4802 * least "...". Thus the apparently 'magic' numbers chosen for
4803 4803 * that statement.
4804 4804 */
4805 4805 }
4806 4806 return (buf);
4807 4807 }
4808 4808
4809 4809 /*
4810 4810 * Called when it is conceptually a ULP that would sent the packet
4811 4811 * e.g., port unreachable and protocol unreachable. Check that the packet
4812 4812 * would have passed the IPsec global policy before sending the error.
4813 4813 *
4814 4814 * Send an ICMP error after patching up the packet appropriately.
4815 4815 * Uses ip_drop_input and bumps the appropriate MIB.
4816 4816 */
4817 4817 void
4818 4818 ip_fanout_send_icmp_v4(mblk_t *mp, uint_t icmp_type, uint_t icmp_code,
4819 4819 ip_recv_attr_t *ira)
4820 4820 {
4821 4821 ipha_t *ipha;
4822 4822 boolean_t secure;
4823 4823 ill_t *ill = ira->ira_ill;
4824 4824 ip_stack_t *ipst = ill->ill_ipst;
4825 4825 netstack_t *ns = ipst->ips_netstack;
4826 4826 ipsec_stack_t *ipss = ns->netstack_ipsec;
4827 4827
4828 4828 secure = ira->ira_flags & IRAF_IPSEC_SECURE;
4829 4829
4830 4830 /*
4831 4831 * We are generating an icmp error for some inbound packet.
4832 4832 * Called from all ip_fanout_(udp, tcp, proto) functions.
4833 4833 * Before we generate an error, check with global policy
4834 4834 * to see whether this is allowed to enter the system. As
4835 4835 * there is no "conn", we are checking with global policy.
4836 4836 */
4837 4837 ipha = (ipha_t *)mp->b_rptr;
4838 4838 if (secure || ipss->ipsec_inbound_v4_policy_present) {
4839 4839 mp = ipsec_check_global_policy(mp, NULL, ipha, NULL, ira, ns);
4840 4840 if (mp == NULL)
4841 4841 return;
4842 4842 }
4843 4843
4844 4844 /* We never send errors for protocols that we do implement */
4845 4845 if (ira->ira_protocol == IPPROTO_ICMP ||
4846 4846 ira->ira_protocol == IPPROTO_IGMP) {
4847 4847 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
4848 4848 ip_drop_input("ip_fanout_send_icmp_v4", mp, ill);
4849 4849 freemsg(mp);
4850 4850 return;
4851 4851 }
4852 4852 /*
4853 4853 * Have to correct checksum since
4854 4854 * the packet might have been
4855 4855 * fragmented and the reassembly code in ip_rput
4856 4856 * does not restore the IP checksum.
4857 4857 */
4858 4858 ipha->ipha_hdr_checksum = 0;
4859 4859 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
4860 4860
4861 4861 switch (icmp_type) {
4862 4862 case ICMP_DEST_UNREACHABLE:
4863 4863 switch (icmp_code) {
4864 4864 case ICMP_PROTOCOL_UNREACHABLE:
4865 4865 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInUnknownProtos);
4866 4866 ip_drop_input("ipIfStatsInUnknownProtos", mp, ill);
4867 4867 break;
4868 4868 case ICMP_PORT_UNREACHABLE:
4869 4869 BUMP_MIB(ill->ill_ip_mib, udpIfStatsNoPorts);
4870 4870 ip_drop_input("ipIfStatsNoPorts", mp, ill);
4871 4871 break;
4872 4872 }
4873 4873
4874 4874 icmp_unreachable(mp, icmp_code, ira);
4875 4875 break;
4876 4876 default:
4877 4877 #ifdef DEBUG
4878 4878 panic("ip_fanout_send_icmp_v4: wrong type");
4879 4879 /*NOTREACHED*/
4880 4880 #else
4881 4881 freemsg(mp);
4882 4882 break;
4883 4883 #endif
4884 4884 }
4885 4885 }
4886 4886
4887 4887 /*
4888 4888 * Used to send an ICMP error message when a packet is received for
4889 4889 * a protocol that is not supported. The mblk passed as argument
4890 4890 * is consumed by this function.
4891 4891 */
4892 4892 void
4893 4893 ip_proto_not_sup(mblk_t *mp, ip_recv_attr_t *ira)
4894 4894 {
4895 4895 ipha_t *ipha;
4896 4896
4897 4897 ipha = (ipha_t *)mp->b_rptr;
4898 4898 if (ira->ira_flags & IRAF_IS_IPV4) {
4899 4899 ASSERT(IPH_HDR_VERSION(ipha) == IP_VERSION);
4900 4900 ip_fanout_send_icmp_v4(mp, ICMP_DEST_UNREACHABLE,
4901 4901 ICMP_PROTOCOL_UNREACHABLE, ira);
4902 4902 } else {
4903 4903 ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
4904 4904 ip_fanout_send_icmp_v6(mp, ICMP6_PARAM_PROB,
4905 4905 ICMP6_PARAMPROB_NEXTHEADER, ira);
4906 4906 }
4907 4907 }
4908 4908
4909 4909 /*
4910 4910 * Deliver a rawip packet to the given conn, possibly applying ipsec policy.
4911 4911 * Handles IPv4 and IPv6.
4912 4912 * We are responsible for disposing of mp, such as by freemsg() or putnext()
4913 4913 * Caller is responsible for dropping references to the conn.
4914 4914 */
4915 4915 void
4916 4916 ip_fanout_proto_conn(conn_t *connp, mblk_t *mp, ipha_t *ipha, ip6_t *ip6h,
4917 4917 ip_recv_attr_t *ira)
4918 4918 {
4919 4919 ill_t *ill = ira->ira_ill;
4920 4920 ip_stack_t *ipst = ill->ill_ipst;
4921 4921 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
4922 4922 boolean_t secure;
4923 4923 uint_t protocol = ira->ira_protocol;
4924 4924 iaflags_t iraflags = ira->ira_flags;
4925 4925 queue_t *rq;
4926 4926
4927 4927 secure = iraflags & IRAF_IPSEC_SECURE;
4928 4928
4929 4929 rq = connp->conn_rq;
4930 4930 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld : !canputnext(rq)) {
4931 4931 switch (protocol) {
4932 4932 case IPPROTO_ICMPV6:
4933 4933 BUMP_MIB(ill->ill_icmp6_mib, ipv6IfIcmpInOverflows);
4934 4934 break;
4935 4935 case IPPROTO_ICMP:
4936 4936 BUMP_MIB(&ipst->ips_icmp_mib, icmpInOverflows);
4937 4937 break;
4938 4938 default:
4939 4939 BUMP_MIB(ill->ill_ip_mib, rawipIfStatsInOverflows);
4940 4940 break;
4941 4941 }
4942 4942 freemsg(mp);
4943 4943 return;
4944 4944 }
4945 4945
4946 4946 ASSERT(!(IPCL_IS_IPTUN(connp)));
4947 4947
4948 4948 if (((iraflags & IRAF_IS_IPV4) ?
4949 4949 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
4950 4950 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
4951 4951 secure) {
4952 4952 mp = ipsec_check_inbound_policy(mp, connp, ipha,
4953 4953 ip6h, ira);
4954 4954 if (mp == NULL) {
4955 4955 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
4956 4956 /* Note that mp is NULL */
4957 4957 ip_drop_input("ipIfStatsInDiscards", mp, ill);
4958 4958 return;
4959 4959 }
4960 4960 }
4961 4961
4962 4962 if (iraflags & IRAF_ICMP_ERROR) {
4963 4963 (connp->conn_recvicmp)(connp, mp, NULL, ira);
4964 4964 } else {
4965 4965 ill_t *rill = ira->ira_rill;
4966 4966
4967 4967 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
4968 4968 ira->ira_ill = ira->ira_rill = NULL;
4969 4969 /* Send it upstream */
4970 4970 (connp->conn_recv)(connp, mp, NULL, ira);
4971 4971 ira->ira_ill = ill;
4972 4972 ira->ira_rill = rill;
4973 4973 }
4974 4974 }
4975 4975
4976 4976 /*
4977 4977 * Handle protocols with which IP is less intimate. There
4978 4978 * can be more than one stream bound to a particular
4979 4979 * protocol. When this is the case, normally each one gets a copy
4980 4980 * of any incoming packets.
4981 4981 *
4982 4982 * IPsec NOTE :
4983 4983 *
4984 4984 * Don't allow a secure packet going up a non-secure connection.
4985 4985 * We don't allow this because
4986 4986 *
4987 4987 * 1) Reply might go out in clear which will be dropped at
4988 4988 * the sending side.
4989 4989 * 2) If the reply goes out in clear it will give the
4990 4990 * adversary enough information for getting the key in
4991 4991 * most of the cases.
4992 4992 *
4993 4993 * Moreover getting a secure packet when we expect clear
4994 4994 * implies that SA's were added without checking for
4995 4995 * policy on both ends. This should not happen once ISAKMP
4996 4996 * is used to negotiate SAs as SAs will be added only after
4997 4997 * verifying the policy.
4998 4998 *
4999 4999 * Zones notes:
5000 5000 * Earlier in ip_input on a system with multiple shared-IP zones we
5001 5001 * duplicate the multicast and broadcast packets and send them up
5002 5002 * with each explicit zoneid that exists on that ill.
5003 5003 * This means that here we can match the zoneid with SO_ALLZONES being special.
5004 5004 */
5005 5005 void
5006 5006 ip_fanout_proto_v4(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
5007 5007 {
5008 5008 mblk_t *mp1;
5009 5009 ipaddr_t laddr;
5010 5010 conn_t *connp, *first_connp, *next_connp;
5011 5011 connf_t *connfp;
5012 5012 ill_t *ill = ira->ira_ill;
5013 5013 ip_stack_t *ipst = ill->ill_ipst;
5014 5014
5015 5015 laddr = ipha->ipha_dst;
5016 5016
5017 5017 connfp = &ipst->ips_ipcl_proto_fanout_v4[ira->ira_protocol];
5018 5018 mutex_enter(&connfp->connf_lock);
5019 5019 connp = connfp->connf_head;
5020 5020 for (connp = connfp->connf_head; connp != NULL;
5021 5021 connp = connp->conn_next) {
5022 5022 /* Note: IPCL_PROTO_MATCH includes conn_wantpacket */
5023 5023 if (IPCL_PROTO_MATCH(connp, ira, ipha) &&
5024 5024 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5025 5025 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp))) {
5026 5026 break;
5027 5027 }
5028 5028 }
5029 5029
5030 5030 if (connp == NULL) {
5031 5031 /*
5032 5032 * No one bound to these addresses. Is
5033 5033 * there a client that wants all
5034 5034 * unclaimed datagrams?
5035 5035 */
5036 5036 mutex_exit(&connfp->connf_lock);
5037 5037 ip_fanout_send_icmp_v4(mp, ICMP_DEST_UNREACHABLE,
5038 5038 ICMP_PROTOCOL_UNREACHABLE, ira);
5039 5039 return;
5040 5040 }
5041 5041
5042 5042 ASSERT(IPCL_IS_NONSTR(connp) || connp->conn_rq != NULL);
5043 5043
5044 5044 CONN_INC_REF(connp);
5045 5045 first_connp = connp;
5046 5046 connp = connp->conn_next;
5047 5047
5048 5048 for (;;) {
5049 5049 while (connp != NULL) {
5050 5050 /* Note: IPCL_PROTO_MATCH includes conn_wantpacket */
5051 5051 if (IPCL_PROTO_MATCH(connp, ira, ipha) &&
5052 5052 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5053 5053 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5054 5054 ira, connp)))
5055 5055 break;
5056 5056 connp = connp->conn_next;
5057 5057 }
5058 5058
5059 5059 if (connp == NULL) {
5060 5060 /* No more interested clients */
5061 5061 connp = first_connp;
5062 5062 break;
5063 5063 }
5064 5064 if (((mp1 = dupmsg(mp)) == NULL) &&
5065 5065 ((mp1 = copymsg(mp)) == NULL)) {
5066 5066 /* Memory allocation failed */
5067 5067 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5068 5068 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5069 5069 connp = first_connp;
5070 5070 break;
5071 5071 }
5072 5072
5073 5073 CONN_INC_REF(connp);
5074 5074 mutex_exit(&connfp->connf_lock);
5075 5075
5076 5076 ip_fanout_proto_conn(connp, mp1, (ipha_t *)mp1->b_rptr, NULL,
5077 5077 ira);
5078 5078
5079 5079 mutex_enter(&connfp->connf_lock);
5080 5080 /* Follow the next pointer before releasing the conn. */
5081 5081 next_connp = connp->conn_next;
5082 5082 CONN_DEC_REF(connp);
5083 5083 connp = next_connp;
5084 5084 }
5085 5085
5086 5086 /* Last one. Send it upstream. */
5087 5087 mutex_exit(&connfp->connf_lock);
5088 5088
5089 5089 ip_fanout_proto_conn(connp, mp, ipha, NULL, ira);
5090 5090
5091 5091 CONN_DEC_REF(connp);
5092 5092 }
5093 5093
5094 5094 /*
5095 5095 * If we have a IPsec NAT-Traversal packet, strip the zero-SPI or
5096 5096 * pass it along to ESP if the SPI is non-zero. Returns the mblk if the mblk
5097 5097 * is not consumed.
5098 5098 *
5099 5099 * One of three things can happen, all of which affect the passed-in mblk:
5100 5100 *
5101 5101 * 1.) The packet is stock UDP and gets its zero-SPI stripped. Return mblk..
5102 5102 *
5103 5103 * 2.) The packet is ESP-in-UDP, gets transformed into an equivalent
5104 5104 * ESP packet, and is passed along to ESP for consumption. Return NULL.
5105 5105 *
5106 5106 * 3.) The packet is an ESP-in-UDP Keepalive. Drop it and return NULL.
5107 5107 */
5108 5108 mblk_t *
5109 5109 zero_spi_check(mblk_t *mp, ip_recv_attr_t *ira)
5110 5110 {
5111 5111 int shift, plen, iph_len;
5112 5112 ipha_t *ipha;
5113 5113 udpha_t *udpha;
5114 5114 uint32_t *spi;
5115 5115 uint32_t esp_ports;
5116 5116 uint8_t *orptr;
5117 5117 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
5118 5118 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
5119 5119
5120 5120 ipha = (ipha_t *)mp->b_rptr;
5121 5121 iph_len = ira->ira_ip_hdr_length;
5122 5122 plen = ira->ira_pktlen;
5123 5123
5124 5124 if (plen - iph_len - sizeof (udpha_t) < sizeof (uint32_t)) {
5125 5125 /*
5126 5126 * Most likely a keepalive for the benefit of an intervening
5127 5127 * NAT. These aren't for us, per se, so drop it.
5128 5128 *
5129 5129 * RFC 3947/8 doesn't say for sure what to do for 2-3
5130 5130 * byte packets (keepalives are 1-byte), but we'll drop them
5131 5131 * also.
5132 5132 */
5133 5133 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
5134 5134 DROPPER(ipss, ipds_esp_nat_t_ka), &ipss->ipsec_dropper);
5135 5135 return (NULL);
5136 5136 }
5137 5137
5138 5138 if (MBLKL(mp) < iph_len + sizeof (udpha_t) + sizeof (*spi)) {
5139 5139 /* might as well pull it all up - it might be ESP. */
5140 5140 if (!pullupmsg(mp, -1)) {
5141 5141 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
5142 5142 DROPPER(ipss, ipds_esp_nomem),
5143 5143 &ipss->ipsec_dropper);
5144 5144 return (NULL);
5145 5145 }
5146 5146
5147 5147 ipha = (ipha_t *)mp->b_rptr;
5148 5148 }
5149 5149 spi = (uint32_t *)(mp->b_rptr + iph_len + sizeof (udpha_t));
5150 5150 if (*spi == 0) {
5151 5151 /* UDP packet - remove 0-spi. */
5152 5152 shift = sizeof (uint32_t);
5153 5153 } else {
5154 5154 /* ESP-in-UDP packet - reduce to ESP. */
5155 5155 ipha->ipha_protocol = IPPROTO_ESP;
5156 5156 shift = sizeof (udpha_t);
5157 5157 }
5158 5158
5159 5159 /* Fix IP header */
5160 5160 ira->ira_pktlen = (plen - shift);
5161 5161 ipha->ipha_length = htons(ira->ira_pktlen);
5162 5162 ipha->ipha_hdr_checksum = 0;
5163 5163
5164 5164 orptr = mp->b_rptr;
5165 5165 mp->b_rptr += shift;
5166 5166
5167 5167 udpha = (udpha_t *)(orptr + iph_len);
5168 5168 if (*spi == 0) {
5169 5169 ASSERT((uint8_t *)ipha == orptr);
5170 5170 udpha->uha_length = htons(plen - shift - iph_len);
5171 5171 iph_len += sizeof (udpha_t); /* For the call to ovbcopy(). */
5172 5172 esp_ports = 0;
5173 5173 } else {
5174 5174 esp_ports = *((uint32_t *)udpha);
5175 5175 ASSERT(esp_ports != 0);
5176 5176 }
5177 5177 ovbcopy(orptr, orptr + shift, iph_len);
5178 5178 if (esp_ports != 0) /* Punt up for ESP processing. */ {
5179 5179 ipha = (ipha_t *)(orptr + shift);
5180 5180
5181 5181 ira->ira_flags |= IRAF_ESP_UDP_PORTS;
5182 5182 ira->ira_esp_udp_ports = esp_ports;
5183 5183 ip_fanout_v4(mp, ipha, ira);
5184 5184 return (NULL);
5185 5185 }
5186 5186 return (mp);
5187 5187 }
5188 5188
5189 5189 /*
5190 5190 * Deliver a udp packet to the given conn, possibly applying ipsec policy.
5191 5191 * Handles IPv4 and IPv6.
5192 5192 * We are responsible for disposing of mp, such as by freemsg() or putnext()
5193 5193 * Caller is responsible for dropping references to the conn.
5194 5194 */
5195 5195 void
5196 5196 ip_fanout_udp_conn(conn_t *connp, mblk_t *mp, ipha_t *ipha, ip6_t *ip6h,
5197 5197 ip_recv_attr_t *ira)
5198 5198 {
5199 5199 ill_t *ill = ira->ira_ill;
5200 5200 ip_stack_t *ipst = ill->ill_ipst;
5201 5201 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
5202 5202 boolean_t secure;
5203 5203 iaflags_t iraflags = ira->ira_flags;
5204 5204
5205 5205 secure = iraflags & IRAF_IPSEC_SECURE;
5206 5206
5207 5207 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld :
5208 5208 !canputnext(connp->conn_rq)) {
5209 5209 BUMP_MIB(ill->ill_ip_mib, udpIfStatsInOverflows);
5210 5210 freemsg(mp);
5211 5211 return;
5212 5212 }
5213 5213
5214 5214 if (((iraflags & IRAF_IS_IPV4) ?
5215 5215 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
5216 5216 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
5217 5217 secure) {
5218 5218 mp = ipsec_check_inbound_policy(mp, connp, ipha,
5219 5219 ip6h, ira);
5220 5220 if (mp == NULL) {
5221 5221 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5222 5222 /* Note that mp is NULL */
5223 5223 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5224 5224 return;
5225 5225 }
5226 5226 }
5227 5227
5228 5228 /*
5229 5229 * Since this code is not used for UDP unicast we don't need a NAT_T
5230 5230 * check. Only ip_fanout_v4 has that check.
5231 5231 */
5232 5232 if (ira->ira_flags & IRAF_ICMP_ERROR) {
5233 5233 (connp->conn_recvicmp)(connp, mp, NULL, ira);
5234 5234 } else {
5235 5235 ill_t *rill = ira->ira_rill;
5236 5236
5237 5237 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
5238 5238 ira->ira_ill = ira->ira_rill = NULL;
5239 5239 /* Send it upstream */
5240 5240 (connp->conn_recv)(connp, mp, NULL, ira);
5241 5241 ira->ira_ill = ill;
5242 5242 ira->ira_rill = rill;
5243 5243 }
5244 5244 }
5245 5245
5246 5246 /*
5247 5247 * Fanout for UDP packets that are multicast or broadcast, and ICMP errors.
5248 5248 * (Unicast fanout is handled in ip_input_v4.)
5249 5249 *
5250 5250 * If SO_REUSEADDR is set all multicast and broadcast packets
5251 5251 * will be delivered to all conns bound to the same port.
5252 5252 *
5253 5253 * If there is at least one matching AF_INET receiver, then we will
5254 5254 * ignore any AF_INET6 receivers.
5255 5255 * In the special case where an AF_INET socket binds to 0.0.0.0/<port> and an
5256 5256 * AF_INET6 socket binds to ::/<port>, only the AF_INET socket receives the IPv4
5257 5257 * packets.
5258 5258 *
5259 5259 * Zones notes:
5260 5260 * Earlier in ip_input on a system with multiple shared-IP zones we
5261 5261 * duplicate the multicast and broadcast packets and send them up
5262 5262 * with each explicit zoneid that exists on that ill.
5263 5263 * This means that here we can match the zoneid with SO_ALLZONES being special.
5264 5264 */
5265 5265 void
5266 5266 ip_fanout_udp_multi_v4(mblk_t *mp, ipha_t *ipha, uint16_t lport, uint16_t fport,
5267 5267 ip_recv_attr_t *ira)
5268 5268 {
5269 5269 ipaddr_t laddr;
5270 5270 in6_addr_t v6faddr;
5271 5271 conn_t *connp;
5272 5272 connf_t *connfp;
5273 5273 ipaddr_t faddr;
5274 5274 ill_t *ill = ira->ira_ill;
5275 5275 ip_stack_t *ipst = ill->ill_ipst;
5276 5276
5277 5277 ASSERT(ira->ira_flags & (IRAF_MULTIBROADCAST|IRAF_ICMP_ERROR));
5278 5278
5279 5279 laddr = ipha->ipha_dst;
5280 5280 faddr = ipha->ipha_src;
5281 5281
5282 5282 connfp = &ipst->ips_ipcl_udp_fanout[IPCL_UDP_HASH(lport, ipst)];
5283 5283 mutex_enter(&connfp->connf_lock);
5284 5284 connp = connfp->connf_head;
5285 5285
5286 5286 /*
5287 5287 * If SO_REUSEADDR has been set on the first we send the
5288 5288 * packet to all clients that have joined the group and
5289 5289 * match the port.
5290 5290 */
5291 5291 while (connp != NULL) {
5292 5292 if ((IPCL_UDP_MATCH(connp, lport, laddr, fport, faddr)) &&
5293 5293 conn_wantpacket(connp, ira, ipha) &&
5294 5294 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5295 5295 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp)))
5296 5296 break;
5297 5297 connp = connp->conn_next;
5298 5298 }
5299 5299
5300 5300 if (connp == NULL)
5301 5301 goto notfound;
5302 5302
5303 5303 CONN_INC_REF(connp);
5304 5304
5305 5305 if (connp->conn_reuseaddr) {
5306 5306 conn_t *first_connp = connp;
5307 5307 conn_t *next_connp;
5308 5308 mblk_t *mp1;
5309 5309
5310 5310 connp = connp->conn_next;
5311 5311 for (;;) {
5312 5312 while (connp != NULL) {
5313 5313 if (IPCL_UDP_MATCH(connp, lport, laddr,
5314 5314 fport, faddr) &&
5315 5315 conn_wantpacket(connp, ira, ipha) &&
5316 5316 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5317 5317 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5318 5318 ira, connp)))
5319 5319 break;
5320 5320 connp = connp->conn_next;
5321 5321 }
5322 5322 if (connp == NULL) {
5323 5323 /* No more interested clients */
5324 5324 connp = first_connp;
5325 5325 break;
5326 5326 }
5327 5327 if (((mp1 = dupmsg(mp)) == NULL) &&
5328 5328 ((mp1 = copymsg(mp)) == NULL)) {
5329 5329 /* Memory allocation failed */
5330 5330 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5331 5331 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5332 5332 connp = first_connp;
5333 5333 break;
5334 5334 }
5335 5335 CONN_INC_REF(connp);
5336 5336 mutex_exit(&connfp->connf_lock);
5337 5337
5338 5338 IP_STAT(ipst, ip_udp_fanmb);
5339 5339 ip_fanout_udp_conn(connp, mp1, (ipha_t *)mp1->b_rptr,
5340 5340 NULL, ira);
5341 5341 mutex_enter(&connfp->connf_lock);
5342 5342 /* Follow the next pointer before releasing the conn */
5343 5343 next_connp = connp->conn_next;
5344 5344 CONN_DEC_REF(connp);
5345 5345 connp = next_connp;
5346 5346 }
5347 5347 }
5348 5348
5349 5349 /* Last one. Send it upstream. */
5350 5350 mutex_exit(&connfp->connf_lock);
5351 5351 IP_STAT(ipst, ip_udp_fanmb);
5352 5352 ip_fanout_udp_conn(connp, mp, ipha, NULL, ira);
5353 5353 CONN_DEC_REF(connp);
5354 5354 return;
5355 5355
5356 5356 notfound:
5357 5357 mutex_exit(&connfp->connf_lock);
5358 5358 /*
5359 5359 * IPv6 endpoints bound to multicast IPv4-mapped addresses
5360 5360 * have already been matched above, since they live in the IPv4
5361 5361 * fanout tables. This implies we only need to
5362 5362 * check for IPv6 in6addr_any endpoints here.
5363 5363 * Thus we compare using ipv6_all_zeros instead of the destination
5364 5364 * address, except for the multicast group membership lookup which
5365 5365 * uses the IPv4 destination.
5366 5366 */
5367 5367 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_src, &v6faddr);
5368 5368 connfp = &ipst->ips_ipcl_udp_fanout[IPCL_UDP_HASH(lport, ipst)];
5369 5369 mutex_enter(&connfp->connf_lock);
5370 5370 connp = connfp->connf_head;
5371 5371 /*
5372 5372 * IPv4 multicast packet being delivered to an AF_INET6
5373 5373 * in6addr_any endpoint.
5374 5374 * Need to check conn_wantpacket(). Note that we use conn_wantpacket()
5375 5375 * and not conn_wantpacket_v6() since any multicast membership is
5376 5376 * for an IPv4-mapped multicast address.
5377 5377 */
5378 5378 while (connp != NULL) {
5379 5379 if (IPCL_UDP_MATCH_V6(connp, lport, ipv6_all_zeros,
5380 5380 fport, v6faddr) &&
5381 5381 conn_wantpacket(connp, ira, ipha) &&
5382 5382 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5383 5383 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp)))
5384 5384 break;
5385 5385 connp = connp->conn_next;
5386 5386 }
5387 5387
5388 5388 if (connp == NULL) {
5389 5389 /*
5390 5390 * No one bound to this port. Is
5391 5391 * there a client that wants all
5392 5392 * unclaimed datagrams?
5393 5393 */
5394 5394 mutex_exit(&connfp->connf_lock);
5395 5395
5396 5396 if (ipst->ips_ipcl_proto_fanout_v4[IPPROTO_UDP].connf_head !=
5397 5397 NULL) {
5398 5398 ASSERT(ira->ira_protocol == IPPROTO_UDP);
5399 5399 ip_fanout_proto_v4(mp, ipha, ira);
5400 5400 } else {
5401 5401 /*
5402 5402 * We used to attempt to send an icmp error here, but
5403 5403 * since this is known to be a multicast packet
5404 5404 * and we don't send icmp errors in response to
5405 5405 * multicast, just drop the packet and give up sooner.
5406 5406 */
5407 5407 BUMP_MIB(ill->ill_ip_mib, udpIfStatsNoPorts);
5408 5408 freemsg(mp);
5409 5409 }
5410 5410 return;
5411 5411 }
5412 5412 CONN_INC_REF(connp);
5413 5413 ASSERT(IPCL_IS_NONSTR(connp) || connp->conn_rq != NULL);
5414 5414
5415 5415 /*
5416 5416 * If SO_REUSEADDR has been set on the first we send the
5417 5417 * packet to all clients that have joined the group and
5418 5418 * match the port.
5419 5419 */
5420 5420 if (connp->conn_reuseaddr) {
5421 5421 conn_t *first_connp = connp;
5422 5422 conn_t *next_connp;
5423 5423 mblk_t *mp1;
5424 5424
5425 5425 connp = connp->conn_next;
5426 5426 for (;;) {
5427 5427 while (connp != NULL) {
5428 5428 if (IPCL_UDP_MATCH_V6(connp, lport,
5429 5429 ipv6_all_zeros, fport, v6faddr) &&
5430 5430 conn_wantpacket(connp, ira, ipha) &&
5431 5431 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5432 5432 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5433 5433 ira, connp)))
5434 5434 break;
5435 5435 connp = connp->conn_next;
5436 5436 }
5437 5437 if (connp == NULL) {
5438 5438 /* No more interested clients */
5439 5439 connp = first_connp;
5440 5440 break;
5441 5441 }
5442 5442 if (((mp1 = dupmsg(mp)) == NULL) &&
5443 5443 ((mp1 = copymsg(mp)) == NULL)) {
5444 5444 /* Memory allocation failed */
5445 5445 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5446 5446 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5447 5447 connp = first_connp;
5448 5448 break;
5449 5449 }
5450 5450 CONN_INC_REF(connp);
5451 5451 mutex_exit(&connfp->connf_lock);
5452 5452
5453 5453 IP_STAT(ipst, ip_udp_fanmb);
5454 5454 ip_fanout_udp_conn(connp, mp1, (ipha_t *)mp1->b_rptr,
5455 5455 NULL, ira);
5456 5456 mutex_enter(&connfp->connf_lock);
5457 5457 /* Follow the next pointer before releasing the conn */
5458 5458 next_connp = connp->conn_next;
5459 5459 CONN_DEC_REF(connp);
5460 5460 connp = next_connp;
5461 5461 }
5462 5462 }
5463 5463
5464 5464 /* Last one. Send it upstream. */
5465 5465 mutex_exit(&connfp->connf_lock);
5466 5466 IP_STAT(ipst, ip_udp_fanmb);
5467 5467 ip_fanout_udp_conn(connp, mp, ipha, NULL, ira);
5468 5468 CONN_DEC_REF(connp);
5469 5469 }
5470 5470
5471 5471 /*
5472 5472 * Split an incoming packet's IPv4 options into the label and the other options.
5473 5473 * If 'allocate' is set it does memory allocation for the ip_pkt_t, including
5474 5474 * clearing out any leftover label or options.
5475 5475 * Otherwise it just makes ipp point into the packet.
5476 5476 *
5477 5477 * Returns zero if ok; ENOMEM if the buffer couldn't be allocated.
5478 5478 */
5479 5479 int
5480 5480 ip_find_hdr_v4(ipha_t *ipha, ip_pkt_t *ipp, boolean_t allocate)
5481 5481 {
5482 5482 uchar_t *opt;
5483 5483 uint32_t totallen;
5484 5484 uint32_t optval;
5485 5485 uint32_t optlen;
5486 5486
5487 5487 ipp->ipp_fields |= IPPF_HOPLIMIT | IPPF_TCLASS | IPPF_ADDR;
5488 5488 ipp->ipp_hoplimit = ipha->ipha_ttl;
5489 5489 ipp->ipp_type_of_service = ipha->ipha_type_of_service;
5490 5490 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &ipp->ipp_addr);
5491 5491
5492 5492 /*
5493 5493 * Get length (in 4 byte octets) of IP header options.
5494 5494 */
5495 5495 totallen = ipha->ipha_version_and_hdr_length -
5496 5496 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
5497 5497
5498 5498 if (totallen == 0) {
5499 5499 if (!allocate)
5500 5500 return (0);
5501 5501
5502 5502 /* Clear out anything from a previous packet */
5503 5503 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
5504 5504 kmem_free(ipp->ipp_ipv4_options,
5505 5505 ipp->ipp_ipv4_options_len);
5506 5506 ipp->ipp_ipv4_options = NULL;
5507 5507 ipp->ipp_ipv4_options_len = 0;
5508 5508 ipp->ipp_fields &= ~IPPF_IPV4_OPTIONS;
5509 5509 }
5510 5510 if (ipp->ipp_fields & IPPF_LABEL_V4) {
5511 5511 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
5512 5512 ipp->ipp_label_v4 = NULL;
5513 5513 ipp->ipp_label_len_v4 = 0;
5514 5514 ipp->ipp_fields &= ~IPPF_LABEL_V4;
5515 5515 }
5516 5516 return (0);
5517 5517 }
5518 5518
5519 5519 totallen <<= 2;
5520 5520 opt = (uchar_t *)&ipha[1];
5521 5521 if (!is_system_labeled()) {
5522 5522
5523 5523 copyall:
5524 5524 if (!allocate) {
5525 5525 if (totallen != 0) {
5526 5526 ipp->ipp_ipv4_options = opt;
5527 5527 ipp->ipp_ipv4_options_len = totallen;
5528 5528 ipp->ipp_fields |= IPPF_IPV4_OPTIONS;
5529 5529 }
5530 5530 return (0);
5531 5531 }
5532 5532 /* Just copy all of options */
5533 5533 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
5534 5534 if (totallen == ipp->ipp_ipv4_options_len) {
5535 5535 bcopy(opt, ipp->ipp_ipv4_options, totallen);
5536 5536 return (0);
5537 5537 }
5538 5538 kmem_free(ipp->ipp_ipv4_options,
5539 5539 ipp->ipp_ipv4_options_len);
5540 5540 ipp->ipp_ipv4_options = NULL;
5541 5541 ipp->ipp_ipv4_options_len = 0;
5542 5542 ipp->ipp_fields &= ~IPPF_IPV4_OPTIONS;
5543 5543 }
5544 5544 if (totallen == 0)
5545 5545 return (0);
5546 5546
5547 5547 ipp->ipp_ipv4_options = kmem_alloc(totallen, KM_NOSLEEP);
5548 5548 if (ipp->ipp_ipv4_options == NULL)
5549 5549 return (ENOMEM);
5550 5550 ipp->ipp_ipv4_options_len = totallen;
5551 5551 ipp->ipp_fields |= IPPF_IPV4_OPTIONS;
5552 5552 bcopy(opt, ipp->ipp_ipv4_options, totallen);
5553 5553 return (0);
5554 5554 }
5555 5555
5556 5556 if (allocate && (ipp->ipp_fields & IPPF_LABEL_V4)) {
5557 5557 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
5558 5558 ipp->ipp_label_v4 = NULL;
5559 5559 ipp->ipp_label_len_v4 = 0;
5560 5560 ipp->ipp_fields &= ~IPPF_LABEL_V4;
5561 5561 }
5562 5562
5563 5563 /*
5564 5564 * Search for CIPSO option.
5565 5565 * We assume CIPSO is first in options if it is present.
5566 5566 * If it isn't, then ipp_opt_ipv4_options will not include the options
5567 5567 * prior to the CIPSO option.
5568 5568 */
5569 5569 while (totallen != 0) {
5570 5570 switch (optval = opt[IPOPT_OPTVAL]) {
5571 5571 case IPOPT_EOL:
5572 5572 return (0);
5573 5573 case IPOPT_NOP:
5574 5574 optlen = 1;
5575 5575 break;
5576 5576 default:
5577 5577 if (totallen <= IPOPT_OLEN)
5578 5578 return (EINVAL);
5579 5579 optlen = opt[IPOPT_OLEN];
5580 5580 if (optlen < 2)
5581 5581 return (EINVAL);
5582 5582 }
5583 5583 if (optlen > totallen)
5584 5584 return (EINVAL);
5585 5585
5586 5586 switch (optval) {
5587 5587 case IPOPT_COMSEC:
5588 5588 if (!allocate) {
5589 5589 ipp->ipp_label_v4 = opt;
5590 5590 ipp->ipp_label_len_v4 = optlen;
5591 5591 ipp->ipp_fields |= IPPF_LABEL_V4;
5592 5592 } else {
5593 5593 ipp->ipp_label_v4 = kmem_alloc(optlen,
5594 5594 KM_NOSLEEP);
5595 5595 if (ipp->ipp_label_v4 == NULL)
5596 5596 return (ENOMEM);
5597 5597 ipp->ipp_label_len_v4 = optlen;
5598 5598 ipp->ipp_fields |= IPPF_LABEL_V4;
5599 5599 bcopy(opt, ipp->ipp_label_v4, optlen);
5600 5600 }
5601 5601 totallen -= optlen;
5602 5602 opt += optlen;
5603 5603
5604 5604 /* Skip padding bytes until we get to a multiple of 4 */
5605 5605 while ((totallen & 3) != 0 && opt[0] == IPOPT_NOP) {
5606 5606 totallen--;
5607 5607 opt++;
5608 5608 }
5609 5609 /* Remaining as ipp_ipv4_options */
5610 5610 goto copyall;
5611 5611 }
5612 5612 totallen -= optlen;
5613 5613 opt += optlen;
5614 5614 }
5615 5615 /* No CIPSO found; return everything as ipp_ipv4_options */
5616 5616 totallen = ipha->ipha_version_and_hdr_length -
5617 5617 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
5618 5618 totallen <<= 2;
5619 5619 opt = (uchar_t *)&ipha[1];
5620 5620 goto copyall;
5621 5621 }
5622 5622
5623 5623 /*
5624 5624 * Efficient versions of lookup for an IRE when we only
5625 5625 * match the address.
5626 5626 * For RTF_REJECT or BLACKHOLE we return IRE_NOROUTE.
5627 5627 * Does not handle multicast addresses.
5628 5628 */
5629 5629 uint_t
5630 5630 ip_type_v4(ipaddr_t addr, ip_stack_t *ipst)
5631 5631 {
5632 5632 ire_t *ire;
5633 5633 uint_t result;
5634 5634
5635 5635 ire = ire_ftable_lookup_simple_v4(addr, 0, ipst, NULL);
5636 5636 ASSERT(ire != NULL);
5637 5637 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))
5638 5638 result = IRE_NOROUTE;
5639 5639 else
5640 5640 result = ire->ire_type;
5641 5641 ire_refrele(ire);
5642 5642 return (result);
5643 5643 }
5644 5644
5645 5645 /*
5646 5646 * Efficient versions of lookup for an IRE when we only
5647 5647 * match the address.
5648 5648 * For RTF_REJECT or BLACKHOLE we return IRE_NOROUTE.
5649 5649 * Does not handle multicast addresses.
5650 5650 */
5651 5651 uint_t
5652 5652 ip_type_v6(const in6_addr_t *addr, ip_stack_t *ipst)
5653 5653 {
5654 5654 ire_t *ire;
5655 5655 uint_t result;
5656 5656
5657 5657 ire = ire_ftable_lookup_simple_v6(addr, 0, ipst, NULL);
5658 5658 ASSERT(ire != NULL);
5659 5659 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))
5660 5660 result = IRE_NOROUTE;
5661 5661 else
5662 5662 result = ire->ire_type;
5663 5663 ire_refrele(ire);
5664 5664 return (result);
5665 5665 }
5666 5666
5667 5667 /*
5668 5668 * Nobody should be sending
5669 5669 * packets up this stream
5670 5670 */
5671 5671 static int
5672 5672 ip_lrput(queue_t *q, mblk_t *mp)
5673 5673 {
5674 5674 switch (mp->b_datap->db_type) {
5675 5675 case M_FLUSH:
5676 5676 /* Turn around */
5677 5677 if (*mp->b_rptr & FLUSHW) {
5678 5678 *mp->b_rptr &= ~FLUSHR;
5679 5679 qreply(q, mp);
5680 5680 return (0);
5681 5681 }
5682 5682 break;
5683 5683 }
5684 5684 freemsg(mp);
5685 5685 return (0);
5686 5686 }
5687 5687
5688 5688 /* Nobody should be sending packets down this stream */
5689 5689 /* ARGSUSED */
5690 5690 int
5691 5691 ip_lwput(queue_t *q, mblk_t *mp)
5692 5692 {
5693 5693 freemsg(mp);
5694 5694 return (0);
5695 5695 }
5696 5696
5697 5697 /*
5698 5698 * Move the first hop in any source route to ipha_dst and remove that part of
5699 5699 * the source route. Called by other protocols. Errors in option formatting
5700 5700 * are ignored - will be handled by ip_output_options. Return the final
5701 5701 * destination (either ipha_dst or the last entry in a source route.)
5702 5702 */
5703 5703 ipaddr_t
5704 5704 ip_massage_options(ipha_t *ipha, netstack_t *ns)
5705 5705 {
5706 5706 ipoptp_t opts;
5707 5707 uchar_t *opt;
5708 5708 uint8_t optval;
5709 5709 uint8_t optlen;
5710 5710 ipaddr_t dst;
5711 5711 int i;
5712 5712 ip_stack_t *ipst = ns->netstack_ip;
5713 5713
5714 5714 ip2dbg(("ip_massage_options\n"));
5715 5715 dst = ipha->ipha_dst;
5716 5716 for (optval = ipoptp_first(&opts, ipha);
5717 5717 optval != IPOPT_EOL;
5718 5718 optval = ipoptp_next(&opts)) {
5719 5719 opt = opts.ipoptp_cur;
5720 5720 switch (optval) {
5721 5721 uint8_t off;
5722 5722 case IPOPT_SSRR:
5723 5723 case IPOPT_LSRR:
5724 5724 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
5725 5725 ip1dbg(("ip_massage_options: bad src route\n"));
5726 5726 break;
5727 5727 }
5728 5728 optlen = opts.ipoptp_len;
5729 5729 off = opt[IPOPT_OFFSET];
5730 5730 off--;
5731 5731 redo_srr:
5732 5732 if (optlen < IP_ADDR_LEN ||
5733 5733 off > optlen - IP_ADDR_LEN) {
5734 5734 /* End of source route */
5735 5735 ip1dbg(("ip_massage_options: end of SR\n"));
5736 5736 break;
5737 5737 }
5738 5738 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
5739 5739 ip1dbg(("ip_massage_options: next hop 0x%x\n",
5740 5740 ntohl(dst)));
5741 5741 /*
5742 5742 * Check if our address is present more than
5743 5743 * once as consecutive hops in source route.
5744 5744 * XXX verify per-interface ip_forwarding
5745 5745 * for source route?
5746 5746 */
5747 5747 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
5748 5748 off += IP_ADDR_LEN;
5749 5749 goto redo_srr;
5750 5750 }
5751 5751 if (dst == htonl(INADDR_LOOPBACK)) {
5752 5752 ip1dbg(("ip_massage_options: loopback addr in "
5753 5753 "source route!\n"));
5754 5754 break;
5755 5755 }
5756 5756 /*
5757 5757 * Update ipha_dst to be the first hop and remove the
5758 5758 * first hop from the source route (by overwriting
5759 5759 * part of the option with NOP options).
5760 5760 */
5761 5761 ipha->ipha_dst = dst;
5762 5762 /* Put the last entry in dst */
5763 5763 off = ((optlen - IP_ADDR_LEN - 3) & ~(IP_ADDR_LEN-1)) +
5764 5764 3;
5765 5765 bcopy(&opt[off], &dst, IP_ADDR_LEN);
5766 5766
5767 5767 ip1dbg(("ip_massage_options: last hop 0x%x\n",
5768 5768 ntohl(dst)));
5769 5769 /* Move down and overwrite */
5770 5770 opt[IP_ADDR_LEN] = opt[0];
5771 5771 opt[IP_ADDR_LEN+1] = opt[IPOPT_OLEN] - IP_ADDR_LEN;
5772 5772 opt[IP_ADDR_LEN+2] = opt[IPOPT_OFFSET];
5773 5773 for (i = 0; i < IP_ADDR_LEN; i++)
5774 5774 opt[i] = IPOPT_NOP;
5775 5775 break;
5776 5776 }
5777 5777 }
5778 5778 return (dst);
5779 5779 }
5780 5780
5781 5781 /*
5782 5782 * Return the network mask
5783 5783 * associated with the specified address.
5784 5784 */
5785 5785 ipaddr_t
5786 5786 ip_net_mask(ipaddr_t addr)
5787 5787 {
5788 5788 uchar_t *up = (uchar_t *)&addr;
5789 5789 ipaddr_t mask = 0;
5790 5790 uchar_t *maskp = (uchar_t *)&mask;
5791 5791
5792 5792 #if defined(__i386) || defined(__amd64)
5793 5793 #define TOTALLY_BRAIN_DAMAGED_C_COMPILER
5794 5794 #endif
5795 5795 #ifdef TOTALLY_BRAIN_DAMAGED_C_COMPILER
5796 5796 maskp[0] = maskp[1] = maskp[2] = maskp[3] = 0;
5797 5797 #endif
5798 5798 if (CLASSD(addr)) {
5799 5799 maskp[0] = 0xF0;
5800 5800 return (mask);
5801 5801 }
5802 5802
5803 5803 /* We assume Class E default netmask to be 32 */
5804 5804 if (CLASSE(addr))
5805 5805 return (0xffffffffU);
5806 5806
5807 5807 if (addr == 0)
5808 5808 return (0);
5809 5809 maskp[0] = 0xFF;
5810 5810 if ((up[0] & 0x80) == 0)
5811 5811 return (mask);
5812 5812
5813 5813 maskp[1] = 0xFF;
5814 5814 if ((up[0] & 0xC0) == 0x80)
5815 5815 return (mask);
5816 5816
5817 5817 maskp[2] = 0xFF;
5818 5818 if ((up[0] & 0xE0) == 0xC0)
5819 5819 return (mask);
5820 5820
5821 5821 /* Otherwise return no mask */
5822 5822 return ((ipaddr_t)0);
5823 5823 }
5824 5824
5825 5825 /* Name/Value Table Lookup Routine */
5826 5826 char *
5827 5827 ip_nv_lookup(nv_t *nv, int value)
5828 5828 {
5829 5829 if (!nv)
5830 5830 return (NULL);
5831 5831 for (; nv->nv_name; nv++) {
5832 5832 if (nv->nv_value == value)
5833 5833 return (nv->nv_name);
5834 5834 }
5835 5835 return ("unknown");
5836 5836 }
5837 5837
5838 5838 static int
5839 5839 ip_wait_for_info_ack(ill_t *ill)
5840 5840 {
5841 5841 int err;
5842 5842
5843 5843 mutex_enter(&ill->ill_lock);
5844 5844 while (ill->ill_state_flags & ILL_LL_SUBNET_PENDING) {
5845 5845 /*
5846 5846 * Return value of 0 indicates a pending signal.
5847 5847 */
5848 5848 err = cv_wait_sig(&ill->ill_cv, &ill->ill_lock);
5849 5849 if (err == 0) {
5850 5850 mutex_exit(&ill->ill_lock);
5851 5851 return (EINTR);
5852 5852 }
5853 5853 }
5854 5854 mutex_exit(&ill->ill_lock);
5855 5855 /*
5856 5856 * ip_rput_other could have set an error in ill_error on
5857 5857 * receipt of M_ERROR.
5858 5858 */
5859 5859 return (ill->ill_error);
5860 5860 }
5861 5861
5862 5862 /*
5863 5863 * This is a module open, i.e. this is a control stream for access
5864 5864 * to a DLPI device. We allocate an ill_t as the instance data in
5865 5865 * this case.
5866 5866 */
5867 5867 static int
5868 5868 ip_modopen(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5869 5869 {
5870 5870 ill_t *ill;
5871 5871 int err;
5872 5872 zoneid_t zoneid;
5873 5873 netstack_t *ns;
5874 5874 ip_stack_t *ipst;
5875 5875
5876 5876 /*
5877 5877 * Prevent unprivileged processes from pushing IP so that
5878 5878 * they can't send raw IP.
5879 5879 */
5880 5880 if (secpolicy_net_rawaccess(credp) != 0)
5881 5881 return (EPERM);
5882 5882
5883 5883 ns = netstack_find_by_cred(credp);
5884 5884 ASSERT(ns != NULL);
5885 5885 ipst = ns->netstack_ip;
5886 5886 ASSERT(ipst != NULL);
5887 5887
5888 5888 /*
5889 5889 * For exclusive stacks we set the zoneid to zero
5890 5890 * to make IP operate as if in the global zone.
5891 5891 */
5892 5892 if (ipst->ips_netstack->netstack_stackid != GLOBAL_NETSTACKID)
5893 5893 zoneid = GLOBAL_ZONEID;
5894 5894 else
5895 5895 zoneid = crgetzoneid(credp);
5896 5896
5897 5897 ill = (ill_t *)mi_open_alloc_sleep(sizeof (ill_t));
5898 5898 q->q_ptr = WR(q)->q_ptr = ill;
5899 5899 ill->ill_ipst = ipst;
5900 5900 ill->ill_zoneid = zoneid;
5901 5901
5902 5902 /*
5903 5903 * ill_init initializes the ill fields and then sends down
5904 5904 * down a DL_INFO_REQ after calling qprocson.
5905 5905 */
5906 5906 err = ill_init(q, ill);
5907 5907
5908 5908 if (err != 0) {
5909 5909 mi_free(ill);
5910 5910 netstack_rele(ipst->ips_netstack);
5911 5911 q->q_ptr = NULL;
5912 5912 WR(q)->q_ptr = NULL;
5913 5913 return (err);
5914 5914 }
5915 5915
5916 5916 /*
5917 5917 * Wait for the DL_INFO_ACK if a DL_INFO_REQ was sent.
5918 5918 *
5919 5919 * ill_init initializes the ipsq marking this thread as
5920 5920 * writer
5921 5921 */
5922 5922 ipsq_exit(ill->ill_phyint->phyint_ipsq);
5923 5923 err = ip_wait_for_info_ack(ill);
5924 5924 if (err == 0)
5925 5925 ill->ill_credp = credp;
5926 5926 else
5927 5927 goto fail;
5928 5928
5929 5929 crhold(credp);
5930 5930
5931 5931 mutex_enter(&ipst->ips_ip_mi_lock);
5932 5932 err = mi_open_link(&ipst->ips_ip_g_head, (IDP)q->q_ptr, devp, flag,
5933 5933 sflag, credp);
5934 5934 mutex_exit(&ipst->ips_ip_mi_lock);
5935 5935 fail:
5936 5936 if (err) {
5937 5937 (void) ip_close(q, 0, credp);
5938 5938 return (err);
5939 5939 }
5940 5940 return (0);
5941 5941 }
5942 5942
5943 5943 /* For /dev/ip aka AF_INET open */
5944 5944 int
5945 5945 ip_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5946 5946 {
5947 5947 return (ip_open(q, devp, flag, sflag, credp, B_FALSE));
5948 5948 }
5949 5949
5950 5950 /* For /dev/ip6 aka AF_INET6 open */
5951 5951 int
5952 5952 ip_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5953 5953 {
5954 5954 return (ip_open(q, devp, flag, sflag, credp, B_TRUE));
5955 5955 }
5956 5956
5957 5957 /* IP open routine. */
5958 5958 int
5959 5959 ip_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp,
5960 5960 boolean_t isv6)
5961 5961 {
5962 5962 conn_t *connp;
5963 5963 major_t maj;
5964 5964 zoneid_t zoneid;
5965 5965 netstack_t *ns;
5966 5966 ip_stack_t *ipst;
5967 5967
5968 5968 /* Allow reopen. */
5969 5969 if (q->q_ptr != NULL)
5970 5970 return (0);
5971 5971
5972 5972 if (sflag & MODOPEN) {
5973 5973 /* This is a module open */
5974 5974 return (ip_modopen(q, devp, flag, sflag, credp));
5975 5975 }
5976 5976
5977 5977 if ((flag & ~(FKLYR)) == IP_HELPER_STR) {
5978 5978 /*
5979 5979 * Non streams based socket looking for a stream
5980 5980 * to access IP
5981 5981 */
5982 5982 return (ip_helper_stream_setup(q, devp, flag, sflag,
5983 5983 credp, isv6));
5984 5984 }
5985 5985
5986 5986 ns = netstack_find_by_cred(credp);
5987 5987 ASSERT(ns != NULL);
5988 5988 ipst = ns->netstack_ip;
5989 5989 ASSERT(ipst != NULL);
5990 5990
5991 5991 /*
5992 5992 * For exclusive stacks we set the zoneid to zero
5993 5993 * to make IP operate as if in the global zone.
5994 5994 */
5995 5995 if (ipst->ips_netstack->netstack_stackid != GLOBAL_NETSTACKID)
5996 5996 zoneid = GLOBAL_ZONEID;
5997 5997 else
5998 5998 zoneid = crgetzoneid(credp);
5999 5999
6000 6000 /*
6001 6001 * We are opening as a device. This is an IP client stream, and we
6002 6002 * allocate an conn_t as the instance data.
6003 6003 */
6004 6004 connp = ipcl_conn_create(IPCL_IPCCONN, KM_SLEEP, ipst->ips_netstack);
6005 6005
6006 6006 /*
6007 6007 * ipcl_conn_create did a netstack_hold. Undo the hold that was
6008 6008 * done by netstack_find_by_cred()
6009 6009 */
6010 6010 netstack_rele(ipst->ips_netstack);
6011 6011
6012 6012 connp->conn_ixa->ixa_flags |= IXAF_MULTICAST_LOOP | IXAF_SET_ULP_CKSUM;
6013 6013 /* conn_allzones can not be set this early, hence no IPCL_ZONEID */
6014 6014 connp->conn_ixa->ixa_zoneid = zoneid;
6015 6015 connp->conn_zoneid = zoneid;
6016 6016
6017 6017 connp->conn_rq = q;
6018 6018 q->q_ptr = WR(q)->q_ptr = connp;
6019 6019
6020 6020 /* Minor tells us which /dev entry was opened */
6021 6021 if (isv6) {
6022 6022 connp->conn_family = AF_INET6;
6023 6023 connp->conn_ipversion = IPV6_VERSION;
6024 6024 connp->conn_ixa->ixa_flags &= ~IXAF_IS_IPV4;
6025 6025 connp->conn_ixa->ixa_src_preferences = IPV6_PREFER_SRC_DEFAULT;
6026 6026 } else {
6027 6027 connp->conn_family = AF_INET;
6028 6028 connp->conn_ipversion = IPV4_VERSION;
6029 6029 connp->conn_ixa->ixa_flags |= IXAF_IS_IPV4;
6030 6030 }
6031 6031
6032 6032 if ((ip_minor_arena_la != NULL) && (flag & SO_SOCKSTR) &&
6033 6033 ((connp->conn_dev = inet_minor_alloc(ip_minor_arena_la)) != 0)) {
6034 6034 connp->conn_minor_arena = ip_minor_arena_la;
6035 6035 } else {
6036 6036 /*
6037 6037 * Either minor numbers in the large arena were exhausted
6038 6038 * or a non socket application is doing the open.
6039 6039 * Try to allocate from the small arena.
6040 6040 */
6041 6041 if ((connp->conn_dev =
6042 6042 inet_minor_alloc(ip_minor_arena_sa)) == 0) {
6043 6043 /* CONN_DEC_REF takes care of netstack_rele() */
6044 6044 q->q_ptr = WR(q)->q_ptr = NULL;
6045 6045 CONN_DEC_REF(connp);
6046 6046 return (EBUSY);
6047 6047 }
6048 6048 connp->conn_minor_arena = ip_minor_arena_sa;
6049 6049 }
6050 6050
6051 6051 maj = getemajor(*devp);
6052 6052 *devp = makedevice(maj, (minor_t)connp->conn_dev);
6053 6053
6054 6054 /*
6055 6055 * connp->conn_cred is crfree()ed in ipcl_conn_destroy()
6056 6056 */
6057 6057 connp->conn_cred = credp;
6058 6058 connp->conn_cpid = curproc->p_pid;
6059 6059 /* Cache things in ixa without an extra refhold */
6060 6060 ASSERT(!(connp->conn_ixa->ixa_free_flags & IXA_FREE_CRED));
6061 6061 connp->conn_ixa->ixa_cred = connp->conn_cred;
6062 6062 connp->conn_ixa->ixa_cpid = connp->conn_cpid;
6063 6063 if (is_system_labeled())
6064 6064 connp->conn_ixa->ixa_tsl = crgetlabel(connp->conn_cred);
6065 6065
6066 6066 /*
6067 6067 * Handle IP_IOC_RTS_REQUEST and other ioctls which use conn_recv
6068 6068 */
6069 6069 connp->conn_recv = ip_conn_input;
6070 6070 connp->conn_recvicmp = ip_conn_input_icmp;
6071 6071
6072 6072 crhold(connp->conn_cred);
6073 6073
6074 6074 /*
6075 6075 * If the caller has the process-wide flag set, then default to MAC
6076 6076 * exempt mode. This allows read-down to unlabeled hosts.
6077 6077 */
6078 6078 if (getpflags(NET_MAC_AWARE, credp) != 0)
6079 6079 connp->conn_mac_mode = CONN_MAC_AWARE;
6080 6080
6081 6081 connp->conn_zone_is_global = (crgetzoneid(credp) == GLOBAL_ZONEID);
6082 6082
6083 6083 connp->conn_rq = q;
6084 6084 connp->conn_wq = WR(q);
6085 6085
6086 6086 /* Non-zero default values */
6087 6087 connp->conn_ixa->ixa_flags |= IXAF_MULTICAST_LOOP;
6088 6088
6089 6089 /*
6090 6090 * Make the conn globally visible to walkers
6091 6091 */
6092 6092 ASSERT(connp->conn_ref == 1);
6093 6093 mutex_enter(&connp->conn_lock);
6094 6094 connp->conn_state_flags &= ~CONN_INCIPIENT;
6095 6095 mutex_exit(&connp->conn_lock);
6096 6096
6097 6097 qprocson(q);
6098 6098
6099 6099 return (0);
6100 6100 }
6101 6101
6102 6102 /*
6103 6103 * Set IPsec policy from an ipsec_req_t. If the req is not "zero" and valid,
6104 6104 * all of them are copied to the conn_t. If the req is "zero", the policy is
6105 6105 * zeroed out. A "zero" policy has zero ipsr_{ah,req,self_encap}_req
6106 6106 * fields.
6107 6107 * We keep only the latest setting of the policy and thus policy setting
6108 6108 * is not incremental/cumulative.
6109 6109 *
6110 6110 * Requests to set policies with multiple alternative actions will
6111 6111 * go through a different API.
6112 6112 */
6113 6113 int
6114 6114 ipsec_set_req(cred_t *cr, conn_t *connp, ipsec_req_t *req)
6115 6115 {
6116 6116 uint_t ah_req = 0;
6117 6117 uint_t esp_req = 0;
6118 6118 uint_t se_req = 0;
6119 6119 ipsec_act_t *actp = NULL;
6120 6120 uint_t nact;
6121 6121 ipsec_policy_head_t *ph;
6122 6122 boolean_t is_pol_reset, is_pol_inserted = B_FALSE;
6123 6123 int error = 0;
6124 6124 netstack_t *ns = connp->conn_netstack;
6125 6125 ip_stack_t *ipst = ns->netstack_ip;
6126 6126 ipsec_stack_t *ipss = ns->netstack_ipsec;
6127 6127
6128 6128 #define REQ_MASK (IPSEC_PREF_REQUIRED|IPSEC_PREF_NEVER)
6129 6129
6130 6130 /*
6131 6131 * The IP_SEC_OPT option does not allow variable length parameters,
6132 6132 * hence a request cannot be NULL.
6133 6133 */
6134 6134 if (req == NULL)
6135 6135 return (EINVAL);
6136 6136
6137 6137 ah_req = req->ipsr_ah_req;
6138 6138 esp_req = req->ipsr_esp_req;
6139 6139 se_req = req->ipsr_self_encap_req;
6140 6140
6141 6141 /* Don't allow setting self-encap without one or more of AH/ESP. */
6142 6142 if (se_req != 0 && esp_req == 0 && ah_req == 0)
6143 6143 return (EINVAL);
6144 6144
6145 6145 /*
6146 6146 * Are we dealing with a request to reset the policy (i.e.
6147 6147 * zero requests).
6148 6148 */
6149 6149 is_pol_reset = ((ah_req & REQ_MASK) == 0 &&
6150 6150 (esp_req & REQ_MASK) == 0 &&
6151 6151 (se_req & REQ_MASK) == 0);
6152 6152
6153 6153 if (!is_pol_reset) {
6154 6154 /*
6155 6155 * If we couldn't load IPsec, fail with "protocol
6156 6156 * not supported".
6157 6157 * IPsec may not have been loaded for a request with zero
6158 6158 * policies, so we don't fail in this case.
6159 6159 */
6160 6160 mutex_enter(&ipss->ipsec_loader_lock);
6161 6161 if (ipss->ipsec_loader_state != IPSEC_LOADER_SUCCEEDED) {
6162 6162 mutex_exit(&ipss->ipsec_loader_lock);
6163 6163 return (EPROTONOSUPPORT);
6164 6164 }
6165 6165 mutex_exit(&ipss->ipsec_loader_lock);
6166 6166
6167 6167 /*
6168 6168 * Test for valid requests. Invalid algorithms
6169 6169 * need to be tested by IPsec code because new
6170 6170 * algorithms can be added dynamically.
6171 6171 */
6172 6172 if ((ah_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0 ||
6173 6173 (esp_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0 ||
6174 6174 (se_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0) {
6175 6175 return (EINVAL);
6176 6176 }
6177 6177
6178 6178 /*
6179 6179 * Only privileged users can issue these
6180 6180 * requests.
6181 6181 */
6182 6182 if (((ah_req & IPSEC_PREF_NEVER) ||
6183 6183 (esp_req & IPSEC_PREF_NEVER) ||
6184 6184 (se_req & IPSEC_PREF_NEVER)) &&
6185 6185 secpolicy_ip_config(cr, B_FALSE) != 0) {
6186 6186 return (EPERM);
6187 6187 }
6188 6188
6189 6189 /*
6190 6190 * The IPSEC_PREF_REQUIRED and IPSEC_PREF_NEVER
6191 6191 * are mutually exclusive.
6192 6192 */
6193 6193 if (((ah_req & REQ_MASK) == REQ_MASK) ||
6194 6194 ((esp_req & REQ_MASK) == REQ_MASK) ||
6195 6195 ((se_req & REQ_MASK) == REQ_MASK)) {
6196 6196 /* Both of them are set */
6197 6197 return (EINVAL);
6198 6198 }
6199 6199 }
6200 6200
6201 6201 ASSERT(MUTEX_HELD(&connp->conn_lock));
6202 6202
6203 6203 /*
6204 6204 * If we have already cached policies in conn_connect(), don't
6205 6205 * let them change now. We cache policies for connections
6206 6206 * whose src,dst [addr, port] is known.
6207 6207 */
6208 6208 if (connp->conn_policy_cached) {
6209 6209 return (EINVAL);
6210 6210 }
6211 6211
6212 6212 /*
6213 6213 * We have a zero policies, reset the connection policy if already
6214 6214 * set. This will cause the connection to inherit the
6215 6215 * global policy, if any.
6216 6216 */
6217 6217 if (is_pol_reset) {
6218 6218 if (connp->conn_policy != NULL) {
6219 6219 IPPH_REFRELE(connp->conn_policy, ipst->ips_netstack);
6220 6220 connp->conn_policy = NULL;
6221 6221 }
6222 6222 connp->conn_in_enforce_policy = B_FALSE;
6223 6223 connp->conn_out_enforce_policy = B_FALSE;
6224 6224 return (0);
6225 6225 }
6226 6226
6227 6227 ph = connp->conn_policy = ipsec_polhead_split(connp->conn_policy,
6228 6228 ipst->ips_netstack);
6229 6229 if (ph == NULL)
6230 6230 goto enomem;
6231 6231
6232 6232 ipsec_actvec_from_req(req, &actp, &nact, ipst->ips_netstack);
6233 6233 if (actp == NULL)
6234 6234 goto enomem;
6235 6235
6236 6236 /*
6237 6237 * Always insert IPv4 policy entries, since they can also apply to
6238 6238 * ipv6 sockets being used in ipv4-compat mode.
6239 6239 */
6240 6240 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V4,
6241 6241 IPSEC_TYPE_INBOUND, ns))
6242 6242 goto enomem;
6243 6243 is_pol_inserted = B_TRUE;
6244 6244 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V4,
6245 6245 IPSEC_TYPE_OUTBOUND, ns))
6246 6246 goto enomem;
6247 6247
6248 6248 /*
6249 6249 * We're looking at a v6 socket, also insert the v6-specific
6250 6250 * entries.
6251 6251 */
6252 6252 if (connp->conn_family == AF_INET6) {
6253 6253 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V6,
6254 6254 IPSEC_TYPE_INBOUND, ns))
6255 6255 goto enomem;
6256 6256 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V6,
6257 6257 IPSEC_TYPE_OUTBOUND, ns))
6258 6258 goto enomem;
6259 6259 }
6260 6260
6261 6261 ipsec_actvec_free(actp, nact);
6262 6262
6263 6263 /*
6264 6264 * If the requests need security, set enforce_policy.
6265 6265 * If the requests are IPSEC_PREF_NEVER, one should
6266 6266 * still set conn_out_enforce_policy so that ip_set_destination
6267 6267 * marks the ip_xmit_attr_t appropriatly. This is needed so that
6268 6268 * for connections that we don't cache policy in at connect time,
6269 6269 * if global policy matches in ip_output_attach_policy, we
6270 6270 * don't wrongly inherit global policy. Similarly, we need
6271 6271 * to set conn_in_enforce_policy also so that we don't verify
6272 6272 * policy wrongly.
6273 6273 */
6274 6274 if ((ah_req & REQ_MASK) != 0 ||
6275 6275 (esp_req & REQ_MASK) != 0 ||
6276 6276 (se_req & REQ_MASK) != 0) {
6277 6277 connp->conn_in_enforce_policy = B_TRUE;
6278 6278 connp->conn_out_enforce_policy = B_TRUE;
6279 6279 }
6280 6280
6281 6281 return (error);
6282 6282 #undef REQ_MASK
6283 6283
6284 6284 /*
6285 6285 * Common memory-allocation-failure exit path.
6286 6286 */
6287 6287 enomem:
6288 6288 if (actp != NULL)
6289 6289 ipsec_actvec_free(actp, nact);
6290 6290 if (is_pol_inserted)
6291 6291 ipsec_polhead_flush(ph, ns);
6292 6292 return (ENOMEM);
6293 6293 }
6294 6294
6295 6295 /*
6296 6296 * Set socket options for joining and leaving multicast groups.
6297 6297 * Common to IPv4 and IPv6; inet6 indicates the type of socket.
6298 6298 * The caller has already check that the option name is consistent with
6299 6299 * the address family of the socket.
6300 6300 */
6301 6301 int
6302 6302 ip_opt_set_multicast_group(conn_t *connp, t_scalar_t name,
6303 6303 uchar_t *invalp, boolean_t inet6, boolean_t checkonly)
6304 6304 {
6305 6305 int *i1 = (int *)invalp;
6306 6306 int error = 0;
6307 6307 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
6308 6308 struct ip_mreq *v4_mreqp;
6309 6309 struct ipv6_mreq *v6_mreqp;
6310 6310 struct group_req *greqp;
6311 6311 ire_t *ire;
6312 6312 boolean_t done = B_FALSE;
6313 6313 ipaddr_t ifaddr;
6314 6314 in6_addr_t v6group;
6315 6315 uint_t ifindex;
6316 6316 boolean_t mcast_opt = B_TRUE;
6317 6317 mcast_record_t fmode;
6318 6318 int (*optfn)(conn_t *, boolean_t, const in6_addr_t *,
6319 6319 ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *);
6320 6320
6321 6321 switch (name) {
6322 6322 case IP_ADD_MEMBERSHIP:
6323 6323 case IPV6_JOIN_GROUP:
6324 6324 mcast_opt = B_FALSE;
6325 6325 /* FALLTHROUGH */
6326 6326 case MCAST_JOIN_GROUP:
6327 6327 fmode = MODE_IS_EXCLUDE;
6328 6328 optfn = ip_opt_add_group;
6329 6329 break;
6330 6330
6331 6331 case IP_DROP_MEMBERSHIP:
6332 6332 case IPV6_LEAVE_GROUP:
6333 6333 mcast_opt = B_FALSE;
6334 6334 /* FALLTHROUGH */
6335 6335 case MCAST_LEAVE_GROUP:
6336 6336 fmode = MODE_IS_INCLUDE;
6337 6337 optfn = ip_opt_delete_group;
6338 6338 break;
6339 6339 default:
6340 6340 ASSERT(0);
6341 6341 }
6342 6342
6343 6343 if (mcast_opt) {
6344 6344 struct sockaddr_in *sin;
6345 6345 struct sockaddr_in6 *sin6;
6346 6346
6347 6347 greqp = (struct group_req *)i1;
6348 6348 if (greqp->gr_group.ss_family == AF_INET) {
6349 6349 sin = (struct sockaddr_in *)&(greqp->gr_group);
6350 6350 IN6_INADDR_TO_V4MAPPED(&sin->sin_addr, &v6group);
6351 6351 } else {
6352 6352 if (!inet6)
6353 6353 return (EINVAL); /* Not on INET socket */
6354 6354
6355 6355 sin6 = (struct sockaddr_in6 *)&(greqp->gr_group);
6356 6356 v6group = sin6->sin6_addr;
6357 6357 }
6358 6358 ifaddr = INADDR_ANY;
6359 6359 ifindex = greqp->gr_interface;
6360 6360 } else if (inet6) {
6361 6361 v6_mreqp = (struct ipv6_mreq *)i1;
6362 6362 v6group = v6_mreqp->ipv6mr_multiaddr;
6363 6363 ifaddr = INADDR_ANY;
6364 6364 ifindex = v6_mreqp->ipv6mr_interface;
6365 6365 } else {
6366 6366 v4_mreqp = (struct ip_mreq *)i1;
6367 6367 IN6_INADDR_TO_V4MAPPED(&v4_mreqp->imr_multiaddr, &v6group);
6368 6368 ifaddr = (ipaddr_t)v4_mreqp->imr_interface.s_addr;
6369 6369 ifindex = 0;
6370 6370 }
6371 6371
6372 6372 /*
6373 6373 * In the multirouting case, we need to replicate
6374 6374 * the request on all interfaces that will take part
6375 6375 * in replication. We do so because multirouting is
6376 6376 * reflective, thus we will probably receive multi-
6377 6377 * casts on those interfaces.
6378 6378 * The ip_multirt_apply_membership() succeeds if
6379 6379 * the operation succeeds on at least one interface.
6380 6380 */
6381 6381 if (IN6_IS_ADDR_V4MAPPED(&v6group)) {
6382 6382 ipaddr_t group;
6383 6383
6384 6384 IN6_V4MAPPED_TO_IPADDR(&v6group, group);
6385 6385
6386 6386 ire = ire_ftable_lookup_v4(group, IP_HOST_MASK, 0,
6387 6387 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6388 6388 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6389 6389 } else {
6390 6390 ire = ire_ftable_lookup_v6(&v6group, &ipv6_all_ones, 0,
6391 6391 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6392 6392 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6393 6393 }
6394 6394 if (ire != NULL) {
6395 6395 if (ire->ire_flags & RTF_MULTIRT) {
6396 6396 error = ip_multirt_apply_membership(optfn, ire, connp,
6397 6397 checkonly, &v6group, fmode, &ipv6_all_zeros);
6398 6398 done = B_TRUE;
6399 6399 }
6400 6400 ire_refrele(ire);
6401 6401 }
6402 6402
6403 6403 if (!done) {
6404 6404 error = optfn(connp, checkonly, &v6group, ifaddr, ifindex,
6405 6405 fmode, &ipv6_all_zeros);
6406 6406 }
6407 6407 return (error);
6408 6408 }
6409 6409
6410 6410 /*
6411 6411 * Set socket options for joining and leaving multicast groups
6412 6412 * for specific sources.
6413 6413 * Common to IPv4 and IPv6; inet6 indicates the type of socket.
6414 6414 * The caller has already check that the option name is consistent with
6415 6415 * the address family of the socket.
6416 6416 */
6417 6417 int
6418 6418 ip_opt_set_multicast_sources(conn_t *connp, t_scalar_t name,
6419 6419 uchar_t *invalp, boolean_t inet6, boolean_t checkonly)
6420 6420 {
6421 6421 int *i1 = (int *)invalp;
6422 6422 int error = 0;
6423 6423 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
6424 6424 struct ip_mreq_source *imreqp;
6425 6425 struct group_source_req *gsreqp;
6426 6426 in6_addr_t v6group, v6src;
6427 6427 uint32_t ifindex;
6428 6428 ipaddr_t ifaddr;
6429 6429 boolean_t mcast_opt = B_TRUE;
6430 6430 mcast_record_t fmode;
6431 6431 ire_t *ire;
6432 6432 boolean_t done = B_FALSE;
6433 6433 int (*optfn)(conn_t *, boolean_t, const in6_addr_t *,
6434 6434 ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *);
6435 6435
6436 6436 switch (name) {
6437 6437 case IP_BLOCK_SOURCE:
6438 6438 mcast_opt = B_FALSE;
6439 6439 /* FALLTHROUGH */
6440 6440 case MCAST_BLOCK_SOURCE:
6441 6441 fmode = MODE_IS_EXCLUDE;
6442 6442 optfn = ip_opt_add_group;
6443 6443 break;
6444 6444
6445 6445 case IP_UNBLOCK_SOURCE:
6446 6446 mcast_opt = B_FALSE;
6447 6447 /* FALLTHROUGH */
6448 6448 case MCAST_UNBLOCK_SOURCE:
6449 6449 fmode = MODE_IS_EXCLUDE;
6450 6450 optfn = ip_opt_delete_group;
6451 6451 break;
6452 6452
6453 6453 case IP_ADD_SOURCE_MEMBERSHIP:
6454 6454 mcast_opt = B_FALSE;
6455 6455 /* FALLTHROUGH */
6456 6456 case MCAST_JOIN_SOURCE_GROUP:
6457 6457 fmode = MODE_IS_INCLUDE;
6458 6458 optfn = ip_opt_add_group;
6459 6459 break;
6460 6460
6461 6461 case IP_DROP_SOURCE_MEMBERSHIP:
6462 6462 mcast_opt = B_FALSE;
6463 6463 /* FALLTHROUGH */
6464 6464 case MCAST_LEAVE_SOURCE_GROUP:
6465 6465 fmode = MODE_IS_INCLUDE;
6466 6466 optfn = ip_opt_delete_group;
6467 6467 break;
6468 6468 default:
6469 6469 ASSERT(0);
6470 6470 }
6471 6471
6472 6472 if (mcast_opt) {
6473 6473 gsreqp = (struct group_source_req *)i1;
6474 6474 ifindex = gsreqp->gsr_interface;
6475 6475 if (gsreqp->gsr_group.ss_family == AF_INET) {
6476 6476 struct sockaddr_in *s;
6477 6477 s = (struct sockaddr_in *)&gsreqp->gsr_group;
6478 6478 IN6_INADDR_TO_V4MAPPED(&s->sin_addr, &v6group);
6479 6479 s = (struct sockaddr_in *)&gsreqp->gsr_source;
6480 6480 IN6_INADDR_TO_V4MAPPED(&s->sin_addr, &v6src);
6481 6481 } else {
6482 6482 struct sockaddr_in6 *s6;
6483 6483
6484 6484 if (!inet6)
6485 6485 return (EINVAL); /* Not on INET socket */
6486 6486
6487 6487 s6 = (struct sockaddr_in6 *)&gsreqp->gsr_group;
6488 6488 v6group = s6->sin6_addr;
6489 6489 s6 = (struct sockaddr_in6 *)&gsreqp->gsr_source;
6490 6490 v6src = s6->sin6_addr;
6491 6491 }
6492 6492 ifaddr = INADDR_ANY;
6493 6493 } else {
6494 6494 imreqp = (struct ip_mreq_source *)i1;
6495 6495 IN6_INADDR_TO_V4MAPPED(&imreqp->imr_multiaddr, &v6group);
6496 6496 IN6_INADDR_TO_V4MAPPED(&imreqp->imr_sourceaddr, &v6src);
6497 6497 ifaddr = (ipaddr_t)imreqp->imr_interface.s_addr;
6498 6498 ifindex = 0;
6499 6499 }
6500 6500
6501 6501 /*
6502 6502 * Handle src being mapped INADDR_ANY by changing it to unspecified.
6503 6503 */
6504 6504 if (IN6_IS_ADDR_V4MAPPED_ANY(&v6src))
6505 6505 v6src = ipv6_all_zeros;
6506 6506
6507 6507 /*
6508 6508 * In the multirouting case, we need to replicate
6509 6509 * the request as noted in the mcast cases above.
6510 6510 */
6511 6511 if (IN6_IS_ADDR_V4MAPPED(&v6group)) {
6512 6512 ipaddr_t group;
6513 6513
6514 6514 IN6_V4MAPPED_TO_IPADDR(&v6group, group);
6515 6515
6516 6516 ire = ire_ftable_lookup_v4(group, IP_HOST_MASK, 0,
6517 6517 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6518 6518 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6519 6519 } else {
6520 6520 ire = ire_ftable_lookup_v6(&v6group, &ipv6_all_ones, 0,
6521 6521 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6522 6522 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6523 6523 }
6524 6524 if (ire != NULL) {
6525 6525 if (ire->ire_flags & RTF_MULTIRT) {
6526 6526 error = ip_multirt_apply_membership(optfn, ire, connp,
6527 6527 checkonly, &v6group, fmode, &v6src);
6528 6528 done = B_TRUE;
6529 6529 }
6530 6530 ire_refrele(ire);
6531 6531 }
6532 6532 if (!done) {
6533 6533 error = optfn(connp, checkonly, &v6group, ifaddr, ifindex,
6534 6534 fmode, &v6src);
6535 6535 }
6536 6536 return (error);
6537 6537 }
6538 6538
6539 6539 /*
6540 6540 * Given a destination address and a pointer to where to put the information
6541 6541 * this routine fills in the mtuinfo.
6542 6542 * The socket must be connected.
6543 6543 * For sctp conn_faddr is the primary address.
6544 6544 */
6545 6545 int
6546 6546 ip_fill_mtuinfo(conn_t *connp, ip_xmit_attr_t *ixa, struct ip6_mtuinfo *mtuinfo)
6547 6547 {
6548 6548 uint32_t pmtu = IP_MAXPACKET;
6549 6549 uint_t scopeid;
6550 6550
6551 6551 if (IN6_IS_ADDR_UNSPECIFIED(&connp->conn_faddr_v6))
6552 6552 return (-1);
6553 6553
6554 6554 /* In case we never sent or called ip_set_destination_v4/v6 */
6555 6555 if (ixa->ixa_ire != NULL)
6556 6556 pmtu = ip_get_pmtu(ixa);
6557 6557
6558 6558 if (ixa->ixa_flags & IXAF_SCOPEID_SET)
6559 6559 scopeid = ixa->ixa_scopeid;
6560 6560 else
6561 6561 scopeid = 0;
6562 6562
6563 6563 bzero(mtuinfo, sizeof (*mtuinfo));
6564 6564 mtuinfo->ip6m_addr.sin6_family = AF_INET6;
6565 6565 mtuinfo->ip6m_addr.sin6_port = connp->conn_fport;
6566 6566 mtuinfo->ip6m_addr.sin6_addr = connp->conn_faddr_v6;
6567 6567 mtuinfo->ip6m_addr.sin6_scope_id = scopeid;
6568 6568 mtuinfo->ip6m_mtu = pmtu;
6569 6569
6570 6570 return (sizeof (struct ip6_mtuinfo));
6571 6571 }
6572 6572
6573 6573 /*
6574 6574 * When the src multihoming is changed from weak to [strong, preferred]
6575 6575 * ip_ire_rebind_walker is called to walk the list of all ire_t entries
6576 6576 * and identify routes that were created by user-applications in the
6577 6577 * unbound state (i.e., without RTA_IFP), and for which an ire_ill is not
6578 6578 * currently defined. These routes are then 'rebound', i.e., their ire_ill
6579 6579 * is selected by finding an interface route for the gateway.
6580 6580 */
6581 6581 /* ARGSUSED */
6582 6582 void
6583 6583 ip_ire_rebind_walker(ire_t *ire, void *notused)
6584 6584 {
6585 6585 if (!ire->ire_unbound || ire->ire_ill != NULL)
6586 6586 return;
6587 6587 ire_rebind(ire);
6588 6588 ire_delete(ire);
6589 6589 }
6590 6590
6591 6591 /*
6592 6592 * When the src multihoming is changed from [strong, preferred] to weak,
6593 6593 * ip_ire_unbind_walker is called to walk the list of all ire_t entries, and
6594 6594 * set any entries that were created by user-applications in the unbound state
6595 6595 * (i.e., without RTA_IFP) back to having a NULL ire_ill.
6596 6596 */
6597 6597 /* ARGSUSED */
6598 6598 void
6599 6599 ip_ire_unbind_walker(ire_t *ire, void *notused)
6600 6600 {
6601 6601 ire_t *new_ire;
6602 6602
6603 6603 if (!ire->ire_unbound || ire->ire_ill == NULL)
6604 6604 return;
6605 6605 if (ire->ire_ipversion == IPV6_VERSION) {
6606 6606 new_ire = ire_create_v6(&ire->ire_addr_v6, &ire->ire_mask_v6,
6607 6607 &ire->ire_gateway_addr_v6, ire->ire_type, NULL,
6608 6608 ire->ire_zoneid, ire->ire_flags, NULL, ire->ire_ipst);
6609 6609 } else {
6610 6610 new_ire = ire_create((uchar_t *)&ire->ire_addr,
6611 6611 (uchar_t *)&ire->ire_mask,
6612 6612 (uchar_t *)&ire->ire_gateway_addr, ire->ire_type, NULL,
6613 6613 ire->ire_zoneid, ire->ire_flags, NULL, ire->ire_ipst);
6614 6614 }
6615 6615 if (new_ire == NULL)
6616 6616 return;
6617 6617 new_ire->ire_unbound = B_TRUE;
6618 6618 /*
6619 6619 * The bound ire must first be deleted so that we don't return
6620 6620 * the existing one on the attempt to add the unbound new_ire.
6621 6621 */
6622 6622 ire_delete(ire);
6623 6623 new_ire = ire_add(new_ire);
6624 6624 if (new_ire != NULL)
6625 6625 ire_refrele(new_ire);
6626 6626 }
6627 6627
6628 6628 /*
6629 6629 * When the settings of ip*_strict_src_multihoming tunables are changed,
6630 6630 * all cached routes need to be recomputed. This recomputation needs to be
6631 6631 * done when going from weaker to stronger modes so that the cached ire
6632 6632 * for the connection does not violate the current ip*_strict_src_multihoming
6633 6633 * setting. It also needs to be done when going from stronger to weaker modes,
6634 6634 * so that we fall back to matching on the longest-matching-route (as opposed
6635 6635 * to a shorter match that may have been selected in the strong mode
6636 6636 * to satisfy src_multihoming settings).
6637 6637 *
6638 6638 * The cached ixa_ire entires for all conn_t entries are marked as
6639 6639 * "verify" so that they will be recomputed for the next packet.
6640 6640 */
6641 6641 void
6642 6642 conn_ire_revalidate(conn_t *connp, void *arg)
6643 6643 {
6644 6644 boolean_t isv6 = (boolean_t)arg;
6645 6645
6646 6646 if ((isv6 && connp->conn_ipversion != IPV6_VERSION) ||
6647 6647 (!isv6 && connp->conn_ipversion != IPV4_VERSION))
6648 6648 return;
6649 6649 connp->conn_ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
6650 6650 }
6651 6651
6652 6652 /*
6653 6653 * Handles both IPv4 and IPv6 reassembly - doing the out-of-order cases,
6654 6654 * When an ipf is passed here for the first time, if
6655 6655 * we already have in-order fragments on the queue, we convert from the fast-
6656 6656 * path reassembly scheme to the hard-case scheme. From then on, additional
6657 6657 * fragments are reassembled here. We keep track of the start and end offsets
6658 6658 * of each piece, and the number of holes in the chain. When the hole count
6659 6659 * goes to zero, we are done!
6660 6660 *
6661 6661 * The ipf_count will be updated to account for any mblk(s) added (pointed to
6662 6662 * by mp) or subtracted (freeb()ed dups), upon return the caller must update
6663 6663 * ipfb_count and ill_frag_count by the difference of ipf_count before and
6664 6664 * after the call to ip_reassemble().
6665 6665 */
6666 6666 int
6667 6667 ip_reassemble(mblk_t *mp, ipf_t *ipf, uint_t start, boolean_t more, ill_t *ill,
6668 6668 size_t msg_len)
6669 6669 {
6670 6670 uint_t end;
6671 6671 mblk_t *next_mp;
6672 6672 mblk_t *mp1;
6673 6673 uint_t offset;
6674 6674 boolean_t incr_dups = B_TRUE;
6675 6675 boolean_t offset_zero_seen = B_FALSE;
6676 6676 boolean_t pkt_boundary_checked = B_FALSE;
6677 6677
6678 6678 /* If start == 0 then ipf_nf_hdr_len has to be set. */
6679 6679 ASSERT(start != 0 || ipf->ipf_nf_hdr_len != 0);
6680 6680
6681 6681 /* Add in byte count */
6682 6682 ipf->ipf_count += msg_len;
6683 6683 if (ipf->ipf_end) {
6684 6684 /*
6685 6685 * We were part way through in-order reassembly, but now there
6686 6686 * is a hole. We walk through messages already queued, and
6687 6687 * mark them for hard case reassembly. We know that up till
6688 6688 * now they were in order starting from offset zero.
6689 6689 */
6690 6690 offset = 0;
6691 6691 for (mp1 = ipf->ipf_mp->b_cont; mp1; mp1 = mp1->b_cont) {
6692 6692 IP_REASS_SET_START(mp1, offset);
6693 6693 if (offset == 0) {
6694 6694 ASSERT(ipf->ipf_nf_hdr_len != 0);
6695 6695 offset = -ipf->ipf_nf_hdr_len;
6696 6696 }
6697 6697 offset += mp1->b_wptr - mp1->b_rptr;
6698 6698 IP_REASS_SET_END(mp1, offset);
6699 6699 }
6700 6700 /* One hole at the end. */
6701 6701 ipf->ipf_hole_cnt = 1;
6702 6702 /* Brand it as a hard case, forever. */
6703 6703 ipf->ipf_end = 0;
6704 6704 }
6705 6705 /* Walk through all the new pieces. */
6706 6706 do {
6707 6707 end = start + (mp->b_wptr - mp->b_rptr);
6708 6708 /*
6709 6709 * If start is 0, decrease 'end' only for the first mblk of
6710 6710 * the fragment. Otherwise 'end' can get wrong value in the
6711 6711 * second pass of the loop if first mblk is exactly the
6712 6712 * size of ipf_nf_hdr_len.
6713 6713 */
6714 6714 if (start == 0 && !offset_zero_seen) {
6715 6715 /* First segment */
6716 6716 ASSERT(ipf->ipf_nf_hdr_len != 0);
6717 6717 end -= ipf->ipf_nf_hdr_len;
6718 6718 offset_zero_seen = B_TRUE;
6719 6719 }
6720 6720 next_mp = mp->b_cont;
6721 6721 /*
6722 6722 * We are checking to see if there is any interesing data
6723 6723 * to process. If there isn't and the mblk isn't the
6724 6724 * one which carries the unfragmentable header then we
6725 6725 * drop it. It's possible to have just the unfragmentable
6726 6726 * header come through without any data. That needs to be
6727 6727 * saved.
6728 6728 *
6729 6729 * If the assert at the top of this function holds then the
6730 6730 * term "ipf->ipf_nf_hdr_len != 0" isn't needed. This code
6731 6731 * is infrequently traveled enough that the test is left in
6732 6732 * to protect against future code changes which break that
6733 6733 * invariant.
6734 6734 */
6735 6735 if (start == end && start != 0 && ipf->ipf_nf_hdr_len != 0) {
6736 6736 /* Empty. Blast it. */
6737 6737 IP_REASS_SET_START(mp, 0);
6738 6738 IP_REASS_SET_END(mp, 0);
6739 6739 /*
6740 6740 * If the ipf points to the mblk we are about to free,
6741 6741 * update ipf to point to the next mblk (or NULL
6742 6742 * if none).
6743 6743 */
6744 6744 if (ipf->ipf_mp->b_cont == mp)
6745 6745 ipf->ipf_mp->b_cont = next_mp;
6746 6746 freeb(mp);
6747 6747 continue;
6748 6748 }
6749 6749 mp->b_cont = NULL;
6750 6750 IP_REASS_SET_START(mp, start);
6751 6751 IP_REASS_SET_END(mp, end);
6752 6752 if (!ipf->ipf_tail_mp) {
6753 6753 ipf->ipf_tail_mp = mp;
6754 6754 ipf->ipf_mp->b_cont = mp;
6755 6755 if (start == 0 || !more) {
6756 6756 ipf->ipf_hole_cnt = 1;
6757 6757 /*
6758 6758 * if the first fragment comes in more than one
6759 6759 * mblk, this loop will be executed for each
6760 6760 * mblk. Need to adjust hole count so exiting
6761 6761 * this routine will leave hole count at 1.
6762 6762 */
6763 6763 if (next_mp)
6764 6764 ipf->ipf_hole_cnt++;
6765 6765 } else
6766 6766 ipf->ipf_hole_cnt = 2;
6767 6767 continue;
6768 6768 } else if (ipf->ipf_last_frag_seen && !more &&
6769 6769 !pkt_boundary_checked) {
6770 6770 /*
6771 6771 * We check datagram boundary only if this fragment
6772 6772 * claims to be the last fragment and we have seen a
6773 6773 * last fragment in the past too. We do this only
6774 6774 * once for a given fragment.
6775 6775 *
6776 6776 * start cannot be 0 here as fragments with start=0
6777 6777 * and MF=0 gets handled as a complete packet. These
6778 6778 * fragments should not reach here.
6779 6779 */
6780 6780
6781 6781 if (start + msgdsize(mp) !=
6782 6782 IP_REASS_END(ipf->ipf_tail_mp)) {
6783 6783 /*
6784 6784 * We have two fragments both of which claim
6785 6785 * to be the last fragment but gives conflicting
6786 6786 * information about the whole datagram size.
6787 6787 * Something fishy is going on. Drop the
6788 6788 * fragment and free up the reassembly list.
6789 6789 */
6790 6790 return (IP_REASS_FAILED);
6791 6791 }
6792 6792
6793 6793 /*
6794 6794 * We shouldn't come to this code block again for this
6795 6795 * particular fragment.
6796 6796 */
6797 6797 pkt_boundary_checked = B_TRUE;
6798 6798 }
6799 6799
6800 6800 /* New stuff at or beyond tail? */
6801 6801 offset = IP_REASS_END(ipf->ipf_tail_mp);
6802 6802 if (start >= offset) {
6803 6803 if (ipf->ipf_last_frag_seen) {
6804 6804 /* current fragment is beyond last fragment */
6805 6805 return (IP_REASS_FAILED);
6806 6806 }
6807 6807 /* Link it on end. */
6808 6808 ipf->ipf_tail_mp->b_cont = mp;
6809 6809 ipf->ipf_tail_mp = mp;
6810 6810 if (more) {
6811 6811 if (start != offset)
6812 6812 ipf->ipf_hole_cnt++;
6813 6813 } else if (start == offset && next_mp == NULL)
6814 6814 ipf->ipf_hole_cnt--;
6815 6815 continue;
6816 6816 }
6817 6817 mp1 = ipf->ipf_mp->b_cont;
6818 6818 offset = IP_REASS_START(mp1);
6819 6819 /* New stuff at the front? */
6820 6820 if (start < offset) {
6821 6821 if (start == 0) {
6822 6822 if (end >= offset) {
6823 6823 /* Nailed the hole at the begining. */
6824 6824 ipf->ipf_hole_cnt--;
6825 6825 }
6826 6826 } else if (end < offset) {
6827 6827 /*
6828 6828 * A hole, stuff, and a hole where there used
6829 6829 * to be just a hole.
6830 6830 */
6831 6831 ipf->ipf_hole_cnt++;
6832 6832 }
6833 6833 mp->b_cont = mp1;
6834 6834 /* Check for overlap. */
6835 6835 while (end > offset) {
6836 6836 if (end < IP_REASS_END(mp1)) {
6837 6837 mp->b_wptr -= end - offset;
6838 6838 IP_REASS_SET_END(mp, offset);
6839 6839 BUMP_MIB(ill->ill_ip_mib,
6840 6840 ipIfStatsReasmPartDups);
6841 6841 break;
6842 6842 }
6843 6843 /* Did we cover another hole? */
6844 6844 if ((mp1->b_cont &&
6845 6845 IP_REASS_END(mp1) !=
6846 6846 IP_REASS_START(mp1->b_cont) &&
6847 6847 end >= IP_REASS_START(mp1->b_cont)) ||
6848 6848 (!ipf->ipf_last_frag_seen && !more)) {
6849 6849 ipf->ipf_hole_cnt--;
6850 6850 }
6851 6851 /* Clip out mp1. */
6852 6852 if ((mp->b_cont = mp1->b_cont) == NULL) {
6853 6853 /*
6854 6854 * After clipping out mp1, this guy
6855 6855 * is now hanging off the end.
6856 6856 */
6857 6857 ipf->ipf_tail_mp = mp;
6858 6858 }
6859 6859 IP_REASS_SET_START(mp1, 0);
6860 6860 IP_REASS_SET_END(mp1, 0);
6861 6861 /* Subtract byte count */
6862 6862 ipf->ipf_count -= mp1->b_datap->db_lim -
6863 6863 mp1->b_datap->db_base;
6864 6864 freeb(mp1);
6865 6865 BUMP_MIB(ill->ill_ip_mib,
6866 6866 ipIfStatsReasmPartDups);
6867 6867 mp1 = mp->b_cont;
6868 6868 if (!mp1)
6869 6869 break;
6870 6870 offset = IP_REASS_START(mp1);
6871 6871 }
6872 6872 ipf->ipf_mp->b_cont = mp;
6873 6873 continue;
6874 6874 }
6875 6875 /*
6876 6876 * The new piece starts somewhere between the start of the head
6877 6877 * and before the end of the tail.
6878 6878 */
6879 6879 for (; mp1; mp1 = mp1->b_cont) {
6880 6880 offset = IP_REASS_END(mp1);
6881 6881 if (start < offset) {
6882 6882 if (end <= offset) {
6883 6883 /* Nothing new. */
6884 6884 IP_REASS_SET_START(mp, 0);
6885 6885 IP_REASS_SET_END(mp, 0);
6886 6886 /* Subtract byte count */
6887 6887 ipf->ipf_count -= mp->b_datap->db_lim -
6888 6888 mp->b_datap->db_base;
6889 6889 if (incr_dups) {
6890 6890 ipf->ipf_num_dups++;
6891 6891 incr_dups = B_FALSE;
6892 6892 }
6893 6893 freeb(mp);
6894 6894 BUMP_MIB(ill->ill_ip_mib,
6895 6895 ipIfStatsReasmDuplicates);
6896 6896 break;
6897 6897 }
6898 6898 /*
6899 6899 * Trim redundant stuff off beginning of new
6900 6900 * piece.
6901 6901 */
6902 6902 IP_REASS_SET_START(mp, offset);
6903 6903 mp->b_rptr += offset - start;
6904 6904 BUMP_MIB(ill->ill_ip_mib,
6905 6905 ipIfStatsReasmPartDups);
6906 6906 start = offset;
6907 6907 if (!mp1->b_cont) {
6908 6908 /*
6909 6909 * After trimming, this guy is now
6910 6910 * hanging off the end.
6911 6911 */
6912 6912 mp1->b_cont = mp;
6913 6913 ipf->ipf_tail_mp = mp;
6914 6914 if (!more) {
6915 6915 ipf->ipf_hole_cnt--;
6916 6916 }
6917 6917 break;
6918 6918 }
6919 6919 }
6920 6920 if (start >= IP_REASS_START(mp1->b_cont))
6921 6921 continue;
6922 6922 /* Fill a hole */
6923 6923 if (start > offset)
6924 6924 ipf->ipf_hole_cnt++;
6925 6925 mp->b_cont = mp1->b_cont;
6926 6926 mp1->b_cont = mp;
6927 6927 mp1 = mp->b_cont;
6928 6928 offset = IP_REASS_START(mp1);
6929 6929 if (end >= offset) {
6930 6930 ipf->ipf_hole_cnt--;
6931 6931 /* Check for overlap. */
6932 6932 while (end > offset) {
6933 6933 if (end < IP_REASS_END(mp1)) {
6934 6934 mp->b_wptr -= end - offset;
6935 6935 IP_REASS_SET_END(mp, offset);
6936 6936 /*
6937 6937 * TODO we might bump
6938 6938 * this up twice if there is
6939 6939 * overlap at both ends.
6940 6940 */
6941 6941 BUMP_MIB(ill->ill_ip_mib,
6942 6942 ipIfStatsReasmPartDups);
6943 6943 break;
6944 6944 }
6945 6945 /* Did we cover another hole? */
6946 6946 if ((mp1->b_cont &&
6947 6947 IP_REASS_END(mp1)
6948 6948 != IP_REASS_START(mp1->b_cont) &&
6949 6949 end >=
6950 6950 IP_REASS_START(mp1->b_cont)) ||
6951 6951 (!ipf->ipf_last_frag_seen &&
6952 6952 !more)) {
6953 6953 ipf->ipf_hole_cnt--;
6954 6954 }
6955 6955 /* Clip out mp1. */
6956 6956 if ((mp->b_cont = mp1->b_cont) ==
6957 6957 NULL) {
6958 6958 /*
6959 6959 * After clipping out mp1,
6960 6960 * this guy is now hanging
6961 6961 * off the end.
6962 6962 */
6963 6963 ipf->ipf_tail_mp = mp;
6964 6964 }
6965 6965 IP_REASS_SET_START(mp1, 0);
6966 6966 IP_REASS_SET_END(mp1, 0);
6967 6967 /* Subtract byte count */
6968 6968 ipf->ipf_count -=
6969 6969 mp1->b_datap->db_lim -
6970 6970 mp1->b_datap->db_base;
6971 6971 freeb(mp1);
6972 6972 BUMP_MIB(ill->ill_ip_mib,
6973 6973 ipIfStatsReasmPartDups);
6974 6974 mp1 = mp->b_cont;
6975 6975 if (!mp1)
6976 6976 break;
6977 6977 offset = IP_REASS_START(mp1);
6978 6978 }
6979 6979 }
6980 6980 break;
6981 6981 }
6982 6982 } while (start = end, mp = next_mp);
6983 6983
6984 6984 /* Fragment just processed could be the last one. Remember this fact */
6985 6985 if (!more)
6986 6986 ipf->ipf_last_frag_seen = B_TRUE;
6987 6987
6988 6988 /* Still got holes? */
6989 6989 if (ipf->ipf_hole_cnt)
6990 6990 return (IP_REASS_PARTIAL);
6991 6991 /* Clean up overloaded fields to avoid upstream disasters. */
6992 6992 for (mp1 = ipf->ipf_mp->b_cont; mp1; mp1 = mp1->b_cont) {
6993 6993 IP_REASS_SET_START(mp1, 0);
6994 6994 IP_REASS_SET_END(mp1, 0);
6995 6995 }
6996 6996 return (IP_REASS_COMPLETE);
6997 6997 }
6998 6998
6999 6999 /*
7000 7000 * Fragmentation reassembly. Each ILL has a hash table for
7001 7001 * queuing packets undergoing reassembly for all IPIFs
7002 7002 * associated with the ILL. The hash is based on the packet
7003 7003 * IP ident field. The ILL frag hash table was allocated
7004 7004 * as a timer block at the time the ILL was created. Whenever
7005 7005 * there is anything on the reassembly queue, the timer will
7006 7006 * be running. Returns the reassembled packet if reassembly completes.
7007 7007 */
7008 7008 mblk_t *
7009 7009 ip_input_fragment(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
7010 7010 {
7011 7011 uint32_t frag_offset_flags;
7012 7012 mblk_t *t_mp;
7013 7013 ipaddr_t dst;
7014 7014 uint8_t proto = ipha->ipha_protocol;
7015 7015 uint32_t sum_val;
7016 7016 uint16_t sum_flags;
7017 7017 ipf_t *ipf;
7018 7018 ipf_t **ipfp;
7019 7019 ipfb_t *ipfb;
7020 7020 uint16_t ident;
7021 7021 uint32_t offset;
7022 7022 ipaddr_t src;
7023 7023 uint_t hdr_length;
7024 7024 uint32_t end;
7025 7025 mblk_t *mp1;
7026 7026 mblk_t *tail_mp;
7027 7027 size_t count;
7028 7028 size_t msg_len;
7029 7029 uint8_t ecn_info = 0;
7030 7030 uint32_t packet_size;
7031 7031 boolean_t pruned = B_FALSE;
7032 7032 ill_t *ill = ira->ira_ill;
7033 7033 ip_stack_t *ipst = ill->ill_ipst;
7034 7034
7035 7035 /*
7036 7036 * Drop the fragmented as early as possible, if
7037 7037 * we don't have resource(s) to re-assemble.
7038 7038 */
7039 7039 if (ipst->ips_ip_reass_queue_bytes == 0) {
7040 7040 freemsg(mp);
7041 7041 return (NULL);
7042 7042 }
7043 7043
7044 7044 /* Check for fragmentation offset; return if there's none */
7045 7045 if ((frag_offset_flags = ntohs(ipha->ipha_fragment_offset_and_flags) &
7046 7046 (IPH_MF | IPH_OFFSET)) == 0)
7047 7047 return (mp);
7048 7048
7049 7049 /*
7050 7050 * We utilize hardware computed checksum info only for UDP since
7051 7051 * IP fragmentation is a normal occurrence for the protocol. In
7052 7052 * addition, checksum offload support for IP fragments carrying
7053 7053 * UDP payload is commonly implemented across network adapters.
7054 7054 */
7055 7055 ASSERT(ira->ira_rill != NULL);
7056 7056 if (proto == IPPROTO_UDP && dohwcksum &&
7057 7057 ILL_HCKSUM_CAPABLE(ira->ira_rill) &&
7058 7058 (DB_CKSUMFLAGS(mp) & (HCK_FULLCKSUM | HCK_PARTIALCKSUM))) {
7059 7059 mblk_t *mp1 = mp->b_cont;
7060 7060 int32_t len;
7061 7061
7062 7062 /* Record checksum information from the packet */
7063 7063 sum_val = (uint32_t)DB_CKSUM16(mp);
7064 7064 sum_flags = DB_CKSUMFLAGS(mp);
7065 7065
7066 7066 /* IP payload offset from beginning of mblk */
7067 7067 offset = ((uchar_t *)ipha + IPH_HDR_LENGTH(ipha)) - mp->b_rptr;
7068 7068
7069 7069 if ((sum_flags & HCK_PARTIALCKSUM) &&
7070 7070 (mp1 == NULL || mp1->b_cont == NULL) &&
7071 7071 offset >= DB_CKSUMSTART(mp) &&
7072 7072 ((len = offset - DB_CKSUMSTART(mp)) & 1) == 0) {
7073 7073 uint32_t adj;
7074 7074 /*
7075 7075 * Partial checksum has been calculated by hardware
7076 7076 * and attached to the packet; in addition, any
7077 7077 * prepended extraneous data is even byte aligned.
7078 7078 * If any such data exists, we adjust the checksum;
7079 7079 * this would also handle any postpended data.
7080 7080 */
7081 7081 IP_ADJCKSUM_PARTIAL(mp->b_rptr + DB_CKSUMSTART(mp),
7082 7082 mp, mp1, len, adj);
7083 7083
7084 7084 /* One's complement subtract extraneous checksum */
7085 7085 if (adj >= sum_val)
7086 7086 sum_val = ~(adj - sum_val) & 0xFFFF;
7087 7087 else
7088 7088 sum_val -= adj;
7089 7089 }
7090 7090 } else {
7091 7091 sum_val = 0;
7092 7092 sum_flags = 0;
7093 7093 }
7094 7094
7095 7095 /* Clear hardware checksumming flag */
7096 7096 DB_CKSUMFLAGS(mp) = 0;
7097 7097
7098 7098 ident = ipha->ipha_ident;
7099 7099 offset = (frag_offset_flags << 3) & 0xFFFF;
7100 7100 src = ipha->ipha_src;
7101 7101 dst = ipha->ipha_dst;
7102 7102 hdr_length = IPH_HDR_LENGTH(ipha);
7103 7103 end = ntohs(ipha->ipha_length) - hdr_length;
7104 7104
7105 7105 /* If end == 0 then we have a packet with no data, so just free it */
7106 7106 if (end == 0) {
7107 7107 freemsg(mp);
7108 7108 return (NULL);
7109 7109 }
7110 7110
7111 7111 /* Record the ECN field info. */
7112 7112 ecn_info = (ipha->ipha_type_of_service & 0x3);
7113 7113 if (offset != 0) {
7114 7114 /*
7115 7115 * If this isn't the first piece, strip the header, and
7116 7116 * add the offset to the end value.
7117 7117 */
7118 7118 mp->b_rptr += hdr_length;
7119 7119 end += offset;
7120 7120 }
7121 7121
7122 7122 /* Handle vnic loopback of fragments */
7123 7123 if (mp->b_datap->db_ref > 2)
7124 7124 msg_len = 0;
7125 7125 else
7126 7126 msg_len = MBLKSIZE(mp);
7127 7127
7128 7128 tail_mp = mp;
7129 7129 while (tail_mp->b_cont != NULL) {
7130 7130 tail_mp = tail_mp->b_cont;
7131 7131 if (tail_mp->b_datap->db_ref <= 2)
7132 7132 msg_len += MBLKSIZE(tail_mp);
7133 7133 }
7134 7134
7135 7135 /* If the reassembly list for this ILL will get too big, prune it */
7136 7136 if ((msg_len + sizeof (*ipf) + ill->ill_frag_count) >=
7137 7137 ipst->ips_ip_reass_queue_bytes) {
7138 7138 DTRACE_PROBE3(ip_reass_queue_bytes, uint_t, msg_len,
7139 7139 uint_t, ill->ill_frag_count,
7140 7140 uint_t, ipst->ips_ip_reass_queue_bytes);
7141 7141 ill_frag_prune(ill,
7142 7142 (ipst->ips_ip_reass_queue_bytes < msg_len) ? 0 :
7143 7143 (ipst->ips_ip_reass_queue_bytes - msg_len));
7144 7144 pruned = B_TRUE;
7145 7145 }
7146 7146
7147 7147 ipfb = &ill->ill_frag_hash_tbl[ILL_FRAG_HASH(src, ident)];
7148 7148 mutex_enter(&ipfb->ipfb_lock);
7149 7149
7150 7150 ipfp = &ipfb->ipfb_ipf;
7151 7151 /* Try to find an existing fragment queue for this packet. */
7152 7152 for (;;) {
7153 7153 ipf = ipfp[0];
7154 7154 if (ipf != NULL) {
7155 7155 /*
7156 7156 * It has to match on ident and src/dst address.
7157 7157 */
7158 7158 if (ipf->ipf_ident == ident &&
7159 7159 ipf->ipf_src == src &&
7160 7160 ipf->ipf_dst == dst &&
7161 7161 ipf->ipf_protocol == proto) {
7162 7162 /*
7163 7163 * If we have received too many
7164 7164 * duplicate fragments for this packet
7165 7165 * free it.
7166 7166 */
7167 7167 if (ipf->ipf_num_dups > ip_max_frag_dups) {
7168 7168 ill_frag_free_pkts(ill, ipfb, ipf, 1);
7169 7169 freemsg(mp);
7170 7170 mutex_exit(&ipfb->ipfb_lock);
7171 7171 return (NULL);
7172 7172 }
7173 7173 /* Found it. */
7174 7174 break;
7175 7175 }
7176 7176 ipfp = &ipf->ipf_hash_next;
7177 7177 continue;
7178 7178 }
7179 7179
7180 7180 /*
7181 7181 * If we pruned the list, do we want to store this new
7182 7182 * fragment?. We apply an optimization here based on the
7183 7183 * fact that most fragments will be received in order.
7184 7184 * So if the offset of this incoming fragment is zero,
7185 7185 * it is the first fragment of a new packet. We will
7186 7186 * keep it. Otherwise drop the fragment, as we have
7187 7187 * probably pruned the packet already (since the
7188 7188 * packet cannot be found).
7189 7189 */
7190 7190 if (pruned && offset != 0) {
7191 7191 mutex_exit(&ipfb->ipfb_lock);
7192 7192 freemsg(mp);
7193 7193 return (NULL);
7194 7194 }
7195 7195
7196 7196 if (ipfb->ipfb_frag_pkts >= MAX_FRAG_PKTS(ipst)) {
7197 7197 /*
7198 7198 * Too many fragmented packets in this hash
7199 7199 * bucket. Free the oldest.
7200 7200 */
7201 7201 ill_frag_free_pkts(ill, ipfb, ipfb->ipfb_ipf, 1);
7202 7202 }
7203 7203
7204 7204 /* New guy. Allocate a frag message. */
7205 7205 mp1 = allocb(sizeof (*ipf), BPRI_MED);
7206 7206 if (mp1 == NULL) {
7207 7207 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7208 7208 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7209 7209 freemsg(mp);
7210 7210 reass_done:
7211 7211 mutex_exit(&ipfb->ipfb_lock);
7212 7212 return (NULL);
7213 7213 }
7214 7214
7215 7215 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmReqds);
7216 7216 mp1->b_cont = mp;
7217 7217
7218 7218 /* Initialize the fragment header. */
7219 7219 ipf = (ipf_t *)mp1->b_rptr;
7220 7220 ipf->ipf_mp = mp1;
7221 7221 ipf->ipf_ptphn = ipfp;
7222 7222 ipfp[0] = ipf;
7223 7223 ipf->ipf_hash_next = NULL;
7224 7224 ipf->ipf_ident = ident;
7225 7225 ipf->ipf_protocol = proto;
7226 7226 ipf->ipf_src = src;
7227 7227 ipf->ipf_dst = dst;
7228 7228 ipf->ipf_nf_hdr_len = 0;
7229 7229 /* Record reassembly start time. */
7230 7230 ipf->ipf_timestamp = gethrestime_sec();
7231 7231 /* Record ipf generation and account for frag header */
7232 7232 ipf->ipf_gen = ill->ill_ipf_gen++;
7233 7233 ipf->ipf_count = MBLKSIZE(mp1);
7234 7234 ipf->ipf_last_frag_seen = B_FALSE;
7235 7235 ipf->ipf_ecn = ecn_info;
7236 7236 ipf->ipf_num_dups = 0;
7237 7237 ipfb->ipfb_frag_pkts++;
7238 7238 ipf->ipf_checksum = 0;
7239 7239 ipf->ipf_checksum_flags = 0;
7240 7240
7241 7241 /* Store checksum value in fragment header */
7242 7242 if (sum_flags != 0) {
7243 7243 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7244 7244 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7245 7245 ipf->ipf_checksum = sum_val;
7246 7246 ipf->ipf_checksum_flags = sum_flags;
7247 7247 }
7248 7248
7249 7249 /*
7250 7250 * We handle reassembly two ways. In the easy case,
7251 7251 * where all the fragments show up in order, we do
7252 7252 * minimal bookkeeping, and just clip new pieces on
7253 7253 * the end. If we ever see a hole, then we go off
7254 7254 * to ip_reassemble which has to mark the pieces and
7255 7255 * keep track of the number of holes, etc. Obviously,
7256 7256 * the point of having both mechanisms is so we can
7257 7257 * handle the easy case as efficiently as possible.
7258 7258 */
7259 7259 if (offset == 0) {
7260 7260 /* Easy case, in-order reassembly so far. */
7261 7261 ipf->ipf_count += msg_len;
7262 7262 ipf->ipf_tail_mp = tail_mp;
7263 7263 /*
7264 7264 * Keep track of next expected offset in
7265 7265 * ipf_end.
7266 7266 */
7267 7267 ipf->ipf_end = end;
7268 7268 ipf->ipf_nf_hdr_len = hdr_length;
7269 7269 } else {
7270 7270 /* Hard case, hole at the beginning. */
7271 7271 ipf->ipf_tail_mp = NULL;
7272 7272 /*
7273 7273 * ipf_end == 0 means that we have given up
7274 7274 * on easy reassembly.
7275 7275 */
7276 7276 ipf->ipf_end = 0;
7277 7277
7278 7278 /* Forget checksum offload from now on */
7279 7279 ipf->ipf_checksum_flags = 0;
7280 7280
7281 7281 /*
7282 7282 * ipf_hole_cnt is set by ip_reassemble.
7283 7283 * ipf_count is updated by ip_reassemble.
7284 7284 * No need to check for return value here
7285 7285 * as we don't expect reassembly to complete
7286 7286 * or fail for the first fragment itself.
7287 7287 */
7288 7288 (void) ip_reassemble(mp, ipf,
7289 7289 (frag_offset_flags & IPH_OFFSET) << 3,
7290 7290 (frag_offset_flags & IPH_MF), ill, msg_len);
7291 7291 }
7292 7292 /* Update per ipfb and ill byte counts */
7293 7293 ipfb->ipfb_count += ipf->ipf_count;
7294 7294 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7295 7295 atomic_add_32(&ill->ill_frag_count, ipf->ipf_count);
7296 7296 /* If the frag timer wasn't already going, start it. */
7297 7297 mutex_enter(&ill->ill_lock);
7298 7298 ill_frag_timer_start(ill);
7299 7299 mutex_exit(&ill->ill_lock);
7300 7300 goto reass_done;
7301 7301 }
7302 7302
7303 7303 /*
7304 7304 * If the packet's flag has changed (it could be coming up
7305 7305 * from an interface different than the previous, therefore
7306 7306 * possibly different checksum capability), then forget about
7307 7307 * any stored checksum states. Otherwise add the value to
7308 7308 * the existing one stored in the fragment header.
7309 7309 */
7310 7310 if (sum_flags != 0 && sum_flags == ipf->ipf_checksum_flags) {
7311 7311 sum_val += ipf->ipf_checksum;
7312 7312 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7313 7313 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7314 7314 ipf->ipf_checksum = sum_val;
7315 7315 } else if (ipf->ipf_checksum_flags != 0) {
7316 7316 /* Forget checksum offload from now on */
7317 7317 ipf->ipf_checksum_flags = 0;
7318 7318 }
7319 7319
7320 7320 /*
7321 7321 * We have a new piece of a datagram which is already being
7322 7322 * reassembled. Update the ECN info if all IP fragments
7323 7323 * are ECN capable. If there is one which is not, clear
7324 7324 * all the info. If there is at least one which has CE
7325 7325 * code point, IP needs to report that up to transport.
7326 7326 */
7327 7327 if (ecn_info != IPH_ECN_NECT && ipf->ipf_ecn != IPH_ECN_NECT) {
7328 7328 if (ecn_info == IPH_ECN_CE)
7329 7329 ipf->ipf_ecn = IPH_ECN_CE;
7330 7330 } else {
7331 7331 ipf->ipf_ecn = IPH_ECN_NECT;
7332 7332 }
7333 7333 if (offset && ipf->ipf_end == offset) {
7334 7334 /* The new fragment fits at the end */
7335 7335 ipf->ipf_tail_mp->b_cont = mp;
7336 7336 /* Update the byte count */
7337 7337 ipf->ipf_count += msg_len;
7338 7338 /* Update per ipfb and ill byte counts */
7339 7339 ipfb->ipfb_count += msg_len;
7340 7340 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7341 7341 atomic_add_32(&ill->ill_frag_count, msg_len);
7342 7342 if (frag_offset_flags & IPH_MF) {
7343 7343 /* More to come. */
7344 7344 ipf->ipf_end = end;
7345 7345 ipf->ipf_tail_mp = tail_mp;
7346 7346 goto reass_done;
7347 7347 }
7348 7348 } else {
7349 7349 /* Go do the hard cases. */
7350 7350 int ret;
7351 7351
7352 7352 if (offset == 0)
7353 7353 ipf->ipf_nf_hdr_len = hdr_length;
7354 7354
7355 7355 /* Save current byte count */
7356 7356 count = ipf->ipf_count;
7357 7357 ret = ip_reassemble(mp, ipf,
7358 7358 (frag_offset_flags & IPH_OFFSET) << 3,
7359 7359 (frag_offset_flags & IPH_MF), ill, msg_len);
7360 7360 /* Count of bytes added and subtracted (freeb()ed) */
7361 7361 count = ipf->ipf_count - count;
7362 7362 if (count) {
7363 7363 /* Update per ipfb and ill byte counts */
7364 7364 ipfb->ipfb_count += count;
7365 7365 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7366 7366 atomic_add_32(&ill->ill_frag_count, count);
7367 7367 }
7368 7368 if (ret == IP_REASS_PARTIAL) {
7369 7369 goto reass_done;
7370 7370 } else if (ret == IP_REASS_FAILED) {
7371 7371 /* Reassembly failed. Free up all resources */
7372 7372 ill_frag_free_pkts(ill, ipfb, ipf, 1);
7373 7373 for (t_mp = mp; t_mp != NULL; t_mp = t_mp->b_cont) {
7374 7374 IP_REASS_SET_START(t_mp, 0);
7375 7375 IP_REASS_SET_END(t_mp, 0);
7376 7376 }
7377 7377 freemsg(mp);
7378 7378 goto reass_done;
7379 7379 }
7380 7380 /* We will reach here iff 'ret' is IP_REASS_COMPLETE */
7381 7381 }
7382 7382 /*
7383 7383 * We have completed reassembly. Unhook the frag header from
7384 7384 * the reassembly list.
7385 7385 *
7386 7386 * Before we free the frag header, record the ECN info
7387 7387 * to report back to the transport.
7388 7388 */
7389 7389 ecn_info = ipf->ipf_ecn;
7390 7390 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmOKs);
7391 7391 ipfp = ipf->ipf_ptphn;
7392 7392
7393 7393 /* We need to supply these to caller */
7394 7394 if ((sum_flags = ipf->ipf_checksum_flags) != 0)
7395 7395 sum_val = ipf->ipf_checksum;
7396 7396 else
7397 7397 sum_val = 0;
7398 7398
7399 7399 mp1 = ipf->ipf_mp;
7400 7400 count = ipf->ipf_count;
7401 7401 ipf = ipf->ipf_hash_next;
7402 7402 if (ipf != NULL)
7403 7403 ipf->ipf_ptphn = ipfp;
7404 7404 ipfp[0] = ipf;
7405 7405 atomic_add_32(&ill->ill_frag_count, -count);
7406 7406 ASSERT(ipfb->ipfb_count >= count);
7407 7407 ipfb->ipfb_count -= count;
7408 7408 ipfb->ipfb_frag_pkts--;
7409 7409 mutex_exit(&ipfb->ipfb_lock);
7410 7410 /* Ditch the frag header. */
7411 7411 mp = mp1->b_cont;
7412 7412
7413 7413 freeb(mp1);
7414 7414
7415 7415 /* Restore original IP length in header. */
7416 7416 packet_size = (uint32_t)msgdsize(mp);
7417 7417 if (packet_size > IP_MAXPACKET) {
7418 7418 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7419 7419 ip_drop_input("Reassembled packet too large", mp, ill);
7420 7420 freemsg(mp);
7421 7421 return (NULL);
7422 7422 }
7423 7423
7424 7424 if (DB_REF(mp) > 1) {
7425 7425 mblk_t *mp2 = copymsg(mp);
7426 7426
7427 7427 if (mp2 == NULL) {
7428 7428 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7429 7429 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7430 7430 freemsg(mp);
7431 7431 return (NULL);
7432 7432 }
7433 7433 freemsg(mp);
7434 7434 mp = mp2;
7435 7435 }
7436 7436 ipha = (ipha_t *)mp->b_rptr;
7437 7437
7438 7438 ipha->ipha_length = htons((uint16_t)packet_size);
7439 7439 /* We're now complete, zip the frag state */
7440 7440 ipha->ipha_fragment_offset_and_flags = 0;
7441 7441 /* Record the ECN info. */
7442 7442 ipha->ipha_type_of_service &= 0xFC;
7443 7443 ipha->ipha_type_of_service |= ecn_info;
7444 7444
7445 7445 /* Update the receive attributes */
7446 7446 ira->ira_pktlen = packet_size;
7447 7447 ira->ira_ip_hdr_length = IPH_HDR_LENGTH(ipha);
7448 7448
7449 7449 /* Reassembly is successful; set checksum information in packet */
7450 7450 DB_CKSUM16(mp) = (uint16_t)sum_val;
7451 7451 DB_CKSUMFLAGS(mp) = sum_flags;
7452 7452 DB_CKSUMSTART(mp) = ira->ira_ip_hdr_length;
7453 7453
7454 7454 return (mp);
7455 7455 }
7456 7456
7457 7457 /*
7458 7458 * Pullup function that should be used for IP input in order to
7459 7459 * ensure we do not loose the L2 source address; we need the l2 source
7460 7460 * address for IP_RECVSLLA and for ndp_input.
7461 7461 *
7462 7462 * We return either NULL or b_rptr.
7463 7463 */
7464 7464 void *
7465 7465 ip_pullup(mblk_t *mp, ssize_t len, ip_recv_attr_t *ira)
7466 7466 {
7467 7467 ill_t *ill = ira->ira_ill;
7468 7468
7469 7469 if (ip_rput_pullups++ == 0) {
7470 7470 (void) mi_strlog(ill->ill_rq, 1, SL_ERROR|SL_TRACE,
7471 7471 "ip_pullup: %s forced us to "
7472 7472 " pullup pkt, hdr len %ld, hdr addr %p",
7473 7473 ill->ill_name, len, (void *)mp->b_rptr);
7474 7474 }
7475 7475 if (!(ira->ira_flags & IRAF_L2SRC_SET))
7476 7476 ip_setl2src(mp, ira, ira->ira_rill);
7477 7477 ASSERT(ira->ira_flags & IRAF_L2SRC_SET);
7478 7478 if (!pullupmsg(mp, len))
7479 7479 return (NULL);
7480 7480 else
7481 7481 return (mp->b_rptr);
7482 7482 }
7483 7483
7484 7484 /*
7485 7485 * Make sure ira_l2src has an address. If we don't have one fill with zeros.
7486 7486 * When called from the ULP ira_rill will be NULL hence the caller has to
7487 7487 * pass in the ill.
7488 7488 */
7489 7489 /* ARGSUSED */
7490 7490 void
7491 7491 ip_setl2src(mblk_t *mp, ip_recv_attr_t *ira, ill_t *ill)
7492 7492 {
7493 7493 const uchar_t *addr;
7494 7494 int alen;
7495 7495
7496 7496 if (ira->ira_flags & IRAF_L2SRC_SET)
7497 7497 return;
7498 7498
7499 7499 ASSERT(ill != NULL);
7500 7500 alen = ill->ill_phys_addr_length;
7501 7501 ASSERT(alen <= sizeof (ira->ira_l2src));
7502 7502 if (ira->ira_mhip != NULL &&
7503 7503 (addr = ira->ira_mhip->mhi_saddr) != NULL) {
7504 7504 bcopy(addr, ira->ira_l2src, alen);
7505 7505 } else if ((ira->ira_flags & IRAF_L2SRC_LOOPBACK) &&
7506 7506 (addr = ill->ill_phys_addr) != NULL) {
7507 7507 bcopy(addr, ira->ira_l2src, alen);
7508 7508 } else {
7509 7509 bzero(ira->ira_l2src, alen);
7510 7510 }
7511 7511 ira->ira_flags |= IRAF_L2SRC_SET;
7512 7512 }
7513 7513
7514 7514 /*
7515 7515 * check ip header length and align it.
7516 7516 */
7517 7517 mblk_t *
7518 7518 ip_check_and_align_header(mblk_t *mp, uint_t min_size, ip_recv_attr_t *ira)
7519 7519 {
7520 7520 ill_t *ill = ira->ira_ill;
7521 7521 ssize_t len;
7522 7522
7523 7523 len = MBLKL(mp);
7524 7524
7525 7525 if (!OK_32PTR(mp->b_rptr))
7526 7526 IP_STAT(ill->ill_ipst, ip_notaligned);
7527 7527 else
7528 7528 IP_STAT(ill->ill_ipst, ip_recv_pullup);
7529 7529
7530 7530 /* Guard against bogus device drivers */
7531 7531 if (len < 0) {
7532 7532 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7533 7533 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7534 7534 freemsg(mp);
7535 7535 return (NULL);
7536 7536 }
7537 7537
7538 7538 if (len == 0) {
7539 7539 /* GLD sometimes sends up mblk with b_rptr == b_wptr! */
7540 7540 mblk_t *mp1 = mp->b_cont;
7541 7541
7542 7542 if (!(ira->ira_flags & IRAF_L2SRC_SET))
7543 7543 ip_setl2src(mp, ira, ira->ira_rill);
7544 7544 ASSERT(ira->ira_flags & IRAF_L2SRC_SET);
7545 7545
7546 7546 freeb(mp);
7547 7547 mp = mp1;
7548 7548 if (mp == NULL)
7549 7549 return (NULL);
7550 7550
7551 7551 if (OK_32PTR(mp->b_rptr) && MBLKL(mp) >= min_size)
7552 7552 return (mp);
7553 7553 }
7554 7554 if (ip_pullup(mp, min_size, ira) == NULL) {
7555 7555 if (msgdsize(mp) < min_size) {
7556 7556 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7557 7557 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7558 7558 } else {
7559 7559 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7560 7560 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7561 7561 }
7562 7562 freemsg(mp);
7563 7563 return (NULL);
7564 7564 }
7565 7565 return (mp);
7566 7566 }
7567 7567
7568 7568 /*
7569 7569 * Common code for IPv4 and IPv6 to check and pullup multi-mblks
7570 7570 */
7571 7571 mblk_t *
7572 7572 ip_check_length(mblk_t *mp, uchar_t *rptr, ssize_t len, uint_t pkt_len,
7573 7573 uint_t min_size, ip_recv_attr_t *ira)
7574 7574 {
7575 7575 ill_t *ill = ira->ira_ill;
7576 7576
7577 7577 /*
7578 7578 * Make sure we have data length consistent
7579 7579 * with the IP header.
7580 7580 */
7581 7581 if (mp->b_cont == NULL) {
7582 7582 /* pkt_len is based on ipha_len, not the mblk length */
7583 7583 if (pkt_len < min_size) {
7584 7584 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7585 7585 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7586 7586 freemsg(mp);
7587 7587 return (NULL);
7588 7588 }
7589 7589 if (len < 0) {
7590 7590 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
7591 7591 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
7592 7592 freemsg(mp);
7593 7593 return (NULL);
7594 7594 }
7595 7595 /* Drop any pad */
7596 7596 mp->b_wptr = rptr + pkt_len;
7597 7597 } else if ((len += msgdsize(mp->b_cont)) != 0) {
7598 7598 ASSERT(pkt_len >= min_size);
7599 7599 if (pkt_len < min_size) {
7600 7600 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7601 7601 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7602 7602 freemsg(mp);
7603 7603 return (NULL);
7604 7604 }
7605 7605 if (len < 0) {
7606 7606 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
7607 7607 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
7608 7608 freemsg(mp);
7609 7609 return (NULL);
7610 7610 }
7611 7611 /* Drop any pad */
7612 7612 (void) adjmsg(mp, -len);
7613 7613 /*
7614 7614 * adjmsg may have freed an mblk from the chain, hence
7615 7615 * invalidate any hw checksum here. This will force IP to
7616 7616 * calculate the checksum in sw, but only for this packet.
7617 7617 */
7618 7618 DB_CKSUMFLAGS(mp) = 0;
7619 7619 IP_STAT(ill->ill_ipst, ip_multimblk);
7620 7620 }
7621 7621 return (mp);
7622 7622 }
7623 7623
7624 7624 /*
7625 7625 * Check that the IPv4 opt_len is consistent with the packet and pullup
7626 7626 * the options.
7627 7627 */
7628 7628 mblk_t *
7629 7629 ip_check_optlen(mblk_t *mp, ipha_t *ipha, uint_t opt_len, uint_t pkt_len,
7630 7630 ip_recv_attr_t *ira)
7631 7631 {
7632 7632 ill_t *ill = ira->ira_ill;
7633 7633 ssize_t len;
7634 7634
7635 7635 /* Assume no IPv6 packets arrive over the IPv4 queue */
7636 7636 if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
7637 7637 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7638 7638 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInWrongIPVersion);
7639 7639 ip_drop_input("IPvN packet on IPv4 ill", mp, ill);
7640 7640 freemsg(mp);
7641 7641 return (NULL);
7642 7642 }
7643 7643
7644 7644 if (opt_len > (15 - IP_SIMPLE_HDR_LENGTH_IN_WORDS)) {
7645 7645 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7646 7646 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7647 7647 freemsg(mp);
7648 7648 return (NULL);
7649 7649 }
7650 7650 /*
7651 7651 * Recompute complete header length and make sure we
7652 7652 * have access to all of it.
7653 7653 */
7654 7654 len = ((size_t)opt_len + IP_SIMPLE_HDR_LENGTH_IN_WORDS) << 2;
7655 7655 if (len > (mp->b_wptr - mp->b_rptr)) {
7656 7656 if (len > pkt_len) {
7657 7657 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7658 7658 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7659 7659 freemsg(mp);
7660 7660 return (NULL);
7661 7661 }
7662 7662 if (ip_pullup(mp, len, ira) == NULL) {
7663 7663 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7664 7664 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7665 7665 freemsg(mp);
7666 7666 return (NULL);
7667 7667 }
7668 7668 }
7669 7669 return (mp);
7670 7670 }
7671 7671
7672 7672 /*
7673 7673 * Returns a new ire, or the same ire, or NULL.
7674 7674 * If a different IRE is returned, then it is held; the caller
7675 7675 * needs to release it.
7676 7676 * In no case is there any hold/release on the ire argument.
7677 7677 */
7678 7678 ire_t *
7679 7679 ip_check_multihome(void *addr, ire_t *ire, ill_t *ill)
7680 7680 {
7681 7681 ire_t *new_ire;
7682 7682 ill_t *ire_ill;
7683 7683 uint_t ifindex;
7684 7684 ip_stack_t *ipst = ill->ill_ipst;
7685 7685 boolean_t strict_check = B_FALSE;
7686 7686
7687 7687 /*
7688 7688 * IPMP common case: if IRE and ILL are in the same group, there's no
7689 7689 * issue (e.g. packet received on an underlying interface matched an
7690 7690 * IRE_LOCAL on its associated group interface).
7691 7691 */
7692 7692 ASSERT(ire->ire_ill != NULL);
7693 7693 if (IS_IN_SAME_ILLGRP(ill, ire->ire_ill))
7694 7694 return (ire);
7695 7695
7696 7696 /*
7697 7697 * Do another ire lookup here, using the ingress ill, to see if the
7698 7698 * interface is in a usesrc group.
7699 7699 * As long as the ills belong to the same group, we don't consider
7700 7700 * them to be arriving on the wrong interface. Thus, if the switch
7701 7701 * is doing inbound load spreading, we won't drop packets when the
7702 7702 * ip*_strict_dst_multihoming switch is on.
7703 7703 * We also need to check for IPIF_UNNUMBERED point2point interfaces
7704 7704 * where the local address may not be unique. In this case we were
7705 7705 * at the mercy of the initial ire lookup and the IRE_LOCAL it
7706 7706 * actually returned. The new lookup, which is more specific, should
7707 7707 * only find the IRE_LOCAL associated with the ingress ill if one
7708 7708 * exists.
7709 7709 */
7710 7710 if (ire->ire_ipversion == IPV4_VERSION) {
7711 7711 if (ipst->ips_ip_strict_dst_multihoming)
7712 7712 strict_check = B_TRUE;
7713 7713 new_ire = ire_ftable_lookup_v4(*((ipaddr_t *)addr), 0, 0,
7714 7714 IRE_LOCAL, ill, ALL_ZONES, NULL,
7715 7715 (MATCH_IRE_TYPE|MATCH_IRE_ILL), 0, ipst, NULL);
7716 7716 } else {
7717 7717 ASSERT(!IN6_IS_ADDR_MULTICAST((in6_addr_t *)addr));
7718 7718 if (ipst->ips_ipv6_strict_dst_multihoming)
7719 7719 strict_check = B_TRUE;
7720 7720 new_ire = ire_ftable_lookup_v6((in6_addr_t *)addr, NULL, NULL,
7721 7721 IRE_LOCAL, ill, ALL_ZONES, NULL,
7722 7722 (MATCH_IRE_TYPE|MATCH_IRE_ILL), 0, ipst, NULL);
7723 7723 }
7724 7724 /*
7725 7725 * If the same ire that was returned in ip_input() is found then this
7726 7726 * is an indication that usesrc groups are in use. The packet
7727 7727 * arrived on a different ill in the group than the one associated with
7728 7728 * the destination address. If a different ire was found then the same
7729 7729 * IP address must be hosted on multiple ills. This is possible with
7730 7730 * unnumbered point2point interfaces. We switch to use this new ire in
7731 7731 * order to have accurate interface statistics.
7732 7732 */
7733 7733 if (new_ire != NULL) {
7734 7734 /* Note: held in one case but not the other? Caller handles */
7735 7735 if (new_ire != ire)
7736 7736 return (new_ire);
7737 7737 /* Unchanged */
7738 7738 ire_refrele(new_ire);
7739 7739 return (ire);
7740 7740 }
7741 7741
7742 7742 /*
7743 7743 * Chase pointers once and store locally.
7744 7744 */
7745 7745 ASSERT(ire->ire_ill != NULL);
7746 7746 ire_ill = ire->ire_ill;
7747 7747 ifindex = ill->ill_usesrc_ifindex;
7748 7748
7749 7749 /*
7750 7750 * Check if it's a legal address on the 'usesrc' interface.
7751 7751 * For IPMP data addresses the IRE_LOCAL is the upper, hence we
7752 7752 * can just check phyint_ifindex.
7753 7753 */
7754 7754 if (ifindex != 0 && ifindex == ire_ill->ill_phyint->phyint_ifindex) {
7755 7755 return (ire);
7756 7756 }
7757 7757
7758 7758 /*
7759 7759 * If the ip*_strict_dst_multihoming switch is on then we can
7760 7760 * only accept this packet if the interface is marked as routing.
7761 7761 */
7762 7762 if (!(strict_check))
7763 7763 return (ire);
7764 7764
7765 7765 if ((ill->ill_flags & ire->ire_ill->ill_flags & ILLF_ROUTER) != 0) {
7766 7766 return (ire);
7767 7767 }
7768 7768 return (NULL);
7769 7769 }
7770 7770
7771 7771 /*
7772 7772 * This function is used to construct a mac_header_info_s from a
7773 7773 * DL_UNITDATA_IND message.
7774 7774 * The address fields in the mhi structure points into the message,
7775 7775 * thus the caller can't use those fields after freeing the message.
7776 7776 *
7777 7777 * We determine whether the packet received is a non-unicast packet
7778 7778 * and in doing so, determine whether or not it is broadcast vs multicast.
7779 7779 * For it to be a broadcast packet, we must have the appropriate mblk_t
7780 7780 * hanging off the ill_t. If this is either not present or doesn't match
7781 7781 * the destination mac address in the DL_UNITDATA_IND, the packet is deemed
7782 7782 * to be multicast. Thus NICs that have no broadcast address (or no
7783 7783 * capability for one, such as point to point links) cannot return as
7784 7784 * the packet being broadcast.
7785 7785 */
7786 7786 void
7787 7787 ip_dlur_to_mhi(ill_t *ill, mblk_t *mb, struct mac_header_info_s *mhip)
7788 7788 {
7789 7789 dl_unitdata_ind_t *ind = (dl_unitdata_ind_t *)mb->b_rptr;
7790 7790 mblk_t *bmp;
7791 7791 uint_t extra_offset;
7792 7792
7793 7793 bzero(mhip, sizeof (struct mac_header_info_s));
7794 7794
7795 7795 mhip->mhi_dsttype = MAC_ADDRTYPE_UNICAST;
7796 7796
7797 7797 if (ill->ill_sap_length < 0)
7798 7798 extra_offset = 0;
7799 7799 else
7800 7800 extra_offset = ill->ill_sap_length;
7801 7801
7802 7802 mhip->mhi_daddr = (uchar_t *)ind + ind->dl_dest_addr_offset +
7803 7803 extra_offset;
7804 7804 mhip->mhi_saddr = (uchar_t *)ind + ind->dl_src_addr_offset +
7805 7805 extra_offset;
7806 7806
7807 7807 if (!ind->dl_group_address)
7808 7808 return;
7809 7809
7810 7810 /* Multicast or broadcast */
7811 7811 mhip->mhi_dsttype = MAC_ADDRTYPE_MULTICAST;
7812 7812
7813 7813 if (ind->dl_dest_addr_offset > sizeof (*ind) &&
7814 7814 ind->dl_dest_addr_offset + ind->dl_dest_addr_length < MBLKL(mb) &&
7815 7815 (bmp = ill->ill_bcast_mp) != NULL) {
7816 7816 dl_unitdata_req_t *dlur;
7817 7817 uint8_t *bphys_addr;
7818 7818
7819 7819 dlur = (dl_unitdata_req_t *)bmp->b_rptr;
7820 7820 bphys_addr = (uchar_t *)dlur + dlur->dl_dest_addr_offset +
7821 7821 extra_offset;
7822 7822
7823 7823 if (bcmp(mhip->mhi_daddr, bphys_addr,
7824 7824 ind->dl_dest_addr_length) == 0)
7825 7825 mhip->mhi_dsttype = MAC_ADDRTYPE_BROADCAST;
7826 7826 }
7827 7827 }
7828 7828
7829 7829 /*
7830 7830 * This function is used to construct a mac_header_info_s from a
7831 7831 * M_DATA fastpath message from a DLPI driver.
7832 7832 * The address fields in the mhi structure points into the message,
7833 7833 * thus the caller can't use those fields after freeing the message.
7834 7834 *
7835 7835 * We determine whether the packet received is a non-unicast packet
7836 7836 * and in doing so, determine whether or not it is broadcast vs multicast.
7837 7837 * For it to be a broadcast packet, we must have the appropriate mblk_t
7838 7838 * hanging off the ill_t. If this is either not present or doesn't match
7839 7839 * the destination mac address in the DL_UNITDATA_IND, the packet is deemed
7840 7840 * to be multicast. Thus NICs that have no broadcast address (or no
7841 7841 * capability for one, such as point to point links) cannot return as
7842 7842 * the packet being broadcast.
7843 7843 */
7844 7844 void
7845 7845 ip_mdata_to_mhi(ill_t *ill, mblk_t *mp, struct mac_header_info_s *mhip)
7846 7846 {
7847 7847 mblk_t *bmp;
7848 7848 struct ether_header *pether;
7849 7849
7850 7850 bzero(mhip, sizeof (struct mac_header_info_s));
7851 7851
7852 7852 mhip->mhi_dsttype = MAC_ADDRTYPE_UNICAST;
7853 7853
7854 7854 pether = (struct ether_header *)((char *)mp->b_rptr
7855 7855 - sizeof (struct ether_header));
7856 7856
7857 7857 /*
7858 7858 * Make sure the interface is an ethernet type, since we don't
7859 7859 * know the header format for anything but Ethernet. Also make
7860 7860 * sure we are pointing correctly above db_base.
7861 7861 */
7862 7862 if (ill->ill_type != IFT_ETHER)
7863 7863 return;
7864 7864
7865 7865 retry:
7866 7866 if ((uchar_t *)pether < mp->b_datap->db_base)
7867 7867 return;
7868 7868
7869 7869 /* Is there a VLAN tag? */
7870 7870 if (ill->ill_isv6) {
7871 7871 if (pether->ether_type != htons(ETHERTYPE_IPV6)) {
7872 7872 pether = (struct ether_header *)((char *)pether - 4);
7873 7873 goto retry;
7874 7874 }
7875 7875 } else {
7876 7876 if (pether->ether_type != htons(ETHERTYPE_IP)) {
7877 7877 pether = (struct ether_header *)((char *)pether - 4);
7878 7878 goto retry;
7879 7879 }
7880 7880 }
7881 7881 mhip->mhi_daddr = (uchar_t *)&pether->ether_dhost;
7882 7882 mhip->mhi_saddr = (uchar_t *)&pether->ether_shost;
7883 7883
7884 7884 if (!(mhip->mhi_daddr[0] & 0x01))
7885 7885 return;
7886 7886
7887 7887 /* Multicast or broadcast */
7888 7888 mhip->mhi_dsttype = MAC_ADDRTYPE_MULTICAST;
7889 7889
7890 7890 if ((bmp = ill->ill_bcast_mp) != NULL) {
7891 7891 dl_unitdata_req_t *dlur;
7892 7892 uint8_t *bphys_addr;
7893 7893 uint_t addrlen;
7894 7894
7895 7895 dlur = (dl_unitdata_req_t *)bmp->b_rptr;
7896 7896 addrlen = dlur->dl_dest_addr_length;
7897 7897 if (ill->ill_sap_length < 0) {
7898 7898 bphys_addr = (uchar_t *)dlur +
7899 7899 dlur->dl_dest_addr_offset;
7900 7900 addrlen += ill->ill_sap_length;
7901 7901 } else {
7902 7902 bphys_addr = (uchar_t *)dlur +
7903 7903 dlur->dl_dest_addr_offset +
7904 7904 ill->ill_sap_length;
7905 7905 addrlen -= ill->ill_sap_length;
7906 7906 }
7907 7907 if (bcmp(mhip->mhi_daddr, bphys_addr, addrlen) == 0)
7908 7908 mhip->mhi_dsttype = MAC_ADDRTYPE_BROADCAST;
7909 7909 }
7910 7910 }
7911 7911
7912 7912 /*
7913 7913 * Handle anything but M_DATA messages
7914 7914 * We see the DL_UNITDATA_IND which are part
7915 7915 * of the data path, and also the other messages from the driver.
7916 7916 */
7917 7917 void
7918 7918 ip_rput_notdata(ill_t *ill, mblk_t *mp)
7919 7919 {
7920 7920 mblk_t *first_mp;
7921 7921 struct iocblk *iocp;
7922 7922 struct mac_header_info_s mhi;
7923 7923
7924 7924 switch (DB_TYPE(mp)) {
7925 7925 case M_PROTO:
7926 7926 case M_PCPROTO: {
7927 7927 if (((dl_unitdata_ind_t *)mp->b_rptr)->dl_primitive !=
7928 7928 DL_UNITDATA_IND) {
7929 7929 /* Go handle anything other than data elsewhere. */
7930 7930 ip_rput_dlpi(ill, mp);
7931 7931 return;
7932 7932 }
7933 7933
7934 7934 first_mp = mp;
7935 7935 mp = first_mp->b_cont;
7936 7936 first_mp->b_cont = NULL;
7937 7937
7938 7938 if (mp == NULL) {
7939 7939 freeb(first_mp);
7940 7940 return;
7941 7941 }
7942 7942 ip_dlur_to_mhi(ill, first_mp, &mhi);
7943 7943 if (ill->ill_isv6)
7944 7944 ip_input_v6(ill, NULL, mp, &mhi);
7945 7945 else
7946 7946 ip_input(ill, NULL, mp, &mhi);
7947 7947
7948 7948 /* Ditch the DLPI header. */
7949 7949 freeb(first_mp);
7950 7950 return;
7951 7951 }
7952 7952 case M_IOCACK:
7953 7953 iocp = (struct iocblk *)mp->b_rptr;
7954 7954 switch (iocp->ioc_cmd) {
7955 7955 case DL_IOC_HDR_INFO:
7956 7956 ill_fastpath_ack(ill, mp);
7957 7957 return;
7958 7958 default:
7959 7959 putnext(ill->ill_rq, mp);
7960 7960 return;
7961 7961 }
7962 7962 /* FALLTHROUGH */
7963 7963 case M_ERROR:
7964 7964 case M_HANGUP:
7965 7965 mutex_enter(&ill->ill_lock);
7966 7966 if (ill->ill_state_flags & ILL_CONDEMNED) {
7967 7967 mutex_exit(&ill->ill_lock);
7968 7968 freemsg(mp);
7969 7969 return;
7970 7970 }
7971 7971 ill_refhold_locked(ill);
7972 7972 mutex_exit(&ill->ill_lock);
7973 7973 qwriter_ip(ill, ill->ill_rq, mp, ip_rput_other, CUR_OP,
7974 7974 B_FALSE);
7975 7975 return;
7976 7976 case M_CTL:
7977 7977 putnext(ill->ill_rq, mp);
7978 7978 return;
7979 7979 case M_IOCNAK:
7980 7980 ip1dbg(("got iocnak "));
7981 7981 iocp = (struct iocblk *)mp->b_rptr;
7982 7982 switch (iocp->ioc_cmd) {
7983 7983 case DL_IOC_HDR_INFO:
7984 7984 ip_rput_other(NULL, ill->ill_rq, mp, NULL);
7985 7985 return;
7986 7986 default:
7987 7987 break;
7988 7988 }
7989 7989 /* FALLTHROUGH */
7990 7990 default:
7991 7991 putnext(ill->ill_rq, mp);
7992 7992 return;
7993 7993 }
7994 7994 }
7995 7995
7996 7996 /* Read side put procedure. Packets coming from the wire arrive here. */
7997 7997 int
7998 7998 ip_rput(queue_t *q, mblk_t *mp)
7999 7999 {
8000 8000 ill_t *ill;
8001 8001 union DL_primitives *dl;
8002 8002
8003 8003 ill = (ill_t *)q->q_ptr;
8004 8004
8005 8005 if (ill->ill_state_flags & (ILL_CONDEMNED | ILL_LL_SUBNET_PENDING)) {
8006 8006 /*
8007 8007 * If things are opening or closing, only accept high-priority
8008 8008 * DLPI messages. (On open ill->ill_ipif has not yet been
8009 8009 * created; on close, things hanging off the ill may have been
8010 8010 * freed already.)
8011 8011 */
8012 8012 dl = (union DL_primitives *)mp->b_rptr;
8013 8013 if (DB_TYPE(mp) != M_PCPROTO ||
8014 8014 dl->dl_primitive == DL_UNITDATA_IND) {
8015 8015 inet_freemsg(mp);
8016 8016 return (0);
8017 8017 }
8018 8018 }
8019 8019 if (DB_TYPE(mp) == M_DATA) {
8020 8020 struct mac_header_info_s mhi;
8021 8021
8022 8022 ip_mdata_to_mhi(ill, mp, &mhi);
8023 8023 ip_input(ill, NULL, mp, &mhi);
8024 8024 } else {
8025 8025 ip_rput_notdata(ill, mp);
8026 8026 }
8027 8027 return (0);
8028 8028 }
8029 8029
8030 8030 /*
8031 8031 * Move the information to a copy.
8032 8032 */
8033 8033 mblk_t *
8034 8034 ip_fix_dbref(mblk_t *mp, ip_recv_attr_t *ira)
8035 8035 {
8036 8036 mblk_t *mp1;
8037 8037 ill_t *ill = ira->ira_ill;
8038 8038 ip_stack_t *ipst = ill->ill_ipst;
8039 8039
8040 8040 IP_STAT(ipst, ip_db_ref);
8041 8041
8042 8042 /* Make sure we have ira_l2src before we loose the original mblk */
8043 8043 if (!(ira->ira_flags & IRAF_L2SRC_SET))
8044 8044 ip_setl2src(mp, ira, ira->ira_rill);
8045 8045
8046 8046 mp1 = copymsg(mp);
8047 8047 if (mp1 == NULL) {
8048 8048 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
8049 8049 ip_drop_input("ipIfStatsInDiscards", mp, ill);
8050 8050 freemsg(mp);
8051 8051 return (NULL);
8052 8052 }
8053 8053 /* preserve the hardware checksum flags and data, if present */
8054 8054 if (DB_CKSUMFLAGS(mp) != 0) {
8055 8055 DB_CKSUMFLAGS(mp1) = DB_CKSUMFLAGS(mp);
8056 8056 DB_CKSUMSTART(mp1) = DB_CKSUMSTART(mp);
8057 8057 DB_CKSUMSTUFF(mp1) = DB_CKSUMSTUFF(mp);
8058 8058 DB_CKSUMEND(mp1) = DB_CKSUMEND(mp);
8059 8059 DB_CKSUM16(mp1) = DB_CKSUM16(mp);
8060 8060 }
8061 8061 freemsg(mp);
8062 8062 return (mp1);
8063 8063 }
8064 8064
8065 8065 static void
8066 8066 ip_dlpi_error(ill_t *ill, t_uscalar_t prim, t_uscalar_t dl_err,
8067 8067 t_uscalar_t err)
8068 8068 {
8069 8069 if (dl_err == DL_SYSERR) {
8070 8070 (void) mi_strlog(ill->ill_rq, 1, SL_CONSOLE|SL_ERROR|SL_TRACE,
8071 8071 "%s: %s failed: DL_SYSERR (errno %u)\n",
8072 8072 ill->ill_name, dl_primstr(prim), err);
8073 8073 return;
8074 8074 }
8075 8075
8076 8076 (void) mi_strlog(ill->ill_rq, 1, SL_CONSOLE|SL_ERROR|SL_TRACE,
8077 8077 "%s: %s failed: %s\n", ill->ill_name, dl_primstr(prim),
8078 8078 dl_errstr(dl_err));
8079 8079 }
8080 8080
8081 8081 /*
8082 8082 * ip_rput_dlpi is called by ip_rput to handle all DLPI messages other
8083 8083 * than DL_UNITDATA_IND messages. If we need to process this message
8084 8084 * exclusively, we call qwriter_ip, in which case we also need to call
8085 8085 * ill_refhold before that, since qwriter_ip does an ill_refrele.
8086 8086 */
8087 8087 void
8088 8088 ip_rput_dlpi(ill_t *ill, mblk_t *mp)
8089 8089 {
8090 8090 dl_ok_ack_t *dloa = (dl_ok_ack_t *)mp->b_rptr;
8091 8091 dl_error_ack_t *dlea = (dl_error_ack_t *)dloa;
8092 8092 queue_t *q = ill->ill_rq;
8093 8093 t_uscalar_t prim = dloa->dl_primitive;
8094 8094 t_uscalar_t reqprim = DL_PRIM_INVAL;
8095 8095
8096 8096 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi",
8097 8097 char *, dl_primstr(prim), ill_t *, ill);
8098 8098 ip1dbg(("ip_rput_dlpi"));
8099 8099
8100 8100 /*
8101 8101 * If we received an ACK but didn't send a request for it, then it
8102 8102 * can't be part of any pending operation; discard up-front.
8103 8103 */
8104 8104 switch (prim) {
8105 8105 case DL_ERROR_ACK:
8106 8106 reqprim = dlea->dl_error_primitive;
8107 8107 ip2dbg(("ip_rput_dlpi(%s): DL_ERROR_ACK for %s (0x%x): %s "
8108 8108 "(0x%x), unix %u\n", ill->ill_name, dl_primstr(reqprim),
8109 8109 reqprim, dl_errstr(dlea->dl_errno), dlea->dl_errno,
8110 8110 dlea->dl_unix_errno));
8111 8111 break;
8112 8112 case DL_OK_ACK:
8113 8113 reqprim = dloa->dl_correct_primitive;
8114 8114 break;
8115 8115 case DL_INFO_ACK:
8116 8116 reqprim = DL_INFO_REQ;
8117 8117 break;
8118 8118 case DL_BIND_ACK:
8119 8119 reqprim = DL_BIND_REQ;
8120 8120 break;
8121 8121 case DL_PHYS_ADDR_ACK:
8122 8122 reqprim = DL_PHYS_ADDR_REQ;
8123 8123 break;
8124 8124 case DL_NOTIFY_ACK:
8125 8125 reqprim = DL_NOTIFY_REQ;
8126 8126 break;
8127 8127 case DL_CAPABILITY_ACK:
8128 8128 reqprim = DL_CAPABILITY_REQ;
8129 8129 break;
8130 8130 }
8131 8131
8132 8132 if (prim != DL_NOTIFY_IND) {
8133 8133 if (reqprim == DL_PRIM_INVAL ||
8134 8134 !ill_dlpi_pending(ill, reqprim)) {
8135 8135 /* Not a DLPI message we support or expected */
8136 8136 freemsg(mp);
8137 8137 return;
8138 8138 }
8139 8139 ip1dbg(("ip_rput: received %s for %s\n", dl_primstr(prim),
8140 8140 dl_primstr(reqprim)));
8141 8141 }
8142 8142
8143 8143 switch (reqprim) {
8144 8144 case DL_UNBIND_REQ:
8145 8145 /*
8146 8146 * NOTE: we mark the unbind as complete even if we got a
8147 8147 * DL_ERROR_ACK, since there's not much else we can do.
8148 8148 */
8149 8149 mutex_enter(&ill->ill_lock);
8150 8150 ill->ill_state_flags &= ~ILL_DL_UNBIND_IN_PROGRESS;
8151 8151 cv_signal(&ill->ill_cv);
8152 8152 mutex_exit(&ill->ill_lock);
8153 8153 break;
8154 8154
8155 8155 case DL_ENABMULTI_REQ:
8156 8156 if (prim == DL_OK_ACK) {
8157 8157 if (ill->ill_dlpi_multicast_state == IDS_INPROGRESS)
8158 8158 ill->ill_dlpi_multicast_state = IDS_OK;
8159 8159 }
8160 8160 break;
8161 8161 }
8162 8162
8163 8163 /*
8164 8164 * The message is one we're waiting for (or DL_NOTIFY_IND), but we
8165 8165 * need to become writer to continue to process it. Because an
8166 8166 * exclusive operation doesn't complete until replies to all queued
8167 8167 * DLPI messages have been received, we know we're in the middle of an
8168 8168 * exclusive operation and pass CUR_OP (except for DL_NOTIFY_IND).
8169 8169 *
8170 8170 * As required by qwriter_ip(), we refhold the ill; it will refrele.
8171 8171 * Since this is on the ill stream we unconditionally bump up the
8172 8172 * refcount without doing ILL_CAN_LOOKUP().
8173 8173 */
8174 8174 ill_refhold(ill);
8175 8175 if (prim == DL_NOTIFY_IND)
8176 8176 qwriter_ip(ill, q, mp, ip_rput_dlpi_writer, NEW_OP, B_FALSE);
8177 8177 else
8178 8178 qwriter_ip(ill, q, mp, ip_rput_dlpi_writer, CUR_OP, B_FALSE);
8179 8179 }
8180 8180
8181 8181 /*
8182 8182 * Handling of DLPI messages that require exclusive access to the ipsq.
8183 8183 *
8184 8184 * Need to do ipsq_pending_mp_get on ioctl completion, which could
8185 8185 * happen here. (along with mi_copy_done)
8186 8186 */
8187 8187 /* ARGSUSED */
8188 8188 static void
8189 8189 ip_rput_dlpi_writer(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
8190 8190 {
8191 8191 dl_ok_ack_t *dloa = (dl_ok_ack_t *)mp->b_rptr;
8192 8192 dl_error_ack_t *dlea = (dl_error_ack_t *)dloa;
8193 8193 int err = 0;
8194 8194 ill_t *ill = (ill_t *)q->q_ptr;
8195 8195 ipif_t *ipif = NULL;
8196 8196 mblk_t *mp1 = NULL;
8197 8197 conn_t *connp = NULL;
8198 8198 t_uscalar_t paddrreq;
8199 8199 mblk_t *mp_hw;
8200 8200 boolean_t success;
8201 8201 boolean_t ioctl_aborted = B_FALSE;
8202 8202 boolean_t log = B_TRUE;
8203 8203
8204 8204 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer",
8205 8205 char *, dl_primstr(dloa->dl_primitive), ill_t *, ill);
8206 8206
8207 8207 ip1dbg(("ip_rput_dlpi_writer .."));
8208 8208 ASSERT(ipsq->ipsq_xop == ill->ill_phyint->phyint_ipsq->ipsq_xop);
8209 8209 ASSERT(IAM_WRITER_ILL(ill));
8210 8210
8211 8211 ipif = ipsq->ipsq_xop->ipx_pending_ipif;
8212 8212 /*
8213 8213 * The current ioctl could have been aborted by the user and a new
8214 8214 * ioctl to bring up another ill could have started. We could still
8215 8215 * get a response from the driver later.
8216 8216 */
8217 8217 if (ipif != NULL && ipif->ipif_ill != ill)
8218 8218 ioctl_aborted = B_TRUE;
8219 8219
8220 8220 switch (dloa->dl_primitive) {
8221 8221 case DL_ERROR_ACK:
8222 8222 ip1dbg(("ip_rput_dlpi_writer: got DL_ERROR_ACK for %s\n",
8223 8223 dl_primstr(dlea->dl_error_primitive)));
8224 8224
8225 8225 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer error",
8226 8226 char *, dl_primstr(dlea->dl_error_primitive),
8227 8227 ill_t *, ill);
8228 8228
8229 8229 switch (dlea->dl_error_primitive) {
8230 8230 case DL_DISABMULTI_REQ:
8231 8231 ill_dlpi_done(ill, dlea->dl_error_primitive);
8232 8232 break;
8233 8233 case DL_PROMISCON_REQ:
8234 8234 case DL_PROMISCOFF_REQ:
8235 8235 case DL_UNBIND_REQ:
8236 8236 case DL_ATTACH_REQ:
8237 8237 case DL_INFO_REQ:
8238 8238 ill_dlpi_done(ill, dlea->dl_error_primitive);
8239 8239 break;
8240 8240 case DL_NOTIFY_REQ:
8241 8241 ill_dlpi_done(ill, DL_NOTIFY_REQ);
8242 8242 log = B_FALSE;
8243 8243 break;
8244 8244 case DL_PHYS_ADDR_REQ:
8245 8245 /*
8246 8246 * For IPv6 only, there are two additional
8247 8247 * phys_addr_req's sent to the driver to get the
8248 8248 * IPv6 token and lla. This allows IP to acquire
8249 8249 * the hardware address format for a given interface
8250 8250 * without having built in knowledge of the hardware
8251 8251 * address. ill_phys_addr_pend keeps track of the last
8252 8252 * DL_PAR sent so we know which response we are
8253 8253 * dealing with. ill_dlpi_done will update
8254 8254 * ill_phys_addr_pend when it sends the next req.
8255 8255 * We don't complete the IOCTL until all three DL_PARs
8256 8256 * have been attempted, so set *_len to 0 and break.
8257 8257 */
8258 8258 paddrreq = ill->ill_phys_addr_pend;
8259 8259 ill_dlpi_done(ill, DL_PHYS_ADDR_REQ);
8260 8260 if (paddrreq == DL_IPV6_TOKEN) {
8261 8261 ill->ill_token_length = 0;
8262 8262 log = B_FALSE;
8263 8263 break;
8264 8264 } else if (paddrreq == DL_IPV6_LINK_LAYER_ADDR) {
8265 8265 ill->ill_nd_lla_len = 0;
8266 8266 log = B_FALSE;
8267 8267 break;
8268 8268 }
8269 8269 /*
8270 8270 * Something went wrong with the DL_PHYS_ADDR_REQ.
8271 8271 * We presumably have an IOCTL hanging out waiting
8272 8272 * for completion. Find it and complete the IOCTL
8273 8273 * with the error noted.
8274 8274 * However, ill_dl_phys was called on an ill queue
8275 8275 * (from SIOCSLIFNAME), thus conn_pending_ill is not
8276 8276 * set. But the ioctl is known to be pending on ill_wq.
8277 8277 */
8278 8278 if (!ill->ill_ifname_pending)
8279 8279 break;
8280 8280 ill->ill_ifname_pending = 0;
8281 8281 if (!ioctl_aborted)
8282 8282 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8283 8283 if (mp1 != NULL) {
8284 8284 /*
8285 8285 * This operation (SIOCSLIFNAME) must have
8286 8286 * happened on the ill. Assert there is no conn
8287 8287 */
8288 8288 ASSERT(connp == NULL);
8289 8289 q = ill->ill_wq;
8290 8290 }
8291 8291 break;
8292 8292 case DL_BIND_REQ:
8293 8293 ill_dlpi_done(ill, DL_BIND_REQ);
8294 8294 if (ill->ill_ifname_pending)
8295 8295 break;
8296 8296 mutex_enter(&ill->ill_lock);
8297 8297 ill->ill_state_flags &= ~ILL_DOWN_IN_PROGRESS;
8298 8298 mutex_exit(&ill->ill_lock);
8299 8299 /*
8300 8300 * Something went wrong with the bind. We presumably
8301 8301 * have an IOCTL hanging out waiting for completion.
8302 8302 * Find it, take down the interface that was coming
8303 8303 * up, and complete the IOCTL with the error noted.
8304 8304 */
8305 8305 if (!ioctl_aborted)
8306 8306 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8307 8307 if (mp1 != NULL) {
8308 8308 /*
8309 8309 * This might be a result of a DL_NOTE_REPLUMB
8310 8310 * notification. In that case, connp is NULL.
8311 8311 */
8312 8312 if (connp != NULL)
8313 8313 q = CONNP_TO_WQ(connp);
8314 8314
8315 8315 (void) ipif_down(ipif, NULL, NULL);
8316 8316 /* error is set below the switch */
8317 8317 }
8318 8318 break;
8319 8319 case DL_ENABMULTI_REQ:
8320 8320 ill_dlpi_done(ill, DL_ENABMULTI_REQ);
8321 8321
8322 8322 if (ill->ill_dlpi_multicast_state == IDS_INPROGRESS)
8323 8323 ill->ill_dlpi_multicast_state = IDS_FAILED;
8324 8324 if (ill->ill_dlpi_multicast_state == IDS_FAILED) {
8325 8325
8326 8326 printf("ip: joining multicasts failed (%d)"
8327 8327 " on %s - will use link layer "
8328 8328 "broadcasts for multicast\n",
8329 8329 dlea->dl_errno, ill->ill_name);
8330 8330
8331 8331 /*
8332 8332 * Set up for multi_bcast; We are the
8333 8333 * writer, so ok to access ill->ill_ipif
8334 8334 * without any lock.
8335 8335 */
8336 8336 mutex_enter(&ill->ill_phyint->phyint_lock);
8337 8337 ill->ill_phyint->phyint_flags |=
8338 8338 PHYI_MULTI_BCAST;
8339 8339 mutex_exit(&ill->ill_phyint->phyint_lock);
8340 8340
8341 8341 }
8342 8342 freemsg(mp); /* Don't want to pass this up */
8343 8343 return;
8344 8344 case DL_CAPABILITY_REQ:
8345 8345 ip1dbg(("ip_rput_dlpi_writer: got DL_ERROR_ACK for "
8346 8346 "DL_CAPABILITY REQ\n"));
8347 8347 if (ill->ill_dlpi_capab_state == IDCS_PROBE_SENT)
8348 8348 ill->ill_dlpi_capab_state = IDCS_FAILED;
8349 8349 ill_capability_done(ill);
8350 8350 freemsg(mp);
8351 8351 return;
8352 8352 }
8353 8353 /*
8354 8354 * Note the error for IOCTL completion (mp1 is set when
8355 8355 * ready to complete ioctl). If ill_ifname_pending_err is
8356 8356 * set, an error occured during plumbing (ill_ifname_pending),
8357 8357 * so we want to report that error.
8358 8358 *
8359 8359 * NOTE: there are two addtional DL_PHYS_ADDR_REQ's
8360 8360 * (DL_IPV6_TOKEN and DL_IPV6_LINK_LAYER_ADDR) that are
8361 8361 * expected to get errack'd if the driver doesn't support
8362 8362 * these flags (e.g. ethernet). log will be set to B_FALSE
8363 8363 * if these error conditions are encountered.
8364 8364 */
8365 8365 if (mp1 != NULL) {
8366 8366 if (ill->ill_ifname_pending_err != 0) {
8367 8367 err = ill->ill_ifname_pending_err;
8368 8368 ill->ill_ifname_pending_err = 0;
8369 8369 } else {
8370 8370 err = dlea->dl_unix_errno ?
8371 8371 dlea->dl_unix_errno : ENXIO;
8372 8372 }
8373 8373 /*
8374 8374 * If we're plumbing an interface and an error hasn't already
8375 8375 * been saved, set ill_ifname_pending_err to the error passed
8376 8376 * up. Ignore the error if log is B_FALSE (see comment above).
8377 8377 */
8378 8378 } else if (log && ill->ill_ifname_pending &&
8379 8379 ill->ill_ifname_pending_err == 0) {
8380 8380 ill->ill_ifname_pending_err = dlea->dl_unix_errno ?
8381 8381 dlea->dl_unix_errno : ENXIO;
8382 8382 }
8383 8383
8384 8384 if (log)
8385 8385 ip_dlpi_error(ill, dlea->dl_error_primitive,
8386 8386 dlea->dl_errno, dlea->dl_unix_errno);
8387 8387 break;
8388 8388 case DL_CAPABILITY_ACK:
8389 8389 ill_capability_ack(ill, mp);
8390 8390 /*
8391 8391 * The message has been handed off to ill_capability_ack
8392 8392 * and must not be freed below
8393 8393 */
8394 8394 mp = NULL;
8395 8395 break;
8396 8396
8397 8397 case DL_INFO_ACK:
8398 8398 /* Call a routine to handle this one. */
8399 8399 ill_dlpi_done(ill, DL_INFO_REQ);
8400 8400 ip_ll_subnet_defaults(ill, mp);
8401 8401 ASSERT(!MUTEX_HELD(&ill->ill_phyint->phyint_ipsq->ipsq_lock));
8402 8402 return;
8403 8403 case DL_BIND_ACK:
8404 8404 /*
8405 8405 * We should have an IOCTL waiting on this unless
8406 8406 * sent by ill_dl_phys, in which case just return
8407 8407 */
8408 8408 ill_dlpi_done(ill, DL_BIND_REQ);
8409 8409
8410 8410 if (ill->ill_ifname_pending) {
8411 8411 DTRACE_PROBE2(ip__rput__dlpi__ifname__pending,
8412 8412 ill_t *, ill, mblk_t *, mp);
8413 8413 break;
8414 8414 }
8415 8415 mutex_enter(&ill->ill_lock);
8416 8416 ill->ill_dl_up = 1;
8417 8417 ill->ill_state_flags &= ~ILL_DOWN_IN_PROGRESS;
8418 8418 mutex_exit(&ill->ill_lock);
8419 8419
8420 8420 if (!ioctl_aborted)
8421 8421 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8422 8422 if (mp1 == NULL) {
8423 8423 DTRACE_PROBE1(ip__rput__dlpi__no__mblk, ill_t *, ill);
8424 8424 break;
8425 8425 }
8426 8426 /*
8427 8427 * mp1 was added by ill_dl_up(). if that is a result of
8428 8428 * a DL_NOTE_REPLUMB notification, connp could be NULL.
8429 8429 */
8430 8430 if (connp != NULL)
8431 8431 q = CONNP_TO_WQ(connp);
8432 8432 /*
8433 8433 * We are exclusive. So nothing can change even after
8434 8434 * we get the pending mp.
8435 8435 */
8436 8436 ip1dbg(("ip_rput_dlpi: bind_ack %s\n", ill->ill_name));
8437 8437 DTRACE_PROBE1(ip__rput__dlpi__bind__ack, ill_t *, ill);
8438 8438 ill_nic_event_dispatch(ill, 0, NE_UP, NULL, 0);
8439 8439
8440 8440 /*
8441 8441 * Now bring up the resolver; when that is complete, we'll
8442 8442 * create IREs. Note that we intentionally mirror what
8443 8443 * ipif_up() would have done, because we got here by way of
8444 8444 * ill_dl_up(), which stopped ipif_up()'s processing.
8445 8445 */
8446 8446 if (ill->ill_isv6) {
8447 8447 /*
8448 8448 * v6 interfaces.
8449 8449 * Unlike ARP which has to do another bind
8450 8450 * and attach, once we get here we are
8451 8451 * done with NDP
8452 8452 */
8453 8453 (void) ipif_resolver_up(ipif, Res_act_initial);
8454 8454 if ((err = ipif_ndp_up(ipif, B_TRUE)) == 0)
8455 8455 err = ipif_up_done_v6(ipif);
8456 8456 } else if (ill->ill_net_type == IRE_IF_RESOLVER) {
8457 8457 /*
8458 8458 * ARP and other v4 external resolvers.
8459 8459 * Leave the pending mblk intact so that
8460 8460 * the ioctl completes in ip_rput().
8461 8461 */
8462 8462 if (connp != NULL)
8463 8463 mutex_enter(&connp->conn_lock);
8464 8464 mutex_enter(&ill->ill_lock);
8465 8465 success = ipsq_pending_mp_add(connp, ipif, q, mp1, 0);
8466 8466 mutex_exit(&ill->ill_lock);
8467 8467 if (connp != NULL)
8468 8468 mutex_exit(&connp->conn_lock);
8469 8469 if (success) {
8470 8470 err = ipif_resolver_up(ipif, Res_act_initial);
8471 8471 if (err == EINPROGRESS) {
8472 8472 freemsg(mp);
8473 8473 return;
8474 8474 }
8475 8475 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8476 8476 } else {
8477 8477 /* The conn has started closing */
8478 8478 err = EINTR;
8479 8479 }
8480 8480 } else {
8481 8481 /*
8482 8482 * This one is complete. Reply to pending ioctl.
8483 8483 */
8484 8484 (void) ipif_resolver_up(ipif, Res_act_initial);
8485 8485 err = ipif_up_done(ipif);
8486 8486 }
8487 8487
8488 8488 if ((err == 0) && (ill->ill_up_ipifs)) {
8489 8489 err = ill_up_ipifs(ill, q, mp1);
8490 8490 if (err == EINPROGRESS) {
8491 8491 freemsg(mp);
8492 8492 return;
8493 8493 }
8494 8494 }
8495 8495
8496 8496 /*
8497 8497 * If we have a moved ipif to bring up, and everything has
8498 8498 * succeeded to this point, bring it up on the IPMP ill.
8499 8499 * Otherwise, leave it down -- the admin can try to bring it
8500 8500 * up by hand if need be.
8501 8501 */
8502 8502 if (ill->ill_move_ipif != NULL) {
8503 8503 if (err != 0) {
8504 8504 ill->ill_move_ipif = NULL;
8505 8505 } else {
8506 8506 ipif = ill->ill_move_ipif;
8507 8507 ill->ill_move_ipif = NULL;
8508 8508 err = ipif_up(ipif, q, mp1);
8509 8509 if (err == EINPROGRESS) {
8510 8510 freemsg(mp);
8511 8511 return;
8512 8512 }
8513 8513 }
8514 8514 }
8515 8515 break;
8516 8516
8517 8517 case DL_NOTIFY_IND: {
8518 8518 dl_notify_ind_t *notify = (dl_notify_ind_t *)mp->b_rptr;
8519 8519 uint_t orig_mtu, orig_mc_mtu;
8520 8520
8521 8521 switch (notify->dl_notification) {
8522 8522 case DL_NOTE_PHYS_ADDR:
8523 8523 err = ill_set_phys_addr(ill, mp);
8524 8524 break;
8525 8525
8526 8526 case DL_NOTE_REPLUMB:
8527 8527 /*
8528 8528 * Directly return after calling ill_replumb().
8529 8529 * Note that we should not free mp as it is reused
8530 8530 * in the ill_replumb() function.
8531 8531 */
8532 8532 err = ill_replumb(ill, mp);
8533 8533 return;
8534 8534
8535 8535 case DL_NOTE_FASTPATH_FLUSH:
8536 8536 nce_flush(ill, B_FALSE);
8537 8537 break;
8538 8538
8539 8539 case DL_NOTE_SDU_SIZE:
8540 8540 case DL_NOTE_SDU_SIZE2:
8541 8541 /*
8542 8542 * The dce and fragmentation code can cope with
8543 8543 * this changing while packets are being sent.
8544 8544 * When packets are sent ip_output will discover
8545 8545 * a change.
8546 8546 *
8547 8547 * Change the MTU size of the interface.
8548 8548 */
8549 8549 mutex_enter(&ill->ill_lock);
8550 8550 orig_mtu = ill->ill_mtu;
8551 8551 orig_mc_mtu = ill->ill_mc_mtu;
8552 8552 switch (notify->dl_notification) {
8553 8553 case DL_NOTE_SDU_SIZE:
8554 8554 ill->ill_current_frag =
8555 8555 (uint_t)notify->dl_data;
8556 8556 ill->ill_mc_mtu = (uint_t)notify->dl_data;
8557 8557 break;
8558 8558 case DL_NOTE_SDU_SIZE2:
8559 8559 ill->ill_current_frag =
8560 8560 (uint_t)notify->dl_data1;
8561 8561 ill->ill_mc_mtu = (uint_t)notify->dl_data2;
8562 8562 break;
8563 8563 }
8564 8564 if (ill->ill_current_frag > ill->ill_max_frag)
8565 8565 ill->ill_max_frag = ill->ill_current_frag;
8566 8566
8567 8567 if (!(ill->ill_flags & ILLF_FIXEDMTU)) {
8568 8568 ill->ill_mtu = ill->ill_current_frag;
8569 8569
8570 8570 /*
8571 8571 * If ill_user_mtu was set (via
8572 8572 * SIOCSLIFLNKINFO), clamp ill_mtu at it.
8573 8573 */
8574 8574 if (ill->ill_user_mtu != 0 &&
8575 8575 ill->ill_user_mtu < ill->ill_mtu)
8576 8576 ill->ill_mtu = ill->ill_user_mtu;
8577 8577
8578 8578 if (ill->ill_user_mtu != 0 &&
8579 8579 ill->ill_user_mtu < ill->ill_mc_mtu)
8580 8580 ill->ill_mc_mtu = ill->ill_user_mtu;
8581 8581
8582 8582 if (ill->ill_isv6) {
8583 8583 if (ill->ill_mtu < IPV6_MIN_MTU)
8584 8584 ill->ill_mtu = IPV6_MIN_MTU;
8585 8585 if (ill->ill_mc_mtu < IPV6_MIN_MTU)
8586 8586 ill->ill_mc_mtu = IPV6_MIN_MTU;
8587 8587 } else {
8588 8588 if (ill->ill_mtu < IP_MIN_MTU)
8589 8589 ill->ill_mtu = IP_MIN_MTU;
8590 8590 if (ill->ill_mc_mtu < IP_MIN_MTU)
8591 8591 ill->ill_mc_mtu = IP_MIN_MTU;
8592 8592 }
8593 8593 } else if (ill->ill_mc_mtu > ill->ill_mtu) {
8594 8594 ill->ill_mc_mtu = ill->ill_mtu;
8595 8595 }
8596 8596
8597 8597 mutex_exit(&ill->ill_lock);
8598 8598 /*
8599 8599 * Make sure all dce_generation checks find out
8600 8600 * that ill_mtu/ill_mc_mtu has changed.
8601 8601 */
8602 8602 if (orig_mtu != ill->ill_mtu ||
8603 8603 orig_mc_mtu != ill->ill_mc_mtu) {
8604 8604 dce_increment_all_generations(ill->ill_isv6,
8605 8605 ill->ill_ipst);
8606 8606 }
8607 8607
8608 8608 /*
8609 8609 * Refresh IPMP meta-interface MTU if necessary.
8610 8610 */
8611 8611 if (IS_UNDER_IPMP(ill))
8612 8612 ipmp_illgrp_refresh_mtu(ill->ill_grp);
8613 8613 break;
8614 8614
8615 8615 case DL_NOTE_LINK_UP:
8616 8616 case DL_NOTE_LINK_DOWN: {
8617 8617 /*
8618 8618 * We are writer. ill / phyint / ipsq assocs stable.
8619 8619 * The RUNNING flag reflects the state of the link.
8620 8620 */
8621 8621 phyint_t *phyint = ill->ill_phyint;
8622 8622 uint64_t new_phyint_flags;
8623 8623 boolean_t changed = B_FALSE;
8624 8624 boolean_t went_up;
8625 8625
8626 8626 went_up = notify->dl_notification == DL_NOTE_LINK_UP;
8627 8627 mutex_enter(&phyint->phyint_lock);
8628 8628
8629 8629 new_phyint_flags = went_up ?
8630 8630 phyint->phyint_flags | PHYI_RUNNING :
8631 8631 phyint->phyint_flags & ~PHYI_RUNNING;
8632 8632
8633 8633 if (IS_IPMP(ill)) {
8634 8634 new_phyint_flags = went_up ?
8635 8635 new_phyint_flags & ~PHYI_FAILED :
8636 8636 new_phyint_flags | PHYI_FAILED;
8637 8637 }
8638 8638
8639 8639 if (new_phyint_flags != phyint->phyint_flags) {
8640 8640 phyint->phyint_flags = new_phyint_flags;
8641 8641 changed = B_TRUE;
8642 8642 }
8643 8643 mutex_exit(&phyint->phyint_lock);
8644 8644 /*
8645 8645 * ill_restart_dad handles the DAD restart and routing
8646 8646 * socket notification logic.
8647 8647 */
8648 8648 if (changed) {
8649 8649 ill_restart_dad(phyint->phyint_illv4, went_up);
8650 8650 ill_restart_dad(phyint->phyint_illv6, went_up);
8651 8651 }
8652 8652 break;
8653 8653 }
8654 8654 case DL_NOTE_PROMISC_ON_PHYS: {
8655 8655 phyint_t *phyint = ill->ill_phyint;
8656 8656
8657 8657 mutex_enter(&phyint->phyint_lock);
8658 8658 phyint->phyint_flags |= PHYI_PROMISC;
8659 8659 mutex_exit(&phyint->phyint_lock);
8660 8660 break;
8661 8661 }
8662 8662 case DL_NOTE_PROMISC_OFF_PHYS: {
8663 8663 phyint_t *phyint = ill->ill_phyint;
8664 8664
8665 8665 mutex_enter(&phyint->phyint_lock);
8666 8666 phyint->phyint_flags &= ~PHYI_PROMISC;
8667 8667 mutex_exit(&phyint->phyint_lock);
8668 8668 break;
8669 8669 }
8670 8670 case DL_NOTE_CAPAB_RENEG:
8671 8671 /*
8672 8672 * Something changed on the driver side.
8673 8673 * It wants us to renegotiate the capabilities
8674 8674 * on this ill. One possible cause is the aggregation
8675 8675 * interface under us where a port got added or
8676 8676 * went away.
8677 8677 *
8678 8678 * If the capability negotiation is already done
8679 8679 * or is in progress, reset the capabilities and
8680 8680 * mark the ill's ill_capab_reneg to be B_TRUE,
8681 8681 * so that when the ack comes back, we can start
8682 8682 * the renegotiation process.
8683 8683 *
8684 8684 * Note that if ill_capab_reneg is already B_TRUE
8685 8685 * (ill_dlpi_capab_state is IDS_UNKNOWN in this case),
8686 8686 * the capability resetting request has been sent
8687 8687 * and the renegotiation has not been started yet;
8688 8688 * nothing needs to be done in this case.
8689 8689 */
8690 8690 ipsq_current_start(ipsq, ill->ill_ipif, 0);
8691 8691 ill_capability_reset(ill, B_TRUE);
8692 8692 ipsq_current_finish(ipsq);
8693 8693 break;
8694 8694
8695 8695 case DL_NOTE_ALLOWED_IPS:
8696 8696 ill_set_allowed_ips(ill, mp);
8697 8697 break;
8698 8698 default:
8699 8699 ip0dbg(("ip_rput_dlpi_writer: unknown notification "
8700 8700 "type 0x%x for DL_NOTIFY_IND\n",
8701 8701 notify->dl_notification));
8702 8702 break;
8703 8703 }
8704 8704
8705 8705 /*
8706 8706 * As this is an asynchronous operation, we
8707 8707 * should not call ill_dlpi_done
8708 8708 */
8709 8709 break;
8710 8710 }
8711 8711 case DL_NOTIFY_ACK: {
8712 8712 dl_notify_ack_t *noteack = (dl_notify_ack_t *)mp->b_rptr;
8713 8713
8714 8714 if (noteack->dl_notifications & DL_NOTE_LINK_UP)
8715 8715 ill->ill_note_link = 1;
8716 8716 ill_dlpi_done(ill, DL_NOTIFY_REQ);
8717 8717 break;
8718 8718 }
8719 8719 case DL_PHYS_ADDR_ACK: {
8720 8720 /*
8721 8721 * As part of plumbing the interface via SIOCSLIFNAME,
8722 8722 * ill_dl_phys() will queue a series of DL_PHYS_ADDR_REQs,
8723 8723 * whose answers we receive here. As each answer is received,
8724 8724 * we call ill_dlpi_done() to dispatch the next request as
8725 8725 * we're processing the current one. Once all answers have
8726 8726 * been received, we use ipsq_pending_mp_get() to dequeue the
8727 8727 * outstanding IOCTL and reply to it. (Because ill_dl_phys()
8728 8728 * is invoked from an ill queue, conn_oper_pending_ill is not
8729 8729 * available, but we know the ioctl is pending on ill_wq.)
8730 8730 */
8731 8731 uint_t paddrlen, paddroff;
8732 8732 uint8_t *addr;
8733 8733
8734 8734 paddrreq = ill->ill_phys_addr_pend;
8735 8735 paddrlen = ((dl_phys_addr_ack_t *)mp->b_rptr)->dl_addr_length;
8736 8736 paddroff = ((dl_phys_addr_ack_t *)mp->b_rptr)->dl_addr_offset;
8737 8737 addr = mp->b_rptr + paddroff;
8738 8738
8739 8739 ill_dlpi_done(ill, DL_PHYS_ADDR_REQ);
8740 8740 if (paddrreq == DL_IPV6_TOKEN) {
8741 8741 /*
8742 8742 * bcopy to low-order bits of ill_token
8743 8743 *
8744 8744 * XXX Temporary hack - currently, all known tokens
8745 8745 * are 64 bits, so I'll cheat for the moment.
8746 8746 */
8747 8747 bcopy(addr, &ill->ill_token.s6_addr32[2], paddrlen);
8748 8748 ill->ill_token_length = paddrlen;
8749 8749 break;
8750 8750 } else if (paddrreq == DL_IPV6_LINK_LAYER_ADDR) {
8751 8751 ASSERT(ill->ill_nd_lla_mp == NULL);
8752 8752 ill_set_ndmp(ill, mp, paddroff, paddrlen);
8753 8753 mp = NULL;
8754 8754 break;
8755 8755 } else if (paddrreq == DL_CURR_DEST_ADDR) {
8756 8756 ASSERT(ill->ill_dest_addr_mp == NULL);
8757 8757 ill->ill_dest_addr_mp = mp;
8758 8758 ill->ill_dest_addr = addr;
8759 8759 mp = NULL;
8760 8760 if (ill->ill_isv6) {
8761 8761 ill_setdesttoken(ill);
8762 8762 ipif_setdestlinklocal(ill->ill_ipif);
8763 8763 }
8764 8764 break;
8765 8765 }
8766 8766
8767 8767 ASSERT(paddrreq == DL_CURR_PHYS_ADDR);
8768 8768 ASSERT(ill->ill_phys_addr_mp == NULL);
8769 8769 if (!ill->ill_ifname_pending)
8770 8770 break;
8771 8771 ill->ill_ifname_pending = 0;
8772 8772 if (!ioctl_aborted)
8773 8773 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8774 8774 if (mp1 != NULL) {
8775 8775 ASSERT(connp == NULL);
8776 8776 q = ill->ill_wq;
8777 8777 }
8778 8778 /*
8779 8779 * If any error acks received during the plumbing sequence,
8780 8780 * ill_ifname_pending_err will be set. Break out and send up
8781 8781 * the error to the pending ioctl.
8782 8782 */
8783 8783 if (ill->ill_ifname_pending_err != 0) {
8784 8784 err = ill->ill_ifname_pending_err;
8785 8785 ill->ill_ifname_pending_err = 0;
8786 8786 break;
8787 8787 }
8788 8788
8789 8789 ill->ill_phys_addr_mp = mp;
8790 8790 ill->ill_phys_addr = (paddrlen == 0 ? NULL : addr);
8791 8791 mp = NULL;
8792 8792
8793 8793 /*
8794 8794 * If paddrlen or ill_phys_addr_length is zero, the DLPI
8795 8795 * provider doesn't support physical addresses. We check both
8796 8796 * paddrlen and ill_phys_addr_length because sppp (PPP) does
8797 8797 * not have physical addresses, but historically adversises a
8798 8798 * physical address length of 0 in its DL_INFO_ACK, but 6 in
8799 8799 * its DL_PHYS_ADDR_ACK.
8800 8800 */
8801 8801 if (paddrlen == 0 || ill->ill_phys_addr_length == 0) {
8802 8802 ill->ill_phys_addr = NULL;
8803 8803 } else if (paddrlen != ill->ill_phys_addr_length) {
8804 8804 ip0dbg(("DL_PHYS_ADDR_ACK: got addrlen %d, expected %d",
8805 8805 paddrlen, ill->ill_phys_addr_length));
8806 8806 err = EINVAL;
8807 8807 break;
8808 8808 }
8809 8809
8810 8810 if (ill->ill_nd_lla_mp == NULL) {
8811 8811 if ((mp_hw = copyb(ill->ill_phys_addr_mp)) == NULL) {
8812 8812 err = ENOMEM;
8813 8813 break;
8814 8814 }
8815 8815 ill_set_ndmp(ill, mp_hw, paddroff, paddrlen);
8816 8816 }
8817 8817
8818 8818 if (ill->ill_isv6) {
8819 8819 ill_setdefaulttoken(ill);
8820 8820 ipif_setlinklocal(ill->ill_ipif);
8821 8821 }
8822 8822 break;
8823 8823 }
8824 8824 case DL_OK_ACK:
8825 8825 ip2dbg(("DL_OK_ACK %s (0x%x)\n",
8826 8826 dl_primstr((int)dloa->dl_correct_primitive),
8827 8827 dloa->dl_correct_primitive));
8828 8828 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer ok",
8829 8829 char *, dl_primstr(dloa->dl_correct_primitive),
8830 8830 ill_t *, ill);
8831 8831
8832 8832 switch (dloa->dl_correct_primitive) {
8833 8833 case DL_ENABMULTI_REQ:
8834 8834 case DL_DISABMULTI_REQ:
8835 8835 ill_dlpi_done(ill, dloa->dl_correct_primitive);
8836 8836 break;
8837 8837 case DL_PROMISCON_REQ:
8838 8838 case DL_PROMISCOFF_REQ:
8839 8839 case DL_UNBIND_REQ:
8840 8840 case DL_ATTACH_REQ:
8841 8841 ill_dlpi_done(ill, dloa->dl_correct_primitive);
8842 8842 break;
8843 8843 }
8844 8844 break;
8845 8845 default:
8846 8846 break;
8847 8847 }
8848 8848
8849 8849 freemsg(mp);
8850 8850 if (mp1 == NULL)
8851 8851 return;
8852 8852
8853 8853 /*
8854 8854 * The operation must complete without EINPROGRESS since
8855 8855 * ipsq_pending_mp_get() has removed the mblk (mp1). Otherwise,
8856 8856 * the operation will be stuck forever inside the IPSQ.
8857 8857 */
8858 8858 ASSERT(err != EINPROGRESS);
8859 8859
8860 8860 DTRACE_PROBE4(ipif__ioctl, char *, "ip_rput_dlpi_writer finish",
8861 8861 int, ipsq->ipsq_xop->ipx_current_ioctl, ill_t *, ill,
8862 8862 ipif_t *, NULL);
8863 8863
8864 8864 switch (ipsq->ipsq_xop->ipx_current_ioctl) {
8865 8865 case 0:
8866 8866 ipsq_current_finish(ipsq);
8867 8867 break;
8868 8868
8869 8869 case SIOCSLIFNAME:
8870 8870 case IF_UNITSEL: {
8871 8871 ill_t *ill_other = ILL_OTHER(ill);
8872 8872
8873 8873 /*
8874 8874 * If SIOCSLIFNAME or IF_UNITSEL is about to succeed, and the
8875 8875 * ill has a peer which is in an IPMP group, then place ill
8876 8876 * into the same group. One catch: although ifconfig plumbs
8877 8877 * the appropriate IPMP meta-interface prior to plumbing this
8878 8878 * ill, it is possible for multiple ifconfig applications to
8879 8879 * race (or for another application to adjust plumbing), in
8880 8880 * which case the IPMP meta-interface we need will be missing.
8881 8881 * If so, kick the phyint out of the group.
8882 8882 */
8883 8883 if (err == 0 && ill_other != NULL && IS_UNDER_IPMP(ill_other)) {
8884 8884 ipmp_grp_t *grp = ill->ill_phyint->phyint_grp;
8885 8885 ipmp_illgrp_t *illg;
8886 8886
8887 8887 illg = ill->ill_isv6 ? grp->gr_v6 : grp->gr_v4;
8888 8888 if (illg == NULL)
8889 8889 ipmp_phyint_leave_grp(ill->ill_phyint);
8890 8890 else
8891 8891 ipmp_ill_join_illgrp(ill, illg);
8892 8892 }
8893 8893
8894 8894 if (ipsq->ipsq_xop->ipx_current_ioctl == IF_UNITSEL)
8895 8895 ip_ioctl_finish(q, mp1, err, NO_COPYOUT, ipsq);
8896 8896 else
8897 8897 ip_ioctl_finish(q, mp1, err, COPYOUT, ipsq);
8898 8898 break;
8899 8899 }
8900 8900 case SIOCLIFADDIF:
8901 8901 ip_ioctl_finish(q, mp1, err, COPYOUT, ipsq);
8902 8902 break;
8903 8903
8904 8904 default:
8905 8905 ip_ioctl_finish(q, mp1, err, NO_COPYOUT, ipsq);
8906 8906 break;
8907 8907 }
8908 8908 }
8909 8909
8910 8910 /*
8911 8911 * ip_rput_other is called by ip_rput to handle messages modifying the global
8912 8912 * state in IP. If 'ipsq' is non-NULL, caller is writer on it.
8913 8913 */
8914 8914 /* ARGSUSED */
8915 8915 void
8916 8916 ip_rput_other(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
8917 8917 {
8918 8918 ill_t *ill = q->q_ptr;
8919 8919 struct iocblk *iocp;
8920 8920
8921 8921 ip1dbg(("ip_rput_other "));
8922 8922 if (ipsq != NULL) {
8923 8923 ASSERT(IAM_WRITER_IPSQ(ipsq));
8924 8924 ASSERT(ipsq->ipsq_xop ==
8925 8925 ill->ill_phyint->phyint_ipsq->ipsq_xop);
8926 8926 }
8927 8927
8928 8928 switch (mp->b_datap->db_type) {
8929 8929 case M_ERROR:
8930 8930 case M_HANGUP:
8931 8931 /*
8932 8932 * The device has a problem. We force the ILL down. It can
8933 8933 * be brought up again manually using SIOCSIFFLAGS (via
8934 8934 * ifconfig or equivalent).
8935 8935 */
8936 8936 ASSERT(ipsq != NULL);
8937 8937 if (mp->b_rptr < mp->b_wptr)
8938 8938 ill->ill_error = (int)(*mp->b_rptr & 0xFF);
8939 8939 if (ill->ill_error == 0)
8940 8940 ill->ill_error = ENXIO;
8941 8941 if (!ill_down_start(q, mp))
8942 8942 return;
8943 8943 ipif_all_down_tail(ipsq, q, mp, NULL);
8944 8944 break;
8945 8945 case M_IOCNAK: {
8946 8946 iocp = (struct iocblk *)mp->b_rptr;
8947 8947
8948 8948 ASSERT(iocp->ioc_cmd == DL_IOC_HDR_INFO);
8949 8949 /*
8950 8950 * If this was the first attempt, turn off the fastpath
8951 8951 * probing.
8952 8952 */
8953 8953 mutex_enter(&ill->ill_lock);
8954 8954 if (ill->ill_dlpi_fastpath_state == IDS_INPROGRESS) {
8955 8955 ill->ill_dlpi_fastpath_state = IDS_FAILED;
8956 8956 mutex_exit(&ill->ill_lock);
8957 8957 /*
8958 8958 * don't flush the nce_t entries: we use them
8959 8959 * as an index to the ncec itself.
8960 8960 */
8961 8961 ip1dbg(("ip_rput: DLPI fastpath off on interface %s\n",
8962 8962 ill->ill_name));
8963 8963 } else {
8964 8964 mutex_exit(&ill->ill_lock);
8965 8965 }
8966 8966 freemsg(mp);
8967 8967 break;
8968 8968 }
8969 8969 default:
8970 8970 ASSERT(0);
8971 8971 break;
8972 8972 }
8973 8973 }
8974 8974
8975 8975 /*
8976 8976 * Update any source route, record route or timestamp options
8977 8977 * When it fails it has consumed the message and BUMPed the MIB.
8978 8978 */
8979 8979 boolean_t
8980 8980 ip_forward_options(mblk_t *mp, ipha_t *ipha, ill_t *dst_ill,
8981 8981 ip_recv_attr_t *ira)
8982 8982 {
8983 8983 ipoptp_t opts;
8984 8984 uchar_t *opt;
8985 8985 uint8_t optval;
8986 8986 uint8_t optlen;
8987 8987 ipaddr_t dst;
8988 8988 ipaddr_t ifaddr;
8989 8989 uint32_t ts;
8990 8990 timestruc_t now;
8991 8991 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
8992 8992
8993 8993 ip2dbg(("ip_forward_options\n"));
8994 8994 dst = ipha->ipha_dst;
8995 8995 for (optval = ipoptp_first(&opts, ipha);
8996 8996 optval != IPOPT_EOL;
8997 8997 optval = ipoptp_next(&opts)) {
8998 8998 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
8999 8999 opt = opts.ipoptp_cur;
9000 9000 optlen = opts.ipoptp_len;
9001 9001 ip2dbg(("ip_forward_options: opt %d, len %d\n",
9002 9002 optval, opts.ipoptp_len));
9003 9003 switch (optval) {
9004 9004 uint32_t off;
9005 9005 case IPOPT_SSRR:
9006 9006 case IPOPT_LSRR:
9007 9007 /* Check if adminstratively disabled */
9008 9008 if (!ipst->ips_ip_forward_src_routed) {
9009 9009 BUMP_MIB(dst_ill->ill_ip_mib,
9010 9010 ipIfStatsForwProhibits);
9011 9011 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED",
9012 9012 mp, dst_ill);
9013 9013 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED,
9014 9014 ira);
9015 9015 return (B_FALSE);
9016 9016 }
9017 9017 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9018 9018 /*
9019 9019 * Must be partial since ip_input_options
9020 9020 * checked for strict.
9021 9021 */
9022 9022 break;
9023 9023 }
9024 9024 off = opt[IPOPT_OFFSET];
9025 9025 off--;
9026 9026 redo_srr:
9027 9027 if (optlen < IP_ADDR_LEN ||
9028 9028 off > optlen - IP_ADDR_LEN) {
9029 9029 /* End of source route */
9030 9030 ip1dbg((
9031 9031 "ip_forward_options: end of SR\n"));
9032 9032 break;
9033 9033 }
9034 9034 /* Pick a reasonable address on the outbound if */
9035 9035 ASSERT(dst_ill != NULL);
9036 9036 if (ip_select_source_v4(dst_ill, INADDR_ANY, dst,
9037 9037 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9038 9038 NULL) != 0) {
9039 9039 /* No source! Shouldn't happen */
9040 9040 ifaddr = INADDR_ANY;
9041 9041 }
9042 9042 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9043 9043 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9044 9044 ip1dbg(("ip_forward_options: next hop 0x%x\n",
9045 9045 ntohl(dst)));
9046 9046
9047 9047 /*
9048 9048 * Check if our address is present more than
9049 9049 * once as consecutive hops in source route.
9050 9050 */
9051 9051 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
9052 9052 off += IP_ADDR_LEN;
9053 9053 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9054 9054 goto redo_srr;
9055 9055 }
9056 9056 ipha->ipha_dst = dst;
9057 9057 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9058 9058 break;
9059 9059 case IPOPT_RR:
9060 9060 off = opt[IPOPT_OFFSET];
9061 9061 off--;
9062 9062 if (optlen < IP_ADDR_LEN ||
9063 9063 off > optlen - IP_ADDR_LEN) {
9064 9064 /* No more room - ignore */
9065 9065 ip1dbg((
9066 9066 "ip_forward_options: end of RR\n"));
9067 9067 break;
9068 9068 }
9069 9069 /* Pick a reasonable address on the outbound if */
9070 9070 ASSERT(dst_ill != NULL);
9071 9071 if (ip_select_source_v4(dst_ill, INADDR_ANY, dst,
9072 9072 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9073 9073 NULL) != 0) {
9074 9074 /* No source! Shouldn't happen */
9075 9075 ifaddr = INADDR_ANY;
9076 9076 }
9077 9077 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9078 9078 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9079 9079 break;
9080 9080 case IPOPT_TS:
9081 9081 /* Insert timestamp if there is room */
9082 9082 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9083 9083 case IPOPT_TS_TSONLY:
9084 9084 off = IPOPT_TS_TIMELEN;
9085 9085 break;
9086 9086 case IPOPT_TS_PRESPEC:
9087 9087 case IPOPT_TS_PRESPEC_RFC791:
9088 9088 /* Verify that the address matched */
9089 9089 off = opt[IPOPT_OFFSET] - 1;
9090 9090 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9091 9091 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9092 9092 /* Not for us */
9093 9093 break;
9094 9094 }
9095 9095 /* FALLTHROUGH */
9096 9096 case IPOPT_TS_TSANDADDR:
9097 9097 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9098 9098 break;
9099 9099 default:
9100 9100 /*
9101 9101 * ip_*put_options should have already
9102 9102 * dropped this packet.
9103 9103 */
9104 9104 cmn_err(CE_PANIC, "ip_forward_options: "
9105 9105 "unknown IT - bug in ip_input_options?\n");
9106 9106 return (B_TRUE); /* Keep "lint" happy */
9107 9107 }
9108 9108 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
9109 9109 /* Increase overflow counter */
9110 9110 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
9111 9111 opt[IPOPT_POS_OV_FLG] =
9112 9112 (uint8_t)((opt[IPOPT_POS_OV_FLG] & 0x0F) |
9113 9113 (off << 4));
9114 9114 break;
9115 9115 }
9116 9116 off = opt[IPOPT_OFFSET] - 1;
9117 9117 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9118 9118 case IPOPT_TS_PRESPEC:
9119 9119 case IPOPT_TS_PRESPEC_RFC791:
9120 9120 case IPOPT_TS_TSANDADDR:
9121 9121 /* Pick a reasonable addr on the outbound if */
9122 9122 ASSERT(dst_ill != NULL);
9123 9123 if (ip_select_source_v4(dst_ill, INADDR_ANY,
9124 9124 dst, INADDR_ANY, ALL_ZONES, ipst, &ifaddr,
9125 9125 NULL, NULL) != 0) {
9126 9126 /* No source! Shouldn't happen */
9127 9127 ifaddr = INADDR_ANY;
9128 9128 }
9129 9129 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9130 9130 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9131 9131 /* FALLTHROUGH */
9132 9132 case IPOPT_TS_TSONLY:
9133 9133 off = opt[IPOPT_OFFSET] - 1;
9134 9134 /* Compute # of milliseconds since midnight */
9135 9135 gethrestime(&now);
9136 9136 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
9137 9137 NSEC2MSEC(now.tv_nsec);
9138 9138 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
9139 9139 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
9140 9140 break;
9141 9141 }
9142 9142 break;
9143 9143 }
9144 9144 }
9145 9145 return (B_TRUE);
9146 9146 }
9147 9147
9148 9148 /*
9149 9149 * Call ill_frag_timeout to do garbage collection. ill_frag_timeout
9150 9150 * returns 'true' if there are still fragments left on the queue, in
9151 9151 * which case we restart the timer.
9152 9152 */
9153 9153 void
9154 9154 ill_frag_timer(void *arg)
9155 9155 {
9156 9156 ill_t *ill = (ill_t *)arg;
9157 9157 boolean_t frag_pending;
9158 9158 ip_stack_t *ipst = ill->ill_ipst;
9159 9159 time_t timeout;
9160 9160
9161 9161 mutex_enter(&ill->ill_lock);
9162 9162 ASSERT(!ill->ill_fragtimer_executing);
9163 9163 if (ill->ill_state_flags & ILL_CONDEMNED) {
9164 9164 ill->ill_frag_timer_id = 0;
9165 9165 mutex_exit(&ill->ill_lock);
9166 9166 return;
9167 9167 }
9168 9168 ill->ill_fragtimer_executing = 1;
9169 9169 mutex_exit(&ill->ill_lock);
9170 9170
9171 9171 timeout = (ill->ill_isv6 ? ipst->ips_ipv6_reassembly_timeout :
9172 9172 ipst->ips_ip_reassembly_timeout);
9173 9173
9174 9174 frag_pending = ill_frag_timeout(ill, timeout);
9175 9175
9176 9176 /*
9177 9177 * Restart the timer, if we have fragments pending or if someone
9178 9178 * wanted us to be scheduled again.
9179 9179 */
9180 9180 mutex_enter(&ill->ill_lock);
9181 9181 ill->ill_fragtimer_executing = 0;
9182 9182 ill->ill_frag_timer_id = 0;
9183 9183 if (frag_pending || ill->ill_fragtimer_needrestart)
9184 9184 ill_frag_timer_start(ill);
9185 9185 mutex_exit(&ill->ill_lock);
9186 9186 }
9187 9187
9188 9188 void
9189 9189 ill_frag_timer_start(ill_t *ill)
9190 9190 {
9191 9191 ip_stack_t *ipst = ill->ill_ipst;
9192 9192 clock_t timeo_ms;
9193 9193
9194 9194 ASSERT(MUTEX_HELD(&ill->ill_lock));
9195 9195
9196 9196 /* If the ill is closing or opening don't proceed */
9197 9197 if (ill->ill_state_flags & ILL_CONDEMNED)
9198 9198 return;
9199 9199
9200 9200 if (ill->ill_fragtimer_executing) {
9201 9201 /*
9202 9202 * ill_frag_timer is currently executing. Just record the
9203 9203 * the fact that we want the timer to be restarted.
9204 9204 * ill_frag_timer will post a timeout before it returns,
9205 9205 * ensuring it will be called again.
9206 9206 */
9207 9207 ill->ill_fragtimer_needrestart = 1;
9208 9208 return;
9209 9209 }
9210 9210
9211 9211 if (ill->ill_frag_timer_id == 0) {
9212 9212 timeo_ms = (ill->ill_isv6 ? ipst->ips_ipv6_reassembly_timeout :
9213 9213 ipst->ips_ip_reassembly_timeout) * SECONDS;
9214 9214
9215 9215 /*
9216 9216 * The timer is neither running nor is the timeout handler
9217 9217 * executing. Post a timeout so that ill_frag_timer will be
9218 9218 * called
9219 9219 */
9220 9220 ill->ill_frag_timer_id = timeout(ill_frag_timer, ill,
9221 9221 MSEC_TO_TICK(timeo_ms >> 1));
9222 9222 ill->ill_fragtimer_needrestart = 0;
9223 9223 }
9224 9224 }
9225 9225
9226 9226 /*
9227 9227 * Update any source route, record route or timestamp options.
9228 9228 * Check that we are at end of strict source route.
9229 9229 * The options have already been checked for sanity in ip_input_options().
9230 9230 */
9231 9231 boolean_t
9232 9232 ip_input_local_options(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
9233 9233 {
9234 9234 ipoptp_t opts;
9235 9235 uchar_t *opt;
9236 9236 uint8_t optval;
9237 9237 uint8_t optlen;
9238 9238 ipaddr_t dst;
9239 9239 ipaddr_t ifaddr;
9240 9240 uint32_t ts;
9241 9241 timestruc_t now;
9242 9242 ill_t *ill = ira->ira_ill;
9243 9243 ip_stack_t *ipst = ill->ill_ipst;
9244 9244
9245 9245 ip2dbg(("ip_input_local_options\n"));
9246 9246
9247 9247 for (optval = ipoptp_first(&opts, ipha);
9248 9248 optval != IPOPT_EOL;
9249 9249 optval = ipoptp_next(&opts)) {
9250 9250 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
9251 9251 opt = opts.ipoptp_cur;
9252 9252 optlen = opts.ipoptp_len;
9253 9253 ip2dbg(("ip_input_local_options: opt %d, len %d\n",
9254 9254 optval, optlen));
9255 9255 switch (optval) {
9256 9256 uint32_t off;
9257 9257 case IPOPT_SSRR:
9258 9258 case IPOPT_LSRR:
9259 9259 off = opt[IPOPT_OFFSET];
9260 9260 off--;
9261 9261 if (optlen < IP_ADDR_LEN ||
9262 9262 off > optlen - IP_ADDR_LEN) {
9263 9263 /* End of source route */
9264 9264 ip1dbg(("ip_input_local_options: end of SR\n"));
9265 9265 break;
9266 9266 }
9267 9267 /*
9268 9268 * This will only happen if two consecutive entries
9269 9269 * in the source route contains our address or if
9270 9270 * it is a packet with a loose source route which
9271 9271 * reaches us before consuming the whole source route
9272 9272 */
9273 9273 ip1dbg(("ip_input_local_options: not end of SR\n"));
9274 9274 if (optval == IPOPT_SSRR) {
9275 9275 goto bad_src_route;
9276 9276 }
9277 9277 /*
9278 9278 * Hack: instead of dropping the packet truncate the
9279 9279 * source route to what has been used by filling the
9280 9280 * rest with IPOPT_NOP.
9281 9281 */
9282 9282 opt[IPOPT_OLEN] = (uint8_t)off;
9283 9283 while (off < optlen) {
9284 9284 opt[off++] = IPOPT_NOP;
9285 9285 }
9286 9286 break;
9287 9287 case IPOPT_RR:
9288 9288 off = opt[IPOPT_OFFSET];
9289 9289 off--;
9290 9290 if (optlen < IP_ADDR_LEN ||
9291 9291 off > optlen - IP_ADDR_LEN) {
9292 9292 /* No more room - ignore */
9293 9293 ip1dbg((
9294 9294 "ip_input_local_options: end of RR\n"));
9295 9295 break;
9296 9296 }
9297 9297 /* Pick a reasonable address on the outbound if */
9298 9298 if (ip_select_source_v4(ill, INADDR_ANY, ipha->ipha_dst,
9299 9299 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9300 9300 NULL) != 0) {
9301 9301 /* No source! Shouldn't happen */
9302 9302 ifaddr = INADDR_ANY;
9303 9303 }
9304 9304 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9305 9305 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9306 9306 break;
9307 9307 case IPOPT_TS:
9308 9308 /* Insert timestamp if there is romm */
9309 9309 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9310 9310 case IPOPT_TS_TSONLY:
9311 9311 off = IPOPT_TS_TIMELEN;
9312 9312 break;
9313 9313 case IPOPT_TS_PRESPEC:
9314 9314 case IPOPT_TS_PRESPEC_RFC791:
9315 9315 /* Verify that the address matched */
9316 9316 off = opt[IPOPT_OFFSET] - 1;
9317 9317 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9318 9318 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9319 9319 /* Not for us */
9320 9320 break;
9321 9321 }
9322 9322 /* FALLTHROUGH */
9323 9323 case IPOPT_TS_TSANDADDR:
9324 9324 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9325 9325 break;
9326 9326 default:
9327 9327 /*
9328 9328 * ip_*put_options should have already
9329 9329 * dropped this packet.
9330 9330 */
9331 9331 cmn_err(CE_PANIC, "ip_input_local_options: "
9332 9332 "unknown IT - bug in ip_input_options?\n");
9333 9333 return (B_TRUE); /* Keep "lint" happy */
9334 9334 }
9335 9335 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
9336 9336 /* Increase overflow counter */
9337 9337 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
9338 9338 opt[IPOPT_POS_OV_FLG] =
9339 9339 (uint8_t)((opt[IPOPT_POS_OV_FLG] & 0x0F) |
9340 9340 (off << 4));
9341 9341 break;
9342 9342 }
9343 9343 off = opt[IPOPT_OFFSET] - 1;
9344 9344 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9345 9345 case IPOPT_TS_PRESPEC:
9346 9346 case IPOPT_TS_PRESPEC_RFC791:
9347 9347 case IPOPT_TS_TSANDADDR:
9348 9348 /* Pick a reasonable addr on the outbound if */
9349 9349 if (ip_select_source_v4(ill, INADDR_ANY,
9350 9350 ipha->ipha_dst, INADDR_ANY, ALL_ZONES, ipst,
9351 9351 &ifaddr, NULL, NULL) != 0) {
9352 9352 /* No source! Shouldn't happen */
9353 9353 ifaddr = INADDR_ANY;
9354 9354 }
9355 9355 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9356 9356 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9357 9357 /* FALLTHROUGH */
9358 9358 case IPOPT_TS_TSONLY:
9359 9359 off = opt[IPOPT_OFFSET] - 1;
9360 9360 /* Compute # of milliseconds since midnight */
9361 9361 gethrestime(&now);
9362 9362 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
9363 9363 NSEC2MSEC(now.tv_nsec);
9364 9364 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
9365 9365 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
9366 9366 break;
9367 9367 }
9368 9368 break;
9369 9369 }
9370 9370 }
9371 9371 return (B_TRUE);
9372 9372
9373 9373 bad_src_route:
9374 9374 /* make sure we clear any indication of a hardware checksum */
9375 9375 DB_CKSUMFLAGS(mp) = 0;
9376 9376 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ill);
9377 9377 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
9378 9378 return (B_FALSE);
9379 9379
9380 9380 }
9381 9381
9382 9382 /*
9383 9383 * Process IP options in an inbound packet. Always returns the nexthop.
9384 9384 * Normally this is the passed in nexthop, but if there is an option
9385 9385 * that effects the nexthop (such as a source route) that will be returned.
9386 9386 * Sets *errorp if there is an error, in which case an ICMP error has been sent
9387 9387 * and mp freed.
9388 9388 */
9389 9389 ipaddr_t
9390 9390 ip_input_options(ipha_t *ipha, ipaddr_t dst, mblk_t *mp,
9391 9391 ip_recv_attr_t *ira, int *errorp)
9392 9392 {
9393 9393 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
9394 9394 ipoptp_t opts;
9395 9395 uchar_t *opt;
9396 9396 uint8_t optval;
9397 9397 uint8_t optlen;
9398 9398 intptr_t code = 0;
9399 9399 ire_t *ire;
9400 9400
9401 9401 ip2dbg(("ip_input_options\n"));
9402 9402 *errorp = 0;
9403 9403 for (optval = ipoptp_first(&opts, ipha);
9404 9404 optval != IPOPT_EOL;
9405 9405 optval = ipoptp_next(&opts)) {
9406 9406 opt = opts.ipoptp_cur;
9407 9407 optlen = opts.ipoptp_len;
9408 9408 ip2dbg(("ip_input_options: opt %d, len %d\n",
9409 9409 optval, optlen));
9410 9410 /*
9411 9411 * Note: we need to verify the checksum before we
9412 9412 * modify anything thus this routine only extracts the next
9413 9413 * hop dst from any source route.
9414 9414 */
9415 9415 switch (optval) {
9416 9416 uint32_t off;
9417 9417 case IPOPT_SSRR:
9418 9418 case IPOPT_LSRR:
9419 9419 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9420 9420 if (optval == IPOPT_SSRR) {
9421 9421 ip1dbg(("ip_input_options: not next"
9422 9422 " strict source route 0x%x\n",
9423 9423 ntohl(dst)));
9424 9424 code = (char *)&ipha->ipha_dst -
9425 9425 (char *)ipha;
9426 9426 goto param_prob; /* RouterReq's */
9427 9427 }
9428 9428 ip2dbg(("ip_input_options: "
9429 9429 "not next source route 0x%x\n",
9430 9430 ntohl(dst)));
9431 9431 break;
9432 9432 }
9433 9433
9434 9434 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9435 9435 ip1dbg((
9436 9436 "ip_input_options: bad option offset\n"));
9437 9437 code = (char *)&opt[IPOPT_OLEN] -
9438 9438 (char *)ipha;
9439 9439 goto param_prob;
9440 9440 }
9441 9441 off = opt[IPOPT_OFFSET];
9442 9442 off--;
9443 9443 redo_srr:
9444 9444 if (optlen < IP_ADDR_LEN ||
9445 9445 off > optlen - IP_ADDR_LEN) {
9446 9446 /* End of source route */
9447 9447 ip1dbg(("ip_input_options: end of SR\n"));
9448 9448 break;
9449 9449 }
9450 9450 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9451 9451 ip1dbg(("ip_input_options: next hop 0x%x\n",
9452 9452 ntohl(dst)));
9453 9453
9454 9454 /*
9455 9455 * Check if our address is present more than
9456 9456 * once as consecutive hops in source route.
9457 9457 * XXX verify per-interface ip_forwarding
9458 9458 * for source route?
9459 9459 */
9460 9460 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
9461 9461 off += IP_ADDR_LEN;
9462 9462 goto redo_srr;
9463 9463 }
9464 9464
9465 9465 if (dst == htonl(INADDR_LOOPBACK)) {
9466 9466 ip1dbg(("ip_input_options: loopback addr in "
9467 9467 "source route!\n"));
9468 9468 goto bad_src_route;
9469 9469 }
9470 9470 /*
9471 9471 * For strict: verify that dst is directly
9472 9472 * reachable.
9473 9473 */
9474 9474 if (optval == IPOPT_SSRR) {
9475 9475 ire = ire_ftable_lookup_v4(dst, 0, 0,
9476 9476 IRE_INTERFACE, NULL, ALL_ZONES,
9477 9477 ira->ira_tsl,
9478 9478 MATCH_IRE_TYPE | MATCH_IRE_SECATTR, 0, ipst,
9479 9479 NULL);
9480 9480 if (ire == NULL) {
9481 9481 ip1dbg(("ip_input_options: SSRR not "
9482 9482 "directly reachable: 0x%x\n",
9483 9483 ntohl(dst)));
9484 9484 goto bad_src_route;
9485 9485 }
9486 9486 ire_refrele(ire);
9487 9487 }
9488 9488 /*
9489 9489 * Defer update of the offset and the record route
9490 9490 * until the packet is forwarded.
9491 9491 */
9492 9492 break;
9493 9493 case IPOPT_RR:
9494 9494 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9495 9495 ip1dbg((
9496 9496 "ip_input_options: bad option offset\n"));
9497 9497 code = (char *)&opt[IPOPT_OLEN] -
9498 9498 (char *)ipha;
9499 9499 goto param_prob;
9500 9500 }
9501 9501 break;
9502 9502 case IPOPT_TS:
9503 9503 /*
9504 9504 * Verify that length >= 5 and that there is either
9505 9505 * room for another timestamp or that the overflow
9506 9506 * counter is not maxed out.
9507 9507 */
9508 9508 code = (char *)&opt[IPOPT_OLEN] - (char *)ipha;
9509 9509 if (optlen < IPOPT_MINLEN_IT) {
9510 9510 goto param_prob;
9511 9511 }
9512 9512 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9513 9513 ip1dbg((
9514 9514 "ip_input_options: bad option offset\n"));
9515 9515 code = (char *)&opt[IPOPT_OFFSET] -
9516 9516 (char *)ipha;
9517 9517 goto param_prob;
9518 9518 }
9519 9519 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9520 9520 case IPOPT_TS_TSONLY:
9521 9521 off = IPOPT_TS_TIMELEN;
9522 9522 break;
9523 9523 case IPOPT_TS_TSANDADDR:
9524 9524 case IPOPT_TS_PRESPEC:
9525 9525 case IPOPT_TS_PRESPEC_RFC791:
9526 9526 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9527 9527 break;
9528 9528 default:
9529 9529 code = (char *)&opt[IPOPT_POS_OV_FLG] -
9530 9530 (char *)ipha;
9531 9531 goto param_prob;
9532 9532 }
9533 9533 if (opt[IPOPT_OFFSET] - 1 + off > optlen &&
9534 9534 (opt[IPOPT_POS_OV_FLG] & 0xF0) == 0xF0) {
9535 9535 /*
9536 9536 * No room and the overflow counter is 15
9537 9537 * already.
9538 9538 */
9539 9539 goto param_prob;
9540 9540 }
9541 9541 break;
9542 9542 }
9543 9543 }
9544 9544
9545 9545 if ((opts.ipoptp_flags & IPOPTP_ERROR) == 0) {
9546 9546 return (dst);
9547 9547 }
9548 9548
9549 9549 ip1dbg(("ip_input_options: error processing IP options."));
9550 9550 code = (char *)&opt[IPOPT_OFFSET] - (char *)ipha;
9551 9551
9552 9552 param_prob:
9553 9553 /* make sure we clear any indication of a hardware checksum */
9554 9554 DB_CKSUMFLAGS(mp) = 0;
9555 9555 ip_drop_input("ICMP_PARAM_PROBLEM", mp, ira->ira_ill);
9556 9556 icmp_param_problem(mp, (uint8_t)code, ira);
9557 9557 *errorp = -1;
9558 9558 return (dst);
9559 9559
9560 9560 bad_src_route:
9561 9561 /* make sure we clear any indication of a hardware checksum */
9562 9562 DB_CKSUMFLAGS(mp) = 0;
9563 9563 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ira->ira_ill);
9564 9564 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
9565 9565 *errorp = -1;
9566 9566 return (dst);
9567 9567 }
9568 9568
9569 9569 /*
9570 9570 * IP & ICMP info in >=14 msg's ...
9571 9571 * - ip fixed part (mib2_ip_t)
9572 9572 * - icmp fixed part (mib2_icmp_t)
9573 9573 * - ipAddrEntryTable (ip 20) all IPv4 ipifs
9574 9574 * - ipRouteEntryTable (ip 21) all IPv4 IREs
9575 9575 * - ipNetToMediaEntryTable (ip 22) all IPv4 Neighbor Cache entries
9576 9576 * - ipRouteAttributeTable (ip 102) labeled routes
9577 9577 * - ip multicast membership (ip_member_t)
9578 9578 * - ip multicast source filtering (ip_grpsrc_t)
9579 9579 * - igmp fixed part (struct igmpstat)
9580 9580 * - multicast routing stats (struct mrtstat)
9581 9581 * - multicast routing vifs (array of struct vifctl)
9582 9582 * - multicast routing routes (array of struct mfcctl)
9583 9583 * - ip6 fixed part (mib2_ipv6IfStatsEntry_t)
9584 9584 * One per ill plus one generic
9585 9585 * - icmp6 fixed part (mib2_ipv6IfIcmpEntry_t)
9586 9586 * One per ill plus one generic
9587 9587 * - ipv6RouteEntry all IPv6 IREs
9588 9588 * - ipv6RouteAttributeTable (ip6 102) labeled routes
9589 9589 * - ipv6NetToMediaEntry all IPv6 Neighbor Cache entries
9590 9590 * - ipv6AddrEntry all IPv6 ipifs
9591 9591 * - ipv6 multicast membership (ipv6_member_t)
9592 9592 * - ipv6 multicast source filtering (ipv6_grpsrc_t)
9593 9593 *
9594 9594 * NOTE: original mpctl is copied for msg's 2..N, since its ctl part is
9595 9595 * already filled in by the caller.
9596 9596 * If legacy_req is true then MIB structures needs to be truncated to their
9597 9597 * legacy sizes before being returned.
9598 9598 * Return value of 0 indicates that no messages were sent and caller
9599 9599 * should free mpctl.
9600 9600 */
9601 9601 int
9602 9602 ip_snmp_get(queue_t *q, mblk_t *mpctl, int level, boolean_t legacy_req)
9603 9603 {
9604 9604 ip_stack_t *ipst;
9605 9605 sctp_stack_t *sctps;
9606 9606
9607 9607 if (q->q_next != NULL) {
9608 9608 ipst = ILLQ_TO_IPST(q);
9609 9609 } else {
9610 9610 ipst = CONNQ_TO_IPST(q);
9611 9611 }
9612 9612 ASSERT(ipst != NULL);
9613 9613 sctps = ipst->ips_netstack->netstack_sctp;
9614 9614
9615 9615 if (mpctl == NULL || mpctl->b_cont == NULL) {
9616 9616 return (0);
9617 9617 }
9618 9618
9619 9619 /*
9620 9620 * For the purposes of the (broken) packet shell use
9621 9621 * of the level we make sure MIB2_TCP/MIB2_UDP can be used
9622 9622 * to make TCP and UDP appear first in the list of mib items.
9623 9623 * TBD: We could expand this and use it in netstat so that
9624 9624 * the kernel doesn't have to produce large tables (connections,
9625 9625 * routes, etc) when netstat only wants the statistics or a particular
9626 9626 * table.
9627 9627 */
|
↓ open down ↓ |
9627 lines elided |
↑ open up ↑ |
9628 9628 if (!(level == MIB2_TCP || level == MIB2_UDP)) {
9629 9629 if ((mpctl = icmp_snmp_get(q, mpctl)) == NULL) {
9630 9630 return (1);
9631 9631 }
9632 9632 }
9633 9633
9634 9634 if (level != MIB2_TCP) {
9635 9635 if ((mpctl = udp_snmp_get(q, mpctl, legacy_req)) == NULL) {
9636 9636 return (1);
9637 9637 }
9638 + if (level == MIB2_UDP) {
9639 + goto done;
9640 + }
9638 9641 }
9639 9642
9640 9643 if (level != MIB2_UDP) {
9641 9644 if ((mpctl = tcp_snmp_get(q, mpctl, legacy_req)) == NULL) {
9642 9645 return (1);
9643 9646 }
9647 + if (level == MIB2_TCP) {
9648 + goto done;
9649 + }
9644 9650 }
9645 9651
9646 9652 if ((mpctl = ip_snmp_get_mib2_ip_traffic_stats(q, mpctl,
9647 9653 ipst, legacy_req)) == NULL) {
9648 9654 return (1);
9649 9655 }
9650 9656
9651 9657 if ((mpctl = ip_snmp_get_mib2_ip6(q, mpctl, ipst,
9652 9658 legacy_req)) == NULL) {
9653 9659 return (1);
9654 9660 }
9655 9661
9656 9662 if ((mpctl = ip_snmp_get_mib2_icmp(q, mpctl, ipst)) == NULL) {
9657 9663 return (1);
9658 9664 }
9659 9665
9660 9666 if ((mpctl = ip_snmp_get_mib2_icmp6(q, mpctl, ipst)) == NULL) {
9661 9667 return (1);
9662 9668 }
9663 9669
9664 9670 if ((mpctl = ip_snmp_get_mib2_igmp(q, mpctl, ipst)) == NULL) {
9665 9671 return (1);
9666 9672 }
9667 9673
9668 9674 if ((mpctl = ip_snmp_get_mib2_multi(q, mpctl, ipst)) == NULL) {
9669 9675 return (1);
9670 9676 }
9671 9677
9672 9678 if ((mpctl = ip_snmp_get_mib2_ip_addr(q, mpctl, ipst,
9673 9679 legacy_req)) == NULL) {
9674 9680 return (1);
9675 9681 }
9676 9682
9677 9683 if ((mpctl = ip_snmp_get_mib2_ip6_addr(q, mpctl, ipst,
9678 9684 legacy_req)) == NULL) {
9679 9685 return (1);
9680 9686 }
9681 9687
9682 9688 if ((mpctl = ip_snmp_get_mib2_ip_group_mem(q, mpctl, ipst)) == NULL) {
9683 9689 return (1);
9684 9690 }
9685 9691
9686 9692 if ((mpctl = ip_snmp_get_mib2_ip6_group_mem(q, mpctl, ipst)) == NULL) {
9687 9693 return (1);
9688 9694 }
9689 9695
9690 9696 if ((mpctl = ip_snmp_get_mib2_ip_group_src(q, mpctl, ipst)) == NULL) {
9691 9697 return (1);
9692 9698 }
9693 9699
9694 9700 if ((mpctl = ip_snmp_get_mib2_ip6_group_src(q, mpctl, ipst)) == NULL) {
9695 9701 return (1);
9696 9702 }
9697 9703
9698 9704 if ((mpctl = ip_snmp_get_mib2_virt_multi(q, mpctl, ipst)) == NULL) {
9699 9705 return (1);
9700 9706 }
9701 9707
9702 9708 if ((mpctl = ip_snmp_get_mib2_multi_rtable(q, mpctl, ipst)) == NULL) {
9703 9709 return (1);
9704 9710 }
9705 9711
9706 9712 mpctl = ip_snmp_get_mib2_ip_route_media(q, mpctl, level, ipst);
9707 9713 if (mpctl == NULL)
9708 9714 return (1);
9709 9715
|
↓ open down ↓ |
56 lines elided |
↑ open up ↑ |
9710 9716 mpctl = ip_snmp_get_mib2_ip6_route_media(q, mpctl, level, ipst);
9711 9717 if (mpctl == NULL)
9712 9718 return (1);
9713 9719
9714 9720 if ((mpctl = sctp_snmp_get_mib2(q, mpctl, sctps)) == NULL) {
9715 9721 return (1);
9716 9722 }
9717 9723 if ((mpctl = ip_snmp_get_mib2_ip_dce(q, mpctl, ipst)) == NULL) {
9718 9724 return (1);
9719 9725 }
9726 +done:
9720 9727 freemsg(mpctl);
9721 9728 return (1);
9722 9729 }
9723 9730
9724 9731 /* Get global (legacy) IPv4 statistics */
9725 9732 static mblk_t *
9726 9733 ip_snmp_get_mib2_ip(queue_t *q, mblk_t *mpctl, mib2_ipIfStatsEntry_t *ipmib,
9727 9734 ip_stack_t *ipst, boolean_t legacy_req)
9728 9735 {
9729 9736 mib2_ip_t old_ip_mib;
9730 9737 struct opthdr *optp;
9731 9738 mblk_t *mp2ctl;
9732 9739 mib2_ipAddrEntry_t mae;
9733 9740
9734 9741 /*
9735 9742 * make a copy of the original message
9736 9743 */
9737 9744 mp2ctl = copymsg(mpctl);
9738 9745
9739 9746 /* fixed length IP structure... */
9740 9747 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9741 9748 optp->level = MIB2_IP;
9742 9749 optp->name = 0;
9743 9750 SET_MIB(old_ip_mib.ipForwarding,
9744 9751 (WE_ARE_FORWARDING(ipst) ? 1 : 2));
9745 9752 SET_MIB(old_ip_mib.ipDefaultTTL,
9746 9753 (uint32_t)ipst->ips_ip_def_ttl);
9747 9754 SET_MIB(old_ip_mib.ipReasmTimeout,
9748 9755 ipst->ips_ip_reassembly_timeout);
9749 9756 SET_MIB(old_ip_mib.ipAddrEntrySize,
9750 9757 (legacy_req) ? LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t) :
9751 9758 sizeof (mib2_ipAddrEntry_t));
9752 9759 SET_MIB(old_ip_mib.ipRouteEntrySize,
9753 9760 sizeof (mib2_ipRouteEntry_t));
9754 9761 SET_MIB(old_ip_mib.ipNetToMediaEntrySize,
9755 9762 sizeof (mib2_ipNetToMediaEntry_t));
9756 9763 SET_MIB(old_ip_mib.ipMemberEntrySize, sizeof (ip_member_t));
9757 9764 SET_MIB(old_ip_mib.ipGroupSourceEntrySize, sizeof (ip_grpsrc_t));
9758 9765 SET_MIB(old_ip_mib.ipRouteAttributeSize,
9759 9766 sizeof (mib2_ipAttributeEntry_t));
9760 9767 SET_MIB(old_ip_mib.transportMLPSize, sizeof (mib2_transportMLPEntry_t));
9761 9768 SET_MIB(old_ip_mib.ipDestEntrySize, sizeof (dest_cache_entry_t));
9762 9769
9763 9770 /*
9764 9771 * Grab the statistics from the new IP MIB
9765 9772 */
9766 9773 SET_MIB(old_ip_mib.ipInReceives,
9767 9774 (uint32_t)ipmib->ipIfStatsHCInReceives);
9768 9775 SET_MIB(old_ip_mib.ipInHdrErrors, ipmib->ipIfStatsInHdrErrors);
9769 9776 SET_MIB(old_ip_mib.ipInAddrErrors, ipmib->ipIfStatsInAddrErrors);
9770 9777 SET_MIB(old_ip_mib.ipForwDatagrams,
9771 9778 (uint32_t)ipmib->ipIfStatsHCOutForwDatagrams);
9772 9779 SET_MIB(old_ip_mib.ipInUnknownProtos,
9773 9780 ipmib->ipIfStatsInUnknownProtos);
9774 9781 SET_MIB(old_ip_mib.ipInDiscards, ipmib->ipIfStatsInDiscards);
9775 9782 SET_MIB(old_ip_mib.ipInDelivers,
9776 9783 (uint32_t)ipmib->ipIfStatsHCInDelivers);
9777 9784 SET_MIB(old_ip_mib.ipOutRequests,
9778 9785 (uint32_t)ipmib->ipIfStatsHCOutRequests);
9779 9786 SET_MIB(old_ip_mib.ipOutDiscards, ipmib->ipIfStatsOutDiscards);
9780 9787 SET_MIB(old_ip_mib.ipOutNoRoutes, ipmib->ipIfStatsOutNoRoutes);
9781 9788 SET_MIB(old_ip_mib.ipReasmReqds, ipmib->ipIfStatsReasmReqds);
9782 9789 SET_MIB(old_ip_mib.ipReasmOKs, ipmib->ipIfStatsReasmOKs);
9783 9790 SET_MIB(old_ip_mib.ipReasmFails, ipmib->ipIfStatsReasmFails);
9784 9791 SET_MIB(old_ip_mib.ipFragOKs, ipmib->ipIfStatsOutFragOKs);
9785 9792 SET_MIB(old_ip_mib.ipFragFails, ipmib->ipIfStatsOutFragFails);
9786 9793 SET_MIB(old_ip_mib.ipFragCreates, ipmib->ipIfStatsOutFragCreates);
9787 9794
9788 9795 /* ipRoutingDiscards is not being used */
9789 9796 SET_MIB(old_ip_mib.ipRoutingDiscards, 0);
9790 9797 SET_MIB(old_ip_mib.tcpInErrs, ipmib->tcpIfStatsInErrs);
9791 9798 SET_MIB(old_ip_mib.udpNoPorts, ipmib->udpIfStatsNoPorts);
9792 9799 SET_MIB(old_ip_mib.ipInCksumErrs, ipmib->ipIfStatsInCksumErrs);
9793 9800 SET_MIB(old_ip_mib.ipReasmDuplicates,
9794 9801 ipmib->ipIfStatsReasmDuplicates);
9795 9802 SET_MIB(old_ip_mib.ipReasmPartDups, ipmib->ipIfStatsReasmPartDups);
9796 9803 SET_MIB(old_ip_mib.ipForwProhibits, ipmib->ipIfStatsForwProhibits);
9797 9804 SET_MIB(old_ip_mib.udpInCksumErrs, ipmib->udpIfStatsInCksumErrs);
9798 9805 SET_MIB(old_ip_mib.udpInOverflows, ipmib->udpIfStatsInOverflows);
9799 9806 SET_MIB(old_ip_mib.rawipInOverflows,
9800 9807 ipmib->rawipIfStatsInOverflows);
9801 9808
9802 9809 SET_MIB(old_ip_mib.ipsecInSucceeded, ipmib->ipsecIfStatsInSucceeded);
9803 9810 SET_MIB(old_ip_mib.ipsecInFailed, ipmib->ipsecIfStatsInFailed);
9804 9811 SET_MIB(old_ip_mib.ipInIPv6, ipmib->ipIfStatsInWrongIPVersion);
9805 9812 SET_MIB(old_ip_mib.ipOutIPv6, ipmib->ipIfStatsOutWrongIPVersion);
9806 9813 SET_MIB(old_ip_mib.ipOutSwitchIPv6,
9807 9814 ipmib->ipIfStatsOutSwitchIPVersion);
9808 9815
9809 9816 if (!snmp_append_data(mpctl->b_cont, (char *)&old_ip_mib,
9810 9817 (int)sizeof (old_ip_mib))) {
9811 9818 ip1dbg(("ip_snmp_get_mib2_ip: failed to allocate %u bytes\n",
9812 9819 (uint_t)sizeof (old_ip_mib)));
9813 9820 }
9814 9821
9815 9822 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9816 9823 ip3dbg(("ip_snmp_get_mib2_ip: level %d, name %d, len %d\n",
9817 9824 (int)optp->level, (int)optp->name, (int)optp->len));
9818 9825 qreply(q, mpctl);
9819 9826 return (mp2ctl);
9820 9827 }
9821 9828
9822 9829 /* Per interface IPv4 statistics */
9823 9830 static mblk_t *
9824 9831 ip_snmp_get_mib2_ip_traffic_stats(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
9825 9832 boolean_t legacy_req)
9826 9833 {
9827 9834 struct opthdr *optp;
9828 9835 mblk_t *mp2ctl;
9829 9836 ill_t *ill;
9830 9837 ill_walk_context_t ctx;
9831 9838 mblk_t *mp_tail = NULL;
9832 9839 mib2_ipIfStatsEntry_t global_ip_mib;
9833 9840 mib2_ipAddrEntry_t mae;
9834 9841
9835 9842 /*
9836 9843 * Make a copy of the original message
9837 9844 */
9838 9845 mp2ctl = copymsg(mpctl);
9839 9846
9840 9847 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9841 9848 optp->level = MIB2_IP;
9842 9849 optp->name = MIB2_IP_TRAFFIC_STATS;
9843 9850 /* Include "unknown interface" ip_mib */
9844 9851 ipst->ips_ip_mib.ipIfStatsIPVersion = MIB2_INETADDRESSTYPE_ipv4;
9845 9852 ipst->ips_ip_mib.ipIfStatsIfIndex =
9846 9853 MIB2_UNKNOWN_INTERFACE; /* Flag to netstat */
9847 9854 SET_MIB(ipst->ips_ip_mib.ipIfStatsForwarding,
9848 9855 (ipst->ips_ip_forwarding ? 1 : 2));
9849 9856 SET_MIB(ipst->ips_ip_mib.ipIfStatsDefaultTTL,
9850 9857 (uint32_t)ipst->ips_ip_def_ttl);
9851 9858 SET_MIB(ipst->ips_ip_mib.ipIfStatsEntrySize,
9852 9859 sizeof (mib2_ipIfStatsEntry_t));
9853 9860 SET_MIB(ipst->ips_ip_mib.ipIfStatsAddrEntrySize,
9854 9861 sizeof (mib2_ipAddrEntry_t));
9855 9862 SET_MIB(ipst->ips_ip_mib.ipIfStatsRouteEntrySize,
9856 9863 sizeof (mib2_ipRouteEntry_t));
9857 9864 SET_MIB(ipst->ips_ip_mib.ipIfStatsNetToMediaEntrySize,
9858 9865 sizeof (mib2_ipNetToMediaEntry_t));
9859 9866 SET_MIB(ipst->ips_ip_mib.ipIfStatsMemberEntrySize,
9860 9867 sizeof (ip_member_t));
9861 9868 SET_MIB(ipst->ips_ip_mib.ipIfStatsGroupSourceEntrySize,
9862 9869 sizeof (ip_grpsrc_t));
9863 9870
9864 9871 bcopy(&ipst->ips_ip_mib, &global_ip_mib, sizeof (global_ip_mib));
9865 9872
9866 9873 if (legacy_req) {
9867 9874 SET_MIB(global_ip_mib.ipIfStatsAddrEntrySize,
9868 9875 LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t));
9869 9876 }
9870 9877
9871 9878 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
9872 9879 (char *)&global_ip_mib, (int)sizeof (global_ip_mib))) {
9873 9880 ip1dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9874 9881 "failed to allocate %u bytes\n",
9875 9882 (uint_t)sizeof (global_ip_mib)));
9876 9883 }
9877 9884
9878 9885 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
9879 9886 ill = ILL_START_WALK_V4(&ctx, ipst);
9880 9887 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
9881 9888 ill->ill_ip_mib->ipIfStatsIfIndex =
9882 9889 ill->ill_phyint->phyint_ifindex;
9883 9890 SET_MIB(ill->ill_ip_mib->ipIfStatsForwarding,
9884 9891 (ipst->ips_ip_forwarding ? 1 : 2));
9885 9892 SET_MIB(ill->ill_ip_mib->ipIfStatsDefaultTTL,
9886 9893 (uint32_t)ipst->ips_ip_def_ttl);
9887 9894
9888 9895 ip_mib2_add_ip_stats(&global_ip_mib, ill->ill_ip_mib);
9889 9896 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
9890 9897 (char *)ill->ill_ip_mib,
9891 9898 (int)sizeof (*ill->ill_ip_mib))) {
9892 9899 ip1dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9893 9900 "failed to allocate %u bytes\n",
9894 9901 (uint_t)sizeof (*ill->ill_ip_mib)));
9895 9902 }
9896 9903 }
9897 9904 rw_exit(&ipst->ips_ill_g_lock);
9898 9905
9899 9906 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9900 9907 ip3dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9901 9908 "level %d, name %d, len %d\n",
9902 9909 (int)optp->level, (int)optp->name, (int)optp->len));
9903 9910 qreply(q, mpctl);
9904 9911
9905 9912 if (mp2ctl == NULL)
9906 9913 return (NULL);
9907 9914
9908 9915 return (ip_snmp_get_mib2_ip(q, mp2ctl, &global_ip_mib, ipst,
9909 9916 legacy_req));
9910 9917 }
9911 9918
9912 9919 /* Global IPv4 ICMP statistics */
9913 9920 static mblk_t *
9914 9921 ip_snmp_get_mib2_icmp(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9915 9922 {
9916 9923 struct opthdr *optp;
9917 9924 mblk_t *mp2ctl;
9918 9925
9919 9926 /*
9920 9927 * Make a copy of the original message
9921 9928 */
9922 9929 mp2ctl = copymsg(mpctl);
9923 9930
9924 9931 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9925 9932 optp->level = MIB2_ICMP;
9926 9933 optp->name = 0;
9927 9934 if (!snmp_append_data(mpctl->b_cont, (char *)&ipst->ips_icmp_mib,
9928 9935 (int)sizeof (ipst->ips_icmp_mib))) {
9929 9936 ip1dbg(("ip_snmp_get_mib2_icmp: failed to allocate %u bytes\n",
9930 9937 (uint_t)sizeof (ipst->ips_icmp_mib)));
9931 9938 }
9932 9939 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9933 9940 ip3dbg(("ip_snmp_get_mib2_icmp: level %d, name %d, len %d\n",
9934 9941 (int)optp->level, (int)optp->name, (int)optp->len));
9935 9942 qreply(q, mpctl);
9936 9943 return (mp2ctl);
9937 9944 }
9938 9945
9939 9946 /* Global IPv4 IGMP statistics */
9940 9947 static mblk_t *
9941 9948 ip_snmp_get_mib2_igmp(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9942 9949 {
9943 9950 struct opthdr *optp;
9944 9951 mblk_t *mp2ctl;
9945 9952
9946 9953 /*
9947 9954 * make a copy of the original message
9948 9955 */
9949 9956 mp2ctl = copymsg(mpctl);
9950 9957
9951 9958 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9952 9959 optp->level = EXPER_IGMP;
9953 9960 optp->name = 0;
9954 9961 if (!snmp_append_data(mpctl->b_cont, (char *)&ipst->ips_igmpstat,
9955 9962 (int)sizeof (ipst->ips_igmpstat))) {
9956 9963 ip1dbg(("ip_snmp_get_mib2_igmp: failed to allocate %u bytes\n",
9957 9964 (uint_t)sizeof (ipst->ips_igmpstat)));
9958 9965 }
9959 9966 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9960 9967 ip3dbg(("ip_snmp_get_mib2_igmp: level %d, name %d, len %d\n",
9961 9968 (int)optp->level, (int)optp->name, (int)optp->len));
9962 9969 qreply(q, mpctl);
9963 9970 return (mp2ctl);
9964 9971 }
9965 9972
9966 9973 /* Global IPv4 Multicast Routing statistics */
9967 9974 static mblk_t *
9968 9975 ip_snmp_get_mib2_multi(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9969 9976 {
9970 9977 struct opthdr *optp;
9971 9978 mblk_t *mp2ctl;
9972 9979
9973 9980 /*
9974 9981 * make a copy of the original message
9975 9982 */
9976 9983 mp2ctl = copymsg(mpctl);
9977 9984
9978 9985 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9979 9986 optp->level = EXPER_DVMRP;
9980 9987 optp->name = 0;
9981 9988 if (!ip_mroute_stats(mpctl->b_cont, ipst)) {
9982 9989 ip0dbg(("ip_mroute_stats: failed\n"));
9983 9990 }
9984 9991 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9985 9992 ip3dbg(("ip_snmp_get_mib2_multi: level %d, name %d, len %d\n",
9986 9993 (int)optp->level, (int)optp->name, (int)optp->len));
9987 9994 qreply(q, mpctl);
9988 9995 return (mp2ctl);
9989 9996 }
9990 9997
9991 9998 /* IPv4 address information */
9992 9999 static mblk_t *
9993 10000 ip_snmp_get_mib2_ip_addr(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
9994 10001 boolean_t legacy_req)
9995 10002 {
9996 10003 struct opthdr *optp;
9997 10004 mblk_t *mp2ctl;
9998 10005 mblk_t *mp_tail = NULL;
9999 10006 ill_t *ill;
10000 10007 ipif_t *ipif;
10001 10008 uint_t bitval;
10002 10009 mib2_ipAddrEntry_t mae;
10003 10010 size_t mae_size;
10004 10011 zoneid_t zoneid;
10005 10012 ill_walk_context_t ctx;
10006 10013
10007 10014 /*
10008 10015 * make a copy of the original message
10009 10016 */
10010 10017 mp2ctl = copymsg(mpctl);
10011 10018
10012 10019 mae_size = (legacy_req) ? LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t) :
10013 10020 sizeof (mib2_ipAddrEntry_t);
10014 10021
10015 10022 /* ipAddrEntryTable */
10016 10023
10017 10024 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10018 10025 optp->level = MIB2_IP;
10019 10026 optp->name = MIB2_IP_ADDR;
10020 10027 zoneid = Q_TO_CONN(q)->conn_zoneid;
10021 10028
10022 10029 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10023 10030 ill = ILL_START_WALK_V4(&ctx, ipst);
10024 10031 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10025 10032 for (ipif = ill->ill_ipif; ipif != NULL;
10026 10033 ipif = ipif->ipif_next) {
10027 10034 if (ipif->ipif_zoneid != zoneid &&
10028 10035 ipif->ipif_zoneid != ALL_ZONES)
10029 10036 continue;
10030 10037 /* Sum of count from dead IRE_LO* and our current */
10031 10038 mae.ipAdEntInfo.ae_ibcnt = ipif->ipif_ib_pkt_count;
10032 10039 if (ipif->ipif_ire_local != NULL) {
10033 10040 mae.ipAdEntInfo.ae_ibcnt +=
10034 10041 ipif->ipif_ire_local->ire_ib_pkt_count;
10035 10042 }
10036 10043 mae.ipAdEntInfo.ae_obcnt = 0;
10037 10044 mae.ipAdEntInfo.ae_focnt = 0;
10038 10045
10039 10046 ipif_get_name(ipif, mae.ipAdEntIfIndex.o_bytes,
10040 10047 OCTET_LENGTH);
10041 10048 mae.ipAdEntIfIndex.o_length =
10042 10049 mi_strlen(mae.ipAdEntIfIndex.o_bytes);
10043 10050 mae.ipAdEntAddr = ipif->ipif_lcl_addr;
10044 10051 mae.ipAdEntNetMask = ipif->ipif_net_mask;
10045 10052 mae.ipAdEntInfo.ae_subnet = ipif->ipif_subnet;
10046 10053 mae.ipAdEntInfo.ae_subnet_len =
10047 10054 ip_mask_to_plen(ipif->ipif_net_mask);
10048 10055 mae.ipAdEntInfo.ae_src_addr = ipif->ipif_lcl_addr;
10049 10056 for (bitval = 1;
10050 10057 bitval &&
10051 10058 !(bitval & ipif->ipif_brd_addr);
10052 10059 bitval <<= 1)
10053 10060 noop;
10054 10061 mae.ipAdEntBcastAddr = bitval;
10055 10062 mae.ipAdEntReasmMaxSize = IP_MAXPACKET;
10056 10063 mae.ipAdEntInfo.ae_mtu = ipif->ipif_ill->ill_mtu;
10057 10064 mae.ipAdEntInfo.ae_metric = ipif->ipif_ill->ill_metric;
10058 10065 mae.ipAdEntInfo.ae_broadcast_addr =
10059 10066 ipif->ipif_brd_addr;
10060 10067 mae.ipAdEntInfo.ae_pp_dst_addr =
10061 10068 ipif->ipif_pp_dst_addr;
10062 10069 mae.ipAdEntInfo.ae_flags = ipif->ipif_flags |
10063 10070 ill->ill_flags | ill->ill_phyint->phyint_flags;
10064 10071 mae.ipAdEntRetransmitTime =
10065 10072 ill->ill_reachable_retrans_time;
10066 10073
10067 10074 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10068 10075 (char *)&mae, (int)mae_size)) {
10069 10076 ip1dbg(("ip_snmp_get_mib2_ip_addr: failed to "
10070 10077 "allocate %u bytes\n", (uint_t)mae_size));
10071 10078 }
10072 10079 }
10073 10080 }
10074 10081 rw_exit(&ipst->ips_ill_g_lock);
10075 10082
10076 10083 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10077 10084 ip3dbg(("ip_snmp_get_mib2_ip_addr: level %d, name %d, len %d\n",
10078 10085 (int)optp->level, (int)optp->name, (int)optp->len));
10079 10086 qreply(q, mpctl);
10080 10087 return (mp2ctl);
10081 10088 }
10082 10089
10083 10090 /* IPv6 address information */
10084 10091 static mblk_t *
10085 10092 ip_snmp_get_mib2_ip6_addr(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
10086 10093 boolean_t legacy_req)
10087 10094 {
10088 10095 struct opthdr *optp;
10089 10096 mblk_t *mp2ctl;
10090 10097 mblk_t *mp_tail = NULL;
10091 10098 ill_t *ill;
10092 10099 ipif_t *ipif;
10093 10100 mib2_ipv6AddrEntry_t mae6;
10094 10101 size_t mae6_size;
10095 10102 zoneid_t zoneid;
10096 10103 ill_walk_context_t ctx;
10097 10104
10098 10105 /*
10099 10106 * make a copy of the original message
10100 10107 */
10101 10108 mp2ctl = copymsg(mpctl);
10102 10109
10103 10110 mae6_size = (legacy_req) ?
10104 10111 LEGACY_MIB_SIZE(&mae6, mib2_ipv6AddrEntry_t) :
10105 10112 sizeof (mib2_ipv6AddrEntry_t);
10106 10113
10107 10114 /* ipv6AddrEntryTable */
10108 10115
10109 10116 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10110 10117 optp->level = MIB2_IP6;
10111 10118 optp->name = MIB2_IP6_ADDR;
10112 10119 zoneid = Q_TO_CONN(q)->conn_zoneid;
10113 10120
10114 10121 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10115 10122 ill = ILL_START_WALK_V6(&ctx, ipst);
10116 10123 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10117 10124 for (ipif = ill->ill_ipif; ipif != NULL;
10118 10125 ipif = ipif->ipif_next) {
10119 10126 if (ipif->ipif_zoneid != zoneid &&
10120 10127 ipif->ipif_zoneid != ALL_ZONES)
10121 10128 continue;
10122 10129 /* Sum of count from dead IRE_LO* and our current */
10123 10130 mae6.ipv6AddrInfo.ae_ibcnt = ipif->ipif_ib_pkt_count;
10124 10131 if (ipif->ipif_ire_local != NULL) {
10125 10132 mae6.ipv6AddrInfo.ae_ibcnt +=
10126 10133 ipif->ipif_ire_local->ire_ib_pkt_count;
10127 10134 }
10128 10135 mae6.ipv6AddrInfo.ae_obcnt = 0;
10129 10136 mae6.ipv6AddrInfo.ae_focnt = 0;
10130 10137
10131 10138 ipif_get_name(ipif, mae6.ipv6AddrIfIndex.o_bytes,
10132 10139 OCTET_LENGTH);
10133 10140 mae6.ipv6AddrIfIndex.o_length =
10134 10141 mi_strlen(mae6.ipv6AddrIfIndex.o_bytes);
10135 10142 mae6.ipv6AddrAddress = ipif->ipif_v6lcl_addr;
10136 10143 mae6.ipv6AddrPfxLength =
10137 10144 ip_mask_to_plen_v6(&ipif->ipif_v6net_mask);
10138 10145 mae6.ipv6AddrInfo.ae_subnet = ipif->ipif_v6subnet;
10139 10146 mae6.ipv6AddrInfo.ae_subnet_len =
10140 10147 mae6.ipv6AddrPfxLength;
10141 10148 mae6.ipv6AddrInfo.ae_src_addr = ipif->ipif_v6lcl_addr;
10142 10149
10143 10150 /* Type: stateless(1), stateful(2), unknown(3) */
10144 10151 if (ipif->ipif_flags & IPIF_ADDRCONF)
10145 10152 mae6.ipv6AddrType = 1;
10146 10153 else
10147 10154 mae6.ipv6AddrType = 2;
10148 10155 /* Anycast: true(1), false(2) */
10149 10156 if (ipif->ipif_flags & IPIF_ANYCAST)
10150 10157 mae6.ipv6AddrAnycastFlag = 1;
10151 10158 else
10152 10159 mae6.ipv6AddrAnycastFlag = 2;
10153 10160
10154 10161 /*
10155 10162 * Address status: preferred(1), deprecated(2),
10156 10163 * invalid(3), inaccessible(4), unknown(5)
10157 10164 */
10158 10165 if (ipif->ipif_flags & IPIF_NOLOCAL)
10159 10166 mae6.ipv6AddrStatus = 3;
10160 10167 else if (ipif->ipif_flags & IPIF_DEPRECATED)
10161 10168 mae6.ipv6AddrStatus = 2;
10162 10169 else
10163 10170 mae6.ipv6AddrStatus = 1;
10164 10171 mae6.ipv6AddrInfo.ae_mtu = ipif->ipif_ill->ill_mtu;
10165 10172 mae6.ipv6AddrInfo.ae_metric =
10166 10173 ipif->ipif_ill->ill_metric;
10167 10174 mae6.ipv6AddrInfo.ae_pp_dst_addr =
10168 10175 ipif->ipif_v6pp_dst_addr;
10169 10176 mae6.ipv6AddrInfo.ae_flags = ipif->ipif_flags |
10170 10177 ill->ill_flags | ill->ill_phyint->phyint_flags;
10171 10178 mae6.ipv6AddrReasmMaxSize = IP_MAXPACKET;
10172 10179 mae6.ipv6AddrIdentifier = ill->ill_token;
10173 10180 mae6.ipv6AddrIdentifierLen = ill->ill_token_length;
10174 10181 mae6.ipv6AddrReachableTime = ill->ill_reachable_time;
10175 10182 mae6.ipv6AddrRetransmitTime =
10176 10183 ill->ill_reachable_retrans_time;
10177 10184 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10178 10185 (char *)&mae6, (int)mae6_size)) {
10179 10186 ip1dbg(("ip_snmp_get_mib2_ip6_addr: failed to "
10180 10187 "allocate %u bytes\n",
10181 10188 (uint_t)mae6_size));
10182 10189 }
10183 10190 }
10184 10191 }
10185 10192 rw_exit(&ipst->ips_ill_g_lock);
10186 10193
10187 10194 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10188 10195 ip3dbg(("ip_snmp_get_mib2_ip6_addr: level %d, name %d, len %d\n",
10189 10196 (int)optp->level, (int)optp->name, (int)optp->len));
10190 10197 qreply(q, mpctl);
10191 10198 return (mp2ctl);
10192 10199 }
10193 10200
10194 10201 /* IPv4 multicast group membership. */
10195 10202 static mblk_t *
10196 10203 ip_snmp_get_mib2_ip_group_mem(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10197 10204 {
10198 10205 struct opthdr *optp;
10199 10206 mblk_t *mp2ctl;
10200 10207 ill_t *ill;
10201 10208 ipif_t *ipif;
10202 10209 ilm_t *ilm;
10203 10210 ip_member_t ipm;
10204 10211 mblk_t *mp_tail = NULL;
10205 10212 ill_walk_context_t ctx;
10206 10213 zoneid_t zoneid;
10207 10214
10208 10215 /*
10209 10216 * make a copy of the original message
10210 10217 */
10211 10218 mp2ctl = copymsg(mpctl);
10212 10219 zoneid = Q_TO_CONN(q)->conn_zoneid;
10213 10220
10214 10221 /* ipGroupMember table */
10215 10222 optp = (struct opthdr *)&mpctl->b_rptr[
10216 10223 sizeof (struct T_optmgmt_ack)];
10217 10224 optp->level = MIB2_IP;
10218 10225 optp->name = EXPER_IP_GROUP_MEMBERSHIP;
10219 10226
10220 10227 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10221 10228 ill = ILL_START_WALK_V4(&ctx, ipst);
10222 10229 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10223 10230 /* Make sure the ill isn't going away. */
10224 10231 if (!ill_check_and_refhold(ill))
10225 10232 continue;
10226 10233 rw_exit(&ipst->ips_ill_g_lock);
10227 10234 rw_enter(&ill->ill_mcast_lock, RW_READER);
10228 10235 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10229 10236 if (ilm->ilm_zoneid != zoneid &&
10230 10237 ilm->ilm_zoneid != ALL_ZONES)
10231 10238 continue;
10232 10239
10233 10240 /* Is there an ipif for ilm_ifaddr? */
10234 10241 for (ipif = ill->ill_ipif; ipif != NULL;
10235 10242 ipif = ipif->ipif_next) {
10236 10243 if (!IPIF_IS_CONDEMNED(ipif) &&
10237 10244 ipif->ipif_lcl_addr == ilm->ilm_ifaddr &&
10238 10245 ilm->ilm_ifaddr != INADDR_ANY)
10239 10246 break;
10240 10247 }
10241 10248 if (ipif != NULL) {
10242 10249 ipif_get_name(ipif,
10243 10250 ipm.ipGroupMemberIfIndex.o_bytes,
10244 10251 OCTET_LENGTH);
10245 10252 } else {
10246 10253 ill_get_name(ill,
10247 10254 ipm.ipGroupMemberIfIndex.o_bytes,
10248 10255 OCTET_LENGTH);
10249 10256 }
10250 10257 ipm.ipGroupMemberIfIndex.o_length =
10251 10258 mi_strlen(ipm.ipGroupMemberIfIndex.o_bytes);
10252 10259
10253 10260 ipm.ipGroupMemberAddress = ilm->ilm_addr;
10254 10261 ipm.ipGroupMemberRefCnt = ilm->ilm_refcnt;
10255 10262 ipm.ipGroupMemberFilterMode = ilm->ilm_fmode;
10256 10263 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10257 10264 (char *)&ipm, (int)sizeof (ipm))) {
10258 10265 ip1dbg(("ip_snmp_get_mib2_ip_group: "
10259 10266 "failed to allocate %u bytes\n",
10260 10267 (uint_t)sizeof (ipm)));
10261 10268 }
10262 10269 }
10263 10270 rw_exit(&ill->ill_mcast_lock);
10264 10271 ill_refrele(ill);
10265 10272 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10266 10273 }
10267 10274 rw_exit(&ipst->ips_ill_g_lock);
10268 10275 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10269 10276 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10270 10277 (int)optp->level, (int)optp->name, (int)optp->len));
10271 10278 qreply(q, mpctl);
10272 10279 return (mp2ctl);
10273 10280 }
10274 10281
10275 10282 /* IPv6 multicast group membership. */
10276 10283 static mblk_t *
10277 10284 ip_snmp_get_mib2_ip6_group_mem(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10278 10285 {
10279 10286 struct opthdr *optp;
10280 10287 mblk_t *mp2ctl;
10281 10288 ill_t *ill;
10282 10289 ilm_t *ilm;
10283 10290 ipv6_member_t ipm6;
10284 10291 mblk_t *mp_tail = NULL;
10285 10292 ill_walk_context_t ctx;
10286 10293 zoneid_t zoneid;
10287 10294
10288 10295 /*
10289 10296 * make a copy of the original message
10290 10297 */
10291 10298 mp2ctl = copymsg(mpctl);
10292 10299 zoneid = Q_TO_CONN(q)->conn_zoneid;
10293 10300
10294 10301 /* ip6GroupMember table */
10295 10302 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10296 10303 optp->level = MIB2_IP6;
10297 10304 optp->name = EXPER_IP6_GROUP_MEMBERSHIP;
10298 10305
10299 10306 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10300 10307 ill = ILL_START_WALK_V6(&ctx, ipst);
10301 10308 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10302 10309 /* Make sure the ill isn't going away. */
10303 10310 if (!ill_check_and_refhold(ill))
10304 10311 continue;
10305 10312 rw_exit(&ipst->ips_ill_g_lock);
10306 10313 /*
10307 10314 * Normally we don't have any members on under IPMP interfaces.
10308 10315 * We report them as a debugging aid.
10309 10316 */
10310 10317 rw_enter(&ill->ill_mcast_lock, RW_READER);
10311 10318 ipm6.ipv6GroupMemberIfIndex = ill->ill_phyint->phyint_ifindex;
10312 10319 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10313 10320 if (ilm->ilm_zoneid != zoneid &&
10314 10321 ilm->ilm_zoneid != ALL_ZONES)
10315 10322 continue; /* not this zone */
10316 10323 ipm6.ipv6GroupMemberAddress = ilm->ilm_v6addr;
10317 10324 ipm6.ipv6GroupMemberRefCnt = ilm->ilm_refcnt;
10318 10325 ipm6.ipv6GroupMemberFilterMode = ilm->ilm_fmode;
10319 10326 if (!snmp_append_data2(mpctl->b_cont,
10320 10327 &mp_tail,
10321 10328 (char *)&ipm6, (int)sizeof (ipm6))) {
10322 10329 ip1dbg(("ip_snmp_get_mib2_ip6_group: "
10323 10330 "failed to allocate %u bytes\n",
10324 10331 (uint_t)sizeof (ipm6)));
10325 10332 }
10326 10333 }
10327 10334 rw_exit(&ill->ill_mcast_lock);
10328 10335 ill_refrele(ill);
10329 10336 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10330 10337 }
10331 10338 rw_exit(&ipst->ips_ill_g_lock);
10332 10339
10333 10340 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10334 10341 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10335 10342 (int)optp->level, (int)optp->name, (int)optp->len));
10336 10343 qreply(q, mpctl);
10337 10344 return (mp2ctl);
10338 10345 }
10339 10346
10340 10347 /* IP multicast filtered sources */
10341 10348 static mblk_t *
10342 10349 ip_snmp_get_mib2_ip_group_src(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10343 10350 {
10344 10351 struct opthdr *optp;
10345 10352 mblk_t *mp2ctl;
10346 10353 ill_t *ill;
10347 10354 ipif_t *ipif;
10348 10355 ilm_t *ilm;
10349 10356 ip_grpsrc_t ips;
10350 10357 mblk_t *mp_tail = NULL;
10351 10358 ill_walk_context_t ctx;
10352 10359 zoneid_t zoneid;
10353 10360 int i;
10354 10361 slist_t *sl;
10355 10362
10356 10363 /*
10357 10364 * make a copy of the original message
10358 10365 */
10359 10366 mp2ctl = copymsg(mpctl);
10360 10367 zoneid = Q_TO_CONN(q)->conn_zoneid;
10361 10368
10362 10369 /* ipGroupSource table */
10363 10370 optp = (struct opthdr *)&mpctl->b_rptr[
10364 10371 sizeof (struct T_optmgmt_ack)];
10365 10372 optp->level = MIB2_IP;
10366 10373 optp->name = EXPER_IP_GROUP_SOURCES;
10367 10374
10368 10375 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10369 10376 ill = ILL_START_WALK_V4(&ctx, ipst);
10370 10377 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10371 10378 /* Make sure the ill isn't going away. */
10372 10379 if (!ill_check_and_refhold(ill))
10373 10380 continue;
10374 10381 rw_exit(&ipst->ips_ill_g_lock);
10375 10382 rw_enter(&ill->ill_mcast_lock, RW_READER);
10376 10383 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10377 10384 sl = ilm->ilm_filter;
10378 10385 if (ilm->ilm_zoneid != zoneid &&
10379 10386 ilm->ilm_zoneid != ALL_ZONES)
10380 10387 continue;
10381 10388 if (SLIST_IS_EMPTY(sl))
10382 10389 continue;
10383 10390
10384 10391 /* Is there an ipif for ilm_ifaddr? */
10385 10392 for (ipif = ill->ill_ipif; ipif != NULL;
10386 10393 ipif = ipif->ipif_next) {
10387 10394 if (!IPIF_IS_CONDEMNED(ipif) &&
10388 10395 ipif->ipif_lcl_addr == ilm->ilm_ifaddr &&
10389 10396 ilm->ilm_ifaddr != INADDR_ANY)
10390 10397 break;
10391 10398 }
10392 10399 if (ipif != NULL) {
10393 10400 ipif_get_name(ipif,
10394 10401 ips.ipGroupSourceIfIndex.o_bytes,
10395 10402 OCTET_LENGTH);
10396 10403 } else {
10397 10404 ill_get_name(ill,
10398 10405 ips.ipGroupSourceIfIndex.o_bytes,
10399 10406 OCTET_LENGTH);
10400 10407 }
10401 10408 ips.ipGroupSourceIfIndex.o_length =
10402 10409 mi_strlen(ips.ipGroupSourceIfIndex.o_bytes);
10403 10410
10404 10411 ips.ipGroupSourceGroup = ilm->ilm_addr;
10405 10412 for (i = 0; i < sl->sl_numsrc; i++) {
10406 10413 if (!IN6_IS_ADDR_V4MAPPED(&sl->sl_addr[i]))
10407 10414 continue;
10408 10415 IN6_V4MAPPED_TO_IPADDR(&sl->sl_addr[i],
10409 10416 ips.ipGroupSourceAddress);
10410 10417 if (snmp_append_data2(mpctl->b_cont, &mp_tail,
10411 10418 (char *)&ips, (int)sizeof (ips)) == 0) {
10412 10419 ip1dbg(("ip_snmp_get_mib2_ip_group_src:"
10413 10420 " failed to allocate %u bytes\n",
10414 10421 (uint_t)sizeof (ips)));
10415 10422 }
10416 10423 }
10417 10424 }
10418 10425 rw_exit(&ill->ill_mcast_lock);
10419 10426 ill_refrele(ill);
10420 10427 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10421 10428 }
10422 10429 rw_exit(&ipst->ips_ill_g_lock);
10423 10430 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10424 10431 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10425 10432 (int)optp->level, (int)optp->name, (int)optp->len));
10426 10433 qreply(q, mpctl);
10427 10434 return (mp2ctl);
10428 10435 }
10429 10436
10430 10437 /* IPv6 multicast filtered sources. */
10431 10438 static mblk_t *
10432 10439 ip_snmp_get_mib2_ip6_group_src(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10433 10440 {
10434 10441 struct opthdr *optp;
10435 10442 mblk_t *mp2ctl;
10436 10443 ill_t *ill;
10437 10444 ilm_t *ilm;
10438 10445 ipv6_grpsrc_t ips6;
10439 10446 mblk_t *mp_tail = NULL;
10440 10447 ill_walk_context_t ctx;
10441 10448 zoneid_t zoneid;
10442 10449 int i;
10443 10450 slist_t *sl;
10444 10451
10445 10452 /*
10446 10453 * make a copy of the original message
10447 10454 */
10448 10455 mp2ctl = copymsg(mpctl);
10449 10456 zoneid = Q_TO_CONN(q)->conn_zoneid;
10450 10457
10451 10458 /* ip6GroupMember table */
10452 10459 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10453 10460 optp->level = MIB2_IP6;
10454 10461 optp->name = EXPER_IP6_GROUP_SOURCES;
10455 10462
10456 10463 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10457 10464 ill = ILL_START_WALK_V6(&ctx, ipst);
10458 10465 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10459 10466 /* Make sure the ill isn't going away. */
10460 10467 if (!ill_check_and_refhold(ill))
10461 10468 continue;
10462 10469 rw_exit(&ipst->ips_ill_g_lock);
10463 10470 /*
10464 10471 * Normally we don't have any members on under IPMP interfaces.
10465 10472 * We report them as a debugging aid.
10466 10473 */
10467 10474 rw_enter(&ill->ill_mcast_lock, RW_READER);
10468 10475 ips6.ipv6GroupSourceIfIndex = ill->ill_phyint->phyint_ifindex;
10469 10476 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10470 10477 sl = ilm->ilm_filter;
10471 10478 if (ilm->ilm_zoneid != zoneid &&
10472 10479 ilm->ilm_zoneid != ALL_ZONES)
10473 10480 continue;
10474 10481 if (SLIST_IS_EMPTY(sl))
10475 10482 continue;
10476 10483 ips6.ipv6GroupSourceGroup = ilm->ilm_v6addr;
10477 10484 for (i = 0; i < sl->sl_numsrc; i++) {
10478 10485 ips6.ipv6GroupSourceAddress = sl->sl_addr[i];
10479 10486 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10480 10487 (char *)&ips6, (int)sizeof (ips6))) {
10481 10488 ip1dbg(("ip_snmp_get_mib2_ip6_"
10482 10489 "group_src: failed to allocate "
10483 10490 "%u bytes\n",
10484 10491 (uint_t)sizeof (ips6)));
10485 10492 }
10486 10493 }
10487 10494 }
10488 10495 rw_exit(&ill->ill_mcast_lock);
10489 10496 ill_refrele(ill);
10490 10497 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10491 10498 }
10492 10499 rw_exit(&ipst->ips_ill_g_lock);
10493 10500
10494 10501 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10495 10502 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10496 10503 (int)optp->level, (int)optp->name, (int)optp->len));
10497 10504 qreply(q, mpctl);
10498 10505 return (mp2ctl);
10499 10506 }
10500 10507
10501 10508 /* Multicast routing virtual interface table. */
10502 10509 static mblk_t *
10503 10510 ip_snmp_get_mib2_virt_multi(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10504 10511 {
10505 10512 struct opthdr *optp;
10506 10513 mblk_t *mp2ctl;
10507 10514
10508 10515 /*
10509 10516 * make a copy of the original message
10510 10517 */
10511 10518 mp2ctl = copymsg(mpctl);
10512 10519
10513 10520 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10514 10521 optp->level = EXPER_DVMRP;
10515 10522 optp->name = EXPER_DVMRP_VIF;
10516 10523 if (!ip_mroute_vif(mpctl->b_cont, ipst)) {
10517 10524 ip0dbg(("ip_mroute_vif: failed\n"));
10518 10525 }
10519 10526 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10520 10527 ip3dbg(("ip_snmp_get_mib2_virt_multi: level %d, name %d, len %d\n",
10521 10528 (int)optp->level, (int)optp->name, (int)optp->len));
10522 10529 qreply(q, mpctl);
10523 10530 return (mp2ctl);
10524 10531 }
10525 10532
10526 10533 /* Multicast routing table. */
10527 10534 static mblk_t *
10528 10535 ip_snmp_get_mib2_multi_rtable(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10529 10536 {
10530 10537 struct opthdr *optp;
10531 10538 mblk_t *mp2ctl;
10532 10539
10533 10540 /*
10534 10541 * make a copy of the original message
10535 10542 */
10536 10543 mp2ctl = copymsg(mpctl);
10537 10544
10538 10545 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10539 10546 optp->level = EXPER_DVMRP;
10540 10547 optp->name = EXPER_DVMRP_MRT;
10541 10548 if (!ip_mroute_mrt(mpctl->b_cont, ipst)) {
10542 10549 ip0dbg(("ip_mroute_mrt: failed\n"));
10543 10550 }
10544 10551 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10545 10552 ip3dbg(("ip_snmp_get_mib2_multi_rtable: level %d, name %d, len %d\n",
10546 10553 (int)optp->level, (int)optp->name, (int)optp->len));
10547 10554 qreply(q, mpctl);
10548 10555 return (mp2ctl);
10549 10556 }
10550 10557
10551 10558 /*
10552 10559 * Return ipRouteEntryTable, ipNetToMediaEntryTable, and ipRouteAttributeTable
10553 10560 * in one IRE walk.
10554 10561 */
10555 10562 static mblk_t *
10556 10563 ip_snmp_get_mib2_ip_route_media(queue_t *q, mblk_t *mpctl, int level,
10557 10564 ip_stack_t *ipst)
10558 10565 {
10559 10566 struct opthdr *optp;
10560 10567 mblk_t *mp2ctl; /* Returned */
10561 10568 mblk_t *mp3ctl; /* nettomedia */
10562 10569 mblk_t *mp4ctl; /* routeattrs */
10563 10570 iproutedata_t ird;
10564 10571 zoneid_t zoneid;
10565 10572
10566 10573 /*
10567 10574 * make copies of the original message
10568 10575 * - mp2ctl is returned unchanged to the caller for its use
10569 10576 * - mpctl is sent upstream as ipRouteEntryTable
10570 10577 * - mp3ctl is sent upstream as ipNetToMediaEntryTable
10571 10578 * - mp4ctl is sent upstream as ipRouteAttributeTable
10572 10579 */
10573 10580 mp2ctl = copymsg(mpctl);
10574 10581 mp3ctl = copymsg(mpctl);
10575 10582 mp4ctl = copymsg(mpctl);
10576 10583 if (mp3ctl == NULL || mp4ctl == NULL) {
10577 10584 freemsg(mp4ctl);
10578 10585 freemsg(mp3ctl);
10579 10586 freemsg(mp2ctl);
10580 10587 freemsg(mpctl);
10581 10588 return (NULL);
10582 10589 }
10583 10590
10584 10591 bzero(&ird, sizeof (ird));
10585 10592
10586 10593 ird.ird_route.lp_head = mpctl->b_cont;
10587 10594 ird.ird_netmedia.lp_head = mp3ctl->b_cont;
10588 10595 ird.ird_attrs.lp_head = mp4ctl->b_cont;
10589 10596 /*
10590 10597 * If the level has been set the special EXPER_IP_AND_ALL_IRES value,
10591 10598 * then also include ire_testhidden IREs and IRE_IF_CLONE. This is
10592 10599 * intended a temporary solution until a proper MIB API is provided
10593 10600 * that provides complete filtering/caller-opt-in.
10594 10601 */
10595 10602 if (level == EXPER_IP_AND_ALL_IRES)
10596 10603 ird.ird_flags |= IRD_REPORT_ALL;
10597 10604
10598 10605 zoneid = Q_TO_CONN(q)->conn_zoneid;
10599 10606 ire_walk_v4(ip_snmp_get2_v4, &ird, zoneid, ipst);
10600 10607
10601 10608 /* ipRouteEntryTable in mpctl */
10602 10609 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10603 10610 optp->level = MIB2_IP;
10604 10611 optp->name = MIB2_IP_ROUTE;
10605 10612 optp->len = msgdsize(ird.ird_route.lp_head);
10606 10613 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10607 10614 (int)optp->level, (int)optp->name, (int)optp->len));
10608 10615 qreply(q, mpctl);
10609 10616
10610 10617 /* ipNetToMediaEntryTable in mp3ctl */
10611 10618 ncec_walk(NULL, ip_snmp_get2_v4_media, &ird, ipst);
10612 10619
10613 10620 optp = (struct opthdr *)&mp3ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10614 10621 optp->level = MIB2_IP;
10615 10622 optp->name = MIB2_IP_MEDIA;
10616 10623 optp->len = msgdsize(ird.ird_netmedia.lp_head);
10617 10624 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10618 10625 (int)optp->level, (int)optp->name, (int)optp->len));
10619 10626 qreply(q, mp3ctl);
10620 10627
10621 10628 /* ipRouteAttributeTable in mp4ctl */
10622 10629 optp = (struct opthdr *)&mp4ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10623 10630 optp->level = MIB2_IP;
10624 10631 optp->name = EXPER_IP_RTATTR;
10625 10632 optp->len = msgdsize(ird.ird_attrs.lp_head);
10626 10633 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10627 10634 (int)optp->level, (int)optp->name, (int)optp->len));
10628 10635 if (optp->len == 0)
10629 10636 freemsg(mp4ctl);
10630 10637 else
10631 10638 qreply(q, mp4ctl);
10632 10639
10633 10640 return (mp2ctl);
10634 10641 }
10635 10642
10636 10643 /*
10637 10644 * Return ipv6RouteEntryTable and ipv6RouteAttributeTable in one IRE walk, and
10638 10645 * ipv6NetToMediaEntryTable in an NDP walk.
10639 10646 */
10640 10647 static mblk_t *
10641 10648 ip_snmp_get_mib2_ip6_route_media(queue_t *q, mblk_t *mpctl, int level,
10642 10649 ip_stack_t *ipst)
10643 10650 {
10644 10651 struct opthdr *optp;
10645 10652 mblk_t *mp2ctl; /* Returned */
10646 10653 mblk_t *mp3ctl; /* nettomedia */
10647 10654 mblk_t *mp4ctl; /* routeattrs */
10648 10655 iproutedata_t ird;
10649 10656 zoneid_t zoneid;
10650 10657
10651 10658 /*
10652 10659 * make copies of the original message
10653 10660 * - mp2ctl is returned unchanged to the caller for its use
10654 10661 * - mpctl is sent upstream as ipv6RouteEntryTable
10655 10662 * - mp3ctl is sent upstream as ipv6NetToMediaEntryTable
10656 10663 * - mp4ctl is sent upstream as ipv6RouteAttributeTable
10657 10664 */
10658 10665 mp2ctl = copymsg(mpctl);
10659 10666 mp3ctl = copymsg(mpctl);
10660 10667 mp4ctl = copymsg(mpctl);
10661 10668 if (mp3ctl == NULL || mp4ctl == NULL) {
10662 10669 freemsg(mp4ctl);
10663 10670 freemsg(mp3ctl);
10664 10671 freemsg(mp2ctl);
10665 10672 freemsg(mpctl);
10666 10673 return (NULL);
10667 10674 }
10668 10675
10669 10676 bzero(&ird, sizeof (ird));
10670 10677
10671 10678 ird.ird_route.lp_head = mpctl->b_cont;
10672 10679 ird.ird_netmedia.lp_head = mp3ctl->b_cont;
10673 10680 ird.ird_attrs.lp_head = mp4ctl->b_cont;
10674 10681 /*
10675 10682 * If the level has been set the special EXPER_IP_AND_ALL_IRES value,
10676 10683 * then also include ire_testhidden IREs and IRE_IF_CLONE. This is
10677 10684 * intended a temporary solution until a proper MIB API is provided
10678 10685 * that provides complete filtering/caller-opt-in.
10679 10686 */
10680 10687 if (level == EXPER_IP_AND_ALL_IRES)
10681 10688 ird.ird_flags |= IRD_REPORT_ALL;
10682 10689
10683 10690 zoneid = Q_TO_CONN(q)->conn_zoneid;
10684 10691 ire_walk_v6(ip_snmp_get2_v6_route, &ird, zoneid, ipst);
10685 10692
10686 10693 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10687 10694 optp->level = MIB2_IP6;
10688 10695 optp->name = MIB2_IP6_ROUTE;
10689 10696 optp->len = msgdsize(ird.ird_route.lp_head);
10690 10697 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10691 10698 (int)optp->level, (int)optp->name, (int)optp->len));
10692 10699 qreply(q, mpctl);
10693 10700
10694 10701 /* ipv6NetToMediaEntryTable in mp3ctl */
10695 10702 ncec_walk(NULL, ip_snmp_get2_v6_media, &ird, ipst);
10696 10703
10697 10704 optp = (struct opthdr *)&mp3ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10698 10705 optp->level = MIB2_IP6;
10699 10706 optp->name = MIB2_IP6_MEDIA;
10700 10707 optp->len = msgdsize(ird.ird_netmedia.lp_head);
10701 10708 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10702 10709 (int)optp->level, (int)optp->name, (int)optp->len));
10703 10710 qreply(q, mp3ctl);
10704 10711
10705 10712 /* ipv6RouteAttributeTable in mp4ctl */
10706 10713 optp = (struct opthdr *)&mp4ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10707 10714 optp->level = MIB2_IP6;
10708 10715 optp->name = EXPER_IP_RTATTR;
10709 10716 optp->len = msgdsize(ird.ird_attrs.lp_head);
10710 10717 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10711 10718 (int)optp->level, (int)optp->name, (int)optp->len));
10712 10719 if (optp->len == 0)
10713 10720 freemsg(mp4ctl);
10714 10721 else
10715 10722 qreply(q, mp4ctl);
10716 10723
10717 10724 return (mp2ctl);
10718 10725 }
10719 10726
10720 10727 /*
10721 10728 * IPv6 mib: One per ill
10722 10729 */
10723 10730 static mblk_t *
10724 10731 ip_snmp_get_mib2_ip6(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
10725 10732 boolean_t legacy_req)
10726 10733 {
10727 10734 struct opthdr *optp;
10728 10735 mblk_t *mp2ctl;
10729 10736 ill_t *ill;
10730 10737 ill_walk_context_t ctx;
10731 10738 mblk_t *mp_tail = NULL;
10732 10739 mib2_ipv6AddrEntry_t mae6;
10733 10740 mib2_ipIfStatsEntry_t *ise;
10734 10741 size_t ise_size, iae_size;
10735 10742
10736 10743 /*
10737 10744 * Make a copy of the original message
10738 10745 */
10739 10746 mp2ctl = copymsg(mpctl);
10740 10747
10741 10748 /* fixed length IPv6 structure ... */
10742 10749
10743 10750 if (legacy_req) {
10744 10751 ise_size = LEGACY_MIB_SIZE(&ipst->ips_ip6_mib,
10745 10752 mib2_ipIfStatsEntry_t);
10746 10753 iae_size = LEGACY_MIB_SIZE(&mae6, mib2_ipv6AddrEntry_t);
10747 10754 } else {
10748 10755 ise_size = sizeof (mib2_ipIfStatsEntry_t);
10749 10756 iae_size = sizeof (mib2_ipv6AddrEntry_t);
10750 10757 }
10751 10758
10752 10759 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10753 10760 optp->level = MIB2_IP6;
10754 10761 optp->name = 0;
10755 10762 /* Include "unknown interface" ip6_mib */
10756 10763 ipst->ips_ip6_mib.ipIfStatsIPVersion = MIB2_INETADDRESSTYPE_ipv6;
10757 10764 ipst->ips_ip6_mib.ipIfStatsIfIndex =
10758 10765 MIB2_UNKNOWN_INTERFACE; /* Flag to netstat */
10759 10766 SET_MIB(ipst->ips_ip6_mib.ipIfStatsForwarding,
10760 10767 ipst->ips_ipv6_forwarding ? 1 : 2);
10761 10768 SET_MIB(ipst->ips_ip6_mib.ipIfStatsDefaultHopLimit,
10762 10769 ipst->ips_ipv6_def_hops);
10763 10770 SET_MIB(ipst->ips_ip6_mib.ipIfStatsEntrySize,
10764 10771 sizeof (mib2_ipIfStatsEntry_t));
10765 10772 SET_MIB(ipst->ips_ip6_mib.ipIfStatsAddrEntrySize,
10766 10773 sizeof (mib2_ipv6AddrEntry_t));
10767 10774 SET_MIB(ipst->ips_ip6_mib.ipIfStatsRouteEntrySize,
10768 10775 sizeof (mib2_ipv6RouteEntry_t));
10769 10776 SET_MIB(ipst->ips_ip6_mib.ipIfStatsNetToMediaEntrySize,
10770 10777 sizeof (mib2_ipv6NetToMediaEntry_t));
10771 10778 SET_MIB(ipst->ips_ip6_mib.ipIfStatsMemberEntrySize,
10772 10779 sizeof (ipv6_member_t));
10773 10780 SET_MIB(ipst->ips_ip6_mib.ipIfStatsGroupSourceEntrySize,
10774 10781 sizeof (ipv6_grpsrc_t));
10775 10782
10776 10783 /*
10777 10784 * Synchronize 64- and 32-bit counters
10778 10785 */
10779 10786 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInReceives,
10780 10787 ipIfStatsHCInReceives);
10781 10788 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInDelivers,
10782 10789 ipIfStatsHCInDelivers);
10783 10790 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutRequests,
10784 10791 ipIfStatsHCOutRequests);
10785 10792 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutForwDatagrams,
10786 10793 ipIfStatsHCOutForwDatagrams);
10787 10794 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutMcastPkts,
10788 10795 ipIfStatsHCOutMcastPkts);
10789 10796 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInMcastPkts,
10790 10797 ipIfStatsHCInMcastPkts);
10791 10798
10792 10799 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10793 10800 (char *)&ipst->ips_ip6_mib, (int)ise_size)) {
10794 10801 ip1dbg(("ip_snmp_get_mib2_ip6: failed to allocate %u bytes\n",
10795 10802 (uint_t)ise_size));
10796 10803 } else if (legacy_req) {
10797 10804 /* Adjust the EntrySize fields for legacy requests. */
10798 10805 ise =
10799 10806 (mib2_ipIfStatsEntry_t *)(mp_tail->b_wptr - (int)ise_size);
10800 10807 SET_MIB(ise->ipIfStatsEntrySize, ise_size);
10801 10808 SET_MIB(ise->ipIfStatsAddrEntrySize, iae_size);
10802 10809 }
10803 10810
10804 10811 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10805 10812 ill = ILL_START_WALK_V6(&ctx, ipst);
10806 10813 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10807 10814 ill->ill_ip_mib->ipIfStatsIfIndex =
10808 10815 ill->ill_phyint->phyint_ifindex;
10809 10816 SET_MIB(ill->ill_ip_mib->ipIfStatsForwarding,
10810 10817 ipst->ips_ipv6_forwarding ? 1 : 2);
10811 10818 SET_MIB(ill->ill_ip_mib->ipIfStatsDefaultHopLimit,
10812 10819 ill->ill_max_hops);
10813 10820
10814 10821 /*
10815 10822 * Synchronize 64- and 32-bit counters
10816 10823 */
10817 10824 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInReceives,
10818 10825 ipIfStatsHCInReceives);
10819 10826 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInDelivers,
10820 10827 ipIfStatsHCInDelivers);
10821 10828 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutRequests,
10822 10829 ipIfStatsHCOutRequests);
10823 10830 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutForwDatagrams,
10824 10831 ipIfStatsHCOutForwDatagrams);
10825 10832 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutMcastPkts,
10826 10833 ipIfStatsHCOutMcastPkts);
10827 10834 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInMcastPkts,
10828 10835 ipIfStatsHCInMcastPkts);
10829 10836
10830 10837 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10831 10838 (char *)ill->ill_ip_mib, (int)ise_size)) {
10832 10839 ip1dbg(("ip_snmp_get_mib2_ip6: failed to allocate "
10833 10840 "%u bytes\n", (uint_t)ise_size));
10834 10841 } else if (legacy_req) {
10835 10842 /* Adjust the EntrySize fields for legacy requests. */
10836 10843 ise = (mib2_ipIfStatsEntry_t *)(mp_tail->b_wptr -
10837 10844 (int)ise_size);
10838 10845 SET_MIB(ise->ipIfStatsEntrySize, ise_size);
10839 10846 SET_MIB(ise->ipIfStatsAddrEntrySize, iae_size);
10840 10847 }
10841 10848 }
10842 10849 rw_exit(&ipst->ips_ill_g_lock);
10843 10850
10844 10851 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10845 10852 ip3dbg(("ip_snmp_get_mib2_ip6: level %d, name %d, len %d\n",
10846 10853 (int)optp->level, (int)optp->name, (int)optp->len));
10847 10854 qreply(q, mpctl);
10848 10855 return (mp2ctl);
10849 10856 }
10850 10857
10851 10858 /*
10852 10859 * ICMPv6 mib: One per ill
10853 10860 */
10854 10861 static mblk_t *
10855 10862 ip_snmp_get_mib2_icmp6(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10856 10863 {
10857 10864 struct opthdr *optp;
10858 10865 mblk_t *mp2ctl;
10859 10866 ill_t *ill;
10860 10867 ill_walk_context_t ctx;
10861 10868 mblk_t *mp_tail = NULL;
10862 10869 /*
10863 10870 * Make a copy of the original message
10864 10871 */
10865 10872 mp2ctl = copymsg(mpctl);
10866 10873
10867 10874 /* fixed length ICMPv6 structure ... */
10868 10875
10869 10876 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10870 10877 optp->level = MIB2_ICMP6;
10871 10878 optp->name = 0;
10872 10879 /* Include "unknown interface" icmp6_mib */
10873 10880 ipst->ips_icmp6_mib.ipv6IfIcmpIfIndex =
10874 10881 MIB2_UNKNOWN_INTERFACE; /* netstat flag */
10875 10882 ipst->ips_icmp6_mib.ipv6IfIcmpEntrySize =
10876 10883 sizeof (mib2_ipv6IfIcmpEntry_t);
10877 10884 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10878 10885 (char *)&ipst->ips_icmp6_mib,
10879 10886 (int)sizeof (ipst->ips_icmp6_mib))) {
10880 10887 ip1dbg(("ip_snmp_get_mib2_icmp6: failed to allocate %u bytes\n",
10881 10888 (uint_t)sizeof (ipst->ips_icmp6_mib)));
10882 10889 }
10883 10890
10884 10891 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10885 10892 ill = ILL_START_WALK_V6(&ctx, ipst);
10886 10893 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10887 10894 ill->ill_icmp6_mib->ipv6IfIcmpIfIndex =
10888 10895 ill->ill_phyint->phyint_ifindex;
10889 10896 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10890 10897 (char *)ill->ill_icmp6_mib,
10891 10898 (int)sizeof (*ill->ill_icmp6_mib))) {
10892 10899 ip1dbg(("ip_snmp_get_mib2_icmp6: failed to allocate "
10893 10900 "%u bytes\n",
10894 10901 (uint_t)sizeof (*ill->ill_icmp6_mib)));
10895 10902 }
10896 10903 }
10897 10904 rw_exit(&ipst->ips_ill_g_lock);
10898 10905
10899 10906 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10900 10907 ip3dbg(("ip_snmp_get_mib2_icmp6: level %d, name %d, len %d\n",
10901 10908 (int)optp->level, (int)optp->name, (int)optp->len));
10902 10909 qreply(q, mpctl);
10903 10910 return (mp2ctl);
10904 10911 }
10905 10912
10906 10913 /*
10907 10914 * ire_walk routine to create both ipRouteEntryTable and
10908 10915 * ipRouteAttributeTable in one IRE walk
10909 10916 */
10910 10917 static void
10911 10918 ip_snmp_get2_v4(ire_t *ire, iproutedata_t *ird)
10912 10919 {
10913 10920 ill_t *ill;
10914 10921 mib2_ipRouteEntry_t *re;
10915 10922 mib2_ipAttributeEntry_t iaes;
10916 10923 tsol_ire_gw_secattr_t *attrp;
10917 10924 tsol_gc_t *gc = NULL;
10918 10925 tsol_gcgrp_t *gcgrp = NULL;
10919 10926 ip_stack_t *ipst = ire->ire_ipst;
10920 10927
10921 10928 ASSERT(ire->ire_ipversion == IPV4_VERSION);
10922 10929
10923 10930 if (!(ird->ird_flags & IRD_REPORT_ALL)) {
10924 10931 if (ire->ire_testhidden)
10925 10932 return;
10926 10933 if (ire->ire_type & IRE_IF_CLONE)
10927 10934 return;
10928 10935 }
10929 10936
10930 10937 if ((re = kmem_zalloc(sizeof (*re), KM_NOSLEEP)) == NULL)
10931 10938 return;
10932 10939
10933 10940 if ((attrp = ire->ire_gw_secattr) != NULL) {
10934 10941 mutex_enter(&attrp->igsa_lock);
10935 10942 if ((gc = attrp->igsa_gc) != NULL) {
10936 10943 gcgrp = gc->gc_grp;
10937 10944 ASSERT(gcgrp != NULL);
10938 10945 rw_enter(&gcgrp->gcgrp_rwlock, RW_READER);
10939 10946 }
10940 10947 mutex_exit(&attrp->igsa_lock);
10941 10948 }
10942 10949 /*
10943 10950 * Return all IRE types for route table... let caller pick and choose
10944 10951 */
10945 10952 re->ipRouteDest = ire->ire_addr;
10946 10953 ill = ire->ire_ill;
10947 10954 re->ipRouteIfIndex.o_length = 0;
10948 10955 if (ill != NULL) {
10949 10956 ill_get_name(ill, re->ipRouteIfIndex.o_bytes, OCTET_LENGTH);
10950 10957 re->ipRouteIfIndex.o_length =
10951 10958 mi_strlen(re->ipRouteIfIndex.o_bytes);
10952 10959 }
10953 10960 re->ipRouteMetric1 = -1;
10954 10961 re->ipRouteMetric2 = -1;
10955 10962 re->ipRouteMetric3 = -1;
10956 10963 re->ipRouteMetric4 = -1;
10957 10964
10958 10965 re->ipRouteNextHop = ire->ire_gateway_addr;
10959 10966 /* indirect(4), direct(3), or invalid(2) */
10960 10967 if (ire->ire_flags & (RTF_REJECT | RTF_BLACKHOLE))
10961 10968 re->ipRouteType = 2;
10962 10969 else if (ire->ire_type & IRE_ONLINK)
10963 10970 re->ipRouteType = 3;
10964 10971 else
10965 10972 re->ipRouteType = 4;
10966 10973
10967 10974 re->ipRouteProto = -1;
10968 10975 re->ipRouteAge = gethrestime_sec() - ire->ire_create_time;
10969 10976 re->ipRouteMask = ire->ire_mask;
10970 10977 re->ipRouteMetric5 = -1;
10971 10978 re->ipRouteInfo.re_max_frag = ire->ire_metrics.iulp_mtu;
10972 10979 if (ire->ire_ill != NULL && re->ipRouteInfo.re_max_frag == 0)
10973 10980 re->ipRouteInfo.re_max_frag = ire->ire_ill->ill_mtu;
10974 10981
10975 10982 re->ipRouteInfo.re_frag_flag = 0;
10976 10983 re->ipRouteInfo.re_rtt = 0;
10977 10984 re->ipRouteInfo.re_src_addr = 0;
10978 10985 re->ipRouteInfo.re_ref = ire->ire_refcnt;
10979 10986 re->ipRouteInfo.re_obpkt = ire->ire_ob_pkt_count;
10980 10987 re->ipRouteInfo.re_ibpkt = ire->ire_ib_pkt_count;
10981 10988 re->ipRouteInfo.re_flags = ire->ire_flags;
10982 10989
10983 10990 /* Add the IRE_IF_CLONE's counters to their parent IRE_INTERFACE */
10984 10991 if (ire->ire_type & IRE_INTERFACE) {
10985 10992 ire_t *child;
10986 10993
10987 10994 rw_enter(&ipst->ips_ire_dep_lock, RW_READER);
10988 10995 child = ire->ire_dep_children;
10989 10996 while (child != NULL) {
10990 10997 re->ipRouteInfo.re_obpkt += child->ire_ob_pkt_count;
10991 10998 re->ipRouteInfo.re_ibpkt += child->ire_ib_pkt_count;
10992 10999 child = child->ire_dep_sib_next;
10993 11000 }
10994 11001 rw_exit(&ipst->ips_ire_dep_lock);
10995 11002 }
10996 11003
10997 11004 if (ire->ire_flags & RTF_DYNAMIC) {
10998 11005 re->ipRouteInfo.re_ire_type = IRE_HOST_REDIRECT;
10999 11006 } else {
11000 11007 re->ipRouteInfo.re_ire_type = ire->ire_type;
11001 11008 }
11002 11009
11003 11010 if (!snmp_append_data2(ird->ird_route.lp_head, &ird->ird_route.lp_tail,
11004 11011 (char *)re, (int)sizeof (*re))) {
11005 11012 ip1dbg(("ip_snmp_get2_v4: failed to allocate %u bytes\n",
11006 11013 (uint_t)sizeof (*re)));
11007 11014 }
11008 11015
11009 11016 if (gc != NULL) {
11010 11017 iaes.iae_routeidx = ird->ird_idx;
11011 11018 iaes.iae_doi = gc->gc_db->gcdb_doi;
11012 11019 iaes.iae_slrange = gc->gc_db->gcdb_slrange;
11013 11020
11014 11021 if (!snmp_append_data2(ird->ird_attrs.lp_head,
11015 11022 &ird->ird_attrs.lp_tail, (char *)&iaes, sizeof (iaes))) {
11016 11023 ip1dbg(("ip_snmp_get2_v4: failed to allocate %u "
11017 11024 "bytes\n", (uint_t)sizeof (iaes)));
11018 11025 }
11019 11026 }
11020 11027
11021 11028 /* bump route index for next pass */
11022 11029 ird->ird_idx++;
11023 11030
11024 11031 kmem_free(re, sizeof (*re));
11025 11032 if (gcgrp != NULL)
11026 11033 rw_exit(&gcgrp->gcgrp_rwlock);
11027 11034 }
11028 11035
11029 11036 /*
11030 11037 * ire_walk routine to create ipv6RouteEntryTable and ipRouteEntryTable.
11031 11038 */
11032 11039 static void
11033 11040 ip_snmp_get2_v6_route(ire_t *ire, iproutedata_t *ird)
11034 11041 {
11035 11042 ill_t *ill;
11036 11043 mib2_ipv6RouteEntry_t *re;
11037 11044 mib2_ipAttributeEntry_t iaes;
11038 11045 tsol_ire_gw_secattr_t *attrp;
11039 11046 tsol_gc_t *gc = NULL;
11040 11047 tsol_gcgrp_t *gcgrp = NULL;
11041 11048 ip_stack_t *ipst = ire->ire_ipst;
11042 11049
11043 11050 ASSERT(ire->ire_ipversion == IPV6_VERSION);
11044 11051
11045 11052 if (!(ird->ird_flags & IRD_REPORT_ALL)) {
11046 11053 if (ire->ire_testhidden)
11047 11054 return;
11048 11055 if (ire->ire_type & IRE_IF_CLONE)
11049 11056 return;
11050 11057 }
11051 11058
11052 11059 if ((re = kmem_zalloc(sizeof (*re), KM_NOSLEEP)) == NULL)
11053 11060 return;
11054 11061
11055 11062 if ((attrp = ire->ire_gw_secattr) != NULL) {
11056 11063 mutex_enter(&attrp->igsa_lock);
11057 11064 if ((gc = attrp->igsa_gc) != NULL) {
11058 11065 gcgrp = gc->gc_grp;
11059 11066 ASSERT(gcgrp != NULL);
11060 11067 rw_enter(&gcgrp->gcgrp_rwlock, RW_READER);
11061 11068 }
11062 11069 mutex_exit(&attrp->igsa_lock);
11063 11070 }
11064 11071 /*
11065 11072 * Return all IRE types for route table... let caller pick and choose
11066 11073 */
11067 11074 re->ipv6RouteDest = ire->ire_addr_v6;
11068 11075 re->ipv6RoutePfxLength = ip_mask_to_plen_v6(&ire->ire_mask_v6);
11069 11076 re->ipv6RouteIndex = 0; /* Unique when multiple with same dest/plen */
11070 11077 re->ipv6RouteIfIndex.o_length = 0;
11071 11078 ill = ire->ire_ill;
11072 11079 if (ill != NULL) {
11073 11080 ill_get_name(ill, re->ipv6RouteIfIndex.o_bytes, OCTET_LENGTH);
11074 11081 re->ipv6RouteIfIndex.o_length =
11075 11082 mi_strlen(re->ipv6RouteIfIndex.o_bytes);
11076 11083 }
11077 11084
11078 11085 ASSERT(!(ire->ire_type & IRE_BROADCAST));
11079 11086
11080 11087 mutex_enter(&ire->ire_lock);
11081 11088 re->ipv6RouteNextHop = ire->ire_gateway_addr_v6;
11082 11089 mutex_exit(&ire->ire_lock);
11083 11090
11084 11091 /* remote(4), local(3), or discard(2) */
11085 11092 if (ire->ire_flags & (RTF_REJECT | RTF_BLACKHOLE))
11086 11093 re->ipv6RouteType = 2;
11087 11094 else if (ire->ire_type & IRE_ONLINK)
11088 11095 re->ipv6RouteType = 3;
11089 11096 else
11090 11097 re->ipv6RouteType = 4;
11091 11098
11092 11099 re->ipv6RouteProtocol = -1;
11093 11100 re->ipv6RoutePolicy = 0;
11094 11101 re->ipv6RouteAge = gethrestime_sec() - ire->ire_create_time;
11095 11102 re->ipv6RouteNextHopRDI = 0;
11096 11103 re->ipv6RouteWeight = 0;
11097 11104 re->ipv6RouteMetric = 0;
11098 11105 re->ipv6RouteInfo.re_max_frag = ire->ire_metrics.iulp_mtu;
11099 11106 if (ire->ire_ill != NULL && re->ipv6RouteInfo.re_max_frag == 0)
11100 11107 re->ipv6RouteInfo.re_max_frag = ire->ire_ill->ill_mtu;
11101 11108
11102 11109 re->ipv6RouteInfo.re_frag_flag = 0;
11103 11110 re->ipv6RouteInfo.re_rtt = 0;
11104 11111 re->ipv6RouteInfo.re_src_addr = ipv6_all_zeros;
11105 11112 re->ipv6RouteInfo.re_obpkt = ire->ire_ob_pkt_count;
11106 11113 re->ipv6RouteInfo.re_ibpkt = ire->ire_ib_pkt_count;
11107 11114 re->ipv6RouteInfo.re_ref = ire->ire_refcnt;
11108 11115 re->ipv6RouteInfo.re_flags = ire->ire_flags;
11109 11116
11110 11117 /* Add the IRE_IF_CLONE's counters to their parent IRE_INTERFACE */
11111 11118 if (ire->ire_type & IRE_INTERFACE) {
11112 11119 ire_t *child;
11113 11120
11114 11121 rw_enter(&ipst->ips_ire_dep_lock, RW_READER);
11115 11122 child = ire->ire_dep_children;
11116 11123 while (child != NULL) {
11117 11124 re->ipv6RouteInfo.re_obpkt += child->ire_ob_pkt_count;
11118 11125 re->ipv6RouteInfo.re_ibpkt += child->ire_ib_pkt_count;
11119 11126 child = child->ire_dep_sib_next;
11120 11127 }
11121 11128 rw_exit(&ipst->ips_ire_dep_lock);
11122 11129 }
11123 11130 if (ire->ire_flags & RTF_DYNAMIC) {
11124 11131 re->ipv6RouteInfo.re_ire_type = IRE_HOST_REDIRECT;
11125 11132 } else {
11126 11133 re->ipv6RouteInfo.re_ire_type = ire->ire_type;
11127 11134 }
11128 11135
11129 11136 if (!snmp_append_data2(ird->ird_route.lp_head, &ird->ird_route.lp_tail,
11130 11137 (char *)re, (int)sizeof (*re))) {
11131 11138 ip1dbg(("ip_snmp_get2_v6: failed to allocate %u bytes\n",
11132 11139 (uint_t)sizeof (*re)));
11133 11140 }
11134 11141
11135 11142 if (gc != NULL) {
11136 11143 iaes.iae_routeidx = ird->ird_idx;
11137 11144 iaes.iae_doi = gc->gc_db->gcdb_doi;
11138 11145 iaes.iae_slrange = gc->gc_db->gcdb_slrange;
11139 11146
11140 11147 if (!snmp_append_data2(ird->ird_attrs.lp_head,
11141 11148 &ird->ird_attrs.lp_tail, (char *)&iaes, sizeof (iaes))) {
11142 11149 ip1dbg(("ip_snmp_get2_v6: failed to allocate %u "
11143 11150 "bytes\n", (uint_t)sizeof (iaes)));
11144 11151 }
11145 11152 }
11146 11153
11147 11154 /* bump route index for next pass */
11148 11155 ird->ird_idx++;
11149 11156
11150 11157 kmem_free(re, sizeof (*re));
11151 11158 if (gcgrp != NULL)
11152 11159 rw_exit(&gcgrp->gcgrp_rwlock);
11153 11160 }
11154 11161
11155 11162 /*
11156 11163 * ncec_walk routine to create ipv6NetToMediaEntryTable
11157 11164 */
11158 11165 static void
11159 11166 ip_snmp_get2_v6_media(ncec_t *ncec, void *ptr)
11160 11167 {
11161 11168 iproutedata_t *ird = ptr;
11162 11169 ill_t *ill;
11163 11170 mib2_ipv6NetToMediaEntry_t ntme;
11164 11171
11165 11172 ill = ncec->ncec_ill;
11166 11173 /* skip arpce entries, and loopback ncec entries */
11167 11174 if (ill->ill_isv6 == B_FALSE || ill->ill_net_type == IRE_LOOPBACK)
11168 11175 return;
11169 11176 /*
11170 11177 * Neighbor cache entry attached to IRE with on-link
11171 11178 * destination.
11172 11179 * We report all IPMP groups on ncec_ill which is normally the upper.
11173 11180 */
11174 11181 ntme.ipv6NetToMediaIfIndex = ill->ill_phyint->phyint_ifindex;
11175 11182 ntme.ipv6NetToMediaNetAddress = ncec->ncec_addr;
11176 11183 ntme.ipv6NetToMediaPhysAddress.o_length = ill->ill_phys_addr_length;
11177 11184 if (ncec->ncec_lladdr != NULL) {
11178 11185 bcopy(ncec->ncec_lladdr, ntme.ipv6NetToMediaPhysAddress.o_bytes,
11179 11186 ntme.ipv6NetToMediaPhysAddress.o_length);
11180 11187 }
11181 11188 /*
11182 11189 * Note: Returns ND_* states. Should be:
11183 11190 * reachable(1), stale(2), delay(3), probe(4),
11184 11191 * invalid(5), unknown(6)
11185 11192 */
11186 11193 ntme.ipv6NetToMediaState = ncec->ncec_state;
11187 11194 ntme.ipv6NetToMediaLastUpdated = 0;
11188 11195
11189 11196 /* other(1), dynamic(2), static(3), local(4) */
11190 11197 if (NCE_MYADDR(ncec)) {
11191 11198 ntme.ipv6NetToMediaType = 4;
11192 11199 } else if (ncec->ncec_flags & NCE_F_PUBLISH) {
11193 11200 ntme.ipv6NetToMediaType = 1; /* proxy */
11194 11201 } else if (ncec->ncec_flags & NCE_F_STATIC) {
11195 11202 ntme.ipv6NetToMediaType = 3;
11196 11203 } else if (ncec->ncec_flags & (NCE_F_MCAST|NCE_F_BCAST)) {
11197 11204 ntme.ipv6NetToMediaType = 1;
11198 11205 } else {
11199 11206 ntme.ipv6NetToMediaType = 2;
11200 11207 }
11201 11208
11202 11209 if (!snmp_append_data2(ird->ird_netmedia.lp_head,
11203 11210 &ird->ird_netmedia.lp_tail, (char *)&ntme, sizeof (ntme))) {
11204 11211 ip1dbg(("ip_snmp_get2_v6_media: failed to allocate %u bytes\n",
11205 11212 (uint_t)sizeof (ntme)));
11206 11213 }
11207 11214 }
11208 11215
11209 11216 int
11210 11217 nce2ace(ncec_t *ncec)
11211 11218 {
11212 11219 int flags = 0;
11213 11220
11214 11221 if (NCE_ISREACHABLE(ncec))
11215 11222 flags |= ACE_F_RESOLVED;
11216 11223 if (ncec->ncec_flags & NCE_F_AUTHORITY)
11217 11224 flags |= ACE_F_AUTHORITY;
11218 11225 if (ncec->ncec_flags & NCE_F_PUBLISH)
11219 11226 flags |= ACE_F_PUBLISH;
11220 11227 if ((ncec->ncec_flags & NCE_F_NONUD) != 0)
11221 11228 flags |= ACE_F_PERMANENT;
11222 11229 if (NCE_MYADDR(ncec))
11223 11230 flags |= (ACE_F_MYADDR | ACE_F_AUTHORITY);
11224 11231 if (ncec->ncec_flags & NCE_F_UNVERIFIED)
11225 11232 flags |= ACE_F_UNVERIFIED;
11226 11233 if (ncec->ncec_flags & NCE_F_AUTHORITY)
11227 11234 flags |= ACE_F_AUTHORITY;
11228 11235 if (ncec->ncec_flags & NCE_F_DELAYED)
11229 11236 flags |= ACE_F_DELAYED;
11230 11237 return (flags);
11231 11238 }
11232 11239
11233 11240 /*
11234 11241 * ncec_walk routine to create ipNetToMediaEntryTable
11235 11242 */
11236 11243 static void
11237 11244 ip_snmp_get2_v4_media(ncec_t *ncec, void *ptr)
11238 11245 {
11239 11246 iproutedata_t *ird = ptr;
11240 11247 ill_t *ill;
11241 11248 mib2_ipNetToMediaEntry_t ntme;
11242 11249 const char *name = "unknown";
11243 11250 ipaddr_t ncec_addr;
11244 11251
11245 11252 ill = ncec->ncec_ill;
11246 11253 if (ill->ill_isv6 || (ncec->ncec_flags & NCE_F_BCAST) ||
11247 11254 ill->ill_net_type == IRE_LOOPBACK)
11248 11255 return;
11249 11256
11250 11257 /* We report all IPMP groups on ncec_ill which is normally the upper. */
11251 11258 name = ill->ill_name;
11252 11259 /* Based on RFC 4293: other(1), inval(2), dyn(3), stat(4) */
11253 11260 if (NCE_MYADDR(ncec)) {
11254 11261 ntme.ipNetToMediaType = 4;
11255 11262 } else if (ncec->ncec_flags & (NCE_F_MCAST|NCE_F_BCAST|NCE_F_PUBLISH)) {
11256 11263 ntme.ipNetToMediaType = 1;
11257 11264 } else {
11258 11265 ntme.ipNetToMediaType = 3;
11259 11266 }
11260 11267 ntme.ipNetToMediaIfIndex.o_length = MIN(OCTET_LENGTH, strlen(name));
11261 11268 bcopy(name, ntme.ipNetToMediaIfIndex.o_bytes,
11262 11269 ntme.ipNetToMediaIfIndex.o_length);
11263 11270
11264 11271 IN6_V4MAPPED_TO_IPADDR(&ncec->ncec_addr, ncec_addr);
11265 11272 bcopy(&ncec_addr, &ntme.ipNetToMediaNetAddress, sizeof (ncec_addr));
11266 11273
11267 11274 ntme.ipNetToMediaInfo.ntm_mask.o_length = sizeof (ipaddr_t);
11268 11275 ncec_addr = INADDR_BROADCAST;
11269 11276 bcopy(&ncec_addr, ntme.ipNetToMediaInfo.ntm_mask.o_bytes,
11270 11277 sizeof (ncec_addr));
11271 11278 /*
11272 11279 * map all the flags to the ACE counterpart.
11273 11280 */
11274 11281 ntme.ipNetToMediaInfo.ntm_flags = nce2ace(ncec);
11275 11282
11276 11283 ntme.ipNetToMediaPhysAddress.o_length =
11277 11284 MIN(OCTET_LENGTH, ill->ill_phys_addr_length);
11278 11285
11279 11286 if (!NCE_ISREACHABLE(ncec))
11280 11287 ntme.ipNetToMediaPhysAddress.o_length = 0;
11281 11288 else {
11282 11289 if (ncec->ncec_lladdr != NULL) {
11283 11290 bcopy(ncec->ncec_lladdr,
11284 11291 ntme.ipNetToMediaPhysAddress.o_bytes,
11285 11292 ntme.ipNetToMediaPhysAddress.o_length);
11286 11293 }
11287 11294 }
11288 11295
11289 11296 if (!snmp_append_data2(ird->ird_netmedia.lp_head,
11290 11297 &ird->ird_netmedia.lp_tail, (char *)&ntme, sizeof (ntme))) {
11291 11298 ip1dbg(("ip_snmp_get2_v4_media: failed to allocate %u bytes\n",
11292 11299 (uint_t)sizeof (ntme)));
11293 11300 }
11294 11301 }
11295 11302
11296 11303 /*
11297 11304 * return (0) if invalid set request, 1 otherwise, including non-tcp requests
11298 11305 */
11299 11306 /* ARGSUSED */
11300 11307 int
11301 11308 ip_snmp_set(queue_t *q, int level, int name, uchar_t *ptr, int len)
11302 11309 {
11303 11310 switch (level) {
11304 11311 case MIB2_IP:
11305 11312 case MIB2_ICMP:
11306 11313 switch (name) {
11307 11314 default:
11308 11315 break;
11309 11316 }
11310 11317 return (1);
11311 11318 default:
11312 11319 return (1);
11313 11320 }
11314 11321 }
11315 11322
11316 11323 /*
11317 11324 * When there exists both a 64- and 32-bit counter of a particular type
11318 11325 * (i.e., InReceives), only the 64-bit counters are added.
11319 11326 */
11320 11327 void
11321 11328 ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *o1, mib2_ipIfStatsEntry_t *o2)
11322 11329 {
11323 11330 UPDATE_MIB(o1, ipIfStatsInHdrErrors, o2->ipIfStatsInHdrErrors);
11324 11331 UPDATE_MIB(o1, ipIfStatsInTooBigErrors, o2->ipIfStatsInTooBigErrors);
11325 11332 UPDATE_MIB(o1, ipIfStatsInNoRoutes, o2->ipIfStatsInNoRoutes);
11326 11333 UPDATE_MIB(o1, ipIfStatsInAddrErrors, o2->ipIfStatsInAddrErrors);
11327 11334 UPDATE_MIB(o1, ipIfStatsInUnknownProtos, o2->ipIfStatsInUnknownProtos);
11328 11335 UPDATE_MIB(o1, ipIfStatsInTruncatedPkts, o2->ipIfStatsInTruncatedPkts);
11329 11336 UPDATE_MIB(o1, ipIfStatsInDiscards, o2->ipIfStatsInDiscards);
11330 11337 UPDATE_MIB(o1, ipIfStatsOutDiscards, o2->ipIfStatsOutDiscards);
11331 11338 UPDATE_MIB(o1, ipIfStatsOutFragOKs, o2->ipIfStatsOutFragOKs);
11332 11339 UPDATE_MIB(o1, ipIfStatsOutFragFails, o2->ipIfStatsOutFragFails);
11333 11340 UPDATE_MIB(o1, ipIfStatsOutFragCreates, o2->ipIfStatsOutFragCreates);
11334 11341 UPDATE_MIB(o1, ipIfStatsReasmReqds, o2->ipIfStatsReasmReqds);
11335 11342 UPDATE_MIB(o1, ipIfStatsReasmOKs, o2->ipIfStatsReasmOKs);
11336 11343 UPDATE_MIB(o1, ipIfStatsReasmFails, o2->ipIfStatsReasmFails);
11337 11344 UPDATE_MIB(o1, ipIfStatsOutNoRoutes, o2->ipIfStatsOutNoRoutes);
11338 11345 UPDATE_MIB(o1, ipIfStatsReasmDuplicates, o2->ipIfStatsReasmDuplicates);
11339 11346 UPDATE_MIB(o1, ipIfStatsReasmPartDups, o2->ipIfStatsReasmPartDups);
11340 11347 UPDATE_MIB(o1, ipIfStatsForwProhibits, o2->ipIfStatsForwProhibits);
11341 11348 UPDATE_MIB(o1, udpInCksumErrs, o2->udpInCksumErrs);
11342 11349 UPDATE_MIB(o1, udpInOverflows, o2->udpInOverflows);
11343 11350 UPDATE_MIB(o1, rawipInOverflows, o2->rawipInOverflows);
11344 11351 UPDATE_MIB(o1, ipIfStatsInWrongIPVersion,
11345 11352 o2->ipIfStatsInWrongIPVersion);
11346 11353 UPDATE_MIB(o1, ipIfStatsOutWrongIPVersion,
11347 11354 o2->ipIfStatsInWrongIPVersion);
11348 11355 UPDATE_MIB(o1, ipIfStatsOutSwitchIPVersion,
11349 11356 o2->ipIfStatsOutSwitchIPVersion);
11350 11357 UPDATE_MIB(o1, ipIfStatsHCInReceives, o2->ipIfStatsHCInReceives);
11351 11358 UPDATE_MIB(o1, ipIfStatsHCInOctets, o2->ipIfStatsHCInOctets);
11352 11359 UPDATE_MIB(o1, ipIfStatsHCInForwDatagrams,
11353 11360 o2->ipIfStatsHCInForwDatagrams);
11354 11361 UPDATE_MIB(o1, ipIfStatsHCInDelivers, o2->ipIfStatsHCInDelivers);
11355 11362 UPDATE_MIB(o1, ipIfStatsHCOutRequests, o2->ipIfStatsHCOutRequests);
11356 11363 UPDATE_MIB(o1, ipIfStatsHCOutForwDatagrams,
11357 11364 o2->ipIfStatsHCOutForwDatagrams);
11358 11365 UPDATE_MIB(o1, ipIfStatsOutFragReqds, o2->ipIfStatsOutFragReqds);
11359 11366 UPDATE_MIB(o1, ipIfStatsHCOutTransmits, o2->ipIfStatsHCOutTransmits);
11360 11367 UPDATE_MIB(o1, ipIfStatsHCOutOctets, o2->ipIfStatsHCOutOctets);
11361 11368 UPDATE_MIB(o1, ipIfStatsHCInMcastPkts, o2->ipIfStatsHCInMcastPkts);
11362 11369 UPDATE_MIB(o1, ipIfStatsHCInMcastOctets, o2->ipIfStatsHCInMcastOctets);
11363 11370 UPDATE_MIB(o1, ipIfStatsHCOutMcastPkts, o2->ipIfStatsHCOutMcastPkts);
11364 11371 UPDATE_MIB(o1, ipIfStatsHCOutMcastOctets,
11365 11372 o2->ipIfStatsHCOutMcastOctets);
11366 11373 UPDATE_MIB(o1, ipIfStatsHCInBcastPkts, o2->ipIfStatsHCInBcastPkts);
11367 11374 UPDATE_MIB(o1, ipIfStatsHCOutBcastPkts, o2->ipIfStatsHCOutBcastPkts);
11368 11375 UPDATE_MIB(o1, ipsecInSucceeded, o2->ipsecInSucceeded);
11369 11376 UPDATE_MIB(o1, ipsecInFailed, o2->ipsecInFailed);
11370 11377 UPDATE_MIB(o1, ipInCksumErrs, o2->ipInCksumErrs);
11371 11378 UPDATE_MIB(o1, tcpInErrs, o2->tcpInErrs);
11372 11379 UPDATE_MIB(o1, udpNoPorts, o2->udpNoPorts);
11373 11380 }
11374 11381
11375 11382 void
11376 11383 ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *o1, mib2_ipv6IfIcmpEntry_t *o2)
11377 11384 {
11378 11385 UPDATE_MIB(o1, ipv6IfIcmpInMsgs, o2->ipv6IfIcmpInMsgs);
11379 11386 UPDATE_MIB(o1, ipv6IfIcmpInErrors, o2->ipv6IfIcmpInErrors);
11380 11387 UPDATE_MIB(o1, ipv6IfIcmpInDestUnreachs, o2->ipv6IfIcmpInDestUnreachs);
11381 11388 UPDATE_MIB(o1, ipv6IfIcmpInAdminProhibs, o2->ipv6IfIcmpInAdminProhibs);
11382 11389 UPDATE_MIB(o1, ipv6IfIcmpInTimeExcds, o2->ipv6IfIcmpInTimeExcds);
11383 11390 UPDATE_MIB(o1, ipv6IfIcmpInParmProblems, o2->ipv6IfIcmpInParmProblems);
11384 11391 UPDATE_MIB(o1, ipv6IfIcmpInPktTooBigs, o2->ipv6IfIcmpInPktTooBigs);
11385 11392 UPDATE_MIB(o1, ipv6IfIcmpInEchos, o2->ipv6IfIcmpInEchos);
11386 11393 UPDATE_MIB(o1, ipv6IfIcmpInEchoReplies, o2->ipv6IfIcmpInEchoReplies);
11387 11394 UPDATE_MIB(o1, ipv6IfIcmpInRouterSolicits,
11388 11395 o2->ipv6IfIcmpInRouterSolicits);
11389 11396 UPDATE_MIB(o1, ipv6IfIcmpInRouterAdvertisements,
11390 11397 o2->ipv6IfIcmpInRouterAdvertisements);
11391 11398 UPDATE_MIB(o1, ipv6IfIcmpInNeighborSolicits,
11392 11399 o2->ipv6IfIcmpInNeighborSolicits);
11393 11400 UPDATE_MIB(o1, ipv6IfIcmpInNeighborAdvertisements,
11394 11401 o2->ipv6IfIcmpInNeighborAdvertisements);
11395 11402 UPDATE_MIB(o1, ipv6IfIcmpInRedirects, o2->ipv6IfIcmpInRedirects);
11396 11403 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembQueries,
11397 11404 o2->ipv6IfIcmpInGroupMembQueries);
11398 11405 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembResponses,
11399 11406 o2->ipv6IfIcmpInGroupMembResponses);
11400 11407 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembReductions,
11401 11408 o2->ipv6IfIcmpInGroupMembReductions);
11402 11409 UPDATE_MIB(o1, ipv6IfIcmpOutMsgs, o2->ipv6IfIcmpOutMsgs);
11403 11410 UPDATE_MIB(o1, ipv6IfIcmpOutErrors, o2->ipv6IfIcmpOutErrors);
11404 11411 UPDATE_MIB(o1, ipv6IfIcmpOutDestUnreachs,
11405 11412 o2->ipv6IfIcmpOutDestUnreachs);
11406 11413 UPDATE_MIB(o1, ipv6IfIcmpOutAdminProhibs,
11407 11414 o2->ipv6IfIcmpOutAdminProhibs);
11408 11415 UPDATE_MIB(o1, ipv6IfIcmpOutTimeExcds, o2->ipv6IfIcmpOutTimeExcds);
11409 11416 UPDATE_MIB(o1, ipv6IfIcmpOutParmProblems,
11410 11417 o2->ipv6IfIcmpOutParmProblems);
11411 11418 UPDATE_MIB(o1, ipv6IfIcmpOutPktTooBigs, o2->ipv6IfIcmpOutPktTooBigs);
11412 11419 UPDATE_MIB(o1, ipv6IfIcmpOutEchos, o2->ipv6IfIcmpOutEchos);
11413 11420 UPDATE_MIB(o1, ipv6IfIcmpOutEchoReplies, o2->ipv6IfIcmpOutEchoReplies);
11414 11421 UPDATE_MIB(o1, ipv6IfIcmpOutRouterSolicits,
11415 11422 o2->ipv6IfIcmpOutRouterSolicits);
11416 11423 UPDATE_MIB(o1, ipv6IfIcmpOutRouterAdvertisements,
11417 11424 o2->ipv6IfIcmpOutRouterAdvertisements);
11418 11425 UPDATE_MIB(o1, ipv6IfIcmpOutNeighborSolicits,
11419 11426 o2->ipv6IfIcmpOutNeighborSolicits);
11420 11427 UPDATE_MIB(o1, ipv6IfIcmpOutNeighborAdvertisements,
11421 11428 o2->ipv6IfIcmpOutNeighborAdvertisements);
11422 11429 UPDATE_MIB(o1, ipv6IfIcmpOutRedirects, o2->ipv6IfIcmpOutRedirects);
11423 11430 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembQueries,
11424 11431 o2->ipv6IfIcmpOutGroupMembQueries);
11425 11432 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembResponses,
11426 11433 o2->ipv6IfIcmpOutGroupMembResponses);
11427 11434 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembReductions,
11428 11435 o2->ipv6IfIcmpOutGroupMembReductions);
11429 11436 UPDATE_MIB(o1, ipv6IfIcmpInOverflows, o2->ipv6IfIcmpInOverflows);
11430 11437 UPDATE_MIB(o1, ipv6IfIcmpBadHoplimit, o2->ipv6IfIcmpBadHoplimit);
11431 11438 UPDATE_MIB(o1, ipv6IfIcmpInBadNeighborAdvertisements,
11432 11439 o2->ipv6IfIcmpInBadNeighborAdvertisements);
11433 11440 UPDATE_MIB(o1, ipv6IfIcmpInBadNeighborSolicitations,
11434 11441 o2->ipv6IfIcmpInBadNeighborSolicitations);
11435 11442 UPDATE_MIB(o1, ipv6IfIcmpInBadRedirects, o2->ipv6IfIcmpInBadRedirects);
11436 11443 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembTotal,
11437 11444 o2->ipv6IfIcmpInGroupMembTotal);
11438 11445 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembBadQueries,
11439 11446 o2->ipv6IfIcmpInGroupMembBadQueries);
11440 11447 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembBadReports,
11441 11448 o2->ipv6IfIcmpInGroupMembBadReports);
11442 11449 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembOurReports,
11443 11450 o2->ipv6IfIcmpInGroupMembOurReports);
11444 11451 }
11445 11452
11446 11453 /*
11447 11454 * Called before the options are updated to check if this packet will
11448 11455 * be source routed from here.
11449 11456 * This routine assumes that the options are well formed i.e. that they
11450 11457 * have already been checked.
11451 11458 */
11452 11459 boolean_t
11453 11460 ip_source_routed(ipha_t *ipha, ip_stack_t *ipst)
11454 11461 {
11455 11462 ipoptp_t opts;
11456 11463 uchar_t *opt;
11457 11464 uint8_t optval;
11458 11465 uint8_t optlen;
11459 11466 ipaddr_t dst;
11460 11467
11461 11468 if (IS_SIMPLE_IPH(ipha)) {
11462 11469 ip2dbg(("not source routed\n"));
11463 11470 return (B_FALSE);
11464 11471 }
11465 11472 dst = ipha->ipha_dst;
11466 11473 for (optval = ipoptp_first(&opts, ipha);
11467 11474 optval != IPOPT_EOL;
11468 11475 optval = ipoptp_next(&opts)) {
11469 11476 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11470 11477 opt = opts.ipoptp_cur;
11471 11478 optlen = opts.ipoptp_len;
11472 11479 ip2dbg(("ip_source_routed: opt %d, len %d\n",
11473 11480 optval, optlen));
11474 11481 switch (optval) {
11475 11482 uint32_t off;
11476 11483 case IPOPT_SSRR:
11477 11484 case IPOPT_LSRR:
11478 11485 /*
11479 11486 * If dst is one of our addresses and there are some
11480 11487 * entries left in the source route return (true).
11481 11488 */
11482 11489 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
11483 11490 ip2dbg(("ip_source_routed: not next"
11484 11491 " source route 0x%x\n",
11485 11492 ntohl(dst)));
11486 11493 return (B_FALSE);
11487 11494 }
11488 11495 off = opt[IPOPT_OFFSET];
11489 11496 off--;
11490 11497 if (optlen < IP_ADDR_LEN ||
11491 11498 off > optlen - IP_ADDR_LEN) {
11492 11499 /* End of source route */
11493 11500 ip1dbg(("ip_source_routed: end of SR\n"));
11494 11501 return (B_FALSE);
11495 11502 }
11496 11503 return (B_TRUE);
11497 11504 }
11498 11505 }
11499 11506 ip2dbg(("not source routed\n"));
11500 11507 return (B_FALSE);
11501 11508 }
11502 11509
11503 11510 /*
11504 11511 * ip_unbind is called by the transports to remove a conn from
11505 11512 * the fanout table.
11506 11513 */
11507 11514 void
11508 11515 ip_unbind(conn_t *connp)
11509 11516 {
11510 11517
11511 11518 ASSERT(!MUTEX_HELD(&connp->conn_lock));
11512 11519
11513 11520 if (is_system_labeled() && connp->conn_anon_port) {
11514 11521 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
11515 11522 connp->conn_mlp_type, connp->conn_proto,
11516 11523 ntohs(connp->conn_lport), B_FALSE);
11517 11524 connp->conn_anon_port = 0;
11518 11525 }
11519 11526 connp->conn_mlp_type = mlptSingle;
11520 11527
11521 11528 ipcl_hash_remove(connp);
11522 11529 }
11523 11530
11524 11531 /*
11525 11532 * Used for deciding the MSS size for the upper layer. Thus
11526 11533 * we need to check the outbound policy values in the conn.
11527 11534 */
11528 11535 int
11529 11536 conn_ipsec_length(conn_t *connp)
11530 11537 {
11531 11538 ipsec_latch_t *ipl;
11532 11539
11533 11540 ipl = connp->conn_latch;
11534 11541 if (ipl == NULL)
11535 11542 return (0);
11536 11543
11537 11544 if (connp->conn_ixa->ixa_ipsec_policy == NULL)
11538 11545 return (0);
11539 11546
11540 11547 return (connp->conn_ixa->ixa_ipsec_policy->ipsp_act->ipa_ovhd);
11541 11548 }
11542 11549
11543 11550 /*
11544 11551 * Returns an estimate of the IPsec headers size. This is used if
11545 11552 * we don't want to call into IPsec to get the exact size.
11546 11553 */
11547 11554 int
11548 11555 ipsec_out_extra_length(ip_xmit_attr_t *ixa)
11549 11556 {
11550 11557 ipsec_action_t *a;
11551 11558
11552 11559 if (!(ixa->ixa_flags & IXAF_IPSEC_SECURE))
11553 11560 return (0);
11554 11561
11555 11562 a = ixa->ixa_ipsec_action;
11556 11563 if (a == NULL) {
11557 11564 ASSERT(ixa->ixa_ipsec_policy != NULL);
11558 11565 a = ixa->ixa_ipsec_policy->ipsp_act;
11559 11566 }
11560 11567 ASSERT(a != NULL);
11561 11568
11562 11569 return (a->ipa_ovhd);
11563 11570 }
11564 11571
11565 11572 /*
11566 11573 * If there are any source route options, return the true final
11567 11574 * destination. Otherwise, return the destination.
11568 11575 */
11569 11576 ipaddr_t
11570 11577 ip_get_dst(ipha_t *ipha)
11571 11578 {
11572 11579 ipoptp_t opts;
11573 11580 uchar_t *opt;
11574 11581 uint8_t optval;
11575 11582 uint8_t optlen;
11576 11583 ipaddr_t dst;
11577 11584 uint32_t off;
11578 11585
11579 11586 dst = ipha->ipha_dst;
11580 11587
11581 11588 if (IS_SIMPLE_IPH(ipha))
11582 11589 return (dst);
11583 11590
11584 11591 for (optval = ipoptp_first(&opts, ipha);
11585 11592 optval != IPOPT_EOL;
11586 11593 optval = ipoptp_next(&opts)) {
11587 11594 opt = opts.ipoptp_cur;
11588 11595 optlen = opts.ipoptp_len;
11589 11596 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11590 11597 switch (optval) {
11591 11598 case IPOPT_SSRR:
11592 11599 case IPOPT_LSRR:
11593 11600 off = opt[IPOPT_OFFSET];
11594 11601 /*
11595 11602 * If one of the conditions is true, it means
11596 11603 * end of options and dst already has the right
11597 11604 * value.
11598 11605 */
11599 11606 if (!(optlen < IP_ADDR_LEN || off > optlen - 3)) {
11600 11607 off = optlen - IP_ADDR_LEN;
11601 11608 bcopy(&opt[off], &dst, IP_ADDR_LEN);
11602 11609 }
11603 11610 return (dst);
11604 11611 default:
11605 11612 break;
11606 11613 }
11607 11614 }
11608 11615
11609 11616 return (dst);
11610 11617 }
11611 11618
11612 11619 /*
11613 11620 * Outbound IP fragmentation routine.
11614 11621 * Assumes the caller has checked whether or not fragmentation should
11615 11622 * be allowed. Here we copy the DF bit from the header to all the generated
11616 11623 * fragments.
11617 11624 */
11618 11625 int
11619 11626 ip_fragment_v4(mblk_t *mp_orig, nce_t *nce, iaflags_t ixaflags,
11620 11627 uint_t pkt_len, uint32_t max_frag, uint32_t xmit_hint, zoneid_t szone,
11621 11628 zoneid_t nolzid, pfirepostfrag_t postfragfn, uintptr_t *ixa_cookie)
11622 11629 {
11623 11630 int i1;
11624 11631 int hdr_len;
11625 11632 mblk_t *hdr_mp;
11626 11633 ipha_t *ipha;
11627 11634 int ip_data_end;
11628 11635 int len;
11629 11636 mblk_t *mp = mp_orig;
11630 11637 int offset;
11631 11638 ill_t *ill = nce->nce_ill;
11632 11639 ip_stack_t *ipst = ill->ill_ipst;
11633 11640 mblk_t *carve_mp;
11634 11641 uint32_t frag_flag;
11635 11642 uint_t priority = mp->b_band;
11636 11643 int error = 0;
11637 11644
11638 11645 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragReqds);
11639 11646
11640 11647 if (pkt_len != msgdsize(mp)) {
11641 11648 ip0dbg(("Packet length mismatch: %d, %ld\n",
11642 11649 pkt_len, msgdsize(mp)));
11643 11650 freemsg(mp);
11644 11651 return (EINVAL);
11645 11652 }
11646 11653
11647 11654 if (max_frag == 0) {
11648 11655 ip1dbg(("ip_fragment_v4: max_frag is zero. Dropping packet\n"));
11649 11656 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11650 11657 ip_drop_output("FragFails: zero max_frag", mp, ill);
11651 11658 freemsg(mp);
11652 11659 return (EINVAL);
11653 11660 }
11654 11661
11655 11662 ASSERT(MBLKL(mp) >= sizeof (ipha_t));
11656 11663 ipha = (ipha_t *)mp->b_rptr;
11657 11664 ASSERT(ntohs(ipha->ipha_length) == pkt_len);
11658 11665 frag_flag = ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_DF;
11659 11666
11660 11667 /*
11661 11668 * Establish the starting offset. May not be zero if we are fragging
11662 11669 * a fragment that is being forwarded.
11663 11670 */
11664 11671 offset = ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_OFFSET;
11665 11672
11666 11673 /* TODO why is this test needed? */
11667 11674 if (((max_frag - ntohs(ipha->ipha_length)) & ~7) < 8) {
11668 11675 /* TODO: notify ulp somehow */
11669 11676 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11670 11677 ip_drop_output("FragFails: bad starting offset", mp, ill);
11671 11678 freemsg(mp);
11672 11679 return (EINVAL);
11673 11680 }
11674 11681
11675 11682 hdr_len = IPH_HDR_LENGTH(ipha);
11676 11683 ipha->ipha_hdr_checksum = 0;
11677 11684
11678 11685 /*
11679 11686 * Establish the number of bytes maximum per frag, after putting
11680 11687 * in the header.
11681 11688 */
11682 11689 len = (max_frag - hdr_len) & ~7;
11683 11690
11684 11691 /* Get a copy of the header for the trailing frags */
11685 11692 hdr_mp = ip_fragment_copyhdr((uchar_t *)ipha, hdr_len, offset, ipst,
11686 11693 mp);
11687 11694 if (hdr_mp == NULL) {
11688 11695 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11689 11696 ip_drop_output("FragFails: no hdr_mp", mp, ill);
11690 11697 freemsg(mp);
11691 11698 return (ENOBUFS);
11692 11699 }
11693 11700
11694 11701 /* Store the starting offset, with the MoreFrags flag. */
11695 11702 i1 = offset | IPH_MF | frag_flag;
11696 11703 ipha->ipha_fragment_offset_and_flags = htons((uint16_t)i1);
11697 11704
11698 11705 /* Establish the ending byte offset, based on the starting offset. */
11699 11706 offset <<= 3;
11700 11707 ip_data_end = offset + ntohs(ipha->ipha_length) - hdr_len;
11701 11708
11702 11709 /* Store the length of the first fragment in the IP header. */
11703 11710 i1 = len + hdr_len;
11704 11711 ASSERT(i1 <= IP_MAXPACKET);
11705 11712 ipha->ipha_length = htons((uint16_t)i1);
11706 11713
11707 11714 /*
11708 11715 * Compute the IP header checksum for the first frag. We have to
11709 11716 * watch out that we stop at the end of the header.
11710 11717 */
11711 11718 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
11712 11719
11713 11720 /*
11714 11721 * Now carve off the first frag. Note that this will include the
11715 11722 * original IP header.
11716 11723 */
11717 11724 if (!(mp = ip_carve_mp(&mp_orig, i1))) {
11718 11725 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11719 11726 ip_drop_output("FragFails: could not carve mp", mp_orig, ill);
11720 11727 freeb(hdr_mp);
11721 11728 freemsg(mp_orig);
11722 11729 return (ENOBUFS);
11723 11730 }
11724 11731
11725 11732 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragCreates);
11726 11733
11727 11734 error = postfragfn(mp, nce, ixaflags, i1, xmit_hint, szone, nolzid,
11728 11735 ixa_cookie);
11729 11736 if (error != 0 && error != EWOULDBLOCK) {
11730 11737 /* No point in sending the other fragments */
11731 11738 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11732 11739 ip_drop_output("FragFails: postfragfn failed", mp_orig, ill);
11733 11740 freeb(hdr_mp);
11734 11741 freemsg(mp_orig);
11735 11742 return (error);
11736 11743 }
11737 11744
11738 11745 /* No need to redo state machine in loop */
11739 11746 ixaflags &= ~IXAF_REACH_CONF;
11740 11747
11741 11748 /* Advance the offset to the second frag starting point. */
11742 11749 offset += len;
11743 11750 /*
11744 11751 * Update hdr_len from the copied header - there might be less options
11745 11752 * in the later fragments.
11746 11753 */
11747 11754 hdr_len = IPH_HDR_LENGTH(hdr_mp->b_rptr);
11748 11755 /* Loop until done. */
11749 11756 for (;;) {
11750 11757 uint16_t offset_and_flags;
11751 11758 uint16_t ip_len;
11752 11759
11753 11760 if (ip_data_end - offset > len) {
11754 11761 /*
11755 11762 * Carve off the appropriate amount from the original
11756 11763 * datagram.
11757 11764 */
11758 11765 if (!(carve_mp = ip_carve_mp(&mp_orig, len))) {
11759 11766 mp = NULL;
11760 11767 break;
11761 11768 }
11762 11769 /*
11763 11770 * More frags after this one. Get another copy
11764 11771 * of the header.
11765 11772 */
11766 11773 if (carve_mp->b_datap->db_ref == 1 &&
11767 11774 hdr_mp->b_wptr - hdr_mp->b_rptr <
11768 11775 carve_mp->b_rptr - carve_mp->b_datap->db_base) {
11769 11776 /* Inline IP header */
11770 11777 carve_mp->b_rptr -= hdr_mp->b_wptr -
11771 11778 hdr_mp->b_rptr;
11772 11779 bcopy(hdr_mp->b_rptr, carve_mp->b_rptr,
11773 11780 hdr_mp->b_wptr - hdr_mp->b_rptr);
11774 11781 mp = carve_mp;
11775 11782 } else {
11776 11783 if (!(mp = copyb(hdr_mp))) {
11777 11784 freemsg(carve_mp);
11778 11785 break;
11779 11786 }
11780 11787 /* Get priority marking, if any. */
11781 11788 mp->b_band = priority;
11782 11789 mp->b_cont = carve_mp;
11783 11790 }
11784 11791 ipha = (ipha_t *)mp->b_rptr;
11785 11792 offset_and_flags = IPH_MF;
11786 11793 } else {
11787 11794 /*
11788 11795 * Last frag. Consume the header. Set len to
11789 11796 * the length of this last piece.
11790 11797 */
11791 11798 len = ip_data_end - offset;
11792 11799
11793 11800 /*
11794 11801 * Carve off the appropriate amount from the original
11795 11802 * datagram.
11796 11803 */
11797 11804 if (!(carve_mp = ip_carve_mp(&mp_orig, len))) {
11798 11805 mp = NULL;
11799 11806 break;
11800 11807 }
11801 11808 if (carve_mp->b_datap->db_ref == 1 &&
11802 11809 hdr_mp->b_wptr - hdr_mp->b_rptr <
11803 11810 carve_mp->b_rptr - carve_mp->b_datap->db_base) {
11804 11811 /* Inline IP header */
11805 11812 carve_mp->b_rptr -= hdr_mp->b_wptr -
11806 11813 hdr_mp->b_rptr;
11807 11814 bcopy(hdr_mp->b_rptr, carve_mp->b_rptr,
11808 11815 hdr_mp->b_wptr - hdr_mp->b_rptr);
11809 11816 mp = carve_mp;
11810 11817 freeb(hdr_mp);
11811 11818 hdr_mp = mp;
11812 11819 } else {
11813 11820 mp = hdr_mp;
11814 11821 /* Get priority marking, if any. */
11815 11822 mp->b_band = priority;
11816 11823 mp->b_cont = carve_mp;
11817 11824 }
11818 11825 ipha = (ipha_t *)mp->b_rptr;
11819 11826 /* A frag of a frag might have IPH_MF non-zero */
11820 11827 offset_and_flags =
11821 11828 ntohs(ipha->ipha_fragment_offset_and_flags) &
11822 11829 IPH_MF;
11823 11830 }
11824 11831 offset_and_flags |= (uint16_t)(offset >> 3);
11825 11832 offset_and_flags |= (uint16_t)frag_flag;
11826 11833 /* Store the offset and flags in the IP header. */
11827 11834 ipha->ipha_fragment_offset_and_flags = htons(offset_and_flags);
11828 11835
11829 11836 /* Store the length in the IP header. */
11830 11837 ip_len = (uint16_t)(len + hdr_len);
11831 11838 ipha->ipha_length = htons(ip_len);
11832 11839
11833 11840 /*
11834 11841 * Set the IP header checksum. Note that mp is just
11835 11842 * the header, so this is easy to pass to ip_csum.
11836 11843 */
11837 11844 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
11838 11845
11839 11846 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragCreates);
11840 11847
11841 11848 error = postfragfn(mp, nce, ixaflags, ip_len, xmit_hint, szone,
11842 11849 nolzid, ixa_cookie);
11843 11850 /* All done if we just consumed the hdr_mp. */
11844 11851 if (mp == hdr_mp) {
11845 11852 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragOKs);
11846 11853 return (error);
11847 11854 }
11848 11855 if (error != 0 && error != EWOULDBLOCK) {
11849 11856 DTRACE_PROBE2(ip__xmit__frag__fail, ill_t *, ill,
11850 11857 mblk_t *, hdr_mp);
11851 11858 /* No point in sending the other fragments */
11852 11859 break;
11853 11860 }
11854 11861
11855 11862 /* Otherwise, advance and loop. */
11856 11863 offset += len;
11857 11864 }
11858 11865 /* Clean up following allocation failure. */
11859 11866 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11860 11867 ip_drop_output("FragFails: loop ended", NULL, ill);
11861 11868 if (mp != hdr_mp)
11862 11869 freeb(hdr_mp);
11863 11870 if (mp != mp_orig)
11864 11871 freemsg(mp_orig);
11865 11872 return (error);
11866 11873 }
11867 11874
11868 11875 /*
11869 11876 * Copy the header plus those options which have the copy bit set
11870 11877 */
11871 11878 static mblk_t *
11872 11879 ip_fragment_copyhdr(uchar_t *rptr, int hdr_len, int offset, ip_stack_t *ipst,
11873 11880 mblk_t *src)
11874 11881 {
11875 11882 mblk_t *mp;
11876 11883 uchar_t *up;
11877 11884
11878 11885 /*
11879 11886 * Quick check if we need to look for options without the copy bit
11880 11887 * set
11881 11888 */
11882 11889 mp = allocb_tmpl(ipst->ips_ip_wroff_extra + hdr_len, src);
11883 11890 if (!mp)
11884 11891 return (mp);
11885 11892 mp->b_rptr += ipst->ips_ip_wroff_extra;
11886 11893 if (hdr_len == IP_SIMPLE_HDR_LENGTH || offset != 0) {
11887 11894 bcopy(rptr, mp->b_rptr, hdr_len);
11888 11895 mp->b_wptr += hdr_len + ipst->ips_ip_wroff_extra;
11889 11896 return (mp);
11890 11897 }
11891 11898 up = mp->b_rptr;
11892 11899 bcopy(rptr, up, IP_SIMPLE_HDR_LENGTH);
11893 11900 up += IP_SIMPLE_HDR_LENGTH;
11894 11901 rptr += IP_SIMPLE_HDR_LENGTH;
11895 11902 hdr_len -= IP_SIMPLE_HDR_LENGTH;
11896 11903 while (hdr_len > 0) {
11897 11904 uint32_t optval;
11898 11905 uint32_t optlen;
11899 11906
11900 11907 optval = *rptr;
11901 11908 if (optval == IPOPT_EOL)
11902 11909 break;
11903 11910 if (optval == IPOPT_NOP)
11904 11911 optlen = 1;
11905 11912 else
11906 11913 optlen = rptr[1];
11907 11914 if (optval & IPOPT_COPY) {
11908 11915 bcopy(rptr, up, optlen);
11909 11916 up += optlen;
11910 11917 }
11911 11918 rptr += optlen;
11912 11919 hdr_len -= optlen;
11913 11920 }
11914 11921 /*
11915 11922 * Make sure that we drop an even number of words by filling
11916 11923 * with EOL to the next word boundary.
11917 11924 */
11918 11925 for (hdr_len = up - (mp->b_rptr + IP_SIMPLE_HDR_LENGTH);
11919 11926 hdr_len & 0x3; hdr_len++)
11920 11927 *up++ = IPOPT_EOL;
11921 11928 mp->b_wptr = up;
11922 11929 /* Update header length */
11923 11930 mp->b_rptr[0] = (uint8_t)((IP_VERSION << 4) | ((up - mp->b_rptr) >> 2));
11924 11931 return (mp);
11925 11932 }
11926 11933
11927 11934 /*
11928 11935 * Update any source route, record route, or timestamp options when
11929 11936 * sending a packet back to ourselves.
11930 11937 * Check that we are at end of strict source route.
11931 11938 * The options have been sanity checked by ip_output_options().
11932 11939 */
11933 11940 void
11934 11941 ip_output_local_options(ipha_t *ipha, ip_stack_t *ipst)
11935 11942 {
11936 11943 ipoptp_t opts;
11937 11944 uchar_t *opt;
11938 11945 uint8_t optval;
11939 11946 uint8_t optlen;
11940 11947 ipaddr_t dst;
11941 11948 uint32_t ts;
11942 11949 timestruc_t now;
11943 11950
11944 11951 for (optval = ipoptp_first(&opts, ipha);
11945 11952 optval != IPOPT_EOL;
11946 11953 optval = ipoptp_next(&opts)) {
11947 11954 opt = opts.ipoptp_cur;
11948 11955 optlen = opts.ipoptp_len;
11949 11956 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11950 11957 switch (optval) {
11951 11958 uint32_t off;
11952 11959 case IPOPT_SSRR:
11953 11960 case IPOPT_LSRR:
11954 11961 off = opt[IPOPT_OFFSET];
11955 11962 off--;
11956 11963 if (optlen < IP_ADDR_LEN ||
11957 11964 off > optlen - IP_ADDR_LEN) {
11958 11965 /* End of source route */
11959 11966 break;
11960 11967 }
11961 11968 /*
11962 11969 * This will only happen if two consecutive entries
11963 11970 * in the source route contains our address or if
11964 11971 * it is a packet with a loose source route which
11965 11972 * reaches us before consuming the whole source route
11966 11973 */
11967 11974
11968 11975 if (optval == IPOPT_SSRR) {
11969 11976 return;
11970 11977 }
11971 11978 /*
11972 11979 * Hack: instead of dropping the packet truncate the
11973 11980 * source route to what has been used by filling the
11974 11981 * rest with IPOPT_NOP.
11975 11982 */
11976 11983 opt[IPOPT_OLEN] = (uint8_t)off;
11977 11984 while (off < optlen) {
11978 11985 opt[off++] = IPOPT_NOP;
11979 11986 }
11980 11987 break;
11981 11988 case IPOPT_RR:
11982 11989 off = opt[IPOPT_OFFSET];
11983 11990 off--;
11984 11991 if (optlen < IP_ADDR_LEN ||
11985 11992 off > optlen - IP_ADDR_LEN) {
11986 11993 /* No more room - ignore */
11987 11994 ip1dbg((
11988 11995 "ip_output_local_options: end of RR\n"));
11989 11996 break;
11990 11997 }
11991 11998 dst = htonl(INADDR_LOOPBACK);
11992 11999 bcopy(&dst, (char *)opt + off, IP_ADDR_LEN);
11993 12000 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
11994 12001 break;
11995 12002 case IPOPT_TS:
11996 12003 /* Insert timestamp if there is romm */
11997 12004 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
11998 12005 case IPOPT_TS_TSONLY:
11999 12006 off = IPOPT_TS_TIMELEN;
12000 12007 break;
12001 12008 case IPOPT_TS_PRESPEC:
12002 12009 case IPOPT_TS_PRESPEC_RFC791:
12003 12010 /* Verify that the address matched */
12004 12011 off = opt[IPOPT_OFFSET] - 1;
12005 12012 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
12006 12013 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
12007 12014 /* Not for us */
12008 12015 break;
12009 12016 }
12010 12017 /* FALLTHROUGH */
12011 12018 case IPOPT_TS_TSANDADDR:
12012 12019 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
12013 12020 break;
12014 12021 default:
12015 12022 /*
12016 12023 * ip_*put_options should have already
12017 12024 * dropped this packet.
12018 12025 */
12019 12026 cmn_err(CE_PANIC, "ip_output_local_options: "
12020 12027 "unknown IT - bug in ip_output_options?\n");
12021 12028 return; /* Keep "lint" happy */
12022 12029 }
12023 12030 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
12024 12031 /* Increase overflow counter */
12025 12032 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
12026 12033 opt[IPOPT_POS_OV_FLG] = (uint8_t)
12027 12034 (opt[IPOPT_POS_OV_FLG] & 0x0F) |
12028 12035 (off << 4);
12029 12036 break;
12030 12037 }
12031 12038 off = opt[IPOPT_OFFSET] - 1;
12032 12039 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
12033 12040 case IPOPT_TS_PRESPEC:
12034 12041 case IPOPT_TS_PRESPEC_RFC791:
12035 12042 case IPOPT_TS_TSANDADDR:
12036 12043 dst = htonl(INADDR_LOOPBACK);
12037 12044 bcopy(&dst, (char *)opt + off, IP_ADDR_LEN);
12038 12045 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
12039 12046 /* FALLTHROUGH */
12040 12047 case IPOPT_TS_TSONLY:
12041 12048 off = opt[IPOPT_OFFSET] - 1;
12042 12049 /* Compute # of milliseconds since midnight */
12043 12050 gethrestime(&now);
12044 12051 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
12045 12052 NSEC2MSEC(now.tv_nsec);
12046 12053 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
12047 12054 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
12048 12055 break;
12049 12056 }
12050 12057 break;
12051 12058 }
12052 12059 }
12053 12060 }
12054 12061
12055 12062 /*
12056 12063 * Prepend an M_DATA fastpath header, and if none present prepend a
12057 12064 * DL_UNITDATA_REQ. Frees the mblk on failure.
12058 12065 *
12059 12066 * nce_dlur_mp and nce_fp_mp can not disappear once they have been set.
12060 12067 * If there is a change to them, the nce will be deleted (condemned) and
12061 12068 * a new nce_t will be created when packets are sent. Thus we need no locks
12062 12069 * to access those fields.
12063 12070 *
12064 12071 * We preserve b_band to support IPQoS. If a DL_UNITDATA_REQ is prepended
12065 12072 * we place b_band in dl_priority.dl_max.
12066 12073 */
12067 12074 static mblk_t *
12068 12075 ip_xmit_attach_llhdr(mblk_t *mp, nce_t *nce)
12069 12076 {
12070 12077 uint_t hlen;
12071 12078 mblk_t *mp1;
12072 12079 uint_t priority;
12073 12080 uchar_t *rptr;
12074 12081
12075 12082 rptr = mp->b_rptr;
12076 12083
12077 12084 ASSERT(DB_TYPE(mp) == M_DATA);
12078 12085 priority = mp->b_band;
12079 12086
12080 12087 ASSERT(nce != NULL);
12081 12088 if ((mp1 = nce->nce_fp_mp) != NULL) {
12082 12089 hlen = MBLKL(mp1);
12083 12090 /*
12084 12091 * Check if we have enough room to prepend fastpath
12085 12092 * header
12086 12093 */
12087 12094 if (hlen != 0 && (rptr - mp->b_datap->db_base) >= hlen) {
12088 12095 rptr -= hlen;
12089 12096 bcopy(mp1->b_rptr, rptr, hlen);
12090 12097 /*
12091 12098 * Set the b_rptr to the start of the link layer
12092 12099 * header
12093 12100 */
12094 12101 mp->b_rptr = rptr;
12095 12102 return (mp);
12096 12103 }
12097 12104 mp1 = copyb(mp1);
12098 12105 if (mp1 == NULL) {
12099 12106 ill_t *ill = nce->nce_ill;
12100 12107
12101 12108 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12102 12109 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12103 12110 freemsg(mp);
12104 12111 return (NULL);
12105 12112 }
12106 12113 mp1->b_band = priority;
12107 12114 mp1->b_cont = mp;
12108 12115 DB_CKSUMSTART(mp1) = DB_CKSUMSTART(mp);
12109 12116 DB_CKSUMSTUFF(mp1) = DB_CKSUMSTUFF(mp);
12110 12117 DB_CKSUMEND(mp1) = DB_CKSUMEND(mp);
12111 12118 DB_CKSUMFLAGS(mp1) = DB_CKSUMFLAGS(mp);
12112 12119 DB_LSOMSS(mp1) = DB_LSOMSS(mp);
12113 12120 DTRACE_PROBE1(ip__xmit__copyb, (mblk_t *), mp1);
12114 12121 /*
12115 12122 * XXX disable ICK_VALID and compute checksum
12116 12123 * here; can happen if nce_fp_mp changes and
12117 12124 * it can't be copied now due to insufficient
12118 12125 * space. (unlikely, fp mp can change, but it
12119 12126 * does not increase in length)
12120 12127 */
12121 12128 return (mp1);
12122 12129 }
12123 12130 mp1 = copyb(nce->nce_dlur_mp);
12124 12131
12125 12132 if (mp1 == NULL) {
12126 12133 ill_t *ill = nce->nce_ill;
12127 12134
12128 12135 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12129 12136 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12130 12137 freemsg(mp);
12131 12138 return (NULL);
12132 12139 }
12133 12140 mp1->b_cont = mp;
12134 12141 if (priority != 0) {
12135 12142 mp1->b_band = priority;
12136 12143 ((dl_unitdata_req_t *)(mp1->b_rptr))->dl_priority.dl_max =
12137 12144 priority;
12138 12145 }
12139 12146 return (mp1);
12140 12147 }
12141 12148
12142 12149 /*
12143 12150 * Finish the outbound IPsec processing. This function is called from
12144 12151 * ipsec_out_process() if the IPsec packet was processed
12145 12152 * synchronously, or from {ah,esp}_kcf_callback_outbound() if it was processed
12146 12153 * asynchronously.
12147 12154 *
12148 12155 * This is common to IPv4 and IPv6.
12149 12156 */
12150 12157 int
12151 12158 ip_output_post_ipsec(mblk_t *mp, ip_xmit_attr_t *ixa)
12152 12159 {
12153 12160 iaflags_t ixaflags = ixa->ixa_flags;
12154 12161 uint_t pktlen;
12155 12162
12156 12163
12157 12164 /* AH/ESP don't update ixa_pktlen when they modify the packet */
12158 12165 if (ixaflags & IXAF_IS_IPV4) {
12159 12166 ipha_t *ipha = (ipha_t *)mp->b_rptr;
12160 12167
12161 12168 ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
12162 12169 pktlen = ntohs(ipha->ipha_length);
12163 12170 } else {
12164 12171 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
12165 12172
12166 12173 ASSERT(IPH_HDR_VERSION(mp->b_rptr) == IPV6_VERSION);
12167 12174 pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
12168 12175 }
12169 12176
12170 12177 /*
12171 12178 * We release any hard reference on the SAs here to make
12172 12179 * sure the SAs can be garbage collected. ipsr_sa has a soft reference
12173 12180 * on the SAs.
12174 12181 * If in the future we want the hard latching of the SAs in the
12175 12182 * ip_xmit_attr_t then we should remove this.
12176 12183 */
12177 12184 if (ixa->ixa_ipsec_esp_sa != NULL) {
12178 12185 IPSA_REFRELE(ixa->ixa_ipsec_esp_sa);
12179 12186 ixa->ixa_ipsec_esp_sa = NULL;
12180 12187 }
12181 12188 if (ixa->ixa_ipsec_ah_sa != NULL) {
12182 12189 IPSA_REFRELE(ixa->ixa_ipsec_ah_sa);
12183 12190 ixa->ixa_ipsec_ah_sa = NULL;
12184 12191 }
12185 12192
12186 12193 /* Do we need to fragment? */
12187 12194 if ((ixa->ixa_flags & IXAF_IPV6_ADD_FRAGHDR) ||
12188 12195 pktlen > ixa->ixa_fragsize) {
12189 12196 if (ixaflags & IXAF_IS_IPV4) {
12190 12197 ASSERT(!(ixa->ixa_flags & IXAF_IPV6_ADD_FRAGHDR));
12191 12198 /*
12192 12199 * We check for the DF case in ipsec_out_process
12193 12200 * hence this only handles the non-DF case.
12194 12201 */
12195 12202 return (ip_fragment_v4(mp, ixa->ixa_nce, ixa->ixa_flags,
12196 12203 pktlen, ixa->ixa_fragsize,
12197 12204 ixa->ixa_xmit_hint, ixa->ixa_zoneid,
12198 12205 ixa->ixa_no_loop_zoneid, ixa->ixa_postfragfn,
12199 12206 &ixa->ixa_cookie));
12200 12207 } else {
12201 12208 mp = ip_fraghdr_add_v6(mp, ixa->ixa_ident, ixa);
12202 12209 if (mp == NULL) {
12203 12210 /* MIB and ip_drop_output already done */
12204 12211 return (ENOMEM);
12205 12212 }
12206 12213 pktlen += sizeof (ip6_frag_t);
12207 12214 if (pktlen > ixa->ixa_fragsize) {
12208 12215 return (ip_fragment_v6(mp, ixa->ixa_nce,
12209 12216 ixa->ixa_flags, pktlen,
12210 12217 ixa->ixa_fragsize, ixa->ixa_xmit_hint,
12211 12218 ixa->ixa_zoneid, ixa->ixa_no_loop_zoneid,
12212 12219 ixa->ixa_postfragfn, &ixa->ixa_cookie));
12213 12220 }
12214 12221 }
12215 12222 }
12216 12223 return ((ixa->ixa_postfragfn)(mp, ixa->ixa_nce, ixa->ixa_flags,
12217 12224 pktlen, ixa->ixa_xmit_hint, ixa->ixa_zoneid,
12218 12225 ixa->ixa_no_loop_zoneid, NULL));
12219 12226 }
12220 12227
12221 12228 /*
12222 12229 * Finish the inbound IPsec processing. This function is called from
12223 12230 * ipsec_out_process() if the IPsec packet was processed
12224 12231 * synchronously, or from {ah,esp}_kcf_callback_outbound() if it was processed
12225 12232 * asynchronously.
12226 12233 *
12227 12234 * This is common to IPv4 and IPv6.
12228 12235 */
12229 12236 void
12230 12237 ip_input_post_ipsec(mblk_t *mp, ip_recv_attr_t *ira)
12231 12238 {
12232 12239 iaflags_t iraflags = ira->ira_flags;
12233 12240
12234 12241 /* Length might have changed */
12235 12242 if (iraflags & IRAF_IS_IPV4) {
12236 12243 ipha_t *ipha = (ipha_t *)mp->b_rptr;
12237 12244
12238 12245 ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
12239 12246 ira->ira_pktlen = ntohs(ipha->ipha_length);
12240 12247 ira->ira_ip_hdr_length = IPH_HDR_LENGTH(ipha);
12241 12248 ira->ira_protocol = ipha->ipha_protocol;
12242 12249
12243 12250 ip_fanout_v4(mp, ipha, ira);
12244 12251 } else {
12245 12252 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
12246 12253 uint8_t *nexthdrp;
12247 12254
12248 12255 ASSERT(IPH_HDR_VERSION(mp->b_rptr) == IPV6_VERSION);
12249 12256 ira->ira_pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
12250 12257 if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &ira->ira_ip_hdr_length,
12251 12258 &nexthdrp)) {
12252 12259 /* Malformed packet */
12253 12260 BUMP_MIB(ira->ira_ill->ill_ip_mib, ipIfStatsInDiscards);
12254 12261 ip_drop_input("ipIfStatsInDiscards", mp, ira->ira_ill);
12255 12262 freemsg(mp);
12256 12263 return;
12257 12264 }
12258 12265 ira->ira_protocol = *nexthdrp;
12259 12266 ip_fanout_v6(mp, ip6h, ira);
12260 12267 }
12261 12268 }
12262 12269
12263 12270 /*
12264 12271 * Select which AH & ESP SA's to use (if any) for the outbound packet.
12265 12272 *
12266 12273 * If this function returns B_TRUE, the requested SA's have been filled
12267 12274 * into the ixa_ipsec_*_sa pointers.
12268 12275 *
12269 12276 * If the function returns B_FALSE, the packet has been "consumed", most
12270 12277 * likely by an ACQUIRE sent up via PF_KEY to a key management daemon.
12271 12278 *
12272 12279 * The SA references created by the protocol-specific "select"
12273 12280 * function will be released in ip_output_post_ipsec.
12274 12281 */
12275 12282 static boolean_t
12276 12283 ipsec_out_select_sa(mblk_t *mp, ip_xmit_attr_t *ixa)
12277 12284 {
12278 12285 boolean_t need_ah_acquire = B_FALSE, need_esp_acquire = B_FALSE;
12279 12286 ipsec_policy_t *pp;
12280 12287 ipsec_action_t *ap;
12281 12288
12282 12289 ASSERT(ixa->ixa_flags & IXAF_IPSEC_SECURE);
12283 12290 ASSERT((ixa->ixa_ipsec_policy != NULL) ||
12284 12291 (ixa->ixa_ipsec_action != NULL));
12285 12292
12286 12293 ap = ixa->ixa_ipsec_action;
12287 12294 if (ap == NULL) {
12288 12295 pp = ixa->ixa_ipsec_policy;
12289 12296 ASSERT(pp != NULL);
12290 12297 ap = pp->ipsp_act;
12291 12298 ASSERT(ap != NULL);
12292 12299 }
12293 12300
12294 12301 /*
12295 12302 * We have an action. now, let's select SA's.
12296 12303 * A side effect of setting ixa_ipsec_*_sa is that it will
12297 12304 * be cached in the conn_t.
12298 12305 */
12299 12306 if (ap->ipa_want_esp) {
12300 12307 if (ixa->ixa_ipsec_esp_sa == NULL) {
12301 12308 need_esp_acquire = !ipsec_outbound_sa(mp, ixa,
12302 12309 IPPROTO_ESP);
12303 12310 }
12304 12311 ASSERT(need_esp_acquire || ixa->ixa_ipsec_esp_sa != NULL);
12305 12312 }
12306 12313
12307 12314 if (ap->ipa_want_ah) {
12308 12315 if (ixa->ixa_ipsec_ah_sa == NULL) {
12309 12316 need_ah_acquire = !ipsec_outbound_sa(mp, ixa,
12310 12317 IPPROTO_AH);
12311 12318 }
12312 12319 ASSERT(need_ah_acquire || ixa->ixa_ipsec_ah_sa != NULL);
12313 12320 /*
12314 12321 * The ESP and AH processing order needs to be preserved
12315 12322 * when both protocols are required (ESP should be applied
12316 12323 * before AH for an outbound packet). Force an ESP ACQUIRE
12317 12324 * when both ESP and AH are required, and an AH ACQUIRE
12318 12325 * is needed.
12319 12326 */
12320 12327 if (ap->ipa_want_esp && need_ah_acquire)
12321 12328 need_esp_acquire = B_TRUE;
12322 12329 }
12323 12330
12324 12331 /*
12325 12332 * Send an ACQUIRE (extended, regular, or both) if we need one.
12326 12333 * Release SAs that got referenced, but will not be used until we
12327 12334 * acquire _all_ of the SAs we need.
12328 12335 */
12329 12336 if (need_ah_acquire || need_esp_acquire) {
12330 12337 if (ixa->ixa_ipsec_ah_sa != NULL) {
12331 12338 IPSA_REFRELE(ixa->ixa_ipsec_ah_sa);
12332 12339 ixa->ixa_ipsec_ah_sa = NULL;
12333 12340 }
12334 12341 if (ixa->ixa_ipsec_esp_sa != NULL) {
12335 12342 IPSA_REFRELE(ixa->ixa_ipsec_esp_sa);
12336 12343 ixa->ixa_ipsec_esp_sa = NULL;
12337 12344 }
12338 12345
12339 12346 sadb_acquire(mp, ixa, need_ah_acquire, need_esp_acquire);
12340 12347 return (B_FALSE);
12341 12348 }
12342 12349
12343 12350 return (B_TRUE);
12344 12351 }
12345 12352
12346 12353 /*
12347 12354 * Handle IPsec output processing.
12348 12355 * This function is only entered once for a given packet.
12349 12356 * We try to do things synchronously, but if we need to have user-level
12350 12357 * set up SAs, or ESP or AH uses asynchronous kEF, then the operation
12351 12358 * will be completed
12352 12359 * - when the SAs are added in esp_add_sa_finish/ah_add_sa_finish
12353 12360 * - when asynchronous ESP is done it will do AH
12354 12361 *
12355 12362 * In all cases we come back in ip_output_post_ipsec() to fragment and
12356 12363 * send out the packet.
12357 12364 */
12358 12365 int
12359 12366 ipsec_out_process(mblk_t *mp, ip_xmit_attr_t *ixa)
12360 12367 {
12361 12368 ill_t *ill = ixa->ixa_nce->nce_ill;
12362 12369 ip_stack_t *ipst = ixa->ixa_ipst;
12363 12370 ipsec_stack_t *ipss;
12364 12371 ipsec_policy_t *pp;
12365 12372 ipsec_action_t *ap;
12366 12373
12367 12374 ASSERT(ixa->ixa_flags & IXAF_IPSEC_SECURE);
12368 12375
12369 12376 ASSERT((ixa->ixa_ipsec_policy != NULL) ||
12370 12377 (ixa->ixa_ipsec_action != NULL));
12371 12378
12372 12379 ipss = ipst->ips_netstack->netstack_ipsec;
12373 12380 if (!ipsec_loaded(ipss)) {
12374 12381 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12375 12382 ip_drop_packet(mp, B_TRUE, ill,
12376 12383 DROPPER(ipss, ipds_ip_ipsec_not_loaded),
12377 12384 &ipss->ipsec_dropper);
12378 12385 return (ENOTSUP);
12379 12386 }
12380 12387
12381 12388 ap = ixa->ixa_ipsec_action;
12382 12389 if (ap == NULL) {
12383 12390 pp = ixa->ixa_ipsec_policy;
12384 12391 ASSERT(pp != NULL);
12385 12392 ap = pp->ipsp_act;
12386 12393 ASSERT(ap != NULL);
12387 12394 }
12388 12395
12389 12396 /* Handle explicit drop action and bypass. */
12390 12397 switch (ap->ipa_act.ipa_type) {
12391 12398 case IPSEC_ACT_DISCARD:
12392 12399 case IPSEC_ACT_REJECT:
12393 12400 ip_drop_packet(mp, B_FALSE, ill,
12394 12401 DROPPER(ipss, ipds_spd_explicit), &ipss->ipsec_spd_dropper);
12395 12402 return (EHOSTUNREACH); /* IPsec policy failure */
12396 12403 case IPSEC_ACT_BYPASS:
12397 12404 return (ip_output_post_ipsec(mp, ixa));
12398 12405 }
12399 12406
12400 12407 /*
12401 12408 * The order of processing is first insert a IP header if needed.
12402 12409 * Then insert the ESP header and then the AH header.
12403 12410 */
12404 12411 if ((ixa->ixa_flags & IXAF_IS_IPV4) && ap->ipa_want_se) {
12405 12412 /*
12406 12413 * First get the outer IP header before sending
12407 12414 * it to ESP.
12408 12415 */
12409 12416 ipha_t *oipha, *iipha;
12410 12417 mblk_t *outer_mp, *inner_mp;
12411 12418
12412 12419 if ((outer_mp = allocb(sizeof (ipha_t), BPRI_HI)) == NULL) {
12413 12420 (void) mi_strlog(ill->ill_rq, 0,
12414 12421 SL_ERROR|SL_TRACE|SL_CONSOLE,
12415 12422 "ipsec_out_process: "
12416 12423 "Self-Encapsulation failed: Out of memory\n");
12417 12424 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12418 12425 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12419 12426 freemsg(mp);
12420 12427 return (ENOBUFS);
12421 12428 }
12422 12429 inner_mp = mp;
12423 12430 ASSERT(inner_mp->b_datap->db_type == M_DATA);
12424 12431 oipha = (ipha_t *)outer_mp->b_rptr;
12425 12432 iipha = (ipha_t *)inner_mp->b_rptr;
12426 12433 *oipha = *iipha;
12427 12434 outer_mp->b_wptr += sizeof (ipha_t);
12428 12435 oipha->ipha_length = htons(ntohs(iipha->ipha_length) +
12429 12436 sizeof (ipha_t));
12430 12437 oipha->ipha_protocol = IPPROTO_ENCAP;
12431 12438 oipha->ipha_version_and_hdr_length =
12432 12439 IP_SIMPLE_HDR_VERSION;
12433 12440 oipha->ipha_hdr_checksum = 0;
12434 12441 oipha->ipha_hdr_checksum = ip_csum_hdr(oipha);
12435 12442 outer_mp->b_cont = inner_mp;
12436 12443 mp = outer_mp;
12437 12444
12438 12445 ixa->ixa_flags |= IXAF_IPSEC_TUNNEL;
12439 12446 }
12440 12447
12441 12448 /* If we need to wait for a SA then we can't return any errno */
12442 12449 if (((ap->ipa_want_ah && (ixa->ixa_ipsec_ah_sa == NULL)) ||
12443 12450 (ap->ipa_want_esp && (ixa->ixa_ipsec_esp_sa == NULL))) &&
12444 12451 !ipsec_out_select_sa(mp, ixa))
12445 12452 return (0);
12446 12453
12447 12454 /*
12448 12455 * By now, we know what SA's to use. Toss over to ESP & AH
12449 12456 * to do the heavy lifting.
12450 12457 */
12451 12458 if (ap->ipa_want_esp) {
12452 12459 ASSERT(ixa->ixa_ipsec_esp_sa != NULL);
12453 12460
12454 12461 mp = ixa->ixa_ipsec_esp_sa->ipsa_output_func(mp, ixa);
12455 12462 if (mp == NULL) {
12456 12463 /*
12457 12464 * Either it failed or is pending. In the former case
12458 12465 * ipIfStatsInDiscards was increased.
12459 12466 */
12460 12467 return (0);
12461 12468 }
12462 12469 }
12463 12470
12464 12471 if (ap->ipa_want_ah) {
12465 12472 ASSERT(ixa->ixa_ipsec_ah_sa != NULL);
12466 12473
12467 12474 mp = ixa->ixa_ipsec_ah_sa->ipsa_output_func(mp, ixa);
12468 12475 if (mp == NULL) {
12469 12476 /*
12470 12477 * Either it failed or is pending. In the former case
12471 12478 * ipIfStatsInDiscards was increased.
12472 12479 */
12473 12480 return (0);
12474 12481 }
12475 12482 }
12476 12483 /*
12477 12484 * We are done with IPsec processing. Send it over
12478 12485 * the wire.
12479 12486 */
12480 12487 return (ip_output_post_ipsec(mp, ixa));
12481 12488 }
12482 12489
12483 12490 /*
12484 12491 * ioctls that go through a down/up sequence may need to wait for the down
12485 12492 * to complete. This involves waiting for the ire and ipif refcnts to go down
12486 12493 * to zero. Subsequently the ioctl is restarted from ipif_ill_refrele_tail.
12487 12494 */
12488 12495 /* ARGSUSED */
12489 12496 void
12490 12497 ip_reprocess_ioctl(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
12491 12498 {
12492 12499 struct iocblk *iocp;
12493 12500 mblk_t *mp1;
12494 12501 ip_ioctl_cmd_t *ipip;
12495 12502 int err;
12496 12503 sin_t *sin;
12497 12504 struct lifreq *lifr;
12498 12505 struct ifreq *ifr;
12499 12506
12500 12507 iocp = (struct iocblk *)mp->b_rptr;
12501 12508 ASSERT(ipsq != NULL);
12502 12509 /* Existence of mp1 verified in ip_wput_nondata */
12503 12510 mp1 = mp->b_cont->b_cont;
12504 12511 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12505 12512 if (ipip->ipi_cmd == SIOCSLIFNAME || ipip->ipi_cmd == IF_UNITSEL) {
12506 12513 /*
12507 12514 * Special case where ipx_current_ipif is not set:
12508 12515 * ill_phyint_reinit merged the v4 and v6 into a single ipsq.
12509 12516 * We are here as were not able to complete the operation in
12510 12517 * ipif_set_values because we could not become exclusive on
12511 12518 * the new ipsq.
12512 12519 */
12513 12520 ill_t *ill = q->q_ptr;
12514 12521 ipsq_current_start(ipsq, ill->ill_ipif, ipip->ipi_cmd);
12515 12522 }
12516 12523 ASSERT(ipsq->ipsq_xop->ipx_current_ipif != NULL);
12517 12524
12518 12525 if (ipip->ipi_cmd_type == IF_CMD) {
12519 12526 /* This a old style SIOC[GS]IF* command */
12520 12527 ifr = (struct ifreq *)mp1->b_rptr;
12521 12528 sin = (sin_t *)&ifr->ifr_addr;
12522 12529 } else if (ipip->ipi_cmd_type == LIF_CMD) {
12523 12530 /* This a new style SIOC[GS]LIF* command */
12524 12531 lifr = (struct lifreq *)mp1->b_rptr;
12525 12532 sin = (sin_t *)&lifr->lifr_addr;
12526 12533 } else {
12527 12534 sin = NULL;
12528 12535 }
12529 12536
12530 12537 err = (*ipip->ipi_func_restart)(ipsq->ipsq_xop->ipx_current_ipif, sin,
12531 12538 q, mp, ipip, mp1->b_rptr);
12532 12539
12533 12540 DTRACE_PROBE4(ipif__ioctl, char *, "ip_reprocess_ioctl finish",
12534 12541 int, ipip->ipi_cmd,
12535 12542 ill_t *, ipsq->ipsq_xop->ipx_current_ipif->ipif_ill,
12536 12543 ipif_t *, ipsq->ipsq_xop->ipx_current_ipif);
12537 12544
12538 12545 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), ipsq);
12539 12546 }
12540 12547
12541 12548 /*
12542 12549 * ioctl processing
12543 12550 *
12544 12551 * ioctl processing starts with ip_sioctl_copyin_setup(), which looks up
12545 12552 * the ioctl command in the ioctl tables, determines the copyin data size
12546 12553 * from the ipi_copyin_size field, and does an mi_copyin() of that size.
12547 12554 *
12548 12555 * ioctl processing then continues when the M_IOCDATA makes its way down to
12549 12556 * ip_wput_nondata(). The ioctl is looked up again in the ioctl table, its
12550 12557 * associated 'conn' is refheld till the end of the ioctl and the general
12551 12558 * ioctl processing function ip_process_ioctl() is called to extract the
12552 12559 * arguments and process the ioctl. To simplify extraction, ioctl commands
12553 12560 * are "typed" based on the arguments they take (e.g., LIF_CMD which takes a
12554 12561 * `struct lifreq'), and a common extract function (e.g., ip_extract_lifreq())
12555 12562 * is used to extract the ioctl's arguments.
12556 12563 *
12557 12564 * ip_process_ioctl determines if the ioctl needs to be serialized, and if
12558 12565 * so goes thru the serialization primitive ipsq_try_enter. Then the
12559 12566 * appropriate function to handle the ioctl is called based on the entry in
12560 12567 * the ioctl table. ioctl completion is encapsulated in ip_ioctl_finish
12561 12568 * which also refreleases the 'conn' that was refheld at the start of the
12562 12569 * ioctl. Finally ipsq_exit is called if needed to exit the ipsq.
12563 12570 *
12564 12571 * Many exclusive ioctls go thru an internal down up sequence as part of
12565 12572 * the operation. For example an attempt to change the IP address of an
12566 12573 * ipif entails ipif_down, set address, ipif_up. Bringing down the interface
12567 12574 * does all the cleanup such as deleting all ires that use this address.
12568 12575 * Then we need to wait till all references to the interface go away.
12569 12576 */
12570 12577 void
12571 12578 ip_process_ioctl(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *arg)
12572 12579 {
12573 12580 struct iocblk *iocp = (struct iocblk *)mp->b_rptr;
12574 12581 ip_ioctl_cmd_t *ipip = arg;
12575 12582 ip_extract_func_t *extract_funcp;
12576 12583 cmd_info_t ci;
12577 12584 int err;
12578 12585 boolean_t entered_ipsq = B_FALSE;
12579 12586
12580 12587 ip3dbg(("ip_process_ioctl: ioctl %X\n", iocp->ioc_cmd));
12581 12588
12582 12589 if (ipip == NULL)
12583 12590 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12584 12591
12585 12592 /*
12586 12593 * SIOCLIFADDIF needs to go thru a special path since the
12587 12594 * ill may not exist yet. This happens in the case of lo0
12588 12595 * which is created using this ioctl.
12589 12596 */
12590 12597 if (ipip->ipi_cmd == SIOCLIFADDIF) {
12591 12598 err = ip_sioctl_addif(NULL, NULL, q, mp, NULL, NULL);
12592 12599 DTRACE_PROBE4(ipif__ioctl, char *, "ip_process_ioctl finish",
12593 12600 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12594 12601 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12595 12602 return;
12596 12603 }
12597 12604
12598 12605 ci.ci_ipif = NULL;
12599 12606 switch (ipip->ipi_cmd_type) {
12600 12607 case MISC_CMD:
12601 12608 case MSFILT_CMD:
12602 12609 /*
12603 12610 * All MISC_CMD ioctls come in here -- e.g. SIOCGLIFCONF.
12604 12611 */
12605 12612 if (ipip->ipi_cmd == IF_UNITSEL) {
12606 12613 /* ioctl comes down the ill */
12607 12614 ci.ci_ipif = ((ill_t *)q->q_ptr)->ill_ipif;
12608 12615 ipif_refhold(ci.ci_ipif);
12609 12616 }
12610 12617 err = 0;
12611 12618 ci.ci_sin = NULL;
12612 12619 ci.ci_sin6 = NULL;
12613 12620 ci.ci_lifr = NULL;
12614 12621 extract_funcp = NULL;
12615 12622 break;
12616 12623
12617 12624 case IF_CMD:
12618 12625 case LIF_CMD:
12619 12626 extract_funcp = ip_extract_lifreq;
12620 12627 break;
12621 12628
12622 12629 case ARP_CMD:
12623 12630 case XARP_CMD:
12624 12631 extract_funcp = ip_extract_arpreq;
12625 12632 break;
12626 12633
12627 12634 default:
12628 12635 ASSERT(0);
12629 12636 }
12630 12637
12631 12638 if (extract_funcp != NULL) {
12632 12639 err = (*extract_funcp)(q, mp, ipip, &ci);
12633 12640 if (err != 0) {
12634 12641 DTRACE_PROBE4(ipif__ioctl,
12635 12642 char *, "ip_process_ioctl finish err",
12636 12643 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12637 12644 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12638 12645 return;
12639 12646 }
12640 12647
12641 12648 /*
12642 12649 * All of the extraction functions return a refheld ipif.
12643 12650 */
12644 12651 ASSERT(ci.ci_ipif != NULL);
12645 12652 }
12646 12653
12647 12654 if (!(ipip->ipi_flags & IPI_WR)) {
12648 12655 /*
12649 12656 * A return value of EINPROGRESS means the ioctl is
12650 12657 * either queued and waiting for some reason or has
12651 12658 * already completed.
12652 12659 */
12653 12660 err = (*ipip->ipi_func)(ci.ci_ipif, ci.ci_sin, q, mp, ipip,
12654 12661 ci.ci_lifr);
12655 12662 if (ci.ci_ipif != NULL) {
12656 12663 DTRACE_PROBE4(ipif__ioctl,
12657 12664 char *, "ip_process_ioctl finish RD",
12658 12665 int, ipip->ipi_cmd, ill_t *, ci.ci_ipif->ipif_ill,
12659 12666 ipif_t *, ci.ci_ipif);
12660 12667 ipif_refrele(ci.ci_ipif);
12661 12668 } else {
12662 12669 DTRACE_PROBE4(ipif__ioctl,
12663 12670 char *, "ip_process_ioctl finish RD",
12664 12671 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12665 12672 }
12666 12673 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12667 12674 return;
12668 12675 }
12669 12676
12670 12677 ASSERT(ci.ci_ipif != NULL);
12671 12678
12672 12679 /*
12673 12680 * If ipsq is non-NULL, we are already being called exclusively
12674 12681 */
12675 12682 ASSERT(ipsq == NULL || IAM_WRITER_IPSQ(ipsq));
12676 12683 if (ipsq == NULL) {
12677 12684 ipsq = ipsq_try_enter(ci.ci_ipif, NULL, q, mp, ip_process_ioctl,
12678 12685 NEW_OP, B_TRUE);
12679 12686 if (ipsq == NULL) {
12680 12687 ipif_refrele(ci.ci_ipif);
12681 12688 return;
12682 12689 }
12683 12690 entered_ipsq = B_TRUE;
12684 12691 }
12685 12692 /*
12686 12693 * Release the ipif so that ipif_down and friends that wait for
12687 12694 * references to go away are not misled about the current ipif_refcnt
12688 12695 * values. We are writer so we can access the ipif even after releasing
12689 12696 * the ipif.
12690 12697 */
12691 12698 ipif_refrele(ci.ci_ipif);
12692 12699
12693 12700 ipsq_current_start(ipsq, ci.ci_ipif, ipip->ipi_cmd);
12694 12701
12695 12702 /*
12696 12703 * A return value of EINPROGRESS means the ioctl is
12697 12704 * either queued and waiting for some reason or has
12698 12705 * already completed.
12699 12706 */
12700 12707 err = (*ipip->ipi_func)(ci.ci_ipif, ci.ci_sin, q, mp, ipip, ci.ci_lifr);
12701 12708
12702 12709 DTRACE_PROBE4(ipif__ioctl, char *, "ip_process_ioctl finish WR",
12703 12710 int, ipip->ipi_cmd,
12704 12711 ill_t *, ci.ci_ipif == NULL ? NULL : ci.ci_ipif->ipif_ill,
12705 12712 ipif_t *, ci.ci_ipif);
12706 12713 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), ipsq);
12707 12714
12708 12715 if (entered_ipsq)
12709 12716 ipsq_exit(ipsq);
12710 12717 }
12711 12718
12712 12719 /*
12713 12720 * Complete the ioctl. Typically ioctls use the mi package and need to
12714 12721 * do mi_copyout/mi_copy_done.
12715 12722 */
12716 12723 void
12717 12724 ip_ioctl_finish(queue_t *q, mblk_t *mp, int err, int mode, ipsq_t *ipsq)
12718 12725 {
12719 12726 conn_t *connp = NULL;
12720 12727
12721 12728 if (err == EINPROGRESS)
12722 12729 return;
12723 12730
12724 12731 if (CONN_Q(q)) {
12725 12732 connp = Q_TO_CONN(q);
12726 12733 ASSERT(connp->conn_ref >= 2);
12727 12734 }
12728 12735
12729 12736 switch (mode) {
12730 12737 case COPYOUT:
12731 12738 if (err == 0)
12732 12739 mi_copyout(q, mp);
12733 12740 else
12734 12741 mi_copy_done(q, mp, err);
12735 12742 break;
12736 12743
12737 12744 case NO_COPYOUT:
12738 12745 mi_copy_done(q, mp, err);
12739 12746 break;
12740 12747
12741 12748 default:
12742 12749 ASSERT(mode == CONN_CLOSE); /* aborted through CONN_CLOSE */
12743 12750 break;
12744 12751 }
12745 12752
12746 12753 /*
12747 12754 * The conn refhold and ioctlref placed on the conn at the start of the
12748 12755 * ioctl are released here.
12749 12756 */
12750 12757 if (connp != NULL) {
12751 12758 CONN_DEC_IOCTLREF(connp);
12752 12759 CONN_OPER_PENDING_DONE(connp);
12753 12760 }
12754 12761
12755 12762 if (ipsq != NULL)
12756 12763 ipsq_current_finish(ipsq);
12757 12764 }
12758 12765
12759 12766 /* Handles all non data messages */
12760 12767 int
12761 12768 ip_wput_nondata(queue_t *q, mblk_t *mp)
12762 12769 {
12763 12770 mblk_t *mp1;
12764 12771 struct iocblk *iocp;
12765 12772 ip_ioctl_cmd_t *ipip;
12766 12773 conn_t *connp;
12767 12774 cred_t *cr;
12768 12775 char *proto_str;
12769 12776
12770 12777 if (CONN_Q(q))
12771 12778 connp = Q_TO_CONN(q);
12772 12779 else
12773 12780 connp = NULL;
12774 12781
12775 12782 switch (DB_TYPE(mp)) {
12776 12783 case M_IOCTL:
12777 12784 /*
12778 12785 * IOCTL processing begins in ip_sioctl_copyin_setup which
12779 12786 * will arrange to copy in associated control structures.
12780 12787 */
12781 12788 ip_sioctl_copyin_setup(q, mp);
12782 12789 return (0);
12783 12790 case M_IOCDATA:
12784 12791 /*
12785 12792 * Ensure that this is associated with one of our trans-
12786 12793 * parent ioctls. If it's not ours, discard it if we're
12787 12794 * running as a driver, or pass it on if we're a module.
12788 12795 */
12789 12796 iocp = (struct iocblk *)mp->b_rptr;
12790 12797 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12791 12798 if (ipip == NULL) {
12792 12799 if (q->q_next == NULL) {
12793 12800 goto nak;
12794 12801 } else {
12795 12802 putnext(q, mp);
12796 12803 }
12797 12804 return (0);
12798 12805 }
12799 12806 if ((q->q_next != NULL) && !(ipip->ipi_flags & IPI_MODOK)) {
12800 12807 /*
12801 12808 * The ioctl is one we recognise, but is not consumed
12802 12809 * by IP as a module and we are a module, so we drop
12803 12810 */
12804 12811 goto nak;
12805 12812 }
12806 12813
12807 12814 /* IOCTL continuation following copyin or copyout. */
12808 12815 if (mi_copy_state(q, mp, NULL) == -1) {
12809 12816 /*
12810 12817 * The copy operation failed. mi_copy_state already
12811 12818 * cleaned up, so we're out of here.
12812 12819 */
12813 12820 return (0);
12814 12821 }
12815 12822 /*
12816 12823 * If we just completed a copy in, we become writer and
12817 12824 * continue processing in ip_sioctl_copyin_done. If it
12818 12825 * was a copy out, we call mi_copyout again. If there is
12819 12826 * nothing more to copy out, it will complete the IOCTL.
12820 12827 */
12821 12828 if (MI_COPY_DIRECTION(mp) == MI_COPY_IN) {
12822 12829 if (!(mp1 = mp->b_cont) || !(mp1 = mp1->b_cont)) {
12823 12830 mi_copy_done(q, mp, EPROTO);
12824 12831 return (0);
12825 12832 }
12826 12833 /*
12827 12834 * Check for cases that need more copying. A return
12828 12835 * value of 0 means a second copyin has been started,
12829 12836 * so we return; a return value of 1 means no more
12830 12837 * copying is needed, so we continue.
12831 12838 */
12832 12839 if (ipip->ipi_cmd_type == MSFILT_CMD &&
12833 12840 MI_COPY_COUNT(mp) == 1) {
12834 12841 if (ip_copyin_msfilter(q, mp) == 0)
12835 12842 return (0);
12836 12843 }
12837 12844 /*
12838 12845 * Refhold the conn, till the ioctl completes. This is
12839 12846 * needed in case the ioctl ends up in the pending mp
12840 12847 * list. Every mp in the ipx_pending_mp list must have
12841 12848 * a refhold on the conn to resume processing. The
12842 12849 * refhold is released when the ioctl completes
12843 12850 * (whether normally or abnormally). An ioctlref is also
12844 12851 * placed on the conn to prevent TCP from removing the
12845 12852 * queue needed to send the ioctl reply back.
12846 12853 * In all cases ip_ioctl_finish is called to finish
12847 12854 * the ioctl and release the refholds.
12848 12855 */
12849 12856 if (connp != NULL) {
12850 12857 /* This is not a reentry */
12851 12858 CONN_INC_REF(connp);
12852 12859 CONN_INC_IOCTLREF(connp);
12853 12860 } else {
12854 12861 if (!(ipip->ipi_flags & IPI_MODOK)) {
12855 12862 mi_copy_done(q, mp, EINVAL);
12856 12863 return (0);
12857 12864 }
12858 12865 }
12859 12866
12860 12867 ip_process_ioctl(NULL, q, mp, ipip);
12861 12868
12862 12869 } else {
12863 12870 mi_copyout(q, mp);
12864 12871 }
12865 12872 return (0);
12866 12873
12867 12874 case M_IOCNAK:
12868 12875 /*
12869 12876 * The only way we could get here is if a resolver didn't like
12870 12877 * an IOCTL we sent it. This shouldn't happen.
12871 12878 */
12872 12879 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
12873 12880 "ip_wput_nondata: unexpected M_IOCNAK, ioc_cmd 0x%x",
12874 12881 ((struct iocblk *)mp->b_rptr)->ioc_cmd);
12875 12882 freemsg(mp);
12876 12883 return (0);
12877 12884 case M_IOCACK:
12878 12885 /* /dev/ip shouldn't see this */
12879 12886 goto nak;
12880 12887 case M_FLUSH:
12881 12888 if (*mp->b_rptr & FLUSHW)
12882 12889 flushq(q, FLUSHALL);
12883 12890 if (q->q_next) {
12884 12891 putnext(q, mp);
12885 12892 return (0);
12886 12893 }
12887 12894 if (*mp->b_rptr & FLUSHR) {
12888 12895 *mp->b_rptr &= ~FLUSHW;
12889 12896 qreply(q, mp);
12890 12897 return (0);
12891 12898 }
12892 12899 freemsg(mp);
12893 12900 return (0);
12894 12901 case M_CTL:
12895 12902 break;
12896 12903 case M_PROTO:
12897 12904 case M_PCPROTO:
12898 12905 /*
12899 12906 * The only PROTO messages we expect are SNMP-related.
12900 12907 */
12901 12908 switch (((union T_primitives *)mp->b_rptr)->type) {
12902 12909 case T_SVR4_OPTMGMT_REQ:
12903 12910 ip2dbg(("ip_wput_nondata: T_SVR4_OPTMGMT_REQ "
12904 12911 "flags %x\n",
12905 12912 ((struct T_optmgmt_req *)mp->b_rptr)->MGMT_flags));
12906 12913
12907 12914 if (connp == NULL) {
12908 12915 proto_str = "T_SVR4_OPTMGMT_REQ";
12909 12916 goto protonak;
12910 12917 }
12911 12918
12912 12919 /*
12913 12920 * All Solaris components should pass a db_credp
12914 12921 * for this TPI message, hence we ASSERT.
12915 12922 * But in case there is some other M_PROTO that looks
12916 12923 * like a TPI message sent by some other kernel
12917 12924 * component, we check and return an error.
12918 12925 */
12919 12926 cr = msg_getcred(mp, NULL);
12920 12927 ASSERT(cr != NULL);
12921 12928 if (cr == NULL) {
12922 12929 mp = mi_tpi_err_ack_alloc(mp, TSYSERR, EINVAL);
12923 12930 if (mp != NULL)
12924 12931 qreply(q, mp);
12925 12932 return (0);
12926 12933 }
12927 12934
12928 12935 if (!snmpcom_req(q, mp, ip_snmp_set, ip_snmp_get, cr)) {
12929 12936 proto_str = "Bad SNMPCOM request?";
12930 12937 goto protonak;
12931 12938 }
12932 12939 return (0);
12933 12940 default:
12934 12941 ip1dbg(("ip_wput_nondata: dropping M_PROTO prim %u\n",
12935 12942 (int)*(uint_t *)mp->b_rptr));
12936 12943 freemsg(mp);
12937 12944 return (0);
12938 12945 }
12939 12946 default:
12940 12947 break;
12941 12948 }
12942 12949 if (q->q_next) {
12943 12950 putnext(q, mp);
12944 12951 } else
12945 12952 freemsg(mp);
12946 12953 return (0);
12947 12954
12948 12955 nak:
12949 12956 iocp->ioc_error = EINVAL;
12950 12957 mp->b_datap->db_type = M_IOCNAK;
12951 12958 iocp->ioc_count = 0;
12952 12959 qreply(q, mp);
12953 12960 return (0);
12954 12961
12955 12962 protonak:
12956 12963 cmn_err(CE_NOTE, "IP doesn't process %s as a module", proto_str);
12957 12964 if ((mp = mi_tpi_err_ack_alloc(mp, TPROTO, EINVAL)) != NULL)
12958 12965 qreply(q, mp);
12959 12966 return (0);
12960 12967 }
12961 12968
12962 12969 /*
12963 12970 * Process IP options in an outbound packet. Verify that the nexthop in a
12964 12971 * strict source route is onlink.
12965 12972 * Returns non-zero if something fails in which case an ICMP error has been
12966 12973 * sent and mp freed.
12967 12974 *
12968 12975 * Assumes the ULP has called ip_massage_options to move nexthop into ipha_dst.
12969 12976 */
12970 12977 int
12971 12978 ip_output_options(mblk_t *mp, ipha_t *ipha, ip_xmit_attr_t *ixa, ill_t *ill)
12972 12979 {
12973 12980 ipoptp_t opts;
12974 12981 uchar_t *opt;
12975 12982 uint8_t optval;
12976 12983 uint8_t optlen;
12977 12984 ipaddr_t dst;
12978 12985 intptr_t code = 0;
12979 12986 ire_t *ire;
12980 12987 ip_stack_t *ipst = ixa->ixa_ipst;
12981 12988 ip_recv_attr_t iras;
12982 12989
12983 12990 ip2dbg(("ip_output_options\n"));
12984 12991
12985 12992 dst = ipha->ipha_dst;
12986 12993 for (optval = ipoptp_first(&opts, ipha);
12987 12994 optval != IPOPT_EOL;
12988 12995 optval = ipoptp_next(&opts)) {
12989 12996 opt = opts.ipoptp_cur;
12990 12997 optlen = opts.ipoptp_len;
12991 12998 ip2dbg(("ip_output_options: opt %d, len %d\n",
12992 12999 optval, optlen));
12993 13000 switch (optval) {
12994 13001 uint32_t off;
12995 13002 case IPOPT_SSRR:
12996 13003 case IPOPT_LSRR:
12997 13004 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
12998 13005 ip1dbg((
12999 13006 "ip_output_options: bad option offset\n"));
13000 13007 code = (char *)&opt[IPOPT_OLEN] -
13001 13008 (char *)ipha;
13002 13009 goto param_prob;
13003 13010 }
13004 13011 off = opt[IPOPT_OFFSET];
13005 13012 ip1dbg(("ip_output_options: next hop 0x%x\n",
13006 13013 ntohl(dst)));
13007 13014 /*
13008 13015 * For strict: verify that dst is directly
13009 13016 * reachable.
13010 13017 */
13011 13018 if (optval == IPOPT_SSRR) {
13012 13019 ire = ire_ftable_lookup_v4(dst, 0, 0,
13013 13020 IRE_INTERFACE, NULL, ALL_ZONES,
13014 13021 ixa->ixa_tsl,
13015 13022 MATCH_IRE_TYPE | MATCH_IRE_SECATTR, 0, ipst,
13016 13023 NULL);
13017 13024 if (ire == NULL) {
13018 13025 ip1dbg(("ip_output_options: SSRR not"
13019 13026 " directly reachable: 0x%x\n",
13020 13027 ntohl(dst)));
13021 13028 goto bad_src_route;
13022 13029 }
13023 13030 ire_refrele(ire);
13024 13031 }
13025 13032 break;
13026 13033 case IPOPT_RR:
13027 13034 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
13028 13035 ip1dbg((
13029 13036 "ip_output_options: bad option offset\n"));
13030 13037 code = (char *)&opt[IPOPT_OLEN] -
13031 13038 (char *)ipha;
13032 13039 goto param_prob;
13033 13040 }
13034 13041 break;
13035 13042 case IPOPT_TS:
13036 13043 /*
13037 13044 * Verify that length >=5 and that there is either
13038 13045 * room for another timestamp or that the overflow
13039 13046 * counter is not maxed out.
13040 13047 */
13041 13048 code = (char *)&opt[IPOPT_OLEN] - (char *)ipha;
13042 13049 if (optlen < IPOPT_MINLEN_IT) {
13043 13050 goto param_prob;
13044 13051 }
13045 13052 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
13046 13053 ip1dbg((
13047 13054 "ip_output_options: bad option offset\n"));
13048 13055 code = (char *)&opt[IPOPT_OFFSET] -
13049 13056 (char *)ipha;
13050 13057 goto param_prob;
13051 13058 }
13052 13059 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
13053 13060 case IPOPT_TS_TSONLY:
13054 13061 off = IPOPT_TS_TIMELEN;
13055 13062 break;
13056 13063 case IPOPT_TS_TSANDADDR:
13057 13064 case IPOPT_TS_PRESPEC:
13058 13065 case IPOPT_TS_PRESPEC_RFC791:
13059 13066 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
13060 13067 break;
13061 13068 default:
13062 13069 code = (char *)&opt[IPOPT_POS_OV_FLG] -
13063 13070 (char *)ipha;
13064 13071 goto param_prob;
13065 13072 }
13066 13073 if (opt[IPOPT_OFFSET] - 1 + off > optlen &&
13067 13074 (opt[IPOPT_POS_OV_FLG] & 0xF0) == 0xF0) {
13068 13075 /*
13069 13076 * No room and the overflow counter is 15
13070 13077 * already.
13071 13078 */
13072 13079 goto param_prob;
13073 13080 }
13074 13081 break;
13075 13082 }
13076 13083 }
13077 13084
13078 13085 if ((opts.ipoptp_flags & IPOPTP_ERROR) == 0)
13079 13086 return (0);
13080 13087
13081 13088 ip1dbg(("ip_output_options: error processing IP options."));
13082 13089 code = (char *)&opt[IPOPT_OFFSET] - (char *)ipha;
13083 13090
13084 13091 param_prob:
13085 13092 bzero(&iras, sizeof (iras));
13086 13093 iras.ira_ill = iras.ira_rill = ill;
13087 13094 iras.ira_ruifindex = ill->ill_phyint->phyint_ifindex;
13088 13095 iras.ira_rifindex = iras.ira_ruifindex;
13089 13096 iras.ira_flags = IRAF_IS_IPV4;
13090 13097
13091 13098 ip_drop_output("ip_output_options", mp, ill);
13092 13099 icmp_param_problem(mp, (uint8_t)code, &iras);
13093 13100 ASSERT(!(iras.ira_flags & IRAF_IPSEC_SECURE));
13094 13101 return (-1);
13095 13102
13096 13103 bad_src_route:
13097 13104 bzero(&iras, sizeof (iras));
13098 13105 iras.ira_ill = iras.ira_rill = ill;
13099 13106 iras.ira_ruifindex = ill->ill_phyint->phyint_ifindex;
13100 13107 iras.ira_rifindex = iras.ira_ruifindex;
13101 13108 iras.ira_flags = IRAF_IS_IPV4;
13102 13109
13103 13110 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ill);
13104 13111 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, &iras);
13105 13112 ASSERT(!(iras.ira_flags & IRAF_IPSEC_SECURE));
13106 13113 return (-1);
13107 13114 }
13108 13115
13109 13116 /*
13110 13117 * The maximum value of conn_drain_list_cnt is CONN_MAXDRAINCNT.
13111 13118 * conn_drain_list_cnt can be changed by setting conn_drain_nthreads
13112 13119 * thru /etc/system.
13113 13120 */
13114 13121 #define CONN_MAXDRAINCNT 64
13115 13122
13116 13123 static void
13117 13124 conn_drain_init(ip_stack_t *ipst)
13118 13125 {
13119 13126 int i, j;
13120 13127 idl_tx_list_t *itl_tx;
13121 13128
13122 13129 ipst->ips_conn_drain_list_cnt = conn_drain_nthreads;
13123 13130
13124 13131 if ((ipst->ips_conn_drain_list_cnt == 0) ||
13125 13132 (ipst->ips_conn_drain_list_cnt > CONN_MAXDRAINCNT)) {
13126 13133 /*
13127 13134 * Default value of the number of drainers is the
13128 13135 * number of cpus, subject to maximum of 8 drainers.
13129 13136 */
13130 13137 if (boot_max_ncpus != -1)
13131 13138 ipst->ips_conn_drain_list_cnt = MIN(boot_max_ncpus, 8);
13132 13139 else
13133 13140 ipst->ips_conn_drain_list_cnt = MIN(max_ncpus, 8);
13134 13141 }
13135 13142
13136 13143 ipst->ips_idl_tx_list =
13137 13144 kmem_zalloc(TX_FANOUT_SIZE * sizeof (idl_tx_list_t), KM_SLEEP);
13138 13145 for (i = 0; i < TX_FANOUT_SIZE; i++) {
13139 13146 itl_tx = &ipst->ips_idl_tx_list[i];
13140 13147 itl_tx->txl_drain_list =
13141 13148 kmem_zalloc(ipst->ips_conn_drain_list_cnt *
13142 13149 sizeof (idl_t), KM_SLEEP);
13143 13150 mutex_init(&itl_tx->txl_lock, NULL, MUTEX_DEFAULT, NULL);
13144 13151 for (j = 0; j < ipst->ips_conn_drain_list_cnt; j++) {
13145 13152 mutex_init(&itl_tx->txl_drain_list[j].idl_lock, NULL,
13146 13153 MUTEX_DEFAULT, NULL);
13147 13154 itl_tx->txl_drain_list[j].idl_itl = itl_tx;
13148 13155 }
13149 13156 }
13150 13157 }
13151 13158
13152 13159 static void
13153 13160 conn_drain_fini(ip_stack_t *ipst)
13154 13161 {
13155 13162 int i;
13156 13163 idl_tx_list_t *itl_tx;
13157 13164
13158 13165 for (i = 0; i < TX_FANOUT_SIZE; i++) {
13159 13166 itl_tx = &ipst->ips_idl_tx_list[i];
13160 13167 kmem_free(itl_tx->txl_drain_list,
13161 13168 ipst->ips_conn_drain_list_cnt * sizeof (idl_t));
13162 13169 }
13163 13170 kmem_free(ipst->ips_idl_tx_list,
13164 13171 TX_FANOUT_SIZE * sizeof (idl_tx_list_t));
13165 13172 ipst->ips_idl_tx_list = NULL;
13166 13173 }
13167 13174
13168 13175 /*
13169 13176 * Flow control has blocked us from proceeding. Insert the given conn in one
13170 13177 * of the conn drain lists. When flow control is unblocked, either ip_wsrv()
13171 13178 * (STREAMS) or ill_flow_enable() (direct) will be called back, which in turn
13172 13179 * will call conn_walk_drain(). See the flow control notes at the top of this
13173 13180 * file for more details.
13174 13181 */
13175 13182 void
13176 13183 conn_drain_insert(conn_t *connp, idl_tx_list_t *tx_list)
13177 13184 {
13178 13185 idl_t *idl = tx_list->txl_drain_list;
13179 13186 uint_t index;
13180 13187 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
13181 13188
13182 13189 mutex_enter(&connp->conn_lock);
13183 13190 if (connp->conn_state_flags & CONN_CLOSING) {
13184 13191 /*
13185 13192 * The conn is closing as a result of which CONN_CLOSING
13186 13193 * is set. Return.
13187 13194 */
13188 13195 mutex_exit(&connp->conn_lock);
13189 13196 return;
13190 13197 } else if (connp->conn_idl == NULL) {
13191 13198 /*
13192 13199 * Assign the next drain list round robin. We dont' use
13193 13200 * a lock, and thus it may not be strictly round robin.
13194 13201 * Atomicity of load/stores is enough to make sure that
13195 13202 * conn_drain_list_index is always within bounds.
13196 13203 */
13197 13204 index = tx_list->txl_drain_index;
13198 13205 ASSERT(index < ipst->ips_conn_drain_list_cnt);
13199 13206 connp->conn_idl = &tx_list->txl_drain_list[index];
13200 13207 index++;
13201 13208 if (index == ipst->ips_conn_drain_list_cnt)
13202 13209 index = 0;
13203 13210 tx_list->txl_drain_index = index;
13204 13211 } else {
13205 13212 ASSERT(connp->conn_idl->idl_itl == tx_list);
13206 13213 }
13207 13214 mutex_exit(&connp->conn_lock);
13208 13215
13209 13216 idl = connp->conn_idl;
13210 13217 mutex_enter(&idl->idl_lock);
13211 13218 if ((connp->conn_drain_prev != NULL) ||
13212 13219 (connp->conn_state_flags & CONN_CLOSING)) {
13213 13220 /*
13214 13221 * The conn is either already in the drain list or closing.
13215 13222 * (We needed to check for CONN_CLOSING again since close can
13216 13223 * sneak in between dropping conn_lock and acquiring idl_lock.)
13217 13224 */
13218 13225 mutex_exit(&idl->idl_lock);
13219 13226 return;
13220 13227 }
13221 13228
13222 13229 /*
13223 13230 * The conn is not in the drain list. Insert it at the
13224 13231 * tail of the drain list. The drain list is circular
13225 13232 * and doubly linked. idl_conn points to the 1st element
13226 13233 * in the list.
13227 13234 */
13228 13235 if (idl->idl_conn == NULL) {
13229 13236 idl->idl_conn = connp;
13230 13237 connp->conn_drain_next = connp;
13231 13238 connp->conn_drain_prev = connp;
13232 13239 } else {
13233 13240 conn_t *head = idl->idl_conn;
13234 13241
13235 13242 connp->conn_drain_next = head;
13236 13243 connp->conn_drain_prev = head->conn_drain_prev;
13237 13244 head->conn_drain_prev->conn_drain_next = connp;
13238 13245 head->conn_drain_prev = connp;
13239 13246 }
13240 13247 /*
13241 13248 * For non streams based sockets assert flow control.
13242 13249 */
13243 13250 conn_setqfull(connp, NULL);
13244 13251 mutex_exit(&idl->idl_lock);
13245 13252 }
13246 13253
13247 13254 static void
13248 13255 conn_drain_remove(conn_t *connp)
13249 13256 {
13250 13257 idl_t *idl = connp->conn_idl;
13251 13258
13252 13259 if (idl != NULL) {
13253 13260 /*
13254 13261 * Remove ourself from the drain list.
13255 13262 */
13256 13263 if (connp->conn_drain_next == connp) {
13257 13264 /* Singleton in the list */
13258 13265 ASSERT(connp->conn_drain_prev == connp);
13259 13266 idl->idl_conn = NULL;
13260 13267 } else {
13261 13268 connp->conn_drain_prev->conn_drain_next =
13262 13269 connp->conn_drain_next;
13263 13270 connp->conn_drain_next->conn_drain_prev =
13264 13271 connp->conn_drain_prev;
13265 13272 if (idl->idl_conn == connp)
13266 13273 idl->idl_conn = connp->conn_drain_next;
13267 13274 }
13268 13275
13269 13276 /*
13270 13277 * NOTE: because conn_idl is associated with a specific drain
13271 13278 * list which in turn is tied to the index the TX ring
13272 13279 * (txl_cookie) hashes to, and because the TX ring can change
13273 13280 * over the lifetime of the conn_t, we must clear conn_idl so
13274 13281 * a subsequent conn_drain_insert() will set conn_idl again
13275 13282 * based on the latest txl_cookie.
13276 13283 */
13277 13284 connp->conn_idl = NULL;
13278 13285 }
13279 13286 connp->conn_drain_next = NULL;
13280 13287 connp->conn_drain_prev = NULL;
13281 13288
13282 13289 conn_clrqfull(connp, NULL);
13283 13290 /*
13284 13291 * For streams based sockets open up flow control.
13285 13292 */
13286 13293 if (!IPCL_IS_NONSTR(connp))
13287 13294 enableok(connp->conn_wq);
13288 13295 }
13289 13296
13290 13297 /*
13291 13298 * This conn is closing, and we are called from ip_close. OR
13292 13299 * this conn is draining because flow-control on the ill has been relieved.
13293 13300 *
13294 13301 * We must also need to remove conn's on this idl from the list, and also
13295 13302 * inform the sockfs upcalls about the change in flow-control.
13296 13303 */
13297 13304 static void
13298 13305 conn_drain(conn_t *connp, boolean_t closing)
13299 13306 {
13300 13307 idl_t *idl;
13301 13308 conn_t *next_connp;
13302 13309
13303 13310 /*
13304 13311 * connp->conn_idl is stable at this point, and no lock is needed
13305 13312 * to check it. If we are called from ip_close, close has already
13306 13313 * set CONN_CLOSING, thus freezing the value of conn_idl, and
13307 13314 * called us only because conn_idl is non-null. If we are called thru
13308 13315 * service, conn_idl could be null, but it cannot change because
13309 13316 * service is single-threaded per queue, and there cannot be another
13310 13317 * instance of service trying to call conn_drain_insert on this conn
13311 13318 * now.
13312 13319 */
13313 13320 ASSERT(!closing || connp == NULL || connp->conn_idl != NULL);
13314 13321
13315 13322 /*
13316 13323 * If the conn doesn't exist or is not on a drain list, bail.
13317 13324 */
13318 13325 if (connp == NULL || connp->conn_idl == NULL ||
13319 13326 connp->conn_drain_prev == NULL) {
13320 13327 return;
13321 13328 }
13322 13329
13323 13330 idl = connp->conn_idl;
13324 13331 ASSERT(MUTEX_HELD(&idl->idl_lock));
13325 13332
13326 13333 if (!closing) {
13327 13334 next_connp = connp->conn_drain_next;
13328 13335 while (next_connp != connp) {
13329 13336 conn_t *delconnp = next_connp;
13330 13337
13331 13338 next_connp = next_connp->conn_drain_next;
13332 13339 conn_drain_remove(delconnp);
13333 13340 }
13334 13341 ASSERT(connp->conn_drain_next == idl->idl_conn);
13335 13342 }
13336 13343 conn_drain_remove(connp);
13337 13344 }
13338 13345
13339 13346 /*
13340 13347 * Write service routine. Shared perimeter entry point.
13341 13348 * The device queue's messages has fallen below the low water mark and STREAMS
13342 13349 * has backenabled the ill_wq. Send sockfs notification about flow-control on
13343 13350 * each waiting conn.
13344 13351 */
13345 13352 int
13346 13353 ip_wsrv(queue_t *q)
13347 13354 {
13348 13355 ill_t *ill;
13349 13356
13350 13357 ill = (ill_t *)q->q_ptr;
13351 13358 if (ill->ill_state_flags == 0) {
13352 13359 ip_stack_t *ipst = ill->ill_ipst;
13353 13360
13354 13361 /*
13355 13362 * The device flow control has opened up.
13356 13363 * Walk through conn drain lists and qenable the
13357 13364 * first conn in each list. This makes sense only
13358 13365 * if the stream is fully plumbed and setup.
13359 13366 * Hence the ill_state_flags check above.
13360 13367 */
13361 13368 ip1dbg(("ip_wsrv: walking\n"));
13362 13369 conn_walk_drain(ipst, &ipst->ips_idl_tx_list[0]);
13363 13370 enableok(ill->ill_wq);
13364 13371 }
13365 13372 return (0);
13366 13373 }
13367 13374
13368 13375 /*
13369 13376 * Callback to disable flow control in IP.
13370 13377 *
13371 13378 * This is a mac client callback added when the DLD_CAPAB_DIRECT capability
13372 13379 * is enabled.
13373 13380 *
13374 13381 * When MAC_TX() is not able to send any more packets, dld sets its queue
13375 13382 * to QFULL and enable the STREAMS flow control. Later, when the underlying
13376 13383 * driver is able to continue to send packets, it calls mac_tx_(ring_)update()
13377 13384 * function and wakes up corresponding mac worker threads, which in turn
13378 13385 * calls this callback function, and disables flow control.
13379 13386 */
13380 13387 void
13381 13388 ill_flow_enable(void *arg, ip_mac_tx_cookie_t cookie)
13382 13389 {
13383 13390 ill_t *ill = (ill_t *)arg;
13384 13391 ip_stack_t *ipst = ill->ill_ipst;
13385 13392 idl_tx_list_t *idl_txl;
13386 13393
13387 13394 idl_txl = &ipst->ips_idl_tx_list[IDLHASHINDEX(cookie)];
13388 13395 mutex_enter(&idl_txl->txl_lock);
13389 13396 /* add code to to set a flag to indicate idl_txl is enabled */
13390 13397 conn_walk_drain(ipst, idl_txl);
13391 13398 mutex_exit(&idl_txl->txl_lock);
13392 13399 }
13393 13400
13394 13401 /*
13395 13402 * Flow control has been relieved and STREAMS has backenabled us; drain
13396 13403 * all the conn lists on `tx_list'.
13397 13404 */
13398 13405 static void
13399 13406 conn_walk_drain(ip_stack_t *ipst, idl_tx_list_t *tx_list)
13400 13407 {
13401 13408 int i;
13402 13409 idl_t *idl;
13403 13410
13404 13411 IP_STAT(ipst, ip_conn_walk_drain);
13405 13412
13406 13413 for (i = 0; i < ipst->ips_conn_drain_list_cnt; i++) {
13407 13414 idl = &tx_list->txl_drain_list[i];
13408 13415 mutex_enter(&idl->idl_lock);
13409 13416 conn_drain(idl->idl_conn, B_FALSE);
13410 13417 mutex_exit(&idl->idl_lock);
13411 13418 }
13412 13419 }
13413 13420
13414 13421 /*
13415 13422 * Determine if the ill and multicast aspects of that packets
13416 13423 * "matches" the conn.
13417 13424 */
13418 13425 boolean_t
13419 13426 conn_wantpacket(conn_t *connp, ip_recv_attr_t *ira, ipha_t *ipha)
13420 13427 {
13421 13428 ill_t *ill = ira->ira_rill;
13422 13429 zoneid_t zoneid = ira->ira_zoneid;
13423 13430 uint_t in_ifindex;
13424 13431 ipaddr_t dst, src;
13425 13432
13426 13433 dst = ipha->ipha_dst;
13427 13434 src = ipha->ipha_src;
13428 13435
13429 13436 /*
13430 13437 * conn_incoming_ifindex is set by IP_BOUND_IF which limits
13431 13438 * unicast, broadcast and multicast reception to
13432 13439 * conn_incoming_ifindex.
13433 13440 * conn_wantpacket is called for unicast, broadcast and
13434 13441 * multicast packets.
13435 13442 */
13436 13443 in_ifindex = connp->conn_incoming_ifindex;
13437 13444
13438 13445 /* mpathd can bind to the under IPMP interface, which we allow */
13439 13446 if (in_ifindex != 0 && in_ifindex != ill->ill_phyint->phyint_ifindex) {
13440 13447 if (!IS_UNDER_IPMP(ill))
13441 13448 return (B_FALSE);
13442 13449
13443 13450 if (in_ifindex != ipmp_ill_get_ipmp_ifindex(ill))
13444 13451 return (B_FALSE);
13445 13452 }
13446 13453
13447 13454 if (!IPCL_ZONE_MATCH(connp, zoneid))
13448 13455 return (B_FALSE);
13449 13456
13450 13457 if (!(ira->ira_flags & IRAF_MULTICAST))
13451 13458 return (B_TRUE);
13452 13459
13453 13460 if (connp->conn_multi_router) {
13454 13461 /* multicast packet and multicast router socket: send up */
13455 13462 return (B_TRUE);
13456 13463 }
13457 13464
13458 13465 if (ipha->ipha_protocol == IPPROTO_PIM ||
13459 13466 ipha->ipha_protocol == IPPROTO_RSVP)
13460 13467 return (B_TRUE);
13461 13468
13462 13469 return (conn_hasmembers_ill_withsrc_v4(connp, dst, src, ira->ira_ill));
13463 13470 }
13464 13471
13465 13472 void
13466 13473 conn_setqfull(conn_t *connp, boolean_t *flow_stopped)
13467 13474 {
13468 13475 if (IPCL_IS_NONSTR(connp)) {
13469 13476 (*connp->conn_upcalls->su_txq_full)
13470 13477 (connp->conn_upper_handle, B_TRUE);
13471 13478 if (flow_stopped != NULL)
13472 13479 *flow_stopped = B_TRUE;
13473 13480 } else {
13474 13481 queue_t *q = connp->conn_wq;
13475 13482
13476 13483 ASSERT(q != NULL);
13477 13484 if (!(q->q_flag & QFULL)) {
13478 13485 mutex_enter(QLOCK(q));
13479 13486 if (!(q->q_flag & QFULL)) {
13480 13487 /* still need to set QFULL */
13481 13488 q->q_flag |= QFULL;
13482 13489 /* set flow_stopped to true under QLOCK */
13483 13490 if (flow_stopped != NULL)
13484 13491 *flow_stopped = B_TRUE;
13485 13492 mutex_exit(QLOCK(q));
13486 13493 } else {
13487 13494 /* flow_stopped is left unchanged */
13488 13495 mutex_exit(QLOCK(q));
13489 13496 }
13490 13497 }
13491 13498 }
13492 13499 }
13493 13500
13494 13501 void
13495 13502 conn_clrqfull(conn_t *connp, boolean_t *flow_stopped)
13496 13503 {
13497 13504 if (IPCL_IS_NONSTR(connp)) {
13498 13505 (*connp->conn_upcalls->su_txq_full)
13499 13506 (connp->conn_upper_handle, B_FALSE);
13500 13507 if (flow_stopped != NULL)
13501 13508 *flow_stopped = B_FALSE;
13502 13509 } else {
13503 13510 queue_t *q = connp->conn_wq;
13504 13511
13505 13512 ASSERT(q != NULL);
13506 13513 if (q->q_flag & QFULL) {
13507 13514 mutex_enter(QLOCK(q));
13508 13515 if (q->q_flag & QFULL) {
13509 13516 q->q_flag &= ~QFULL;
13510 13517 /* set flow_stopped to false under QLOCK */
13511 13518 if (flow_stopped != NULL)
13512 13519 *flow_stopped = B_FALSE;
13513 13520 mutex_exit(QLOCK(q));
13514 13521 if (q->q_flag & QWANTW)
13515 13522 qbackenable(q, 0);
13516 13523 } else {
13517 13524 /* flow_stopped is left unchanged */
13518 13525 mutex_exit(QLOCK(q));
13519 13526 }
13520 13527 }
13521 13528 }
13522 13529
13523 13530 mutex_enter(&connp->conn_lock);
13524 13531 connp->conn_blocked = B_FALSE;
13525 13532 mutex_exit(&connp->conn_lock);
13526 13533 }
13527 13534
13528 13535 /*
13529 13536 * Return the length in bytes of the IPv4 headers (base header, label, and
13530 13537 * other IP options) that will be needed based on the
13531 13538 * ip_pkt_t structure passed by the caller.
13532 13539 *
13533 13540 * The returned length does not include the length of the upper level
13534 13541 * protocol (ULP) header.
13535 13542 * The caller needs to check that the length doesn't exceed the max for IPv4.
13536 13543 */
13537 13544 int
13538 13545 ip_total_hdrs_len_v4(const ip_pkt_t *ipp)
13539 13546 {
13540 13547 int len;
13541 13548
13542 13549 len = IP_SIMPLE_HDR_LENGTH;
13543 13550 if (ipp->ipp_fields & IPPF_LABEL_V4) {
13544 13551 ASSERT(ipp->ipp_label_len_v4 != 0);
13545 13552 /* We need to round up here */
13546 13553 len += (ipp->ipp_label_len_v4 + 3) & ~3;
13547 13554 }
13548 13555
13549 13556 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
13550 13557 ASSERT(ipp->ipp_ipv4_options_len != 0);
13551 13558 ASSERT((ipp->ipp_ipv4_options_len & 3) == 0);
13552 13559 len += ipp->ipp_ipv4_options_len;
13553 13560 }
13554 13561 return (len);
13555 13562 }
13556 13563
13557 13564 /*
13558 13565 * All-purpose routine to build an IPv4 header with options based
13559 13566 * on the abstract ip_pkt_t.
13560 13567 *
13561 13568 * The caller has to set the source and destination address as well as
13562 13569 * ipha_length. The caller has to massage any source route and compensate
13563 13570 * for the ULP pseudo-header checksum due to the source route.
13564 13571 */
13565 13572 void
13566 13573 ip_build_hdrs_v4(uchar_t *buf, uint_t buf_len, const ip_pkt_t *ipp,
13567 13574 uint8_t protocol)
13568 13575 {
13569 13576 ipha_t *ipha = (ipha_t *)buf;
13570 13577 uint8_t *cp;
13571 13578
13572 13579 /* Initialize IPv4 header */
13573 13580 ipha->ipha_type_of_service = ipp->ipp_type_of_service;
13574 13581 ipha->ipha_length = 0; /* Caller will set later */
13575 13582 ipha->ipha_ident = 0;
13576 13583 ipha->ipha_fragment_offset_and_flags = 0;
13577 13584 ipha->ipha_ttl = ipp->ipp_unicast_hops;
13578 13585 ipha->ipha_protocol = protocol;
13579 13586 ipha->ipha_hdr_checksum = 0;
13580 13587
13581 13588 if ((ipp->ipp_fields & IPPF_ADDR) &&
13582 13589 IN6_IS_ADDR_V4MAPPED(&ipp->ipp_addr))
13583 13590 ipha->ipha_src = ipp->ipp_addr_v4;
13584 13591
13585 13592 cp = (uint8_t *)&ipha[1];
13586 13593 if (ipp->ipp_fields & IPPF_LABEL_V4) {
13587 13594 ASSERT(ipp->ipp_label_len_v4 != 0);
13588 13595 bcopy(ipp->ipp_label_v4, cp, ipp->ipp_label_len_v4);
13589 13596 cp += ipp->ipp_label_len_v4;
13590 13597 /* We need to round up here */
13591 13598 while ((uintptr_t)cp & 0x3) {
13592 13599 *cp++ = IPOPT_NOP;
13593 13600 }
13594 13601 }
13595 13602
13596 13603 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
13597 13604 ASSERT(ipp->ipp_ipv4_options_len != 0);
13598 13605 ASSERT((ipp->ipp_ipv4_options_len & 3) == 0);
13599 13606 bcopy(ipp->ipp_ipv4_options, cp, ipp->ipp_ipv4_options_len);
13600 13607 cp += ipp->ipp_ipv4_options_len;
13601 13608 }
13602 13609 ipha->ipha_version_and_hdr_length =
13603 13610 (uint8_t)((IP_VERSION << 4) + buf_len / 4);
13604 13611
13605 13612 ASSERT((int)(cp - buf) == buf_len);
13606 13613 }
13607 13614
13608 13615 /* Allocate the private structure */
13609 13616 static int
13610 13617 ip_priv_alloc(void **bufp)
13611 13618 {
13612 13619 void *buf;
13613 13620
13614 13621 if ((buf = kmem_alloc(sizeof (ip_priv_t), KM_NOSLEEP)) == NULL)
13615 13622 return (ENOMEM);
13616 13623
13617 13624 *bufp = buf;
13618 13625 return (0);
13619 13626 }
13620 13627
13621 13628 /* Function to delete the private structure */
13622 13629 void
13623 13630 ip_priv_free(void *buf)
13624 13631 {
13625 13632 ASSERT(buf != NULL);
13626 13633 kmem_free(buf, sizeof (ip_priv_t));
13627 13634 }
13628 13635
13629 13636 /*
13630 13637 * The entry point for IPPF processing.
13631 13638 * If the classifier (IPGPC_CLASSIFY) is not loaded and configured, the
13632 13639 * routine just returns.
13633 13640 *
13634 13641 * When called, ip_process generates an ipp_packet_t structure
13635 13642 * which holds the state information for this packet and invokes the
13636 13643 * the classifier (via ipp_packet_process). The classification, depending on
13637 13644 * configured filters, results in a list of actions for this packet. Invoking
13638 13645 * an action may cause the packet to be dropped, in which case we return NULL.
13639 13646 * proc indicates the callout position for
13640 13647 * this packet and ill is the interface this packet arrived on or will leave
13641 13648 * on (inbound and outbound resp.).
13642 13649 *
13643 13650 * We do the processing on the rill (mapped to the upper if ipmp), but MIB
13644 13651 * on the ill corrsponding to the destination IP address.
13645 13652 */
13646 13653 mblk_t *
13647 13654 ip_process(ip_proc_t proc, mblk_t *mp, ill_t *rill, ill_t *ill)
13648 13655 {
13649 13656 ip_priv_t *priv;
13650 13657 ipp_action_id_t aid;
13651 13658 int rc = 0;
13652 13659 ipp_packet_t *pp;
13653 13660
13654 13661 /* If the classifier is not loaded, return */
13655 13662 if ((aid = ipp_action_lookup(IPGPC_CLASSIFY)) == IPP_ACTION_INVAL) {
13656 13663 return (mp);
13657 13664 }
13658 13665
13659 13666 ASSERT(mp != NULL);
13660 13667
13661 13668 /* Allocate the packet structure */
13662 13669 rc = ipp_packet_alloc(&pp, "ip", aid);
13663 13670 if (rc != 0)
13664 13671 goto drop;
13665 13672
13666 13673 /* Allocate the private structure */
13667 13674 rc = ip_priv_alloc((void **)&priv);
13668 13675 if (rc != 0) {
13669 13676 ipp_packet_free(pp);
13670 13677 goto drop;
13671 13678 }
13672 13679 priv->proc = proc;
13673 13680 priv->ill_index = ill_get_upper_ifindex(rill);
13674 13681
13675 13682 ipp_packet_set_private(pp, priv, ip_priv_free);
13676 13683 ipp_packet_set_data(pp, mp);
13677 13684
13678 13685 /* Invoke the classifier */
13679 13686 rc = ipp_packet_process(&pp);
13680 13687 if (pp != NULL) {
13681 13688 mp = ipp_packet_get_data(pp);
13682 13689 ipp_packet_free(pp);
13683 13690 if (rc != 0)
13684 13691 goto drop;
13685 13692 return (mp);
13686 13693 } else {
13687 13694 /* No mp to trace in ip_drop_input/ip_drop_output */
13688 13695 mp = NULL;
13689 13696 }
13690 13697 drop:
13691 13698 if (proc == IPP_LOCAL_IN || proc == IPP_FWD_IN) {
13692 13699 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
13693 13700 ip_drop_input("ip_process", mp, ill);
13694 13701 } else {
13695 13702 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
13696 13703 ip_drop_output("ip_process", mp, ill);
13697 13704 }
13698 13705 freemsg(mp);
13699 13706 return (NULL);
13700 13707 }
13701 13708
13702 13709 /*
13703 13710 * Propagate a multicast group membership operation (add/drop) on
13704 13711 * all the interfaces crossed by the related multirt routes.
13705 13712 * The call is considered successful if the operation succeeds
13706 13713 * on at least one interface.
13707 13714 *
13708 13715 * This assumes that a set of IRE_HOST/RTF_MULTIRT has been created for the
13709 13716 * multicast addresses with the ire argument being the first one.
13710 13717 * We walk the bucket to find all the of those.
13711 13718 *
13712 13719 * Common to IPv4 and IPv6.
13713 13720 */
13714 13721 static int
13715 13722 ip_multirt_apply_membership(int (*fn)(conn_t *, boolean_t,
13716 13723 const in6_addr_t *, ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *),
13717 13724 ire_t *ire, conn_t *connp, boolean_t checkonly, const in6_addr_t *v6group,
13718 13725 mcast_record_t fmode, const in6_addr_t *v6src)
13719 13726 {
13720 13727 ire_t *ire_gw;
13721 13728 irb_t *irb;
13722 13729 int ifindex;
13723 13730 int error = 0;
13724 13731 int result;
13725 13732 ip_stack_t *ipst = ire->ire_ipst;
13726 13733 ipaddr_t group;
13727 13734 boolean_t isv6;
13728 13735 int match_flags;
13729 13736
13730 13737 if (IN6_IS_ADDR_V4MAPPED(v6group)) {
13731 13738 IN6_V4MAPPED_TO_IPADDR(v6group, group);
13732 13739 isv6 = B_FALSE;
13733 13740 } else {
13734 13741 isv6 = B_TRUE;
13735 13742 }
13736 13743
13737 13744 irb = ire->ire_bucket;
13738 13745 ASSERT(irb != NULL);
13739 13746
13740 13747 result = 0;
13741 13748 irb_refhold(irb);
13742 13749 for (; ire != NULL; ire = ire->ire_next) {
13743 13750 if ((ire->ire_flags & RTF_MULTIRT) == 0)
13744 13751 continue;
13745 13752
13746 13753 /* We handle -ifp routes by matching on the ill if set */
13747 13754 match_flags = MATCH_IRE_TYPE;
13748 13755 if (ire->ire_ill != NULL)
13749 13756 match_flags |= MATCH_IRE_ILL;
13750 13757
13751 13758 if (isv6) {
13752 13759 if (!IN6_ARE_ADDR_EQUAL(&ire->ire_addr_v6, v6group))
13753 13760 continue;
13754 13761
13755 13762 ire_gw = ire_ftable_lookup_v6(&ire->ire_gateway_addr_v6,
13756 13763 0, 0, IRE_INTERFACE, ire->ire_ill, ALL_ZONES, NULL,
13757 13764 match_flags, 0, ipst, NULL);
13758 13765 } else {
13759 13766 if (ire->ire_addr != group)
13760 13767 continue;
13761 13768
13762 13769 ire_gw = ire_ftable_lookup_v4(ire->ire_gateway_addr,
13763 13770 0, 0, IRE_INTERFACE, ire->ire_ill, ALL_ZONES, NULL,
13764 13771 match_flags, 0, ipst, NULL);
13765 13772 }
13766 13773 /* No interface route exists for the gateway; skip this ire. */
13767 13774 if (ire_gw == NULL)
13768 13775 continue;
13769 13776 if (ire_gw->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
13770 13777 ire_refrele(ire_gw);
13771 13778 continue;
13772 13779 }
13773 13780 ASSERT(ire_gw->ire_ill != NULL); /* IRE_INTERFACE */
13774 13781 ifindex = ire_gw->ire_ill->ill_phyint->phyint_ifindex;
13775 13782
13776 13783 /*
13777 13784 * The operation is considered a success if
13778 13785 * it succeeds at least once on any one interface.
13779 13786 */
13780 13787 error = fn(connp, checkonly, v6group, INADDR_ANY, ifindex,
13781 13788 fmode, v6src);
13782 13789 if (error == 0)
13783 13790 result = CGTP_MCAST_SUCCESS;
13784 13791
13785 13792 ire_refrele(ire_gw);
13786 13793 }
13787 13794 irb_refrele(irb);
13788 13795 /*
13789 13796 * Consider the call as successful if we succeeded on at least
13790 13797 * one interface. Otherwise, return the last encountered error.
13791 13798 */
13792 13799 return (result == CGTP_MCAST_SUCCESS ? 0 : error);
13793 13800 }
13794 13801
13795 13802 /*
13796 13803 * Return the expected CGTP hooks version number.
13797 13804 */
13798 13805 int
13799 13806 ip_cgtp_filter_supported(void)
13800 13807 {
13801 13808 return (ip_cgtp_filter_rev);
13802 13809 }
13803 13810
13804 13811 /*
13805 13812 * CGTP hooks can be registered by invoking this function.
13806 13813 * Checks that the version number matches.
13807 13814 */
13808 13815 int
13809 13816 ip_cgtp_filter_register(netstackid_t stackid, cgtp_filter_ops_t *ops)
13810 13817 {
13811 13818 netstack_t *ns;
13812 13819 ip_stack_t *ipst;
13813 13820
13814 13821 if (ops->cfo_filter_rev != CGTP_FILTER_REV)
13815 13822 return (ENOTSUP);
13816 13823
13817 13824 ns = netstack_find_by_stackid(stackid);
13818 13825 if (ns == NULL)
13819 13826 return (EINVAL);
13820 13827 ipst = ns->netstack_ip;
13821 13828 ASSERT(ipst != NULL);
13822 13829
13823 13830 if (ipst->ips_ip_cgtp_filter_ops != NULL) {
13824 13831 netstack_rele(ns);
13825 13832 return (EALREADY);
13826 13833 }
13827 13834
13828 13835 ipst->ips_ip_cgtp_filter_ops = ops;
13829 13836
13830 13837 ill_set_inputfn_all(ipst);
13831 13838
13832 13839 netstack_rele(ns);
13833 13840 return (0);
13834 13841 }
13835 13842
13836 13843 /*
13837 13844 * CGTP hooks can be unregistered by invoking this function.
13838 13845 * Returns ENXIO if there was no registration.
13839 13846 * Returns EBUSY if the ndd variable has not been turned off.
13840 13847 */
13841 13848 int
13842 13849 ip_cgtp_filter_unregister(netstackid_t stackid)
13843 13850 {
13844 13851 netstack_t *ns;
13845 13852 ip_stack_t *ipst;
13846 13853
13847 13854 ns = netstack_find_by_stackid(stackid);
13848 13855 if (ns == NULL)
13849 13856 return (EINVAL);
13850 13857 ipst = ns->netstack_ip;
13851 13858 ASSERT(ipst != NULL);
13852 13859
13853 13860 if (ipst->ips_ip_cgtp_filter) {
13854 13861 netstack_rele(ns);
13855 13862 return (EBUSY);
13856 13863 }
13857 13864
13858 13865 if (ipst->ips_ip_cgtp_filter_ops == NULL) {
13859 13866 netstack_rele(ns);
13860 13867 return (ENXIO);
13861 13868 }
13862 13869 ipst->ips_ip_cgtp_filter_ops = NULL;
13863 13870
13864 13871 ill_set_inputfn_all(ipst);
13865 13872
13866 13873 netstack_rele(ns);
13867 13874 return (0);
13868 13875 }
13869 13876
13870 13877 /*
13871 13878 * Check whether there is a CGTP filter registration.
13872 13879 * Returns non-zero if there is a registration, otherwise returns zero.
13873 13880 * Note: returns zero if bad stackid.
13874 13881 */
13875 13882 int
13876 13883 ip_cgtp_filter_is_registered(netstackid_t stackid)
13877 13884 {
13878 13885 netstack_t *ns;
13879 13886 ip_stack_t *ipst;
13880 13887 int ret;
13881 13888
13882 13889 ns = netstack_find_by_stackid(stackid);
13883 13890 if (ns == NULL)
13884 13891 return (0);
13885 13892 ipst = ns->netstack_ip;
13886 13893 ASSERT(ipst != NULL);
13887 13894
13888 13895 if (ipst->ips_ip_cgtp_filter_ops != NULL)
13889 13896 ret = 1;
13890 13897 else
13891 13898 ret = 0;
13892 13899
13893 13900 netstack_rele(ns);
13894 13901 return (ret);
13895 13902 }
13896 13903
13897 13904 static int
13898 13905 ip_squeue_switch(int val)
13899 13906 {
13900 13907 int rval;
13901 13908
13902 13909 switch (val) {
13903 13910 case IP_SQUEUE_ENTER_NODRAIN:
13904 13911 rval = SQ_NODRAIN;
13905 13912 break;
13906 13913 case IP_SQUEUE_ENTER:
13907 13914 rval = SQ_PROCESS;
13908 13915 break;
13909 13916 case IP_SQUEUE_FILL:
13910 13917 default:
13911 13918 rval = SQ_FILL;
13912 13919 break;
13913 13920 }
13914 13921 return (rval);
13915 13922 }
13916 13923
13917 13924 static void *
13918 13925 ip_kstat2_init(netstackid_t stackid, ip_stat_t *ip_statisticsp)
13919 13926 {
13920 13927 kstat_t *ksp;
13921 13928
13922 13929 ip_stat_t template = {
13923 13930 { "ip_udp_fannorm", KSTAT_DATA_UINT64 },
13924 13931 { "ip_udp_fanmb", KSTAT_DATA_UINT64 },
13925 13932 { "ip_recv_pullup", KSTAT_DATA_UINT64 },
13926 13933 { "ip_db_ref", KSTAT_DATA_UINT64 },
13927 13934 { "ip_notaligned", KSTAT_DATA_UINT64 },
13928 13935 { "ip_multimblk", KSTAT_DATA_UINT64 },
13929 13936 { "ip_opt", KSTAT_DATA_UINT64 },
13930 13937 { "ipsec_proto_ahesp", KSTAT_DATA_UINT64 },
13931 13938 { "ip_conn_flputbq", KSTAT_DATA_UINT64 },
13932 13939 { "ip_conn_walk_drain", KSTAT_DATA_UINT64 },
13933 13940 { "ip_out_sw_cksum", KSTAT_DATA_UINT64 },
13934 13941 { "ip_out_sw_cksum_bytes", KSTAT_DATA_UINT64 },
13935 13942 { "ip_in_sw_cksum", KSTAT_DATA_UINT64 },
13936 13943 { "ip_ire_reclaim_calls", KSTAT_DATA_UINT64 },
13937 13944 { "ip_ire_reclaim_deleted", KSTAT_DATA_UINT64 },
13938 13945 { "ip_nce_reclaim_calls", KSTAT_DATA_UINT64 },
13939 13946 { "ip_nce_reclaim_deleted", KSTAT_DATA_UINT64 },
13940 13947 { "ip_nce_mcast_reclaim_calls", KSTAT_DATA_UINT64 },
13941 13948 { "ip_nce_mcast_reclaim_deleted", KSTAT_DATA_UINT64 },
13942 13949 { "ip_nce_mcast_reclaim_tqfail", KSTAT_DATA_UINT64 },
13943 13950 { "ip_dce_reclaim_calls", KSTAT_DATA_UINT64 },
13944 13951 { "ip_dce_reclaim_deleted", KSTAT_DATA_UINT64 },
13945 13952 { "ip_tcp_in_full_hw_cksum_err", KSTAT_DATA_UINT64 },
13946 13953 { "ip_tcp_in_part_hw_cksum_err", KSTAT_DATA_UINT64 },
13947 13954 { "ip_tcp_in_sw_cksum_err", KSTAT_DATA_UINT64 },
13948 13955 { "ip_udp_in_full_hw_cksum_err", KSTAT_DATA_UINT64 },
13949 13956 { "ip_udp_in_part_hw_cksum_err", KSTAT_DATA_UINT64 },
13950 13957 { "ip_udp_in_sw_cksum_err", KSTAT_DATA_UINT64 },
13951 13958 { "conn_in_recvdstaddr", KSTAT_DATA_UINT64 },
13952 13959 { "conn_in_recvopts", KSTAT_DATA_UINT64 },
13953 13960 { "conn_in_recvif", KSTAT_DATA_UINT64 },
13954 13961 { "conn_in_recvslla", KSTAT_DATA_UINT64 },
13955 13962 { "conn_in_recvucred", KSTAT_DATA_UINT64 },
13956 13963 { "conn_in_recvttl", KSTAT_DATA_UINT64 },
13957 13964 { "conn_in_recvhopopts", KSTAT_DATA_UINT64 },
13958 13965 { "conn_in_recvhoplimit", KSTAT_DATA_UINT64 },
13959 13966 { "conn_in_recvdstopts", KSTAT_DATA_UINT64 },
13960 13967 { "conn_in_recvrthdrdstopts", KSTAT_DATA_UINT64 },
13961 13968 { "conn_in_recvrthdr", KSTAT_DATA_UINT64 },
13962 13969 { "conn_in_recvpktinfo", KSTAT_DATA_UINT64 },
13963 13970 { "conn_in_recvtclass", KSTAT_DATA_UINT64 },
13964 13971 { "conn_in_timestamp", KSTAT_DATA_UINT64 },
13965 13972 };
13966 13973
13967 13974 ksp = kstat_create_netstack("ip", 0, "ipstat", "net",
13968 13975 KSTAT_TYPE_NAMED, sizeof (template) / sizeof (kstat_named_t),
13969 13976 KSTAT_FLAG_VIRTUAL, stackid);
13970 13977
13971 13978 if (ksp == NULL)
13972 13979 return (NULL);
13973 13980
13974 13981 bcopy(&template, ip_statisticsp, sizeof (template));
13975 13982 ksp->ks_data = (void *)ip_statisticsp;
13976 13983 ksp->ks_private = (void *)(uintptr_t)stackid;
13977 13984
13978 13985 kstat_install(ksp);
13979 13986 return (ksp);
13980 13987 }
13981 13988
13982 13989 static void
13983 13990 ip_kstat2_fini(netstackid_t stackid, kstat_t *ksp)
13984 13991 {
13985 13992 if (ksp != NULL) {
13986 13993 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
13987 13994 kstat_delete_netstack(ksp, stackid);
13988 13995 }
13989 13996 }
13990 13997
13991 13998 static void *
13992 13999 ip_kstat_init(netstackid_t stackid, ip_stack_t *ipst)
13993 14000 {
13994 14001 kstat_t *ksp;
13995 14002
13996 14003 ip_named_kstat_t template = {
13997 14004 { "forwarding", KSTAT_DATA_UINT32, 0 },
13998 14005 { "defaultTTL", KSTAT_DATA_UINT32, 0 },
13999 14006 { "inReceives", KSTAT_DATA_UINT64, 0 },
14000 14007 { "inHdrErrors", KSTAT_DATA_UINT32, 0 },
14001 14008 { "inAddrErrors", KSTAT_DATA_UINT32, 0 },
14002 14009 { "forwDatagrams", KSTAT_DATA_UINT64, 0 },
14003 14010 { "inUnknownProtos", KSTAT_DATA_UINT32, 0 },
14004 14011 { "inDiscards", KSTAT_DATA_UINT32, 0 },
14005 14012 { "inDelivers", KSTAT_DATA_UINT64, 0 },
14006 14013 { "outRequests", KSTAT_DATA_UINT64, 0 },
14007 14014 { "outDiscards", KSTAT_DATA_UINT32, 0 },
14008 14015 { "outNoRoutes", KSTAT_DATA_UINT32, 0 },
14009 14016 { "reasmTimeout", KSTAT_DATA_UINT32, 0 },
14010 14017 { "reasmReqds", KSTAT_DATA_UINT32, 0 },
14011 14018 { "reasmOKs", KSTAT_DATA_UINT32, 0 },
14012 14019 { "reasmFails", KSTAT_DATA_UINT32, 0 },
14013 14020 { "fragOKs", KSTAT_DATA_UINT32, 0 },
14014 14021 { "fragFails", KSTAT_DATA_UINT32, 0 },
14015 14022 { "fragCreates", KSTAT_DATA_UINT32, 0 },
14016 14023 { "addrEntrySize", KSTAT_DATA_INT32, 0 },
14017 14024 { "routeEntrySize", KSTAT_DATA_INT32, 0 },
14018 14025 { "netToMediaEntrySize", KSTAT_DATA_INT32, 0 },
14019 14026 { "routingDiscards", KSTAT_DATA_UINT32, 0 },
14020 14027 { "inErrs", KSTAT_DATA_UINT32, 0 },
14021 14028 { "noPorts", KSTAT_DATA_UINT32, 0 },
14022 14029 { "inCksumErrs", KSTAT_DATA_UINT32, 0 },
14023 14030 { "reasmDuplicates", KSTAT_DATA_UINT32, 0 },
14024 14031 { "reasmPartDups", KSTAT_DATA_UINT32, 0 },
14025 14032 { "forwProhibits", KSTAT_DATA_UINT32, 0 },
14026 14033 { "udpInCksumErrs", KSTAT_DATA_UINT32, 0 },
14027 14034 { "udpInOverflows", KSTAT_DATA_UINT32, 0 },
14028 14035 { "rawipInOverflows", KSTAT_DATA_UINT32, 0 },
14029 14036 { "ipsecInSucceeded", KSTAT_DATA_UINT32, 0 },
14030 14037 { "ipsecInFailed", KSTAT_DATA_INT32, 0 },
14031 14038 { "memberEntrySize", KSTAT_DATA_INT32, 0 },
14032 14039 { "inIPv6", KSTAT_DATA_UINT32, 0 },
14033 14040 { "outIPv6", KSTAT_DATA_UINT32, 0 },
14034 14041 { "outSwitchIPv6", KSTAT_DATA_UINT32, 0 },
14035 14042 };
14036 14043
14037 14044 ksp = kstat_create_netstack("ip", 0, "ip", "mib2", KSTAT_TYPE_NAMED,
14038 14045 NUM_OF_FIELDS(ip_named_kstat_t), 0, stackid);
14039 14046 if (ksp == NULL || ksp->ks_data == NULL)
14040 14047 return (NULL);
14041 14048
14042 14049 template.forwarding.value.ui32 = WE_ARE_FORWARDING(ipst) ? 1:2;
14043 14050 template.defaultTTL.value.ui32 = (uint32_t)ipst->ips_ip_def_ttl;
14044 14051 template.reasmTimeout.value.ui32 = ipst->ips_ip_reassembly_timeout;
14045 14052 template.addrEntrySize.value.i32 = sizeof (mib2_ipAddrEntry_t);
14046 14053 template.routeEntrySize.value.i32 = sizeof (mib2_ipRouteEntry_t);
14047 14054
14048 14055 template.netToMediaEntrySize.value.i32 =
14049 14056 sizeof (mib2_ipNetToMediaEntry_t);
14050 14057
14051 14058 template.memberEntrySize.value.i32 = sizeof (ipv6_member_t);
14052 14059
14053 14060 bcopy(&template, ksp->ks_data, sizeof (template));
14054 14061 ksp->ks_update = ip_kstat_update;
14055 14062 ksp->ks_private = (void *)(uintptr_t)stackid;
14056 14063
14057 14064 kstat_install(ksp);
14058 14065 return (ksp);
14059 14066 }
14060 14067
14061 14068 static void
14062 14069 ip_kstat_fini(netstackid_t stackid, kstat_t *ksp)
14063 14070 {
14064 14071 if (ksp != NULL) {
14065 14072 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
14066 14073 kstat_delete_netstack(ksp, stackid);
14067 14074 }
14068 14075 }
14069 14076
14070 14077 static int
14071 14078 ip_kstat_update(kstat_t *kp, int rw)
14072 14079 {
14073 14080 ip_named_kstat_t *ipkp;
14074 14081 mib2_ipIfStatsEntry_t ipmib;
14075 14082 ill_walk_context_t ctx;
14076 14083 ill_t *ill;
14077 14084 netstackid_t stackid = (zoneid_t)(uintptr_t)kp->ks_private;
14078 14085 netstack_t *ns;
14079 14086 ip_stack_t *ipst;
14080 14087
14081 14088 if (kp->ks_data == NULL)
14082 14089 return (EIO);
14083 14090
14084 14091 if (rw == KSTAT_WRITE)
14085 14092 return (EACCES);
14086 14093
14087 14094 ns = netstack_find_by_stackid(stackid);
14088 14095 if (ns == NULL)
14089 14096 return (-1);
14090 14097 ipst = ns->netstack_ip;
14091 14098 if (ipst == NULL) {
14092 14099 netstack_rele(ns);
14093 14100 return (-1);
14094 14101 }
14095 14102 ipkp = (ip_named_kstat_t *)kp->ks_data;
14096 14103
14097 14104 bcopy(&ipst->ips_ip_mib, &ipmib, sizeof (ipmib));
14098 14105 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
14099 14106 ill = ILL_START_WALK_V4(&ctx, ipst);
14100 14107 for (; ill != NULL; ill = ill_next(&ctx, ill))
14101 14108 ip_mib2_add_ip_stats(&ipmib, ill->ill_ip_mib);
14102 14109 rw_exit(&ipst->ips_ill_g_lock);
14103 14110
14104 14111 ipkp->forwarding.value.ui32 = ipmib.ipIfStatsForwarding;
14105 14112 ipkp->defaultTTL.value.ui32 = ipmib.ipIfStatsDefaultTTL;
14106 14113 ipkp->inReceives.value.ui64 = ipmib.ipIfStatsHCInReceives;
14107 14114 ipkp->inHdrErrors.value.ui32 = ipmib.ipIfStatsInHdrErrors;
14108 14115 ipkp->inAddrErrors.value.ui32 = ipmib.ipIfStatsInAddrErrors;
14109 14116 ipkp->forwDatagrams.value.ui64 = ipmib.ipIfStatsHCOutForwDatagrams;
14110 14117 ipkp->inUnknownProtos.value.ui32 = ipmib.ipIfStatsInUnknownProtos;
14111 14118 ipkp->inDiscards.value.ui32 = ipmib.ipIfStatsInDiscards;
14112 14119 ipkp->inDelivers.value.ui64 = ipmib.ipIfStatsHCInDelivers;
14113 14120 ipkp->outRequests.value.ui64 = ipmib.ipIfStatsHCOutRequests;
14114 14121 ipkp->outDiscards.value.ui32 = ipmib.ipIfStatsOutDiscards;
14115 14122 ipkp->outNoRoutes.value.ui32 = ipmib.ipIfStatsOutNoRoutes;
14116 14123 ipkp->reasmTimeout.value.ui32 = ipst->ips_ip_reassembly_timeout;
14117 14124 ipkp->reasmReqds.value.ui32 = ipmib.ipIfStatsReasmReqds;
14118 14125 ipkp->reasmOKs.value.ui32 = ipmib.ipIfStatsReasmOKs;
14119 14126 ipkp->reasmFails.value.ui32 = ipmib.ipIfStatsReasmFails;
14120 14127 ipkp->fragOKs.value.ui32 = ipmib.ipIfStatsOutFragOKs;
14121 14128 ipkp->fragFails.value.ui32 = ipmib.ipIfStatsOutFragFails;
14122 14129 ipkp->fragCreates.value.ui32 = ipmib.ipIfStatsOutFragCreates;
14123 14130
14124 14131 ipkp->routingDiscards.value.ui32 = 0;
14125 14132 ipkp->inErrs.value.ui32 = ipmib.tcpIfStatsInErrs;
14126 14133 ipkp->noPorts.value.ui32 = ipmib.udpIfStatsNoPorts;
14127 14134 ipkp->inCksumErrs.value.ui32 = ipmib.ipIfStatsInCksumErrs;
14128 14135 ipkp->reasmDuplicates.value.ui32 = ipmib.ipIfStatsReasmDuplicates;
14129 14136 ipkp->reasmPartDups.value.ui32 = ipmib.ipIfStatsReasmPartDups;
14130 14137 ipkp->forwProhibits.value.ui32 = ipmib.ipIfStatsForwProhibits;
14131 14138 ipkp->udpInCksumErrs.value.ui32 = ipmib.udpIfStatsInCksumErrs;
14132 14139 ipkp->udpInOverflows.value.ui32 = ipmib.udpIfStatsInOverflows;
14133 14140 ipkp->rawipInOverflows.value.ui32 = ipmib.rawipIfStatsInOverflows;
14134 14141 ipkp->ipsecInSucceeded.value.ui32 = ipmib.ipsecIfStatsInSucceeded;
14135 14142 ipkp->ipsecInFailed.value.i32 = ipmib.ipsecIfStatsInFailed;
14136 14143
14137 14144 ipkp->inIPv6.value.ui32 = ipmib.ipIfStatsInWrongIPVersion;
14138 14145 ipkp->outIPv6.value.ui32 = ipmib.ipIfStatsOutWrongIPVersion;
14139 14146 ipkp->outSwitchIPv6.value.ui32 = ipmib.ipIfStatsOutSwitchIPVersion;
14140 14147
14141 14148 netstack_rele(ns);
14142 14149
14143 14150 return (0);
14144 14151 }
14145 14152
14146 14153 static void *
14147 14154 icmp_kstat_init(netstackid_t stackid)
14148 14155 {
14149 14156 kstat_t *ksp;
14150 14157
14151 14158 icmp_named_kstat_t template = {
14152 14159 { "inMsgs", KSTAT_DATA_UINT32 },
14153 14160 { "inErrors", KSTAT_DATA_UINT32 },
14154 14161 { "inDestUnreachs", KSTAT_DATA_UINT32 },
14155 14162 { "inTimeExcds", KSTAT_DATA_UINT32 },
14156 14163 { "inParmProbs", KSTAT_DATA_UINT32 },
14157 14164 { "inSrcQuenchs", KSTAT_DATA_UINT32 },
14158 14165 { "inRedirects", KSTAT_DATA_UINT32 },
14159 14166 { "inEchos", KSTAT_DATA_UINT32 },
14160 14167 { "inEchoReps", KSTAT_DATA_UINT32 },
14161 14168 { "inTimestamps", KSTAT_DATA_UINT32 },
14162 14169 { "inTimestampReps", KSTAT_DATA_UINT32 },
14163 14170 { "inAddrMasks", KSTAT_DATA_UINT32 },
14164 14171 { "inAddrMaskReps", KSTAT_DATA_UINT32 },
14165 14172 { "outMsgs", KSTAT_DATA_UINT32 },
14166 14173 { "outErrors", KSTAT_DATA_UINT32 },
14167 14174 { "outDestUnreachs", KSTAT_DATA_UINT32 },
14168 14175 { "outTimeExcds", KSTAT_DATA_UINT32 },
14169 14176 { "outParmProbs", KSTAT_DATA_UINT32 },
14170 14177 { "outSrcQuenchs", KSTAT_DATA_UINT32 },
14171 14178 { "outRedirects", KSTAT_DATA_UINT32 },
14172 14179 { "outEchos", KSTAT_DATA_UINT32 },
14173 14180 { "outEchoReps", KSTAT_DATA_UINT32 },
14174 14181 { "outTimestamps", KSTAT_DATA_UINT32 },
14175 14182 { "outTimestampReps", KSTAT_DATA_UINT32 },
14176 14183 { "outAddrMasks", KSTAT_DATA_UINT32 },
14177 14184 { "outAddrMaskReps", KSTAT_DATA_UINT32 },
14178 14185 { "inChksumErrs", KSTAT_DATA_UINT32 },
14179 14186 { "inUnknowns", KSTAT_DATA_UINT32 },
14180 14187 { "inFragNeeded", KSTAT_DATA_UINT32 },
14181 14188 { "outFragNeeded", KSTAT_DATA_UINT32 },
14182 14189 { "outDrops", KSTAT_DATA_UINT32 },
14183 14190 { "inOverFlows", KSTAT_DATA_UINT32 },
14184 14191 { "inBadRedirects", KSTAT_DATA_UINT32 },
14185 14192 };
14186 14193
14187 14194 ksp = kstat_create_netstack("ip", 0, "icmp", "mib2", KSTAT_TYPE_NAMED,
14188 14195 NUM_OF_FIELDS(icmp_named_kstat_t), 0, stackid);
14189 14196 if (ksp == NULL || ksp->ks_data == NULL)
14190 14197 return (NULL);
14191 14198
14192 14199 bcopy(&template, ksp->ks_data, sizeof (template));
14193 14200
14194 14201 ksp->ks_update = icmp_kstat_update;
14195 14202 ksp->ks_private = (void *)(uintptr_t)stackid;
14196 14203
14197 14204 kstat_install(ksp);
14198 14205 return (ksp);
14199 14206 }
14200 14207
14201 14208 static void
14202 14209 icmp_kstat_fini(netstackid_t stackid, kstat_t *ksp)
14203 14210 {
14204 14211 if (ksp != NULL) {
14205 14212 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
14206 14213 kstat_delete_netstack(ksp, stackid);
14207 14214 }
14208 14215 }
14209 14216
14210 14217 static int
14211 14218 icmp_kstat_update(kstat_t *kp, int rw)
14212 14219 {
14213 14220 icmp_named_kstat_t *icmpkp;
14214 14221 netstackid_t stackid = (zoneid_t)(uintptr_t)kp->ks_private;
14215 14222 netstack_t *ns;
14216 14223 ip_stack_t *ipst;
14217 14224
14218 14225 if (kp->ks_data == NULL)
14219 14226 return (EIO);
14220 14227
14221 14228 if (rw == KSTAT_WRITE)
14222 14229 return (EACCES);
14223 14230
14224 14231 ns = netstack_find_by_stackid(stackid);
14225 14232 if (ns == NULL)
14226 14233 return (-1);
14227 14234 ipst = ns->netstack_ip;
14228 14235 if (ipst == NULL) {
14229 14236 netstack_rele(ns);
14230 14237 return (-1);
14231 14238 }
14232 14239 icmpkp = (icmp_named_kstat_t *)kp->ks_data;
14233 14240
14234 14241 icmpkp->inMsgs.value.ui32 = ipst->ips_icmp_mib.icmpInMsgs;
14235 14242 icmpkp->inErrors.value.ui32 = ipst->ips_icmp_mib.icmpInErrors;
14236 14243 icmpkp->inDestUnreachs.value.ui32 =
14237 14244 ipst->ips_icmp_mib.icmpInDestUnreachs;
14238 14245 icmpkp->inTimeExcds.value.ui32 = ipst->ips_icmp_mib.icmpInTimeExcds;
14239 14246 icmpkp->inParmProbs.value.ui32 = ipst->ips_icmp_mib.icmpInParmProbs;
14240 14247 icmpkp->inSrcQuenchs.value.ui32 = ipst->ips_icmp_mib.icmpInSrcQuenchs;
14241 14248 icmpkp->inRedirects.value.ui32 = ipst->ips_icmp_mib.icmpInRedirects;
14242 14249 icmpkp->inEchos.value.ui32 = ipst->ips_icmp_mib.icmpInEchos;
14243 14250 icmpkp->inEchoReps.value.ui32 = ipst->ips_icmp_mib.icmpInEchoReps;
14244 14251 icmpkp->inTimestamps.value.ui32 = ipst->ips_icmp_mib.icmpInTimestamps;
14245 14252 icmpkp->inTimestampReps.value.ui32 =
14246 14253 ipst->ips_icmp_mib.icmpInTimestampReps;
14247 14254 icmpkp->inAddrMasks.value.ui32 = ipst->ips_icmp_mib.icmpInAddrMasks;
14248 14255 icmpkp->inAddrMaskReps.value.ui32 =
14249 14256 ipst->ips_icmp_mib.icmpInAddrMaskReps;
14250 14257 icmpkp->outMsgs.value.ui32 = ipst->ips_icmp_mib.icmpOutMsgs;
14251 14258 icmpkp->outErrors.value.ui32 = ipst->ips_icmp_mib.icmpOutErrors;
14252 14259 icmpkp->outDestUnreachs.value.ui32 =
14253 14260 ipst->ips_icmp_mib.icmpOutDestUnreachs;
14254 14261 icmpkp->outTimeExcds.value.ui32 = ipst->ips_icmp_mib.icmpOutTimeExcds;
14255 14262 icmpkp->outParmProbs.value.ui32 = ipst->ips_icmp_mib.icmpOutParmProbs;
14256 14263 icmpkp->outSrcQuenchs.value.ui32 =
14257 14264 ipst->ips_icmp_mib.icmpOutSrcQuenchs;
14258 14265 icmpkp->outRedirects.value.ui32 = ipst->ips_icmp_mib.icmpOutRedirects;
14259 14266 icmpkp->outEchos.value.ui32 = ipst->ips_icmp_mib.icmpOutEchos;
14260 14267 icmpkp->outEchoReps.value.ui32 = ipst->ips_icmp_mib.icmpOutEchoReps;
14261 14268 icmpkp->outTimestamps.value.ui32 =
14262 14269 ipst->ips_icmp_mib.icmpOutTimestamps;
14263 14270 icmpkp->outTimestampReps.value.ui32 =
14264 14271 ipst->ips_icmp_mib.icmpOutTimestampReps;
14265 14272 icmpkp->outAddrMasks.value.ui32 =
14266 14273 ipst->ips_icmp_mib.icmpOutAddrMasks;
14267 14274 icmpkp->outAddrMaskReps.value.ui32 =
14268 14275 ipst->ips_icmp_mib.icmpOutAddrMaskReps;
14269 14276 icmpkp->inCksumErrs.value.ui32 = ipst->ips_icmp_mib.icmpInCksumErrs;
14270 14277 icmpkp->inUnknowns.value.ui32 = ipst->ips_icmp_mib.icmpInUnknowns;
14271 14278 icmpkp->inFragNeeded.value.ui32 = ipst->ips_icmp_mib.icmpInFragNeeded;
14272 14279 icmpkp->outFragNeeded.value.ui32 =
14273 14280 ipst->ips_icmp_mib.icmpOutFragNeeded;
14274 14281 icmpkp->outDrops.value.ui32 = ipst->ips_icmp_mib.icmpOutDrops;
14275 14282 icmpkp->inOverflows.value.ui32 = ipst->ips_icmp_mib.icmpInOverflows;
14276 14283 icmpkp->inBadRedirects.value.ui32 =
14277 14284 ipst->ips_icmp_mib.icmpInBadRedirects;
14278 14285
14279 14286 netstack_rele(ns);
14280 14287 return (0);
14281 14288 }
14282 14289
14283 14290 /*
14284 14291 * This is the fanout function for raw socket opened for SCTP. Note
14285 14292 * that it is called after SCTP checks that there is no socket which
14286 14293 * wants a packet. Then before SCTP handles this out of the blue packet,
14287 14294 * this function is called to see if there is any raw socket for SCTP.
14288 14295 * If there is and it is bound to the correct address, the packet will
14289 14296 * be sent to that socket. Note that only one raw socket can be bound to
14290 14297 * a port. This is assured in ipcl_sctp_hash_insert();
14291 14298 */
14292 14299 void
14293 14300 ip_fanout_sctp_raw(mblk_t *mp, ipha_t *ipha, ip6_t *ip6h, uint32_t ports,
14294 14301 ip_recv_attr_t *ira)
14295 14302 {
14296 14303 conn_t *connp;
14297 14304 queue_t *rq;
14298 14305 boolean_t secure;
14299 14306 ill_t *ill = ira->ira_ill;
14300 14307 ip_stack_t *ipst = ill->ill_ipst;
14301 14308 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
14302 14309 sctp_stack_t *sctps = ipst->ips_netstack->netstack_sctp;
14303 14310 iaflags_t iraflags = ira->ira_flags;
14304 14311 ill_t *rill = ira->ira_rill;
14305 14312
14306 14313 secure = iraflags & IRAF_IPSEC_SECURE;
14307 14314
14308 14315 connp = ipcl_classify_raw(mp, IPPROTO_SCTP, ports, ipha, ip6h,
14309 14316 ira, ipst);
14310 14317 if (connp == NULL) {
14311 14318 /*
14312 14319 * Although raw sctp is not summed, OOB chunks must be.
14313 14320 * Drop the packet here if the sctp checksum failed.
14314 14321 */
14315 14322 if (iraflags & IRAF_SCTP_CSUM_ERR) {
14316 14323 SCTPS_BUMP_MIB(sctps, sctpChecksumError);
14317 14324 freemsg(mp);
14318 14325 return;
14319 14326 }
14320 14327 ira->ira_ill = ira->ira_rill = NULL;
14321 14328 sctp_ootb_input(mp, ira, ipst);
14322 14329 ira->ira_ill = ill;
14323 14330 ira->ira_rill = rill;
14324 14331 return;
14325 14332 }
14326 14333 rq = connp->conn_rq;
14327 14334 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld : !canputnext(rq)) {
14328 14335 CONN_DEC_REF(connp);
14329 14336 BUMP_MIB(ill->ill_ip_mib, rawipIfStatsInOverflows);
14330 14337 freemsg(mp);
14331 14338 return;
14332 14339 }
14333 14340 if (((iraflags & IRAF_IS_IPV4) ?
14334 14341 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
14335 14342 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
14336 14343 secure) {
14337 14344 mp = ipsec_check_inbound_policy(mp, connp, ipha,
14338 14345 ip6h, ira);
14339 14346 if (mp == NULL) {
14340 14347 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
14341 14348 /* Note that mp is NULL */
14342 14349 ip_drop_input("ipIfStatsInDiscards", mp, ill);
14343 14350 CONN_DEC_REF(connp);
14344 14351 return;
14345 14352 }
14346 14353 }
14347 14354
14348 14355 if (iraflags & IRAF_ICMP_ERROR) {
14349 14356 (connp->conn_recvicmp)(connp, mp, NULL, ira);
14350 14357 } else {
14351 14358 ill_t *rill = ira->ira_rill;
14352 14359
14353 14360 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
14354 14361 /* This is the SOCK_RAW, IPPROTO_SCTP case. */
14355 14362 ira->ira_ill = ira->ira_rill = NULL;
14356 14363 (connp->conn_recv)(connp, mp, NULL, ira);
14357 14364 ira->ira_ill = ill;
14358 14365 ira->ira_rill = rill;
14359 14366 }
14360 14367 CONN_DEC_REF(connp);
14361 14368 }
14362 14369
14363 14370 /*
14364 14371 * Free a packet that has the link-layer dl_unitdata_req_t or fast-path
14365 14372 * header before the ip payload.
14366 14373 */
14367 14374 static void
14368 14375 ip_xmit_flowctl_drop(ill_t *ill, mblk_t *mp, boolean_t is_fp_mp, int fp_mp_len)
14369 14376 {
14370 14377 int len = (mp->b_wptr - mp->b_rptr);
14371 14378 mblk_t *ip_mp;
14372 14379
14373 14380 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14374 14381 if (is_fp_mp || len != fp_mp_len) {
14375 14382 if (len > fp_mp_len) {
14376 14383 /*
14377 14384 * fastpath header and ip header in the first mblk
14378 14385 */
14379 14386 mp->b_rptr += fp_mp_len;
14380 14387 } else {
14381 14388 /*
14382 14389 * ip_xmit_attach_llhdr had to prepend an mblk to
14383 14390 * attach the fastpath header before ip header.
14384 14391 */
14385 14392 ip_mp = mp->b_cont;
14386 14393 freeb(mp);
14387 14394 mp = ip_mp;
14388 14395 mp->b_rptr += (fp_mp_len - len);
14389 14396 }
14390 14397 } else {
14391 14398 ip_mp = mp->b_cont;
14392 14399 freeb(mp);
14393 14400 mp = ip_mp;
14394 14401 }
14395 14402 ip_drop_output("ipIfStatsOutDiscards - flow ctl", mp, ill);
14396 14403 freemsg(mp);
14397 14404 }
14398 14405
14399 14406 /*
14400 14407 * Normal post fragmentation function.
14401 14408 *
14402 14409 * Send a packet using the passed in nce. This handles both IPv4 and IPv6
14403 14410 * using the same state machine.
14404 14411 *
14405 14412 * We return an error on failure. In particular we return EWOULDBLOCK
14406 14413 * when the driver flow controls. In that case this ensures that ip_wsrv runs
14407 14414 * (currently by canputnext failure resulting in backenabling from GLD.)
14408 14415 * This allows the callers of conn_ip_output() to use EWOULDBLOCK as an
14409 14416 * indication that they can flow control until ip_wsrv() tells then to restart.
14410 14417 *
14411 14418 * If the nce passed by caller is incomplete, this function
14412 14419 * queues the packet and if necessary, sends ARP request and bails.
14413 14420 * If the Neighbor Cache passed is fully resolved, we simply prepend
14414 14421 * the link-layer header to the packet, do ipsec hw acceleration
14415 14422 * work if necessary, and send the packet out on the wire.
14416 14423 */
14417 14424 /* ARGSUSED6 */
14418 14425 int
14419 14426 ip_xmit(mblk_t *mp, nce_t *nce, iaflags_t ixaflags, uint_t pkt_len,
14420 14427 uint32_t xmit_hint, zoneid_t szone, zoneid_t nolzid, uintptr_t *ixacookie)
14421 14428 {
14422 14429 queue_t *wq;
14423 14430 ill_t *ill = nce->nce_ill;
14424 14431 ip_stack_t *ipst = ill->ill_ipst;
14425 14432 uint64_t delta;
14426 14433 boolean_t isv6 = ill->ill_isv6;
14427 14434 boolean_t fp_mp;
14428 14435 ncec_t *ncec = nce->nce_common;
14429 14436 int64_t now = LBOLT_FASTPATH64;
14430 14437 boolean_t is_probe;
14431 14438
14432 14439 DTRACE_PROBE1(ip__xmit, nce_t *, nce);
14433 14440
14434 14441 ASSERT(mp != NULL);
14435 14442 ASSERT(mp->b_datap->db_type == M_DATA);
14436 14443 ASSERT(pkt_len == msgdsize(mp));
14437 14444
14438 14445 /*
14439 14446 * If we have already been here and are coming back after ARP/ND.
14440 14447 * the IXAF_NO_TRACE flag is set. We skip FW_HOOKS, DTRACE and ipobs
14441 14448 * in that case since they have seen the packet when it came here
14442 14449 * the first time.
14443 14450 */
14444 14451 if (ixaflags & IXAF_NO_TRACE)
14445 14452 goto sendit;
14446 14453
14447 14454 if (ixaflags & IXAF_IS_IPV4) {
14448 14455 ipha_t *ipha = (ipha_t *)mp->b_rptr;
14449 14456
14450 14457 ASSERT(!isv6);
14451 14458 ASSERT(pkt_len == ntohs(((ipha_t *)mp->b_rptr)->ipha_length));
14452 14459 if (HOOKS4_INTERESTED_PHYSICAL_OUT(ipst) &&
14453 14460 !(ixaflags & IXAF_NO_PFHOOK)) {
14454 14461 int error;
14455 14462
14456 14463 FW_HOOKS(ipst->ips_ip4_physical_out_event,
14457 14464 ipst->ips_ipv4firewall_physical_out,
14458 14465 NULL, ill, ipha, mp, mp, 0, ipst, error);
14459 14466 DTRACE_PROBE1(ip4__physical__out__end,
14460 14467 mblk_t *, mp);
14461 14468 if (mp == NULL)
14462 14469 return (error);
14463 14470
14464 14471 /* The length could have changed */
14465 14472 pkt_len = msgdsize(mp);
14466 14473 }
14467 14474 if (ipst->ips_ip4_observe.he_interested) {
14468 14475 /*
14469 14476 * Note that for TX the zoneid is the sending
14470 14477 * zone, whether or not MLP is in play.
14471 14478 * Since the szone argument is the IP zoneid (i.e.,
14472 14479 * zero for exclusive-IP zones) and ipobs wants
14473 14480 * the system zoneid, we map it here.
14474 14481 */
14475 14482 szone = IP_REAL_ZONEID(szone, ipst);
14476 14483
14477 14484 /*
14478 14485 * On the outbound path the destination zone will be
14479 14486 * unknown as we're sending this packet out on the
14480 14487 * wire.
14481 14488 */
14482 14489 ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone, ALL_ZONES,
14483 14490 ill, ipst);
14484 14491 }
14485 14492 DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
14486 14493 void_ip_t *, ipha, __dtrace_ipsr_ill_t *, ill,
14487 14494 ipha_t *, ipha, ip6_t *, NULL, int, 0);
14488 14495 } else {
14489 14496 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
14490 14497
14491 14498 ASSERT(isv6);
14492 14499 ASSERT(pkt_len ==
14493 14500 ntohs(((ip6_t *)mp->b_rptr)->ip6_plen) + IPV6_HDR_LEN);
14494 14501 if (HOOKS6_INTERESTED_PHYSICAL_OUT(ipst) &&
14495 14502 !(ixaflags & IXAF_NO_PFHOOK)) {
14496 14503 int error;
14497 14504
14498 14505 FW_HOOKS6(ipst->ips_ip6_physical_out_event,
14499 14506 ipst->ips_ipv6firewall_physical_out,
14500 14507 NULL, ill, ip6h, mp, mp, 0, ipst, error);
14501 14508 DTRACE_PROBE1(ip6__physical__out__end,
14502 14509 mblk_t *, mp);
14503 14510 if (mp == NULL)
14504 14511 return (error);
14505 14512
14506 14513 /* The length could have changed */
14507 14514 pkt_len = msgdsize(mp);
14508 14515 }
14509 14516 if (ipst->ips_ip6_observe.he_interested) {
14510 14517 /* See above */
14511 14518 szone = IP_REAL_ZONEID(szone, ipst);
14512 14519
14513 14520 ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone, ALL_ZONES,
14514 14521 ill, ipst);
14515 14522 }
14516 14523 DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
14517 14524 void_ip_t *, ip6h, __dtrace_ipsr_ill_t *, ill,
14518 14525 ipha_t *, NULL, ip6_t *, ip6h, int, 0);
14519 14526 }
14520 14527
14521 14528 sendit:
14522 14529 /*
14523 14530 * We check the state without a lock because the state can never
14524 14531 * move "backwards" to initial or incomplete.
14525 14532 */
14526 14533 switch (ncec->ncec_state) {
14527 14534 case ND_REACHABLE:
14528 14535 case ND_STALE:
14529 14536 case ND_DELAY:
14530 14537 case ND_PROBE:
14531 14538 mp = ip_xmit_attach_llhdr(mp, nce);
14532 14539 if (mp == NULL) {
14533 14540 /*
14534 14541 * ip_xmit_attach_llhdr has increased
14535 14542 * ipIfStatsOutDiscards and called ip_drop_output()
14536 14543 */
14537 14544 return (ENOBUFS);
14538 14545 }
14539 14546 /*
14540 14547 * check if nce_fastpath completed and we tagged on a
14541 14548 * copy of nce_fp_mp in ip_xmit_attach_llhdr().
14542 14549 */
14543 14550 fp_mp = (mp->b_datap->db_type == M_DATA);
14544 14551
14545 14552 if (fp_mp &&
14546 14553 (ill->ill_capabilities & ILL_CAPAB_DLD_DIRECT)) {
14547 14554 ill_dld_direct_t *idd;
14548 14555
14549 14556 idd = &ill->ill_dld_capab->idc_direct;
14550 14557 /*
14551 14558 * Send the packet directly to DLD, where it
14552 14559 * may be queued depending on the availability
14553 14560 * of transmit resources at the media layer.
14554 14561 * Return value should be taken into
14555 14562 * account and flow control the TCP.
14556 14563 */
14557 14564 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
14558 14565 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
14559 14566 pkt_len);
14560 14567
14561 14568 if (ixaflags & IXAF_NO_DEV_FLOW_CTL) {
14562 14569 (void) idd->idd_tx_df(idd->idd_tx_dh, mp,
14563 14570 (uintptr_t)xmit_hint, IP_DROP_ON_NO_DESC);
14564 14571 } else {
14565 14572 uintptr_t cookie;
14566 14573
14567 14574 if ((cookie = idd->idd_tx_df(idd->idd_tx_dh,
14568 14575 mp, (uintptr_t)xmit_hint, 0)) != 0) {
14569 14576 if (ixacookie != NULL)
14570 14577 *ixacookie = cookie;
14571 14578 return (EWOULDBLOCK);
14572 14579 }
14573 14580 }
14574 14581 } else {
14575 14582 wq = ill->ill_wq;
14576 14583
14577 14584 if (!(ixaflags & IXAF_NO_DEV_FLOW_CTL) &&
14578 14585 !canputnext(wq)) {
14579 14586 if (ixacookie != NULL)
14580 14587 *ixacookie = 0;
14581 14588 ip_xmit_flowctl_drop(ill, mp, fp_mp,
14582 14589 nce->nce_fp_mp != NULL ?
14583 14590 MBLKL(nce->nce_fp_mp) : 0);
14584 14591 return (EWOULDBLOCK);
14585 14592 }
14586 14593 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
14587 14594 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
14588 14595 pkt_len);
14589 14596 putnext(wq, mp);
14590 14597 }
14591 14598
14592 14599 /*
14593 14600 * The rest of this function implements Neighbor Unreachability
14594 14601 * detection. Determine if the ncec is eligible for NUD.
14595 14602 */
14596 14603 if (ncec->ncec_flags & NCE_F_NONUD)
14597 14604 return (0);
14598 14605
14599 14606 ASSERT(ncec->ncec_state != ND_INCOMPLETE);
14600 14607
14601 14608 /*
14602 14609 * Check for upper layer advice
14603 14610 */
14604 14611 if (ixaflags & IXAF_REACH_CONF) {
14605 14612 timeout_id_t tid;
14606 14613
14607 14614 /*
14608 14615 * It should be o.k. to check the state without
14609 14616 * a lock here, at most we lose an advice.
14610 14617 */
14611 14618 ncec->ncec_last = TICK_TO_MSEC(now);
14612 14619 if (ncec->ncec_state != ND_REACHABLE) {
14613 14620 mutex_enter(&ncec->ncec_lock);
14614 14621 ncec->ncec_state = ND_REACHABLE;
14615 14622 tid = ncec->ncec_timeout_id;
14616 14623 ncec->ncec_timeout_id = 0;
14617 14624 mutex_exit(&ncec->ncec_lock);
14618 14625 (void) untimeout(tid);
14619 14626 if (ip_debug > 2) {
14620 14627 /* ip1dbg */
14621 14628 pr_addr_dbg("ip_xmit: state"
14622 14629 " for %s changed to"
14623 14630 " REACHABLE\n", AF_INET6,
14624 14631 &ncec->ncec_addr);
14625 14632 }
14626 14633 }
14627 14634 return (0);
14628 14635 }
14629 14636
14630 14637 delta = TICK_TO_MSEC(now) - ncec->ncec_last;
14631 14638 ip1dbg(("ip_xmit: delta = %" PRId64
14632 14639 " ill_reachable_time = %d \n", delta,
14633 14640 ill->ill_reachable_time));
14634 14641 if (delta > (uint64_t)ill->ill_reachable_time) {
14635 14642 mutex_enter(&ncec->ncec_lock);
14636 14643 switch (ncec->ncec_state) {
14637 14644 case ND_REACHABLE:
14638 14645 ASSERT((ncec->ncec_flags & NCE_F_NONUD) == 0);
14639 14646 /* FALLTHROUGH */
14640 14647 case ND_STALE:
14641 14648 /*
14642 14649 * ND_REACHABLE is identical to
14643 14650 * ND_STALE in this specific case. If
14644 14651 * reachable time has expired for this
14645 14652 * neighbor (delta is greater than
14646 14653 * reachable time), conceptually, the
14647 14654 * neighbor cache is no longer in
14648 14655 * REACHABLE state, but already in
14649 14656 * STALE state. So the correct
14650 14657 * transition here is to ND_DELAY.
14651 14658 */
14652 14659 ncec->ncec_state = ND_DELAY;
14653 14660 mutex_exit(&ncec->ncec_lock);
14654 14661 nce_restart_timer(ncec,
14655 14662 ipst->ips_delay_first_probe_time);
14656 14663 if (ip_debug > 3) {
14657 14664 /* ip2dbg */
14658 14665 pr_addr_dbg("ip_xmit: state"
14659 14666 " for %s changed to"
14660 14667 " DELAY\n", AF_INET6,
14661 14668 &ncec->ncec_addr);
14662 14669 }
14663 14670 break;
14664 14671 case ND_DELAY:
14665 14672 case ND_PROBE:
14666 14673 mutex_exit(&ncec->ncec_lock);
14667 14674 /* Timers have already started */
14668 14675 break;
14669 14676 case ND_UNREACHABLE:
14670 14677 /*
14671 14678 * nce_timer has detected that this ncec
14672 14679 * is unreachable and initiated deleting
14673 14680 * this ncec.
14674 14681 * This is a harmless race where we found the
14675 14682 * ncec before it was deleted and have
14676 14683 * just sent out a packet using this
14677 14684 * unreachable ncec.
14678 14685 */
14679 14686 mutex_exit(&ncec->ncec_lock);
14680 14687 break;
14681 14688 default:
14682 14689 ASSERT(0);
14683 14690 mutex_exit(&ncec->ncec_lock);
14684 14691 }
14685 14692 }
14686 14693 return (0);
14687 14694
14688 14695 case ND_INCOMPLETE:
14689 14696 /*
14690 14697 * the state could have changed since we didn't hold the lock.
14691 14698 * Re-verify state under lock.
14692 14699 */
14693 14700 is_probe = ipmp_packet_is_probe(mp, nce->nce_ill);
14694 14701 mutex_enter(&ncec->ncec_lock);
14695 14702 if (NCE_ISREACHABLE(ncec)) {
14696 14703 mutex_exit(&ncec->ncec_lock);
14697 14704 goto sendit;
14698 14705 }
14699 14706 /* queue the packet */
14700 14707 nce_queue_mp(ncec, mp, is_probe);
14701 14708 mutex_exit(&ncec->ncec_lock);
14702 14709 DTRACE_PROBE2(ip__xmit__incomplete,
14703 14710 (ncec_t *), ncec, (mblk_t *), mp);
14704 14711 return (0);
14705 14712
14706 14713 case ND_INITIAL:
14707 14714 /*
14708 14715 * State could have changed since we didn't hold the lock, so
14709 14716 * re-verify state.
14710 14717 */
14711 14718 is_probe = ipmp_packet_is_probe(mp, nce->nce_ill);
14712 14719 mutex_enter(&ncec->ncec_lock);
14713 14720 if (NCE_ISREACHABLE(ncec)) {
14714 14721 mutex_exit(&ncec->ncec_lock);
14715 14722 goto sendit;
14716 14723 }
14717 14724 nce_queue_mp(ncec, mp, is_probe);
14718 14725 if (ncec->ncec_state == ND_INITIAL) {
14719 14726 ncec->ncec_state = ND_INCOMPLETE;
14720 14727 mutex_exit(&ncec->ncec_lock);
14721 14728 /*
14722 14729 * figure out the source we want to use
14723 14730 * and resolve it.
14724 14731 */
14725 14732 ip_ndp_resolve(ncec);
14726 14733 } else {
14727 14734 mutex_exit(&ncec->ncec_lock);
14728 14735 }
14729 14736 return (0);
14730 14737
14731 14738 case ND_UNREACHABLE:
14732 14739 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14733 14740 ip_drop_output("ipIfStatsOutDiscards - ND_UNREACHABLE",
14734 14741 mp, ill);
14735 14742 freemsg(mp);
14736 14743 return (0);
14737 14744
14738 14745 default:
14739 14746 ASSERT(0);
14740 14747 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14741 14748 ip_drop_output("ipIfStatsOutDiscards - ND_other",
14742 14749 mp, ill);
14743 14750 freemsg(mp);
14744 14751 return (ENETUNREACH);
14745 14752 }
14746 14753 }
14747 14754
14748 14755 /*
14749 14756 * Return B_TRUE if the buffers differ in length or content.
14750 14757 * This is used for comparing extension header buffers.
14751 14758 * Note that an extension header would be declared different
14752 14759 * even if all that changed was the next header value in that header i.e.
14753 14760 * what really changed is the next extension header.
14754 14761 */
14755 14762 boolean_t
14756 14763 ip_cmpbuf(const void *abuf, uint_t alen, boolean_t b_valid, const void *bbuf,
14757 14764 uint_t blen)
14758 14765 {
14759 14766 if (!b_valid)
14760 14767 blen = 0;
14761 14768
14762 14769 if (alen != blen)
14763 14770 return (B_TRUE);
14764 14771 if (alen == 0)
14765 14772 return (B_FALSE); /* Both zero length */
14766 14773 return (bcmp(abuf, bbuf, alen));
14767 14774 }
14768 14775
14769 14776 /*
14770 14777 * Preallocate memory for ip_savebuf(). Returns B_TRUE if ok.
14771 14778 * Return B_FALSE if memory allocation fails - don't change any state!
14772 14779 */
14773 14780 boolean_t
14774 14781 ip_allocbuf(void **dstp, uint_t *dstlenp, boolean_t src_valid,
14775 14782 const void *src, uint_t srclen)
14776 14783 {
14777 14784 void *dst;
14778 14785
14779 14786 if (!src_valid)
14780 14787 srclen = 0;
14781 14788
14782 14789 ASSERT(*dstlenp == 0);
14783 14790 if (src != NULL && srclen != 0) {
14784 14791 dst = mi_alloc(srclen, BPRI_MED);
14785 14792 if (dst == NULL)
14786 14793 return (B_FALSE);
14787 14794 } else {
14788 14795 dst = NULL;
14789 14796 }
14790 14797 if (*dstp != NULL)
14791 14798 mi_free(*dstp);
14792 14799 *dstp = dst;
14793 14800 *dstlenp = dst == NULL ? 0 : srclen;
14794 14801 return (B_TRUE);
14795 14802 }
14796 14803
14797 14804 /*
14798 14805 * Replace what is in *dst, *dstlen with the source.
14799 14806 * Assumes ip_allocbuf has already been called.
14800 14807 */
14801 14808 void
14802 14809 ip_savebuf(void **dstp, uint_t *dstlenp, boolean_t src_valid,
14803 14810 const void *src, uint_t srclen)
14804 14811 {
14805 14812 if (!src_valid)
14806 14813 srclen = 0;
14807 14814
14808 14815 ASSERT(*dstlenp == srclen);
14809 14816 if (src != NULL && srclen != 0)
14810 14817 bcopy(src, *dstp, srclen);
14811 14818 }
14812 14819
14813 14820 /*
14814 14821 * Free the storage pointed to by the members of an ip_pkt_t.
14815 14822 */
14816 14823 void
14817 14824 ip_pkt_free(ip_pkt_t *ipp)
14818 14825 {
14819 14826 uint_t fields = ipp->ipp_fields;
14820 14827
14821 14828 if (fields & IPPF_HOPOPTS) {
14822 14829 kmem_free(ipp->ipp_hopopts, ipp->ipp_hopoptslen);
14823 14830 ipp->ipp_hopopts = NULL;
14824 14831 ipp->ipp_hopoptslen = 0;
14825 14832 }
14826 14833 if (fields & IPPF_RTHDRDSTOPTS) {
14827 14834 kmem_free(ipp->ipp_rthdrdstopts, ipp->ipp_rthdrdstoptslen);
14828 14835 ipp->ipp_rthdrdstopts = NULL;
14829 14836 ipp->ipp_rthdrdstoptslen = 0;
14830 14837 }
14831 14838 if (fields & IPPF_DSTOPTS) {
14832 14839 kmem_free(ipp->ipp_dstopts, ipp->ipp_dstoptslen);
14833 14840 ipp->ipp_dstopts = NULL;
14834 14841 ipp->ipp_dstoptslen = 0;
14835 14842 }
14836 14843 if (fields & IPPF_RTHDR) {
14837 14844 kmem_free(ipp->ipp_rthdr, ipp->ipp_rthdrlen);
14838 14845 ipp->ipp_rthdr = NULL;
14839 14846 ipp->ipp_rthdrlen = 0;
14840 14847 }
14841 14848 if (fields & IPPF_IPV4_OPTIONS) {
14842 14849 kmem_free(ipp->ipp_ipv4_options, ipp->ipp_ipv4_options_len);
14843 14850 ipp->ipp_ipv4_options = NULL;
14844 14851 ipp->ipp_ipv4_options_len = 0;
14845 14852 }
14846 14853 if (fields & IPPF_LABEL_V4) {
14847 14854 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
14848 14855 ipp->ipp_label_v4 = NULL;
14849 14856 ipp->ipp_label_len_v4 = 0;
14850 14857 }
14851 14858 if (fields & IPPF_LABEL_V6) {
14852 14859 kmem_free(ipp->ipp_label_v6, ipp->ipp_label_len_v6);
14853 14860 ipp->ipp_label_v6 = NULL;
14854 14861 ipp->ipp_label_len_v6 = 0;
14855 14862 }
14856 14863 ipp->ipp_fields &= ~(IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14857 14864 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6);
14858 14865 }
14859 14866
14860 14867 /*
14861 14868 * Copy from src to dst and allocate as needed.
14862 14869 * Returns zero or ENOMEM.
14863 14870 *
14864 14871 * The caller must initialize dst to zero.
14865 14872 */
14866 14873 int
14867 14874 ip_pkt_copy(ip_pkt_t *src, ip_pkt_t *dst, int kmflag)
14868 14875 {
14869 14876 uint_t fields = src->ipp_fields;
14870 14877
14871 14878 /* Start with fields that don't require memory allocation */
14872 14879 dst->ipp_fields = fields &
14873 14880 ~(IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14874 14881 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6);
14875 14882
14876 14883 dst->ipp_addr = src->ipp_addr;
14877 14884 dst->ipp_unicast_hops = src->ipp_unicast_hops;
14878 14885 dst->ipp_hoplimit = src->ipp_hoplimit;
14879 14886 dst->ipp_tclass = src->ipp_tclass;
14880 14887 dst->ipp_type_of_service = src->ipp_type_of_service;
14881 14888
14882 14889 if (!(fields & (IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14883 14890 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6)))
14884 14891 return (0);
14885 14892
14886 14893 if (fields & IPPF_HOPOPTS) {
14887 14894 dst->ipp_hopopts = kmem_alloc(src->ipp_hopoptslen, kmflag);
14888 14895 if (dst->ipp_hopopts == NULL) {
14889 14896 ip_pkt_free(dst);
14890 14897 return (ENOMEM);
14891 14898 }
14892 14899 dst->ipp_fields |= IPPF_HOPOPTS;
14893 14900 bcopy(src->ipp_hopopts, dst->ipp_hopopts,
14894 14901 src->ipp_hopoptslen);
14895 14902 dst->ipp_hopoptslen = src->ipp_hopoptslen;
14896 14903 }
14897 14904 if (fields & IPPF_RTHDRDSTOPTS) {
14898 14905 dst->ipp_rthdrdstopts = kmem_alloc(src->ipp_rthdrdstoptslen,
14899 14906 kmflag);
14900 14907 if (dst->ipp_rthdrdstopts == NULL) {
14901 14908 ip_pkt_free(dst);
14902 14909 return (ENOMEM);
14903 14910 }
14904 14911 dst->ipp_fields |= IPPF_RTHDRDSTOPTS;
14905 14912 bcopy(src->ipp_rthdrdstopts, dst->ipp_rthdrdstopts,
14906 14913 src->ipp_rthdrdstoptslen);
14907 14914 dst->ipp_rthdrdstoptslen = src->ipp_rthdrdstoptslen;
14908 14915 }
14909 14916 if (fields & IPPF_DSTOPTS) {
14910 14917 dst->ipp_dstopts = kmem_alloc(src->ipp_dstoptslen, kmflag);
14911 14918 if (dst->ipp_dstopts == NULL) {
14912 14919 ip_pkt_free(dst);
14913 14920 return (ENOMEM);
14914 14921 }
14915 14922 dst->ipp_fields |= IPPF_DSTOPTS;
14916 14923 bcopy(src->ipp_dstopts, dst->ipp_dstopts,
14917 14924 src->ipp_dstoptslen);
14918 14925 dst->ipp_dstoptslen = src->ipp_dstoptslen;
14919 14926 }
14920 14927 if (fields & IPPF_RTHDR) {
14921 14928 dst->ipp_rthdr = kmem_alloc(src->ipp_rthdrlen, kmflag);
14922 14929 if (dst->ipp_rthdr == NULL) {
14923 14930 ip_pkt_free(dst);
14924 14931 return (ENOMEM);
14925 14932 }
14926 14933 dst->ipp_fields |= IPPF_RTHDR;
14927 14934 bcopy(src->ipp_rthdr, dst->ipp_rthdr,
14928 14935 src->ipp_rthdrlen);
14929 14936 dst->ipp_rthdrlen = src->ipp_rthdrlen;
14930 14937 }
14931 14938 if (fields & IPPF_IPV4_OPTIONS) {
14932 14939 dst->ipp_ipv4_options = kmem_alloc(src->ipp_ipv4_options_len,
14933 14940 kmflag);
14934 14941 if (dst->ipp_ipv4_options == NULL) {
14935 14942 ip_pkt_free(dst);
14936 14943 return (ENOMEM);
14937 14944 }
14938 14945 dst->ipp_fields |= IPPF_IPV4_OPTIONS;
14939 14946 bcopy(src->ipp_ipv4_options, dst->ipp_ipv4_options,
14940 14947 src->ipp_ipv4_options_len);
14941 14948 dst->ipp_ipv4_options_len = src->ipp_ipv4_options_len;
14942 14949 }
14943 14950 if (fields & IPPF_LABEL_V4) {
14944 14951 dst->ipp_label_v4 = kmem_alloc(src->ipp_label_len_v4, kmflag);
14945 14952 if (dst->ipp_label_v4 == NULL) {
14946 14953 ip_pkt_free(dst);
14947 14954 return (ENOMEM);
14948 14955 }
14949 14956 dst->ipp_fields |= IPPF_LABEL_V4;
14950 14957 bcopy(src->ipp_label_v4, dst->ipp_label_v4,
14951 14958 src->ipp_label_len_v4);
14952 14959 dst->ipp_label_len_v4 = src->ipp_label_len_v4;
14953 14960 }
14954 14961 if (fields & IPPF_LABEL_V6) {
14955 14962 dst->ipp_label_v6 = kmem_alloc(src->ipp_label_len_v6, kmflag);
14956 14963 if (dst->ipp_label_v6 == NULL) {
14957 14964 ip_pkt_free(dst);
14958 14965 return (ENOMEM);
14959 14966 }
14960 14967 dst->ipp_fields |= IPPF_LABEL_V6;
14961 14968 bcopy(src->ipp_label_v6, dst->ipp_label_v6,
14962 14969 src->ipp_label_len_v6);
14963 14970 dst->ipp_label_len_v6 = src->ipp_label_len_v6;
14964 14971 }
14965 14972 if (fields & IPPF_FRAGHDR) {
14966 14973 dst->ipp_fraghdr = kmem_alloc(src->ipp_fraghdrlen, kmflag);
14967 14974 if (dst->ipp_fraghdr == NULL) {
14968 14975 ip_pkt_free(dst);
14969 14976 return (ENOMEM);
14970 14977 }
14971 14978 dst->ipp_fields |= IPPF_FRAGHDR;
14972 14979 bcopy(src->ipp_fraghdr, dst->ipp_fraghdr,
14973 14980 src->ipp_fraghdrlen);
14974 14981 dst->ipp_fraghdrlen = src->ipp_fraghdrlen;
14975 14982 }
14976 14983 return (0);
14977 14984 }
14978 14985
14979 14986 /*
14980 14987 * Returns INADDR_ANY if no source route
14981 14988 */
14982 14989 ipaddr_t
14983 14990 ip_pkt_source_route_v4(const ip_pkt_t *ipp)
14984 14991 {
14985 14992 ipaddr_t nexthop = INADDR_ANY;
14986 14993 ipoptp_t opts;
14987 14994 uchar_t *opt;
14988 14995 uint8_t optval;
14989 14996 uint8_t optlen;
14990 14997 uint32_t totallen;
14991 14998
14992 14999 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
14993 15000 return (INADDR_ANY);
14994 15001
14995 15002 totallen = ipp->ipp_ipv4_options_len;
14996 15003 if (totallen & 0x3)
14997 15004 return (INADDR_ANY);
14998 15005
14999 15006 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
15000 15007 optval != IPOPT_EOL;
15001 15008 optval = ipoptp_next(&opts)) {
15002 15009 opt = opts.ipoptp_cur;
15003 15010 switch (optval) {
15004 15011 uint8_t off;
15005 15012 case IPOPT_SSRR:
15006 15013 case IPOPT_LSRR:
15007 15014 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
15008 15015 break;
15009 15016 }
15010 15017 optlen = opts.ipoptp_len;
15011 15018 off = opt[IPOPT_OFFSET];
15012 15019 off--;
15013 15020 if (optlen < IP_ADDR_LEN ||
15014 15021 off > optlen - IP_ADDR_LEN) {
15015 15022 /* End of source route */
15016 15023 break;
15017 15024 }
15018 15025 bcopy((char *)opt + off, &nexthop, IP_ADDR_LEN);
15019 15026 if (nexthop == htonl(INADDR_LOOPBACK)) {
15020 15027 /* Ignore */
15021 15028 nexthop = INADDR_ANY;
15022 15029 break;
15023 15030 }
15024 15031 break;
15025 15032 }
15026 15033 }
15027 15034 return (nexthop);
15028 15035 }
15029 15036
15030 15037 /*
15031 15038 * Reverse a source route.
15032 15039 */
15033 15040 void
15034 15041 ip_pkt_source_route_reverse_v4(ip_pkt_t *ipp)
15035 15042 {
15036 15043 ipaddr_t tmp;
15037 15044 ipoptp_t opts;
15038 15045 uchar_t *opt;
15039 15046 uint8_t optval;
15040 15047 uint32_t totallen;
15041 15048
15042 15049 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
15043 15050 return;
15044 15051
15045 15052 totallen = ipp->ipp_ipv4_options_len;
15046 15053 if (totallen & 0x3)
15047 15054 return;
15048 15055
15049 15056 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
15050 15057 optval != IPOPT_EOL;
15051 15058 optval = ipoptp_next(&opts)) {
15052 15059 uint8_t off1, off2;
15053 15060
15054 15061 opt = opts.ipoptp_cur;
15055 15062 switch (optval) {
15056 15063 case IPOPT_SSRR:
15057 15064 case IPOPT_LSRR:
15058 15065 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
15059 15066 break;
15060 15067 }
15061 15068 off1 = IPOPT_MINOFF_SR - 1;
15062 15069 off2 = opt[IPOPT_OFFSET] - IP_ADDR_LEN - 1;
15063 15070 while (off2 > off1) {
15064 15071 bcopy(opt + off2, &tmp, IP_ADDR_LEN);
15065 15072 bcopy(opt + off1, opt + off2, IP_ADDR_LEN);
15066 15073 bcopy(&tmp, opt + off2, IP_ADDR_LEN);
15067 15074 off2 -= IP_ADDR_LEN;
15068 15075 off1 += IP_ADDR_LEN;
15069 15076 }
15070 15077 opt[IPOPT_OFFSET] = IPOPT_MINOFF_SR;
15071 15078 break;
15072 15079 }
15073 15080 }
15074 15081 }
15075 15082
15076 15083 /*
15077 15084 * Returns NULL if no routing header
15078 15085 */
15079 15086 in6_addr_t *
15080 15087 ip_pkt_source_route_v6(const ip_pkt_t *ipp)
15081 15088 {
15082 15089 in6_addr_t *nexthop = NULL;
15083 15090 ip6_rthdr0_t *rthdr;
15084 15091
15085 15092 if (!(ipp->ipp_fields & IPPF_RTHDR))
15086 15093 return (NULL);
15087 15094
15088 15095 rthdr = (ip6_rthdr0_t *)ipp->ipp_rthdr;
15089 15096 if (rthdr->ip6r0_segleft == 0)
15090 15097 return (NULL);
15091 15098
15092 15099 nexthop = (in6_addr_t *)((char *)rthdr + sizeof (*rthdr));
15093 15100 return (nexthop);
15094 15101 }
15095 15102
15096 15103 zoneid_t
15097 15104 ip_get_zoneid_v4(ipaddr_t addr, mblk_t *mp, ip_recv_attr_t *ira,
15098 15105 zoneid_t lookup_zoneid)
15099 15106 {
15100 15107 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
15101 15108 ire_t *ire;
15102 15109 int ire_flags = MATCH_IRE_TYPE;
15103 15110 zoneid_t zoneid = ALL_ZONES;
15104 15111
15105 15112 if (is_system_labeled() && !tsol_can_accept_raw(mp, ira, B_FALSE))
15106 15113 return (ALL_ZONES);
15107 15114
15108 15115 if (lookup_zoneid != ALL_ZONES)
15109 15116 ire_flags |= MATCH_IRE_ZONEONLY;
15110 15117 ire = ire_ftable_lookup_v4(addr, 0, 0, IRE_LOCAL | IRE_LOOPBACK,
15111 15118 NULL, lookup_zoneid, NULL, ire_flags, 0, ipst, NULL);
15112 15119 if (ire != NULL) {
15113 15120 zoneid = IP_REAL_ZONEID(ire->ire_zoneid, ipst);
15114 15121 ire_refrele(ire);
15115 15122 }
15116 15123 return (zoneid);
15117 15124 }
15118 15125
15119 15126 zoneid_t
15120 15127 ip_get_zoneid_v6(in6_addr_t *addr, mblk_t *mp, const ill_t *ill,
15121 15128 ip_recv_attr_t *ira, zoneid_t lookup_zoneid)
15122 15129 {
15123 15130 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
15124 15131 ire_t *ire;
15125 15132 int ire_flags = MATCH_IRE_TYPE;
15126 15133 zoneid_t zoneid = ALL_ZONES;
15127 15134
15128 15135 if (is_system_labeled() && !tsol_can_accept_raw(mp, ira, B_FALSE))
15129 15136 return (ALL_ZONES);
15130 15137
15131 15138 if (IN6_IS_ADDR_LINKLOCAL(addr))
15132 15139 ire_flags |= MATCH_IRE_ILL;
15133 15140
15134 15141 if (lookup_zoneid != ALL_ZONES)
15135 15142 ire_flags |= MATCH_IRE_ZONEONLY;
15136 15143 ire = ire_ftable_lookup_v6(addr, NULL, NULL, IRE_LOCAL | IRE_LOOPBACK,
15137 15144 ill, lookup_zoneid, NULL, ire_flags, 0, ipst, NULL);
15138 15145 if (ire != NULL) {
15139 15146 zoneid = IP_REAL_ZONEID(ire->ire_zoneid, ipst);
15140 15147 ire_refrele(ire);
15141 15148 }
15142 15149 return (zoneid);
15143 15150 }
15144 15151
15145 15152 /*
15146 15153 * IP obserability hook support functions.
15147 15154 */
15148 15155 static void
15149 15156 ipobs_init(ip_stack_t *ipst)
15150 15157 {
15151 15158 netid_t id;
15152 15159
15153 15160 id = net_getnetidbynetstackid(ipst->ips_netstack->netstack_stackid);
15154 15161
15155 15162 ipst->ips_ip4_observe_pr = net_protocol_lookup(id, NHF_INET);
15156 15163 VERIFY(ipst->ips_ip4_observe_pr != NULL);
15157 15164
15158 15165 ipst->ips_ip6_observe_pr = net_protocol_lookup(id, NHF_INET6);
15159 15166 VERIFY(ipst->ips_ip6_observe_pr != NULL);
15160 15167 }
15161 15168
15162 15169 static void
15163 15170 ipobs_fini(ip_stack_t *ipst)
15164 15171 {
15165 15172
15166 15173 VERIFY(net_protocol_release(ipst->ips_ip4_observe_pr) == 0);
15167 15174 VERIFY(net_protocol_release(ipst->ips_ip6_observe_pr) == 0);
15168 15175 }
15169 15176
15170 15177 /*
15171 15178 * hook_pkt_observe_t is composed in network byte order so that the
15172 15179 * entire mblk_t chain handed into hook_run can be used as-is.
15173 15180 * The caveat is that use of the fields, such as the zone fields,
15174 15181 * requires conversion into host byte order first.
15175 15182 */
15176 15183 void
15177 15184 ipobs_hook(mblk_t *mp, int htype, zoneid_t zsrc, zoneid_t zdst,
15178 15185 const ill_t *ill, ip_stack_t *ipst)
15179 15186 {
15180 15187 hook_pkt_observe_t *hdr;
15181 15188 uint64_t grifindex;
15182 15189 mblk_t *imp;
15183 15190
15184 15191 imp = allocb(sizeof (*hdr), BPRI_HI);
15185 15192 if (imp == NULL)
15186 15193 return;
15187 15194
15188 15195 hdr = (hook_pkt_observe_t *)imp->b_rptr;
15189 15196 /*
15190 15197 * b_wptr is set to make the apparent size of the data in the mblk_t
15191 15198 * to exclude the pointers at the end of hook_pkt_observer_t.
15192 15199 */
15193 15200 imp->b_wptr = imp->b_rptr + sizeof (dl_ipnetinfo_t);
15194 15201 imp->b_cont = mp;
15195 15202
15196 15203 ASSERT(DB_TYPE(mp) == M_DATA);
15197 15204
15198 15205 if (IS_UNDER_IPMP(ill))
15199 15206 grifindex = ipmp_ill_get_ipmp_ifindex(ill);
15200 15207 else
15201 15208 grifindex = 0;
15202 15209
15203 15210 hdr->hpo_version = 1;
15204 15211 hdr->hpo_htype = htons(htype);
15205 15212 hdr->hpo_pktlen = htonl((ulong_t)msgdsize(mp));
15206 15213 hdr->hpo_ifindex = htonl(ill->ill_phyint->phyint_ifindex);
15207 15214 hdr->hpo_grifindex = htonl(grifindex);
15208 15215 hdr->hpo_zsrc = htonl(zsrc);
15209 15216 hdr->hpo_zdst = htonl(zdst);
15210 15217 hdr->hpo_pkt = imp;
15211 15218 hdr->hpo_ctx = ipst->ips_netstack;
15212 15219
15213 15220 if (ill->ill_isv6) {
15214 15221 hdr->hpo_family = AF_INET6;
15215 15222 (void) hook_run(ipst->ips_ipv6_net_data->netd_hooks,
15216 15223 ipst->ips_ipv6observing, (hook_data_t)hdr);
15217 15224 } else {
15218 15225 hdr->hpo_family = AF_INET;
15219 15226 (void) hook_run(ipst->ips_ipv4_net_data->netd_hooks,
15220 15227 ipst->ips_ipv4observing, (hook_data_t)hdr);
15221 15228 }
15222 15229
15223 15230 imp->b_cont = NULL;
15224 15231 freemsg(imp);
15225 15232 }
15226 15233
15227 15234 /*
15228 15235 * Utility routine that checks if `v4srcp' is a valid address on underlying
15229 15236 * interface `ill'. If `ipifp' is non-NULL, it's set to a held ipif
15230 15237 * associated with `v4srcp' on success. NOTE: if this is not called from
15231 15238 * inside the IPSQ (ill_g_lock is not held), `ill' may be removed from the
15232 15239 * group during or after this lookup.
15233 15240 */
15234 15241 boolean_t
15235 15242 ipif_lookup_testaddr_v4(ill_t *ill, const in_addr_t *v4srcp, ipif_t **ipifp)
15236 15243 {
15237 15244 ipif_t *ipif;
15238 15245
15239 15246 ipif = ipif_lookup_addr_exact(*v4srcp, ill, ill->ill_ipst);
15240 15247 if (ipif != NULL) {
15241 15248 if (ipifp != NULL)
15242 15249 *ipifp = ipif;
15243 15250 else
15244 15251 ipif_refrele(ipif);
15245 15252 return (B_TRUE);
15246 15253 }
15247 15254
15248 15255 ip1dbg(("ipif_lookup_testaddr_v4: cannot find ipif for src %x\n",
15249 15256 *v4srcp));
15250 15257 return (B_FALSE);
15251 15258 }
15252 15259
15253 15260 /*
15254 15261 * Transport protocol call back function for CPU state change.
15255 15262 */
15256 15263 /* ARGSUSED */
15257 15264 static int
15258 15265 ip_tp_cpu_update(cpu_setup_t what, int id, void *arg)
15259 15266 {
15260 15267 processorid_t cpu_seqid;
15261 15268 netstack_handle_t nh;
15262 15269 netstack_t *ns;
15263 15270
15264 15271 ASSERT(MUTEX_HELD(&cpu_lock));
15265 15272
15266 15273 switch (what) {
15267 15274 case CPU_CONFIG:
15268 15275 case CPU_ON:
15269 15276 case CPU_INIT:
15270 15277 case CPU_CPUPART_IN:
15271 15278 cpu_seqid = cpu[id]->cpu_seqid;
15272 15279 netstack_next_init(&nh);
15273 15280 while ((ns = netstack_next(&nh)) != NULL) {
15274 15281 tcp_stack_cpu_add(ns->netstack_tcp, cpu_seqid);
15275 15282 sctp_stack_cpu_add(ns->netstack_sctp, cpu_seqid);
15276 15283 udp_stack_cpu_add(ns->netstack_udp, cpu_seqid);
15277 15284 netstack_rele(ns);
15278 15285 }
15279 15286 netstack_next_fini(&nh);
15280 15287 break;
15281 15288 case CPU_UNCONFIG:
15282 15289 case CPU_OFF:
15283 15290 case CPU_CPUPART_OUT:
15284 15291 /*
15285 15292 * Nothing to do. We don't remove the per CPU stats from
15286 15293 * the IP stack even when the CPU goes offline.
15287 15294 */
15288 15295 break;
15289 15296 default:
15290 15297 break;
15291 15298 }
15292 15299 return (0);
15293 15300 }
|
↓ open down ↓ |
5564 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX