Print this page
OS-7184 prototype
| 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 #include <inet/cc.h>
99 99
100 100 #include <net/pfkeyv2.h>
101 101 #include <inet/sadb.h>
102 102 #include <inet/ipsec_impl.h>
103 103 #include <inet/iptun/iptun_impl.h>
104 104 #include <inet/ipdrop.h>
105 105 #include <inet/ip_netinfo.h>
106 106 #include <inet/ilb_ip.h>
107 107
108 108 #include <sys/ethernet.h>
109 109 #include <net/if_types.h>
110 110 #include <sys/cpuvar.h>
111 111
112 112 #include <ipp/ipp.h>
113 113 #include <ipp/ipp_impl.h>
114 114 #include <ipp/ipgpc/ipgpc.h>
115 115
116 116 #include <sys/pattr.h>
117 117 #include <inet/ipclassifier.h>
118 118 #include <inet/sctp_ip.h>
119 119 #include <inet/sctp/sctp_impl.h>
120 120 #include <inet/udp_impl.h>
121 121 #include <inet/rawip_impl.h>
122 122 #include <inet/rts_impl.h>
123 123
124 124 #include <sys/tsol/label.h>
125 125 #include <sys/tsol/tnet.h>
126 126
127 127 #include <sys/squeue_impl.h>
128 128 #include <inet/ip_arp.h>
129 129
130 130 #include <sys/clock_impl.h> /* For LBOLT_FASTPATH{,64} */
131 131
132 132 /*
133 133 * Values for squeue switch:
134 134 * IP_SQUEUE_ENTER_NODRAIN: SQ_NODRAIN
135 135 * IP_SQUEUE_ENTER: SQ_PROCESS
136 136 * IP_SQUEUE_FILL: SQ_FILL
137 137 */
138 138 int ip_squeue_enter = IP_SQUEUE_ENTER; /* Setable in /etc/system */
139 139
140 140 int ip_squeue_flag;
141 141
142 142 /*
143 143 * Setable in /etc/system
144 144 */
145 145 int ip_poll_normal_ms = 100;
146 146 int ip_poll_normal_ticks = 0;
147 147 int ip_modclose_ackwait_ms = 3000;
148 148
149 149 /*
150 150 * It would be nice to have these present only in DEBUG systems, but the
151 151 * current design of the global symbol checking logic requires them to be
152 152 * unconditionally present.
153 153 */
154 154 uint_t ip_thread_data; /* TSD key for debug support */
155 155 krwlock_t ip_thread_rwlock;
156 156 list_t ip_thread_list;
157 157
158 158 /*
159 159 * Structure to represent a linked list of msgblks. Used by ip_snmp_ functions.
160 160 */
161 161
162 162 struct listptr_s {
163 163 mblk_t *lp_head; /* pointer to the head of the list */
164 164 mblk_t *lp_tail; /* pointer to the tail of the list */
165 165 };
166 166
167 167 typedef struct listptr_s listptr_t;
168 168
169 169 /*
170 170 * This is used by ip_snmp_get_mib2_ip_route_media and
171 171 * ip_snmp_get_mib2_ip6_route_media to carry the lists of return data.
172 172 */
173 173 typedef struct iproutedata_s {
174 174 uint_t ird_idx;
175 175 uint_t ird_flags; /* see below */
176 176 listptr_t ird_route; /* ipRouteEntryTable */
177 177 listptr_t ird_netmedia; /* ipNetToMediaEntryTable */
178 178 listptr_t ird_attrs; /* ipRouteAttributeTable */
179 179 } iproutedata_t;
180 180
181 181 /* Include ire_testhidden and IRE_IF_CLONE routes */
182 182 #define IRD_REPORT_ALL 0x01
183 183
184 184 /*
185 185 * Cluster specific hooks. These should be NULL when booted as a non-cluster
186 186 */
187 187
188 188 /*
189 189 * Hook functions to enable cluster networking
190 190 * On non-clustered systems these vectors must always be NULL.
191 191 *
192 192 * Hook function to Check ip specified ip address is a shared ip address
193 193 * in the cluster
194 194 *
195 195 */
196 196 int (*cl_inet_isclusterwide)(netstackid_t stack_id, uint8_t protocol,
197 197 sa_family_t addr_family, uint8_t *laddrp, void *args) = NULL;
198 198
199 199 /*
200 200 * Hook function to generate cluster wide ip fragment identifier
201 201 */
202 202 uint32_t (*cl_inet_ipident)(netstackid_t stack_id, uint8_t protocol,
203 203 sa_family_t addr_family, uint8_t *laddrp, uint8_t *faddrp,
204 204 void *args) = NULL;
205 205
206 206 /*
207 207 * Hook function to generate cluster wide SPI.
208 208 */
209 209 void (*cl_inet_getspi)(netstackid_t, uint8_t, uint8_t *, size_t,
210 210 void *) = NULL;
211 211
212 212 /*
213 213 * Hook function to verify if the SPI is already utlized.
214 214 */
215 215
216 216 int (*cl_inet_checkspi)(netstackid_t, uint8_t, uint32_t, void *) = NULL;
217 217
218 218 /*
219 219 * Hook function to delete the SPI from the cluster wide repository.
220 220 */
221 221
222 222 void (*cl_inet_deletespi)(netstackid_t, uint8_t, uint32_t, void *) = NULL;
223 223
224 224 /*
225 225 * Hook function to inform the cluster when packet received on an IDLE SA
226 226 */
227 227
228 228 void (*cl_inet_idlesa)(netstackid_t, uint8_t, uint32_t, sa_family_t,
229 229 in6_addr_t, in6_addr_t, void *) = NULL;
230 230
231 231 /*
232 232 * Synchronization notes:
233 233 *
234 234 * IP is a fully D_MP STREAMS module/driver. Thus it does not depend on any
235 235 * MT level protection given by STREAMS. IP uses a combination of its own
236 236 * internal serialization mechanism and standard Solaris locking techniques.
237 237 * The internal serialization is per phyint. This is used to serialize
238 238 * plumbing operations, IPMP operations, most set ioctls, etc.
239 239 *
240 240 * Plumbing is a long sequence of operations involving message
241 241 * exchanges between IP, ARP and device drivers. Many set ioctls are typically
242 242 * involved in plumbing operations. A natural model is to serialize these
243 243 * ioctls one per ill. For example plumbing of hme0 and qfe0 can go on in
244 244 * parallel without any interference. But various set ioctls on hme0 are best
245 245 * serialized, along with IPMP operations and processing of DLPI control
246 246 * messages received from drivers on a per phyint basis. This serialization is
247 247 * provided by the ipsq_t and primitives operating on this. Details can
248 248 * be found in ip_if.c above the core primitives operating on ipsq_t.
249 249 *
250 250 * Lookups of an ipif or ill by a thread return a refheld ipif / ill.
251 251 * Simiarly lookup of an ire by a thread also returns a refheld ire.
252 252 * In addition ipif's and ill's referenced by the ire are also indirectly
253 253 * refheld. Thus no ipif or ill can vanish as long as an ipif is refheld
254 254 * directly or indirectly. For example an SIOCSLIFADDR ioctl that changes the
255 255 * address of an ipif has to go through the ipsq_t. This ensures that only
256 256 * one such exclusive operation proceeds at any time on the ipif. It then
257 257 * waits for all refcnts
258 258 * associated with this ipif to come down to zero. The address is changed
259 259 * only after the ipif has been quiesced. Then the ipif is brought up again.
260 260 * More details are described above the comment in ip_sioctl_flags.
261 261 *
262 262 * Packet processing is based mostly on IREs and are fully multi-threaded
263 263 * using standard Solaris MT techniques.
264 264 *
265 265 * There are explicit locks in IP to handle:
266 266 * - The ip_g_head list maintained by mi_open_link() and friends.
267 267 *
268 268 * - The reassembly data structures (one lock per hash bucket)
269 269 *
270 270 * - conn_lock is meant to protect conn_t fields. The fields actually
271 271 * protected by conn_lock are documented in the conn_t definition.
272 272 *
273 273 * - ire_lock to protect some of the fields of the ire, IRE tables
274 274 * (one lock per hash bucket). Refer to ip_ire.c for details.
275 275 *
276 276 * - ndp_g_lock and ncec_lock for protecting NCEs.
277 277 *
278 278 * - ill_lock protects fields of the ill and ipif. Details in ip.h
279 279 *
280 280 * - ill_g_lock: This is a global reader/writer lock. Protects the following
281 281 * * The AVL tree based global multi list of all ills.
282 282 * * The linked list of all ipifs of an ill
283 283 * * The <ipsq-xop> mapping
284 284 * * <ill-phyint> association
285 285 * Insertion/deletion of an ill in the system, insertion/deletion of an ipif
286 286 * into an ill, changing the <ipsq-xop> mapping of an ill, changing the
287 287 * <ill-phyint> assoc of an ill will all have to hold the ill_g_lock as
288 288 * writer for the actual duration of the insertion/deletion/change.
289 289 *
290 290 * - ill_lock: This is a per ill mutex.
291 291 * It protects some members of the ill_t struct; see ip.h for details.
292 292 * It also protects the <ill-phyint> assoc.
293 293 * It also protects the list of ipifs hanging off the ill.
294 294 *
295 295 * - ipsq_lock: This is a per ipsq_t mutex lock.
296 296 * This protects some members of the ipsq_t struct; see ip.h for details.
297 297 * It also protects the <ipsq-ipxop> mapping
298 298 *
299 299 * - ipx_lock: This is a per ipxop_t mutex lock.
300 300 * This protects some members of the ipxop_t struct; see ip.h for details.
301 301 *
302 302 * - phyint_lock: This is a per phyint mutex lock. Protects just the
303 303 * phyint_flags
304 304 *
305 305 * - ip_addr_avail_lock: This is used to ensure the uniqueness of IP addresses.
306 306 * This lock is held in ipif_up_done and the ipif is marked IPIF_UP and the
307 307 * uniqueness check also done atomically.
308 308 *
309 309 * - ill_g_usesrc_lock: This readers/writer lock protects the usesrc
310 310 * group list linked by ill_usesrc_grp_next. It also protects the
311 311 * ill_usesrc_ifindex field. It is taken as a writer when a member of the
312 312 * group is being added or deleted. This lock is taken as a reader when
313 313 * walking the list/group(eg: to get the number of members in a usesrc group).
314 314 * Note, it is only necessary to take this lock if the ill_usesrc_grp_next
315 315 * field is changing state i.e from NULL to non-NULL or vice-versa. For
316 316 * example, it is not necessary to take this lock in the initial portion
317 317 * of ip_sioctl_slifusesrc or at all in ip_sioctl_flags since these
318 318 * operations are executed exclusively and that ensures that the "usesrc
319 319 * group state" cannot change. The "usesrc group state" change can happen
320 320 * only in the latter part of ip_sioctl_slifusesrc and in ill_delete.
321 321 *
322 322 * Changing <ill-phyint>, <ipsq-xop> assocications:
323 323 *
324 324 * To change the <ill-phyint> association, the ill_g_lock must be held
325 325 * as writer, and the ill_locks of both the v4 and v6 instance of the ill
326 326 * must be held.
327 327 *
328 328 * To change the <ipsq-xop> association, the ill_g_lock must be held as
329 329 * writer, the ipsq_lock must be held, and one must be writer on the ipsq.
330 330 * This is only done when ills are added or removed from IPMP groups.
331 331 *
332 332 * To add or delete an ipif from the list of ipifs hanging off the ill,
333 333 * ill_g_lock (writer) and ill_lock must be held and the thread must be
334 334 * a writer on the associated ipsq.
335 335 *
336 336 * To add or delete an ill to the system, the ill_g_lock must be held as
337 337 * writer and the thread must be a writer on the associated ipsq.
338 338 *
339 339 * To add or delete an ilm to an ill, the ill_lock must be held and the thread
340 340 * must be a writer on the associated ipsq.
341 341 *
342 342 * Lock hierarchy
343 343 *
344 344 * Some lock hierarchy scenarios are listed below.
345 345 *
346 346 * ill_g_lock -> conn_lock -> ill_lock -> ipsq_lock -> ipx_lock
347 347 * ill_g_lock -> ill_lock(s) -> phyint_lock
348 348 * ill_g_lock -> ndp_g_lock -> ill_lock -> ncec_lock
349 349 * ill_g_lock -> ip_addr_avail_lock
350 350 * conn_lock -> irb_lock -> ill_lock -> ire_lock
351 351 * ill_g_lock -> ip_g_nd_lock
352 352 * ill_g_lock -> ips_ipmp_lock -> ill_lock -> nce_lock
353 353 * ill_g_lock -> ndp_g_lock -> ill_lock -> ncec_lock -> nce_lock
354 354 * arl_lock -> ill_lock
355 355 * ips_ire_dep_lock -> irb_lock
356 356 *
357 357 * When more than 1 ill lock is needed to be held, all ill lock addresses
358 358 * are sorted on address and locked starting from highest addressed lock
359 359 * downward.
360 360 *
361 361 * Multicast scenarios
362 362 * ips_ill_g_lock -> ill_mcast_lock
363 363 * conn_ilg_lock -> ips_ill_g_lock -> ill_lock
364 364 * ill_mcast_serializer -> ill_mcast_lock -> ips_ipmp_lock -> ill_lock
365 365 * ill_mcast_serializer -> ill_mcast_lock -> connf_lock -> conn_lock
366 366 * ill_mcast_serializer -> ill_mcast_lock -> conn_ilg_lock
367 367 * ill_mcast_serializer -> ill_mcast_lock -> ips_igmp_timer_lock
368 368 *
369 369 * IPsec scenarios
370 370 *
371 371 * ipsa_lock -> ill_g_lock -> ill_lock
372 372 * ill_g_usesrc_lock -> ill_g_lock -> ill_lock
373 373 *
374 374 * Trusted Solaris scenarios
375 375 *
376 376 * igsa_lock -> gcgrp_rwlock -> gcgrp_lock
377 377 * igsa_lock -> gcdb_lock
378 378 * gcgrp_rwlock -> ire_lock
379 379 * gcgrp_rwlock -> gcdb_lock
380 380 *
381 381 * squeue(sq_lock), flow related (ft_lock, fe_lock) locking
382 382 *
383 383 * cpu_lock --> ill_lock --> sqset_lock --> sq_lock
384 384 * sq_lock -> conn_lock -> QLOCK(q)
385 385 * ill_lock -> ft_lock -> fe_lock
386 386 *
387 387 * Routing/forwarding table locking notes:
388 388 *
389 389 * Lock acquisition order: Radix tree lock, irb_lock.
390 390 * Requirements:
391 391 * i. Walker must not hold any locks during the walker callback.
392 392 * ii Walker must not see a truncated tree during the walk because of any node
393 393 * deletion.
394 394 * iii Existing code assumes ire_bucket is valid if it is non-null and is used
395 395 * in many places in the code to walk the irb list. Thus even if all the
396 396 * ires in a bucket have been deleted, we still can't free the radix node
397 397 * until the ires have actually been inactive'd (freed).
398 398 *
399 399 * Tree traversal - Need to hold the global tree lock in read mode.
400 400 * Before dropping the global tree lock, need to either increment the ire_refcnt
401 401 * to ensure that the radix node can't be deleted.
402 402 *
403 403 * Tree add - Need to hold the global tree lock in write mode to add a
404 404 * radix node. To prevent the node from being deleted, increment the
405 405 * irb_refcnt, after the node is added to the tree. The ire itself is
406 406 * added later while holding the irb_lock, but not the tree lock.
407 407 *
408 408 * Tree delete - Need to hold the global tree lock and irb_lock in write mode.
409 409 * All associated ires must be inactive (i.e. freed), and irb_refcnt
410 410 * must be zero.
411 411 *
412 412 * Walker - Increment irb_refcnt before calling the walker callback. Hold the
413 413 * global tree lock (read mode) for traversal.
414 414 *
415 415 * IRE dependencies - In some cases we hold ips_ire_dep_lock across ire_refrele
416 416 * hence we will acquire irb_lock while holding ips_ire_dep_lock.
417 417 *
418 418 * IPsec notes :
419 419 *
420 420 * IP interacts with the IPsec code (AH/ESP) by storing IPsec attributes
421 421 * in the ip_xmit_attr_t ip_recv_attr_t. For outbound datagrams, the
422 422 * ip_xmit_attr_t has the
423 423 * information used by the IPsec code for applying the right level of
424 424 * protection. The information initialized by IP in the ip_xmit_attr_t
425 425 * is determined by the per-socket policy or global policy in the system.
426 426 * For inbound datagrams, the ip_recv_attr_t
427 427 * starts out with nothing in it. It gets filled
428 428 * with the right information if it goes through the AH/ESP code, which
429 429 * happens if the incoming packet is secure. The information initialized
430 430 * by AH/ESP, is later used by IP (during fanouts to ULP) to see whether
431 431 * the policy requirements needed by per-socket policy or global policy
432 432 * is met or not.
433 433 *
434 434 * For fully connected sockets i.e dst, src [addr, port] is known,
435 435 * conn_policy_cached is set indicating that policy has been cached.
436 436 * conn_in_enforce_policy may or may not be set depending on whether
437 437 * there is a global policy match or per-socket policy match.
438 438 * Policy inheriting happpens in ip_policy_set once the destination is known.
439 439 * Once the right policy is set on the conn_t, policy cannot change for
440 440 * this socket. This makes life simpler for TCP (UDP ?) where
441 441 * re-transmissions go out with the same policy. For symmetry, policy
442 442 * is cached for fully connected UDP sockets also. Thus if policy is cached,
443 443 * it also implies that policy is latched i.e policy cannot change
444 444 * on these sockets. As we have the right policy on the conn, we don't
445 445 * have to lookup global policy for every outbound and inbound datagram
446 446 * and thus serving as an optimization. Note that a global policy change
447 447 * does not affect fully connected sockets if they have policy. If fully
448 448 * connected sockets did not have any policy associated with it, global
449 449 * policy change may affect them.
450 450 *
451 451 * IP Flow control notes:
452 452 * ---------------------
453 453 * Non-TCP streams are flow controlled by IP. The way this is accomplished
454 454 * differs when ILL_CAPAB_DLD_DIRECT is enabled for that IP instance. When
455 455 * ILL_DIRECT_CAPABLE(ill) is TRUE, IP can do direct function calls into
456 456 * GLDv3. Otherwise packets are sent down to lower layers using STREAMS
457 457 * functions.
458 458 *
459 459 * Per Tx ring udp flow control:
460 460 * This is applicable only when ILL_CAPAB_DLD_DIRECT capability is set in
461 461 * the ill (i.e. ILL_DIRECT_CAPABLE(ill) is true).
462 462 *
463 463 * The underlying link can expose multiple Tx rings to the GLDv3 mac layer.
464 464 * To achieve best performance, outgoing traffic need to be fanned out among
465 465 * these Tx ring. mac_tx() is called (via str_mdata_fastpath_put()) to send
466 466 * traffic out of the NIC and it takes a fanout hint. UDP connections pass
467 467 * the address of connp as fanout hint to mac_tx(). Under flow controlled
468 468 * condition, mac_tx() returns a non-NULL cookie (ip_mac_tx_cookie_t). This
469 469 * cookie points to a specific Tx ring that is blocked. The cookie is used to
470 470 * hash into an idl_tx_list[] entry in idl_tx_list[] array. Each idl_tx_list_t
471 471 * point to drain_lists (idl_t's). These drain list will store the blocked UDP
472 472 * connp's. The drain list is not a single list but a configurable number of
473 473 * lists.
474 474 *
475 475 * The diagram below shows idl_tx_list_t's and their drain_lists. ip_stack_t
476 476 * has an array of idl_tx_list_t. The size of the array is TX_FANOUT_SIZE
477 477 * which is equal to 128. This array in turn contains a pointer to idl_t[],
478 478 * the ip drain list. The idl_t[] array size is MIN(max_ncpus, 8). The drain
479 479 * list will point to the list of connp's that are flow controlled.
480 480 *
481 481 * --------------- ------- ------- -------
482 482 * |->|drain_list[0]|-->|connp|-->|connp|-->|connp|-->
483 483 * | --------------- ------- ------- -------
484 484 * | --------------- ------- ------- -------
485 485 * |->|drain_list[1]|-->|connp|-->|connp|-->|connp|-->
486 486 * ---------------- | --------------- ------- ------- -------
487 487 * |idl_tx_list[0]|->| --------------- ------- ------- -------
488 488 * ---------------- |->|drain_list[2]|-->|connp|-->|connp|-->|connp|-->
489 489 * | --------------- ------- ------- -------
490 490 * . . . . .
491 491 * | --------------- ------- ------- -------
492 492 * |->|drain_list[n]|-->|connp|-->|connp|-->|connp|-->
493 493 * --------------- ------- ------- -------
494 494 * --------------- ------- ------- -------
495 495 * |->|drain_list[0]|-->|connp|-->|connp|-->|connp|-->
496 496 * | --------------- ------- ------- -------
497 497 * | --------------- ------- ------- -------
498 498 * ---------------- |->|drain_list[1]|-->|connp|-->|connp|-->|connp|-->
499 499 * |idl_tx_list[1]|->| --------------- ------- ------- -------
500 500 * ---------------- | . . . .
501 501 * | --------------- ------- ------- -------
502 502 * |->|drain_list[n]|-->|connp|-->|connp|-->|connp|-->
503 503 * --------------- ------- ------- -------
504 504 * .....
505 505 * ----------------
506 506 * |idl_tx_list[n]|-> ...
507 507 * ----------------
508 508 *
509 509 * When mac_tx() returns a cookie, the cookie is hashed into an index into
510 510 * ips_idl_tx_list[], and conn_drain_insert() is called with the idl_tx_list
511 511 * to insert the conn onto. conn_drain_insert() asserts flow control for the
512 512 * sockets via su_txq_full() (non-STREAMS) or QFULL on conn_wq (STREAMS).
513 513 * Further, conn_blocked is set to indicate that the conn is blocked.
514 514 *
515 515 * GLDv3 calls ill_flow_enable() when flow control is relieved. The cookie
516 516 * passed in the call to ill_flow_enable() identifies the blocked Tx ring and
517 517 * is again hashed to locate the appropriate idl_tx_list, which is then
518 518 * drained via conn_walk_drain(). conn_walk_drain() goes through each conn in
519 519 * the drain list and calls conn_drain_remove() to clear flow control (via
520 520 * calling su_txq_full() or clearing QFULL), and remove the conn from the
521 521 * drain list.
522 522 *
523 523 * Note that the drain list is not a single list but a (configurable) array of
524 524 * lists (8 elements by default). Synchronization between drain insertion and
525 525 * flow control wakeup is handled by using idl_txl->txl_lock, and only
526 526 * conn_drain_insert() and conn_drain_remove() manipulate the drain list.
527 527 *
528 528 * Flow control via STREAMS is used when ILL_DIRECT_CAPABLE() returns FALSE.
529 529 * On the send side, if the packet cannot be sent down to the driver by IP
530 530 * (canput() fails), ip_xmit() drops the packet and returns EWOULDBLOCK to the
531 531 * caller, who may then invoke ixa_check_drain_insert() to insert the conn on
532 532 * the 0'th drain list. When ip_wsrv() runs on the ill_wq because flow
533 533 * control has been relieved, the blocked conns in the 0'th drain list are
534 534 * drained as in the non-STREAMS case.
535 535 *
536 536 * In both the STREAMS and non-STREAMS cases, the sockfs upcall to set QFULL
537 537 * is done when the conn is inserted into the drain list (conn_drain_insert())
538 538 * and cleared when the conn is removed from the it (conn_drain_remove()).
539 539 *
540 540 * IPQOS notes:
541 541 *
542 542 * IPQoS Policies are applied to packets using IPPF (IP Policy framework)
543 543 * and IPQoS modules. IPPF includes hooks in IP at different control points
544 544 * (callout positions) which direct packets to IPQoS modules for policy
545 545 * processing. Policies, if present, are global.
546 546 *
547 547 * The callout positions are located in the following paths:
548 548 * o local_in (packets destined for this host)
549 549 * o local_out (packets orginating from this host )
550 550 * o fwd_in (packets forwarded by this m/c - inbound)
551 551 * o fwd_out (packets forwarded by this m/c - outbound)
552 552 * Hooks at these callout points can be enabled/disabled using the ndd variable
553 553 * ip_policy_mask (a bit mask with the 4 LSB indicating the callout positions).
554 554 * By default all the callout positions are enabled.
555 555 *
556 556 * Outbound (local_out)
557 557 * Hooks are placed in ire_send_wire_v4 and ire_send_wire_v6.
558 558 *
559 559 * Inbound (local_in)
560 560 * Hooks are placed in ip_fanout_v4 and ip_fanout_v6.
561 561 *
562 562 * Forwarding (in and out)
563 563 * Hooks are placed in ire_recv_forward_v4/v6.
564 564 *
565 565 * IP Policy Framework processing (IPPF processing)
566 566 * Policy processing for a packet is initiated by ip_process, which ascertains
567 567 * that the classifier (ipgpc) is loaded and configured, failing which the
568 568 * packet resumes normal processing in IP. If the clasifier is present, the
569 569 * packet is acted upon by one or more IPQoS modules (action instances), per
570 570 * filters configured in ipgpc and resumes normal IP processing thereafter.
571 571 * An action instance can drop a packet in course of its processing.
572 572 *
573 573 * Zones notes:
574 574 *
575 575 * The partitioning rules for networking are as follows:
576 576 * 1) Packets coming from a zone must have a source address belonging to that
577 577 * zone.
578 578 * 2) Packets coming from a zone can only be sent on a physical interface on
579 579 * which the zone has an IP address.
580 580 * 3) Between two zones on the same machine, packet delivery is only allowed if
581 581 * there's a matching route for the destination and zone in the forwarding
582 582 * table.
583 583 * 4) The TCP and UDP port spaces are per-zone; that is, two processes in
584 584 * different zones can bind to the same port with the wildcard address
585 585 * (INADDR_ANY).
586 586 *
587 587 * The granularity of interface partitioning is at the logical interface level.
588 588 * Therefore, every zone has its own IP addresses, and incoming packets can be
589 589 * attributed to a zone unambiguously. A logical interface is placed into a zone
590 590 * using the SIOCSLIFZONE ioctl; this sets the ipif_zoneid field in the ipif_t
591 591 * structure. Rule (1) is implemented by modifying the source address selection
592 592 * algorithm so that the list of eligible addresses is filtered based on the
593 593 * sending process zone.
594 594 *
595 595 * The Internet Routing Entries (IREs) are either exclusive to a zone or shared
596 596 * across all zones, depending on their type. Here is the break-up:
597 597 *
598 598 * IRE type Shared/exclusive
599 599 * -------- ----------------
600 600 * IRE_BROADCAST Exclusive
601 601 * IRE_DEFAULT (default routes) Shared (*)
602 602 * IRE_LOCAL Exclusive (x)
603 603 * IRE_LOOPBACK Exclusive
604 604 * IRE_PREFIX (net routes) Shared (*)
605 605 * IRE_IF_NORESOLVER (interface routes) Exclusive
606 606 * IRE_IF_RESOLVER (interface routes) Exclusive
607 607 * IRE_IF_CLONE (interface routes) Exclusive
608 608 * IRE_HOST (host routes) Shared (*)
609 609 *
610 610 * (*) A zone can only use a default or off-subnet route if the gateway is
611 611 * directly reachable from the zone, that is, if the gateway's address matches
612 612 * one of the zone's logical interfaces.
613 613 *
614 614 * (x) IRE_LOCAL are handled a bit differently.
615 615 * When ip_restrict_interzone_loopback is set (the default),
616 616 * ire_route_recursive restricts loopback using an IRE_LOCAL
617 617 * between zone to the case when L2 would have conceptually looped the packet
618 618 * back, i.e. the loopback which is required since neither Ethernet drivers
619 619 * nor Ethernet hardware loops them back. This is the case when the normal
620 620 * routes (ignoring IREs with different zoneids) would send out the packet on
621 621 * the same ill as the ill with which is IRE_LOCAL is associated.
622 622 *
623 623 * Multiple zones can share a common broadcast address; typically all zones
624 624 * share the 255.255.255.255 address. Incoming as well as locally originated
625 625 * broadcast packets must be dispatched to all the zones on the broadcast
626 626 * network. For directed broadcasts (e.g. 10.16.72.255) this is not trivial
627 627 * since some zones may not be on the 10.16.72/24 network. To handle this, each
628 628 * zone has its own set of IRE_BROADCAST entries; then, broadcast packets are
629 629 * sent to every zone that has an IRE_BROADCAST entry for the destination
630 630 * address on the input ill, see ip_input_broadcast().
631 631 *
632 632 * Applications in different zones can join the same multicast group address.
633 633 * The same logic applies for multicast as for broadcast. ip_input_multicast
634 634 * dispatches packets to all zones that have members on the physical interface.
635 635 */
636 636
637 637 /*
638 638 * Squeue Fanout flags:
639 639 * 0: No fanout.
640 640 * 1: Fanout across all squeues
641 641 */
642 642 boolean_t ip_squeue_fanout = 0;
643 643
644 644 /*
645 645 * Maximum dups allowed per packet.
646 646 */
647 647 uint_t ip_max_frag_dups = 10;
648 648
649 649 static int ip_open(queue_t *q, dev_t *devp, int flag, int sflag,
650 650 cred_t *credp, boolean_t isv6);
651 651 static mblk_t *ip_xmit_attach_llhdr(mblk_t *, nce_t *);
652 652
653 653 static boolean_t icmp_inbound_verify_v4(mblk_t *, icmph_t *, ip_recv_attr_t *);
654 654 static void icmp_inbound_too_big_v4(icmph_t *, ip_recv_attr_t *);
655 655 static void icmp_inbound_error_fanout_v4(mblk_t *, icmph_t *,
656 656 ip_recv_attr_t *);
657 657 static void icmp_options_update(ipha_t *);
658 658 static void icmp_param_problem(mblk_t *, uint8_t, ip_recv_attr_t *);
659 659 static void icmp_pkt(mblk_t *, void *, size_t, ip_recv_attr_t *);
660 660 static mblk_t *icmp_pkt_err_ok(mblk_t *, ip_recv_attr_t *);
661 661 static void icmp_redirect_v4(mblk_t *mp, ipha_t *, icmph_t *,
662 662 ip_recv_attr_t *);
663 663 static void icmp_send_redirect(mblk_t *, ipaddr_t, ip_recv_attr_t *);
664 664 static void icmp_send_reply_v4(mblk_t *, ipha_t *, icmph_t *,
665 665 ip_recv_attr_t *);
666 666
667 667 mblk_t *ip_dlpi_alloc(size_t, t_uscalar_t);
668 668 char *ip_dot_addr(ipaddr_t, char *);
669 669 mblk_t *ip_carve_mp(mblk_t **, ssize_t);
670 670 static char *ip_dot_saddr(uchar_t *, char *);
671 671 static int ip_lrput(queue_t *, mblk_t *);
672 672 ipaddr_t ip_net_mask(ipaddr_t);
673 673 char *ip_nv_lookup(nv_t *, int);
674 674 int ip_rput(queue_t *, mblk_t *);
675 675 static void ip_rput_dlpi_writer(ipsq_t *dummy_sq, queue_t *q, mblk_t *mp,
676 676 void *dummy_arg);
677 677 int ip_snmp_get(queue_t *, mblk_t *, int, boolean_t);
678 678 static mblk_t *ip_snmp_get_mib2_ip(queue_t *, mblk_t *,
679 679 mib2_ipIfStatsEntry_t *, ip_stack_t *, boolean_t);
680 680 static mblk_t *ip_snmp_get_mib2_ip_traffic_stats(queue_t *, mblk_t *,
681 681 ip_stack_t *, boolean_t);
682 682 static mblk_t *ip_snmp_get_mib2_ip6(queue_t *, mblk_t *, ip_stack_t *,
683 683 boolean_t);
684 684 static mblk_t *ip_snmp_get_mib2_icmp(queue_t *, mblk_t *, ip_stack_t *ipst);
685 685 static mblk_t *ip_snmp_get_mib2_icmp6(queue_t *, mblk_t *, ip_stack_t *ipst);
686 686 static mblk_t *ip_snmp_get_mib2_igmp(queue_t *, mblk_t *, ip_stack_t *ipst);
687 687 static mblk_t *ip_snmp_get_mib2_multi(queue_t *, mblk_t *, ip_stack_t *ipst);
688 688 static mblk_t *ip_snmp_get_mib2_ip_addr(queue_t *, mblk_t *,
689 689 ip_stack_t *ipst, boolean_t);
690 690 static mblk_t *ip_snmp_get_mib2_ip6_addr(queue_t *, mblk_t *,
691 691 ip_stack_t *ipst, boolean_t);
692 692 static mblk_t *ip_snmp_get_mib2_ip_group_src(queue_t *, mblk_t *,
693 693 ip_stack_t *ipst);
694 694 static mblk_t *ip_snmp_get_mib2_ip6_group_src(queue_t *, mblk_t *,
695 695 ip_stack_t *ipst);
696 696 static mblk_t *ip_snmp_get_mib2_ip_group_mem(queue_t *, mblk_t *,
697 697 ip_stack_t *ipst);
698 698 static mblk_t *ip_snmp_get_mib2_ip6_group_mem(queue_t *, mblk_t *,
699 699 ip_stack_t *ipst);
700 700 static mblk_t *ip_snmp_get_mib2_virt_multi(queue_t *, mblk_t *,
701 701 ip_stack_t *ipst);
702 702 static mblk_t *ip_snmp_get_mib2_multi_rtable(queue_t *, mblk_t *,
703 703 ip_stack_t *ipst);
704 704 static mblk_t *ip_snmp_get_mib2_ip_route_media(queue_t *, mblk_t *, int,
705 705 ip_stack_t *ipst);
706 706 static mblk_t *ip_snmp_get_mib2_ip6_route_media(queue_t *, mblk_t *, int,
707 707 ip_stack_t *ipst);
708 708 static void ip_snmp_get2_v4(ire_t *, iproutedata_t *);
709 709 static void ip_snmp_get2_v6_route(ire_t *, iproutedata_t *);
710 710 static void ip_snmp_get2_v4_media(ncec_t *, void *);
711 711 static void ip_snmp_get2_v6_media(ncec_t *, void *);
712 712 int ip_snmp_set(queue_t *, int, int, uchar_t *, int);
713 713
714 714 static mblk_t *ip_fragment_copyhdr(uchar_t *, int, int, ip_stack_t *,
715 715 mblk_t *);
716 716
717 717 static void conn_drain_init(ip_stack_t *);
718 718 static void conn_drain_fini(ip_stack_t *);
719 719 static void conn_drain(conn_t *connp, boolean_t closing);
720 720
721 721 static void conn_walk_drain(ip_stack_t *, idl_tx_list_t *);
722 722 static void conn_walk_sctp(pfv_t, void *, zoneid_t, netstack_t *);
723 723
724 724 static void *ip_stack_init(netstackid_t stackid, netstack_t *ns);
725 725 static void ip_stack_shutdown(netstackid_t stackid, void *arg);
726 726 static void ip_stack_fini(netstackid_t stackid, void *arg);
727 727
728 728 static int ip_multirt_apply_membership(int (*fn)(conn_t *, boolean_t,
729 729 const in6_addr_t *, ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *),
730 730 ire_t *, conn_t *, boolean_t, const in6_addr_t *, mcast_record_t,
731 731 const in6_addr_t *);
732 732
733 733 static int ip_squeue_switch(int);
734 734
735 735 static void *ip_kstat_init(netstackid_t, ip_stack_t *);
736 736 static void ip_kstat_fini(netstackid_t, kstat_t *);
737 737 static int ip_kstat_update(kstat_t *kp, int rw);
738 738 static void *icmp_kstat_init(netstackid_t);
739 739 static void icmp_kstat_fini(netstackid_t, kstat_t *);
740 740 static int icmp_kstat_update(kstat_t *kp, int rw);
741 741 static void *ip_kstat2_init(netstackid_t, ip_stat_t *);
742 742 static void ip_kstat2_fini(netstackid_t, kstat_t *);
743 743
744 744 static void ipobs_init(ip_stack_t *);
745 745 static void ipobs_fini(ip_stack_t *);
746 746
747 747 static int ip_tp_cpu_update(cpu_setup_t, int, void *);
748 748
749 749 ipaddr_t ip_g_all_ones = IP_HOST_MASK;
750 750
751 751 static long ip_rput_pullups;
752 752 int dohwcksum = 1; /* use h/w cksum if supported by the hardware */
753 753
754 754 vmem_t *ip_minor_arena_sa; /* for minor nos. from INET_MIN_DEV+2 thru 2^^18-1 */
755 755 vmem_t *ip_minor_arena_la; /* for minor nos. from 2^^18 thru 2^^32-1 */
756 756
757 757 int ip_debug;
758 758
759 759 /*
760 760 * Multirouting/CGTP stuff
761 761 */
762 762 int ip_cgtp_filter_rev = CGTP_FILTER_REV; /* CGTP hooks version */
763 763
764 764 /*
765 765 * IP tunables related declarations. Definitions are in ip_tunables.c
766 766 */
767 767 extern mod_prop_info_t ip_propinfo_tbl[];
768 768 extern int ip_propinfo_count;
769 769
770 770 /*
771 771 * Table of IP ioctls encoding the various properties of the ioctl and
772 772 * indexed based on the last byte of the ioctl command. Occasionally there
773 773 * is a clash, and there is more than 1 ioctl with the same last byte.
774 774 * In such a case 1 ioctl is encoded in the ndx table and the remaining
775 775 * ioctls are encoded in the misc table. An entry in the ndx table is
776 776 * retrieved by indexing on the last byte of the ioctl command and comparing
777 777 * the ioctl command with the value in the ndx table. In the event of a
778 778 * mismatch the misc table is then searched sequentially for the desired
779 779 * ioctl command.
780 780 *
781 781 * Entry: <command> <copyin_size> <flags> <cmd_type> <function> <restart_func>
782 782 */
783 783 ip_ioctl_cmd_t ip_ndx_ioctl_table[] = {
784 784 /* 000 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
785 785 /* 001 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
786 786 /* 002 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
787 787 /* 003 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
788 788 /* 004 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
789 789 /* 005 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
790 790 /* 006 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
791 791 /* 007 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
792 792 /* 008 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
793 793 /* 009 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
794 794
795 795 /* 010 */ { SIOCADDRT, sizeof (struct rtentry), IPI_PRIV,
796 796 MISC_CMD, ip_siocaddrt, NULL },
797 797 /* 011 */ { SIOCDELRT, sizeof (struct rtentry), IPI_PRIV,
798 798 MISC_CMD, ip_siocdelrt, NULL },
799 799
800 800 /* 012 */ { SIOCSIFADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
801 801 IF_CMD, ip_sioctl_addr, ip_sioctl_addr_restart },
802 802 /* 013 */ { SIOCGIFADDR, sizeof (struct ifreq), IPI_GET_CMD,
803 803 IF_CMD, ip_sioctl_get_addr, NULL },
804 804
805 805 /* 014 */ { SIOCSIFDSTADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
806 806 IF_CMD, ip_sioctl_dstaddr, ip_sioctl_dstaddr_restart },
807 807 /* 015 */ { SIOCGIFDSTADDR, sizeof (struct ifreq),
808 808 IPI_GET_CMD, IF_CMD, ip_sioctl_get_dstaddr, NULL },
809 809
810 810 /* 016 */ { SIOCSIFFLAGS, sizeof (struct ifreq),
811 811 IPI_PRIV | IPI_WR,
812 812 IF_CMD, ip_sioctl_flags, ip_sioctl_flags_restart },
813 813 /* 017 */ { SIOCGIFFLAGS, sizeof (struct ifreq),
814 814 IPI_MODOK | IPI_GET_CMD,
815 815 IF_CMD, ip_sioctl_get_flags, NULL },
816 816
817 817 /* 018 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
818 818 /* 019 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
819 819
820 820 /* copyin size cannot be coded for SIOCGIFCONF */
821 821 /* 020 */ { O_SIOCGIFCONF, 0, IPI_GET_CMD,
822 822 MISC_CMD, ip_sioctl_get_ifconf, NULL },
823 823
824 824 /* 021 */ { SIOCSIFMTU, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
825 825 IF_CMD, ip_sioctl_mtu, NULL },
826 826 /* 022 */ { SIOCGIFMTU, sizeof (struct ifreq), IPI_GET_CMD,
827 827 IF_CMD, ip_sioctl_get_mtu, NULL },
828 828 /* 023 */ { SIOCGIFBRDADDR, sizeof (struct ifreq),
829 829 IPI_GET_CMD, IF_CMD, ip_sioctl_get_brdaddr, NULL },
830 830 /* 024 */ { SIOCSIFBRDADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
831 831 IF_CMD, ip_sioctl_brdaddr, NULL },
832 832 /* 025 */ { SIOCGIFNETMASK, sizeof (struct ifreq),
833 833 IPI_GET_CMD, IF_CMD, ip_sioctl_get_netmask, NULL },
834 834 /* 026 */ { SIOCSIFNETMASK, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
835 835 IF_CMD, ip_sioctl_netmask, ip_sioctl_netmask_restart },
836 836 /* 027 */ { SIOCGIFMETRIC, sizeof (struct ifreq),
837 837 IPI_GET_CMD, IF_CMD, ip_sioctl_get_metric, NULL },
838 838 /* 028 */ { SIOCSIFMETRIC, sizeof (struct ifreq), IPI_PRIV,
839 839 IF_CMD, ip_sioctl_metric, NULL },
840 840 /* 029 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
841 841
842 842 /* See 166-168 below for extended SIOC*XARP ioctls */
843 843 /* 030 */ { SIOCSARP, sizeof (struct arpreq), IPI_PRIV | IPI_WR,
844 844 ARP_CMD, ip_sioctl_arp, NULL },
845 845 /* 031 */ { SIOCGARP, sizeof (struct arpreq), IPI_GET_CMD,
846 846 ARP_CMD, ip_sioctl_arp, NULL },
847 847 /* 032 */ { SIOCDARP, sizeof (struct arpreq), IPI_PRIV | IPI_WR,
848 848 ARP_CMD, ip_sioctl_arp, NULL },
849 849
850 850 /* 033 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
851 851 /* 034 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
852 852 /* 035 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
853 853 /* 036 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
854 854 /* 037 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
855 855 /* 038 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
856 856 /* 039 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
857 857 /* 040 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
858 858 /* 041 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
859 859 /* 042 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
860 860 /* 043 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
861 861 /* 044 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
862 862 /* 045 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
863 863 /* 046 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
864 864 /* 047 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
865 865 /* 048 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
866 866 /* 049 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
867 867 /* 050 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
868 868 /* 051 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
869 869 /* 052 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
870 870 /* 053 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
871 871
872 872 /* 054 */ { IF_UNITSEL, sizeof (int), IPI_PRIV | IPI_WR | IPI_MODOK,
873 873 MISC_CMD, if_unitsel, if_unitsel_restart },
874 874
875 875 /* 055 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
876 876 /* 056 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
877 877 /* 057 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
878 878 /* 058 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
879 879 /* 059 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
880 880 /* 060 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
881 881 /* 061 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
882 882 /* 062 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
883 883 /* 063 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
884 884 /* 064 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
885 885 /* 065 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
886 886 /* 066 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
887 887 /* 067 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
888 888 /* 068 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
889 889 /* 069 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
890 890 /* 070 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
891 891 /* 071 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
892 892 /* 072 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
893 893
894 894 /* 073 */ { SIOCSIFNAME, sizeof (struct ifreq),
895 895 IPI_PRIV | IPI_WR | IPI_MODOK,
896 896 IF_CMD, ip_sioctl_sifname, NULL },
897 897
898 898 /* 074 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
899 899 /* 075 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
900 900 /* 076 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
901 901 /* 077 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
902 902 /* 078 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
903 903 /* 079 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
904 904 /* 080 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
905 905 /* 081 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
906 906 /* 082 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
907 907 /* 083 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
908 908 /* 084 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
909 909 /* 085 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
910 910 /* 086 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
911 911
912 912 /* 087 */ { SIOCGIFNUM, sizeof (int), IPI_GET_CMD,
913 913 MISC_CMD, ip_sioctl_get_ifnum, NULL },
914 914 /* 088 */ { SIOCGIFMUXID, sizeof (struct ifreq), IPI_GET_CMD,
915 915 IF_CMD, ip_sioctl_get_muxid, NULL },
916 916 /* 089 */ { SIOCSIFMUXID, sizeof (struct ifreq),
917 917 IPI_PRIV | IPI_WR, IF_CMD, ip_sioctl_muxid, NULL },
918 918
919 919 /* Both if and lif variants share same func */
920 920 /* 090 */ { SIOCGIFINDEX, sizeof (struct ifreq), IPI_GET_CMD,
921 921 IF_CMD, ip_sioctl_get_lifindex, NULL },
922 922 /* Both if and lif variants share same func */
923 923 /* 091 */ { SIOCSIFINDEX, sizeof (struct ifreq),
924 924 IPI_PRIV | IPI_WR, IF_CMD, ip_sioctl_slifindex, NULL },
925 925
926 926 /* copyin size cannot be coded for SIOCGIFCONF */
927 927 /* 092 */ { SIOCGIFCONF, 0, IPI_GET_CMD,
928 928 MISC_CMD, ip_sioctl_get_ifconf, NULL },
929 929 /* 093 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
930 930 /* 094 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
931 931 /* 095 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
932 932 /* 096 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
933 933 /* 097 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
934 934 /* 098 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
935 935 /* 099 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
936 936 /* 100 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
937 937 /* 101 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
938 938 /* 102 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
939 939 /* 103 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
940 940 /* 104 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
941 941 /* 105 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
942 942 /* 106 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
943 943 /* 107 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
944 944 /* 108 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
945 945 /* 109 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
946 946
947 947 /* 110 */ { SIOCLIFREMOVEIF, sizeof (struct lifreq),
948 948 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_removeif,
949 949 ip_sioctl_removeif_restart },
950 950 /* 111 */ { SIOCLIFADDIF, sizeof (struct lifreq),
951 951 IPI_GET_CMD | IPI_PRIV | IPI_WR,
952 952 LIF_CMD, ip_sioctl_addif, NULL },
953 953 #define SIOCLIFADDR_NDX 112
954 954 /* 112 */ { SIOCSLIFADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
955 955 LIF_CMD, ip_sioctl_addr, ip_sioctl_addr_restart },
956 956 /* 113 */ { SIOCGLIFADDR, sizeof (struct lifreq),
957 957 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_addr, NULL },
958 958 /* 114 */ { SIOCSLIFDSTADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
959 959 LIF_CMD, ip_sioctl_dstaddr, ip_sioctl_dstaddr_restart },
960 960 /* 115 */ { SIOCGLIFDSTADDR, sizeof (struct lifreq),
961 961 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_dstaddr, NULL },
962 962 /* 116 */ { SIOCSLIFFLAGS, sizeof (struct lifreq),
963 963 IPI_PRIV | IPI_WR,
964 964 LIF_CMD, ip_sioctl_flags, ip_sioctl_flags_restart },
965 965 /* 117 */ { SIOCGLIFFLAGS, sizeof (struct lifreq),
966 966 IPI_GET_CMD | IPI_MODOK,
967 967 LIF_CMD, ip_sioctl_get_flags, NULL },
968 968
969 969 /* 118 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
970 970 /* 119 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
971 971
972 972 /* 120 */ { O_SIOCGLIFCONF, 0, IPI_GET_CMD, MISC_CMD,
973 973 ip_sioctl_get_lifconf, NULL },
974 974 /* 121 */ { SIOCSLIFMTU, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
975 975 LIF_CMD, ip_sioctl_mtu, NULL },
976 976 /* 122 */ { SIOCGLIFMTU, sizeof (struct lifreq), IPI_GET_CMD,
977 977 LIF_CMD, ip_sioctl_get_mtu, NULL },
978 978 /* 123 */ { SIOCGLIFBRDADDR, sizeof (struct lifreq),
979 979 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_brdaddr, NULL },
980 980 /* 124 */ { SIOCSLIFBRDADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
981 981 LIF_CMD, ip_sioctl_brdaddr, NULL },
982 982 /* 125 */ { SIOCGLIFNETMASK, sizeof (struct lifreq),
983 983 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_netmask, NULL },
984 984 /* 126 */ { SIOCSLIFNETMASK, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
985 985 LIF_CMD, ip_sioctl_netmask, ip_sioctl_netmask_restart },
986 986 /* 127 */ { SIOCGLIFMETRIC, sizeof (struct lifreq),
987 987 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_metric, NULL },
988 988 /* 128 */ { SIOCSLIFMETRIC, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
989 989 LIF_CMD, ip_sioctl_metric, NULL },
990 990 /* 129 */ { SIOCSLIFNAME, sizeof (struct lifreq),
991 991 IPI_PRIV | IPI_WR | IPI_MODOK,
992 992 LIF_CMD, ip_sioctl_slifname,
993 993 ip_sioctl_slifname_restart },
994 994
995 995 /* 130 */ { SIOCGLIFNUM, sizeof (struct lifnum), IPI_GET_CMD,
996 996 MISC_CMD, ip_sioctl_get_lifnum, NULL },
997 997 /* 131 */ { SIOCGLIFMUXID, sizeof (struct lifreq),
998 998 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_muxid, NULL },
999 999 /* 132 */ { SIOCSLIFMUXID, sizeof (struct lifreq),
1000 1000 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_muxid, NULL },
1001 1001 /* 133 */ { SIOCGLIFINDEX, sizeof (struct lifreq),
1002 1002 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lifindex, 0 },
1003 1003 /* 134 */ { SIOCSLIFINDEX, sizeof (struct lifreq),
1004 1004 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_slifindex, 0 },
1005 1005 /* 135 */ { SIOCSLIFTOKEN, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1006 1006 LIF_CMD, ip_sioctl_token, NULL },
1007 1007 /* 136 */ { SIOCGLIFTOKEN, sizeof (struct lifreq),
1008 1008 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_token, NULL },
1009 1009 /* 137 */ { SIOCSLIFSUBNET, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1010 1010 LIF_CMD, ip_sioctl_subnet, ip_sioctl_subnet_restart },
1011 1011 /* 138 */ { SIOCGLIFSUBNET, sizeof (struct lifreq),
1012 1012 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_subnet, NULL },
1013 1013 /* 139 */ { SIOCSLIFLNKINFO, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1014 1014 LIF_CMD, ip_sioctl_lnkinfo, NULL },
1015 1015
1016 1016 /* 140 */ { SIOCGLIFLNKINFO, sizeof (struct lifreq),
1017 1017 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lnkinfo, NULL },
1018 1018 /* 141 */ { SIOCLIFDELND, sizeof (struct lifreq), IPI_PRIV,
1019 1019 LIF_CMD, ip_siocdelndp_v6, NULL },
1020 1020 /* 142 */ { SIOCLIFGETND, sizeof (struct lifreq), IPI_GET_CMD,
1021 1021 LIF_CMD, ip_siocqueryndp_v6, NULL },
1022 1022 /* 143 */ { SIOCLIFSETND, sizeof (struct lifreq), IPI_PRIV,
1023 1023 LIF_CMD, ip_siocsetndp_v6, NULL },
1024 1024 /* 144 */ { SIOCTMYADDR, sizeof (struct sioc_addrreq), IPI_GET_CMD,
1025 1025 MISC_CMD, ip_sioctl_tmyaddr, NULL },
1026 1026 /* 145 */ { SIOCTONLINK, sizeof (struct sioc_addrreq), IPI_GET_CMD,
1027 1027 MISC_CMD, ip_sioctl_tonlink, NULL },
1028 1028 /* 146 */ { SIOCTMYSITE, sizeof (struct sioc_addrreq), 0,
1029 1029 MISC_CMD, ip_sioctl_tmysite, NULL },
1030 1030 /* 147 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1031 1031 /* 148 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1032 1032
1033 1033 /* Old *IPSECONFIG ioctls are now deprecated, now see spdsock.c */
1034 1034 /* 149 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1035 1035 /* 150 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1036 1036 /* 151 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1037 1037 /* 152 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1038 1038
1039 1039 /* 153 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1040 1040
1041 1041 /* 154 */ { SIOCGLIFBINDING, sizeof (struct lifreq), IPI_GET_CMD,
1042 1042 LIF_CMD, ip_sioctl_get_binding, NULL },
1043 1043 /* 155 */ { SIOCSLIFGROUPNAME, sizeof (struct lifreq),
1044 1044 IPI_PRIV | IPI_WR,
1045 1045 LIF_CMD, ip_sioctl_groupname, ip_sioctl_groupname },
1046 1046 /* 156 */ { SIOCGLIFGROUPNAME, sizeof (struct lifreq),
1047 1047 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_groupname, NULL },
1048 1048 /* 157 */ { SIOCGLIFGROUPINFO, sizeof (lifgroupinfo_t),
1049 1049 IPI_GET_CMD, MISC_CMD, ip_sioctl_groupinfo, NULL },
1050 1050
1051 1051 /* Leave 158-160 unused; used to be SIOC*IFARP ioctls */
1052 1052 /* 158 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1053 1053 /* 159 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1054 1054 /* 160 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1055 1055
1056 1056 /* 161 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1057 1057
1058 1058 /* These are handled in ip_sioctl_copyin_setup itself */
1059 1059 /* 162 */ { SIOCGIP6ADDRPOLICY, 0, IPI_NULL_BCONT,
1060 1060 MISC_CMD, NULL, NULL },
1061 1061 /* 163 */ { SIOCSIP6ADDRPOLICY, 0, IPI_PRIV | IPI_NULL_BCONT,
1062 1062 MISC_CMD, NULL, NULL },
1063 1063 /* 164 */ { SIOCGDSTINFO, 0, IPI_GET_CMD, MISC_CMD, NULL, NULL },
1064 1064
1065 1065 /* 165 */ { SIOCGLIFCONF, 0, IPI_GET_CMD, MISC_CMD,
1066 1066 ip_sioctl_get_lifconf, NULL },
1067 1067
1068 1068 /* 166 */ { SIOCSXARP, sizeof (struct xarpreq), IPI_PRIV | IPI_WR,
1069 1069 XARP_CMD, ip_sioctl_arp, NULL },
1070 1070 /* 167 */ { SIOCGXARP, sizeof (struct xarpreq), IPI_GET_CMD,
1071 1071 XARP_CMD, ip_sioctl_arp, NULL },
1072 1072 /* 168 */ { SIOCDXARP, sizeof (struct xarpreq), IPI_PRIV | IPI_WR,
1073 1073 XARP_CMD, ip_sioctl_arp, NULL },
1074 1074
1075 1075 /* SIOCPOPSOCKFS is not handled by IP */
1076 1076 /* 169 */ { IPI_DONTCARE /* SIOCPOPSOCKFS */, 0, 0, 0, NULL, NULL },
1077 1077
1078 1078 /* 170 */ { SIOCGLIFZONE, sizeof (struct lifreq),
1079 1079 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lifzone, NULL },
1080 1080 /* 171 */ { SIOCSLIFZONE, sizeof (struct lifreq),
1081 1081 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_slifzone,
1082 1082 ip_sioctl_slifzone_restart },
1083 1083 /* 172-174 are SCTP ioctls and not handled by IP */
1084 1084 /* 172 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1085 1085 /* 173 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1086 1086 /* 174 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1087 1087 /* 175 */ { SIOCGLIFUSESRC, sizeof (struct lifreq),
1088 1088 IPI_GET_CMD, LIF_CMD,
1089 1089 ip_sioctl_get_lifusesrc, 0 },
1090 1090 /* 176 */ { SIOCSLIFUSESRC, sizeof (struct lifreq),
1091 1091 IPI_PRIV | IPI_WR,
1092 1092 LIF_CMD, ip_sioctl_slifusesrc,
1093 1093 NULL },
1094 1094 /* 177 */ { SIOCGLIFSRCOF, 0, IPI_GET_CMD, MISC_CMD,
1095 1095 ip_sioctl_get_lifsrcof, NULL },
1096 1096 /* 178 */ { SIOCGMSFILTER, sizeof (struct group_filter), IPI_GET_CMD,
1097 1097 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1098 1098 /* 179 */ { SIOCSMSFILTER, sizeof (struct group_filter), 0,
1099 1099 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1100 1100 /* 180 */ { SIOCGIPMSFILTER, sizeof (struct ip_msfilter), IPI_GET_CMD,
1101 1101 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1102 1102 /* 181 */ { SIOCSIPMSFILTER, sizeof (struct ip_msfilter), 0,
1103 1103 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1104 1104 /* 182 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1105 1105 /* SIOCSENABLESDP is handled by SDP */
1106 1106 /* 183 */ { IPI_DONTCARE /* SIOCSENABLESDP */, 0, 0, 0, NULL, NULL },
1107 1107 /* 184 */ { IPI_DONTCARE /* SIOCSQPTR */, 0, 0, 0, NULL, NULL },
1108 1108 /* 185 */ { SIOCGIFHWADDR, sizeof (struct ifreq), IPI_GET_CMD,
1109 1109 IF_CMD, ip_sioctl_get_ifhwaddr, NULL },
1110 1110 /* 186 */ { IPI_DONTCARE /* SIOCGSTAMP */, 0, 0, 0, NULL, NULL },
1111 1111 /* 187 */ { SIOCILB, 0, IPI_PRIV | IPI_GET_CMD, MISC_CMD,
1112 1112 ip_sioctl_ilb_cmd, NULL },
1113 1113 /* 188 */ { SIOCGETPROP, 0, IPI_GET_CMD, 0, NULL, NULL },
1114 1114 /* 189 */ { SIOCSETPROP, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL},
1115 1115 /* 190 */ { SIOCGLIFDADSTATE, sizeof (struct lifreq),
1116 1116 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_dadstate, NULL },
1117 1117 /* 191 */ { SIOCSLIFPREFIX, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1118 1118 LIF_CMD, ip_sioctl_prefix, ip_sioctl_prefix_restart },
1119 1119 /* 192 */ { SIOCGLIFHWADDR, sizeof (struct lifreq), IPI_GET_CMD,
1120 1120 LIF_CMD, ip_sioctl_get_lifhwaddr, NULL }
1121 1121 };
1122 1122
1123 1123 int ip_ndx_ioctl_count = sizeof (ip_ndx_ioctl_table) / sizeof (ip_ioctl_cmd_t);
1124 1124
1125 1125 ip_ioctl_cmd_t ip_misc_ioctl_table[] = {
1126 1126 { I_LINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1127 1127 { I_UNLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1128 1128 { I_PLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1129 1129 { I_PUNLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1130 1130 { ND_GET, 0, 0, 0, NULL, NULL },
1131 1131 { ND_SET, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1132 1132 { IP_IOCTL, 0, 0, 0, NULL, NULL },
1133 1133 { SIOCGETVIFCNT, sizeof (struct sioc_vif_req), IPI_GET_CMD,
1134 1134 MISC_CMD, mrt_ioctl},
1135 1135 { SIOCGETSGCNT, sizeof (struct sioc_sg_req), IPI_GET_CMD,
1136 1136 MISC_CMD, mrt_ioctl},
1137 1137 { SIOCGETLSGCNT, sizeof (struct sioc_lsg_req), IPI_GET_CMD,
1138 1138 MISC_CMD, mrt_ioctl}
1139 1139 };
1140 1140
1141 1141 int ip_misc_ioctl_count =
1142 1142 sizeof (ip_misc_ioctl_table) / sizeof (ip_ioctl_cmd_t);
1143 1143
1144 1144 int conn_drain_nthreads; /* Number of drainers reqd. */
1145 1145 /* Settable in /etc/system */
1146 1146 /* Defined in ip_ire.c */
1147 1147 extern uint32_t ip_ire_max_bucket_cnt, ip6_ire_max_bucket_cnt;
1148 1148 extern uint32_t ip_ire_min_bucket_cnt, ip6_ire_min_bucket_cnt;
1149 1149 extern uint32_t ip_ire_mem_ratio, ip_ire_cpu_ratio;
1150 1150
1151 1151 static nv_t ire_nv_arr[] = {
1152 1152 { IRE_BROADCAST, "BROADCAST" },
1153 1153 { IRE_LOCAL, "LOCAL" },
1154 1154 { IRE_LOOPBACK, "LOOPBACK" },
1155 1155 { IRE_DEFAULT, "DEFAULT" },
1156 1156 { IRE_PREFIX, "PREFIX" },
1157 1157 { IRE_IF_NORESOLVER, "IF_NORESOL" },
1158 1158 { IRE_IF_RESOLVER, "IF_RESOLV" },
1159 1159 { IRE_IF_CLONE, "IF_CLONE" },
1160 1160 { IRE_HOST, "HOST" },
1161 1161 { IRE_MULTICAST, "MULTICAST" },
1162 1162 { IRE_NOROUTE, "NOROUTE" },
1163 1163 { 0 }
1164 1164 };
1165 1165
1166 1166 nv_t *ire_nv_tbl = ire_nv_arr;
1167 1167
1168 1168 /* Simple ICMP IP Header Template */
1169 1169 static ipha_t icmp_ipha = {
1170 1170 IP_SIMPLE_HDR_VERSION, 0, 0, 0, 0, 0, IPPROTO_ICMP
1171 1171 };
1172 1172
1173 1173 struct module_info ip_mod_info = {
1174 1174 IP_MOD_ID, IP_MOD_NAME, IP_MOD_MINPSZ, IP_MOD_MAXPSZ, IP_MOD_HIWAT,
1175 1175 IP_MOD_LOWAT
1176 1176 };
1177 1177
1178 1178 /*
1179 1179 * Duplicate static symbols within a module confuses mdb; so we avoid the
1180 1180 * problem by making the symbols here distinct from those in udp.c.
1181 1181 */
1182 1182
1183 1183 /*
1184 1184 * Entry points for IP as a device and as a module.
1185 1185 * We have separate open functions for the /dev/ip and /dev/ip6 devices.
1186 1186 */
1187 1187 static struct qinit iprinitv4 = {
1188 1188 ip_rput, NULL, ip_openv4, ip_close, NULL, &ip_mod_info
1189 1189 };
1190 1190
1191 1191 struct qinit iprinitv6 = {
1192 1192 ip_rput_v6, NULL, ip_openv6, ip_close, NULL, &ip_mod_info
1193 1193 };
1194 1194
1195 1195 static struct qinit ipwinit = {
1196 1196 ip_wput_nondata, ip_wsrv, NULL, NULL, NULL, &ip_mod_info
1197 1197 };
1198 1198
1199 1199 static struct qinit iplrinit = {
1200 1200 ip_lrput, NULL, ip_openv4, ip_close, NULL, &ip_mod_info
1201 1201 };
1202 1202
1203 1203 static struct qinit iplwinit = {
1204 1204 ip_lwput, NULL, NULL, NULL, NULL, &ip_mod_info
1205 1205 };
1206 1206
1207 1207 /* For AF_INET aka /dev/ip */
1208 1208 struct streamtab ipinfov4 = {
1209 1209 &iprinitv4, &ipwinit, &iplrinit, &iplwinit
1210 1210 };
1211 1211
1212 1212 /* For AF_INET6 aka /dev/ip6 */
1213 1213 struct streamtab ipinfov6 = {
1214 1214 &iprinitv6, &ipwinit, &iplrinit, &iplwinit
1215 1215 };
1216 1216
1217 1217 #ifdef DEBUG
1218 1218 boolean_t skip_sctp_cksum = B_FALSE;
1219 1219 #endif
1220 1220
1221 1221 /*
1222 1222 * Generate an ICMP fragmentation needed message.
1223 1223 * When called from ip_output side a minimal ip_recv_attr_t needs to be
1224 1224 * constructed by the caller.
1225 1225 */
1226 1226 void
1227 1227 icmp_frag_needed(mblk_t *mp, int mtu, ip_recv_attr_t *ira)
1228 1228 {
1229 1229 icmph_t icmph;
1230 1230 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
1231 1231
1232 1232 mp = icmp_pkt_err_ok(mp, ira);
1233 1233 if (mp == NULL)
1234 1234 return;
1235 1235
1236 1236 bzero(&icmph, sizeof (icmph_t));
1237 1237 icmph.icmph_type = ICMP_DEST_UNREACHABLE;
1238 1238 icmph.icmph_code = ICMP_FRAGMENTATION_NEEDED;
1239 1239 icmph.icmph_du_mtu = htons((uint16_t)mtu);
1240 1240 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutFragNeeded);
1241 1241 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDestUnreachs);
1242 1242
1243 1243 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
1244 1244 }
1245 1245
1246 1246 /*
1247 1247 * icmp_inbound_v4 deals with ICMP messages that are handled by IP.
1248 1248 * If the ICMP message is consumed by IP, i.e., it should not be delivered
1249 1249 * to any IPPROTO_ICMP raw sockets, then it returns NULL.
1250 1250 * Likewise, if the ICMP error is misformed (too short, etc), then it
1251 1251 * returns NULL. The caller uses this to determine whether or not to send
1252 1252 * to raw sockets.
1253 1253 *
1254 1254 * All error messages are passed to the matching transport stream.
1255 1255 *
1256 1256 * The following cases are handled by icmp_inbound:
1257 1257 * 1) It needs to send a reply back and possibly delivering it
1258 1258 * to the "interested" upper clients.
1259 1259 * 2) Return the mblk so that the caller can pass it to the RAW socket clients.
1260 1260 * 3) It needs to change some values in IP only.
1261 1261 * 4) It needs to change some values in IP and upper layers e.g TCP
1262 1262 * by delivering an error to the upper layers.
1263 1263 *
1264 1264 * We handle the above three cases in the context of IPsec in the
1265 1265 * following way :
1266 1266 *
1267 1267 * 1) Send the reply back in the same way as the request came in.
1268 1268 * If it came in encrypted, it goes out encrypted. If it came in
1269 1269 * clear, it goes out in clear. Thus, this will prevent chosen
1270 1270 * plain text attack.
1271 1271 * 2) The client may or may not expect things to come in secure.
1272 1272 * If it comes in secure, the policy constraints are checked
1273 1273 * before delivering it to the upper layers. If it comes in
1274 1274 * clear, ipsec_inbound_accept_clear will decide whether to
1275 1275 * accept this in clear or not. In both the cases, if the returned
1276 1276 * message (IP header + 8 bytes) that caused the icmp message has
1277 1277 * AH/ESP headers, it is sent up to AH/ESP for validation before
1278 1278 * sending up. If there are only 8 bytes of returned message, then
1279 1279 * upper client will not be notified.
1280 1280 * 3) Check with global policy to see whether it matches the constaints.
1281 1281 * But this will be done only if icmp_accept_messages_in_clear is
1282 1282 * zero.
1283 1283 * 4) If we need to change both in IP and ULP, then the decision taken
1284 1284 * while affecting the values in IP and while delivering up to TCP
1285 1285 * should be the same.
1286 1286 *
1287 1287 * There are two cases.
1288 1288 *
1289 1289 * a) If we reject data at the IP layer (ipsec_check_global_policy()
1290 1290 * failed), we will not deliver it to the ULP, even though they
1291 1291 * are *willing* to accept in *clear*. This is fine as our global
1292 1292 * disposition to icmp messages asks us reject the datagram.
1293 1293 *
1294 1294 * b) If we accept data at the IP layer (ipsec_check_global_policy()
1295 1295 * succeeded or icmp_accept_messages_in_clear is 1), and not able
1296 1296 * to deliver it to ULP (policy failed), it can lead to
1297 1297 * consistency problems. The cases known at this time are
1298 1298 * ICMP_DESTINATION_UNREACHABLE messages with following code
1299 1299 * values :
1300 1300 *
1301 1301 * - ICMP_FRAGMENTATION_NEEDED : IP adapts to the new value
1302 1302 * and Upper layer rejects. Then the communication will
1303 1303 * come to a stop. This is solved by making similar decisions
1304 1304 * at both levels. Currently, when we are unable to deliver
1305 1305 * to the Upper Layer (due to policy failures) while IP has
1306 1306 * adjusted dce_pmtu, the next outbound datagram would
1307 1307 * generate a local ICMP_FRAGMENTATION_NEEDED message - which
1308 1308 * will be with the right level of protection. Thus the right
1309 1309 * value will be communicated even if we are not able to
1310 1310 * communicate when we get from the wire initially. But this
1311 1311 * assumes there would be at least one outbound datagram after
1312 1312 * IP has adjusted its dce_pmtu value. To make things
1313 1313 * simpler, we accept in clear after the validation of
1314 1314 * AH/ESP headers.
1315 1315 *
1316 1316 * - Other ICMP ERRORS : We may not be able to deliver it to the
1317 1317 * upper layer depending on the level of protection the upper
1318 1318 * layer expects and the disposition in ipsec_inbound_accept_clear().
1319 1319 * ipsec_inbound_accept_clear() decides whether a given ICMP error
1320 1320 * should be accepted in clear when the Upper layer expects secure.
1321 1321 * Thus the communication may get aborted by some bad ICMP
1322 1322 * packets.
1323 1323 */
1324 1324 mblk_t *
1325 1325 icmp_inbound_v4(mblk_t *mp, ip_recv_attr_t *ira)
1326 1326 {
1327 1327 icmph_t *icmph;
1328 1328 ipha_t *ipha; /* Outer header */
1329 1329 int ip_hdr_length; /* Outer header length */
1330 1330 boolean_t interested;
1331 1331 ipif_t *ipif;
1332 1332 uint32_t ts;
1333 1333 uint32_t *tsp;
1334 1334 timestruc_t now;
1335 1335 ill_t *ill = ira->ira_ill;
1336 1336 ip_stack_t *ipst = ill->ill_ipst;
1337 1337 zoneid_t zoneid = ira->ira_zoneid;
1338 1338 int len_needed;
1339 1339 mblk_t *mp_ret = NULL;
1340 1340
1341 1341 ipha = (ipha_t *)mp->b_rptr;
1342 1342
1343 1343 BUMP_MIB(&ipst->ips_icmp_mib, icmpInMsgs);
1344 1344
1345 1345 ip_hdr_length = ira->ira_ip_hdr_length;
1346 1346 if ((mp->b_wptr - mp->b_rptr) < (ip_hdr_length + ICMPH_SIZE)) {
1347 1347 if (ira->ira_pktlen < (ip_hdr_length + ICMPH_SIZE)) {
1348 1348 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
1349 1349 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
1350 1350 freemsg(mp);
1351 1351 return (NULL);
1352 1352 }
1353 1353 /* Last chance to get real. */
1354 1354 ipha = ip_pullup(mp, ip_hdr_length + ICMPH_SIZE, ira);
1355 1355 if (ipha == NULL) {
1356 1356 BUMP_MIB(&ipst->ips_icmp_mib, icmpInErrors);
1357 1357 freemsg(mp);
1358 1358 return (NULL);
1359 1359 }
1360 1360 }
1361 1361
1362 1362 /* The IP header will always be a multiple of four bytes */
1363 1363 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1364 1364 ip2dbg(("icmp_inbound_v4: type %d code %d\n", icmph->icmph_type,
1365 1365 icmph->icmph_code));
1366 1366
1367 1367 /*
1368 1368 * We will set "interested" to "true" if we should pass a copy to
1369 1369 * the transport or if we handle the packet locally.
1370 1370 */
1371 1371 interested = B_FALSE;
1372 1372 switch (icmph->icmph_type) {
1373 1373 case ICMP_ECHO_REPLY:
1374 1374 BUMP_MIB(&ipst->ips_icmp_mib, icmpInEchoReps);
1375 1375 break;
1376 1376 case ICMP_DEST_UNREACHABLE:
1377 1377 if (icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED)
1378 1378 BUMP_MIB(&ipst->ips_icmp_mib, icmpInFragNeeded);
1379 1379 interested = B_TRUE; /* Pass up to transport */
1380 1380 BUMP_MIB(&ipst->ips_icmp_mib, icmpInDestUnreachs);
1381 1381 break;
1382 1382 case ICMP_SOURCE_QUENCH:
1383 1383 interested = B_TRUE; /* Pass up to transport */
1384 1384 BUMP_MIB(&ipst->ips_icmp_mib, icmpInSrcQuenchs);
1385 1385 break;
1386 1386 case ICMP_REDIRECT:
1387 1387 if (!ipst->ips_ip_ignore_redirect)
1388 1388 interested = B_TRUE;
1389 1389 BUMP_MIB(&ipst->ips_icmp_mib, icmpInRedirects);
1390 1390 break;
1391 1391 case ICMP_ECHO_REQUEST:
1392 1392 /*
1393 1393 * Whether to respond to echo requests that come in as IP
1394 1394 * broadcasts or as IP multicast is subject to debate
1395 1395 * (what isn't?). We aim to please, you pick it.
1396 1396 * Default is do it.
1397 1397 */
1398 1398 if (ira->ira_flags & IRAF_MULTICAST) {
1399 1399 /* multicast: respond based on tunable */
1400 1400 interested = ipst->ips_ip_g_resp_to_echo_mcast;
1401 1401 } else if (ira->ira_flags & IRAF_BROADCAST) {
1402 1402 /* broadcast: respond based on tunable */
1403 1403 interested = ipst->ips_ip_g_resp_to_echo_bcast;
1404 1404 } else {
1405 1405 /* unicast: always respond */
1406 1406 interested = B_TRUE;
1407 1407 }
1408 1408 BUMP_MIB(&ipst->ips_icmp_mib, icmpInEchos);
1409 1409 if (!interested) {
1410 1410 /* We never pass these to RAW sockets */
1411 1411 freemsg(mp);
1412 1412 return (NULL);
1413 1413 }
1414 1414
1415 1415 /* Check db_ref to make sure we can modify the packet. */
1416 1416 if (mp->b_datap->db_ref > 1) {
1417 1417 mblk_t *mp1;
1418 1418
1419 1419 mp1 = copymsg(mp);
1420 1420 freemsg(mp);
1421 1421 if (!mp1) {
1422 1422 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1423 1423 return (NULL);
1424 1424 }
1425 1425 mp = mp1;
1426 1426 ipha = (ipha_t *)mp->b_rptr;
1427 1427 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1428 1428 }
1429 1429 icmph->icmph_type = ICMP_ECHO_REPLY;
1430 1430 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutEchoReps);
1431 1431 icmp_send_reply_v4(mp, ipha, icmph, ira);
1432 1432 return (NULL);
1433 1433
1434 1434 case ICMP_ROUTER_ADVERTISEMENT:
1435 1435 case ICMP_ROUTER_SOLICITATION:
1436 1436 break;
1437 1437 case ICMP_TIME_EXCEEDED:
1438 1438 interested = B_TRUE; /* Pass up to transport */
1439 1439 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimeExcds);
1440 1440 break;
1441 1441 case ICMP_PARAM_PROBLEM:
1442 1442 interested = B_TRUE; /* Pass up to transport */
1443 1443 BUMP_MIB(&ipst->ips_icmp_mib, icmpInParmProbs);
1444 1444 break;
1445 1445 case ICMP_TIME_STAMP_REQUEST:
1446 1446 /* Response to Time Stamp Requests is local policy. */
1447 1447 if (ipst->ips_ip_g_resp_to_timestamp) {
1448 1448 if (ira->ira_flags & IRAF_MULTIBROADCAST)
1449 1449 interested =
1450 1450 ipst->ips_ip_g_resp_to_timestamp_bcast;
1451 1451 else
1452 1452 interested = B_TRUE;
1453 1453 }
1454 1454 if (!interested) {
1455 1455 /* We never pass these to RAW sockets */
1456 1456 freemsg(mp);
1457 1457 return (NULL);
1458 1458 }
1459 1459
1460 1460 /* Make sure we have enough of the packet */
1461 1461 len_needed = ip_hdr_length + ICMPH_SIZE +
1462 1462 3 * sizeof (uint32_t);
1463 1463
1464 1464 if (mp->b_wptr - mp->b_rptr < len_needed) {
1465 1465 ipha = ip_pullup(mp, len_needed, ira);
1466 1466 if (ipha == NULL) {
1467 1467 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1468 1468 ip_drop_input("ipIfStatsInDiscards - ip_pullup",
1469 1469 mp, ill);
1470 1470 freemsg(mp);
1471 1471 return (NULL);
1472 1472 }
1473 1473 /* Refresh following the pullup. */
1474 1474 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1475 1475 }
1476 1476 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimestamps);
1477 1477 /* Check db_ref to make sure we can modify the packet. */
1478 1478 if (mp->b_datap->db_ref > 1) {
1479 1479 mblk_t *mp1;
1480 1480
1481 1481 mp1 = copymsg(mp);
1482 1482 freemsg(mp);
1483 1483 if (!mp1) {
1484 1484 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1485 1485 return (NULL);
1486 1486 }
1487 1487 mp = mp1;
1488 1488 ipha = (ipha_t *)mp->b_rptr;
1489 1489 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1490 1490 }
1491 1491 icmph->icmph_type = ICMP_TIME_STAMP_REPLY;
1492 1492 tsp = (uint32_t *)&icmph[1];
1493 1493 tsp++; /* Skip past 'originate time' */
1494 1494 /* Compute # of milliseconds since midnight */
1495 1495 gethrestime(&now);
1496 1496 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
1497 1497 NSEC2MSEC(now.tv_nsec);
1498 1498 *tsp++ = htonl(ts); /* Lay in 'receive time' */
1499 1499 *tsp++ = htonl(ts); /* Lay in 'send time' */
1500 1500 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutTimestampReps);
1501 1501 icmp_send_reply_v4(mp, ipha, icmph, ira);
1502 1502 return (NULL);
1503 1503
1504 1504 case ICMP_TIME_STAMP_REPLY:
1505 1505 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimestampReps);
1506 1506 break;
1507 1507 case ICMP_INFO_REQUEST:
1508 1508 /* Per RFC 1122 3.2.2.7, ignore this. */
1509 1509 case ICMP_INFO_REPLY:
1510 1510 break;
1511 1511 case ICMP_ADDRESS_MASK_REQUEST:
1512 1512 if (ira->ira_flags & IRAF_MULTIBROADCAST) {
1513 1513 interested =
1514 1514 ipst->ips_ip_respond_to_address_mask_broadcast;
1515 1515 } else {
1516 1516 interested = B_TRUE;
1517 1517 }
1518 1518 if (!interested) {
1519 1519 /* We never pass these to RAW sockets */
1520 1520 freemsg(mp);
1521 1521 return (NULL);
1522 1522 }
1523 1523 len_needed = ip_hdr_length + ICMPH_SIZE + IP_ADDR_LEN;
1524 1524 if (mp->b_wptr - mp->b_rptr < len_needed) {
1525 1525 ipha = ip_pullup(mp, len_needed, ira);
1526 1526 if (ipha == NULL) {
1527 1527 BUMP_MIB(ill->ill_ip_mib,
1528 1528 ipIfStatsInTruncatedPkts);
1529 1529 ip_drop_input("ipIfStatsInTruncatedPkts", mp,
1530 1530 ill);
1531 1531 freemsg(mp);
1532 1532 return (NULL);
1533 1533 }
1534 1534 /* Refresh following the pullup. */
1535 1535 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1536 1536 }
1537 1537 BUMP_MIB(&ipst->ips_icmp_mib, icmpInAddrMasks);
1538 1538 /* Check db_ref to make sure we can modify the packet. */
1539 1539 if (mp->b_datap->db_ref > 1) {
1540 1540 mblk_t *mp1;
1541 1541
1542 1542 mp1 = copymsg(mp);
1543 1543 freemsg(mp);
1544 1544 if (!mp1) {
1545 1545 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1546 1546 return (NULL);
1547 1547 }
1548 1548 mp = mp1;
1549 1549 ipha = (ipha_t *)mp->b_rptr;
1550 1550 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1551 1551 }
1552 1552 /*
1553 1553 * Need the ipif with the mask be the same as the source
1554 1554 * address of the mask reply. For unicast we have a specific
1555 1555 * ipif. For multicast/broadcast we only handle onlink
1556 1556 * senders, and use the source address to pick an ipif.
1557 1557 */
1558 1558 ipif = ipif_lookup_addr(ipha->ipha_dst, ill, zoneid, ipst);
1559 1559 if (ipif == NULL) {
1560 1560 /* Broadcast or multicast */
1561 1561 ipif = ipif_lookup_remote(ill, ipha->ipha_src, zoneid);
1562 1562 if (ipif == NULL) {
1563 1563 freemsg(mp);
1564 1564 return (NULL);
1565 1565 }
1566 1566 }
1567 1567 icmph->icmph_type = ICMP_ADDRESS_MASK_REPLY;
1568 1568 bcopy(&ipif->ipif_net_mask, &icmph[1], IP_ADDR_LEN);
1569 1569 ipif_refrele(ipif);
1570 1570 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutAddrMaskReps);
1571 1571 icmp_send_reply_v4(mp, ipha, icmph, ira);
1572 1572 return (NULL);
1573 1573
1574 1574 case ICMP_ADDRESS_MASK_REPLY:
1575 1575 BUMP_MIB(&ipst->ips_icmp_mib, icmpInAddrMaskReps);
1576 1576 break;
1577 1577 default:
1578 1578 interested = B_TRUE; /* Pass up to transport */
1579 1579 BUMP_MIB(&ipst->ips_icmp_mib, icmpInUnknowns);
1580 1580 break;
1581 1581 }
1582 1582 /*
1583 1583 * See if there is an ICMP client to avoid an extra copymsg/freemsg
1584 1584 * if there isn't one.
1585 1585 */
1586 1586 if (ipst->ips_ipcl_proto_fanout_v4[IPPROTO_ICMP].connf_head != NULL) {
1587 1587 /* If there is an ICMP client and we want one too, copy it. */
1588 1588
1589 1589 if (!interested) {
1590 1590 /* Caller will deliver to RAW sockets */
1591 1591 return (mp);
1592 1592 }
1593 1593 mp_ret = copymsg(mp);
1594 1594 if (mp_ret == NULL) {
1595 1595 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1596 1596 ip_drop_input("ipIfStatsInDiscards - copymsg", mp, ill);
1597 1597 }
1598 1598 } else if (!interested) {
1599 1599 /* Neither we nor raw sockets are interested. Drop packet now */
1600 1600 freemsg(mp);
1601 1601 return (NULL);
1602 1602 }
1603 1603
1604 1604 /*
1605 1605 * ICMP error or redirect packet. Make sure we have enough of
1606 1606 * the header and that db_ref == 1 since we might end up modifying
1607 1607 * the packet.
1608 1608 */
1609 1609 if (mp->b_cont != NULL) {
1610 1610 if (ip_pullup(mp, -1, ira) == NULL) {
1611 1611 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1612 1612 ip_drop_input("ipIfStatsInDiscards - ip_pullup",
1613 1613 mp, ill);
1614 1614 freemsg(mp);
1615 1615 return (mp_ret);
1616 1616 }
1617 1617 }
1618 1618
1619 1619 if (mp->b_datap->db_ref > 1) {
1620 1620 mblk_t *mp1;
1621 1621
1622 1622 mp1 = copymsg(mp);
1623 1623 if (mp1 == NULL) {
1624 1624 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1625 1625 ip_drop_input("ipIfStatsInDiscards - copymsg", mp, ill);
1626 1626 freemsg(mp);
1627 1627 return (mp_ret);
1628 1628 }
1629 1629 freemsg(mp);
1630 1630 mp = mp1;
1631 1631 }
1632 1632
1633 1633 /*
1634 1634 * In case mp has changed, verify the message before any further
1635 1635 * processes.
1636 1636 */
1637 1637 ipha = (ipha_t *)mp->b_rptr;
1638 1638 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1639 1639 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
1640 1640 freemsg(mp);
1641 1641 return (mp_ret);
1642 1642 }
1643 1643
1644 1644 switch (icmph->icmph_type) {
1645 1645 case ICMP_REDIRECT:
1646 1646 icmp_redirect_v4(mp, ipha, icmph, ira);
1647 1647 break;
1648 1648 case ICMP_DEST_UNREACHABLE:
1649 1649 if (icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED) {
1650 1650 /* Update DCE and adjust MTU is icmp header if needed */
1651 1651 icmp_inbound_too_big_v4(icmph, ira);
1652 1652 }
1653 1653 /* FALLTHROUGH */
1654 1654 default:
1655 1655 icmp_inbound_error_fanout_v4(mp, icmph, ira);
1656 1656 break;
1657 1657 }
1658 1658 return (mp_ret);
1659 1659 }
1660 1660
1661 1661 /*
1662 1662 * Send an ICMP echo, timestamp or address mask reply.
1663 1663 * The caller has already updated the payload part of the packet.
1664 1664 * We handle the ICMP checksum, IP source address selection and feed
1665 1665 * the packet into ip_output_simple.
1666 1666 */
1667 1667 static void
1668 1668 icmp_send_reply_v4(mblk_t *mp, ipha_t *ipha, icmph_t *icmph,
1669 1669 ip_recv_attr_t *ira)
1670 1670 {
1671 1671 uint_t ip_hdr_length = ira->ira_ip_hdr_length;
1672 1672 ill_t *ill = ira->ira_ill;
1673 1673 ip_stack_t *ipst = ill->ill_ipst;
1674 1674 ip_xmit_attr_t ixas;
1675 1675
1676 1676 /* Send out an ICMP packet */
1677 1677 icmph->icmph_checksum = 0;
1678 1678 icmph->icmph_checksum = IP_CSUM(mp, ip_hdr_length, 0);
1679 1679 /* Reset time to live. */
1680 1680 ipha->ipha_ttl = ipst->ips_ip_def_ttl;
1681 1681 {
1682 1682 /* Swap source and destination addresses */
1683 1683 ipaddr_t tmp;
1684 1684
1685 1685 tmp = ipha->ipha_src;
1686 1686 ipha->ipha_src = ipha->ipha_dst;
1687 1687 ipha->ipha_dst = tmp;
1688 1688 }
1689 1689 ipha->ipha_ident = 0;
1690 1690 if (!IS_SIMPLE_IPH(ipha))
1691 1691 icmp_options_update(ipha);
1692 1692
1693 1693 bzero(&ixas, sizeof (ixas));
1694 1694 ixas.ixa_flags = IXAF_BASIC_SIMPLE_V4;
1695 1695 ixas.ixa_zoneid = ira->ira_zoneid;
1696 1696 ixas.ixa_cred = kcred;
1697 1697 ixas.ixa_cpid = NOPID;
1698 1698 ixas.ixa_tsl = ira->ira_tsl; /* Behave as a multi-level responder */
1699 1699 ixas.ixa_ifindex = 0;
1700 1700 ixas.ixa_ipst = ipst;
1701 1701 ixas.ixa_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1702 1702
1703 1703 if (!(ira->ira_flags & IRAF_IPSEC_SECURE)) {
1704 1704 /*
1705 1705 * This packet should go out the same way as it
1706 1706 * came in i.e in clear, independent of the IPsec policy
1707 1707 * for transmitting packets.
1708 1708 */
1709 1709 ixas.ixa_flags |= IXAF_NO_IPSEC;
1710 1710 } else {
1711 1711 if (!ipsec_in_to_out(ira, &ixas, mp, ipha, NULL)) {
1712 1712 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1713 1713 /* Note: mp already consumed and ip_drop_packet done */
1714 1714 return;
1715 1715 }
1716 1716 }
1717 1717 if (ira->ira_flags & IRAF_MULTIBROADCAST) {
1718 1718 /*
1719 1719 * Not one or our addresses (IRE_LOCALs), thus we let
1720 1720 * ip_output_simple pick the source.
1721 1721 */
1722 1722 ipha->ipha_src = INADDR_ANY;
1723 1723 ixas.ixa_flags |= IXAF_SET_SOURCE;
1724 1724 }
1725 1725 /* Should we send with DF and use dce_pmtu? */
1726 1726 if (ipst->ips_ipv4_icmp_return_pmtu) {
1727 1727 ixas.ixa_flags |= IXAF_PMTU_DISCOVERY;
1728 1728 ipha->ipha_fragment_offset_and_flags |= IPH_DF_HTONS;
1729 1729 }
1730 1730
1731 1731 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutMsgs);
1732 1732
1733 1733 (void) ip_output_simple(mp, &ixas);
1734 1734 ixa_cleanup(&ixas);
1735 1735 }
1736 1736
1737 1737 /*
1738 1738 * Verify the ICMP messages for either for ICMP error or redirect packet.
1739 1739 * The caller should have fully pulled up the message. If it's a redirect
1740 1740 * packet, only basic checks on IP header will be done; otherwise, verify
1741 1741 * the packet by looking at the included ULP header.
1742 1742 *
1743 1743 * Called before icmp_inbound_error_fanout_v4 is called.
1744 1744 */
1745 1745 static boolean_t
1746 1746 icmp_inbound_verify_v4(mblk_t *mp, icmph_t *icmph, ip_recv_attr_t *ira)
1747 1747 {
1748 1748 ill_t *ill = ira->ira_ill;
1749 1749 int hdr_length;
1750 1750 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
1751 1751 conn_t *connp;
1752 1752 ipha_t *ipha; /* Inner IP header */
1753 1753
1754 1754 ipha = (ipha_t *)&icmph[1];
1755 1755 if ((uchar_t *)ipha + IP_SIMPLE_HDR_LENGTH > mp->b_wptr)
1756 1756 goto truncated;
1757 1757
1758 1758 hdr_length = IPH_HDR_LENGTH(ipha);
1759 1759
1760 1760 if ((IPH_HDR_VERSION(ipha) != IPV4_VERSION))
1761 1761 goto discard_pkt;
1762 1762
1763 1763 if (hdr_length < sizeof (ipha_t))
1764 1764 goto truncated;
1765 1765
1766 1766 if ((uchar_t *)ipha + hdr_length > mp->b_wptr)
1767 1767 goto truncated;
1768 1768
1769 1769 /*
1770 1770 * Stop here for ICMP_REDIRECT.
1771 1771 */
1772 1772 if (icmph->icmph_type == ICMP_REDIRECT)
1773 1773 return (B_TRUE);
1774 1774
1775 1775 /*
1776 1776 * ICMP errors only.
1777 1777 */
1778 1778 switch (ipha->ipha_protocol) {
1779 1779 case IPPROTO_UDP:
1780 1780 /*
1781 1781 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1782 1782 * transport header.
1783 1783 */
1784 1784 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1785 1785 mp->b_wptr)
1786 1786 goto truncated;
1787 1787 break;
1788 1788 case IPPROTO_TCP: {
1789 1789 tcpha_t *tcpha;
1790 1790
1791 1791 /*
1792 1792 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1793 1793 * transport header.
1794 1794 */
1795 1795 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1796 1796 mp->b_wptr)
1797 1797 goto truncated;
1798 1798
1799 1799 tcpha = (tcpha_t *)((uchar_t *)ipha + hdr_length);
1800 1800 connp = ipcl_tcp_lookup_reversed_ipv4(ipha, tcpha, TCPS_LISTEN,
1801 1801 ipst);
1802 1802 if (connp == NULL)
1803 1803 goto discard_pkt;
1804 1804
1805 1805 if ((connp->conn_verifyicmp != NULL) &&
1806 1806 !connp->conn_verifyicmp(connp, tcpha, icmph, NULL, ira)) {
1807 1807 CONN_DEC_REF(connp);
1808 1808 goto discard_pkt;
1809 1809 }
1810 1810 CONN_DEC_REF(connp);
1811 1811 break;
1812 1812 }
1813 1813 case IPPROTO_SCTP:
1814 1814 /*
1815 1815 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1816 1816 * transport header.
1817 1817 */
1818 1818 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1819 1819 mp->b_wptr)
1820 1820 goto truncated;
1821 1821 break;
1822 1822 case IPPROTO_ESP:
1823 1823 case IPPROTO_AH:
1824 1824 break;
1825 1825 case IPPROTO_ENCAP:
1826 1826 if ((uchar_t *)ipha + hdr_length + sizeof (ipha_t) >
1827 1827 mp->b_wptr)
1828 1828 goto truncated;
1829 1829 break;
1830 1830 default:
1831 1831 break;
1832 1832 }
1833 1833
1834 1834 return (B_TRUE);
1835 1835
1836 1836 discard_pkt:
1837 1837 /* Bogus ICMP error. */
1838 1838 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1839 1839 return (B_FALSE);
1840 1840
1841 1841 truncated:
1842 1842 /* We pulled up everthing already. Must be truncated */
1843 1843 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
1844 1844 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
1845 1845 return (B_FALSE);
1846 1846 }
1847 1847
1848 1848 /* Table from RFC 1191 */
1849 1849 static int icmp_frag_size_table[] =
1850 1850 { 32000, 17914, 8166, 4352, 2002, 1496, 1006, 508, 296, 68 };
1851 1851
1852 1852 /*
1853 1853 * Process received ICMP Packet too big.
1854 1854 * Just handles the DCE create/update, including using the above table of
1855 1855 * PMTU guesses. The caller is responsible for validating the packet before
1856 1856 * passing it in and also to fanout the ICMP error to any matching transport
1857 1857 * conns. Assumes the message has been fully pulled up and verified.
1858 1858 *
1859 1859 * Before getting here, the caller has called icmp_inbound_verify_v4()
1860 1860 * that should have verified with ULP to prevent undoing the changes we're
1861 1861 * going to make to DCE. For example, TCP might have verified that the packet
1862 1862 * which generated error is in the send window.
1863 1863 *
1864 1864 * In some cases modified this MTU in the ICMP header packet; the caller
1865 1865 * should pass to the matching ULP after this returns.
1866 1866 */
1867 1867 static void
1868 1868 icmp_inbound_too_big_v4(icmph_t *icmph, ip_recv_attr_t *ira)
1869 1869 {
1870 1870 dce_t *dce;
1871 1871 int old_mtu;
1872 1872 int mtu, orig_mtu;
1873 1873 ipaddr_t dst;
1874 1874 boolean_t disable_pmtud;
1875 1875 ill_t *ill = ira->ira_ill;
1876 1876 ip_stack_t *ipst = ill->ill_ipst;
1877 1877 uint_t hdr_length;
1878 1878 ipha_t *ipha;
1879 1879
1880 1880 /* Caller already pulled up everything. */
1881 1881 ipha = (ipha_t *)&icmph[1];
1882 1882 ASSERT(icmph->icmph_type == ICMP_DEST_UNREACHABLE &&
1883 1883 icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED);
1884 1884 ASSERT(ill != NULL);
1885 1885
1886 1886 hdr_length = IPH_HDR_LENGTH(ipha);
1887 1887
1888 1888 /*
1889 1889 * We handle path MTU for source routed packets since the DCE
1890 1890 * is looked up using the final destination.
1891 1891 */
1892 1892 dst = ip_get_dst(ipha);
1893 1893
1894 1894 dce = dce_lookup_and_add_v4(dst, ipst);
1895 1895 if (dce == NULL) {
1896 1896 /* Couldn't add a unique one - ENOMEM */
1897 1897 ip1dbg(("icmp_inbound_too_big_v4: no dce for 0x%x\n",
1898 1898 ntohl(dst)));
1899 1899 return;
1900 1900 }
1901 1901
1902 1902 /* Check for MTU discovery advice as described in RFC 1191 */
1903 1903 mtu = ntohs(icmph->icmph_du_mtu);
1904 1904 orig_mtu = mtu;
1905 1905 disable_pmtud = B_FALSE;
1906 1906
1907 1907 mutex_enter(&dce->dce_lock);
1908 1908 if (dce->dce_flags & DCEF_PMTU)
1909 1909 old_mtu = dce->dce_pmtu;
1910 1910 else
1911 1911 old_mtu = ill->ill_mtu;
1912 1912
1913 1913 if (icmph->icmph_du_zero != 0 || mtu < ipst->ips_ip_pmtu_min) {
1914 1914 uint32_t length;
1915 1915 int i;
1916 1916
1917 1917 /*
1918 1918 * Use the table from RFC 1191 to figure out
1919 1919 * the next "plateau" based on the length in
1920 1920 * the original IP packet.
1921 1921 */
1922 1922 length = ntohs(ipha->ipha_length);
1923 1923 DTRACE_PROBE2(ip4__pmtu__guess, dce_t *, dce,
1924 1924 uint32_t, length);
1925 1925 if (old_mtu <= length &&
1926 1926 old_mtu >= length - hdr_length) {
1927 1927 /*
1928 1928 * Handle broken BSD 4.2 systems that
1929 1929 * return the wrong ipha_length in ICMP
1930 1930 * errors.
1931 1931 */
1932 1932 ip1dbg(("Wrong mtu: sent %d, dce %d\n",
1933 1933 length, old_mtu));
1934 1934 length -= hdr_length;
1935 1935 }
1936 1936 for (i = 0; i < A_CNT(icmp_frag_size_table); i++) {
1937 1937 if (length > icmp_frag_size_table[i])
1938 1938 break;
1939 1939 }
1940 1940 if (i == A_CNT(icmp_frag_size_table)) {
1941 1941 /* Smaller than IP_MIN_MTU! */
1942 1942 ip1dbg(("Too big for packet size %d\n",
1943 1943 length));
1944 1944 disable_pmtud = B_TRUE;
1945 1945 mtu = ipst->ips_ip_pmtu_min;
1946 1946 } else {
1947 1947 mtu = icmp_frag_size_table[i];
1948 1948 ip1dbg(("Calculated mtu %d, packet size %d, "
1949 1949 "before %d\n", mtu, length, old_mtu));
1950 1950 if (mtu < ipst->ips_ip_pmtu_min) {
1951 1951 mtu = ipst->ips_ip_pmtu_min;
1952 1952 disable_pmtud = B_TRUE;
1953 1953 }
1954 1954 }
1955 1955 }
1956 1956 if (disable_pmtud)
1957 1957 dce->dce_flags |= DCEF_TOO_SMALL_PMTU;
1958 1958 else
1959 1959 dce->dce_flags &= ~DCEF_TOO_SMALL_PMTU;
1960 1960
1961 1961 dce->dce_pmtu = MIN(old_mtu, mtu);
1962 1962 /* Prepare to send the new max frag size for the ULP. */
1963 1963 icmph->icmph_du_zero = 0;
1964 1964 icmph->icmph_du_mtu = htons((uint16_t)dce->dce_pmtu);
1965 1965 DTRACE_PROBE4(ip4__pmtu__change, icmph_t *, icmph, dce_t *,
1966 1966 dce, int, orig_mtu, int, mtu);
1967 1967
1968 1968 /* We now have a PMTU for sure */
1969 1969 dce->dce_flags |= DCEF_PMTU;
1970 1970 dce->dce_last_change_time = TICK_TO_SEC(ddi_get_lbolt64());
1971 1971 mutex_exit(&dce->dce_lock);
1972 1972 /*
1973 1973 * After dropping the lock the new value is visible to everyone.
1974 1974 * Then we bump the generation number so any cached values reinspect
1975 1975 * the dce_t.
1976 1976 */
1977 1977 dce_increment_generation(dce);
1978 1978 dce_refrele(dce);
1979 1979 }
1980 1980
1981 1981 /*
1982 1982 * If the packet in error is Self-Encapsulated, icmp_inbound_error_fanout_v4
1983 1983 * calls this function.
1984 1984 */
1985 1985 static mblk_t *
1986 1986 icmp_inbound_self_encap_error_v4(mblk_t *mp, ipha_t *ipha, ipha_t *in_ipha)
1987 1987 {
1988 1988 int length;
1989 1989
1990 1990 ASSERT(mp->b_datap->db_type == M_DATA);
1991 1991
1992 1992 /* icmp_inbound_v4 has already pulled up the whole error packet */
1993 1993 ASSERT(mp->b_cont == NULL);
1994 1994
1995 1995 /*
1996 1996 * The length that we want to overlay is the inner header
1997 1997 * and what follows it.
1998 1998 */
1999 1999 length = msgdsize(mp) - ((uchar_t *)in_ipha - mp->b_rptr);
2000 2000
2001 2001 /*
2002 2002 * Overlay the inner header and whatever follows it over the
2003 2003 * outer header.
2004 2004 */
2005 2005 bcopy((uchar_t *)in_ipha, (uchar_t *)ipha, length);
2006 2006
2007 2007 /* Adjust for what we removed */
2008 2008 mp->b_wptr -= (uchar_t *)in_ipha - (uchar_t *)ipha;
2009 2009 return (mp);
2010 2010 }
2011 2011
2012 2012 /*
2013 2013 * Try to pass the ICMP message upstream in case the ULP cares.
2014 2014 *
2015 2015 * If the packet that caused the ICMP error is secure, we send
2016 2016 * it to AH/ESP to make sure that the attached packet has a
2017 2017 * valid association. ipha in the code below points to the
2018 2018 * IP header of the packet that caused the error.
2019 2019 *
2020 2020 * For IPsec cases, we let the next-layer-up (which has access to
2021 2021 * cached policy on the conn_t, or can query the SPD directly)
2022 2022 * subtract out any IPsec overhead if they must. We therefore make no
2023 2023 * adjustments here for IPsec overhead.
2024 2024 *
2025 2025 * IFN could have been generated locally or by some router.
2026 2026 *
2027 2027 * LOCAL : ire_send_wire (before calling ipsec_out_process) can call
2028 2028 * icmp_frag_needed/icmp_pkt2big_v6 to generated a local IFN.
2029 2029 * This happens because IP adjusted its value of MTU on an
2030 2030 * earlier IFN message and could not tell the upper layer,
2031 2031 * the new adjusted value of MTU e.g. Packet was encrypted
2032 2032 * or there was not enough information to fanout to upper
2033 2033 * layers. Thus on the next outbound datagram, ire_send_wire
2034 2034 * generates the IFN, where IPsec processing has *not* been
2035 2035 * done.
2036 2036 *
2037 2037 * Note that we retain ixa_fragsize across IPsec thus once
2038 2038 * we have picking ixa_fragsize and entered ipsec_out_process we do
2039 2039 * no change the fragsize even if the path MTU changes before
2040 2040 * we reach ip_output_post_ipsec.
2041 2041 *
2042 2042 * In the local case, IRAF_LOOPBACK will be set indicating
2043 2043 * that IFN was generated locally.
2044 2044 *
2045 2045 * ROUTER : IFN could be secure or non-secure.
2046 2046 *
2047 2047 * * SECURE : We use the IPSEC_IN to fanout to AH/ESP if the
2048 2048 * packet in error has AH/ESP headers to validate the AH/ESP
2049 2049 * headers. AH/ESP will verify whether there is a valid SA or
2050 2050 * not and send it back. We will fanout again if we have more
2051 2051 * data in the packet.
2052 2052 *
2053 2053 * If the packet in error does not have AH/ESP, we handle it
2054 2054 * like any other case.
2055 2055 *
2056 2056 * * NON_SECURE : If the packet in error has AH/ESP headers, we send it
2057 2057 * up to AH/ESP for validation. AH/ESP will verify whether there is a
2058 2058 * valid SA or not and send it back. We will fanout again if
2059 2059 * we have more data in the packet.
2060 2060 *
2061 2061 * If the packet in error does not have AH/ESP, we handle it
2062 2062 * like any other case.
2063 2063 *
2064 2064 * The caller must have called icmp_inbound_verify_v4.
2065 2065 */
2066 2066 static void
2067 2067 icmp_inbound_error_fanout_v4(mblk_t *mp, icmph_t *icmph, ip_recv_attr_t *ira)
2068 2068 {
2069 2069 uint16_t *up; /* Pointer to ports in ULP header */
2070 2070 uint32_t ports; /* reversed ports for fanout */
2071 2071 ipha_t ripha; /* With reversed addresses */
2072 2072 ipha_t *ipha; /* Inner IP header */
2073 2073 uint_t hdr_length; /* Inner IP header length */
2074 2074 tcpha_t *tcpha;
2075 2075 conn_t *connp;
2076 2076 ill_t *ill = ira->ira_ill;
2077 2077 ip_stack_t *ipst = ill->ill_ipst;
2078 2078 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
2079 2079 ill_t *rill = ira->ira_rill;
2080 2080
2081 2081 /* Caller already pulled up everything. */
2082 2082 ipha = (ipha_t *)&icmph[1];
2083 2083 ASSERT((uchar_t *)&ipha[1] <= mp->b_wptr);
2084 2084 ASSERT(mp->b_cont == NULL);
2085 2085
2086 2086 hdr_length = IPH_HDR_LENGTH(ipha);
2087 2087 ira->ira_protocol = ipha->ipha_protocol;
2088 2088
2089 2089 /*
2090 2090 * We need a separate IP header with the source and destination
2091 2091 * addresses reversed to do fanout/classification because the ipha in
2092 2092 * the ICMP error is in the form we sent it out.
2093 2093 */
2094 2094 ripha.ipha_src = ipha->ipha_dst;
2095 2095 ripha.ipha_dst = ipha->ipha_src;
2096 2096 ripha.ipha_protocol = ipha->ipha_protocol;
2097 2097 ripha.ipha_version_and_hdr_length = ipha->ipha_version_and_hdr_length;
2098 2098
2099 2099 ip2dbg(("icmp_inbound_error_v4: proto %d %x to %x: %d/%d\n",
2100 2100 ripha.ipha_protocol, ntohl(ipha->ipha_src),
2101 2101 ntohl(ipha->ipha_dst),
2102 2102 icmph->icmph_type, icmph->icmph_code));
2103 2103
2104 2104 switch (ipha->ipha_protocol) {
2105 2105 case IPPROTO_UDP:
2106 2106 up = (uint16_t *)((uchar_t *)ipha + hdr_length);
2107 2107
2108 2108 /* Attempt to find a client stream based on port. */
2109 2109 ip2dbg(("icmp_inbound_error_v4: UDP ports %d to %d\n",
2110 2110 ntohs(up[0]), ntohs(up[1])));
2111 2111
2112 2112 /* Note that we send error to all matches. */
2113 2113 ira->ira_flags |= IRAF_ICMP_ERROR;
2114 2114 ip_fanout_udp_multi_v4(mp, &ripha, up[0], up[1], ira);
2115 2115 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2116 2116 return;
2117 2117
2118 2118 case IPPROTO_TCP:
2119 2119 /*
2120 2120 * Find a TCP client stream for this packet.
2121 2121 * Note that we do a reverse lookup since the header is
2122 2122 * in the form we sent it out.
2123 2123 */
2124 2124 tcpha = (tcpha_t *)((uchar_t *)ipha + hdr_length);
2125 2125 connp = ipcl_tcp_lookup_reversed_ipv4(ipha, tcpha, TCPS_LISTEN,
2126 2126 ipst);
2127 2127 if (connp == NULL)
2128 2128 goto discard_pkt;
2129 2129
2130 2130 if (CONN_INBOUND_POLICY_PRESENT(connp, ipss) ||
2131 2131 (ira->ira_flags & IRAF_IPSEC_SECURE)) {
2132 2132 mp = ipsec_check_inbound_policy(mp, connp,
2133 2133 ipha, NULL, ira);
2134 2134 if (mp == NULL) {
2135 2135 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2136 2136 /* Note that mp is NULL */
2137 2137 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2138 2138 CONN_DEC_REF(connp);
2139 2139 return;
2140 2140 }
2141 2141 }
2142 2142
2143 2143 ira->ira_flags |= IRAF_ICMP_ERROR;
2144 2144 ira->ira_ill = ira->ira_rill = NULL;
2145 2145 if (IPCL_IS_TCP(connp)) {
2146 2146 SQUEUE_ENTER_ONE(connp->conn_sqp, mp,
2147 2147 connp->conn_recvicmp, connp, ira, SQ_FILL,
2148 2148 SQTAG_TCP_INPUT_ICMP_ERR);
2149 2149 } else {
2150 2150 /* Not TCP; must be SOCK_RAW, IPPROTO_TCP */
2151 2151 (connp->conn_recv)(connp, mp, NULL, ira);
2152 2152 CONN_DEC_REF(connp);
2153 2153 }
2154 2154 ira->ira_ill = ill;
2155 2155 ira->ira_rill = rill;
2156 2156 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2157 2157 return;
2158 2158
2159 2159 case IPPROTO_SCTP:
2160 2160 up = (uint16_t *)((uchar_t *)ipha + hdr_length);
2161 2161 /* Find a SCTP client stream for this packet. */
2162 2162 ((uint16_t *)&ports)[0] = up[1];
2163 2163 ((uint16_t *)&ports)[1] = up[0];
2164 2164
2165 2165 ira->ira_flags |= IRAF_ICMP_ERROR;
2166 2166 ip_fanout_sctp(mp, &ripha, NULL, ports, ira);
2167 2167 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2168 2168 return;
2169 2169
2170 2170 case IPPROTO_ESP:
2171 2171 case IPPROTO_AH:
2172 2172 if (!ipsec_loaded(ipss)) {
2173 2173 ip_proto_not_sup(mp, ira);
2174 2174 return;
2175 2175 }
2176 2176
2177 2177 if (ipha->ipha_protocol == IPPROTO_ESP)
2178 2178 mp = ipsecesp_icmp_error(mp, ira);
2179 2179 else
2180 2180 mp = ipsecah_icmp_error(mp, ira);
2181 2181 if (mp == NULL)
2182 2182 return;
2183 2183
2184 2184 /* Just in case ipsec didn't preserve the NULL b_cont */
2185 2185 if (mp->b_cont != NULL) {
2186 2186 if (!pullupmsg(mp, -1))
2187 2187 goto discard_pkt;
2188 2188 }
2189 2189
2190 2190 /*
2191 2191 * Note that ira_pktlen and ira_ip_hdr_length are no longer
2192 2192 * correct, but we don't use them any more here.
2193 2193 *
2194 2194 * If succesful, the mp has been modified to not include
2195 2195 * the ESP/AH header so we can fanout to the ULP's icmp
2196 2196 * error handler.
2197 2197 */
2198 2198 if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH)
2199 2199 goto truncated;
2200 2200
2201 2201 /* Verify the modified message before any further processes. */
2202 2202 ipha = (ipha_t *)mp->b_rptr;
2203 2203 hdr_length = IPH_HDR_LENGTH(ipha);
2204 2204 icmph = (icmph_t *)&mp->b_rptr[hdr_length];
2205 2205 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
2206 2206 freemsg(mp);
2207 2207 return;
2208 2208 }
2209 2209
2210 2210 icmp_inbound_error_fanout_v4(mp, icmph, ira);
2211 2211 return;
2212 2212
2213 2213 case IPPROTO_ENCAP: {
2214 2214 /* Look for self-encapsulated packets that caused an error */
2215 2215 ipha_t *in_ipha;
2216 2216
2217 2217 /*
2218 2218 * Caller has verified that length has to be
2219 2219 * at least the size of IP header.
2220 2220 */
2221 2221 ASSERT(hdr_length >= sizeof (ipha_t));
2222 2222 /*
2223 2223 * Check the sanity of the inner IP header like
2224 2224 * we did for the outer header.
2225 2225 */
2226 2226 in_ipha = (ipha_t *)((uchar_t *)ipha + hdr_length);
2227 2227 if ((IPH_HDR_VERSION(in_ipha) != IPV4_VERSION)) {
2228 2228 goto discard_pkt;
2229 2229 }
2230 2230 if (IPH_HDR_LENGTH(in_ipha) < sizeof (ipha_t)) {
2231 2231 goto discard_pkt;
2232 2232 }
2233 2233 /* Check for Self-encapsulated tunnels */
2234 2234 if (in_ipha->ipha_src == ipha->ipha_src &&
2235 2235 in_ipha->ipha_dst == ipha->ipha_dst) {
2236 2236
2237 2237 mp = icmp_inbound_self_encap_error_v4(mp, ipha,
2238 2238 in_ipha);
2239 2239 if (mp == NULL)
2240 2240 goto discard_pkt;
2241 2241
2242 2242 /*
2243 2243 * Just in case self_encap didn't preserve the NULL
2244 2244 * b_cont
2245 2245 */
2246 2246 if (mp->b_cont != NULL) {
2247 2247 if (!pullupmsg(mp, -1))
2248 2248 goto discard_pkt;
2249 2249 }
2250 2250 /*
2251 2251 * Note that ira_pktlen and ira_ip_hdr_length are no
2252 2252 * longer correct, but we don't use them any more here.
2253 2253 */
2254 2254 if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH)
2255 2255 goto truncated;
2256 2256
2257 2257 /*
2258 2258 * Verify the modified message before any further
2259 2259 * processes.
2260 2260 */
2261 2261 ipha = (ipha_t *)mp->b_rptr;
2262 2262 hdr_length = IPH_HDR_LENGTH(ipha);
2263 2263 icmph = (icmph_t *)&mp->b_rptr[hdr_length];
2264 2264 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
2265 2265 freemsg(mp);
2266 2266 return;
2267 2267 }
2268 2268
2269 2269 /*
2270 2270 * The packet in error is self-encapsualted.
2271 2271 * And we are finding it further encapsulated
2272 2272 * which we could not have possibly generated.
2273 2273 */
2274 2274 if (ipha->ipha_protocol == IPPROTO_ENCAP) {
2275 2275 goto discard_pkt;
2276 2276 }
2277 2277 icmp_inbound_error_fanout_v4(mp, icmph, ira);
2278 2278 return;
2279 2279 }
2280 2280 /* No self-encapsulated */
2281 2281 }
2282 2282 /* FALLTHROUGH */
2283 2283 case IPPROTO_IPV6:
2284 2284 if ((connp = ipcl_iptun_classify_v4(&ripha.ipha_src,
2285 2285 &ripha.ipha_dst, ipst)) != NULL) {
2286 2286 ira->ira_flags |= IRAF_ICMP_ERROR;
2287 2287 connp->conn_recvicmp(connp, mp, NULL, ira);
2288 2288 CONN_DEC_REF(connp);
2289 2289 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2290 2290 return;
2291 2291 }
2292 2292 /*
2293 2293 * No IP tunnel is interested, fallthrough and see
2294 2294 * if a raw socket will want it.
2295 2295 */
2296 2296 /* FALLTHROUGH */
2297 2297 default:
2298 2298 ira->ira_flags |= IRAF_ICMP_ERROR;
2299 2299 ip_fanout_proto_v4(mp, &ripha, ira);
2300 2300 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2301 2301 return;
2302 2302 }
2303 2303 /* NOTREACHED */
2304 2304 discard_pkt:
2305 2305 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2306 2306 ip1dbg(("icmp_inbound_error_fanout_v4: drop pkt\n"));
2307 2307 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2308 2308 freemsg(mp);
2309 2309 return;
2310 2310
2311 2311 truncated:
2312 2312 /* We pulled up everthing already. Must be truncated */
2313 2313 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
2314 2314 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
2315 2315 freemsg(mp);
2316 2316 }
2317 2317
2318 2318 /*
2319 2319 * Common IP options parser.
2320 2320 *
2321 2321 * Setup routine: fill in *optp with options-parsing state, then
2322 2322 * tail-call ipoptp_next to return the first option.
2323 2323 */
2324 2324 uint8_t
2325 2325 ipoptp_first(ipoptp_t *optp, ipha_t *ipha)
2326 2326 {
2327 2327 uint32_t totallen; /* total length of all options */
2328 2328
2329 2329 totallen = ipha->ipha_version_and_hdr_length -
2330 2330 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
2331 2331 totallen <<= 2;
2332 2332 optp->ipoptp_next = (uint8_t *)(&ipha[1]);
2333 2333 optp->ipoptp_end = optp->ipoptp_next + totallen;
2334 2334 optp->ipoptp_flags = 0;
2335 2335 return (ipoptp_next(optp));
2336 2336 }
2337 2337
2338 2338 /* Like above but without an ipha_t */
2339 2339 uint8_t
2340 2340 ipoptp_first2(ipoptp_t *optp, uint32_t totallen, uint8_t *opt)
2341 2341 {
2342 2342 optp->ipoptp_next = opt;
2343 2343 optp->ipoptp_end = optp->ipoptp_next + totallen;
2344 2344 optp->ipoptp_flags = 0;
2345 2345 return (ipoptp_next(optp));
2346 2346 }
2347 2347
2348 2348 /*
2349 2349 * Common IP options parser: extract next option.
2350 2350 */
2351 2351 uint8_t
2352 2352 ipoptp_next(ipoptp_t *optp)
2353 2353 {
2354 2354 uint8_t *end = optp->ipoptp_end;
2355 2355 uint8_t *cur = optp->ipoptp_next;
2356 2356 uint8_t opt, len, pointer;
2357 2357
2358 2358 /*
2359 2359 * If cur > end already, then the ipoptp_end or ipoptp_next pointer
2360 2360 * has been corrupted.
2361 2361 */
2362 2362 ASSERT(cur <= end);
2363 2363
2364 2364 if (cur == end)
2365 2365 return (IPOPT_EOL);
2366 2366
2367 2367 opt = cur[IPOPT_OPTVAL];
2368 2368
2369 2369 /*
2370 2370 * Skip any NOP options.
2371 2371 */
2372 2372 while (opt == IPOPT_NOP) {
2373 2373 cur++;
2374 2374 if (cur == end)
2375 2375 return (IPOPT_EOL);
2376 2376 opt = cur[IPOPT_OPTVAL];
2377 2377 }
2378 2378
2379 2379 if (opt == IPOPT_EOL)
2380 2380 return (IPOPT_EOL);
2381 2381
2382 2382 /*
2383 2383 * Option requiring a length.
2384 2384 */
2385 2385 if ((cur + 1) >= end) {
2386 2386 optp->ipoptp_flags |= IPOPTP_ERROR;
2387 2387 return (IPOPT_EOL);
2388 2388 }
2389 2389 len = cur[IPOPT_OLEN];
2390 2390 if (len < 2) {
2391 2391 optp->ipoptp_flags |= IPOPTP_ERROR;
2392 2392 return (IPOPT_EOL);
2393 2393 }
2394 2394 optp->ipoptp_cur = cur;
2395 2395 optp->ipoptp_len = len;
2396 2396 optp->ipoptp_next = cur + len;
2397 2397 if (cur + len > end) {
2398 2398 optp->ipoptp_flags |= IPOPTP_ERROR;
2399 2399 return (IPOPT_EOL);
2400 2400 }
2401 2401
2402 2402 /*
2403 2403 * For the options which require a pointer field, make sure
2404 2404 * its there, and make sure it points to either something
2405 2405 * inside this option, or the end of the option.
2406 2406 */
2407 2407 switch (opt) {
2408 2408 case IPOPT_RR:
2409 2409 case IPOPT_TS:
2410 2410 case IPOPT_LSRR:
2411 2411 case IPOPT_SSRR:
2412 2412 if (len <= IPOPT_OFFSET) {
2413 2413 optp->ipoptp_flags |= IPOPTP_ERROR;
2414 2414 return (opt);
2415 2415 }
2416 2416 pointer = cur[IPOPT_OFFSET];
2417 2417 if (pointer - 1 > len) {
2418 2418 optp->ipoptp_flags |= IPOPTP_ERROR;
2419 2419 return (opt);
2420 2420 }
2421 2421 break;
2422 2422 }
2423 2423
2424 2424 /*
2425 2425 * Sanity check the pointer field based on the type of the
2426 2426 * option.
2427 2427 */
2428 2428 switch (opt) {
2429 2429 case IPOPT_RR:
2430 2430 case IPOPT_SSRR:
2431 2431 case IPOPT_LSRR:
2432 2432 if (pointer < IPOPT_MINOFF_SR)
2433 2433 optp->ipoptp_flags |= IPOPTP_ERROR;
2434 2434 break;
2435 2435 case IPOPT_TS:
2436 2436 if (pointer < IPOPT_MINOFF_IT)
2437 2437 optp->ipoptp_flags |= IPOPTP_ERROR;
2438 2438 /*
2439 2439 * Note that the Internet Timestamp option also
2440 2440 * contains two four bit fields (the Overflow field,
2441 2441 * and the Flag field), which follow the pointer
2442 2442 * field. We don't need to check that these fields
2443 2443 * fall within the length of the option because this
2444 2444 * was implicitely done above. We've checked that the
2445 2445 * pointer value is at least IPOPT_MINOFF_IT, and that
2446 2446 * it falls within the option. Since IPOPT_MINOFF_IT >
2447 2447 * IPOPT_POS_OV_FLG, we don't need the explicit check.
2448 2448 */
2449 2449 ASSERT(len > IPOPT_POS_OV_FLG);
2450 2450 break;
2451 2451 }
2452 2452
2453 2453 return (opt);
2454 2454 }
2455 2455
2456 2456 /*
2457 2457 * Use the outgoing IP header to create an IP_OPTIONS option the way
2458 2458 * it was passed down from the application.
2459 2459 *
2460 2460 * This is compatible with BSD in that it returns
2461 2461 * the reverse source route with the final destination
2462 2462 * as the last entry. The first 4 bytes of the option
2463 2463 * will contain the final destination.
2464 2464 */
2465 2465 int
2466 2466 ip_opt_get_user(conn_t *connp, uchar_t *buf)
2467 2467 {
2468 2468 ipoptp_t opts;
2469 2469 uchar_t *opt;
2470 2470 uint8_t optval;
2471 2471 uint8_t optlen;
2472 2472 uint32_t len = 0;
2473 2473 uchar_t *buf1 = buf;
2474 2474 uint32_t totallen;
2475 2475 ipaddr_t dst;
2476 2476 ip_pkt_t *ipp = &connp->conn_xmit_ipp;
2477 2477
2478 2478 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
2479 2479 return (0);
2480 2480
2481 2481 totallen = ipp->ipp_ipv4_options_len;
2482 2482 if (totallen & 0x3)
2483 2483 return (0);
2484 2484
2485 2485 buf += IP_ADDR_LEN; /* Leave room for final destination */
2486 2486 len += IP_ADDR_LEN;
2487 2487 bzero(buf1, IP_ADDR_LEN);
2488 2488
2489 2489 dst = connp->conn_faddr_v4;
2490 2490
2491 2491 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
2492 2492 optval != IPOPT_EOL;
2493 2493 optval = ipoptp_next(&opts)) {
2494 2494 int off;
2495 2495
2496 2496 opt = opts.ipoptp_cur;
2497 2497 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
2498 2498 break;
2499 2499 }
2500 2500 optlen = opts.ipoptp_len;
2501 2501
2502 2502 switch (optval) {
2503 2503 case IPOPT_SSRR:
2504 2504 case IPOPT_LSRR:
2505 2505
2506 2506 /*
2507 2507 * Insert destination as the first entry in the source
2508 2508 * route and move down the entries on step.
2509 2509 * The last entry gets placed at buf1.
2510 2510 */
2511 2511 buf[IPOPT_OPTVAL] = optval;
2512 2512 buf[IPOPT_OLEN] = optlen;
2513 2513 buf[IPOPT_OFFSET] = optlen;
2514 2514
2515 2515 off = optlen - IP_ADDR_LEN;
2516 2516 if (off < 0) {
2517 2517 /* No entries in source route */
2518 2518 break;
2519 2519 }
2520 2520 /* Last entry in source route if not already set */
2521 2521 if (dst == INADDR_ANY)
2522 2522 bcopy(opt + off, buf1, IP_ADDR_LEN);
2523 2523 off -= IP_ADDR_LEN;
2524 2524
2525 2525 while (off > 0) {
2526 2526 bcopy(opt + off,
2527 2527 buf + off + IP_ADDR_LEN,
2528 2528 IP_ADDR_LEN);
2529 2529 off -= IP_ADDR_LEN;
2530 2530 }
2531 2531 /* ipha_dst into first slot */
2532 2532 bcopy(&dst, buf + off + IP_ADDR_LEN,
2533 2533 IP_ADDR_LEN);
2534 2534 buf += optlen;
2535 2535 len += optlen;
2536 2536 break;
2537 2537
2538 2538 default:
2539 2539 bcopy(opt, buf, optlen);
2540 2540 buf += optlen;
2541 2541 len += optlen;
2542 2542 break;
2543 2543 }
2544 2544 }
2545 2545 done:
2546 2546 /* Pad the resulting options */
2547 2547 while (len & 0x3) {
2548 2548 *buf++ = IPOPT_EOL;
2549 2549 len++;
2550 2550 }
2551 2551 return (len);
2552 2552 }
2553 2553
2554 2554 /*
2555 2555 * Update any record route or timestamp options to include this host.
2556 2556 * Reverse any source route option.
2557 2557 * This routine assumes that the options are well formed i.e. that they
2558 2558 * have already been checked.
2559 2559 */
2560 2560 static void
2561 2561 icmp_options_update(ipha_t *ipha)
2562 2562 {
2563 2563 ipoptp_t opts;
2564 2564 uchar_t *opt;
2565 2565 uint8_t optval;
2566 2566 ipaddr_t src; /* Our local address */
2567 2567 ipaddr_t dst;
2568 2568
2569 2569 ip2dbg(("icmp_options_update\n"));
2570 2570 src = ipha->ipha_src;
2571 2571 dst = ipha->ipha_dst;
2572 2572
2573 2573 for (optval = ipoptp_first(&opts, ipha);
2574 2574 optval != IPOPT_EOL;
2575 2575 optval = ipoptp_next(&opts)) {
2576 2576 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
2577 2577 opt = opts.ipoptp_cur;
2578 2578 ip2dbg(("icmp_options_update: opt %d, len %d\n",
2579 2579 optval, opts.ipoptp_len));
2580 2580 switch (optval) {
2581 2581 int off1, off2;
2582 2582 case IPOPT_SSRR:
2583 2583 case IPOPT_LSRR:
2584 2584 /*
2585 2585 * Reverse the source route. The first entry
2586 2586 * should be the next to last one in the current
2587 2587 * source route (the last entry is our address).
2588 2588 * The last entry should be the final destination.
2589 2589 */
2590 2590 off1 = IPOPT_MINOFF_SR - 1;
2591 2591 off2 = opt[IPOPT_OFFSET] - IP_ADDR_LEN - 1;
2592 2592 if (off2 < 0) {
2593 2593 /* No entries in source route */
2594 2594 ip1dbg((
2595 2595 "icmp_options_update: bad src route\n"));
2596 2596 break;
2597 2597 }
2598 2598 bcopy((char *)opt + off2, &dst, IP_ADDR_LEN);
2599 2599 bcopy(&ipha->ipha_dst, (char *)opt + off2, IP_ADDR_LEN);
2600 2600 bcopy(&dst, &ipha->ipha_dst, IP_ADDR_LEN);
2601 2601 off2 -= IP_ADDR_LEN;
2602 2602
2603 2603 while (off1 < off2) {
2604 2604 bcopy((char *)opt + off1, &src, IP_ADDR_LEN);
2605 2605 bcopy((char *)opt + off2, (char *)opt + off1,
2606 2606 IP_ADDR_LEN);
2607 2607 bcopy(&src, (char *)opt + off2, IP_ADDR_LEN);
2608 2608 off1 += IP_ADDR_LEN;
2609 2609 off2 -= IP_ADDR_LEN;
2610 2610 }
2611 2611 opt[IPOPT_OFFSET] = IPOPT_MINOFF_SR;
2612 2612 break;
2613 2613 }
2614 2614 }
2615 2615 }
2616 2616
2617 2617 /*
2618 2618 * Process received ICMP Redirect messages.
2619 2619 * Assumes the caller has verified that the headers are in the pulled up mblk.
2620 2620 * Consumes mp.
2621 2621 */
2622 2622 static void
2623 2623 icmp_redirect_v4(mblk_t *mp, ipha_t *ipha, icmph_t *icmph, ip_recv_attr_t *ira)
2624 2624 {
2625 2625 ire_t *ire, *nire;
2626 2626 ire_t *prev_ire;
2627 2627 ipaddr_t src, dst, gateway;
2628 2628 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2629 2629 ipha_t *inner_ipha; /* Inner IP header */
2630 2630
2631 2631 /* Caller already pulled up everything. */
2632 2632 inner_ipha = (ipha_t *)&icmph[1];
2633 2633 src = ipha->ipha_src;
2634 2634 dst = inner_ipha->ipha_dst;
2635 2635 gateway = icmph->icmph_rd_gateway;
2636 2636 /* Make sure the new gateway is reachable somehow. */
2637 2637 ire = ire_ftable_lookup_v4(gateway, 0, 0, IRE_ONLINK, NULL,
2638 2638 ALL_ZONES, NULL, MATCH_IRE_TYPE, 0, ipst, NULL);
2639 2639 /*
2640 2640 * Make sure we had a route for the dest in question and that
2641 2641 * that route was pointing to the old gateway (the source of the
2642 2642 * redirect packet.)
2643 2643 * We do longest match and then compare ire_gateway_addr below.
2644 2644 */
2645 2645 prev_ire = ire_ftable_lookup_v4(dst, 0, 0, 0, NULL, ALL_ZONES,
2646 2646 NULL, MATCH_IRE_DSTONLY, 0, ipst, NULL);
2647 2647 /*
2648 2648 * Check that
2649 2649 * the redirect was not from ourselves
2650 2650 * the new gateway and the old gateway are directly reachable
2651 2651 */
2652 2652 if (prev_ire == NULL || ire == NULL ||
2653 2653 (prev_ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK)) ||
2654 2654 (prev_ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) ||
2655 2655 !(ire->ire_type & IRE_IF_ALL) ||
2656 2656 prev_ire->ire_gateway_addr != src) {
2657 2657 BUMP_MIB(&ipst->ips_icmp_mib, icmpInBadRedirects);
2658 2658 ip_drop_input("icmpInBadRedirects - ire", mp, ira->ira_ill);
2659 2659 freemsg(mp);
2660 2660 if (ire != NULL)
2661 2661 ire_refrele(ire);
2662 2662 if (prev_ire != NULL)
2663 2663 ire_refrele(prev_ire);
2664 2664 return;
2665 2665 }
2666 2666
2667 2667 ire_refrele(prev_ire);
2668 2668 ire_refrele(ire);
2669 2669
2670 2670 /*
2671 2671 * TODO: more precise handling for cases 0, 2, 3, the latter two
2672 2672 * require TOS routing
2673 2673 */
2674 2674 switch (icmph->icmph_code) {
2675 2675 case 0:
2676 2676 case 1:
2677 2677 /* TODO: TOS specificity for cases 2 and 3 */
2678 2678 case 2:
2679 2679 case 3:
2680 2680 break;
2681 2681 default:
2682 2682 BUMP_MIB(&ipst->ips_icmp_mib, icmpInBadRedirects);
2683 2683 ip_drop_input("icmpInBadRedirects - code", mp, ira->ira_ill);
2684 2684 freemsg(mp);
2685 2685 return;
2686 2686 }
2687 2687 /*
2688 2688 * Create a Route Association. This will allow us to remember that
2689 2689 * someone we believe told us to use the particular gateway.
2690 2690 */
2691 2691 ire = ire_create(
2692 2692 (uchar_t *)&dst, /* dest addr */
2693 2693 (uchar_t *)&ip_g_all_ones, /* mask */
2694 2694 (uchar_t *)&gateway, /* gateway addr */
2695 2695 IRE_HOST,
2696 2696 NULL, /* ill */
2697 2697 ALL_ZONES,
2698 2698 (RTF_DYNAMIC | RTF_GATEWAY | RTF_HOST),
2699 2699 NULL, /* tsol_gc_t */
2700 2700 ipst);
2701 2701
2702 2702 if (ire == NULL) {
2703 2703 freemsg(mp);
2704 2704 return;
2705 2705 }
2706 2706 nire = ire_add(ire);
2707 2707 /* Check if it was a duplicate entry */
2708 2708 if (nire != NULL && nire != ire) {
2709 2709 ASSERT(nire->ire_identical_ref > 1);
2710 2710 ire_delete(nire);
2711 2711 ire_refrele(nire);
2712 2712 nire = NULL;
2713 2713 }
2714 2714 ire = nire;
2715 2715 if (ire != NULL) {
2716 2716 ire_refrele(ire); /* Held in ire_add */
2717 2717
2718 2718 /* tell routing sockets that we received a redirect */
2719 2719 ip_rts_change(RTM_REDIRECT, dst, gateway, IP_HOST_MASK, 0, src,
2720 2720 (RTF_DYNAMIC | RTF_GATEWAY | RTF_HOST), 0,
2721 2721 (RTA_DST | RTA_GATEWAY | RTA_NETMASK | RTA_AUTHOR), ipst);
2722 2722 }
2723 2723
2724 2724 /*
2725 2725 * Delete any existing IRE_HOST type redirect ires for this destination.
2726 2726 * This together with the added IRE has the effect of
2727 2727 * modifying an existing redirect.
2728 2728 */
2729 2729 prev_ire = ire_ftable_lookup_v4(dst, 0, src, IRE_HOST, NULL,
2730 2730 ALL_ZONES, NULL, (MATCH_IRE_GW | MATCH_IRE_TYPE), 0, ipst, NULL);
2731 2731 if (prev_ire != NULL) {
2732 2732 if (prev_ire ->ire_flags & RTF_DYNAMIC)
2733 2733 ire_delete(prev_ire);
2734 2734 ire_refrele(prev_ire);
2735 2735 }
2736 2736
2737 2737 freemsg(mp);
2738 2738 }
2739 2739
2740 2740 /*
2741 2741 * Generate an ICMP parameter problem message.
2742 2742 * When called from ip_output side a minimal ip_recv_attr_t needs to be
2743 2743 * constructed by the caller.
2744 2744 */
2745 2745 static void
2746 2746 icmp_param_problem(mblk_t *mp, uint8_t ptr, ip_recv_attr_t *ira)
2747 2747 {
2748 2748 icmph_t icmph;
2749 2749 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2750 2750
2751 2751 mp = icmp_pkt_err_ok(mp, ira);
2752 2752 if (mp == NULL)
2753 2753 return;
2754 2754
2755 2755 bzero(&icmph, sizeof (icmph_t));
2756 2756 icmph.icmph_type = ICMP_PARAM_PROBLEM;
2757 2757 icmph.icmph_pp_ptr = ptr;
2758 2758 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutParmProbs);
2759 2759 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
2760 2760 }
2761 2761
2762 2762 /*
2763 2763 * Build and ship an IPv4 ICMP message using the packet data in mp, and
2764 2764 * the ICMP header pointed to by "stuff". (May be called as writer.)
2765 2765 * Note: assumes that icmp_pkt_err_ok has been called to verify that
2766 2766 * an icmp error packet can be sent.
2767 2767 * Assigns an appropriate source address to the packet. If ipha_dst is
2768 2768 * one of our addresses use it for source. Otherwise let ip_output_simple
2769 2769 * pick the source address.
2770 2770 */
2771 2771 static void
2772 2772 icmp_pkt(mblk_t *mp, void *stuff, size_t len, ip_recv_attr_t *ira)
2773 2773 {
2774 2774 ipaddr_t dst;
2775 2775 icmph_t *icmph;
2776 2776 ipha_t *ipha;
2777 2777 uint_t len_needed;
2778 2778 size_t msg_len;
2779 2779 mblk_t *mp1;
2780 2780 ipaddr_t src;
2781 2781 ire_t *ire;
2782 2782 ip_xmit_attr_t ixas;
2783 2783 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2784 2784
2785 2785 ipha = (ipha_t *)mp->b_rptr;
2786 2786
2787 2787 bzero(&ixas, sizeof (ixas));
2788 2788 ixas.ixa_flags = IXAF_BASIC_SIMPLE_V4;
2789 2789 ixas.ixa_zoneid = ira->ira_zoneid;
2790 2790 ixas.ixa_ifindex = 0;
2791 2791 ixas.ixa_ipst = ipst;
2792 2792 ixas.ixa_cred = kcred;
2793 2793 ixas.ixa_cpid = NOPID;
2794 2794 ixas.ixa_tsl = ira->ira_tsl; /* Behave as a multi-level responder */
2795 2795 ixas.ixa_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
2796 2796
2797 2797 if (ira->ira_flags & IRAF_IPSEC_SECURE) {
2798 2798 /*
2799 2799 * Apply IPsec based on how IPsec was applied to
2800 2800 * the packet that had the error.
2801 2801 *
2802 2802 * If it was an outbound packet that caused the ICMP
2803 2803 * error, then the caller will have setup the IRA
2804 2804 * appropriately.
2805 2805 */
2806 2806 if (!ipsec_in_to_out(ira, &ixas, mp, ipha, NULL)) {
2807 2807 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
2808 2808 /* Note: mp already consumed and ip_drop_packet done */
2809 2809 return;
2810 2810 }
2811 2811 } else {
2812 2812 /*
2813 2813 * This is in clear. The icmp message we are building
2814 2814 * here should go out in clear, independent of our policy.
2815 2815 */
2816 2816 ixas.ixa_flags |= IXAF_NO_IPSEC;
2817 2817 }
2818 2818
2819 2819 /* Remember our eventual destination */
2820 2820 dst = ipha->ipha_src;
2821 2821
2822 2822 /*
2823 2823 * If the packet was for one of our unicast addresses, make
2824 2824 * sure we respond with that as the source. Otherwise
2825 2825 * have ip_output_simple pick the source address.
2826 2826 */
2827 2827 ire = ire_ftable_lookup_v4(ipha->ipha_dst, 0, 0,
2828 2828 (IRE_LOCAL|IRE_LOOPBACK), NULL, ira->ira_zoneid, NULL,
2829 2829 MATCH_IRE_TYPE|MATCH_IRE_ZONEONLY, 0, ipst, NULL);
2830 2830 if (ire != NULL) {
2831 2831 ire_refrele(ire);
2832 2832 src = ipha->ipha_dst;
2833 2833 } else {
2834 2834 src = INADDR_ANY;
2835 2835 ixas.ixa_flags |= IXAF_SET_SOURCE;
2836 2836 }
2837 2837
2838 2838 /*
2839 2839 * Check if we can send back more then 8 bytes in addition to
2840 2840 * the IP header. We try to send 64 bytes of data and the internal
2841 2841 * header in the special cases of ipv4 encapsulated ipv4 or ipv6.
2842 2842 */
2843 2843 len_needed = IPH_HDR_LENGTH(ipha);
2844 2844 if (ipha->ipha_protocol == IPPROTO_ENCAP ||
2845 2845 ipha->ipha_protocol == IPPROTO_IPV6) {
2846 2846 if (!pullupmsg(mp, -1)) {
2847 2847 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
2848 2848 ip_drop_output("ipIfStatsOutDiscards", mp, NULL);
2849 2849 freemsg(mp);
2850 2850 return;
2851 2851 }
2852 2852 ipha = (ipha_t *)mp->b_rptr;
2853 2853
2854 2854 if (ipha->ipha_protocol == IPPROTO_ENCAP) {
2855 2855 len_needed += IPH_HDR_LENGTH(((uchar_t *)ipha +
2856 2856 len_needed));
2857 2857 } else {
2858 2858 ip6_t *ip6h = (ip6_t *)((uchar_t *)ipha + len_needed);
2859 2859
2860 2860 ASSERT(ipha->ipha_protocol == IPPROTO_IPV6);
2861 2861 len_needed += ip_hdr_length_v6(mp, ip6h);
2862 2862 }
2863 2863 }
2864 2864 len_needed += ipst->ips_ip_icmp_return;
2865 2865 msg_len = msgdsize(mp);
2866 2866 if (msg_len > len_needed) {
2867 2867 (void) adjmsg(mp, len_needed - msg_len);
2868 2868 msg_len = len_needed;
2869 2869 }
2870 2870 mp1 = allocb(sizeof (icmp_ipha) + len, BPRI_MED);
2871 2871 if (mp1 == NULL) {
2872 2872 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutErrors);
2873 2873 freemsg(mp);
2874 2874 return;
2875 2875 }
2876 2876 mp1->b_cont = mp;
2877 2877 mp = mp1;
2878 2878
2879 2879 /*
2880 2880 * Set IXAF_TRUSTED_ICMP so we can let the ICMP messages this
2881 2881 * node generates be accepted in peace by all on-host destinations.
2882 2882 * If we do NOT assume that all on-host destinations trust
2883 2883 * self-generated ICMP messages, then rework here, ip6.c, and spd.c.
2884 2884 * (Look for IXAF_TRUSTED_ICMP).
2885 2885 */
2886 2886 ixas.ixa_flags |= IXAF_TRUSTED_ICMP;
2887 2887
2888 2888 ipha = (ipha_t *)mp->b_rptr;
2889 2889 mp1->b_wptr = (uchar_t *)ipha + (sizeof (icmp_ipha) + len);
2890 2890 *ipha = icmp_ipha;
2891 2891 ipha->ipha_src = src;
2892 2892 ipha->ipha_dst = dst;
2893 2893 ipha->ipha_ttl = ipst->ips_ip_def_ttl;
2894 2894 msg_len += sizeof (icmp_ipha) + len;
2895 2895 if (msg_len > IP_MAXPACKET) {
2896 2896 (void) adjmsg(mp, IP_MAXPACKET - msg_len);
2897 2897 msg_len = IP_MAXPACKET;
2898 2898 }
2899 2899 ipha->ipha_length = htons((uint16_t)msg_len);
2900 2900 icmph = (icmph_t *)&ipha[1];
2901 2901 bcopy(stuff, icmph, len);
2902 2902 icmph->icmph_checksum = 0;
2903 2903 icmph->icmph_checksum = IP_CSUM(mp, (int32_t)sizeof (ipha_t), 0);
2904 2904 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutMsgs);
2905 2905
2906 2906 (void) ip_output_simple(mp, &ixas);
2907 2907 ixa_cleanup(&ixas);
2908 2908 }
2909 2909
2910 2910 /*
2911 2911 * Determine if an ICMP error packet can be sent given the rate limit.
2912 2912 * The limit consists of an average frequency (icmp_pkt_err_interval measured
2913 2913 * in milliseconds) and a burst size. Burst size number of packets can
2914 2914 * be sent arbitrarely closely spaced.
2915 2915 * The state is tracked using two variables to implement an approximate
2916 2916 * token bucket filter:
2917 2917 * icmp_pkt_err_last - lbolt value when the last burst started
2918 2918 * icmp_pkt_err_sent - number of packets sent in current burst
2919 2919 */
2920 2920 boolean_t
2921 2921 icmp_err_rate_limit(ip_stack_t *ipst)
2922 2922 {
2923 2923 clock_t now = TICK_TO_MSEC(ddi_get_lbolt());
2924 2924 uint_t refilled; /* Number of packets refilled in tbf since last */
2925 2925 /* Guard against changes by loading into local variable */
2926 2926 uint_t err_interval = ipst->ips_ip_icmp_err_interval;
2927 2927
2928 2928 if (err_interval == 0)
2929 2929 return (B_FALSE);
2930 2930
2931 2931 if (ipst->ips_icmp_pkt_err_last > now) {
2932 2932 /* 100HZ lbolt in ms for 32bit arch wraps every 49.7 days */
2933 2933 ipst->ips_icmp_pkt_err_last = 0;
2934 2934 ipst->ips_icmp_pkt_err_sent = 0;
2935 2935 }
2936 2936 /*
2937 2937 * If we are in a burst update the token bucket filter.
2938 2938 * Update the "last" time to be close to "now" but make sure
2939 2939 * we don't loose precision.
2940 2940 */
2941 2941 if (ipst->ips_icmp_pkt_err_sent != 0) {
2942 2942 refilled = (now - ipst->ips_icmp_pkt_err_last)/err_interval;
2943 2943 if (refilled > ipst->ips_icmp_pkt_err_sent) {
2944 2944 ipst->ips_icmp_pkt_err_sent = 0;
2945 2945 } else {
2946 2946 ipst->ips_icmp_pkt_err_sent -= refilled;
2947 2947 ipst->ips_icmp_pkt_err_last += refilled * err_interval;
2948 2948 }
2949 2949 }
2950 2950 if (ipst->ips_icmp_pkt_err_sent == 0) {
2951 2951 /* Start of new burst */
2952 2952 ipst->ips_icmp_pkt_err_last = now;
2953 2953 }
2954 2954 if (ipst->ips_icmp_pkt_err_sent < ipst->ips_ip_icmp_err_burst) {
2955 2955 ipst->ips_icmp_pkt_err_sent++;
2956 2956 ip1dbg(("icmp_err_rate_limit: %d sent in burst\n",
2957 2957 ipst->ips_icmp_pkt_err_sent));
2958 2958 return (B_FALSE);
2959 2959 }
2960 2960 ip1dbg(("icmp_err_rate_limit: dropped\n"));
2961 2961 return (B_TRUE);
2962 2962 }
2963 2963
2964 2964 /*
2965 2965 * Check if it is ok to send an IPv4 ICMP error packet in
2966 2966 * response to the IPv4 packet in mp.
2967 2967 * Free the message and return null if no
2968 2968 * ICMP error packet should be sent.
2969 2969 */
2970 2970 static mblk_t *
2971 2971 icmp_pkt_err_ok(mblk_t *mp, ip_recv_attr_t *ira)
2972 2972 {
2973 2973 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2974 2974 icmph_t *icmph;
2975 2975 ipha_t *ipha;
2976 2976 uint_t len_needed;
2977 2977
2978 2978 if (!mp)
2979 2979 return (NULL);
2980 2980 ipha = (ipha_t *)mp->b_rptr;
2981 2981 if (ip_csum_hdr(ipha)) {
2982 2982 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsInCksumErrs);
2983 2983 ip_drop_input("ipIfStatsInCksumErrs", mp, NULL);
2984 2984 freemsg(mp);
2985 2985 return (NULL);
2986 2986 }
2987 2987 if (ip_type_v4(ipha->ipha_dst, ipst) == IRE_BROADCAST ||
2988 2988 ip_type_v4(ipha->ipha_src, ipst) == IRE_BROADCAST ||
2989 2989 CLASSD(ipha->ipha_dst) ||
2990 2990 CLASSD(ipha->ipha_src) ||
2991 2991 (ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_OFFSET)) {
2992 2992 /* Note: only errors to the fragment with offset 0 */
2993 2993 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
2994 2994 freemsg(mp);
2995 2995 return (NULL);
2996 2996 }
2997 2997 if (ipha->ipha_protocol == IPPROTO_ICMP) {
2998 2998 /*
2999 2999 * Check the ICMP type. RFC 1122 sez: don't send ICMP
3000 3000 * errors in response to any ICMP errors.
3001 3001 */
3002 3002 len_needed = IPH_HDR_LENGTH(ipha) + ICMPH_SIZE;
3003 3003 if (mp->b_wptr - mp->b_rptr < len_needed) {
3004 3004 if (!pullupmsg(mp, len_needed)) {
3005 3005 BUMP_MIB(&ipst->ips_icmp_mib, icmpInErrors);
3006 3006 freemsg(mp);
3007 3007 return (NULL);
3008 3008 }
3009 3009 ipha = (ipha_t *)mp->b_rptr;
3010 3010 }
3011 3011 icmph = (icmph_t *)
3012 3012 (&((char *)ipha)[IPH_HDR_LENGTH(ipha)]);
3013 3013 switch (icmph->icmph_type) {
3014 3014 case ICMP_DEST_UNREACHABLE:
3015 3015 case ICMP_SOURCE_QUENCH:
3016 3016 case ICMP_TIME_EXCEEDED:
3017 3017 case ICMP_PARAM_PROBLEM:
3018 3018 case ICMP_REDIRECT:
3019 3019 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
3020 3020 freemsg(mp);
3021 3021 return (NULL);
3022 3022 default:
3023 3023 break;
3024 3024 }
3025 3025 }
3026 3026 /*
3027 3027 * If this is a labeled system, then check to see if we're allowed to
3028 3028 * send a response to this particular sender. If not, then just drop.
3029 3029 */
3030 3030 if (is_system_labeled() && !tsol_can_reply_error(mp, ira)) {
3031 3031 ip2dbg(("icmp_pkt_err_ok: can't respond to packet\n"));
3032 3032 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
3033 3033 freemsg(mp);
3034 3034 return (NULL);
3035 3035 }
3036 3036 if (icmp_err_rate_limit(ipst)) {
3037 3037 /*
3038 3038 * Only send ICMP error packets every so often.
3039 3039 * This should be done on a per port/source basis,
3040 3040 * but for now this will suffice.
3041 3041 */
3042 3042 freemsg(mp);
3043 3043 return (NULL);
3044 3044 }
3045 3045 return (mp);
3046 3046 }
3047 3047
3048 3048 /*
3049 3049 * Called when a packet was sent out the same link that it arrived on.
3050 3050 * Check if it is ok to send a redirect and then send it.
3051 3051 */
3052 3052 void
3053 3053 ip_send_potential_redirect_v4(mblk_t *mp, ipha_t *ipha, ire_t *ire,
3054 3054 ip_recv_attr_t *ira)
3055 3055 {
3056 3056 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3057 3057 ipaddr_t src, nhop;
3058 3058 mblk_t *mp1;
3059 3059 ire_t *nhop_ire;
3060 3060
3061 3061 /*
3062 3062 * Check the source address to see if it originated
3063 3063 * on the same logical subnet it is going back out on.
3064 3064 * If so, we should be able to send it a redirect.
3065 3065 * Avoid sending a redirect if the destination
3066 3066 * is directly connected (i.e., we matched an IRE_ONLINK),
3067 3067 * or if the packet was source routed out this interface.
3068 3068 *
3069 3069 * We avoid sending a redirect if the
3070 3070 * destination is directly connected
3071 3071 * because it is possible that multiple
3072 3072 * IP subnets may have been configured on
3073 3073 * the link, and the source may not
3074 3074 * be on the same subnet as ip destination,
3075 3075 * even though they are on the same
3076 3076 * physical link.
3077 3077 */
3078 3078 if ((ire->ire_type & IRE_ONLINK) ||
3079 3079 ip_source_routed(ipha, ipst))
3080 3080 return;
3081 3081
3082 3082 nhop_ire = ire_nexthop(ire);
3083 3083 if (nhop_ire == NULL)
3084 3084 return;
3085 3085
3086 3086 nhop = nhop_ire->ire_addr;
3087 3087
3088 3088 if (nhop_ire->ire_type & IRE_IF_CLONE) {
3089 3089 ire_t *ire2;
3090 3090
3091 3091 /* Follow ire_dep_parent to find non-clone IRE_INTERFACE */
3092 3092 mutex_enter(&nhop_ire->ire_lock);
3093 3093 ire2 = nhop_ire->ire_dep_parent;
3094 3094 if (ire2 != NULL)
3095 3095 ire_refhold(ire2);
3096 3096 mutex_exit(&nhop_ire->ire_lock);
3097 3097 ire_refrele(nhop_ire);
3098 3098 nhop_ire = ire2;
3099 3099 }
3100 3100 if (nhop_ire == NULL)
3101 3101 return;
3102 3102
3103 3103 ASSERT(!(nhop_ire->ire_type & IRE_IF_CLONE));
3104 3104
3105 3105 src = ipha->ipha_src;
3106 3106
3107 3107 /*
3108 3108 * We look at the interface ire for the nexthop,
3109 3109 * to see if ipha_src is in the same subnet
3110 3110 * as the nexthop.
3111 3111 */
3112 3112 if ((src & nhop_ire->ire_mask) == (nhop & nhop_ire->ire_mask)) {
3113 3113 /*
3114 3114 * The source is directly connected.
3115 3115 */
3116 3116 mp1 = copymsg(mp);
3117 3117 if (mp1 != NULL) {
3118 3118 icmp_send_redirect(mp1, nhop, ira);
3119 3119 }
3120 3120 }
3121 3121 ire_refrele(nhop_ire);
3122 3122 }
3123 3123
3124 3124 /*
3125 3125 * Generate an ICMP redirect message.
3126 3126 */
3127 3127 static void
3128 3128 icmp_send_redirect(mblk_t *mp, ipaddr_t gateway, ip_recv_attr_t *ira)
3129 3129 {
3130 3130 icmph_t icmph;
3131 3131 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3132 3132
3133 3133 mp = icmp_pkt_err_ok(mp, ira);
3134 3134 if (mp == NULL)
3135 3135 return;
3136 3136
3137 3137 bzero(&icmph, sizeof (icmph_t));
3138 3138 icmph.icmph_type = ICMP_REDIRECT;
3139 3139 icmph.icmph_code = 1;
3140 3140 icmph.icmph_rd_gateway = gateway;
3141 3141 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutRedirects);
3142 3142 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3143 3143 }
3144 3144
3145 3145 /*
3146 3146 * Generate an ICMP time exceeded message.
3147 3147 */
3148 3148 void
3149 3149 icmp_time_exceeded(mblk_t *mp, uint8_t code, ip_recv_attr_t *ira)
3150 3150 {
3151 3151 icmph_t icmph;
3152 3152 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3153 3153
3154 3154 mp = icmp_pkt_err_ok(mp, ira);
3155 3155 if (mp == NULL)
3156 3156 return;
3157 3157
3158 3158 bzero(&icmph, sizeof (icmph_t));
3159 3159 icmph.icmph_type = ICMP_TIME_EXCEEDED;
3160 3160 icmph.icmph_code = code;
3161 3161 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutTimeExcds);
3162 3162 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3163 3163 }
3164 3164
3165 3165 /*
3166 3166 * Generate an ICMP unreachable message.
3167 3167 * When called from ip_output side a minimal ip_recv_attr_t needs to be
3168 3168 * constructed by the caller.
3169 3169 */
3170 3170 void
3171 3171 icmp_unreachable(mblk_t *mp, uint8_t code, ip_recv_attr_t *ira)
3172 3172 {
3173 3173 icmph_t icmph;
3174 3174 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3175 3175
3176 3176 mp = icmp_pkt_err_ok(mp, ira);
3177 3177 if (mp == NULL)
3178 3178 return;
3179 3179
3180 3180 bzero(&icmph, sizeof (icmph_t));
3181 3181 icmph.icmph_type = ICMP_DEST_UNREACHABLE;
3182 3182 icmph.icmph_code = code;
3183 3183 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDestUnreachs);
3184 3184 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3185 3185 }
3186 3186
3187 3187 /*
3188 3188 * Latch in the IPsec state for a stream based the policy in the listener
3189 3189 * and the actions in the ip_recv_attr_t.
3190 3190 * Called directly from TCP and SCTP.
3191 3191 */
3192 3192 boolean_t
3193 3193 ip_ipsec_policy_inherit(conn_t *connp, conn_t *lconnp, ip_recv_attr_t *ira)
3194 3194 {
3195 3195 ASSERT(lconnp->conn_policy != NULL);
3196 3196 ASSERT(connp->conn_policy == NULL);
3197 3197
3198 3198 IPPH_REFHOLD(lconnp->conn_policy);
3199 3199 connp->conn_policy = lconnp->conn_policy;
3200 3200
3201 3201 if (ira->ira_ipsec_action != NULL) {
3202 3202 if (connp->conn_latch == NULL) {
3203 3203 connp->conn_latch = iplatch_create();
3204 3204 if (connp->conn_latch == NULL)
3205 3205 return (B_FALSE);
3206 3206 }
3207 3207 ipsec_latch_inbound(connp, ira);
3208 3208 }
3209 3209 return (B_TRUE);
3210 3210 }
3211 3211
3212 3212 /*
3213 3213 * Verify whether or not the IP address is a valid local address.
3214 3214 * Could be a unicast, including one for a down interface.
3215 3215 * If allow_mcbc then a multicast or broadcast address is also
3216 3216 * acceptable.
3217 3217 *
3218 3218 * In the case of a broadcast/multicast address, however, the
3219 3219 * upper protocol is expected to reset the src address
3220 3220 * to zero when we return IPVL_MCAST/IPVL_BCAST so that
3221 3221 * no packets are emitted with broadcast/multicast address as
3222 3222 * source address (that violates hosts requirements RFC 1122)
3223 3223 * The addresses valid for bind are:
3224 3224 * (1) - INADDR_ANY (0)
3225 3225 * (2) - IP address of an UP interface
3226 3226 * (3) - IP address of a DOWN interface
3227 3227 * (4) - valid local IP broadcast addresses. In this case
3228 3228 * the conn will only receive packets destined to
3229 3229 * the specified broadcast address.
3230 3230 * (5) - a multicast address. In this case
3231 3231 * the conn will only receive packets destined to
3232 3232 * the specified multicast address. Note: the
3233 3233 * application still has to issue an
3234 3234 * IP_ADD_MEMBERSHIP socket option.
3235 3235 *
3236 3236 * In all the above cases, the bound address must be valid in the current zone.
3237 3237 * When the address is loopback, multicast or broadcast, there might be many
3238 3238 * matching IREs so bind has to look up based on the zone.
3239 3239 */
3240 3240 ip_laddr_t
3241 3241 ip_laddr_verify_v4(ipaddr_t src_addr, zoneid_t zoneid,
3242 3242 ip_stack_t *ipst, boolean_t allow_mcbc)
3243 3243 {
3244 3244 ire_t *src_ire;
3245 3245
3246 3246 ASSERT(src_addr != INADDR_ANY);
3247 3247
3248 3248 src_ire = ire_ftable_lookup_v4(src_addr, 0, 0, 0,
3249 3249 NULL, zoneid, NULL, MATCH_IRE_ZONEONLY, 0, ipst, NULL);
3250 3250
3251 3251 /*
3252 3252 * If an address other than in6addr_any is requested,
3253 3253 * we verify that it is a valid address for bind
3254 3254 * Note: Following code is in if-else-if form for
3255 3255 * readability compared to a condition check.
3256 3256 */
3257 3257 if (src_ire != NULL && (src_ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK))) {
3258 3258 /*
3259 3259 * (2) Bind to address of local UP interface
3260 3260 */
3261 3261 ire_refrele(src_ire);
3262 3262 return (IPVL_UNICAST_UP);
3263 3263 } else if (src_ire != NULL && src_ire->ire_type & IRE_BROADCAST) {
3264 3264 /*
3265 3265 * (4) Bind to broadcast address
3266 3266 */
3267 3267 ire_refrele(src_ire);
3268 3268 if (allow_mcbc)
3269 3269 return (IPVL_BCAST);
3270 3270 else
3271 3271 return (IPVL_BAD);
3272 3272 } else if (CLASSD(src_addr)) {
3273 3273 /* (5) bind to multicast address. */
3274 3274 if (src_ire != NULL)
3275 3275 ire_refrele(src_ire);
3276 3276
3277 3277 if (allow_mcbc)
3278 3278 return (IPVL_MCAST);
3279 3279 else
3280 3280 return (IPVL_BAD);
3281 3281 } else {
3282 3282 ipif_t *ipif;
3283 3283
3284 3284 /*
3285 3285 * (3) Bind to address of local DOWN interface?
3286 3286 * (ipif_lookup_addr() looks up all interfaces
3287 3287 * but we do not get here for UP interfaces
3288 3288 * - case (2) above)
3289 3289 */
3290 3290 if (src_ire != NULL)
3291 3291 ire_refrele(src_ire);
3292 3292
3293 3293 ipif = ipif_lookup_addr(src_addr, NULL, zoneid, ipst);
3294 3294 if (ipif == NULL)
3295 3295 return (IPVL_BAD);
3296 3296
3297 3297 /* Not a useful source? */
3298 3298 if (ipif->ipif_flags & (IPIF_NOLOCAL | IPIF_ANYCAST)) {
3299 3299 ipif_refrele(ipif);
3300 3300 return (IPVL_BAD);
3301 3301 }
3302 3302 ipif_refrele(ipif);
3303 3303 return (IPVL_UNICAST_DOWN);
3304 3304 }
3305 3305 }
3306 3306
3307 3307 /*
3308 3308 * Insert in the bind fanout for IPv4 and IPv6.
3309 3309 * The caller should already have used ip_laddr_verify_v*() before calling
3310 3310 * this.
3311 3311 */
3312 3312 int
3313 3313 ip_laddr_fanout_insert(conn_t *connp)
3314 3314 {
3315 3315 int error;
3316 3316
3317 3317 /*
3318 3318 * Allow setting new policies. For example, disconnects result
3319 3319 * in us being called. As we would have set conn_policy_cached
3320 3320 * to B_TRUE before, we should set it to B_FALSE, so that policy
3321 3321 * can change after the disconnect.
3322 3322 */
3323 3323 connp->conn_policy_cached = B_FALSE;
3324 3324
3325 3325 error = ipcl_bind_insert(connp);
3326 3326 if (error != 0) {
3327 3327 if (connp->conn_anon_port) {
3328 3328 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
3329 3329 connp->conn_mlp_type, connp->conn_proto,
3330 3330 ntohs(connp->conn_lport), B_FALSE);
3331 3331 }
3332 3332 connp->conn_mlp_type = mlptSingle;
3333 3333 }
3334 3334 return (error);
3335 3335 }
3336 3336
3337 3337 /*
3338 3338 * Verify that both the source and destination addresses are valid. If
3339 3339 * IPDF_VERIFY_DST is not set, then the destination address may be unreachable,
3340 3340 * i.e. have no route to it. Protocols like TCP want to verify destination
3341 3341 * reachability, while tunnels do not.
3342 3342 *
3343 3343 * Determine the route, the interface, and (optionally) the source address
3344 3344 * to use to reach a given destination.
3345 3345 * Note that we allow connect to broadcast and multicast addresses when
3346 3346 * IPDF_ALLOW_MCBC is set.
3347 3347 * first_hop and dst_addr are normally the same, but if source routing
3348 3348 * they will differ; in that case the first_hop is what we'll use for the
3349 3349 * routing lookup but the dce and label checks will be done on dst_addr,
3350 3350 *
3351 3351 * If uinfo is set, then we fill in the best available information
3352 3352 * we have for the destination. This is based on (in priority order) any
3353 3353 * metrics and path MTU stored in a dce_t, route metrics, and finally the
3354 3354 * ill_mtu/ill_mc_mtu.
3355 3355 *
3356 3356 * Tsol note: If we have a source route then dst_addr != firsthop. But we
3357 3357 * always do the label check on dst_addr.
3358 3358 */
3359 3359 int
3360 3360 ip_set_destination_v4(ipaddr_t *src_addrp, ipaddr_t dst_addr, ipaddr_t firsthop,
3361 3361 ip_xmit_attr_t *ixa, iulp_t *uinfo, uint32_t flags, uint_t mac_mode)
3362 3362 {
3363 3363 ire_t *ire = NULL;
3364 3364 int error = 0;
3365 3365 ipaddr_t setsrc; /* RTF_SETSRC */
3366 3366 zoneid_t zoneid = ixa->ixa_zoneid; /* Honors SO_ALLZONES */
3367 3367 ip_stack_t *ipst = ixa->ixa_ipst;
3368 3368 dce_t *dce;
3369 3369 uint_t pmtu;
3370 3370 uint_t generation;
3371 3371 nce_t *nce;
3372 3372 ill_t *ill = NULL;
3373 3373 boolean_t multirt = B_FALSE;
3374 3374
3375 3375 ASSERT(ixa->ixa_flags & IXAF_IS_IPV4);
3376 3376
3377 3377 /*
3378 3378 * We never send to zero; the ULPs map it to the loopback address.
3379 3379 * We can't allow it since we use zero to mean unitialized in some
3380 3380 * places.
3381 3381 */
3382 3382 ASSERT(dst_addr != INADDR_ANY);
3383 3383
3384 3384 if (is_system_labeled()) {
3385 3385 ts_label_t *tsl = NULL;
3386 3386
3387 3387 error = tsol_check_dest(ixa->ixa_tsl, &dst_addr, IPV4_VERSION,
3388 3388 mac_mode, (flags & IPDF_ZONE_IS_GLOBAL) != 0, &tsl);
3389 3389 if (error != 0)
3390 3390 return (error);
3391 3391 if (tsl != NULL) {
3392 3392 /* Update the label */
3393 3393 ip_xmit_attr_replace_tsl(ixa, tsl);
3394 3394 }
3395 3395 }
3396 3396
3397 3397 setsrc = INADDR_ANY;
3398 3398 /*
3399 3399 * Select a route; For IPMP interfaces, we would only select
3400 3400 * a "hidden" route (i.e., going through a specific under_ill)
3401 3401 * if ixa_ifindex has been specified.
3402 3402 */
3403 3403 ire = ip_select_route_v4(firsthop, *src_addrp, ixa,
3404 3404 &generation, &setsrc, &error, &multirt);
3405 3405 ASSERT(ire != NULL); /* IRE_NOROUTE if none found */
3406 3406 if (error != 0)
3407 3407 goto bad_addr;
3408 3408
3409 3409 /*
3410 3410 * ire can't be a broadcast or multicast unless IPDF_ALLOW_MCBC is set.
3411 3411 * If IPDF_VERIFY_DST is set, the destination must be reachable;
3412 3412 * Otherwise the destination needn't be reachable.
3413 3413 *
3414 3414 * If we match on a reject or black hole, then we've got a
3415 3415 * local failure. May as well fail out the connect() attempt,
3416 3416 * since it's never going to succeed.
3417 3417 */
3418 3418 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
3419 3419 /*
3420 3420 * If we're verifying destination reachability, we always want
3421 3421 * to complain here.
3422 3422 *
3423 3423 * If we're not verifying destination reachability but the
3424 3424 * destination has a route, we still want to fail on the
3425 3425 * temporary address and broadcast address tests.
3426 3426 *
3427 3427 * In both cases do we let the code continue so some reasonable
3428 3428 * information is returned to the caller. That enables the
3429 3429 * caller to use (and even cache) the IRE. conn_ip_ouput will
3430 3430 * use the generation mismatch path to check for the unreachable
3431 3431 * case thereby avoiding any specific check in the main path.
3432 3432 */
3433 3433 ASSERT(generation == IRE_GENERATION_VERIFY);
3434 3434 if (flags & IPDF_VERIFY_DST) {
3435 3435 /*
3436 3436 * Set errno but continue to set up ixa_ire to be
3437 3437 * the RTF_REJECT|RTF_BLACKHOLE IRE.
3438 3438 * That allows callers to use ip_output to get an
3439 3439 * ICMP error back.
3440 3440 */
3441 3441 if (!(ire->ire_type & IRE_HOST))
3442 3442 error = ENETUNREACH;
3443 3443 else
3444 3444 error = EHOSTUNREACH;
3445 3445 }
3446 3446 }
3447 3447
3448 3448 if ((ire->ire_type & (IRE_BROADCAST|IRE_MULTICAST)) &&
3449 3449 !(flags & IPDF_ALLOW_MCBC)) {
3450 3450 ire_refrele(ire);
3451 3451 ire = ire_reject(ipst, B_FALSE);
3452 3452 generation = IRE_GENERATION_VERIFY;
3453 3453 error = ENETUNREACH;
3454 3454 }
3455 3455
3456 3456 /* Cache things */
3457 3457 if (ixa->ixa_ire != NULL)
3458 3458 ire_refrele_notr(ixa->ixa_ire);
3459 3459 #ifdef DEBUG
3460 3460 ire_refhold_notr(ire);
3461 3461 ire_refrele(ire);
3462 3462 #endif
3463 3463 ixa->ixa_ire = ire;
3464 3464 ixa->ixa_ire_generation = generation;
3465 3465
3466 3466 /*
3467 3467 * Ensure that ixa_dce is always set any time that ixa_ire is set,
3468 3468 * since some callers will send a packet to conn_ip_output() even if
3469 3469 * there's an error.
3470 3470 */
3471 3471 if (flags & IPDF_UNIQUE_DCE) {
3472 3472 /* Fallback to the default dce if allocation fails */
3473 3473 dce = dce_lookup_and_add_v4(dst_addr, ipst);
3474 3474 if (dce != NULL)
3475 3475 generation = dce->dce_generation;
3476 3476 else
3477 3477 dce = dce_lookup_v4(dst_addr, ipst, &generation);
3478 3478 } else {
3479 3479 dce = dce_lookup_v4(dst_addr, ipst, &generation);
3480 3480 }
3481 3481 ASSERT(dce != NULL);
3482 3482 if (ixa->ixa_dce != NULL)
3483 3483 dce_refrele_notr(ixa->ixa_dce);
3484 3484 #ifdef DEBUG
3485 3485 dce_refhold_notr(dce);
3486 3486 dce_refrele(dce);
3487 3487 #endif
3488 3488 ixa->ixa_dce = dce;
3489 3489 ixa->ixa_dce_generation = generation;
3490 3490
3491 3491 /*
3492 3492 * For multicast with multirt we have a flag passed back from
3493 3493 * ire_lookup_multi_ill_v4 since we don't have an IRE for each
3494 3494 * possible multicast address.
3495 3495 * We also need a flag for multicast since we can't check
3496 3496 * whether RTF_MULTIRT is set in ixa_ire for multicast.
3497 3497 */
3498 3498 if (multirt) {
3499 3499 ixa->ixa_postfragfn = ip_postfrag_multirt_v4;
3500 3500 ixa->ixa_flags |= IXAF_MULTIRT_MULTICAST;
3501 3501 } else {
3502 3502 ixa->ixa_postfragfn = ire->ire_postfragfn;
3503 3503 ixa->ixa_flags &= ~IXAF_MULTIRT_MULTICAST;
3504 3504 }
3505 3505 if (!(ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))) {
3506 3506 /* Get an nce to cache. */
3507 3507 nce = ire_to_nce(ire, firsthop, NULL);
3508 3508 if (nce == NULL) {
3509 3509 /* Allocation failure? */
3510 3510 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3511 3511 } else {
3512 3512 if (ixa->ixa_nce != NULL)
3513 3513 nce_refrele(ixa->ixa_nce);
3514 3514 ixa->ixa_nce = nce;
3515 3515 }
3516 3516 }
3517 3517
3518 3518 /*
3519 3519 * If the source address is a loopback address, the
3520 3520 * destination had best be local or multicast.
3521 3521 * If we are sending to an IRE_LOCAL using a loopback source then
3522 3522 * it had better be the same zoneid.
3523 3523 */
3524 3524 if (*src_addrp == htonl(INADDR_LOOPBACK)) {
3525 3525 if ((ire->ire_type & IRE_LOCAL) && ire->ire_zoneid != zoneid) {
3526 3526 ire = NULL; /* Stored in ixa_ire */
3527 3527 error = EADDRNOTAVAIL;
3528 3528 goto bad_addr;
3529 3529 }
3530 3530 if (!(ire->ire_type & (IRE_LOOPBACK|IRE_LOCAL|IRE_MULTICAST))) {
3531 3531 ire = NULL; /* Stored in ixa_ire */
3532 3532 error = EADDRNOTAVAIL;
3533 3533 goto bad_addr;
3534 3534 }
3535 3535 }
3536 3536 if (ire->ire_type & IRE_BROADCAST) {
3537 3537 /*
3538 3538 * If the ULP didn't have a specified source, then we
3539 3539 * make sure we reselect the source when sending
3540 3540 * broadcasts out different interfaces.
3541 3541 */
3542 3542 if (flags & IPDF_SELECT_SRC)
3543 3543 ixa->ixa_flags |= IXAF_SET_SOURCE;
3544 3544 else
3545 3545 ixa->ixa_flags &= ~IXAF_SET_SOURCE;
3546 3546 }
3547 3547
3548 3548 /*
3549 3549 * Does the caller want us to pick a source address?
3550 3550 */
3551 3551 if (flags & IPDF_SELECT_SRC) {
3552 3552 ipaddr_t src_addr;
3553 3553
3554 3554 /*
3555 3555 * We use use ire_nexthop_ill to avoid the under ipmp
3556 3556 * interface for source address selection. Note that for ipmp
3557 3557 * probe packets, ixa_ifindex would have been specified, and
3558 3558 * the ip_select_route() invocation would have picked an ire
3559 3559 * will ire_ill pointing at an under interface.
3560 3560 */
3561 3561 ill = ire_nexthop_ill(ire);
3562 3562
3563 3563 /* If unreachable we have no ill but need some source */
3564 3564 if (ill == NULL) {
3565 3565 src_addr = htonl(INADDR_LOOPBACK);
3566 3566 /* Make sure we look for a better source address */
3567 3567 generation = SRC_GENERATION_VERIFY;
3568 3568 } else {
3569 3569 error = ip_select_source_v4(ill, setsrc, dst_addr,
3570 3570 ixa->ixa_multicast_ifaddr, zoneid,
3571 3571 ipst, &src_addr, &generation, NULL);
3572 3572 if (error != 0) {
3573 3573 ire = NULL; /* Stored in ixa_ire */
3574 3574 goto bad_addr;
3575 3575 }
3576 3576 }
3577 3577
3578 3578 /*
3579 3579 * We allow the source address to to down.
3580 3580 * However, we check that we don't use the loopback address
3581 3581 * as a source when sending out on the wire.
3582 3582 */
3583 3583 if ((src_addr == htonl(INADDR_LOOPBACK)) &&
3584 3584 !(ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK|IRE_MULTICAST)) &&
3585 3585 !(ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))) {
3586 3586 ire = NULL; /* Stored in ixa_ire */
3587 3587 error = EADDRNOTAVAIL;
3588 3588 goto bad_addr;
3589 3589 }
3590 3590
3591 3591 *src_addrp = src_addr;
3592 3592 ixa->ixa_src_generation = generation;
3593 3593 }
3594 3594
3595 3595 /*
3596 3596 * Make sure we don't leave an unreachable ixa_nce in place
3597 3597 * since ip_select_route is used when we unplumb i.e., remove
3598 3598 * references on ixa_ire, ixa_nce, and ixa_dce.
3599 3599 */
3600 3600 nce = ixa->ixa_nce;
3601 3601 if (nce != NULL && nce->nce_is_condemned) {
3602 3602 nce_refrele(nce);
3603 3603 ixa->ixa_nce = NULL;
3604 3604 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3605 3605 }
3606 3606
3607 3607 /*
3608 3608 * The caller has set IXAF_PMTU_DISCOVERY if path MTU is desired.
3609 3609 * However, we can't do it for IPv4 multicast or broadcast.
3610 3610 */
3611 3611 if (ire->ire_type & (IRE_BROADCAST|IRE_MULTICAST))
3612 3612 ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
3613 3613
3614 3614 /*
3615 3615 * Set initial value for fragmentation limit. Either conn_ip_output
3616 3616 * or ULP might updates it when there are routing changes.
3617 3617 * Handles a NULL ixa_ire->ire_ill or a NULL ixa_nce for RTF_REJECT.
3618 3618 */
3619 3619 pmtu = ip_get_pmtu(ixa);
3620 3620 ixa->ixa_fragsize = pmtu;
3621 3621 /* Make sure ixa_fragsize and ixa_pmtu remain identical */
3622 3622 if (ixa->ixa_flags & IXAF_VERIFY_PMTU)
3623 3623 ixa->ixa_pmtu = pmtu;
3624 3624
3625 3625 /*
3626 3626 * Extract information useful for some transports.
3627 3627 * First we look for DCE metrics. Then we take what we have in
3628 3628 * the metrics in the route, where the offlink is used if we have
3629 3629 * one.
3630 3630 */
3631 3631 if (uinfo != NULL) {
3632 3632 bzero(uinfo, sizeof (*uinfo));
3633 3633
3634 3634 if (dce->dce_flags & DCEF_UINFO)
3635 3635 *uinfo = dce->dce_uinfo;
3636 3636
3637 3637 rts_merge_metrics(uinfo, &ire->ire_metrics);
3638 3638
3639 3639 /* Allow ire_metrics to decrease the path MTU from above */
3640 3640 if (uinfo->iulp_mtu == 0 || uinfo->iulp_mtu > pmtu)
3641 3641 uinfo->iulp_mtu = pmtu;
3642 3642
3643 3643 uinfo->iulp_localnet = (ire->ire_type & IRE_ONLINK) != 0;
3644 3644 uinfo->iulp_loopback = (ire->ire_type & IRE_LOOPBACK) != 0;
3645 3645 uinfo->iulp_local = (ire->ire_type & IRE_LOCAL) != 0;
3646 3646 }
3647 3647
3648 3648 if (ill != NULL)
3649 3649 ill_refrele(ill);
3650 3650
3651 3651 return (error);
3652 3652
3653 3653 bad_addr:
3654 3654 if (ire != NULL)
3655 3655 ire_refrele(ire);
3656 3656
3657 3657 if (ill != NULL)
3658 3658 ill_refrele(ill);
3659 3659
3660 3660 /*
3661 3661 * Make sure we don't leave an unreachable ixa_nce in place
3662 3662 * since ip_select_route is used when we unplumb i.e., remove
3663 3663 * references on ixa_ire, ixa_nce, and ixa_dce.
3664 3664 */
3665 3665 nce = ixa->ixa_nce;
3666 3666 if (nce != NULL && nce->nce_is_condemned) {
3667 3667 nce_refrele(nce);
3668 3668 ixa->ixa_nce = NULL;
3669 3669 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3670 3670 }
3671 3671
3672 3672 return (error);
3673 3673 }
3674 3674
3675 3675
3676 3676 /*
3677 3677 * Get the base MTU for the case when path MTU discovery is not used.
3678 3678 * Takes the MTU of the IRE into account.
3679 3679 */
3680 3680 uint_t
3681 3681 ip_get_base_mtu(ill_t *ill, ire_t *ire)
3682 3682 {
3683 3683 uint_t mtu;
3684 3684 uint_t iremtu = ire->ire_metrics.iulp_mtu;
3685 3685
3686 3686 if (ire->ire_type & (IRE_MULTICAST|IRE_BROADCAST))
3687 3687 mtu = ill->ill_mc_mtu;
3688 3688 else
3689 3689 mtu = ill->ill_mtu;
3690 3690
3691 3691 if (iremtu != 0 && iremtu < mtu)
3692 3692 mtu = iremtu;
3693 3693
3694 3694 return (mtu);
3695 3695 }
3696 3696
3697 3697 /*
3698 3698 * Get the PMTU for the attributes. Handles both IPv4 and IPv6.
3699 3699 * Assumes that ixa_ire, dce, and nce have already been set up.
3700 3700 *
3701 3701 * The caller has set IXAF_PMTU_DISCOVERY if path MTU discovery is desired.
3702 3702 * We avoid path MTU discovery if it is disabled with ndd.
3703 3703 * Furtermore, if the path MTU is too small, then we don't set DF for IPv4.
3704 3704 *
3705 3705 * NOTE: We also used to turn it off for source routed packets. That
3706 3706 * is no longer required since the dce is per final destination.
3707 3707 */
3708 3708 uint_t
3709 3709 ip_get_pmtu(ip_xmit_attr_t *ixa)
3710 3710 {
3711 3711 ip_stack_t *ipst = ixa->ixa_ipst;
3712 3712 dce_t *dce;
3713 3713 nce_t *nce;
3714 3714 ire_t *ire;
3715 3715 uint_t pmtu;
3716 3716
3717 3717 ire = ixa->ixa_ire;
3718 3718 dce = ixa->ixa_dce;
3719 3719 nce = ixa->ixa_nce;
3720 3720
3721 3721 /*
3722 3722 * If path MTU discovery has been turned off by ndd, then we ignore
3723 3723 * any dce_pmtu and for IPv4 we will not set DF.
3724 3724 */
3725 3725 if (!ipst->ips_ip_path_mtu_discovery)
3726 3726 ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
3727 3727
3728 3728 pmtu = IP_MAXPACKET;
3729 3729 /*
3730 3730 * Decide whether whether IPv4 sets DF
3731 3731 * For IPv6 "no DF" means to use the 1280 mtu
3732 3732 */
3733 3733 if (ixa->ixa_flags & IXAF_PMTU_DISCOVERY) {
3734 3734 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3735 3735 } else {
3736 3736 ixa->ixa_flags &= ~IXAF_PMTU_IPV4_DF;
3737 3737 if (!(ixa->ixa_flags & IXAF_IS_IPV4))
3738 3738 pmtu = IPV6_MIN_MTU;
3739 3739 }
3740 3740
3741 3741 /* Check if the PMTU is to old before we use it */
3742 3742 if ((dce->dce_flags & DCEF_PMTU) &&
3743 3743 TICK_TO_SEC(ddi_get_lbolt64()) - dce->dce_last_change_time >
3744 3744 ipst->ips_ip_pathmtu_interval) {
3745 3745 /*
3746 3746 * Older than 20 minutes. Drop the path MTU information.
3747 3747 */
3748 3748 mutex_enter(&dce->dce_lock);
3749 3749 dce->dce_flags &= ~(DCEF_PMTU|DCEF_TOO_SMALL_PMTU);
3750 3750 dce->dce_last_change_time = TICK_TO_SEC(ddi_get_lbolt64());
3751 3751 mutex_exit(&dce->dce_lock);
3752 3752 dce_increment_generation(dce);
3753 3753 }
3754 3754
3755 3755 /* The metrics on the route can lower the path MTU */
3756 3756 if (ire->ire_metrics.iulp_mtu != 0 &&
3757 3757 ire->ire_metrics.iulp_mtu < pmtu)
3758 3758 pmtu = ire->ire_metrics.iulp_mtu;
3759 3759
3760 3760 /*
3761 3761 * If the path MTU is smaller than some minimum, we still use dce_pmtu
3762 3762 * above (would be 576 for IPv4 and 1280 for IPv6), but we clear
3763 3763 * IXAF_PMTU_IPV4_DF so that we avoid setting DF for IPv4.
3764 3764 */
3765 3765 if (ixa->ixa_flags & IXAF_PMTU_DISCOVERY) {
3766 3766 if (dce->dce_flags & DCEF_PMTU) {
3767 3767 if (dce->dce_pmtu < pmtu)
3768 3768 pmtu = dce->dce_pmtu;
3769 3769
3770 3770 if (dce->dce_flags & DCEF_TOO_SMALL_PMTU) {
3771 3771 ixa->ixa_flags |= IXAF_PMTU_TOO_SMALL;
3772 3772 ixa->ixa_flags &= ~IXAF_PMTU_IPV4_DF;
3773 3773 } else {
3774 3774 ixa->ixa_flags &= ~IXAF_PMTU_TOO_SMALL;
3775 3775 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3776 3776 }
3777 3777 } else {
3778 3778 ixa->ixa_flags &= ~IXAF_PMTU_TOO_SMALL;
3779 3779 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3780 3780 }
3781 3781 }
3782 3782
3783 3783 /*
3784 3784 * If we have an IRE_LOCAL we use the loopback mtu instead of
3785 3785 * the ill for going out the wire i.e., IRE_LOCAL gets the same
3786 3786 * mtu as IRE_LOOPBACK.
3787 3787 */
3788 3788 if (ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK)) {
3789 3789 uint_t loopback_mtu;
3790 3790
3791 3791 loopback_mtu = (ire->ire_ipversion == IPV6_VERSION) ?
3792 3792 ip_loopback_mtu_v6plus : ip_loopback_mtuplus;
3793 3793
3794 3794 if (loopback_mtu < pmtu)
3795 3795 pmtu = loopback_mtu;
3796 3796 } else if (nce != NULL) {
3797 3797 /*
3798 3798 * Make sure we don't exceed the interface MTU.
3799 3799 * In the case of RTF_REJECT or RTF_BLACKHOLE we might not have
3800 3800 * an ill. We'd use the above IP_MAXPACKET in that case just
3801 3801 * to tell the transport something larger than zero.
3802 3802 */
3803 3803 if (ire->ire_type & (IRE_MULTICAST|IRE_BROADCAST)) {
3804 3804 if (nce->nce_common->ncec_ill->ill_mc_mtu < pmtu)
3805 3805 pmtu = nce->nce_common->ncec_ill->ill_mc_mtu;
3806 3806 if (nce->nce_common->ncec_ill != nce->nce_ill &&
3807 3807 nce->nce_ill->ill_mc_mtu < pmtu) {
3808 3808 /*
3809 3809 * for interfaces in an IPMP group, the mtu of
3810 3810 * the nce_ill (under_ill) could be different
3811 3811 * from the mtu of the ncec_ill, so we take the
3812 3812 * min of the two.
3813 3813 */
3814 3814 pmtu = nce->nce_ill->ill_mc_mtu;
3815 3815 }
3816 3816 } else {
3817 3817 if (nce->nce_common->ncec_ill->ill_mtu < pmtu)
3818 3818 pmtu = nce->nce_common->ncec_ill->ill_mtu;
3819 3819 if (nce->nce_common->ncec_ill != nce->nce_ill &&
3820 3820 nce->nce_ill->ill_mtu < pmtu) {
3821 3821 /*
3822 3822 * for interfaces in an IPMP group, the mtu of
3823 3823 * the nce_ill (under_ill) could be different
3824 3824 * from the mtu of the ncec_ill, so we take the
3825 3825 * min of the two.
3826 3826 */
3827 3827 pmtu = nce->nce_ill->ill_mtu;
3828 3828 }
3829 3829 }
3830 3830 }
3831 3831
3832 3832 /*
3833 3833 * Handle the IPV6_USE_MIN_MTU socket option or ancillary data.
3834 3834 * Only applies to IPv6.
3835 3835 */
3836 3836 if (!(ixa->ixa_flags & IXAF_IS_IPV4)) {
3837 3837 if (ixa->ixa_flags & IXAF_USE_MIN_MTU) {
3838 3838 switch (ixa->ixa_use_min_mtu) {
3839 3839 case IPV6_USE_MIN_MTU_MULTICAST:
3840 3840 if (ire->ire_type & IRE_MULTICAST)
3841 3841 pmtu = IPV6_MIN_MTU;
3842 3842 break;
3843 3843 case IPV6_USE_MIN_MTU_ALWAYS:
3844 3844 pmtu = IPV6_MIN_MTU;
3845 3845 break;
3846 3846 case IPV6_USE_MIN_MTU_NEVER:
3847 3847 break;
3848 3848 }
3849 3849 } else {
3850 3850 /* Default is IPV6_USE_MIN_MTU_MULTICAST */
3851 3851 if (ire->ire_type & IRE_MULTICAST)
3852 3852 pmtu = IPV6_MIN_MTU;
3853 3853 }
3854 3854 }
3855 3855
3856 3856 /*
3857 3857 * For multirouted IPv6 packets, the IP layer will insert a 8-byte
3858 3858 * fragment header in every packet. We compensate for those cases by
3859 3859 * returning a smaller path MTU to the ULP.
3860 3860 *
3861 3861 * In the case of CGTP then ip_output will add a fragment header.
3862 3862 * Make sure there is room for it by telling a smaller number
3863 3863 * to the transport.
3864 3864 *
3865 3865 * When IXAF_IPV6_ADDR_FRAGHDR we subtract the frag hdr here
3866 3866 * so the ULPs consistently see a iulp_pmtu and ip_get_pmtu()
3867 3867 * which is the size of the packets it can send.
3868 3868 */
3869 3869 if (!(ixa->ixa_flags & IXAF_IS_IPV4)) {
3870 3870 if ((ire->ire_flags & RTF_MULTIRT) ||
3871 3871 (ixa->ixa_flags & IXAF_MULTIRT_MULTICAST)) {
3872 3872 pmtu -= sizeof (ip6_frag_t);
3873 3873 ixa->ixa_flags |= IXAF_IPV6_ADD_FRAGHDR;
3874 3874 }
3875 3875 }
3876 3876
3877 3877 return (pmtu);
3878 3878 }
3879 3879
3880 3880 /*
3881 3881 * Carve "len" bytes out of an mblk chain, consuming any we empty, and duping
3882 3882 * the final piece where we don't. Return a pointer to the first mblk in the
3883 3883 * result, and update the pointer to the next mblk to chew on. If anything
3884 3884 * goes wrong (i.e., dupb fails), we waste everything in sight and return a
3885 3885 * NULL pointer.
3886 3886 */
3887 3887 mblk_t *
3888 3888 ip_carve_mp(mblk_t **mpp, ssize_t len)
3889 3889 {
3890 3890 mblk_t *mp0;
3891 3891 mblk_t *mp1;
3892 3892 mblk_t *mp2;
3893 3893
3894 3894 if (!len || !mpp || !(mp0 = *mpp))
3895 3895 return (NULL);
3896 3896 /* If we aren't going to consume the first mblk, we need a dup. */
3897 3897 if (mp0->b_wptr - mp0->b_rptr > len) {
3898 3898 mp1 = dupb(mp0);
3899 3899 if (mp1) {
3900 3900 /* Partition the data between the two mblks. */
3901 3901 mp1->b_wptr = mp1->b_rptr + len;
3902 3902 mp0->b_rptr = mp1->b_wptr;
3903 3903 /*
3904 3904 * after adjustments if mblk not consumed is now
3905 3905 * unaligned, try to align it. If this fails free
3906 3906 * all messages and let upper layer recover.
3907 3907 */
3908 3908 if (!OK_32PTR(mp0->b_rptr)) {
3909 3909 if (!pullupmsg(mp0, -1)) {
3910 3910 freemsg(mp0);
3911 3911 freemsg(mp1);
3912 3912 *mpp = NULL;
3913 3913 return (NULL);
3914 3914 }
3915 3915 }
3916 3916 }
3917 3917 return (mp1);
3918 3918 }
3919 3919 /* Eat through as many mblks as we need to get len bytes. */
3920 3920 len -= mp0->b_wptr - mp0->b_rptr;
3921 3921 for (mp2 = mp1 = mp0; (mp2 = mp2->b_cont) != 0 && len; mp1 = mp2) {
3922 3922 if (mp2->b_wptr - mp2->b_rptr > len) {
3923 3923 /*
3924 3924 * We won't consume the entire last mblk. Like
3925 3925 * above, dup and partition it.
3926 3926 */
3927 3927 mp1->b_cont = dupb(mp2);
3928 3928 mp1 = mp1->b_cont;
3929 3929 if (!mp1) {
3930 3930 /*
3931 3931 * Trouble. Rather than go to a lot of
3932 3932 * trouble to clean up, we free the messages.
3933 3933 * This won't be any worse than losing it on
3934 3934 * the wire.
3935 3935 */
3936 3936 freemsg(mp0);
3937 3937 freemsg(mp2);
3938 3938 *mpp = NULL;
3939 3939 return (NULL);
3940 3940 }
3941 3941 mp1->b_wptr = mp1->b_rptr + len;
3942 3942 mp2->b_rptr = mp1->b_wptr;
3943 3943 /*
3944 3944 * after adjustments if mblk not consumed is now
3945 3945 * unaligned, try to align it. If this fails free
3946 3946 * all messages and let upper layer recover.
3947 3947 */
3948 3948 if (!OK_32PTR(mp2->b_rptr)) {
3949 3949 if (!pullupmsg(mp2, -1)) {
3950 3950 freemsg(mp0);
3951 3951 freemsg(mp2);
3952 3952 *mpp = NULL;
3953 3953 return (NULL);
3954 3954 }
3955 3955 }
3956 3956 *mpp = mp2;
3957 3957 return (mp0);
3958 3958 }
3959 3959 /* Decrement len by the amount we just got. */
3960 3960 len -= mp2->b_wptr - mp2->b_rptr;
3961 3961 }
3962 3962 /*
3963 3963 * len should be reduced to zero now. If not our caller has
3964 3964 * screwed up.
3965 3965 */
3966 3966 if (len) {
3967 3967 /* Shouldn't happen! */
3968 3968 freemsg(mp0);
3969 3969 *mpp = NULL;
3970 3970 return (NULL);
3971 3971 }
3972 3972 /*
3973 3973 * We consumed up to exactly the end of an mblk. Detach the part
3974 3974 * we are returning from the rest of the chain.
3975 3975 */
3976 3976 mp1->b_cont = NULL;
3977 3977 *mpp = mp2;
3978 3978 return (mp0);
3979 3979 }
3980 3980
3981 3981 /* The ill stream is being unplumbed. Called from ip_close */
3982 3982 int
3983 3983 ip_modclose(ill_t *ill)
3984 3984 {
3985 3985 boolean_t success;
3986 3986 ipsq_t *ipsq;
3987 3987 ipif_t *ipif;
3988 3988 queue_t *q = ill->ill_rq;
3989 3989 ip_stack_t *ipst = ill->ill_ipst;
3990 3990 int i;
3991 3991 arl_ill_common_t *ai = ill->ill_common;
3992 3992
3993 3993 /*
3994 3994 * The punlink prior to this may have initiated a capability
3995 3995 * negotiation. But ipsq_enter will block until that finishes or
3996 3996 * times out.
3997 3997 */
3998 3998 success = ipsq_enter(ill, B_FALSE, NEW_OP);
3999 3999
4000 4000 /*
4001 4001 * Open/close/push/pop is guaranteed to be single threaded
4002 4002 * per stream by STREAMS. FS guarantees that all references
4003 4003 * from top are gone before close is called. So there can't
4004 4004 * be another close thread that has set CONDEMNED on this ill.
4005 4005 * and cause ipsq_enter to return failure.
4006 4006 */
4007 4007 ASSERT(success);
4008 4008 ipsq = ill->ill_phyint->phyint_ipsq;
4009 4009
4010 4010 /*
4011 4011 * Mark it condemned. No new reference will be made to this ill.
4012 4012 * Lookup functions will return an error. Threads that try to
4013 4013 * increment the refcnt must check for ILL_CAN_LOOKUP. This ensures
4014 4014 * that the refcnt will drop down to zero.
4015 4015 */
4016 4016 mutex_enter(&ill->ill_lock);
4017 4017 ill->ill_state_flags |= ILL_CONDEMNED;
4018 4018 for (ipif = ill->ill_ipif; ipif != NULL;
4019 4019 ipif = ipif->ipif_next) {
4020 4020 ipif->ipif_state_flags |= IPIF_CONDEMNED;
4021 4021 }
4022 4022 /*
4023 4023 * Wake up anybody waiting to enter the ipsq. ipsq_enter
4024 4024 * returns error if ILL_CONDEMNED is set
4025 4025 */
4026 4026 cv_broadcast(&ill->ill_cv);
4027 4027 mutex_exit(&ill->ill_lock);
4028 4028
4029 4029 /*
4030 4030 * Send all the deferred DLPI messages downstream which came in
4031 4031 * during the small window right before ipsq_enter(). We do this
4032 4032 * without waiting for the ACKs because all the ACKs for M_PROTO
4033 4033 * messages are ignored in ip_rput() when ILL_CONDEMNED is set.
4034 4034 */
4035 4035 ill_dlpi_send_deferred(ill);
4036 4036
4037 4037 /*
4038 4038 * Shut down fragmentation reassembly.
4039 4039 * ill_frag_timer won't start a timer again.
4040 4040 * Now cancel any existing timer
4041 4041 */
4042 4042 (void) untimeout(ill->ill_frag_timer_id);
4043 4043 (void) ill_frag_timeout(ill, 0);
4044 4044
4045 4045 /*
4046 4046 * Call ill_delete to bring down the ipifs, ilms and ill on
4047 4047 * this ill. Then wait for the refcnts to drop to zero.
4048 4048 * ill_is_freeable checks whether the ill is really quiescent.
4049 4049 * Then make sure that threads that are waiting to enter the
4050 4050 * ipsq have seen the error returned by ipsq_enter and have
4051 4051 * gone away. Then we call ill_delete_tail which does the
4052 4052 * DL_UNBIND_REQ with the driver and then qprocsoff.
4053 4053 */
4054 4054 ill_delete(ill);
4055 4055 mutex_enter(&ill->ill_lock);
4056 4056 while (!ill_is_freeable(ill))
4057 4057 cv_wait(&ill->ill_cv, &ill->ill_lock);
4058 4058
4059 4059 while (ill->ill_waiters)
4060 4060 cv_wait(&ill->ill_cv, &ill->ill_lock);
4061 4061
4062 4062 mutex_exit(&ill->ill_lock);
4063 4063
4064 4064 /*
4065 4065 * ill_delete_tail drops reference on ill_ipst, but we need to keep
4066 4066 * it held until the end of the function since the cleanup
4067 4067 * below needs to be able to use the ip_stack_t.
4068 4068 */
4069 4069 netstack_hold(ipst->ips_netstack);
4070 4070
4071 4071 /* qprocsoff is done via ill_delete_tail */
4072 4072 ill_delete_tail(ill);
4073 4073 /*
4074 4074 * synchronously wait for arp stream to unbind. After this, we
4075 4075 * cannot get any data packets up from the driver.
4076 4076 */
4077 4077 arp_unbind_complete(ill);
4078 4078 ASSERT(ill->ill_ipst == NULL);
4079 4079
4080 4080 /*
4081 4081 * Walk through all conns and qenable those that have queued data.
4082 4082 * Close synchronization needs this to
4083 4083 * be done to ensure that all upper layers blocked
4084 4084 * due to flow control to the closing device
4085 4085 * get unblocked.
4086 4086 */
4087 4087 ip1dbg(("ip_wsrv: walking\n"));
4088 4088 for (i = 0; i < TX_FANOUT_SIZE; i++) {
4089 4089 conn_walk_drain(ipst, &ipst->ips_idl_tx_list[i]);
4090 4090 }
4091 4091
4092 4092 /*
4093 4093 * ai can be null if this is an IPv6 ill, or if the IPv4
4094 4094 * stream is being torn down before ARP was plumbed (e.g.,
4095 4095 * /sbin/ifconfig plumbing a stream twice, and encountering
4096 4096 * an error
4097 4097 */
4098 4098 if (ai != NULL) {
4099 4099 ASSERT(!ill->ill_isv6);
4100 4100 mutex_enter(&ai->ai_lock);
4101 4101 ai->ai_ill = NULL;
4102 4102 if (ai->ai_arl == NULL) {
4103 4103 mutex_destroy(&ai->ai_lock);
4104 4104 kmem_free(ai, sizeof (*ai));
4105 4105 } else {
4106 4106 cv_signal(&ai->ai_ill_unplumb_done);
4107 4107 mutex_exit(&ai->ai_lock);
4108 4108 }
4109 4109 }
4110 4110
4111 4111 mutex_enter(&ipst->ips_ip_mi_lock);
4112 4112 mi_close_unlink(&ipst->ips_ip_g_head, (IDP)ill);
4113 4113 mutex_exit(&ipst->ips_ip_mi_lock);
4114 4114
4115 4115 /*
4116 4116 * credp could be null if the open didn't succeed and ip_modopen
|
↓ open down ↓ |
4116 lines elided |
↑ open up ↑ |
4117 4117 * itself calls ip_close.
4118 4118 */
4119 4119 if (ill->ill_credp != NULL)
4120 4120 crfree(ill->ill_credp);
4121 4121
4122 4122 mutex_destroy(&ill->ill_saved_ire_lock);
4123 4123 mutex_destroy(&ill->ill_lock);
4124 4124 rw_destroy(&ill->ill_mcast_lock);
4125 4125 mutex_destroy(&ill->ill_mcast_serializer);
4126 4126 list_destroy(&ill->ill_nce);
4127 - cv_destroy(&ill->ill_dlpi_capab_cv);
4128 - mutex_destroy(&ill->ill_dlpi_capab_lock);
4129 4127
4130 4128 /*
4131 4129 * Now we are done with the module close pieces that
4132 4130 * need the netstack_t.
4133 4131 */
4134 4132 netstack_rele(ipst->ips_netstack);
4135 4133
4136 4134 mi_close_free((IDP)ill);
4137 4135 q->q_ptr = WR(q)->q_ptr = NULL;
4138 4136
4139 4137 ipsq_exit(ipsq);
4140 4138
4141 4139 return (0);
4142 4140 }
4143 4141
4144 4142 /*
4145 4143 * This is called as part of close() for IP, UDP, ICMP, and RTS
4146 4144 * in order to quiesce the conn.
4147 4145 */
4148 4146 void
4149 4147 ip_quiesce_conn(conn_t *connp)
4150 4148 {
4151 4149 boolean_t drain_cleanup_reqd = B_FALSE;
4152 4150 boolean_t conn_ioctl_cleanup_reqd = B_FALSE;
4153 4151 boolean_t ilg_cleanup_reqd = B_FALSE;
4154 4152 ip_stack_t *ipst;
4155 4153
4156 4154 ASSERT(!IPCL_IS_TCP(connp));
4157 4155 ipst = connp->conn_netstack->netstack_ip;
4158 4156
4159 4157 /*
4160 4158 * Mark the conn as closing, and this conn must not be
4161 4159 * inserted in future into any list. Eg. conn_drain_insert(),
4162 4160 * won't insert this conn into the conn_drain_list.
4163 4161 *
4164 4162 * conn_idl, and conn_ilg cannot get set henceforth.
4165 4163 */
4166 4164 mutex_enter(&connp->conn_lock);
4167 4165 ASSERT(!(connp->conn_state_flags & CONN_QUIESCED));
4168 4166 connp->conn_state_flags |= CONN_CLOSING;
4169 4167 if (connp->conn_idl != NULL)
4170 4168 drain_cleanup_reqd = B_TRUE;
4171 4169 if (connp->conn_oper_pending_ill != NULL)
4172 4170 conn_ioctl_cleanup_reqd = B_TRUE;
4173 4171 if (connp->conn_dhcpinit_ill != NULL) {
4174 4172 ASSERT(connp->conn_dhcpinit_ill->ill_dhcpinit != 0);
4175 4173 atomic_dec_32(&connp->conn_dhcpinit_ill->ill_dhcpinit);
4176 4174 ill_set_inputfn(connp->conn_dhcpinit_ill);
4177 4175 connp->conn_dhcpinit_ill = NULL;
4178 4176 }
4179 4177 if (connp->conn_ilg != NULL)
4180 4178 ilg_cleanup_reqd = B_TRUE;
4181 4179 mutex_exit(&connp->conn_lock);
4182 4180
4183 4181 if (conn_ioctl_cleanup_reqd)
4184 4182 conn_ioctl_cleanup(connp);
4185 4183
4186 4184 if (is_system_labeled() && connp->conn_anon_port) {
4187 4185 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
4188 4186 connp->conn_mlp_type, connp->conn_proto,
4189 4187 ntohs(connp->conn_lport), B_FALSE);
4190 4188 connp->conn_anon_port = 0;
4191 4189 }
4192 4190 connp->conn_mlp_type = mlptSingle;
4193 4191
4194 4192 /*
4195 4193 * Remove this conn from any fanout list it is on.
4196 4194 * and then wait for any threads currently operating
4197 4195 * on this endpoint to finish
4198 4196 */
4199 4197 ipcl_hash_remove(connp);
4200 4198
4201 4199 /*
4202 4200 * Remove this conn from the drain list, and do any other cleanup that
4203 4201 * may be required. (TCP conns are never flow controlled, and
4204 4202 * conn_idl will be NULL.)
4205 4203 */
4206 4204 if (drain_cleanup_reqd && connp->conn_idl != NULL) {
4207 4205 idl_t *idl = connp->conn_idl;
4208 4206
4209 4207 mutex_enter(&idl->idl_lock);
4210 4208 conn_drain(connp, B_TRUE);
4211 4209 mutex_exit(&idl->idl_lock);
4212 4210 }
4213 4211
4214 4212 if (connp == ipst->ips_ip_g_mrouter)
4215 4213 (void) ip_mrouter_done(ipst);
4216 4214
4217 4215 if (ilg_cleanup_reqd)
4218 4216 ilg_delete_all(connp);
4219 4217
4220 4218 /*
4221 4219 * Now conn refcnt can increase only thru CONN_INC_REF_LOCKED.
4222 4220 * callers from write side can't be there now because close
4223 4221 * is in progress. The only other caller is ipcl_walk
4224 4222 * which checks for the condemned flag.
4225 4223 */
4226 4224 mutex_enter(&connp->conn_lock);
4227 4225 connp->conn_state_flags |= CONN_CONDEMNED;
4228 4226 while (connp->conn_ref != 1)
4229 4227 cv_wait(&connp->conn_cv, &connp->conn_lock);
4230 4228 connp->conn_state_flags |= CONN_QUIESCED;
4231 4229 mutex_exit(&connp->conn_lock);
4232 4230 }
4233 4231
4234 4232 /* ARGSUSED */
4235 4233 int
4236 4234 ip_close(queue_t *q, int flags, cred_t *credp __unused)
4237 4235 {
4238 4236 conn_t *connp;
4239 4237
4240 4238 /*
4241 4239 * Call the appropriate delete routine depending on whether this is
4242 4240 * a module or device.
4243 4241 */
4244 4242 if (WR(q)->q_next != NULL) {
4245 4243 /* This is a module close */
4246 4244 return (ip_modclose((ill_t *)q->q_ptr));
4247 4245 }
4248 4246
4249 4247 connp = q->q_ptr;
4250 4248 ip_quiesce_conn(connp);
4251 4249
4252 4250 qprocsoff(q);
4253 4251
4254 4252 /*
4255 4253 * Now we are truly single threaded on this stream, and can
4256 4254 * delete the things hanging off the connp, and finally the connp.
4257 4255 * We removed this connp from the fanout list, it cannot be
4258 4256 * accessed thru the fanouts, and we already waited for the
4259 4257 * conn_ref to drop to 0. We are already in close, so
4260 4258 * there cannot be any other thread from the top. qprocsoff
4261 4259 * has completed, and service has completed or won't run in
4262 4260 * future.
4263 4261 */
4264 4262 ASSERT(connp->conn_ref == 1);
4265 4263
4266 4264 inet_minor_free(connp->conn_minor_arena, connp->conn_dev);
4267 4265
4268 4266 connp->conn_ref--;
4269 4267 ipcl_conn_destroy(connp);
4270 4268
4271 4269 q->q_ptr = WR(q)->q_ptr = NULL;
4272 4270 return (0);
4273 4271 }
4274 4272
4275 4273 /*
4276 4274 * Wapper around putnext() so that ip_rts_request can merely use
4277 4275 * conn_recv.
4278 4276 */
4279 4277 /*ARGSUSED2*/
4280 4278 static void
4281 4279 ip_conn_input(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira)
4282 4280 {
4283 4281 conn_t *connp = (conn_t *)arg1;
4284 4282
4285 4283 putnext(connp->conn_rq, mp);
4286 4284 }
4287 4285
4288 4286 /* Dummy in case ICMP error delivery is attempted to a /dev/ip instance */
4289 4287 /* ARGSUSED */
4290 4288 static void
4291 4289 ip_conn_input_icmp(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira)
4292 4290 {
4293 4291 freemsg(mp);
4294 4292 }
4295 4293
4296 4294 /*
4297 4295 * Called when the module is about to be unloaded
4298 4296 */
4299 4297 void
4300 4298 ip_ddi_destroy(void)
4301 4299 {
4302 4300 /* This needs to be called before destroying any transports. */
4303 4301 mutex_enter(&cpu_lock);
4304 4302 unregister_cpu_setup_func(ip_tp_cpu_update, NULL);
4305 4303 mutex_exit(&cpu_lock);
4306 4304
4307 4305 tnet_fini();
4308 4306
4309 4307 icmp_ddi_g_destroy();
4310 4308 rts_ddi_g_destroy();
4311 4309 udp_ddi_g_destroy();
4312 4310 sctp_ddi_g_destroy();
4313 4311 tcp_ddi_g_destroy();
4314 4312 ilb_ddi_g_destroy();
4315 4313 dce_g_destroy();
4316 4314 ipsec_policy_g_destroy();
4317 4315 ipcl_g_destroy();
4318 4316 ip_net_g_destroy();
4319 4317 ip_ire_g_fini();
4320 4318 inet_minor_destroy(ip_minor_arena_sa);
4321 4319 #if defined(_LP64)
4322 4320 inet_minor_destroy(ip_minor_arena_la);
4323 4321 #endif
4324 4322
4325 4323 #ifdef DEBUG
4326 4324 list_destroy(&ip_thread_list);
4327 4325 rw_destroy(&ip_thread_rwlock);
4328 4326 tsd_destroy(&ip_thread_data);
4329 4327 #endif
4330 4328
4331 4329 netstack_unregister(NS_IP);
4332 4330 }
4333 4331
4334 4332 /*
4335 4333 * First step in cleanup.
4336 4334 */
4337 4335 /* ARGSUSED */
4338 4336 static void
4339 4337 ip_stack_shutdown(netstackid_t stackid, void *arg)
4340 4338 {
4341 4339 ip_stack_t *ipst = (ip_stack_t *)arg;
4342 4340 kt_did_t ktid;
4343 4341
4344 4342 #ifdef NS_DEBUG
4345 4343 printf("ip_stack_shutdown(%p, stack %d)\n", (void *)ipst, stackid);
4346 4344 #endif
4347 4345
4348 4346 /*
4349 4347 * Perform cleanup for special interfaces (loopback and IPMP).
4350 4348 */
4351 4349 ip_interface_cleanup(ipst);
4352 4350
4353 4351 /*
4354 4352 * The *_hook_shutdown()s start the process of notifying any
4355 4353 * consumers that things are going away.... nothing is destroyed.
4356 4354 */
4357 4355 ipv4_hook_shutdown(ipst);
4358 4356 ipv6_hook_shutdown(ipst);
4359 4357 arp_hook_shutdown(ipst);
4360 4358
4361 4359 mutex_enter(&ipst->ips_capab_taskq_lock);
4362 4360 ktid = ipst->ips_capab_taskq_thread->t_did;
4363 4361 ipst->ips_capab_taskq_quit = B_TRUE;
4364 4362 cv_signal(&ipst->ips_capab_taskq_cv);
4365 4363 mutex_exit(&ipst->ips_capab_taskq_lock);
4366 4364
4367 4365 /*
4368 4366 * In rare occurrences, particularly on virtual hardware where CPUs can
4369 4367 * be de-scheduled, the thread that we just signaled will not run until
4370 4368 * after we have gotten through parts of ip_stack_fini. If that happens
4371 4369 * then we'll try to grab the ips_capab_taskq_lock as part of returning
4372 4370 * from cv_wait which no longer exists.
4373 4371 */
4374 4372 thread_join(ktid);
4375 4373 }
4376 4374
4377 4375 /*
4378 4376 * Free the IP stack instance.
4379 4377 */
4380 4378 static void
4381 4379 ip_stack_fini(netstackid_t stackid, void *arg)
4382 4380 {
4383 4381 ip_stack_t *ipst = (ip_stack_t *)arg;
4384 4382 int ret;
4385 4383
4386 4384 #ifdef NS_DEBUG
4387 4385 printf("ip_stack_fini(%p, stack %d)\n", (void *)ipst, stackid);
4388 4386 #endif
4389 4387 /*
4390 4388 * At this point, all of the notifications that the events and
4391 4389 * protocols are going away have been run, meaning that we can
4392 4390 * now set about starting to clean things up.
4393 4391 */
4394 4392 ipobs_fini(ipst);
4395 4393 ipv4_hook_destroy(ipst);
4396 4394 ipv6_hook_destroy(ipst);
4397 4395 arp_hook_destroy(ipst);
4398 4396 ip_net_destroy(ipst);
4399 4397
4400 4398 ipmp_destroy(ipst);
4401 4399
4402 4400 ip_kstat_fini(stackid, ipst->ips_ip_mibkp);
4403 4401 ipst->ips_ip_mibkp = NULL;
4404 4402 icmp_kstat_fini(stackid, ipst->ips_icmp_mibkp);
4405 4403 ipst->ips_icmp_mibkp = NULL;
4406 4404 ip_kstat2_fini(stackid, ipst->ips_ip_kstat);
4407 4405 ipst->ips_ip_kstat = NULL;
4408 4406 bzero(&ipst->ips_ip_statistics, sizeof (ipst->ips_ip_statistics));
4409 4407 ip6_kstat_fini(stackid, ipst->ips_ip6_kstat);
4410 4408 ipst->ips_ip6_kstat = NULL;
4411 4409 bzero(&ipst->ips_ip6_statistics, sizeof (ipst->ips_ip6_statistics));
4412 4410
4413 4411 kmem_free(ipst->ips_propinfo_tbl,
4414 4412 ip_propinfo_count * sizeof (mod_prop_info_t));
4415 4413 ipst->ips_propinfo_tbl = NULL;
4416 4414
4417 4415 dce_stack_destroy(ipst);
4418 4416 ip_mrouter_stack_destroy(ipst);
4419 4417
4420 4418 /*
4421 4419 * Quiesce all of our timers. Note we set the quiesce flags before we
4422 4420 * call untimeout. The slowtimers may actually kick off another instance
4423 4421 * of the non-slow timers.
4424 4422 */
4425 4423 mutex_enter(&ipst->ips_igmp_timer_lock);
4426 4424 ipst->ips_igmp_timer_quiesce = B_TRUE;
4427 4425 mutex_exit(&ipst->ips_igmp_timer_lock);
4428 4426
4429 4427 mutex_enter(&ipst->ips_mld_timer_lock);
4430 4428 ipst->ips_mld_timer_quiesce = B_TRUE;
4431 4429 mutex_exit(&ipst->ips_mld_timer_lock);
4432 4430
4433 4431 mutex_enter(&ipst->ips_igmp_slowtimeout_lock);
4434 4432 ipst->ips_igmp_slowtimeout_quiesce = B_TRUE;
4435 4433 mutex_exit(&ipst->ips_igmp_slowtimeout_lock);
4436 4434
4437 4435 mutex_enter(&ipst->ips_mld_slowtimeout_lock);
4438 4436 ipst->ips_mld_slowtimeout_quiesce = B_TRUE;
4439 4437 mutex_exit(&ipst->ips_mld_slowtimeout_lock);
4440 4438
4441 4439 ret = untimeout(ipst->ips_igmp_timeout_id);
4442 4440 if (ret == -1) {
4443 4441 ASSERT(ipst->ips_igmp_timeout_id == 0);
4444 4442 } else {
4445 4443 ASSERT(ipst->ips_igmp_timeout_id != 0);
4446 4444 ipst->ips_igmp_timeout_id = 0;
4447 4445 }
4448 4446 ret = untimeout(ipst->ips_igmp_slowtimeout_id);
4449 4447 if (ret == -1) {
4450 4448 ASSERT(ipst->ips_igmp_slowtimeout_id == 0);
4451 4449 } else {
4452 4450 ASSERT(ipst->ips_igmp_slowtimeout_id != 0);
4453 4451 ipst->ips_igmp_slowtimeout_id = 0;
4454 4452 }
4455 4453 ret = untimeout(ipst->ips_mld_timeout_id);
4456 4454 if (ret == -1) {
4457 4455 ASSERT(ipst->ips_mld_timeout_id == 0);
4458 4456 } else {
4459 4457 ASSERT(ipst->ips_mld_timeout_id != 0);
4460 4458 ipst->ips_mld_timeout_id = 0;
4461 4459 }
4462 4460 ret = untimeout(ipst->ips_mld_slowtimeout_id);
4463 4461 if (ret == -1) {
4464 4462 ASSERT(ipst->ips_mld_slowtimeout_id == 0);
4465 4463 } else {
4466 4464 ASSERT(ipst->ips_mld_slowtimeout_id != 0);
4467 4465 ipst->ips_mld_slowtimeout_id = 0;
4468 4466 }
4469 4467
4470 4468 ip_ire_fini(ipst);
4471 4469 ip6_asp_free(ipst);
4472 4470 conn_drain_fini(ipst);
4473 4471 ipcl_destroy(ipst);
4474 4472
4475 4473 mutex_destroy(&ipst->ips_ndp4->ndp_g_lock);
4476 4474 mutex_destroy(&ipst->ips_ndp6->ndp_g_lock);
4477 4475 kmem_free(ipst->ips_ndp4, sizeof (ndp_g_t));
4478 4476 ipst->ips_ndp4 = NULL;
4479 4477 kmem_free(ipst->ips_ndp6, sizeof (ndp_g_t));
4480 4478 ipst->ips_ndp6 = NULL;
4481 4479
4482 4480 if (ipst->ips_loopback_ksp != NULL) {
4483 4481 kstat_delete_netstack(ipst->ips_loopback_ksp, stackid);
4484 4482 ipst->ips_loopback_ksp = NULL;
4485 4483 }
4486 4484
4487 4485 mutex_destroy(&ipst->ips_capab_taskq_lock);
4488 4486 cv_destroy(&ipst->ips_capab_taskq_cv);
4489 4487
4490 4488 rw_destroy(&ipst->ips_srcid_lock);
4491 4489
4492 4490 mutex_destroy(&ipst->ips_ip_mi_lock);
4493 4491 rw_destroy(&ipst->ips_ill_g_usesrc_lock);
4494 4492
4495 4493 mutex_destroy(&ipst->ips_igmp_timer_lock);
4496 4494 mutex_destroy(&ipst->ips_mld_timer_lock);
4497 4495 mutex_destroy(&ipst->ips_igmp_slowtimeout_lock);
4498 4496 mutex_destroy(&ipst->ips_mld_slowtimeout_lock);
4499 4497 mutex_destroy(&ipst->ips_ip_addr_avail_lock);
4500 4498 rw_destroy(&ipst->ips_ill_g_lock);
4501 4499
4502 4500 kmem_free(ipst->ips_phyint_g_list, sizeof (phyint_list_t));
4503 4501 ipst->ips_phyint_g_list = NULL;
4504 4502 kmem_free(ipst->ips_ill_g_heads, sizeof (ill_g_head_t) * MAX_G_HEADS);
4505 4503 ipst->ips_ill_g_heads = NULL;
4506 4504
4507 4505 ldi_ident_release(ipst->ips_ldi_ident);
4508 4506 kmem_free(ipst, sizeof (*ipst));
4509 4507 }
4510 4508
4511 4509 /*
4512 4510 * This function is called from the TSD destructor, and is used to debug
4513 4511 * reference count issues in IP. See block comment in <inet/ip_if.h> for
4514 4512 * details.
4515 4513 */
4516 4514 static void
4517 4515 ip_thread_exit(void *phash)
4518 4516 {
4519 4517 th_hash_t *thh = phash;
4520 4518
4521 4519 rw_enter(&ip_thread_rwlock, RW_WRITER);
4522 4520 list_remove(&ip_thread_list, thh);
4523 4521 rw_exit(&ip_thread_rwlock);
4524 4522 mod_hash_destroy_hash(thh->thh_hash);
4525 4523 kmem_free(thh, sizeof (*thh));
4526 4524 }
4527 4525
4528 4526 /*
4529 4527 * Called when the IP kernel module is loaded into the kernel
4530 4528 */
4531 4529 void
4532 4530 ip_ddi_init(void)
4533 4531 {
4534 4532 ip_squeue_flag = ip_squeue_switch(ip_squeue_enter);
4535 4533
4536 4534 /*
4537 4535 * For IP and TCP the minor numbers should start from 2 since we have 4
4538 4536 * initial devices: ip, ip6, tcp, tcp6.
4539 4537 */
4540 4538 /*
4541 4539 * If this is a 64-bit kernel, then create two separate arenas -
4542 4540 * one for TLIs in the range of INET_MIN_DEV+2 through 2^^18-1, and the
4543 4541 * other for socket apps in the range 2^^18 through 2^^32-1.
4544 4542 */
4545 4543 ip_minor_arena_la = NULL;
4546 4544 ip_minor_arena_sa = NULL;
4547 4545 #if defined(_LP64)
4548 4546 if ((ip_minor_arena_sa = inet_minor_create("ip_minor_arena_sa",
4549 4547 INET_MIN_DEV + 2, MAXMIN32, KM_SLEEP)) == NULL) {
4550 4548 cmn_err(CE_PANIC,
4551 4549 "ip_ddi_init: ip_minor_arena_sa creation failed\n");
4552 4550 }
4553 4551 if ((ip_minor_arena_la = inet_minor_create("ip_minor_arena_la",
4554 4552 MAXMIN32 + 1, MAXMIN64, KM_SLEEP)) == NULL) {
4555 4553 cmn_err(CE_PANIC,
4556 4554 "ip_ddi_init: ip_minor_arena_la creation failed\n");
4557 4555 }
4558 4556 #else
4559 4557 if ((ip_minor_arena_sa = inet_minor_create("ip_minor_arena_sa",
4560 4558 INET_MIN_DEV + 2, MAXMIN, KM_SLEEP)) == NULL) {
4561 4559 cmn_err(CE_PANIC,
4562 4560 "ip_ddi_init: ip_minor_arena_sa creation failed\n");
4563 4561 }
4564 4562 #endif
4565 4563 ip_poll_normal_ticks = MSEC_TO_TICK_ROUNDUP(ip_poll_normal_ms);
4566 4564
4567 4565 ipcl_g_init();
4568 4566 ip_ire_g_init();
4569 4567 ip_net_g_init();
4570 4568
4571 4569 #ifdef DEBUG
4572 4570 tsd_create(&ip_thread_data, ip_thread_exit);
4573 4571 rw_init(&ip_thread_rwlock, NULL, RW_DEFAULT, NULL);
4574 4572 list_create(&ip_thread_list, sizeof (th_hash_t),
4575 4573 offsetof(th_hash_t, thh_link));
4576 4574 #endif
4577 4575 ipsec_policy_g_init();
4578 4576 tcp_ddi_g_init();
4579 4577 sctp_ddi_g_init();
4580 4578 dce_g_init();
4581 4579
4582 4580 /*
4583 4581 * We want to be informed each time a stack is created or
4584 4582 * destroyed in the kernel, so we can maintain the
4585 4583 * set of udp_stack_t's.
4586 4584 */
4587 4585 netstack_register(NS_IP, ip_stack_init, ip_stack_shutdown,
4588 4586 ip_stack_fini);
4589 4587
4590 4588 tnet_init();
4591 4589
4592 4590 udp_ddi_g_init();
4593 4591 rts_ddi_g_init();
4594 4592 icmp_ddi_g_init();
4595 4593 ilb_ddi_g_init();
4596 4594
4597 4595 /* This needs to be called after all transports are initialized. */
4598 4596 mutex_enter(&cpu_lock);
4599 4597 register_cpu_setup_func(ip_tp_cpu_update, NULL);
4600 4598 mutex_exit(&cpu_lock);
4601 4599 }
4602 4600
4603 4601 /*
4604 4602 * Initialize the IP stack instance.
4605 4603 */
4606 4604 static void *
4607 4605 ip_stack_init(netstackid_t stackid, netstack_t *ns)
4608 4606 {
4609 4607 ip_stack_t *ipst;
4610 4608 size_t arrsz;
4611 4609 major_t major;
4612 4610
4613 4611 #ifdef NS_DEBUG
4614 4612 printf("ip_stack_init(stack %d)\n", stackid);
4615 4613 #endif
4616 4614
4617 4615 ipst = (ip_stack_t *)kmem_zalloc(sizeof (*ipst), KM_SLEEP);
4618 4616 ipst->ips_netstack = ns;
4619 4617
4620 4618 ipst->ips_ill_g_heads = kmem_zalloc(sizeof (ill_g_head_t) * MAX_G_HEADS,
4621 4619 KM_SLEEP);
4622 4620 ipst->ips_phyint_g_list = kmem_zalloc(sizeof (phyint_list_t),
4623 4621 KM_SLEEP);
4624 4622 ipst->ips_ndp4 = kmem_zalloc(sizeof (ndp_g_t), KM_SLEEP);
4625 4623 ipst->ips_ndp6 = kmem_zalloc(sizeof (ndp_g_t), KM_SLEEP);
4626 4624 mutex_init(&ipst->ips_ndp4->ndp_g_lock, NULL, MUTEX_DEFAULT, NULL);
4627 4625 mutex_init(&ipst->ips_ndp6->ndp_g_lock, NULL, MUTEX_DEFAULT, NULL);
4628 4626
4629 4627 mutex_init(&ipst->ips_igmp_timer_lock, NULL, MUTEX_DEFAULT, NULL);
4630 4628 ipst->ips_igmp_deferred_next = INFINITY;
4631 4629 mutex_init(&ipst->ips_mld_timer_lock, NULL, MUTEX_DEFAULT, NULL);
4632 4630 ipst->ips_mld_deferred_next = INFINITY;
4633 4631 mutex_init(&ipst->ips_igmp_slowtimeout_lock, NULL, MUTEX_DEFAULT, NULL);
4634 4632 mutex_init(&ipst->ips_mld_slowtimeout_lock, NULL, MUTEX_DEFAULT, NULL);
4635 4633 mutex_init(&ipst->ips_ip_mi_lock, NULL, MUTEX_DEFAULT, NULL);
4636 4634 mutex_init(&ipst->ips_ip_addr_avail_lock, NULL, MUTEX_DEFAULT, NULL);
4637 4635 rw_init(&ipst->ips_ill_g_lock, NULL, RW_DEFAULT, NULL);
4638 4636 rw_init(&ipst->ips_ill_g_usesrc_lock, NULL, RW_DEFAULT, NULL);
4639 4637
4640 4638 ipcl_init(ipst);
4641 4639 ip_ire_init(ipst);
4642 4640 ip6_asp_init(ipst);
4643 4641 ipif_init(ipst);
4644 4642 conn_drain_init(ipst);
4645 4643 ip_mrouter_stack_init(ipst);
4646 4644 dce_stack_init(ipst);
4647 4645
4648 4646 ipst->ips_ip_multirt_log_interval = 1000;
4649 4647
4650 4648 ipst->ips_ill_index = 1;
4651 4649
4652 4650 ipst->ips_saved_ip_forwarding = -1;
4653 4651 ipst->ips_reg_vif_num = ALL_VIFS; /* Index to Register vif */
4654 4652
4655 4653 arrsz = ip_propinfo_count * sizeof (mod_prop_info_t);
4656 4654 ipst->ips_propinfo_tbl = (mod_prop_info_t *)kmem_alloc(arrsz, KM_SLEEP);
4657 4655 bcopy(ip_propinfo_tbl, ipst->ips_propinfo_tbl, arrsz);
4658 4656
4659 4657 ipst->ips_ip_mibkp = ip_kstat_init(stackid, ipst);
4660 4658 ipst->ips_icmp_mibkp = icmp_kstat_init(stackid);
4661 4659 ipst->ips_ip_kstat = ip_kstat2_init(stackid, &ipst->ips_ip_statistics);
4662 4660 ipst->ips_ip6_kstat =
4663 4661 ip6_kstat_init(stackid, &ipst->ips_ip6_statistics);
4664 4662
4665 4663 ipst->ips_ip_src_id = 1;
4666 4664 rw_init(&ipst->ips_srcid_lock, NULL, RW_DEFAULT, NULL);
4667 4665
4668 4666 ipst->ips_src_generation = SRC_GENERATION_INITIAL;
4669 4667
4670 4668 ip_net_init(ipst, ns);
4671 4669 ipv4_hook_init(ipst);
4672 4670 ipv6_hook_init(ipst);
4673 4671 arp_hook_init(ipst);
4674 4672 ipmp_init(ipst);
4675 4673 ipobs_init(ipst);
4676 4674
4677 4675 /*
4678 4676 * Create the taskq dispatcher thread and initialize related stuff.
4679 4677 */
4680 4678 mutex_init(&ipst->ips_capab_taskq_lock, NULL, MUTEX_DEFAULT, NULL);
4681 4679 cv_init(&ipst->ips_capab_taskq_cv, NULL, CV_DEFAULT, NULL);
4682 4680 ipst->ips_capab_taskq_thread = thread_create(NULL, 0,
4683 4681 ill_taskq_dispatch, ipst, 0, &p0, TS_RUN, minclsyspri);
4684 4682
4685 4683 major = mod_name_to_major(INET_NAME);
4686 4684 (void) ldi_ident_from_major(major, &ipst->ips_ldi_ident);
4687 4685 return (ipst);
4688 4686 }
4689 4687
4690 4688 /*
4691 4689 * Allocate and initialize a DLPI template of the specified length. (May be
4692 4690 * called as writer.)
4693 4691 */
4694 4692 mblk_t *
4695 4693 ip_dlpi_alloc(size_t len, t_uscalar_t prim)
4696 4694 {
4697 4695 mblk_t *mp;
4698 4696
4699 4697 mp = allocb(len, BPRI_MED);
4700 4698 if (!mp)
4701 4699 return (NULL);
4702 4700
4703 4701 /*
4704 4702 * DLPIv2 says that DL_INFO_REQ and DL_TOKEN_REQ (the latter
4705 4703 * of which we don't seem to use) are sent with M_PCPROTO, and
4706 4704 * that other DLPI are M_PROTO.
4707 4705 */
4708 4706 if (prim == DL_INFO_REQ) {
4709 4707 mp->b_datap->db_type = M_PCPROTO;
4710 4708 } else {
4711 4709 mp->b_datap->db_type = M_PROTO;
4712 4710 }
4713 4711
4714 4712 mp->b_wptr = mp->b_rptr + len;
4715 4713 bzero(mp->b_rptr, len);
4716 4714 ((dl_unitdata_req_t *)mp->b_rptr)->dl_primitive = prim;
4717 4715 return (mp);
4718 4716 }
4719 4717
4720 4718 /*
4721 4719 * Allocate and initialize a DLPI notification. (May be called as writer.)
4722 4720 */
4723 4721 mblk_t *
4724 4722 ip_dlnotify_alloc(uint_t notification, uint_t data)
4725 4723 {
4726 4724 dl_notify_ind_t *notifyp;
4727 4725 mblk_t *mp;
4728 4726
4729 4727 if ((mp = ip_dlpi_alloc(DL_NOTIFY_IND_SIZE, DL_NOTIFY_IND)) == NULL)
4730 4728 return (NULL);
4731 4729
4732 4730 notifyp = (dl_notify_ind_t *)mp->b_rptr;
4733 4731 notifyp->dl_notification = notification;
4734 4732 notifyp->dl_data = data;
4735 4733 return (mp);
4736 4734 }
4737 4735
4738 4736 mblk_t *
4739 4737 ip_dlnotify_alloc2(uint_t notification, uint_t data1, uint_t data2)
4740 4738 {
4741 4739 dl_notify_ind_t *notifyp;
4742 4740 mblk_t *mp;
4743 4741
4744 4742 if ((mp = ip_dlpi_alloc(DL_NOTIFY_IND_SIZE, DL_NOTIFY_IND)) == NULL)
4745 4743 return (NULL);
4746 4744
4747 4745 notifyp = (dl_notify_ind_t *)mp->b_rptr;
4748 4746 notifyp->dl_notification = notification;
4749 4747 notifyp->dl_data1 = data1;
4750 4748 notifyp->dl_data2 = data2;
4751 4749 return (mp);
4752 4750 }
4753 4751
4754 4752 /*
4755 4753 * Debug formatting routine. Returns a character string representation of the
4756 4754 * addr in buf, of the form xxx.xxx.xxx.xxx. This routine takes the address
4757 4755 * in the form of a ipaddr_t and calls ip_dot_saddr with a pointer.
4758 4756 *
4759 4757 * Once the ndd table-printing interfaces are removed, this can be changed to
4760 4758 * standard dotted-decimal form.
4761 4759 */
4762 4760 char *
4763 4761 ip_dot_addr(ipaddr_t addr, char *buf)
4764 4762 {
4765 4763 uint8_t *ap = (uint8_t *)&addr;
4766 4764
4767 4765 (void) mi_sprintf(buf, "%03d.%03d.%03d.%03d",
4768 4766 ap[0] & 0xFF, ap[1] & 0xFF, ap[2] & 0xFF, ap[3] & 0xFF);
4769 4767 return (buf);
4770 4768 }
4771 4769
4772 4770 /*
4773 4771 * Write the given MAC address as a printable string in the usual colon-
4774 4772 * separated format.
4775 4773 */
4776 4774 const char *
4777 4775 mac_colon_addr(const uint8_t *addr, size_t alen, char *buf, size_t buflen)
4778 4776 {
4779 4777 char *bp;
4780 4778
4781 4779 if (alen == 0 || buflen < 4)
4782 4780 return ("?");
4783 4781 bp = buf;
4784 4782 for (;;) {
4785 4783 /*
4786 4784 * If there are more MAC address bytes available, but we won't
4787 4785 * have any room to print them, then add "..." to the string
4788 4786 * instead. See below for the 'magic number' explanation.
4789 4787 */
4790 4788 if ((alen == 2 && buflen < 6) || (alen > 2 && buflen < 7)) {
4791 4789 (void) strcpy(bp, "...");
4792 4790 break;
4793 4791 }
4794 4792 (void) sprintf(bp, "%02x", *addr++);
4795 4793 bp += 2;
4796 4794 if (--alen == 0)
4797 4795 break;
4798 4796 *bp++ = ':';
4799 4797 buflen -= 3;
4800 4798 /*
4801 4799 * At this point, based on the first 'if' statement above,
4802 4800 * either alen == 1 and buflen >= 3, or alen > 1 and
4803 4801 * buflen >= 4. The first case leaves room for the final "xx"
4804 4802 * number and trailing NUL byte. The second leaves room for at
4805 4803 * least "...". Thus the apparently 'magic' numbers chosen for
4806 4804 * that statement.
4807 4805 */
4808 4806 }
4809 4807 return (buf);
4810 4808 }
4811 4809
4812 4810 /*
4813 4811 * Called when it is conceptually a ULP that would sent the packet
4814 4812 * e.g., port unreachable and protocol unreachable. Check that the packet
4815 4813 * would have passed the IPsec global policy before sending the error.
4816 4814 *
4817 4815 * Send an ICMP error after patching up the packet appropriately.
4818 4816 * Uses ip_drop_input and bumps the appropriate MIB.
4819 4817 */
4820 4818 void
4821 4819 ip_fanout_send_icmp_v4(mblk_t *mp, uint_t icmp_type, uint_t icmp_code,
4822 4820 ip_recv_attr_t *ira)
4823 4821 {
4824 4822 ipha_t *ipha;
4825 4823 boolean_t secure;
4826 4824 ill_t *ill = ira->ira_ill;
4827 4825 ip_stack_t *ipst = ill->ill_ipst;
4828 4826 netstack_t *ns = ipst->ips_netstack;
4829 4827 ipsec_stack_t *ipss = ns->netstack_ipsec;
4830 4828
4831 4829 secure = ira->ira_flags & IRAF_IPSEC_SECURE;
4832 4830
4833 4831 /*
4834 4832 * We are generating an icmp error for some inbound packet.
4835 4833 * Called from all ip_fanout_(udp, tcp, proto) functions.
4836 4834 * Before we generate an error, check with global policy
4837 4835 * to see whether this is allowed to enter the system. As
4838 4836 * there is no "conn", we are checking with global policy.
4839 4837 */
4840 4838 ipha = (ipha_t *)mp->b_rptr;
4841 4839 if (secure || ipss->ipsec_inbound_v4_policy_present) {
4842 4840 mp = ipsec_check_global_policy(mp, NULL, ipha, NULL, ira, ns);
4843 4841 if (mp == NULL)
4844 4842 return;
4845 4843 }
4846 4844
4847 4845 /* We never send errors for protocols that we do implement */
4848 4846 if (ira->ira_protocol == IPPROTO_ICMP ||
4849 4847 ira->ira_protocol == IPPROTO_IGMP) {
4850 4848 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
4851 4849 ip_drop_input("ip_fanout_send_icmp_v4", mp, ill);
4852 4850 freemsg(mp);
4853 4851 return;
4854 4852 }
4855 4853 /*
4856 4854 * Have to correct checksum since
4857 4855 * the packet might have been
4858 4856 * fragmented and the reassembly code in ip_rput
4859 4857 * does not restore the IP checksum.
4860 4858 */
4861 4859 ipha->ipha_hdr_checksum = 0;
4862 4860 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
4863 4861
4864 4862 switch (icmp_type) {
4865 4863 case ICMP_DEST_UNREACHABLE:
4866 4864 switch (icmp_code) {
4867 4865 case ICMP_PROTOCOL_UNREACHABLE:
4868 4866 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInUnknownProtos);
4869 4867 ip_drop_input("ipIfStatsInUnknownProtos", mp, ill);
4870 4868 break;
4871 4869 case ICMP_PORT_UNREACHABLE:
4872 4870 BUMP_MIB(ill->ill_ip_mib, udpIfStatsNoPorts);
4873 4871 ip_drop_input("ipIfStatsNoPorts", mp, ill);
4874 4872 break;
4875 4873 }
4876 4874
4877 4875 icmp_unreachable(mp, icmp_code, ira);
4878 4876 break;
4879 4877 default:
4880 4878 #ifdef DEBUG
4881 4879 panic("ip_fanout_send_icmp_v4: wrong type");
4882 4880 /*NOTREACHED*/
4883 4881 #else
4884 4882 freemsg(mp);
4885 4883 break;
4886 4884 #endif
4887 4885 }
4888 4886 }
4889 4887
4890 4888 /*
4891 4889 * Used to send an ICMP error message when a packet is received for
4892 4890 * a protocol that is not supported. The mblk passed as argument
4893 4891 * is consumed by this function.
4894 4892 */
4895 4893 void
4896 4894 ip_proto_not_sup(mblk_t *mp, ip_recv_attr_t *ira)
4897 4895 {
4898 4896 ipha_t *ipha;
4899 4897
4900 4898 ipha = (ipha_t *)mp->b_rptr;
4901 4899 if (ira->ira_flags & IRAF_IS_IPV4) {
4902 4900 ASSERT(IPH_HDR_VERSION(ipha) == IP_VERSION);
4903 4901 ip_fanout_send_icmp_v4(mp, ICMP_DEST_UNREACHABLE,
4904 4902 ICMP_PROTOCOL_UNREACHABLE, ira);
4905 4903 } else {
4906 4904 ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
4907 4905 ip_fanout_send_icmp_v6(mp, ICMP6_PARAM_PROB,
4908 4906 ICMP6_PARAMPROB_NEXTHEADER, ira);
4909 4907 }
4910 4908 }
4911 4909
4912 4910 /*
4913 4911 * Deliver a rawip packet to the given conn, possibly applying ipsec policy.
4914 4912 * Handles IPv4 and IPv6.
4915 4913 * We are responsible for disposing of mp, such as by freemsg() or putnext()
4916 4914 * Caller is responsible for dropping references to the conn.
4917 4915 */
4918 4916 void
4919 4917 ip_fanout_proto_conn(conn_t *connp, mblk_t *mp, ipha_t *ipha, ip6_t *ip6h,
4920 4918 ip_recv_attr_t *ira)
4921 4919 {
4922 4920 ill_t *ill = ira->ira_ill;
4923 4921 ip_stack_t *ipst = ill->ill_ipst;
4924 4922 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
4925 4923 boolean_t secure;
4926 4924 uint_t protocol = ira->ira_protocol;
4927 4925 iaflags_t iraflags = ira->ira_flags;
4928 4926 queue_t *rq;
4929 4927
4930 4928 secure = iraflags & IRAF_IPSEC_SECURE;
4931 4929
4932 4930 rq = connp->conn_rq;
4933 4931 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld : !canputnext(rq)) {
4934 4932 switch (protocol) {
4935 4933 case IPPROTO_ICMPV6:
4936 4934 BUMP_MIB(ill->ill_icmp6_mib, ipv6IfIcmpInOverflows);
4937 4935 break;
4938 4936 case IPPROTO_ICMP:
4939 4937 BUMP_MIB(&ipst->ips_icmp_mib, icmpInOverflows);
4940 4938 break;
4941 4939 default:
4942 4940 BUMP_MIB(ill->ill_ip_mib, rawipIfStatsInOverflows);
4943 4941 break;
4944 4942 }
4945 4943 freemsg(mp);
4946 4944 return;
4947 4945 }
4948 4946
4949 4947 ASSERT(!(IPCL_IS_IPTUN(connp)));
4950 4948
4951 4949 if (((iraflags & IRAF_IS_IPV4) ?
4952 4950 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
4953 4951 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
4954 4952 secure) {
4955 4953 mp = ipsec_check_inbound_policy(mp, connp, ipha,
4956 4954 ip6h, ira);
4957 4955 if (mp == NULL) {
4958 4956 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
4959 4957 /* Note that mp is NULL */
4960 4958 ip_drop_input("ipIfStatsInDiscards", mp, ill);
4961 4959 return;
4962 4960 }
4963 4961 }
4964 4962
4965 4963 if (iraflags & IRAF_ICMP_ERROR) {
4966 4964 (connp->conn_recvicmp)(connp, mp, NULL, ira);
4967 4965 } else {
4968 4966 ill_t *rill = ira->ira_rill;
4969 4967
4970 4968 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
4971 4969 ira->ira_ill = ira->ira_rill = NULL;
4972 4970 /* Send it upstream */
4973 4971 (connp->conn_recv)(connp, mp, NULL, ira);
4974 4972 ira->ira_ill = ill;
4975 4973 ira->ira_rill = rill;
4976 4974 }
4977 4975 }
4978 4976
4979 4977 /*
4980 4978 * Handle protocols with which IP is less intimate. There
4981 4979 * can be more than one stream bound to a particular
4982 4980 * protocol. When this is the case, normally each one gets a copy
4983 4981 * of any incoming packets.
4984 4982 *
4985 4983 * IPsec NOTE :
4986 4984 *
4987 4985 * Don't allow a secure packet going up a non-secure connection.
4988 4986 * We don't allow this because
4989 4987 *
4990 4988 * 1) Reply might go out in clear which will be dropped at
4991 4989 * the sending side.
4992 4990 * 2) If the reply goes out in clear it will give the
4993 4991 * adversary enough information for getting the key in
4994 4992 * most of the cases.
4995 4993 *
4996 4994 * Moreover getting a secure packet when we expect clear
4997 4995 * implies that SA's were added without checking for
4998 4996 * policy on both ends. This should not happen once ISAKMP
4999 4997 * is used to negotiate SAs as SAs will be added only after
5000 4998 * verifying the policy.
5001 4999 *
5002 5000 * Zones notes:
5003 5001 * Earlier in ip_input on a system with multiple shared-IP zones we
5004 5002 * duplicate the multicast and broadcast packets and send them up
5005 5003 * with each explicit zoneid that exists on that ill.
5006 5004 * This means that here we can match the zoneid with SO_ALLZONES being special.
5007 5005 */
5008 5006 void
5009 5007 ip_fanout_proto_v4(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
5010 5008 {
5011 5009 mblk_t *mp1;
5012 5010 ipaddr_t laddr;
5013 5011 conn_t *connp, *first_connp, *next_connp;
5014 5012 connf_t *connfp;
5015 5013 ill_t *ill = ira->ira_ill;
5016 5014 ip_stack_t *ipst = ill->ill_ipst;
5017 5015
5018 5016 laddr = ipha->ipha_dst;
5019 5017
5020 5018 connfp = &ipst->ips_ipcl_proto_fanout_v4[ira->ira_protocol];
5021 5019 mutex_enter(&connfp->connf_lock);
5022 5020 connp = connfp->connf_head;
5023 5021 for (connp = connfp->connf_head; connp != NULL;
5024 5022 connp = connp->conn_next) {
5025 5023 /* Note: IPCL_PROTO_MATCH includes conn_wantpacket */
5026 5024 if (IPCL_PROTO_MATCH(connp, ira, ipha) &&
5027 5025 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5028 5026 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp))) {
5029 5027 break;
5030 5028 }
5031 5029 }
5032 5030
5033 5031 if (connp == NULL) {
5034 5032 /*
5035 5033 * No one bound to these addresses. Is
5036 5034 * there a client that wants all
5037 5035 * unclaimed datagrams?
5038 5036 */
5039 5037 mutex_exit(&connfp->connf_lock);
5040 5038 ip_fanout_send_icmp_v4(mp, ICMP_DEST_UNREACHABLE,
5041 5039 ICMP_PROTOCOL_UNREACHABLE, ira);
5042 5040 return;
5043 5041 }
5044 5042
5045 5043 ASSERT(IPCL_IS_NONSTR(connp) || connp->conn_rq != NULL);
5046 5044
5047 5045 CONN_INC_REF(connp);
5048 5046 first_connp = connp;
5049 5047 connp = connp->conn_next;
5050 5048
5051 5049 for (;;) {
5052 5050 while (connp != NULL) {
5053 5051 /* Note: IPCL_PROTO_MATCH includes conn_wantpacket */
5054 5052 if (IPCL_PROTO_MATCH(connp, ira, ipha) &&
5055 5053 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5056 5054 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5057 5055 ira, connp)))
5058 5056 break;
5059 5057 connp = connp->conn_next;
5060 5058 }
5061 5059
5062 5060 if (connp == NULL) {
5063 5061 /* No more interested clients */
5064 5062 connp = first_connp;
5065 5063 break;
5066 5064 }
5067 5065 if (((mp1 = dupmsg(mp)) == NULL) &&
5068 5066 ((mp1 = copymsg(mp)) == NULL)) {
5069 5067 /* Memory allocation failed */
5070 5068 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5071 5069 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5072 5070 connp = first_connp;
5073 5071 break;
5074 5072 }
5075 5073
5076 5074 CONN_INC_REF(connp);
5077 5075 mutex_exit(&connfp->connf_lock);
5078 5076
5079 5077 ip_fanout_proto_conn(connp, mp1, (ipha_t *)mp1->b_rptr, NULL,
5080 5078 ira);
5081 5079
5082 5080 mutex_enter(&connfp->connf_lock);
5083 5081 /* Follow the next pointer before releasing the conn. */
5084 5082 next_connp = connp->conn_next;
5085 5083 CONN_DEC_REF(connp);
5086 5084 connp = next_connp;
5087 5085 }
5088 5086
5089 5087 /* Last one. Send it upstream. */
5090 5088 mutex_exit(&connfp->connf_lock);
5091 5089
5092 5090 ip_fanout_proto_conn(connp, mp, ipha, NULL, ira);
5093 5091
5094 5092 CONN_DEC_REF(connp);
5095 5093 }
5096 5094
5097 5095 /*
5098 5096 * If we have a IPsec NAT-Traversal packet, strip the zero-SPI or
5099 5097 * pass it along to ESP if the SPI is non-zero. Returns the mblk if the mblk
5100 5098 * is not consumed.
5101 5099 *
5102 5100 * One of three things can happen, all of which affect the passed-in mblk:
5103 5101 *
5104 5102 * 1.) The packet is stock UDP and gets its zero-SPI stripped. Return mblk..
5105 5103 *
5106 5104 * 2.) The packet is ESP-in-UDP, gets transformed into an equivalent
5107 5105 * ESP packet, and is passed along to ESP for consumption. Return NULL.
5108 5106 *
5109 5107 * 3.) The packet is an ESP-in-UDP Keepalive. Drop it and return NULL.
5110 5108 */
5111 5109 mblk_t *
5112 5110 zero_spi_check(mblk_t *mp, ip_recv_attr_t *ira)
5113 5111 {
5114 5112 int shift, plen, iph_len;
5115 5113 ipha_t *ipha;
5116 5114 udpha_t *udpha;
5117 5115 uint32_t *spi;
5118 5116 uint32_t esp_ports;
5119 5117 uint8_t *orptr;
5120 5118 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
5121 5119 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
5122 5120
5123 5121 ipha = (ipha_t *)mp->b_rptr;
5124 5122 iph_len = ira->ira_ip_hdr_length;
5125 5123 plen = ira->ira_pktlen;
5126 5124
5127 5125 if (plen - iph_len - sizeof (udpha_t) < sizeof (uint32_t)) {
5128 5126 /*
5129 5127 * Most likely a keepalive for the benefit of an intervening
5130 5128 * NAT. These aren't for us, per se, so drop it.
5131 5129 *
5132 5130 * RFC 3947/8 doesn't say for sure what to do for 2-3
5133 5131 * byte packets (keepalives are 1-byte), but we'll drop them
5134 5132 * also.
5135 5133 */
5136 5134 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
5137 5135 DROPPER(ipss, ipds_esp_nat_t_ka), &ipss->ipsec_dropper);
5138 5136 return (NULL);
5139 5137 }
5140 5138
5141 5139 if (MBLKL(mp) < iph_len + sizeof (udpha_t) + sizeof (*spi)) {
5142 5140 /* might as well pull it all up - it might be ESP. */
5143 5141 if (!pullupmsg(mp, -1)) {
5144 5142 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
5145 5143 DROPPER(ipss, ipds_esp_nomem),
5146 5144 &ipss->ipsec_dropper);
5147 5145 return (NULL);
5148 5146 }
5149 5147
5150 5148 ipha = (ipha_t *)mp->b_rptr;
5151 5149 }
5152 5150 spi = (uint32_t *)(mp->b_rptr + iph_len + sizeof (udpha_t));
5153 5151 if (*spi == 0) {
5154 5152 /* UDP packet - remove 0-spi. */
5155 5153 shift = sizeof (uint32_t);
5156 5154 } else {
5157 5155 /* ESP-in-UDP packet - reduce to ESP. */
5158 5156 ipha->ipha_protocol = IPPROTO_ESP;
5159 5157 shift = sizeof (udpha_t);
5160 5158 }
5161 5159
5162 5160 /* Fix IP header */
5163 5161 ira->ira_pktlen = (plen - shift);
5164 5162 ipha->ipha_length = htons(ira->ira_pktlen);
5165 5163 ipha->ipha_hdr_checksum = 0;
5166 5164
5167 5165 orptr = mp->b_rptr;
5168 5166 mp->b_rptr += shift;
5169 5167
5170 5168 udpha = (udpha_t *)(orptr + iph_len);
5171 5169 if (*spi == 0) {
5172 5170 ASSERT((uint8_t *)ipha == orptr);
5173 5171 udpha->uha_length = htons(plen - shift - iph_len);
5174 5172 iph_len += sizeof (udpha_t); /* For the call to ovbcopy(). */
5175 5173 esp_ports = 0;
5176 5174 } else {
5177 5175 esp_ports = *((uint32_t *)udpha);
5178 5176 ASSERT(esp_ports != 0);
5179 5177 }
5180 5178 ovbcopy(orptr, orptr + shift, iph_len);
5181 5179 if (esp_ports != 0) /* Punt up for ESP processing. */ {
5182 5180 ipha = (ipha_t *)(orptr + shift);
5183 5181
5184 5182 ira->ira_flags |= IRAF_ESP_UDP_PORTS;
5185 5183 ira->ira_esp_udp_ports = esp_ports;
5186 5184 ip_fanout_v4(mp, ipha, ira);
5187 5185 return (NULL);
5188 5186 }
5189 5187 return (mp);
5190 5188 }
5191 5189
5192 5190 /*
5193 5191 * Deliver a udp packet to the given conn, possibly applying ipsec policy.
5194 5192 * Handles IPv4 and IPv6.
5195 5193 * We are responsible for disposing of mp, such as by freemsg() or putnext()
5196 5194 * Caller is responsible for dropping references to the conn.
5197 5195 */
5198 5196 void
5199 5197 ip_fanout_udp_conn(conn_t *connp, mblk_t *mp, ipha_t *ipha, ip6_t *ip6h,
5200 5198 ip_recv_attr_t *ira)
5201 5199 {
5202 5200 ill_t *ill = ira->ira_ill;
5203 5201 ip_stack_t *ipst = ill->ill_ipst;
5204 5202 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
5205 5203 boolean_t secure;
5206 5204 iaflags_t iraflags = ira->ira_flags;
5207 5205
5208 5206 secure = iraflags & IRAF_IPSEC_SECURE;
5209 5207
5210 5208 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld :
5211 5209 !canputnext(connp->conn_rq)) {
5212 5210 BUMP_MIB(ill->ill_ip_mib, udpIfStatsInOverflows);
5213 5211 freemsg(mp);
5214 5212 return;
5215 5213 }
5216 5214
5217 5215 if (((iraflags & IRAF_IS_IPV4) ?
5218 5216 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
5219 5217 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
5220 5218 secure) {
5221 5219 mp = ipsec_check_inbound_policy(mp, connp, ipha,
5222 5220 ip6h, ira);
5223 5221 if (mp == NULL) {
5224 5222 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5225 5223 /* Note that mp is NULL */
5226 5224 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5227 5225 return;
5228 5226 }
5229 5227 }
5230 5228
5231 5229 /*
5232 5230 * Since this code is not used for UDP unicast we don't need a NAT_T
5233 5231 * check. Only ip_fanout_v4 has that check.
5234 5232 */
5235 5233 if (ira->ira_flags & IRAF_ICMP_ERROR) {
5236 5234 (connp->conn_recvicmp)(connp, mp, NULL, ira);
5237 5235 } else {
5238 5236 ill_t *rill = ira->ira_rill;
5239 5237
5240 5238 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
5241 5239 ira->ira_ill = ira->ira_rill = NULL;
5242 5240 /* Send it upstream */
5243 5241 (connp->conn_recv)(connp, mp, NULL, ira);
5244 5242 ira->ira_ill = ill;
5245 5243 ira->ira_rill = rill;
5246 5244 }
5247 5245 }
5248 5246
5249 5247 /*
5250 5248 * Fanout for UDP packets that are multicast or broadcast, and ICMP errors.
5251 5249 * (Unicast fanout is handled in ip_input_v4.)
5252 5250 *
5253 5251 * If SO_REUSEADDR is set all multicast and broadcast packets
5254 5252 * will be delivered to all conns bound to the same port.
5255 5253 *
5256 5254 * If there is at least one matching AF_INET receiver, then we will
5257 5255 * ignore any AF_INET6 receivers.
5258 5256 * In the special case where an AF_INET socket binds to 0.0.0.0/<port> and an
5259 5257 * AF_INET6 socket binds to ::/<port>, only the AF_INET socket receives the IPv4
5260 5258 * packets.
5261 5259 *
5262 5260 * Zones notes:
5263 5261 * Earlier in ip_input on a system with multiple shared-IP zones we
5264 5262 * duplicate the multicast and broadcast packets and send them up
5265 5263 * with each explicit zoneid that exists on that ill.
5266 5264 * This means that here we can match the zoneid with SO_ALLZONES being special.
5267 5265 */
5268 5266 void
5269 5267 ip_fanout_udp_multi_v4(mblk_t *mp, ipha_t *ipha, uint16_t lport, uint16_t fport,
5270 5268 ip_recv_attr_t *ira)
5271 5269 {
5272 5270 ipaddr_t laddr;
5273 5271 in6_addr_t v6faddr;
5274 5272 conn_t *connp;
5275 5273 connf_t *connfp;
5276 5274 ipaddr_t faddr;
5277 5275 ill_t *ill = ira->ira_ill;
5278 5276 ip_stack_t *ipst = ill->ill_ipst;
5279 5277
5280 5278 ASSERT(ira->ira_flags & (IRAF_MULTIBROADCAST|IRAF_ICMP_ERROR));
5281 5279
5282 5280 laddr = ipha->ipha_dst;
5283 5281 faddr = ipha->ipha_src;
5284 5282
5285 5283 connfp = &ipst->ips_ipcl_udp_fanout[IPCL_UDP_HASH(lport, ipst)];
5286 5284 mutex_enter(&connfp->connf_lock);
5287 5285 connp = connfp->connf_head;
5288 5286
5289 5287 /*
5290 5288 * If SO_REUSEADDR has been set on the first we send the
5291 5289 * packet to all clients that have joined the group and
5292 5290 * match the port.
5293 5291 */
5294 5292 while (connp != NULL) {
5295 5293 if ((IPCL_UDP_MATCH(connp, lport, laddr, fport, faddr)) &&
5296 5294 conn_wantpacket(connp, ira, ipha) &&
5297 5295 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5298 5296 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp)))
5299 5297 break;
5300 5298 connp = connp->conn_next;
5301 5299 }
5302 5300
5303 5301 if (connp == NULL)
5304 5302 goto notfound;
5305 5303
5306 5304 CONN_INC_REF(connp);
5307 5305
5308 5306 if (connp->conn_reuseaddr) {
5309 5307 conn_t *first_connp = connp;
5310 5308 conn_t *next_connp;
5311 5309 mblk_t *mp1;
5312 5310
5313 5311 connp = connp->conn_next;
5314 5312 for (;;) {
5315 5313 while (connp != NULL) {
5316 5314 if (IPCL_UDP_MATCH(connp, lport, laddr,
5317 5315 fport, faddr) &&
5318 5316 conn_wantpacket(connp, ira, ipha) &&
5319 5317 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5320 5318 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5321 5319 ira, connp)))
5322 5320 break;
5323 5321 connp = connp->conn_next;
5324 5322 }
5325 5323 if (connp == NULL) {
5326 5324 /* No more interested clients */
5327 5325 connp = first_connp;
5328 5326 break;
5329 5327 }
5330 5328 if (((mp1 = dupmsg(mp)) == NULL) &&
5331 5329 ((mp1 = copymsg(mp)) == NULL)) {
5332 5330 /* Memory allocation failed */
5333 5331 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5334 5332 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5335 5333 connp = first_connp;
5336 5334 break;
5337 5335 }
5338 5336 CONN_INC_REF(connp);
5339 5337 mutex_exit(&connfp->connf_lock);
5340 5338
5341 5339 IP_STAT(ipst, ip_udp_fanmb);
5342 5340 ip_fanout_udp_conn(connp, mp1, (ipha_t *)mp1->b_rptr,
5343 5341 NULL, ira);
5344 5342 mutex_enter(&connfp->connf_lock);
5345 5343 /* Follow the next pointer before releasing the conn */
5346 5344 next_connp = connp->conn_next;
5347 5345 CONN_DEC_REF(connp);
5348 5346 connp = next_connp;
5349 5347 }
5350 5348 }
5351 5349
5352 5350 /* Last one. Send it upstream. */
5353 5351 mutex_exit(&connfp->connf_lock);
5354 5352 IP_STAT(ipst, ip_udp_fanmb);
5355 5353 ip_fanout_udp_conn(connp, mp, ipha, NULL, ira);
5356 5354 CONN_DEC_REF(connp);
5357 5355 return;
5358 5356
5359 5357 notfound:
5360 5358 mutex_exit(&connfp->connf_lock);
5361 5359 /*
5362 5360 * IPv6 endpoints bound to multicast IPv4-mapped addresses
5363 5361 * have already been matched above, since they live in the IPv4
5364 5362 * fanout tables. This implies we only need to
5365 5363 * check for IPv6 in6addr_any endpoints here.
5366 5364 * Thus we compare using ipv6_all_zeros instead of the destination
5367 5365 * address, except for the multicast group membership lookup which
5368 5366 * uses the IPv4 destination.
5369 5367 */
5370 5368 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_src, &v6faddr);
5371 5369 connfp = &ipst->ips_ipcl_udp_fanout[IPCL_UDP_HASH(lport, ipst)];
5372 5370 mutex_enter(&connfp->connf_lock);
5373 5371 connp = connfp->connf_head;
5374 5372 /*
5375 5373 * IPv4 multicast packet being delivered to an AF_INET6
5376 5374 * in6addr_any endpoint.
5377 5375 * Need to check conn_wantpacket(). Note that we use conn_wantpacket()
5378 5376 * and not conn_wantpacket_v6() since any multicast membership is
5379 5377 * for an IPv4-mapped multicast address.
5380 5378 */
5381 5379 while (connp != NULL) {
5382 5380 if (IPCL_UDP_MATCH_V6(connp, lport, ipv6_all_zeros,
5383 5381 fport, v6faddr) &&
5384 5382 conn_wantpacket(connp, ira, ipha) &&
5385 5383 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5386 5384 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp)))
5387 5385 break;
5388 5386 connp = connp->conn_next;
5389 5387 }
5390 5388
5391 5389 if (connp == NULL) {
5392 5390 /*
5393 5391 * No one bound to this port. Is
5394 5392 * there a client that wants all
5395 5393 * unclaimed datagrams?
5396 5394 */
5397 5395 mutex_exit(&connfp->connf_lock);
5398 5396
5399 5397 if (ipst->ips_ipcl_proto_fanout_v4[IPPROTO_UDP].connf_head !=
5400 5398 NULL) {
5401 5399 ASSERT(ira->ira_protocol == IPPROTO_UDP);
5402 5400 ip_fanout_proto_v4(mp, ipha, ira);
5403 5401 } else {
5404 5402 /*
5405 5403 * We used to attempt to send an icmp error here, but
5406 5404 * since this is known to be a multicast packet
5407 5405 * and we don't send icmp errors in response to
5408 5406 * multicast, just drop the packet and give up sooner.
5409 5407 */
5410 5408 BUMP_MIB(ill->ill_ip_mib, udpIfStatsNoPorts);
5411 5409 freemsg(mp);
5412 5410 }
5413 5411 return;
5414 5412 }
5415 5413 CONN_INC_REF(connp);
5416 5414 ASSERT(IPCL_IS_NONSTR(connp) || connp->conn_rq != NULL);
5417 5415
5418 5416 /*
5419 5417 * If SO_REUSEADDR has been set on the first we send the
5420 5418 * packet to all clients that have joined the group and
5421 5419 * match the port.
5422 5420 */
5423 5421 if (connp->conn_reuseaddr) {
5424 5422 conn_t *first_connp = connp;
5425 5423 conn_t *next_connp;
5426 5424 mblk_t *mp1;
5427 5425
5428 5426 connp = connp->conn_next;
5429 5427 for (;;) {
5430 5428 while (connp != NULL) {
5431 5429 if (IPCL_UDP_MATCH_V6(connp, lport,
5432 5430 ipv6_all_zeros, fport, v6faddr) &&
5433 5431 conn_wantpacket(connp, ira, ipha) &&
5434 5432 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5435 5433 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5436 5434 ira, connp)))
5437 5435 break;
5438 5436 connp = connp->conn_next;
5439 5437 }
5440 5438 if (connp == NULL) {
5441 5439 /* No more interested clients */
5442 5440 connp = first_connp;
5443 5441 break;
5444 5442 }
5445 5443 if (((mp1 = dupmsg(mp)) == NULL) &&
5446 5444 ((mp1 = copymsg(mp)) == NULL)) {
5447 5445 /* Memory allocation failed */
5448 5446 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5449 5447 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5450 5448 connp = first_connp;
5451 5449 break;
5452 5450 }
5453 5451 CONN_INC_REF(connp);
5454 5452 mutex_exit(&connfp->connf_lock);
5455 5453
5456 5454 IP_STAT(ipst, ip_udp_fanmb);
5457 5455 ip_fanout_udp_conn(connp, mp1, (ipha_t *)mp1->b_rptr,
5458 5456 NULL, ira);
5459 5457 mutex_enter(&connfp->connf_lock);
5460 5458 /* Follow the next pointer before releasing the conn */
5461 5459 next_connp = connp->conn_next;
5462 5460 CONN_DEC_REF(connp);
5463 5461 connp = next_connp;
5464 5462 }
5465 5463 }
5466 5464
5467 5465 /* Last one. Send it upstream. */
5468 5466 mutex_exit(&connfp->connf_lock);
5469 5467 IP_STAT(ipst, ip_udp_fanmb);
5470 5468 ip_fanout_udp_conn(connp, mp, ipha, NULL, ira);
5471 5469 CONN_DEC_REF(connp);
5472 5470 }
5473 5471
5474 5472 /*
5475 5473 * Split an incoming packet's IPv4 options into the label and the other options.
5476 5474 * If 'allocate' is set it does memory allocation for the ip_pkt_t, including
5477 5475 * clearing out any leftover label or options.
5478 5476 * Otherwise it just makes ipp point into the packet.
5479 5477 *
5480 5478 * Returns zero if ok; ENOMEM if the buffer couldn't be allocated.
5481 5479 */
5482 5480 int
5483 5481 ip_find_hdr_v4(ipha_t *ipha, ip_pkt_t *ipp, boolean_t allocate)
5484 5482 {
5485 5483 uchar_t *opt;
5486 5484 uint32_t totallen;
5487 5485 uint32_t optval;
5488 5486 uint32_t optlen;
5489 5487
5490 5488 ipp->ipp_fields |= IPPF_HOPLIMIT | IPPF_TCLASS | IPPF_ADDR;
5491 5489 ipp->ipp_hoplimit = ipha->ipha_ttl;
5492 5490 ipp->ipp_type_of_service = ipha->ipha_type_of_service;
5493 5491 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &ipp->ipp_addr);
5494 5492
5495 5493 /*
5496 5494 * Get length (in 4 byte octets) of IP header options.
5497 5495 */
5498 5496 totallen = ipha->ipha_version_and_hdr_length -
5499 5497 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
5500 5498
5501 5499 if (totallen == 0) {
5502 5500 if (!allocate)
5503 5501 return (0);
5504 5502
5505 5503 /* Clear out anything from a previous packet */
5506 5504 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
5507 5505 kmem_free(ipp->ipp_ipv4_options,
5508 5506 ipp->ipp_ipv4_options_len);
5509 5507 ipp->ipp_ipv4_options = NULL;
5510 5508 ipp->ipp_ipv4_options_len = 0;
5511 5509 ipp->ipp_fields &= ~IPPF_IPV4_OPTIONS;
5512 5510 }
5513 5511 if (ipp->ipp_fields & IPPF_LABEL_V4) {
5514 5512 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
5515 5513 ipp->ipp_label_v4 = NULL;
5516 5514 ipp->ipp_label_len_v4 = 0;
5517 5515 ipp->ipp_fields &= ~IPPF_LABEL_V4;
5518 5516 }
5519 5517 return (0);
5520 5518 }
5521 5519
5522 5520 totallen <<= 2;
5523 5521 opt = (uchar_t *)&ipha[1];
5524 5522 if (!is_system_labeled()) {
5525 5523
5526 5524 copyall:
5527 5525 if (!allocate) {
5528 5526 if (totallen != 0) {
5529 5527 ipp->ipp_ipv4_options = opt;
5530 5528 ipp->ipp_ipv4_options_len = totallen;
5531 5529 ipp->ipp_fields |= IPPF_IPV4_OPTIONS;
5532 5530 }
5533 5531 return (0);
5534 5532 }
5535 5533 /* Just copy all of options */
5536 5534 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
5537 5535 if (totallen == ipp->ipp_ipv4_options_len) {
5538 5536 bcopy(opt, ipp->ipp_ipv4_options, totallen);
5539 5537 return (0);
5540 5538 }
5541 5539 kmem_free(ipp->ipp_ipv4_options,
5542 5540 ipp->ipp_ipv4_options_len);
5543 5541 ipp->ipp_ipv4_options = NULL;
5544 5542 ipp->ipp_ipv4_options_len = 0;
5545 5543 ipp->ipp_fields &= ~IPPF_IPV4_OPTIONS;
5546 5544 }
5547 5545 if (totallen == 0)
5548 5546 return (0);
5549 5547
5550 5548 ipp->ipp_ipv4_options = kmem_alloc(totallen, KM_NOSLEEP);
5551 5549 if (ipp->ipp_ipv4_options == NULL)
5552 5550 return (ENOMEM);
5553 5551 ipp->ipp_ipv4_options_len = totallen;
5554 5552 ipp->ipp_fields |= IPPF_IPV4_OPTIONS;
5555 5553 bcopy(opt, ipp->ipp_ipv4_options, totallen);
5556 5554 return (0);
5557 5555 }
5558 5556
5559 5557 if (allocate && (ipp->ipp_fields & IPPF_LABEL_V4)) {
5560 5558 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
5561 5559 ipp->ipp_label_v4 = NULL;
5562 5560 ipp->ipp_label_len_v4 = 0;
5563 5561 ipp->ipp_fields &= ~IPPF_LABEL_V4;
5564 5562 }
5565 5563
5566 5564 /*
5567 5565 * Search for CIPSO option.
5568 5566 * We assume CIPSO is first in options if it is present.
5569 5567 * If it isn't, then ipp_opt_ipv4_options will not include the options
5570 5568 * prior to the CIPSO option.
5571 5569 */
5572 5570 while (totallen != 0) {
5573 5571 switch (optval = opt[IPOPT_OPTVAL]) {
5574 5572 case IPOPT_EOL:
5575 5573 return (0);
5576 5574 case IPOPT_NOP:
5577 5575 optlen = 1;
5578 5576 break;
5579 5577 default:
5580 5578 if (totallen <= IPOPT_OLEN)
5581 5579 return (EINVAL);
5582 5580 optlen = opt[IPOPT_OLEN];
5583 5581 if (optlen < 2)
5584 5582 return (EINVAL);
5585 5583 }
5586 5584 if (optlen > totallen)
5587 5585 return (EINVAL);
5588 5586
5589 5587 switch (optval) {
5590 5588 case IPOPT_COMSEC:
5591 5589 if (!allocate) {
5592 5590 ipp->ipp_label_v4 = opt;
5593 5591 ipp->ipp_label_len_v4 = optlen;
5594 5592 ipp->ipp_fields |= IPPF_LABEL_V4;
5595 5593 } else {
5596 5594 ipp->ipp_label_v4 = kmem_alloc(optlen,
5597 5595 KM_NOSLEEP);
5598 5596 if (ipp->ipp_label_v4 == NULL)
5599 5597 return (ENOMEM);
5600 5598 ipp->ipp_label_len_v4 = optlen;
5601 5599 ipp->ipp_fields |= IPPF_LABEL_V4;
5602 5600 bcopy(opt, ipp->ipp_label_v4, optlen);
5603 5601 }
5604 5602 totallen -= optlen;
5605 5603 opt += optlen;
5606 5604
5607 5605 /* Skip padding bytes until we get to a multiple of 4 */
5608 5606 while ((totallen & 3) != 0 && opt[0] == IPOPT_NOP) {
5609 5607 totallen--;
5610 5608 opt++;
5611 5609 }
5612 5610 /* Remaining as ipp_ipv4_options */
5613 5611 goto copyall;
5614 5612 }
5615 5613 totallen -= optlen;
5616 5614 opt += optlen;
5617 5615 }
5618 5616 /* No CIPSO found; return everything as ipp_ipv4_options */
5619 5617 totallen = ipha->ipha_version_and_hdr_length -
5620 5618 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
5621 5619 totallen <<= 2;
5622 5620 opt = (uchar_t *)&ipha[1];
5623 5621 goto copyall;
5624 5622 }
5625 5623
5626 5624 /*
5627 5625 * Efficient versions of lookup for an IRE when we only
5628 5626 * match the address.
5629 5627 * For RTF_REJECT or BLACKHOLE we return IRE_NOROUTE.
5630 5628 * Does not handle multicast addresses.
5631 5629 */
5632 5630 uint_t
5633 5631 ip_type_v4(ipaddr_t addr, ip_stack_t *ipst)
5634 5632 {
5635 5633 ire_t *ire;
5636 5634 uint_t result;
5637 5635
5638 5636 ire = ire_ftable_lookup_simple_v4(addr, 0, ipst, NULL);
5639 5637 ASSERT(ire != NULL);
5640 5638 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))
5641 5639 result = IRE_NOROUTE;
5642 5640 else
5643 5641 result = ire->ire_type;
5644 5642 ire_refrele(ire);
5645 5643 return (result);
5646 5644 }
5647 5645
5648 5646 /*
5649 5647 * Efficient versions of lookup for an IRE when we only
5650 5648 * match the address.
5651 5649 * For RTF_REJECT or BLACKHOLE we return IRE_NOROUTE.
5652 5650 * Does not handle multicast addresses.
5653 5651 */
5654 5652 uint_t
5655 5653 ip_type_v6(const in6_addr_t *addr, ip_stack_t *ipst)
5656 5654 {
5657 5655 ire_t *ire;
5658 5656 uint_t result;
5659 5657
5660 5658 ire = ire_ftable_lookup_simple_v6(addr, 0, ipst, NULL);
5661 5659 ASSERT(ire != NULL);
5662 5660 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))
5663 5661 result = IRE_NOROUTE;
5664 5662 else
5665 5663 result = ire->ire_type;
5666 5664 ire_refrele(ire);
5667 5665 return (result);
5668 5666 }
5669 5667
5670 5668 /*
5671 5669 * Nobody should be sending
5672 5670 * packets up this stream
5673 5671 */
5674 5672 static int
5675 5673 ip_lrput(queue_t *q, mblk_t *mp)
5676 5674 {
5677 5675 switch (mp->b_datap->db_type) {
5678 5676 case M_FLUSH:
5679 5677 /* Turn around */
5680 5678 if (*mp->b_rptr & FLUSHW) {
5681 5679 *mp->b_rptr &= ~FLUSHR;
5682 5680 qreply(q, mp);
5683 5681 return (0);
5684 5682 }
5685 5683 break;
5686 5684 }
5687 5685 freemsg(mp);
5688 5686 return (0);
5689 5687 }
5690 5688
5691 5689 /* Nobody should be sending packets down this stream */
5692 5690 /* ARGSUSED */
5693 5691 int
5694 5692 ip_lwput(queue_t *q, mblk_t *mp)
5695 5693 {
5696 5694 freemsg(mp);
5697 5695 return (0);
5698 5696 }
5699 5697
5700 5698 /*
5701 5699 * Move the first hop in any source route to ipha_dst and remove that part of
5702 5700 * the source route. Called by other protocols. Errors in option formatting
5703 5701 * are ignored - will be handled by ip_output_options. Return the final
5704 5702 * destination (either ipha_dst or the last entry in a source route.)
5705 5703 */
5706 5704 ipaddr_t
5707 5705 ip_massage_options(ipha_t *ipha, netstack_t *ns)
5708 5706 {
5709 5707 ipoptp_t opts;
5710 5708 uchar_t *opt;
5711 5709 uint8_t optval;
5712 5710 uint8_t optlen;
5713 5711 ipaddr_t dst;
5714 5712 int i;
5715 5713 ip_stack_t *ipst = ns->netstack_ip;
5716 5714
5717 5715 ip2dbg(("ip_massage_options\n"));
5718 5716 dst = ipha->ipha_dst;
5719 5717 for (optval = ipoptp_first(&opts, ipha);
5720 5718 optval != IPOPT_EOL;
5721 5719 optval = ipoptp_next(&opts)) {
5722 5720 opt = opts.ipoptp_cur;
5723 5721 switch (optval) {
5724 5722 uint8_t off;
5725 5723 case IPOPT_SSRR:
5726 5724 case IPOPT_LSRR:
5727 5725 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
5728 5726 ip1dbg(("ip_massage_options: bad src route\n"));
5729 5727 break;
5730 5728 }
5731 5729 optlen = opts.ipoptp_len;
5732 5730 off = opt[IPOPT_OFFSET];
5733 5731 off--;
5734 5732 redo_srr:
5735 5733 if (optlen < IP_ADDR_LEN ||
5736 5734 off > optlen - IP_ADDR_LEN) {
5737 5735 /* End of source route */
5738 5736 ip1dbg(("ip_massage_options: end of SR\n"));
5739 5737 break;
5740 5738 }
5741 5739 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
5742 5740 ip1dbg(("ip_massage_options: next hop 0x%x\n",
5743 5741 ntohl(dst)));
5744 5742 /*
5745 5743 * Check if our address is present more than
5746 5744 * once as consecutive hops in source route.
5747 5745 * XXX verify per-interface ip_forwarding
5748 5746 * for source route?
5749 5747 */
5750 5748 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
5751 5749 off += IP_ADDR_LEN;
5752 5750 goto redo_srr;
5753 5751 }
5754 5752 if (dst == htonl(INADDR_LOOPBACK)) {
5755 5753 ip1dbg(("ip_massage_options: loopback addr in "
5756 5754 "source route!\n"));
5757 5755 break;
5758 5756 }
5759 5757 /*
5760 5758 * Update ipha_dst to be the first hop and remove the
5761 5759 * first hop from the source route (by overwriting
5762 5760 * part of the option with NOP options).
5763 5761 */
5764 5762 ipha->ipha_dst = dst;
5765 5763 /* Put the last entry in dst */
5766 5764 off = ((optlen - IP_ADDR_LEN - 3) & ~(IP_ADDR_LEN-1)) +
5767 5765 3;
5768 5766 bcopy(&opt[off], &dst, IP_ADDR_LEN);
5769 5767
5770 5768 ip1dbg(("ip_massage_options: last hop 0x%x\n",
5771 5769 ntohl(dst)));
5772 5770 /* Move down and overwrite */
5773 5771 opt[IP_ADDR_LEN] = opt[0];
5774 5772 opt[IP_ADDR_LEN+1] = opt[IPOPT_OLEN] - IP_ADDR_LEN;
5775 5773 opt[IP_ADDR_LEN+2] = opt[IPOPT_OFFSET];
5776 5774 for (i = 0; i < IP_ADDR_LEN; i++)
5777 5775 opt[i] = IPOPT_NOP;
5778 5776 break;
5779 5777 }
5780 5778 }
5781 5779 return (dst);
5782 5780 }
5783 5781
5784 5782 /*
5785 5783 * Return the network mask
5786 5784 * associated with the specified address.
5787 5785 */
5788 5786 ipaddr_t
5789 5787 ip_net_mask(ipaddr_t addr)
5790 5788 {
5791 5789 uchar_t *up = (uchar_t *)&addr;
5792 5790 ipaddr_t mask = 0;
5793 5791 uchar_t *maskp = (uchar_t *)&mask;
5794 5792
5795 5793 #if defined(__i386) || defined(__amd64)
5796 5794 #define TOTALLY_BRAIN_DAMAGED_C_COMPILER
5797 5795 #endif
5798 5796 #ifdef TOTALLY_BRAIN_DAMAGED_C_COMPILER
5799 5797 maskp[0] = maskp[1] = maskp[2] = maskp[3] = 0;
5800 5798 #endif
5801 5799 if (CLASSD(addr)) {
5802 5800 maskp[0] = 0xF0;
5803 5801 return (mask);
5804 5802 }
5805 5803
5806 5804 /* We assume Class E default netmask to be 32 */
5807 5805 if (CLASSE(addr))
5808 5806 return (0xffffffffU);
5809 5807
5810 5808 if (addr == 0)
5811 5809 return (0);
5812 5810 maskp[0] = 0xFF;
5813 5811 if ((up[0] & 0x80) == 0)
5814 5812 return (mask);
5815 5813
5816 5814 maskp[1] = 0xFF;
5817 5815 if ((up[0] & 0xC0) == 0x80)
5818 5816 return (mask);
5819 5817
5820 5818 maskp[2] = 0xFF;
5821 5819 if ((up[0] & 0xE0) == 0xC0)
5822 5820 return (mask);
5823 5821
5824 5822 /* Otherwise return no mask */
5825 5823 return ((ipaddr_t)0);
5826 5824 }
5827 5825
5828 5826 /* Name/Value Table Lookup Routine */
5829 5827 char *
5830 5828 ip_nv_lookup(nv_t *nv, int value)
5831 5829 {
5832 5830 if (!nv)
5833 5831 return (NULL);
5834 5832 for (; nv->nv_name; nv++) {
5835 5833 if (nv->nv_value == value)
5836 5834 return (nv->nv_name);
5837 5835 }
5838 5836 return ("unknown");
5839 5837 }
5840 5838
5841 5839 static int
5842 5840 ip_wait_for_info_ack(ill_t *ill)
5843 5841 {
5844 5842 int err;
5845 5843
5846 5844 mutex_enter(&ill->ill_lock);
5847 5845 while (ill->ill_state_flags & ILL_LL_SUBNET_PENDING) {
5848 5846 /*
5849 5847 * Return value of 0 indicates a pending signal.
5850 5848 */
5851 5849 err = cv_wait_sig(&ill->ill_cv, &ill->ill_lock);
5852 5850 if (err == 0) {
5853 5851 mutex_exit(&ill->ill_lock);
5854 5852 return (EINTR);
5855 5853 }
5856 5854 }
5857 5855 mutex_exit(&ill->ill_lock);
5858 5856 /*
5859 5857 * ip_rput_other could have set an error in ill_error on
5860 5858 * receipt of M_ERROR.
5861 5859 */
5862 5860 return (ill->ill_error);
5863 5861 }
5864 5862
5865 5863 /*
5866 5864 * This is a module open, i.e. this is a control stream for access
5867 5865 * to a DLPI device. We allocate an ill_t as the instance data in
5868 5866 * this case.
5869 5867 */
5870 5868 static int
5871 5869 ip_modopen(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5872 5870 {
5873 5871 ill_t *ill;
5874 5872 int err;
5875 5873 zoneid_t zoneid;
5876 5874 netstack_t *ns;
5877 5875 ip_stack_t *ipst;
5878 5876
5879 5877 /*
5880 5878 * Prevent unprivileged processes from pushing IP so that
5881 5879 * they can't send raw IP.
5882 5880 */
5883 5881 if (secpolicy_net_rawaccess(credp) != 0)
5884 5882 return (EPERM);
5885 5883
5886 5884 ns = netstack_find_by_cred(credp);
5887 5885 ASSERT(ns != NULL);
5888 5886 ipst = ns->netstack_ip;
5889 5887 ASSERT(ipst != NULL);
5890 5888
5891 5889 /*
5892 5890 * For exclusive stacks we set the zoneid to zero
5893 5891 * to make IP operate as if in the global zone.
5894 5892 */
5895 5893 if (ipst->ips_netstack->netstack_stackid != GLOBAL_NETSTACKID)
5896 5894 zoneid = GLOBAL_ZONEID;
5897 5895 else
5898 5896 zoneid = crgetzoneid(credp);
5899 5897
5900 5898 ill = (ill_t *)mi_open_alloc_sleep(sizeof (ill_t));
5901 5899 q->q_ptr = WR(q)->q_ptr = ill;
5902 5900 ill->ill_ipst = ipst;
5903 5901 ill->ill_zoneid = zoneid;
5904 5902
5905 5903 /*
5906 5904 * ill_init initializes the ill fields and then sends down
5907 5905 * down a DL_INFO_REQ after calling qprocson.
5908 5906 */
5909 5907 err = ill_init(q, ill);
5910 5908
5911 5909 if (err != 0) {
5912 5910 mi_free(ill);
5913 5911 netstack_rele(ipst->ips_netstack);
5914 5912 q->q_ptr = NULL;
5915 5913 WR(q)->q_ptr = NULL;
5916 5914 return (err);
5917 5915 }
5918 5916
5919 5917 /*
5920 5918 * Wait for the DL_INFO_ACK if a DL_INFO_REQ was sent.
5921 5919 *
5922 5920 * ill_init initializes the ipsq marking this thread as
5923 5921 * writer
5924 5922 */
5925 5923 ipsq_exit(ill->ill_phyint->phyint_ipsq);
5926 5924 err = ip_wait_for_info_ack(ill);
5927 5925 if (err == 0)
5928 5926 ill->ill_credp = credp;
5929 5927 else
5930 5928 goto fail;
5931 5929
5932 5930 crhold(credp);
5933 5931
5934 5932 mutex_enter(&ipst->ips_ip_mi_lock);
5935 5933 err = mi_open_link(&ipst->ips_ip_g_head, (IDP)q->q_ptr, devp, flag,
5936 5934 sflag, credp);
5937 5935 mutex_exit(&ipst->ips_ip_mi_lock);
5938 5936 fail:
5939 5937 if (err) {
5940 5938 (void) ip_close(q, 0, credp);
5941 5939 return (err);
5942 5940 }
5943 5941 return (0);
5944 5942 }
5945 5943
5946 5944 /* For /dev/ip aka AF_INET open */
5947 5945 int
5948 5946 ip_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5949 5947 {
5950 5948 return (ip_open(q, devp, flag, sflag, credp, B_FALSE));
5951 5949 }
5952 5950
5953 5951 /* For /dev/ip6 aka AF_INET6 open */
5954 5952 int
5955 5953 ip_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5956 5954 {
5957 5955 return (ip_open(q, devp, flag, sflag, credp, B_TRUE));
5958 5956 }
5959 5957
5960 5958 /* IP open routine. */
5961 5959 int
5962 5960 ip_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp,
5963 5961 boolean_t isv6)
5964 5962 {
5965 5963 conn_t *connp;
5966 5964 major_t maj;
5967 5965 zoneid_t zoneid;
5968 5966 netstack_t *ns;
5969 5967 ip_stack_t *ipst;
5970 5968
5971 5969 /* Allow reopen. */
5972 5970 if (q->q_ptr != NULL)
5973 5971 return (0);
5974 5972
5975 5973 if (sflag & MODOPEN) {
5976 5974 /* This is a module open */
5977 5975 return (ip_modopen(q, devp, flag, sflag, credp));
5978 5976 }
5979 5977
5980 5978 if ((flag & ~(FKLYR)) == IP_HELPER_STR) {
5981 5979 /*
5982 5980 * Non streams based socket looking for a stream
5983 5981 * to access IP
5984 5982 */
5985 5983 return (ip_helper_stream_setup(q, devp, flag, sflag,
5986 5984 credp, isv6));
5987 5985 }
5988 5986
5989 5987 ns = netstack_find_by_cred(credp);
5990 5988 ASSERT(ns != NULL);
5991 5989 ipst = ns->netstack_ip;
5992 5990 ASSERT(ipst != NULL);
5993 5991
5994 5992 /*
5995 5993 * For exclusive stacks we set the zoneid to zero
5996 5994 * to make IP operate as if in the global zone.
5997 5995 */
5998 5996 if (ipst->ips_netstack->netstack_stackid != GLOBAL_NETSTACKID)
5999 5997 zoneid = GLOBAL_ZONEID;
6000 5998 else
6001 5999 zoneid = crgetzoneid(credp);
6002 6000
6003 6001 /*
6004 6002 * We are opening as a device. This is an IP client stream, and we
6005 6003 * allocate an conn_t as the instance data.
6006 6004 */
6007 6005 connp = ipcl_conn_create(IPCL_IPCCONN, KM_SLEEP, ipst->ips_netstack);
6008 6006
6009 6007 /*
6010 6008 * ipcl_conn_create did a netstack_hold. Undo the hold that was
6011 6009 * done by netstack_find_by_cred()
6012 6010 */
6013 6011 netstack_rele(ipst->ips_netstack);
6014 6012
6015 6013 connp->conn_ixa->ixa_flags |= IXAF_MULTICAST_LOOP | IXAF_SET_ULP_CKSUM;
6016 6014 /* conn_allzones can not be set this early, hence no IPCL_ZONEID */
6017 6015 connp->conn_ixa->ixa_zoneid = zoneid;
6018 6016 connp->conn_zoneid = zoneid;
6019 6017
6020 6018 connp->conn_rq = q;
6021 6019 q->q_ptr = WR(q)->q_ptr = connp;
6022 6020
6023 6021 /* Minor tells us which /dev entry was opened */
6024 6022 if (isv6) {
6025 6023 connp->conn_family = AF_INET6;
6026 6024 connp->conn_ipversion = IPV6_VERSION;
6027 6025 connp->conn_ixa->ixa_flags &= ~IXAF_IS_IPV4;
6028 6026 connp->conn_ixa->ixa_src_preferences = IPV6_PREFER_SRC_DEFAULT;
6029 6027 } else {
6030 6028 connp->conn_family = AF_INET;
6031 6029 connp->conn_ipversion = IPV4_VERSION;
6032 6030 connp->conn_ixa->ixa_flags |= IXAF_IS_IPV4;
6033 6031 }
6034 6032
6035 6033 if ((ip_minor_arena_la != NULL) && (flag & SO_SOCKSTR) &&
6036 6034 ((connp->conn_dev = inet_minor_alloc(ip_minor_arena_la)) != 0)) {
6037 6035 connp->conn_minor_arena = ip_minor_arena_la;
6038 6036 } else {
6039 6037 /*
6040 6038 * Either minor numbers in the large arena were exhausted
6041 6039 * or a non socket application is doing the open.
6042 6040 * Try to allocate from the small arena.
6043 6041 */
6044 6042 if ((connp->conn_dev =
6045 6043 inet_minor_alloc(ip_minor_arena_sa)) == 0) {
6046 6044 /* CONN_DEC_REF takes care of netstack_rele() */
6047 6045 q->q_ptr = WR(q)->q_ptr = NULL;
6048 6046 CONN_DEC_REF(connp);
6049 6047 return (EBUSY);
6050 6048 }
6051 6049 connp->conn_minor_arena = ip_minor_arena_sa;
6052 6050 }
6053 6051
6054 6052 maj = getemajor(*devp);
6055 6053 *devp = makedevice(maj, (minor_t)connp->conn_dev);
6056 6054
6057 6055 /*
6058 6056 * connp->conn_cred is crfree()ed in ipcl_conn_destroy()
6059 6057 */
6060 6058 connp->conn_cred = credp;
6061 6059 connp->conn_cpid = curproc->p_pid;
6062 6060 /* Cache things in ixa without an extra refhold */
6063 6061 ASSERT(!(connp->conn_ixa->ixa_free_flags & IXA_FREE_CRED));
6064 6062 connp->conn_ixa->ixa_cred = connp->conn_cred;
6065 6063 connp->conn_ixa->ixa_cpid = connp->conn_cpid;
6066 6064 if (is_system_labeled())
6067 6065 connp->conn_ixa->ixa_tsl = crgetlabel(connp->conn_cred);
6068 6066
6069 6067 /*
6070 6068 * Handle IP_IOC_RTS_REQUEST and other ioctls which use conn_recv
6071 6069 */
6072 6070 connp->conn_recv = ip_conn_input;
6073 6071 connp->conn_recvicmp = ip_conn_input_icmp;
6074 6072
6075 6073 crhold(connp->conn_cred);
6076 6074
6077 6075 /*
6078 6076 * If the caller has the process-wide flag set, then default to MAC
6079 6077 * exempt mode. This allows read-down to unlabeled hosts.
6080 6078 */
6081 6079 if (getpflags(NET_MAC_AWARE, credp) != 0)
6082 6080 connp->conn_mac_mode = CONN_MAC_AWARE;
6083 6081
6084 6082 connp->conn_zone_is_global = (crgetzoneid(credp) == GLOBAL_ZONEID);
6085 6083
6086 6084 connp->conn_rq = q;
6087 6085 connp->conn_wq = WR(q);
6088 6086
6089 6087 /* Non-zero default values */
6090 6088 connp->conn_ixa->ixa_flags |= IXAF_MULTICAST_LOOP;
6091 6089
6092 6090 /*
6093 6091 * Make the conn globally visible to walkers
6094 6092 */
6095 6093 ASSERT(connp->conn_ref == 1);
6096 6094 mutex_enter(&connp->conn_lock);
6097 6095 connp->conn_state_flags &= ~CONN_INCIPIENT;
6098 6096 mutex_exit(&connp->conn_lock);
6099 6097
6100 6098 qprocson(q);
6101 6099
6102 6100 return (0);
6103 6101 }
6104 6102
6105 6103 /*
6106 6104 * Set IPsec policy from an ipsec_req_t. If the req is not "zero" and valid,
6107 6105 * all of them are copied to the conn_t. If the req is "zero", the policy is
6108 6106 * zeroed out. A "zero" policy has zero ipsr_{ah,req,self_encap}_req
6109 6107 * fields.
6110 6108 * We keep only the latest setting of the policy and thus policy setting
6111 6109 * is not incremental/cumulative.
6112 6110 *
6113 6111 * Requests to set policies with multiple alternative actions will
6114 6112 * go through a different API.
6115 6113 */
6116 6114 int
6117 6115 ipsec_set_req(cred_t *cr, conn_t *connp, ipsec_req_t *req)
6118 6116 {
6119 6117 uint_t ah_req = 0;
6120 6118 uint_t esp_req = 0;
6121 6119 uint_t se_req = 0;
6122 6120 ipsec_act_t *actp = NULL;
6123 6121 uint_t nact;
6124 6122 ipsec_policy_head_t *ph;
6125 6123 boolean_t is_pol_reset, is_pol_inserted = B_FALSE;
6126 6124 int error = 0;
6127 6125 netstack_t *ns = connp->conn_netstack;
6128 6126 ip_stack_t *ipst = ns->netstack_ip;
6129 6127 ipsec_stack_t *ipss = ns->netstack_ipsec;
6130 6128
6131 6129 #define REQ_MASK (IPSEC_PREF_REQUIRED|IPSEC_PREF_NEVER)
6132 6130
6133 6131 /*
6134 6132 * The IP_SEC_OPT option does not allow variable length parameters,
6135 6133 * hence a request cannot be NULL.
6136 6134 */
6137 6135 if (req == NULL)
6138 6136 return (EINVAL);
6139 6137
6140 6138 ah_req = req->ipsr_ah_req;
6141 6139 esp_req = req->ipsr_esp_req;
6142 6140 se_req = req->ipsr_self_encap_req;
6143 6141
6144 6142 /* Don't allow setting self-encap without one or more of AH/ESP. */
6145 6143 if (se_req != 0 && esp_req == 0 && ah_req == 0)
6146 6144 return (EINVAL);
6147 6145
6148 6146 /*
6149 6147 * Are we dealing with a request to reset the policy (i.e.
6150 6148 * zero requests).
6151 6149 */
6152 6150 is_pol_reset = ((ah_req & REQ_MASK) == 0 &&
6153 6151 (esp_req & REQ_MASK) == 0 &&
6154 6152 (se_req & REQ_MASK) == 0);
6155 6153
6156 6154 if (!is_pol_reset) {
6157 6155 /*
6158 6156 * If we couldn't load IPsec, fail with "protocol
6159 6157 * not supported".
6160 6158 * IPsec may not have been loaded for a request with zero
6161 6159 * policies, so we don't fail in this case.
6162 6160 */
6163 6161 mutex_enter(&ipss->ipsec_loader_lock);
6164 6162 if (ipss->ipsec_loader_state != IPSEC_LOADER_SUCCEEDED) {
6165 6163 mutex_exit(&ipss->ipsec_loader_lock);
6166 6164 return (EPROTONOSUPPORT);
6167 6165 }
6168 6166 mutex_exit(&ipss->ipsec_loader_lock);
6169 6167
6170 6168 /*
6171 6169 * Test for valid requests. Invalid algorithms
6172 6170 * need to be tested by IPsec code because new
6173 6171 * algorithms can be added dynamically.
6174 6172 */
6175 6173 if ((ah_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0 ||
6176 6174 (esp_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0 ||
6177 6175 (se_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0) {
6178 6176 return (EINVAL);
6179 6177 }
6180 6178
6181 6179 /*
6182 6180 * Only privileged users can issue these
6183 6181 * requests.
6184 6182 */
6185 6183 if (((ah_req & IPSEC_PREF_NEVER) ||
6186 6184 (esp_req & IPSEC_PREF_NEVER) ||
6187 6185 (se_req & IPSEC_PREF_NEVER)) &&
6188 6186 secpolicy_ip_config(cr, B_FALSE) != 0) {
6189 6187 return (EPERM);
6190 6188 }
6191 6189
6192 6190 /*
6193 6191 * The IPSEC_PREF_REQUIRED and IPSEC_PREF_NEVER
6194 6192 * are mutually exclusive.
6195 6193 */
6196 6194 if (((ah_req & REQ_MASK) == REQ_MASK) ||
6197 6195 ((esp_req & REQ_MASK) == REQ_MASK) ||
6198 6196 ((se_req & REQ_MASK) == REQ_MASK)) {
6199 6197 /* Both of them are set */
6200 6198 return (EINVAL);
6201 6199 }
6202 6200 }
6203 6201
6204 6202 ASSERT(MUTEX_HELD(&connp->conn_lock));
6205 6203
6206 6204 /*
6207 6205 * If we have already cached policies in conn_connect(), don't
6208 6206 * let them change now. We cache policies for connections
6209 6207 * whose src,dst [addr, port] is known.
6210 6208 */
6211 6209 if (connp->conn_policy_cached) {
6212 6210 return (EINVAL);
6213 6211 }
6214 6212
6215 6213 /*
6216 6214 * We have a zero policies, reset the connection policy if already
6217 6215 * set. This will cause the connection to inherit the
6218 6216 * global policy, if any.
6219 6217 */
6220 6218 if (is_pol_reset) {
6221 6219 if (connp->conn_policy != NULL) {
6222 6220 IPPH_REFRELE(connp->conn_policy, ipst->ips_netstack);
6223 6221 connp->conn_policy = NULL;
6224 6222 }
6225 6223 connp->conn_in_enforce_policy = B_FALSE;
6226 6224 connp->conn_out_enforce_policy = B_FALSE;
6227 6225 return (0);
6228 6226 }
6229 6227
6230 6228 ph = connp->conn_policy = ipsec_polhead_split(connp->conn_policy,
6231 6229 ipst->ips_netstack);
6232 6230 if (ph == NULL)
6233 6231 goto enomem;
6234 6232
6235 6233 ipsec_actvec_from_req(req, &actp, &nact, ipst->ips_netstack);
6236 6234 if (actp == NULL)
6237 6235 goto enomem;
6238 6236
6239 6237 /*
6240 6238 * Always insert IPv4 policy entries, since they can also apply to
6241 6239 * ipv6 sockets being used in ipv4-compat mode.
6242 6240 */
6243 6241 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V4,
6244 6242 IPSEC_TYPE_INBOUND, ns))
6245 6243 goto enomem;
6246 6244 is_pol_inserted = B_TRUE;
6247 6245 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V4,
6248 6246 IPSEC_TYPE_OUTBOUND, ns))
6249 6247 goto enomem;
6250 6248
6251 6249 /*
6252 6250 * We're looking at a v6 socket, also insert the v6-specific
6253 6251 * entries.
6254 6252 */
6255 6253 if (connp->conn_family == AF_INET6) {
6256 6254 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V6,
6257 6255 IPSEC_TYPE_INBOUND, ns))
6258 6256 goto enomem;
6259 6257 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V6,
6260 6258 IPSEC_TYPE_OUTBOUND, ns))
6261 6259 goto enomem;
6262 6260 }
6263 6261
6264 6262 ipsec_actvec_free(actp, nact);
6265 6263
6266 6264 /*
6267 6265 * If the requests need security, set enforce_policy.
6268 6266 * If the requests are IPSEC_PREF_NEVER, one should
6269 6267 * still set conn_out_enforce_policy so that ip_set_destination
6270 6268 * marks the ip_xmit_attr_t appropriatly. This is needed so that
6271 6269 * for connections that we don't cache policy in at connect time,
6272 6270 * if global policy matches in ip_output_attach_policy, we
6273 6271 * don't wrongly inherit global policy. Similarly, we need
6274 6272 * to set conn_in_enforce_policy also so that we don't verify
6275 6273 * policy wrongly.
6276 6274 */
6277 6275 if ((ah_req & REQ_MASK) != 0 ||
6278 6276 (esp_req & REQ_MASK) != 0 ||
6279 6277 (se_req & REQ_MASK) != 0) {
6280 6278 connp->conn_in_enforce_policy = B_TRUE;
6281 6279 connp->conn_out_enforce_policy = B_TRUE;
6282 6280 }
6283 6281
6284 6282 return (error);
6285 6283 #undef REQ_MASK
6286 6284
6287 6285 /*
6288 6286 * Common memory-allocation-failure exit path.
6289 6287 */
6290 6288 enomem:
6291 6289 if (actp != NULL)
6292 6290 ipsec_actvec_free(actp, nact);
6293 6291 if (is_pol_inserted)
6294 6292 ipsec_polhead_flush(ph, ns);
6295 6293 return (ENOMEM);
6296 6294 }
6297 6295
6298 6296 /*
6299 6297 * Set socket options for joining and leaving multicast groups.
6300 6298 * Common to IPv4 and IPv6; inet6 indicates the type of socket.
6301 6299 * The caller has already check that the option name is consistent with
6302 6300 * the address family of the socket.
6303 6301 */
6304 6302 int
6305 6303 ip_opt_set_multicast_group(conn_t *connp, t_scalar_t name,
6306 6304 uchar_t *invalp, boolean_t inet6, boolean_t checkonly)
6307 6305 {
6308 6306 int *i1 = (int *)invalp;
6309 6307 int error = 0;
6310 6308 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
6311 6309 struct ip_mreq *v4_mreqp;
6312 6310 struct ipv6_mreq *v6_mreqp;
6313 6311 struct group_req *greqp;
6314 6312 ire_t *ire;
6315 6313 boolean_t done = B_FALSE;
6316 6314 ipaddr_t ifaddr;
6317 6315 in6_addr_t v6group;
6318 6316 uint_t ifindex;
6319 6317 boolean_t mcast_opt = B_TRUE;
6320 6318 mcast_record_t fmode;
6321 6319 int (*optfn)(conn_t *, boolean_t, const in6_addr_t *,
6322 6320 ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *);
6323 6321
6324 6322 switch (name) {
6325 6323 case IP_ADD_MEMBERSHIP:
6326 6324 case IPV6_JOIN_GROUP:
6327 6325 mcast_opt = B_FALSE;
6328 6326 /* FALLTHROUGH */
6329 6327 case MCAST_JOIN_GROUP:
6330 6328 fmode = MODE_IS_EXCLUDE;
6331 6329 optfn = ip_opt_add_group;
6332 6330 break;
6333 6331
6334 6332 case IP_DROP_MEMBERSHIP:
6335 6333 case IPV6_LEAVE_GROUP:
6336 6334 mcast_opt = B_FALSE;
6337 6335 /* FALLTHROUGH */
6338 6336 case MCAST_LEAVE_GROUP:
6339 6337 fmode = MODE_IS_INCLUDE;
6340 6338 optfn = ip_opt_delete_group;
6341 6339 break;
6342 6340 default:
6343 6341 ASSERT(0);
6344 6342 }
6345 6343
6346 6344 if (mcast_opt) {
6347 6345 struct sockaddr_in *sin;
6348 6346 struct sockaddr_in6 *sin6;
6349 6347
6350 6348 greqp = (struct group_req *)i1;
6351 6349 if (greqp->gr_group.ss_family == AF_INET) {
6352 6350 sin = (struct sockaddr_in *)&(greqp->gr_group);
6353 6351 IN6_INADDR_TO_V4MAPPED(&sin->sin_addr, &v6group);
6354 6352 } else {
6355 6353 if (!inet6)
6356 6354 return (EINVAL); /* Not on INET socket */
6357 6355
6358 6356 sin6 = (struct sockaddr_in6 *)&(greqp->gr_group);
6359 6357 v6group = sin6->sin6_addr;
6360 6358 }
6361 6359 ifaddr = INADDR_ANY;
6362 6360 ifindex = greqp->gr_interface;
6363 6361 } else if (inet6) {
6364 6362 v6_mreqp = (struct ipv6_mreq *)i1;
6365 6363 v6group = v6_mreqp->ipv6mr_multiaddr;
6366 6364 ifaddr = INADDR_ANY;
6367 6365 ifindex = v6_mreqp->ipv6mr_interface;
6368 6366 } else {
6369 6367 v4_mreqp = (struct ip_mreq *)i1;
6370 6368 IN6_INADDR_TO_V4MAPPED(&v4_mreqp->imr_multiaddr, &v6group);
6371 6369 ifaddr = (ipaddr_t)v4_mreqp->imr_interface.s_addr;
6372 6370 ifindex = 0;
6373 6371 }
6374 6372
6375 6373 /*
6376 6374 * In the multirouting case, we need to replicate
6377 6375 * the request on all interfaces that will take part
6378 6376 * in replication. We do so because multirouting is
6379 6377 * reflective, thus we will probably receive multi-
6380 6378 * casts on those interfaces.
6381 6379 * The ip_multirt_apply_membership() succeeds if
6382 6380 * the operation succeeds on at least one interface.
6383 6381 */
6384 6382 if (IN6_IS_ADDR_V4MAPPED(&v6group)) {
6385 6383 ipaddr_t group;
6386 6384
6387 6385 IN6_V4MAPPED_TO_IPADDR(&v6group, group);
6388 6386
6389 6387 ire = ire_ftable_lookup_v4(group, IP_HOST_MASK, 0,
6390 6388 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6391 6389 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6392 6390 } else {
6393 6391 ire = ire_ftable_lookup_v6(&v6group, &ipv6_all_ones, 0,
6394 6392 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6395 6393 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6396 6394 }
6397 6395 if (ire != NULL) {
6398 6396 if (ire->ire_flags & RTF_MULTIRT) {
6399 6397 error = ip_multirt_apply_membership(optfn, ire, connp,
6400 6398 checkonly, &v6group, fmode, &ipv6_all_zeros);
6401 6399 done = B_TRUE;
6402 6400 }
6403 6401 ire_refrele(ire);
6404 6402 }
6405 6403
6406 6404 if (!done) {
6407 6405 error = optfn(connp, checkonly, &v6group, ifaddr, ifindex,
6408 6406 fmode, &ipv6_all_zeros);
6409 6407 }
6410 6408 return (error);
6411 6409 }
6412 6410
6413 6411 /*
6414 6412 * Set socket options for joining and leaving multicast groups
6415 6413 * for specific sources.
6416 6414 * Common to IPv4 and IPv6; inet6 indicates the type of socket.
6417 6415 * The caller has already check that the option name is consistent with
6418 6416 * the address family of the socket.
6419 6417 */
6420 6418 int
6421 6419 ip_opt_set_multicast_sources(conn_t *connp, t_scalar_t name,
6422 6420 uchar_t *invalp, boolean_t inet6, boolean_t checkonly)
6423 6421 {
6424 6422 int *i1 = (int *)invalp;
6425 6423 int error = 0;
6426 6424 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
6427 6425 struct ip_mreq_source *imreqp;
6428 6426 struct group_source_req *gsreqp;
6429 6427 in6_addr_t v6group, v6src;
6430 6428 uint32_t ifindex;
6431 6429 ipaddr_t ifaddr;
6432 6430 boolean_t mcast_opt = B_TRUE;
6433 6431 mcast_record_t fmode;
6434 6432 ire_t *ire;
6435 6433 boolean_t done = B_FALSE;
6436 6434 int (*optfn)(conn_t *, boolean_t, const in6_addr_t *,
6437 6435 ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *);
6438 6436
6439 6437 switch (name) {
6440 6438 case IP_BLOCK_SOURCE:
6441 6439 mcast_opt = B_FALSE;
6442 6440 /* FALLTHROUGH */
6443 6441 case MCAST_BLOCK_SOURCE:
6444 6442 fmode = MODE_IS_EXCLUDE;
6445 6443 optfn = ip_opt_add_group;
6446 6444 break;
6447 6445
6448 6446 case IP_UNBLOCK_SOURCE:
6449 6447 mcast_opt = B_FALSE;
6450 6448 /* FALLTHROUGH */
6451 6449 case MCAST_UNBLOCK_SOURCE:
6452 6450 fmode = MODE_IS_EXCLUDE;
6453 6451 optfn = ip_opt_delete_group;
6454 6452 break;
6455 6453
6456 6454 case IP_ADD_SOURCE_MEMBERSHIP:
6457 6455 mcast_opt = B_FALSE;
6458 6456 /* FALLTHROUGH */
6459 6457 case MCAST_JOIN_SOURCE_GROUP:
6460 6458 fmode = MODE_IS_INCLUDE;
6461 6459 optfn = ip_opt_add_group;
6462 6460 break;
6463 6461
6464 6462 case IP_DROP_SOURCE_MEMBERSHIP:
6465 6463 mcast_opt = B_FALSE;
6466 6464 /* FALLTHROUGH */
6467 6465 case MCAST_LEAVE_SOURCE_GROUP:
6468 6466 fmode = MODE_IS_INCLUDE;
6469 6467 optfn = ip_opt_delete_group;
6470 6468 break;
6471 6469 default:
6472 6470 ASSERT(0);
6473 6471 }
6474 6472
6475 6473 if (mcast_opt) {
6476 6474 gsreqp = (struct group_source_req *)i1;
6477 6475 ifindex = gsreqp->gsr_interface;
6478 6476 if (gsreqp->gsr_group.ss_family == AF_INET) {
6479 6477 struct sockaddr_in *s;
6480 6478 s = (struct sockaddr_in *)&gsreqp->gsr_group;
6481 6479 IN6_INADDR_TO_V4MAPPED(&s->sin_addr, &v6group);
6482 6480 s = (struct sockaddr_in *)&gsreqp->gsr_source;
6483 6481 IN6_INADDR_TO_V4MAPPED(&s->sin_addr, &v6src);
6484 6482 } else {
6485 6483 struct sockaddr_in6 *s6;
6486 6484
6487 6485 if (!inet6)
6488 6486 return (EINVAL); /* Not on INET socket */
6489 6487
6490 6488 s6 = (struct sockaddr_in6 *)&gsreqp->gsr_group;
6491 6489 v6group = s6->sin6_addr;
6492 6490 s6 = (struct sockaddr_in6 *)&gsreqp->gsr_source;
6493 6491 v6src = s6->sin6_addr;
6494 6492 }
6495 6493 ifaddr = INADDR_ANY;
6496 6494 } else {
6497 6495 imreqp = (struct ip_mreq_source *)i1;
6498 6496 IN6_INADDR_TO_V4MAPPED(&imreqp->imr_multiaddr, &v6group);
6499 6497 IN6_INADDR_TO_V4MAPPED(&imreqp->imr_sourceaddr, &v6src);
6500 6498 ifaddr = (ipaddr_t)imreqp->imr_interface.s_addr;
6501 6499 ifindex = 0;
6502 6500 }
6503 6501
6504 6502 /*
6505 6503 * Handle src being mapped INADDR_ANY by changing it to unspecified.
6506 6504 */
6507 6505 if (IN6_IS_ADDR_V4MAPPED_ANY(&v6src))
6508 6506 v6src = ipv6_all_zeros;
6509 6507
6510 6508 /*
6511 6509 * In the multirouting case, we need to replicate
6512 6510 * the request as noted in the mcast cases above.
6513 6511 */
6514 6512 if (IN6_IS_ADDR_V4MAPPED(&v6group)) {
6515 6513 ipaddr_t group;
6516 6514
6517 6515 IN6_V4MAPPED_TO_IPADDR(&v6group, group);
6518 6516
6519 6517 ire = ire_ftable_lookup_v4(group, IP_HOST_MASK, 0,
6520 6518 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6521 6519 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6522 6520 } else {
6523 6521 ire = ire_ftable_lookup_v6(&v6group, &ipv6_all_ones, 0,
6524 6522 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6525 6523 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6526 6524 }
6527 6525 if (ire != NULL) {
6528 6526 if (ire->ire_flags & RTF_MULTIRT) {
6529 6527 error = ip_multirt_apply_membership(optfn, ire, connp,
6530 6528 checkonly, &v6group, fmode, &v6src);
6531 6529 done = B_TRUE;
6532 6530 }
6533 6531 ire_refrele(ire);
6534 6532 }
6535 6533 if (!done) {
6536 6534 error = optfn(connp, checkonly, &v6group, ifaddr, ifindex,
6537 6535 fmode, &v6src);
6538 6536 }
6539 6537 return (error);
6540 6538 }
6541 6539
6542 6540 /*
6543 6541 * Given a destination address and a pointer to where to put the information
6544 6542 * this routine fills in the mtuinfo.
6545 6543 * The socket must be connected.
6546 6544 * For sctp conn_faddr is the primary address.
6547 6545 */
6548 6546 int
6549 6547 ip_fill_mtuinfo(conn_t *connp, ip_xmit_attr_t *ixa, struct ip6_mtuinfo *mtuinfo)
6550 6548 {
6551 6549 uint32_t pmtu = IP_MAXPACKET;
6552 6550 uint_t scopeid;
6553 6551
6554 6552 if (IN6_IS_ADDR_UNSPECIFIED(&connp->conn_faddr_v6))
6555 6553 return (-1);
6556 6554
6557 6555 /* In case we never sent or called ip_set_destination_v4/v6 */
6558 6556 if (ixa->ixa_ire != NULL)
6559 6557 pmtu = ip_get_pmtu(ixa);
6560 6558
6561 6559 if (ixa->ixa_flags & IXAF_SCOPEID_SET)
6562 6560 scopeid = ixa->ixa_scopeid;
6563 6561 else
6564 6562 scopeid = 0;
6565 6563
6566 6564 bzero(mtuinfo, sizeof (*mtuinfo));
6567 6565 mtuinfo->ip6m_addr.sin6_family = AF_INET6;
6568 6566 mtuinfo->ip6m_addr.sin6_port = connp->conn_fport;
6569 6567 mtuinfo->ip6m_addr.sin6_addr = connp->conn_faddr_v6;
6570 6568 mtuinfo->ip6m_addr.sin6_scope_id = scopeid;
6571 6569 mtuinfo->ip6m_mtu = pmtu;
6572 6570
6573 6571 return (sizeof (struct ip6_mtuinfo));
6574 6572 }
6575 6573
6576 6574 /*
6577 6575 * When the src multihoming is changed from weak to [strong, preferred]
6578 6576 * ip_ire_rebind_walker is called to walk the list of all ire_t entries
6579 6577 * and identify routes that were created by user-applications in the
6580 6578 * unbound state (i.e., without RTA_IFP), and for which an ire_ill is not
6581 6579 * currently defined. These routes are then 'rebound', i.e., their ire_ill
6582 6580 * is selected by finding an interface route for the gateway.
6583 6581 */
6584 6582 /* ARGSUSED */
6585 6583 void
6586 6584 ip_ire_rebind_walker(ire_t *ire, void *notused)
6587 6585 {
6588 6586 if (!ire->ire_unbound || ire->ire_ill != NULL)
6589 6587 return;
6590 6588 ire_rebind(ire);
6591 6589 ire_delete(ire);
6592 6590 }
6593 6591
6594 6592 /*
6595 6593 * When the src multihoming is changed from [strong, preferred] to weak,
6596 6594 * ip_ire_unbind_walker is called to walk the list of all ire_t entries, and
6597 6595 * set any entries that were created by user-applications in the unbound state
6598 6596 * (i.e., without RTA_IFP) back to having a NULL ire_ill.
6599 6597 */
6600 6598 /* ARGSUSED */
6601 6599 void
6602 6600 ip_ire_unbind_walker(ire_t *ire, void *notused)
6603 6601 {
6604 6602 ire_t *new_ire;
6605 6603
6606 6604 if (!ire->ire_unbound || ire->ire_ill == NULL)
6607 6605 return;
6608 6606 if (ire->ire_ipversion == IPV6_VERSION) {
6609 6607 new_ire = ire_create_v6(&ire->ire_addr_v6, &ire->ire_mask_v6,
6610 6608 &ire->ire_gateway_addr_v6, ire->ire_type, NULL,
6611 6609 ire->ire_zoneid, ire->ire_flags, NULL, ire->ire_ipst);
6612 6610 } else {
6613 6611 new_ire = ire_create((uchar_t *)&ire->ire_addr,
6614 6612 (uchar_t *)&ire->ire_mask,
6615 6613 (uchar_t *)&ire->ire_gateway_addr, ire->ire_type, NULL,
6616 6614 ire->ire_zoneid, ire->ire_flags, NULL, ire->ire_ipst);
6617 6615 }
6618 6616 if (new_ire == NULL)
6619 6617 return;
6620 6618 new_ire->ire_unbound = B_TRUE;
6621 6619 /*
6622 6620 * The bound ire must first be deleted so that we don't return
6623 6621 * the existing one on the attempt to add the unbound new_ire.
6624 6622 */
6625 6623 ire_delete(ire);
6626 6624 new_ire = ire_add(new_ire);
6627 6625 if (new_ire != NULL)
6628 6626 ire_refrele(new_ire);
6629 6627 }
6630 6628
6631 6629 /*
6632 6630 * When the settings of ip*_strict_src_multihoming tunables are changed,
6633 6631 * all cached routes need to be recomputed. This recomputation needs to be
6634 6632 * done when going from weaker to stronger modes so that the cached ire
6635 6633 * for the connection does not violate the current ip*_strict_src_multihoming
6636 6634 * setting. It also needs to be done when going from stronger to weaker modes,
6637 6635 * so that we fall back to matching on the longest-matching-route (as opposed
6638 6636 * to a shorter match that may have been selected in the strong mode
6639 6637 * to satisfy src_multihoming settings).
6640 6638 *
6641 6639 * The cached ixa_ire entires for all conn_t entries are marked as
6642 6640 * "verify" so that they will be recomputed for the next packet.
6643 6641 */
6644 6642 void
6645 6643 conn_ire_revalidate(conn_t *connp, void *arg)
6646 6644 {
6647 6645 boolean_t isv6 = (boolean_t)arg;
6648 6646
6649 6647 if ((isv6 && connp->conn_ipversion != IPV6_VERSION) ||
6650 6648 (!isv6 && connp->conn_ipversion != IPV4_VERSION))
6651 6649 return;
6652 6650 connp->conn_ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
6653 6651 }
6654 6652
6655 6653 /*
6656 6654 * Handles both IPv4 and IPv6 reassembly - doing the out-of-order cases,
6657 6655 * When an ipf is passed here for the first time, if
6658 6656 * we already have in-order fragments on the queue, we convert from the fast-
6659 6657 * path reassembly scheme to the hard-case scheme. From then on, additional
6660 6658 * fragments are reassembled here. We keep track of the start and end offsets
6661 6659 * of each piece, and the number of holes in the chain. When the hole count
6662 6660 * goes to zero, we are done!
6663 6661 *
6664 6662 * The ipf_count will be updated to account for any mblk(s) added (pointed to
6665 6663 * by mp) or subtracted (freeb()ed dups), upon return the caller must update
6666 6664 * ipfb_count and ill_frag_count by the difference of ipf_count before and
6667 6665 * after the call to ip_reassemble().
6668 6666 */
6669 6667 int
6670 6668 ip_reassemble(mblk_t *mp, ipf_t *ipf, uint_t start, boolean_t more, ill_t *ill,
6671 6669 size_t msg_len)
6672 6670 {
6673 6671 uint_t end;
6674 6672 mblk_t *next_mp;
6675 6673 mblk_t *mp1;
6676 6674 uint_t offset;
6677 6675 boolean_t incr_dups = B_TRUE;
6678 6676 boolean_t offset_zero_seen = B_FALSE;
6679 6677 boolean_t pkt_boundary_checked = B_FALSE;
6680 6678
6681 6679 /* If start == 0 then ipf_nf_hdr_len has to be set. */
6682 6680 ASSERT(start != 0 || ipf->ipf_nf_hdr_len != 0);
6683 6681
6684 6682 /* Add in byte count */
6685 6683 ipf->ipf_count += msg_len;
6686 6684 if (ipf->ipf_end) {
6687 6685 /*
6688 6686 * We were part way through in-order reassembly, but now there
6689 6687 * is a hole. We walk through messages already queued, and
6690 6688 * mark them for hard case reassembly. We know that up till
6691 6689 * now they were in order starting from offset zero.
6692 6690 */
6693 6691 offset = 0;
6694 6692 for (mp1 = ipf->ipf_mp->b_cont; mp1; mp1 = mp1->b_cont) {
6695 6693 IP_REASS_SET_START(mp1, offset);
6696 6694 if (offset == 0) {
6697 6695 ASSERT(ipf->ipf_nf_hdr_len != 0);
6698 6696 offset = -ipf->ipf_nf_hdr_len;
6699 6697 }
6700 6698 offset += mp1->b_wptr - mp1->b_rptr;
6701 6699 IP_REASS_SET_END(mp1, offset);
6702 6700 }
6703 6701 /* One hole at the end. */
6704 6702 ipf->ipf_hole_cnt = 1;
6705 6703 /* Brand it as a hard case, forever. */
6706 6704 ipf->ipf_end = 0;
6707 6705 }
6708 6706 /* Walk through all the new pieces. */
6709 6707 do {
6710 6708 end = start + (mp->b_wptr - mp->b_rptr);
6711 6709 /*
6712 6710 * If start is 0, decrease 'end' only for the first mblk of
6713 6711 * the fragment. Otherwise 'end' can get wrong value in the
6714 6712 * second pass of the loop if first mblk is exactly the
6715 6713 * size of ipf_nf_hdr_len.
6716 6714 */
6717 6715 if (start == 0 && !offset_zero_seen) {
6718 6716 /* First segment */
6719 6717 ASSERT(ipf->ipf_nf_hdr_len != 0);
6720 6718 end -= ipf->ipf_nf_hdr_len;
6721 6719 offset_zero_seen = B_TRUE;
6722 6720 }
6723 6721 next_mp = mp->b_cont;
6724 6722 /*
6725 6723 * We are checking to see if there is any interesing data
6726 6724 * to process. If there isn't and the mblk isn't the
6727 6725 * one which carries the unfragmentable header then we
6728 6726 * drop it. It's possible to have just the unfragmentable
6729 6727 * header come through without any data. That needs to be
6730 6728 * saved.
6731 6729 *
6732 6730 * If the assert at the top of this function holds then the
6733 6731 * term "ipf->ipf_nf_hdr_len != 0" isn't needed. This code
6734 6732 * is infrequently traveled enough that the test is left in
6735 6733 * to protect against future code changes which break that
6736 6734 * invariant.
6737 6735 */
6738 6736 if (start == end && start != 0 && ipf->ipf_nf_hdr_len != 0) {
6739 6737 /* Empty. Blast it. */
6740 6738 IP_REASS_SET_START(mp, 0);
6741 6739 IP_REASS_SET_END(mp, 0);
6742 6740 /*
6743 6741 * If the ipf points to the mblk we are about to free,
6744 6742 * update ipf to point to the next mblk (or NULL
6745 6743 * if none).
6746 6744 */
6747 6745 if (ipf->ipf_mp->b_cont == mp)
6748 6746 ipf->ipf_mp->b_cont = next_mp;
6749 6747 freeb(mp);
6750 6748 continue;
6751 6749 }
6752 6750 mp->b_cont = NULL;
6753 6751 IP_REASS_SET_START(mp, start);
6754 6752 IP_REASS_SET_END(mp, end);
6755 6753 if (!ipf->ipf_tail_mp) {
6756 6754 ipf->ipf_tail_mp = mp;
6757 6755 ipf->ipf_mp->b_cont = mp;
6758 6756 if (start == 0 || !more) {
6759 6757 ipf->ipf_hole_cnt = 1;
6760 6758 /*
6761 6759 * if the first fragment comes in more than one
6762 6760 * mblk, this loop will be executed for each
6763 6761 * mblk. Need to adjust hole count so exiting
6764 6762 * this routine will leave hole count at 1.
6765 6763 */
6766 6764 if (next_mp)
6767 6765 ipf->ipf_hole_cnt++;
6768 6766 } else
6769 6767 ipf->ipf_hole_cnt = 2;
6770 6768 continue;
6771 6769 } else if (ipf->ipf_last_frag_seen && !more &&
6772 6770 !pkt_boundary_checked) {
6773 6771 /*
6774 6772 * We check datagram boundary only if this fragment
6775 6773 * claims to be the last fragment and we have seen a
6776 6774 * last fragment in the past too. We do this only
6777 6775 * once for a given fragment.
6778 6776 *
6779 6777 * start cannot be 0 here as fragments with start=0
6780 6778 * and MF=0 gets handled as a complete packet. These
6781 6779 * fragments should not reach here.
6782 6780 */
6783 6781
6784 6782 if (start + msgdsize(mp) !=
6785 6783 IP_REASS_END(ipf->ipf_tail_mp)) {
6786 6784 /*
6787 6785 * We have two fragments both of which claim
6788 6786 * to be the last fragment but gives conflicting
6789 6787 * information about the whole datagram size.
6790 6788 * Something fishy is going on. Drop the
6791 6789 * fragment and free up the reassembly list.
6792 6790 */
6793 6791 return (IP_REASS_FAILED);
6794 6792 }
6795 6793
6796 6794 /*
6797 6795 * We shouldn't come to this code block again for this
6798 6796 * particular fragment.
6799 6797 */
6800 6798 pkt_boundary_checked = B_TRUE;
6801 6799 }
6802 6800
6803 6801 /* New stuff at or beyond tail? */
6804 6802 offset = IP_REASS_END(ipf->ipf_tail_mp);
6805 6803 if (start >= offset) {
6806 6804 if (ipf->ipf_last_frag_seen) {
6807 6805 /* current fragment is beyond last fragment */
6808 6806 return (IP_REASS_FAILED);
6809 6807 }
6810 6808 /* Link it on end. */
6811 6809 ipf->ipf_tail_mp->b_cont = mp;
6812 6810 ipf->ipf_tail_mp = mp;
6813 6811 if (more) {
6814 6812 if (start != offset)
6815 6813 ipf->ipf_hole_cnt++;
6816 6814 } else if (start == offset && next_mp == NULL)
6817 6815 ipf->ipf_hole_cnt--;
6818 6816 continue;
6819 6817 }
6820 6818 mp1 = ipf->ipf_mp->b_cont;
6821 6819 offset = IP_REASS_START(mp1);
6822 6820 /* New stuff at the front? */
6823 6821 if (start < offset) {
6824 6822 if (start == 0) {
6825 6823 if (end >= offset) {
6826 6824 /* Nailed the hole at the begining. */
6827 6825 ipf->ipf_hole_cnt--;
6828 6826 }
6829 6827 } else if (end < offset) {
6830 6828 /*
6831 6829 * A hole, stuff, and a hole where there used
6832 6830 * to be just a hole.
6833 6831 */
6834 6832 ipf->ipf_hole_cnt++;
6835 6833 }
6836 6834 mp->b_cont = mp1;
6837 6835 /* Check for overlap. */
6838 6836 while (end > offset) {
6839 6837 if (end < IP_REASS_END(mp1)) {
6840 6838 mp->b_wptr -= end - offset;
6841 6839 IP_REASS_SET_END(mp, offset);
6842 6840 BUMP_MIB(ill->ill_ip_mib,
6843 6841 ipIfStatsReasmPartDups);
6844 6842 break;
6845 6843 }
6846 6844 /* Did we cover another hole? */
6847 6845 if ((mp1->b_cont &&
6848 6846 IP_REASS_END(mp1) !=
6849 6847 IP_REASS_START(mp1->b_cont) &&
6850 6848 end >= IP_REASS_START(mp1->b_cont)) ||
6851 6849 (!ipf->ipf_last_frag_seen && !more)) {
6852 6850 ipf->ipf_hole_cnt--;
6853 6851 }
6854 6852 /* Clip out mp1. */
6855 6853 if ((mp->b_cont = mp1->b_cont) == NULL) {
6856 6854 /*
6857 6855 * After clipping out mp1, this guy
6858 6856 * is now hanging off the end.
6859 6857 */
6860 6858 ipf->ipf_tail_mp = mp;
6861 6859 }
6862 6860 IP_REASS_SET_START(mp1, 0);
6863 6861 IP_REASS_SET_END(mp1, 0);
6864 6862 /* Subtract byte count */
6865 6863 ipf->ipf_count -= mp1->b_datap->db_lim -
6866 6864 mp1->b_datap->db_base;
6867 6865 freeb(mp1);
6868 6866 BUMP_MIB(ill->ill_ip_mib,
6869 6867 ipIfStatsReasmPartDups);
6870 6868 mp1 = mp->b_cont;
6871 6869 if (!mp1)
6872 6870 break;
6873 6871 offset = IP_REASS_START(mp1);
6874 6872 }
6875 6873 ipf->ipf_mp->b_cont = mp;
6876 6874 continue;
6877 6875 }
6878 6876 /*
6879 6877 * The new piece starts somewhere between the start of the head
6880 6878 * and before the end of the tail.
6881 6879 */
6882 6880 for (; mp1; mp1 = mp1->b_cont) {
6883 6881 offset = IP_REASS_END(mp1);
6884 6882 if (start < offset) {
6885 6883 if (end <= offset) {
6886 6884 /* Nothing new. */
6887 6885 IP_REASS_SET_START(mp, 0);
6888 6886 IP_REASS_SET_END(mp, 0);
6889 6887 /* Subtract byte count */
6890 6888 ipf->ipf_count -= mp->b_datap->db_lim -
6891 6889 mp->b_datap->db_base;
6892 6890 if (incr_dups) {
6893 6891 ipf->ipf_num_dups++;
6894 6892 incr_dups = B_FALSE;
6895 6893 }
6896 6894 freeb(mp);
6897 6895 BUMP_MIB(ill->ill_ip_mib,
6898 6896 ipIfStatsReasmDuplicates);
6899 6897 break;
6900 6898 }
6901 6899 /*
6902 6900 * Trim redundant stuff off beginning of new
6903 6901 * piece.
6904 6902 */
6905 6903 IP_REASS_SET_START(mp, offset);
6906 6904 mp->b_rptr += offset - start;
6907 6905 BUMP_MIB(ill->ill_ip_mib,
6908 6906 ipIfStatsReasmPartDups);
6909 6907 start = offset;
6910 6908 if (!mp1->b_cont) {
6911 6909 /*
6912 6910 * After trimming, this guy is now
6913 6911 * hanging off the end.
6914 6912 */
6915 6913 mp1->b_cont = mp;
6916 6914 ipf->ipf_tail_mp = mp;
6917 6915 if (!more) {
6918 6916 ipf->ipf_hole_cnt--;
6919 6917 }
6920 6918 break;
6921 6919 }
6922 6920 }
6923 6921 if (start >= IP_REASS_START(mp1->b_cont))
6924 6922 continue;
6925 6923 /* Fill a hole */
6926 6924 if (start > offset)
6927 6925 ipf->ipf_hole_cnt++;
6928 6926 mp->b_cont = mp1->b_cont;
6929 6927 mp1->b_cont = mp;
6930 6928 mp1 = mp->b_cont;
6931 6929 offset = IP_REASS_START(mp1);
6932 6930 if (end >= offset) {
6933 6931 ipf->ipf_hole_cnt--;
6934 6932 /* Check for overlap. */
6935 6933 while (end > offset) {
6936 6934 if (end < IP_REASS_END(mp1)) {
6937 6935 mp->b_wptr -= end - offset;
6938 6936 IP_REASS_SET_END(mp, offset);
6939 6937 /*
6940 6938 * TODO we might bump
6941 6939 * this up twice if there is
6942 6940 * overlap at both ends.
6943 6941 */
6944 6942 BUMP_MIB(ill->ill_ip_mib,
6945 6943 ipIfStatsReasmPartDups);
6946 6944 break;
6947 6945 }
6948 6946 /* Did we cover another hole? */
6949 6947 if ((mp1->b_cont &&
6950 6948 IP_REASS_END(mp1)
6951 6949 != IP_REASS_START(mp1->b_cont) &&
6952 6950 end >=
6953 6951 IP_REASS_START(mp1->b_cont)) ||
6954 6952 (!ipf->ipf_last_frag_seen &&
6955 6953 !more)) {
6956 6954 ipf->ipf_hole_cnt--;
6957 6955 }
6958 6956 /* Clip out mp1. */
6959 6957 if ((mp->b_cont = mp1->b_cont) ==
6960 6958 NULL) {
6961 6959 /*
6962 6960 * After clipping out mp1,
6963 6961 * this guy is now hanging
6964 6962 * off the end.
6965 6963 */
6966 6964 ipf->ipf_tail_mp = mp;
6967 6965 }
6968 6966 IP_REASS_SET_START(mp1, 0);
6969 6967 IP_REASS_SET_END(mp1, 0);
6970 6968 /* Subtract byte count */
6971 6969 ipf->ipf_count -=
6972 6970 mp1->b_datap->db_lim -
6973 6971 mp1->b_datap->db_base;
6974 6972 freeb(mp1);
6975 6973 BUMP_MIB(ill->ill_ip_mib,
6976 6974 ipIfStatsReasmPartDups);
6977 6975 mp1 = mp->b_cont;
6978 6976 if (!mp1)
6979 6977 break;
6980 6978 offset = IP_REASS_START(mp1);
6981 6979 }
6982 6980 }
6983 6981 break;
6984 6982 }
6985 6983 } while (start = end, mp = next_mp);
6986 6984
6987 6985 /* Fragment just processed could be the last one. Remember this fact */
6988 6986 if (!more)
6989 6987 ipf->ipf_last_frag_seen = B_TRUE;
6990 6988
6991 6989 /* Still got holes? */
6992 6990 if (ipf->ipf_hole_cnt)
6993 6991 return (IP_REASS_PARTIAL);
6994 6992 /* Clean up overloaded fields to avoid upstream disasters. */
6995 6993 for (mp1 = ipf->ipf_mp->b_cont; mp1; mp1 = mp1->b_cont) {
6996 6994 IP_REASS_SET_START(mp1, 0);
6997 6995 IP_REASS_SET_END(mp1, 0);
6998 6996 }
6999 6997 return (IP_REASS_COMPLETE);
7000 6998 }
7001 6999
7002 7000 /*
7003 7001 * Fragmentation reassembly. Each ILL has a hash table for
7004 7002 * queuing packets undergoing reassembly for all IPIFs
7005 7003 * associated with the ILL. The hash is based on the packet
7006 7004 * IP ident field. The ILL frag hash table was allocated
7007 7005 * as a timer block at the time the ILL was created. Whenever
7008 7006 * there is anything on the reassembly queue, the timer will
7009 7007 * be running. Returns the reassembled packet if reassembly completes.
7010 7008 */
7011 7009 mblk_t *
7012 7010 ip_input_fragment(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
7013 7011 {
7014 7012 uint32_t frag_offset_flags;
7015 7013 mblk_t *t_mp;
7016 7014 ipaddr_t dst;
7017 7015 uint8_t proto = ipha->ipha_protocol;
7018 7016 uint32_t sum_val;
7019 7017 uint16_t sum_flags;
7020 7018 ipf_t *ipf;
7021 7019 ipf_t **ipfp;
7022 7020 ipfb_t *ipfb;
7023 7021 uint16_t ident;
7024 7022 uint32_t offset;
7025 7023 ipaddr_t src;
7026 7024 uint_t hdr_length;
7027 7025 uint32_t end;
7028 7026 mblk_t *mp1;
7029 7027 mblk_t *tail_mp;
7030 7028 size_t count;
7031 7029 size_t msg_len;
7032 7030 uint8_t ecn_info = 0;
7033 7031 uint32_t packet_size;
7034 7032 boolean_t pruned = B_FALSE;
7035 7033 ill_t *ill = ira->ira_ill;
7036 7034 ip_stack_t *ipst = ill->ill_ipst;
7037 7035
7038 7036 /*
7039 7037 * Drop the fragmented as early as possible, if
7040 7038 * we don't have resource(s) to re-assemble.
7041 7039 */
7042 7040 if (ipst->ips_ip_reass_queue_bytes == 0) {
7043 7041 freemsg(mp);
7044 7042 return (NULL);
7045 7043 }
7046 7044
7047 7045 /* Check for fragmentation offset; return if there's none */
7048 7046 if ((frag_offset_flags = ntohs(ipha->ipha_fragment_offset_and_flags) &
7049 7047 (IPH_MF | IPH_OFFSET)) == 0)
7050 7048 return (mp);
7051 7049
7052 7050 /*
7053 7051 * We utilize hardware computed checksum info only for UDP since
7054 7052 * IP fragmentation is a normal occurrence for the protocol. In
7055 7053 * addition, checksum offload support for IP fragments carrying
7056 7054 * UDP payload is commonly implemented across network adapters.
7057 7055 */
7058 7056 ASSERT(ira->ira_rill != NULL);
7059 7057 if (proto == IPPROTO_UDP && dohwcksum &&
7060 7058 ILL_HCKSUM_CAPABLE(ira->ira_rill) &&
7061 7059 (DB_CKSUMFLAGS(mp) & (HCK_FULLCKSUM | HCK_PARTIALCKSUM))) {
7062 7060 mblk_t *mp1 = mp->b_cont;
7063 7061 int32_t len;
7064 7062
7065 7063 /* Record checksum information from the packet */
7066 7064 sum_val = (uint32_t)DB_CKSUM16(mp);
7067 7065 sum_flags = DB_CKSUMFLAGS(mp);
7068 7066
7069 7067 /* IP payload offset from beginning of mblk */
7070 7068 offset = ((uchar_t *)ipha + IPH_HDR_LENGTH(ipha)) - mp->b_rptr;
7071 7069
7072 7070 if ((sum_flags & HCK_PARTIALCKSUM) &&
7073 7071 (mp1 == NULL || mp1->b_cont == NULL) &&
7074 7072 offset >= DB_CKSUMSTART(mp) &&
7075 7073 ((len = offset - DB_CKSUMSTART(mp)) & 1) == 0) {
7076 7074 uint32_t adj;
7077 7075 /*
7078 7076 * Partial checksum has been calculated by hardware
7079 7077 * and attached to the packet; in addition, any
7080 7078 * prepended extraneous data is even byte aligned.
7081 7079 * If any such data exists, we adjust the checksum;
7082 7080 * this would also handle any postpended data.
7083 7081 */
7084 7082 IP_ADJCKSUM_PARTIAL(mp->b_rptr + DB_CKSUMSTART(mp),
7085 7083 mp, mp1, len, adj);
7086 7084
7087 7085 /* One's complement subtract extraneous checksum */
7088 7086 if (adj >= sum_val)
7089 7087 sum_val = ~(adj - sum_val) & 0xFFFF;
7090 7088 else
7091 7089 sum_val -= adj;
7092 7090 }
7093 7091 } else {
7094 7092 sum_val = 0;
7095 7093 sum_flags = 0;
7096 7094 }
7097 7095
7098 7096 /* Clear hardware checksumming flag */
7099 7097 DB_CKSUMFLAGS(mp) = 0;
7100 7098
7101 7099 ident = ipha->ipha_ident;
7102 7100 offset = (frag_offset_flags << 3) & 0xFFFF;
7103 7101 src = ipha->ipha_src;
7104 7102 dst = ipha->ipha_dst;
7105 7103 hdr_length = IPH_HDR_LENGTH(ipha);
7106 7104 end = ntohs(ipha->ipha_length) - hdr_length;
7107 7105
7108 7106 /* If end == 0 then we have a packet with no data, so just free it */
7109 7107 if (end == 0) {
7110 7108 freemsg(mp);
7111 7109 return (NULL);
7112 7110 }
7113 7111
7114 7112 /* Record the ECN field info. */
7115 7113 ecn_info = (ipha->ipha_type_of_service & 0x3);
7116 7114 if (offset != 0) {
7117 7115 /*
7118 7116 * If this isn't the first piece, strip the header, and
7119 7117 * add the offset to the end value.
7120 7118 */
7121 7119 mp->b_rptr += hdr_length;
7122 7120 end += offset;
7123 7121 }
7124 7122
7125 7123 /* Handle vnic loopback of fragments */
7126 7124 if (mp->b_datap->db_ref > 2)
7127 7125 msg_len = 0;
7128 7126 else
7129 7127 msg_len = MBLKSIZE(mp);
7130 7128
7131 7129 tail_mp = mp;
7132 7130 while (tail_mp->b_cont != NULL) {
7133 7131 tail_mp = tail_mp->b_cont;
7134 7132 if (tail_mp->b_datap->db_ref <= 2)
7135 7133 msg_len += MBLKSIZE(tail_mp);
7136 7134 }
7137 7135
7138 7136 /* If the reassembly list for this ILL will get too big, prune it */
7139 7137 if ((msg_len + sizeof (*ipf) + ill->ill_frag_count) >=
7140 7138 ipst->ips_ip_reass_queue_bytes) {
7141 7139 DTRACE_PROBE3(ip_reass_queue_bytes, uint_t, msg_len,
7142 7140 uint_t, ill->ill_frag_count,
7143 7141 uint_t, ipst->ips_ip_reass_queue_bytes);
7144 7142 ill_frag_prune(ill,
7145 7143 (ipst->ips_ip_reass_queue_bytes < msg_len) ? 0 :
7146 7144 (ipst->ips_ip_reass_queue_bytes - msg_len));
7147 7145 pruned = B_TRUE;
7148 7146 }
7149 7147
7150 7148 ipfb = &ill->ill_frag_hash_tbl[ILL_FRAG_HASH(src, ident)];
7151 7149 mutex_enter(&ipfb->ipfb_lock);
7152 7150
7153 7151 ipfp = &ipfb->ipfb_ipf;
7154 7152 /* Try to find an existing fragment queue for this packet. */
7155 7153 for (;;) {
7156 7154 ipf = ipfp[0];
7157 7155 if (ipf != NULL) {
7158 7156 /*
7159 7157 * It has to match on ident and src/dst address.
7160 7158 */
7161 7159 if (ipf->ipf_ident == ident &&
7162 7160 ipf->ipf_src == src &&
7163 7161 ipf->ipf_dst == dst &&
7164 7162 ipf->ipf_protocol == proto) {
7165 7163 /*
7166 7164 * If we have received too many
7167 7165 * duplicate fragments for this packet
7168 7166 * free it.
7169 7167 */
7170 7168 if (ipf->ipf_num_dups > ip_max_frag_dups) {
7171 7169 ill_frag_free_pkts(ill, ipfb, ipf, 1);
7172 7170 freemsg(mp);
7173 7171 mutex_exit(&ipfb->ipfb_lock);
7174 7172 return (NULL);
7175 7173 }
7176 7174 /* Found it. */
7177 7175 break;
7178 7176 }
7179 7177 ipfp = &ipf->ipf_hash_next;
7180 7178 continue;
7181 7179 }
7182 7180
7183 7181 /*
7184 7182 * If we pruned the list, do we want to store this new
7185 7183 * fragment?. We apply an optimization here based on the
7186 7184 * fact that most fragments will be received in order.
7187 7185 * So if the offset of this incoming fragment is zero,
7188 7186 * it is the first fragment of a new packet. We will
7189 7187 * keep it. Otherwise drop the fragment, as we have
7190 7188 * probably pruned the packet already (since the
7191 7189 * packet cannot be found).
7192 7190 */
7193 7191 if (pruned && offset != 0) {
7194 7192 mutex_exit(&ipfb->ipfb_lock);
7195 7193 freemsg(mp);
7196 7194 return (NULL);
7197 7195 }
7198 7196
7199 7197 if (ipfb->ipfb_frag_pkts >= MAX_FRAG_PKTS(ipst)) {
7200 7198 /*
7201 7199 * Too many fragmented packets in this hash
7202 7200 * bucket. Free the oldest.
7203 7201 */
7204 7202 ill_frag_free_pkts(ill, ipfb, ipfb->ipfb_ipf, 1);
7205 7203 }
7206 7204
7207 7205 /* New guy. Allocate a frag message. */
7208 7206 mp1 = allocb(sizeof (*ipf), BPRI_MED);
7209 7207 if (mp1 == NULL) {
7210 7208 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7211 7209 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7212 7210 freemsg(mp);
7213 7211 reass_done:
7214 7212 mutex_exit(&ipfb->ipfb_lock);
7215 7213 return (NULL);
7216 7214 }
7217 7215
7218 7216 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmReqds);
7219 7217 mp1->b_cont = mp;
7220 7218
7221 7219 /* Initialize the fragment header. */
7222 7220 ipf = (ipf_t *)mp1->b_rptr;
7223 7221 ipf->ipf_mp = mp1;
7224 7222 ipf->ipf_ptphn = ipfp;
7225 7223 ipfp[0] = ipf;
7226 7224 ipf->ipf_hash_next = NULL;
7227 7225 ipf->ipf_ident = ident;
7228 7226 ipf->ipf_protocol = proto;
7229 7227 ipf->ipf_src = src;
7230 7228 ipf->ipf_dst = dst;
7231 7229 ipf->ipf_nf_hdr_len = 0;
7232 7230 /* Record reassembly start time. */
7233 7231 ipf->ipf_timestamp = gethrestime_sec();
7234 7232 /* Record ipf generation and account for frag header */
7235 7233 ipf->ipf_gen = ill->ill_ipf_gen++;
7236 7234 ipf->ipf_count = MBLKSIZE(mp1);
7237 7235 ipf->ipf_last_frag_seen = B_FALSE;
7238 7236 ipf->ipf_ecn = ecn_info;
7239 7237 ipf->ipf_num_dups = 0;
7240 7238 ipfb->ipfb_frag_pkts++;
7241 7239 ipf->ipf_checksum = 0;
7242 7240 ipf->ipf_checksum_flags = 0;
7243 7241
7244 7242 /* Store checksum value in fragment header */
7245 7243 if (sum_flags != 0) {
7246 7244 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7247 7245 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7248 7246 ipf->ipf_checksum = sum_val;
7249 7247 ipf->ipf_checksum_flags = sum_flags;
7250 7248 }
7251 7249
7252 7250 /*
7253 7251 * We handle reassembly two ways. In the easy case,
7254 7252 * where all the fragments show up in order, we do
7255 7253 * minimal bookkeeping, and just clip new pieces on
7256 7254 * the end. If we ever see a hole, then we go off
7257 7255 * to ip_reassemble which has to mark the pieces and
7258 7256 * keep track of the number of holes, etc. Obviously,
7259 7257 * the point of having both mechanisms is so we can
7260 7258 * handle the easy case as efficiently as possible.
7261 7259 */
7262 7260 if (offset == 0) {
7263 7261 /* Easy case, in-order reassembly so far. */
7264 7262 ipf->ipf_count += msg_len;
7265 7263 ipf->ipf_tail_mp = tail_mp;
7266 7264 /*
7267 7265 * Keep track of next expected offset in
7268 7266 * ipf_end.
7269 7267 */
7270 7268 ipf->ipf_end = end;
7271 7269 ipf->ipf_nf_hdr_len = hdr_length;
7272 7270 } else {
7273 7271 /* Hard case, hole at the beginning. */
7274 7272 ipf->ipf_tail_mp = NULL;
7275 7273 /*
7276 7274 * ipf_end == 0 means that we have given up
7277 7275 * on easy reassembly.
7278 7276 */
7279 7277 ipf->ipf_end = 0;
7280 7278
7281 7279 /* Forget checksum offload from now on */
7282 7280 ipf->ipf_checksum_flags = 0;
7283 7281
7284 7282 /*
7285 7283 * ipf_hole_cnt is set by ip_reassemble.
7286 7284 * ipf_count is updated by ip_reassemble.
7287 7285 * No need to check for return value here
7288 7286 * as we don't expect reassembly to complete
7289 7287 * or fail for the first fragment itself.
7290 7288 */
7291 7289 (void) ip_reassemble(mp, ipf,
7292 7290 (frag_offset_flags & IPH_OFFSET) << 3,
7293 7291 (frag_offset_flags & IPH_MF), ill, msg_len);
7294 7292 }
7295 7293 /* Update per ipfb and ill byte counts */
7296 7294 ipfb->ipfb_count += ipf->ipf_count;
7297 7295 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7298 7296 atomic_add_32(&ill->ill_frag_count, ipf->ipf_count);
7299 7297 /* If the frag timer wasn't already going, start it. */
7300 7298 mutex_enter(&ill->ill_lock);
7301 7299 ill_frag_timer_start(ill);
7302 7300 mutex_exit(&ill->ill_lock);
7303 7301 goto reass_done;
7304 7302 }
7305 7303
7306 7304 /*
7307 7305 * If the packet's flag has changed (it could be coming up
7308 7306 * from an interface different than the previous, therefore
7309 7307 * possibly different checksum capability), then forget about
7310 7308 * any stored checksum states. Otherwise add the value to
7311 7309 * the existing one stored in the fragment header.
7312 7310 */
7313 7311 if (sum_flags != 0 && sum_flags == ipf->ipf_checksum_flags) {
7314 7312 sum_val += ipf->ipf_checksum;
7315 7313 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7316 7314 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7317 7315 ipf->ipf_checksum = sum_val;
7318 7316 } else if (ipf->ipf_checksum_flags != 0) {
7319 7317 /* Forget checksum offload from now on */
7320 7318 ipf->ipf_checksum_flags = 0;
7321 7319 }
7322 7320
7323 7321 /*
7324 7322 * We have a new piece of a datagram which is already being
7325 7323 * reassembled. Update the ECN info if all IP fragments
7326 7324 * are ECN capable. If there is one which is not, clear
7327 7325 * all the info. If there is at least one which has CE
7328 7326 * code point, IP needs to report that up to transport.
7329 7327 */
7330 7328 if (ecn_info != IPH_ECN_NECT && ipf->ipf_ecn != IPH_ECN_NECT) {
7331 7329 if (ecn_info == IPH_ECN_CE)
7332 7330 ipf->ipf_ecn = IPH_ECN_CE;
7333 7331 } else {
7334 7332 ipf->ipf_ecn = IPH_ECN_NECT;
7335 7333 }
7336 7334 if (offset && ipf->ipf_end == offset) {
7337 7335 /* The new fragment fits at the end */
7338 7336 ipf->ipf_tail_mp->b_cont = mp;
7339 7337 /* Update the byte count */
7340 7338 ipf->ipf_count += msg_len;
7341 7339 /* Update per ipfb and ill byte counts */
7342 7340 ipfb->ipfb_count += msg_len;
7343 7341 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7344 7342 atomic_add_32(&ill->ill_frag_count, msg_len);
7345 7343 if (frag_offset_flags & IPH_MF) {
7346 7344 /* More to come. */
7347 7345 ipf->ipf_end = end;
7348 7346 ipf->ipf_tail_mp = tail_mp;
7349 7347 goto reass_done;
7350 7348 }
7351 7349 } else {
7352 7350 /* Go do the hard cases. */
7353 7351 int ret;
7354 7352
7355 7353 if (offset == 0)
7356 7354 ipf->ipf_nf_hdr_len = hdr_length;
7357 7355
7358 7356 /* Save current byte count */
7359 7357 count = ipf->ipf_count;
7360 7358 ret = ip_reassemble(mp, ipf,
7361 7359 (frag_offset_flags & IPH_OFFSET) << 3,
7362 7360 (frag_offset_flags & IPH_MF), ill, msg_len);
7363 7361 /* Count of bytes added and subtracted (freeb()ed) */
7364 7362 count = ipf->ipf_count - count;
7365 7363 if (count) {
7366 7364 /* Update per ipfb and ill byte counts */
7367 7365 ipfb->ipfb_count += count;
7368 7366 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7369 7367 atomic_add_32(&ill->ill_frag_count, count);
7370 7368 }
7371 7369 if (ret == IP_REASS_PARTIAL) {
7372 7370 goto reass_done;
7373 7371 } else if (ret == IP_REASS_FAILED) {
7374 7372 /* Reassembly failed. Free up all resources */
7375 7373 ill_frag_free_pkts(ill, ipfb, ipf, 1);
7376 7374 for (t_mp = mp; t_mp != NULL; t_mp = t_mp->b_cont) {
7377 7375 IP_REASS_SET_START(t_mp, 0);
7378 7376 IP_REASS_SET_END(t_mp, 0);
7379 7377 }
7380 7378 freemsg(mp);
7381 7379 goto reass_done;
7382 7380 }
7383 7381 /* We will reach here iff 'ret' is IP_REASS_COMPLETE */
7384 7382 }
7385 7383 /*
7386 7384 * We have completed reassembly. Unhook the frag header from
7387 7385 * the reassembly list.
7388 7386 *
7389 7387 * Before we free the frag header, record the ECN info
7390 7388 * to report back to the transport.
7391 7389 */
7392 7390 ecn_info = ipf->ipf_ecn;
7393 7391 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmOKs);
7394 7392 ipfp = ipf->ipf_ptphn;
7395 7393
7396 7394 /* We need to supply these to caller */
7397 7395 if ((sum_flags = ipf->ipf_checksum_flags) != 0)
7398 7396 sum_val = ipf->ipf_checksum;
7399 7397 else
7400 7398 sum_val = 0;
7401 7399
7402 7400 mp1 = ipf->ipf_mp;
7403 7401 count = ipf->ipf_count;
7404 7402 ipf = ipf->ipf_hash_next;
7405 7403 if (ipf != NULL)
7406 7404 ipf->ipf_ptphn = ipfp;
7407 7405 ipfp[0] = ipf;
7408 7406 atomic_add_32(&ill->ill_frag_count, -count);
7409 7407 ASSERT(ipfb->ipfb_count >= count);
7410 7408 ipfb->ipfb_count -= count;
7411 7409 ipfb->ipfb_frag_pkts--;
7412 7410 mutex_exit(&ipfb->ipfb_lock);
7413 7411 /* Ditch the frag header. */
7414 7412 mp = mp1->b_cont;
7415 7413
7416 7414 freeb(mp1);
7417 7415
7418 7416 /* Restore original IP length in header. */
7419 7417 packet_size = (uint32_t)msgdsize(mp);
7420 7418 if (packet_size > IP_MAXPACKET) {
7421 7419 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7422 7420 ip_drop_input("Reassembled packet too large", mp, ill);
7423 7421 freemsg(mp);
7424 7422 return (NULL);
7425 7423 }
7426 7424
7427 7425 if (DB_REF(mp) > 1) {
7428 7426 mblk_t *mp2 = copymsg(mp);
7429 7427
7430 7428 if (mp2 == NULL) {
7431 7429 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7432 7430 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7433 7431 freemsg(mp);
7434 7432 return (NULL);
7435 7433 }
7436 7434 freemsg(mp);
7437 7435 mp = mp2;
7438 7436 }
7439 7437 ipha = (ipha_t *)mp->b_rptr;
7440 7438
7441 7439 ipha->ipha_length = htons((uint16_t)packet_size);
7442 7440 /* We're now complete, zip the frag state */
7443 7441 ipha->ipha_fragment_offset_and_flags = 0;
7444 7442 /* Record the ECN info. */
7445 7443 ipha->ipha_type_of_service &= 0xFC;
7446 7444 ipha->ipha_type_of_service |= ecn_info;
7447 7445
7448 7446 /* Update the receive attributes */
7449 7447 ira->ira_pktlen = packet_size;
7450 7448 ira->ira_ip_hdr_length = IPH_HDR_LENGTH(ipha);
7451 7449
7452 7450 /* Reassembly is successful; set checksum information in packet */
7453 7451 DB_CKSUM16(mp) = (uint16_t)sum_val;
7454 7452 DB_CKSUMFLAGS(mp) = sum_flags;
7455 7453 DB_CKSUMSTART(mp) = ira->ira_ip_hdr_length;
7456 7454
7457 7455 return (mp);
7458 7456 }
7459 7457
7460 7458 /*
7461 7459 * Pullup function that should be used for IP input in order to
7462 7460 * ensure we do not loose the L2 source address; we need the l2 source
7463 7461 * address for IP_RECVSLLA and for ndp_input.
7464 7462 *
7465 7463 * We return either NULL or b_rptr.
7466 7464 */
7467 7465 void *
7468 7466 ip_pullup(mblk_t *mp, ssize_t len, ip_recv_attr_t *ira)
7469 7467 {
7470 7468 ill_t *ill = ira->ira_ill;
7471 7469
7472 7470 if (ip_rput_pullups++ == 0) {
7473 7471 (void) mi_strlog(ill->ill_rq, 1, SL_ERROR|SL_TRACE,
7474 7472 "ip_pullup: %s forced us to "
7475 7473 " pullup pkt, hdr len %ld, hdr addr %p",
7476 7474 ill->ill_name, len, (void *)mp->b_rptr);
7477 7475 }
7478 7476 if (!(ira->ira_flags & IRAF_L2SRC_SET))
7479 7477 ip_setl2src(mp, ira, ira->ira_rill);
7480 7478 ASSERT(ira->ira_flags & IRAF_L2SRC_SET);
7481 7479 if (!pullupmsg(mp, len))
7482 7480 return (NULL);
7483 7481 else
7484 7482 return (mp->b_rptr);
7485 7483 }
7486 7484
7487 7485 /*
7488 7486 * Make sure ira_l2src has an address. If we don't have one fill with zeros.
7489 7487 * When called from the ULP ira_rill will be NULL hence the caller has to
7490 7488 * pass in the ill.
7491 7489 */
7492 7490 /* ARGSUSED */
7493 7491 void
7494 7492 ip_setl2src(mblk_t *mp, ip_recv_attr_t *ira, ill_t *ill)
7495 7493 {
7496 7494 const uchar_t *addr;
7497 7495 int alen;
7498 7496
7499 7497 if (ira->ira_flags & IRAF_L2SRC_SET)
7500 7498 return;
7501 7499
7502 7500 ASSERT(ill != NULL);
7503 7501 alen = ill->ill_phys_addr_length;
7504 7502 ASSERT(alen <= sizeof (ira->ira_l2src));
7505 7503 if (ira->ira_mhip != NULL &&
7506 7504 (addr = ira->ira_mhip->mhi_saddr) != NULL) {
7507 7505 bcopy(addr, ira->ira_l2src, alen);
7508 7506 } else if ((ira->ira_flags & IRAF_L2SRC_LOOPBACK) &&
7509 7507 (addr = ill->ill_phys_addr) != NULL) {
7510 7508 bcopy(addr, ira->ira_l2src, alen);
7511 7509 } else {
7512 7510 bzero(ira->ira_l2src, alen);
7513 7511 }
7514 7512 ira->ira_flags |= IRAF_L2SRC_SET;
7515 7513 }
7516 7514
7517 7515 /*
7518 7516 * check ip header length and align it.
7519 7517 */
7520 7518 mblk_t *
7521 7519 ip_check_and_align_header(mblk_t *mp, uint_t min_size, ip_recv_attr_t *ira)
7522 7520 {
7523 7521 ill_t *ill = ira->ira_ill;
7524 7522 ssize_t len;
7525 7523
7526 7524 len = MBLKL(mp);
7527 7525
7528 7526 if (!OK_32PTR(mp->b_rptr))
7529 7527 IP_STAT(ill->ill_ipst, ip_notaligned);
7530 7528 else
7531 7529 IP_STAT(ill->ill_ipst, ip_recv_pullup);
7532 7530
7533 7531 /* Guard against bogus device drivers */
7534 7532 if (len < 0) {
7535 7533 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7536 7534 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7537 7535 freemsg(mp);
7538 7536 return (NULL);
7539 7537 }
7540 7538
7541 7539 if (len == 0) {
7542 7540 /* GLD sometimes sends up mblk with b_rptr == b_wptr! */
7543 7541 mblk_t *mp1 = mp->b_cont;
7544 7542
7545 7543 if (!(ira->ira_flags & IRAF_L2SRC_SET))
7546 7544 ip_setl2src(mp, ira, ira->ira_rill);
7547 7545 ASSERT(ira->ira_flags & IRAF_L2SRC_SET);
7548 7546
7549 7547 freeb(mp);
7550 7548 mp = mp1;
7551 7549 if (mp == NULL)
7552 7550 return (NULL);
7553 7551
7554 7552 if (OK_32PTR(mp->b_rptr) && MBLKL(mp) >= min_size)
7555 7553 return (mp);
7556 7554 }
7557 7555 if (ip_pullup(mp, min_size, ira) == NULL) {
7558 7556 if (msgdsize(mp) < min_size) {
7559 7557 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7560 7558 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7561 7559 } else {
7562 7560 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7563 7561 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7564 7562 }
7565 7563 freemsg(mp);
7566 7564 return (NULL);
7567 7565 }
7568 7566 return (mp);
7569 7567 }
7570 7568
7571 7569 /*
7572 7570 * Common code for IPv4 and IPv6 to check and pullup multi-mblks
7573 7571 */
7574 7572 mblk_t *
7575 7573 ip_check_length(mblk_t *mp, uchar_t *rptr, ssize_t len, uint_t pkt_len,
7576 7574 uint_t min_size, ip_recv_attr_t *ira)
7577 7575 {
7578 7576 ill_t *ill = ira->ira_ill;
7579 7577
7580 7578 /*
7581 7579 * Make sure we have data length consistent
7582 7580 * with the IP header.
7583 7581 */
7584 7582 if (mp->b_cont == NULL) {
7585 7583 /* pkt_len is based on ipha_len, not the mblk length */
7586 7584 if (pkt_len < min_size) {
7587 7585 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7588 7586 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7589 7587 freemsg(mp);
7590 7588 return (NULL);
7591 7589 }
7592 7590 if (len < 0) {
7593 7591 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
7594 7592 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
7595 7593 freemsg(mp);
7596 7594 return (NULL);
7597 7595 }
7598 7596 /* Drop any pad */
7599 7597 mp->b_wptr = rptr + pkt_len;
7600 7598 } else if ((len += msgdsize(mp->b_cont)) != 0) {
7601 7599 ASSERT(pkt_len >= min_size);
7602 7600 if (pkt_len < min_size) {
7603 7601 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7604 7602 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7605 7603 freemsg(mp);
7606 7604 return (NULL);
7607 7605 }
7608 7606 if (len < 0) {
7609 7607 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
7610 7608 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
7611 7609 freemsg(mp);
7612 7610 return (NULL);
7613 7611 }
7614 7612 /* Drop any pad */
7615 7613 (void) adjmsg(mp, -len);
7616 7614 /*
7617 7615 * adjmsg may have freed an mblk from the chain, hence
7618 7616 * invalidate any hw checksum here. This will force IP to
7619 7617 * calculate the checksum in sw, but only for this packet.
7620 7618 */
7621 7619 DB_CKSUMFLAGS(mp) = 0;
7622 7620 IP_STAT(ill->ill_ipst, ip_multimblk);
7623 7621 }
7624 7622 return (mp);
7625 7623 }
7626 7624
7627 7625 /*
7628 7626 * Check that the IPv4 opt_len is consistent with the packet and pullup
7629 7627 * the options.
7630 7628 */
7631 7629 mblk_t *
7632 7630 ip_check_optlen(mblk_t *mp, ipha_t *ipha, uint_t opt_len, uint_t pkt_len,
7633 7631 ip_recv_attr_t *ira)
7634 7632 {
7635 7633 ill_t *ill = ira->ira_ill;
7636 7634 ssize_t len;
7637 7635
7638 7636 /* Assume no IPv6 packets arrive over the IPv4 queue */
7639 7637 if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
7640 7638 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7641 7639 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInWrongIPVersion);
7642 7640 ip_drop_input("IPvN packet on IPv4 ill", mp, ill);
7643 7641 freemsg(mp);
7644 7642 return (NULL);
7645 7643 }
7646 7644
7647 7645 if (opt_len > (15 - IP_SIMPLE_HDR_LENGTH_IN_WORDS)) {
7648 7646 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7649 7647 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7650 7648 freemsg(mp);
7651 7649 return (NULL);
7652 7650 }
7653 7651 /*
7654 7652 * Recompute complete header length and make sure we
7655 7653 * have access to all of it.
7656 7654 */
7657 7655 len = ((size_t)opt_len + IP_SIMPLE_HDR_LENGTH_IN_WORDS) << 2;
7658 7656 if (len > (mp->b_wptr - mp->b_rptr)) {
7659 7657 if (len > pkt_len) {
7660 7658 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7661 7659 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7662 7660 freemsg(mp);
7663 7661 return (NULL);
7664 7662 }
7665 7663 if (ip_pullup(mp, len, ira) == NULL) {
7666 7664 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7667 7665 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7668 7666 freemsg(mp);
7669 7667 return (NULL);
7670 7668 }
7671 7669 }
7672 7670 return (mp);
7673 7671 }
7674 7672
7675 7673 /*
7676 7674 * Returns a new ire, or the same ire, or NULL.
7677 7675 * If a different IRE is returned, then it is held; the caller
7678 7676 * needs to release it.
7679 7677 * In no case is there any hold/release on the ire argument.
7680 7678 */
7681 7679 ire_t *
7682 7680 ip_check_multihome(void *addr, ire_t *ire, ill_t *ill)
7683 7681 {
7684 7682 ire_t *new_ire;
7685 7683 ill_t *ire_ill;
7686 7684 uint_t ifindex;
7687 7685 ip_stack_t *ipst = ill->ill_ipst;
7688 7686 boolean_t strict_check = B_FALSE;
7689 7687
7690 7688 /*
7691 7689 * IPMP common case: if IRE and ILL are in the same group, there's no
7692 7690 * issue (e.g. packet received on an underlying interface matched an
7693 7691 * IRE_LOCAL on its associated group interface).
7694 7692 */
7695 7693 ASSERT(ire->ire_ill != NULL);
7696 7694 if (IS_IN_SAME_ILLGRP(ill, ire->ire_ill))
7697 7695 return (ire);
7698 7696
7699 7697 /*
7700 7698 * Do another ire lookup here, using the ingress ill, to see if the
7701 7699 * interface is in a usesrc group.
7702 7700 * As long as the ills belong to the same group, we don't consider
7703 7701 * them to be arriving on the wrong interface. Thus, if the switch
7704 7702 * is doing inbound load spreading, we won't drop packets when the
7705 7703 * ip*_strict_dst_multihoming switch is on.
7706 7704 * We also need to check for IPIF_UNNUMBERED point2point interfaces
7707 7705 * where the local address may not be unique. In this case we were
7708 7706 * at the mercy of the initial ire lookup and the IRE_LOCAL it
7709 7707 * actually returned. The new lookup, which is more specific, should
7710 7708 * only find the IRE_LOCAL associated with the ingress ill if one
7711 7709 * exists.
7712 7710 */
7713 7711 if (ire->ire_ipversion == IPV4_VERSION) {
7714 7712 if (ipst->ips_ip_strict_dst_multihoming)
7715 7713 strict_check = B_TRUE;
7716 7714 new_ire = ire_ftable_lookup_v4(*((ipaddr_t *)addr), 0, 0,
7717 7715 IRE_LOCAL, ill, ALL_ZONES, NULL,
7718 7716 (MATCH_IRE_TYPE|MATCH_IRE_ILL), 0, ipst, NULL);
7719 7717 } else {
7720 7718 ASSERT(!IN6_IS_ADDR_MULTICAST((in6_addr_t *)addr));
7721 7719 if (ipst->ips_ipv6_strict_dst_multihoming)
7722 7720 strict_check = B_TRUE;
7723 7721 new_ire = ire_ftable_lookup_v6((in6_addr_t *)addr, NULL, NULL,
7724 7722 IRE_LOCAL, ill, ALL_ZONES, NULL,
7725 7723 (MATCH_IRE_TYPE|MATCH_IRE_ILL), 0, ipst, NULL);
7726 7724 }
7727 7725 /*
7728 7726 * If the same ire that was returned in ip_input() is found then this
7729 7727 * is an indication that usesrc groups are in use. The packet
7730 7728 * arrived on a different ill in the group than the one associated with
7731 7729 * the destination address. If a different ire was found then the same
7732 7730 * IP address must be hosted on multiple ills. This is possible with
7733 7731 * unnumbered point2point interfaces. We switch to use this new ire in
7734 7732 * order to have accurate interface statistics.
7735 7733 */
7736 7734 if (new_ire != NULL) {
7737 7735 /* Note: held in one case but not the other? Caller handles */
7738 7736 if (new_ire != ire)
7739 7737 return (new_ire);
7740 7738 /* Unchanged */
7741 7739 ire_refrele(new_ire);
7742 7740 return (ire);
7743 7741 }
7744 7742
7745 7743 /*
7746 7744 * Chase pointers once and store locally.
7747 7745 */
7748 7746 ASSERT(ire->ire_ill != NULL);
7749 7747 ire_ill = ire->ire_ill;
7750 7748 ifindex = ill->ill_usesrc_ifindex;
7751 7749
7752 7750 /*
7753 7751 * Check if it's a legal address on the 'usesrc' interface.
7754 7752 * For IPMP data addresses the IRE_LOCAL is the upper, hence we
7755 7753 * can just check phyint_ifindex.
7756 7754 */
7757 7755 if (ifindex != 0 && ifindex == ire_ill->ill_phyint->phyint_ifindex) {
7758 7756 return (ire);
7759 7757 }
7760 7758
7761 7759 /*
7762 7760 * If the ip*_strict_dst_multihoming switch is on then we can
7763 7761 * only accept this packet if the interface is marked as routing.
7764 7762 */
7765 7763 if (!(strict_check))
7766 7764 return (ire);
7767 7765
7768 7766 if ((ill->ill_flags & ire->ire_ill->ill_flags & ILLF_ROUTER) != 0) {
7769 7767 return (ire);
7770 7768 }
7771 7769 return (NULL);
7772 7770 }
7773 7771
7774 7772 /*
7775 7773 * This function is used to construct a mac_header_info_s from a
7776 7774 * DL_UNITDATA_IND message.
7777 7775 * The address fields in the mhi structure points into the message,
7778 7776 * thus the caller can't use those fields after freeing the message.
7779 7777 *
7780 7778 * We determine whether the packet received is a non-unicast packet
7781 7779 * and in doing so, determine whether or not it is broadcast vs multicast.
7782 7780 * For it to be a broadcast packet, we must have the appropriate mblk_t
7783 7781 * hanging off the ill_t. If this is either not present or doesn't match
7784 7782 * the destination mac address in the DL_UNITDATA_IND, the packet is deemed
7785 7783 * to be multicast. Thus NICs that have no broadcast address (or no
7786 7784 * capability for one, such as point to point links) cannot return as
7787 7785 * the packet being broadcast.
7788 7786 */
7789 7787 void
7790 7788 ip_dlur_to_mhi(ill_t *ill, mblk_t *mb, struct mac_header_info_s *mhip)
7791 7789 {
7792 7790 dl_unitdata_ind_t *ind = (dl_unitdata_ind_t *)mb->b_rptr;
7793 7791 mblk_t *bmp;
7794 7792 uint_t extra_offset;
7795 7793
7796 7794 bzero(mhip, sizeof (struct mac_header_info_s));
7797 7795
7798 7796 mhip->mhi_dsttype = MAC_ADDRTYPE_UNICAST;
7799 7797
7800 7798 if (ill->ill_sap_length < 0)
7801 7799 extra_offset = 0;
7802 7800 else
7803 7801 extra_offset = ill->ill_sap_length;
7804 7802
7805 7803 mhip->mhi_daddr = (uchar_t *)ind + ind->dl_dest_addr_offset +
7806 7804 extra_offset;
7807 7805 mhip->mhi_saddr = (uchar_t *)ind + ind->dl_src_addr_offset +
7808 7806 extra_offset;
7809 7807
7810 7808 if (!ind->dl_group_address)
7811 7809 return;
7812 7810
7813 7811 /* Multicast or broadcast */
7814 7812 mhip->mhi_dsttype = MAC_ADDRTYPE_MULTICAST;
7815 7813
7816 7814 if (ind->dl_dest_addr_offset > sizeof (*ind) &&
7817 7815 ind->dl_dest_addr_offset + ind->dl_dest_addr_length < MBLKL(mb) &&
7818 7816 (bmp = ill->ill_bcast_mp) != NULL) {
7819 7817 dl_unitdata_req_t *dlur;
7820 7818 uint8_t *bphys_addr;
7821 7819
7822 7820 dlur = (dl_unitdata_req_t *)bmp->b_rptr;
7823 7821 bphys_addr = (uchar_t *)dlur + dlur->dl_dest_addr_offset +
7824 7822 extra_offset;
7825 7823
7826 7824 if (bcmp(mhip->mhi_daddr, bphys_addr,
7827 7825 ind->dl_dest_addr_length) == 0)
7828 7826 mhip->mhi_dsttype = MAC_ADDRTYPE_BROADCAST;
7829 7827 }
7830 7828 }
7831 7829
7832 7830 /*
7833 7831 * This function is used to construct a mac_header_info_s from a
7834 7832 * M_DATA fastpath message from a DLPI driver.
7835 7833 * The address fields in the mhi structure points into the message,
7836 7834 * thus the caller can't use those fields after freeing the message.
7837 7835 *
7838 7836 * We determine whether the packet received is a non-unicast packet
7839 7837 * and in doing so, determine whether or not it is broadcast vs multicast.
7840 7838 * For it to be a broadcast packet, we must have the appropriate mblk_t
7841 7839 * hanging off the ill_t. If this is either not present or doesn't match
7842 7840 * the destination mac address in the DL_UNITDATA_IND, the packet is deemed
7843 7841 * to be multicast. Thus NICs that have no broadcast address (or no
7844 7842 * capability for one, such as point to point links) cannot return as
7845 7843 * the packet being broadcast.
7846 7844 */
7847 7845 void
7848 7846 ip_mdata_to_mhi(ill_t *ill, mblk_t *mp, struct mac_header_info_s *mhip)
7849 7847 {
7850 7848 mblk_t *bmp;
7851 7849 struct ether_header *pether;
7852 7850
7853 7851 bzero(mhip, sizeof (struct mac_header_info_s));
7854 7852
7855 7853 mhip->mhi_dsttype = MAC_ADDRTYPE_UNICAST;
7856 7854
7857 7855 pether = (struct ether_header *)((char *)mp->b_rptr
7858 7856 - sizeof (struct ether_header));
7859 7857
7860 7858 /*
7861 7859 * Make sure the interface is an ethernet type, since we don't
7862 7860 * know the header format for anything but Ethernet. Also make
7863 7861 * sure we are pointing correctly above db_base.
7864 7862 */
7865 7863 if (ill->ill_type != IFT_ETHER)
7866 7864 return;
7867 7865
7868 7866 retry:
7869 7867 if ((uchar_t *)pether < mp->b_datap->db_base)
7870 7868 return;
7871 7869
7872 7870 /* Is there a VLAN tag? */
7873 7871 if (ill->ill_isv6) {
7874 7872 if (pether->ether_type != htons(ETHERTYPE_IPV6)) {
7875 7873 pether = (struct ether_header *)((char *)pether - 4);
7876 7874 goto retry;
7877 7875 }
7878 7876 } else {
7879 7877 if (pether->ether_type != htons(ETHERTYPE_IP)) {
7880 7878 pether = (struct ether_header *)((char *)pether - 4);
7881 7879 goto retry;
7882 7880 }
7883 7881 }
7884 7882 mhip->mhi_daddr = (uchar_t *)&pether->ether_dhost;
7885 7883 mhip->mhi_saddr = (uchar_t *)&pether->ether_shost;
7886 7884
7887 7885 if (!(mhip->mhi_daddr[0] & 0x01))
7888 7886 return;
7889 7887
7890 7888 /* Multicast or broadcast */
7891 7889 mhip->mhi_dsttype = MAC_ADDRTYPE_MULTICAST;
7892 7890
7893 7891 if ((bmp = ill->ill_bcast_mp) != NULL) {
7894 7892 dl_unitdata_req_t *dlur;
7895 7893 uint8_t *bphys_addr;
7896 7894 uint_t addrlen;
7897 7895
7898 7896 dlur = (dl_unitdata_req_t *)bmp->b_rptr;
7899 7897 addrlen = dlur->dl_dest_addr_length;
7900 7898 if (ill->ill_sap_length < 0) {
7901 7899 bphys_addr = (uchar_t *)dlur +
7902 7900 dlur->dl_dest_addr_offset;
7903 7901 addrlen += ill->ill_sap_length;
7904 7902 } else {
7905 7903 bphys_addr = (uchar_t *)dlur +
7906 7904 dlur->dl_dest_addr_offset +
7907 7905 ill->ill_sap_length;
7908 7906 addrlen -= ill->ill_sap_length;
7909 7907 }
7910 7908 if (bcmp(mhip->mhi_daddr, bphys_addr, addrlen) == 0)
7911 7909 mhip->mhi_dsttype = MAC_ADDRTYPE_BROADCAST;
7912 7910 }
7913 7911 }
7914 7912
7915 7913 /*
7916 7914 * Handle anything but M_DATA messages
7917 7915 * We see the DL_UNITDATA_IND which are part
7918 7916 * of the data path, and also the other messages from the driver.
7919 7917 */
7920 7918 void
7921 7919 ip_rput_notdata(ill_t *ill, mblk_t *mp)
7922 7920 {
7923 7921 mblk_t *first_mp;
7924 7922 struct iocblk *iocp;
7925 7923 struct mac_header_info_s mhi;
7926 7924
7927 7925 switch (DB_TYPE(mp)) {
7928 7926 case M_PROTO:
7929 7927 case M_PCPROTO: {
7930 7928 if (((dl_unitdata_ind_t *)mp->b_rptr)->dl_primitive !=
7931 7929 DL_UNITDATA_IND) {
7932 7930 /* Go handle anything other than data elsewhere. */
7933 7931 ip_rput_dlpi(ill, mp);
7934 7932 return;
7935 7933 }
7936 7934
7937 7935 first_mp = mp;
7938 7936 mp = first_mp->b_cont;
7939 7937 first_mp->b_cont = NULL;
7940 7938
7941 7939 if (mp == NULL) {
7942 7940 freeb(first_mp);
7943 7941 return;
7944 7942 }
7945 7943 ip_dlur_to_mhi(ill, first_mp, &mhi);
7946 7944 if (ill->ill_isv6)
7947 7945 ip_input_v6(ill, NULL, mp, &mhi);
7948 7946 else
7949 7947 ip_input(ill, NULL, mp, &mhi);
7950 7948
7951 7949 /* Ditch the DLPI header. */
7952 7950 freeb(first_mp);
7953 7951 return;
7954 7952 }
7955 7953 case M_IOCACK:
7956 7954 iocp = (struct iocblk *)mp->b_rptr;
7957 7955 switch (iocp->ioc_cmd) {
7958 7956 case DL_IOC_HDR_INFO:
7959 7957 ill_fastpath_ack(ill, mp);
7960 7958 return;
7961 7959 default:
7962 7960 putnext(ill->ill_rq, mp);
7963 7961 return;
7964 7962 }
7965 7963 /* FALLTHROUGH */
7966 7964 case M_ERROR:
7967 7965 case M_HANGUP:
7968 7966 mutex_enter(&ill->ill_lock);
7969 7967 if (ill->ill_state_flags & ILL_CONDEMNED) {
7970 7968 mutex_exit(&ill->ill_lock);
7971 7969 freemsg(mp);
7972 7970 return;
7973 7971 }
7974 7972 ill_refhold_locked(ill);
7975 7973 mutex_exit(&ill->ill_lock);
7976 7974 qwriter_ip(ill, ill->ill_rq, mp, ip_rput_other, CUR_OP,
7977 7975 B_FALSE);
7978 7976 return;
7979 7977 case M_CTL:
7980 7978 putnext(ill->ill_rq, mp);
7981 7979 return;
7982 7980 case M_IOCNAK:
7983 7981 ip1dbg(("got iocnak "));
7984 7982 iocp = (struct iocblk *)mp->b_rptr;
7985 7983 switch (iocp->ioc_cmd) {
7986 7984 case DL_IOC_HDR_INFO:
7987 7985 ip_rput_other(NULL, ill->ill_rq, mp, NULL);
7988 7986 return;
7989 7987 default:
7990 7988 break;
7991 7989 }
7992 7990 /* FALLTHROUGH */
7993 7991 default:
7994 7992 putnext(ill->ill_rq, mp);
7995 7993 return;
7996 7994 }
7997 7995 }
7998 7996
7999 7997 /* Read side put procedure. Packets coming from the wire arrive here. */
8000 7998 int
8001 7999 ip_rput(queue_t *q, mblk_t *mp)
8002 8000 {
8003 8001 ill_t *ill;
8004 8002 union DL_primitives *dl;
8005 8003
8006 8004 ill = (ill_t *)q->q_ptr;
8007 8005
8008 8006 if (ill->ill_state_flags & (ILL_CONDEMNED | ILL_LL_SUBNET_PENDING)) {
8009 8007 /*
8010 8008 * If things are opening or closing, only accept high-priority
8011 8009 * DLPI messages. (On open ill->ill_ipif has not yet been
8012 8010 * created; on close, things hanging off the ill may have been
8013 8011 * freed already.)
8014 8012 */
8015 8013 dl = (union DL_primitives *)mp->b_rptr;
8016 8014 if (DB_TYPE(mp) != M_PCPROTO ||
8017 8015 dl->dl_primitive == DL_UNITDATA_IND) {
8018 8016 inet_freemsg(mp);
8019 8017 return (0);
8020 8018 }
8021 8019 }
8022 8020 if (DB_TYPE(mp) == M_DATA) {
8023 8021 struct mac_header_info_s mhi;
8024 8022
8025 8023 ip_mdata_to_mhi(ill, mp, &mhi);
8026 8024 ip_input(ill, NULL, mp, &mhi);
8027 8025 } else {
8028 8026 ip_rput_notdata(ill, mp);
8029 8027 }
8030 8028 return (0);
8031 8029 }
8032 8030
8033 8031 /*
8034 8032 * Move the information to a copy.
8035 8033 */
8036 8034 mblk_t *
8037 8035 ip_fix_dbref(mblk_t *mp, ip_recv_attr_t *ira)
8038 8036 {
8039 8037 mblk_t *mp1;
8040 8038 ill_t *ill = ira->ira_ill;
8041 8039 ip_stack_t *ipst = ill->ill_ipst;
8042 8040
8043 8041 IP_STAT(ipst, ip_db_ref);
8044 8042
8045 8043 /* Make sure we have ira_l2src before we loose the original mblk */
8046 8044 if (!(ira->ira_flags & IRAF_L2SRC_SET))
8047 8045 ip_setl2src(mp, ira, ira->ira_rill);
8048 8046
8049 8047 mp1 = copymsg(mp);
8050 8048 if (mp1 == NULL) {
8051 8049 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
8052 8050 ip_drop_input("ipIfStatsInDiscards", mp, ill);
8053 8051 freemsg(mp);
8054 8052 return (NULL);
8055 8053 }
8056 8054 /* preserve the hardware checksum flags and data, if present */
8057 8055 if (DB_CKSUMFLAGS(mp) != 0) {
8058 8056 DB_CKSUMFLAGS(mp1) = DB_CKSUMFLAGS(mp);
8059 8057 DB_CKSUMSTART(mp1) = DB_CKSUMSTART(mp);
8060 8058 DB_CKSUMSTUFF(mp1) = DB_CKSUMSTUFF(mp);
8061 8059 DB_CKSUMEND(mp1) = DB_CKSUMEND(mp);
8062 8060 DB_CKSUM16(mp1) = DB_CKSUM16(mp);
8063 8061 }
8064 8062 freemsg(mp);
8065 8063 return (mp1);
8066 8064 }
8067 8065
8068 8066 static void
8069 8067 ip_dlpi_error(ill_t *ill, t_uscalar_t prim, t_uscalar_t dl_err,
8070 8068 t_uscalar_t err)
8071 8069 {
8072 8070 if (dl_err == DL_SYSERR) {
8073 8071 (void) mi_strlog(ill->ill_rq, 1, SL_CONSOLE|SL_ERROR|SL_TRACE,
8074 8072 "%s: %s failed: DL_SYSERR (errno %u)\n",
8075 8073 ill->ill_name, dl_primstr(prim), err);
8076 8074 return;
8077 8075 }
8078 8076
8079 8077 (void) mi_strlog(ill->ill_rq, 1, SL_CONSOLE|SL_ERROR|SL_TRACE,
8080 8078 "%s: %s failed: %s\n", ill->ill_name, dl_primstr(prim),
8081 8079 dl_errstr(dl_err));
8082 8080 }
8083 8081
8084 8082 /*
8085 8083 * ip_rput_dlpi is called by ip_rput to handle all DLPI messages other
8086 8084 * than DL_UNITDATA_IND messages. If we need to process this message
8087 8085 * exclusively, we call qwriter_ip, in which case we also need to call
8088 8086 * ill_refhold before that, since qwriter_ip does an ill_refrele.
8089 8087 */
8090 8088 void
8091 8089 ip_rput_dlpi(ill_t *ill, mblk_t *mp)
8092 8090 {
8093 8091 dl_ok_ack_t *dloa = (dl_ok_ack_t *)mp->b_rptr;
8094 8092 dl_error_ack_t *dlea = (dl_error_ack_t *)dloa;
8095 8093 queue_t *q = ill->ill_rq;
8096 8094 t_uscalar_t prim = dloa->dl_primitive;
8097 8095 t_uscalar_t reqprim = DL_PRIM_INVAL;
8098 8096
8099 8097 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi",
8100 8098 char *, dl_primstr(prim), ill_t *, ill);
8101 8099 ip1dbg(("ip_rput_dlpi"));
8102 8100
8103 8101 /*
8104 8102 * If we received an ACK but didn't send a request for it, then it
8105 8103 * can't be part of any pending operation; discard up-front.
8106 8104 */
8107 8105 switch (prim) {
8108 8106 case DL_ERROR_ACK:
8109 8107 reqprim = dlea->dl_error_primitive;
8110 8108 ip2dbg(("ip_rput_dlpi(%s): DL_ERROR_ACK for %s (0x%x): %s "
8111 8109 "(0x%x), unix %u\n", ill->ill_name, dl_primstr(reqprim),
8112 8110 reqprim, dl_errstr(dlea->dl_errno), dlea->dl_errno,
8113 8111 dlea->dl_unix_errno));
8114 8112 break;
8115 8113 case DL_OK_ACK:
8116 8114 reqprim = dloa->dl_correct_primitive;
8117 8115 break;
8118 8116 case DL_INFO_ACK:
8119 8117 reqprim = DL_INFO_REQ;
8120 8118 break;
8121 8119 case DL_BIND_ACK:
8122 8120 reqprim = DL_BIND_REQ;
8123 8121 break;
8124 8122 case DL_PHYS_ADDR_ACK:
8125 8123 reqprim = DL_PHYS_ADDR_REQ;
8126 8124 break;
8127 8125 case DL_NOTIFY_ACK:
8128 8126 reqprim = DL_NOTIFY_REQ;
8129 8127 break;
8130 8128 case DL_CAPABILITY_ACK:
8131 8129 reqprim = DL_CAPABILITY_REQ;
8132 8130 break;
8133 8131 }
8134 8132
8135 8133 if (prim != DL_NOTIFY_IND) {
8136 8134 if (reqprim == DL_PRIM_INVAL ||
8137 8135 !ill_dlpi_pending(ill, reqprim)) {
8138 8136 /* Not a DLPI message we support or expected */
8139 8137 freemsg(mp);
8140 8138 return;
8141 8139 }
8142 8140 ip1dbg(("ip_rput: received %s for %s\n", dl_primstr(prim),
8143 8141 dl_primstr(reqprim)));
8144 8142 }
8145 8143
8146 8144 switch (reqprim) {
8147 8145 case DL_UNBIND_REQ:
8148 8146 /*
8149 8147 * NOTE: we mark the unbind as complete even if we got a
8150 8148 * DL_ERROR_ACK, since there's not much else we can do.
8151 8149 */
8152 8150 mutex_enter(&ill->ill_lock);
8153 8151 ill->ill_state_flags &= ~ILL_DL_UNBIND_IN_PROGRESS;
8154 8152 cv_signal(&ill->ill_cv);
8155 8153 mutex_exit(&ill->ill_lock);
8156 8154 break;
8157 8155
8158 8156 case DL_ENABMULTI_REQ:
8159 8157 if (prim == DL_OK_ACK) {
8160 8158 if (ill->ill_dlpi_multicast_state == IDS_INPROGRESS)
8161 8159 ill->ill_dlpi_multicast_state = IDS_OK;
8162 8160 }
8163 8161 break;
8164 8162 }
8165 8163
8166 8164 /*
8167 8165 * The message is one we're waiting for (or DL_NOTIFY_IND), but we
8168 8166 * need to become writer to continue to process it. Because an
8169 8167 * exclusive operation doesn't complete until replies to all queued
8170 8168 * DLPI messages have been received, we know we're in the middle of an
8171 8169 * exclusive operation and pass CUR_OP (except for DL_NOTIFY_IND).
8172 8170 *
8173 8171 * As required by qwriter_ip(), we refhold the ill; it will refrele.
8174 8172 * Since this is on the ill stream we unconditionally bump up the
8175 8173 * refcount without doing ILL_CAN_LOOKUP().
8176 8174 */
8177 8175 ill_refhold(ill);
8178 8176 if (prim == DL_NOTIFY_IND)
8179 8177 qwriter_ip(ill, q, mp, ip_rput_dlpi_writer, NEW_OP, B_FALSE);
8180 8178 else
8181 8179 qwriter_ip(ill, q, mp, ip_rput_dlpi_writer, CUR_OP, B_FALSE);
8182 8180 }
8183 8181
8184 8182 /*
8185 8183 * Handling of DLPI messages that require exclusive access to the ipsq.
8186 8184 *
8187 8185 * Need to do ipsq_pending_mp_get on ioctl completion, which could
8188 8186 * happen here. (along with mi_copy_done)
8189 8187 */
8190 8188 /* ARGSUSED */
8191 8189 static void
8192 8190 ip_rput_dlpi_writer(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
8193 8191 {
8194 8192 dl_ok_ack_t *dloa = (dl_ok_ack_t *)mp->b_rptr;
8195 8193 dl_error_ack_t *dlea = (dl_error_ack_t *)dloa;
8196 8194 int err = 0;
8197 8195 ill_t *ill = (ill_t *)q->q_ptr;
8198 8196 ipif_t *ipif = NULL;
8199 8197 mblk_t *mp1 = NULL;
8200 8198 conn_t *connp = NULL;
8201 8199 t_uscalar_t paddrreq;
8202 8200 mblk_t *mp_hw;
8203 8201 boolean_t ioctl_aborted = B_FALSE;
8204 8202 boolean_t log = B_TRUE;
8205 8203
8206 8204 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer",
8207 8205 char *, dl_primstr(dloa->dl_primitive), ill_t *, ill);
8208 8206
8209 8207 ip1dbg(("ip_rput_dlpi_writer .."));
8210 8208 ASSERT(ipsq->ipsq_xop == ill->ill_phyint->phyint_ipsq->ipsq_xop);
8211 8209 ASSERT(IAM_WRITER_ILL(ill));
8212 8210
8213 8211 ipif = ipsq->ipsq_xop->ipx_pending_ipif;
8214 8212 /*
8215 8213 * The current ioctl could have been aborted by the user and a new
8216 8214 * ioctl to bring up another ill could have started. We could still
8217 8215 * get a response from the driver later.
8218 8216 */
8219 8217 if (ipif != NULL && ipif->ipif_ill != ill)
8220 8218 ioctl_aborted = B_TRUE;
8221 8219
8222 8220 switch (dloa->dl_primitive) {
8223 8221 case DL_ERROR_ACK:
8224 8222 ip1dbg(("ip_rput_dlpi_writer: got DL_ERROR_ACK for %s\n",
8225 8223 dl_primstr(dlea->dl_error_primitive)));
8226 8224
8227 8225 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer error",
8228 8226 char *, dl_primstr(dlea->dl_error_primitive),
8229 8227 ill_t *, ill);
8230 8228
8231 8229 switch (dlea->dl_error_primitive) {
8232 8230 case DL_DISABMULTI_REQ:
8233 8231 ill_dlpi_done(ill, dlea->dl_error_primitive);
8234 8232 break;
8235 8233 case DL_PROMISCON_REQ:
8236 8234 case DL_PROMISCOFF_REQ:
8237 8235 case DL_UNBIND_REQ:
8238 8236 case DL_ATTACH_REQ:
8239 8237 case DL_INFO_REQ:
8240 8238 ill_dlpi_done(ill, dlea->dl_error_primitive);
8241 8239 break;
8242 8240 case DL_NOTIFY_REQ:
8243 8241 ill_dlpi_done(ill, DL_NOTIFY_REQ);
8244 8242 log = B_FALSE;
8245 8243 break;
8246 8244 case DL_PHYS_ADDR_REQ:
8247 8245 /*
8248 8246 * For IPv6 only, there are two additional
8249 8247 * phys_addr_req's sent to the driver to get the
8250 8248 * IPv6 token and lla. This allows IP to acquire
8251 8249 * the hardware address format for a given interface
8252 8250 * without having built in knowledge of the hardware
8253 8251 * address. ill_phys_addr_pend keeps track of the last
8254 8252 * DL_PAR sent so we know which response we are
8255 8253 * dealing with. ill_dlpi_done will update
8256 8254 * ill_phys_addr_pend when it sends the next req.
8257 8255 * We don't complete the IOCTL until all three DL_PARs
8258 8256 * have been attempted, so set *_len to 0 and break.
8259 8257 */
8260 8258 paddrreq = ill->ill_phys_addr_pend;
8261 8259 ill_dlpi_done(ill, DL_PHYS_ADDR_REQ);
8262 8260 if (paddrreq == DL_IPV6_TOKEN) {
8263 8261 ill->ill_token_length = 0;
8264 8262 log = B_FALSE;
8265 8263 break;
8266 8264 } else if (paddrreq == DL_IPV6_LINK_LAYER_ADDR) {
8267 8265 ill->ill_nd_lla_len = 0;
8268 8266 log = B_FALSE;
8269 8267 break;
8270 8268 }
8271 8269 /*
8272 8270 * Something went wrong with the DL_PHYS_ADDR_REQ.
8273 8271 * We presumably have an IOCTL hanging out waiting
8274 8272 * for completion. Find it and complete the IOCTL
8275 8273 * with the error noted.
8276 8274 * However, ill_dl_phys was called on an ill queue
8277 8275 * (from SIOCSLIFNAME), thus conn_pending_ill is not
8278 8276 * set. But the ioctl is known to be pending on ill_wq.
8279 8277 */
8280 8278 if (!ill->ill_ifname_pending)
8281 8279 break;
8282 8280 ill->ill_ifname_pending = 0;
8283 8281 if (!ioctl_aborted)
8284 8282 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8285 8283 if (mp1 != NULL) {
8286 8284 /*
8287 8285 * This operation (SIOCSLIFNAME) must have
8288 8286 * happened on the ill. Assert there is no conn
8289 8287 */
8290 8288 ASSERT(connp == NULL);
8291 8289 q = ill->ill_wq;
8292 8290 }
8293 8291 break;
8294 8292 case DL_BIND_REQ:
8295 8293 ill_dlpi_done(ill, DL_BIND_REQ);
8296 8294 if (ill->ill_ifname_pending)
8297 8295 break;
8298 8296 mutex_enter(&ill->ill_lock);
8299 8297 ill->ill_state_flags &= ~ILL_DOWN_IN_PROGRESS;
8300 8298 mutex_exit(&ill->ill_lock);
8301 8299 /*
8302 8300 * Something went wrong with the bind. If this was the
8303 8301 * result of a DL_NOTE_REPLUMB, then we presumably
8304 8302 * have an IOCTL hanging out waiting for completion.
8305 8303 * Find it, take down the interface that was coming
8306 8304 * up, and complete the IOCTL with the error noted.
8307 8305 */
8308 8306 if (!ioctl_aborted)
8309 8307 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8310 8308 if (mp1 != NULL) {
8311 8309 /*
8312 8310 * This might be a result of a DL_NOTE_REPLUMB
8313 8311 * notification. In that case, connp is NULL.
8314 8312 */
8315 8313 if (connp != NULL)
8316 8314 q = CONNP_TO_WQ(connp);
8317 8315
8318 8316 (void) ipif_down(ipif, NULL, NULL);
8319 8317 /* error is set below the switch */
8320 8318 } else {
8321 8319 /*
8322 8320 * There's no pending IOCTL, so the bind was
8323 8321 * most likely started by ill_dl_up(). We save
8324 8322 * the error and let it take care of responding
8325 8323 * to the IOCTL.
8326 8324 */
8327 8325 ill->ill_dl_bind_err = dlea->dl_unix_errno ?
8328 8326 dlea->dl_unix_errno : ENXIO;
8329 8327 }
8330 8328 break;
8331 8329 case DL_ENABMULTI_REQ:
8332 8330 ill_dlpi_done(ill, DL_ENABMULTI_REQ);
8333 8331
8334 8332 if (ill->ill_dlpi_multicast_state == IDS_INPROGRESS)
8335 8333 ill->ill_dlpi_multicast_state = IDS_FAILED;
8336 8334 if (ill->ill_dlpi_multicast_state == IDS_FAILED) {
8337 8335
8338 8336 printf("ip: joining multicasts failed (%d)"
8339 8337 " on %s - will use link layer "
8340 8338 "broadcasts for multicast\n",
8341 8339 dlea->dl_errno, ill->ill_name);
8342 8340
8343 8341 /*
8344 8342 * Set up for multi_bcast; We are the
8345 8343 * writer, so ok to access ill->ill_ipif
8346 8344 * without any lock.
8347 8345 */
8348 8346 mutex_enter(&ill->ill_phyint->phyint_lock);
8349 8347 ill->ill_phyint->phyint_flags |=
8350 8348 PHYI_MULTI_BCAST;
8351 8349 mutex_exit(&ill->ill_phyint->phyint_lock);
8352 8350
8353 8351 }
8354 8352 freemsg(mp); /* Don't want to pass this up */
8355 8353 return;
8356 8354 case DL_CAPABILITY_REQ:
8357 8355 ip1dbg(("ip_rput_dlpi_writer: got DL_ERROR_ACK for "
8358 8356 "DL_CAPABILITY REQ\n"));
8359 8357 if (ill->ill_dlpi_capab_state == IDCS_PROBE_SENT)
8360 8358 ill->ill_dlpi_capab_state = IDCS_FAILED;
8361 8359 ill_capability_done(ill);
8362 8360 freemsg(mp);
8363 8361 return;
8364 8362 }
8365 8363 /*
8366 8364 * Note the error for IOCTL completion (mp1 is set when
8367 8365 * ready to complete ioctl). If ill_ifname_pending_err is
8368 8366 * set, an error occured during plumbing (ill_ifname_pending),
8369 8367 * so we want to report that error.
8370 8368 *
8371 8369 * NOTE: there are two addtional DL_PHYS_ADDR_REQ's
8372 8370 * (DL_IPV6_TOKEN and DL_IPV6_LINK_LAYER_ADDR) that are
8373 8371 * expected to get errack'd if the driver doesn't support
8374 8372 * these flags (e.g. ethernet). log will be set to B_FALSE
8375 8373 * if these error conditions are encountered.
8376 8374 */
8377 8375 if (mp1 != NULL) {
8378 8376 if (ill->ill_ifname_pending_err != 0) {
8379 8377 err = ill->ill_ifname_pending_err;
8380 8378 ill->ill_ifname_pending_err = 0;
8381 8379 } else {
8382 8380 err = dlea->dl_unix_errno ?
8383 8381 dlea->dl_unix_errno : ENXIO;
8384 8382 }
8385 8383 /*
8386 8384 * If we're plumbing an interface and an error hasn't already
8387 8385 * been saved, set ill_ifname_pending_err to the error passed
8388 8386 * up. Ignore the error if log is B_FALSE (see comment above).
8389 8387 */
8390 8388 } else if (log && ill->ill_ifname_pending &&
8391 8389 ill->ill_ifname_pending_err == 0) {
8392 8390 ill->ill_ifname_pending_err = dlea->dl_unix_errno ?
8393 8391 dlea->dl_unix_errno : ENXIO;
8394 8392 }
8395 8393
8396 8394 if (log)
8397 8395 ip_dlpi_error(ill, dlea->dl_error_primitive,
8398 8396 dlea->dl_errno, dlea->dl_unix_errno);
8399 8397 break;
8400 8398 case DL_CAPABILITY_ACK:
8401 8399 ill_capability_ack(ill, mp);
8402 8400 /*
8403 8401 * The message has been handed off to ill_capability_ack
8404 8402 * and must not be freed below
8405 8403 */
8406 8404 mp = NULL;
8407 8405 break;
8408 8406
8409 8407 case DL_INFO_ACK:
8410 8408 /* Call a routine to handle this one. */
8411 8409 ill_dlpi_done(ill, DL_INFO_REQ);
8412 8410 ip_ll_subnet_defaults(ill, mp);
8413 8411 ASSERT(!MUTEX_HELD(&ill->ill_phyint->phyint_ipsq->ipsq_lock));
8414 8412 return;
8415 8413 case DL_BIND_ACK:
8416 8414 /*
8417 8415 * We should have an IOCTL waiting on this unless
8418 8416 * sent by ill_dl_phys, in which case just return
8419 8417 */
8420 8418 ill_dlpi_done(ill, DL_BIND_REQ);
8421 8419
8422 8420 if (ill->ill_ifname_pending) {
8423 8421 DTRACE_PROBE2(ip__rput__dlpi__ifname__pending,
8424 8422 ill_t *, ill, mblk_t *, mp);
8425 8423 break;
8426 8424 }
8427 8425 mutex_enter(&ill->ill_lock);
8428 8426 ill->ill_dl_up = 1;
8429 8427 ill->ill_state_flags &= ~ILL_DOWN_IN_PROGRESS;
8430 8428 mutex_exit(&ill->ill_lock);
8431 8429
8432 8430 if (!ioctl_aborted)
8433 8431 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8434 8432 if (mp1 == NULL) {
8435 8433 DTRACE_PROBE1(ip__rput__dlpi__no__mblk, ill_t *, ill);
8436 8434 break;
8437 8435 }
8438 8436 /*
8439 8437 * mp1 was added by ill_dl_up(). if that is a result of
8440 8438 * a DL_NOTE_REPLUMB notification, connp could be NULL.
8441 8439 */
8442 8440 if (connp != NULL)
8443 8441 q = CONNP_TO_WQ(connp);
8444 8442 /*
8445 8443 * We are exclusive. So nothing can change even after
8446 8444 * we get the pending mp.
8447 8445 */
8448 8446 ip1dbg(("ip_rput_dlpi: bind_ack %s\n", ill->ill_name));
8449 8447 DTRACE_PROBE1(ip__rput__dlpi__bind__ack, ill_t *, ill);
8450 8448 ill_nic_event_dispatch(ill, 0, NE_UP, NULL, 0);
8451 8449
8452 8450 if (ill->ill_up_ipifs) {
8453 8451 err = ill_up_ipifs(ill, q, mp1);
8454 8452 if (err == EINPROGRESS) {
8455 8453 freemsg(mp);
8456 8454 return;
8457 8455 }
8458 8456 }
8459 8457
8460 8458 break;
8461 8459
8462 8460 case DL_NOTIFY_IND: {
8463 8461 dl_notify_ind_t *notify = (dl_notify_ind_t *)mp->b_rptr;
8464 8462 uint_t orig_mtu, orig_mc_mtu;
8465 8463
8466 8464 switch (notify->dl_notification) {
8467 8465 case DL_NOTE_PHYS_ADDR:
8468 8466 err = ill_set_phys_addr(ill, mp);
8469 8467 break;
8470 8468
8471 8469 case DL_NOTE_REPLUMB:
8472 8470 /*
8473 8471 * Directly return after calling ill_replumb().
8474 8472 * Note that we should not free mp as it is reused
8475 8473 * in the ill_replumb() function.
8476 8474 */
8477 8475 err = ill_replumb(ill, mp);
8478 8476 return;
8479 8477
8480 8478 case DL_NOTE_FASTPATH_FLUSH:
8481 8479 nce_flush(ill, B_FALSE);
8482 8480 break;
8483 8481
8484 8482 case DL_NOTE_SDU_SIZE:
8485 8483 case DL_NOTE_SDU_SIZE2:
8486 8484 /*
8487 8485 * The dce and fragmentation code can cope with
8488 8486 * this changing while packets are being sent.
8489 8487 * When packets are sent ip_output will discover
8490 8488 * a change.
8491 8489 *
8492 8490 * Change the MTU size of the interface.
8493 8491 */
8494 8492 mutex_enter(&ill->ill_lock);
8495 8493 orig_mtu = ill->ill_mtu;
8496 8494 orig_mc_mtu = ill->ill_mc_mtu;
8497 8495 switch (notify->dl_notification) {
8498 8496 case DL_NOTE_SDU_SIZE:
8499 8497 ill->ill_current_frag =
8500 8498 (uint_t)notify->dl_data;
8501 8499 ill->ill_mc_mtu = (uint_t)notify->dl_data;
8502 8500 break;
8503 8501 case DL_NOTE_SDU_SIZE2:
8504 8502 ill->ill_current_frag =
8505 8503 (uint_t)notify->dl_data1;
8506 8504 ill->ill_mc_mtu = (uint_t)notify->dl_data2;
8507 8505 break;
8508 8506 }
8509 8507 if (ill->ill_current_frag > ill->ill_max_frag)
8510 8508 ill->ill_max_frag = ill->ill_current_frag;
8511 8509
8512 8510 if (!(ill->ill_flags & ILLF_FIXEDMTU)) {
8513 8511 ill->ill_mtu = ill->ill_current_frag;
8514 8512
8515 8513 /*
8516 8514 * If ill_user_mtu was set (via
8517 8515 * SIOCSLIFLNKINFO), clamp ill_mtu at it.
8518 8516 */
8519 8517 if (ill->ill_user_mtu != 0 &&
8520 8518 ill->ill_user_mtu < ill->ill_mtu)
8521 8519 ill->ill_mtu = ill->ill_user_mtu;
8522 8520
8523 8521 if (ill->ill_user_mtu != 0 &&
8524 8522 ill->ill_user_mtu < ill->ill_mc_mtu)
8525 8523 ill->ill_mc_mtu = ill->ill_user_mtu;
8526 8524
8527 8525 if (ill->ill_isv6) {
8528 8526 if (ill->ill_mtu < IPV6_MIN_MTU)
8529 8527 ill->ill_mtu = IPV6_MIN_MTU;
8530 8528 if (ill->ill_mc_mtu < IPV6_MIN_MTU)
8531 8529 ill->ill_mc_mtu = IPV6_MIN_MTU;
8532 8530 } else {
8533 8531 if (ill->ill_mtu < IP_MIN_MTU)
8534 8532 ill->ill_mtu = IP_MIN_MTU;
8535 8533 if (ill->ill_mc_mtu < IP_MIN_MTU)
8536 8534 ill->ill_mc_mtu = IP_MIN_MTU;
8537 8535 }
8538 8536 } else if (ill->ill_mc_mtu > ill->ill_mtu) {
8539 8537 ill->ill_mc_mtu = ill->ill_mtu;
8540 8538 }
8541 8539
8542 8540 mutex_exit(&ill->ill_lock);
8543 8541 /*
8544 8542 * Make sure all dce_generation checks find out
8545 8543 * that ill_mtu/ill_mc_mtu has changed.
8546 8544 */
8547 8545 if (orig_mtu != ill->ill_mtu ||
8548 8546 orig_mc_mtu != ill->ill_mc_mtu) {
8549 8547 dce_increment_all_generations(ill->ill_isv6,
8550 8548 ill->ill_ipst);
8551 8549 }
8552 8550
8553 8551 /*
8554 8552 * Refresh IPMP meta-interface MTU if necessary.
8555 8553 */
8556 8554 if (IS_UNDER_IPMP(ill))
8557 8555 ipmp_illgrp_refresh_mtu(ill->ill_grp);
8558 8556 break;
8559 8557
8560 8558 case DL_NOTE_LINK_UP:
8561 8559 case DL_NOTE_LINK_DOWN: {
8562 8560 /*
8563 8561 * We are writer. ill / phyint / ipsq assocs stable.
8564 8562 * The RUNNING flag reflects the state of the link.
8565 8563 */
8566 8564 phyint_t *phyint = ill->ill_phyint;
8567 8565 uint64_t new_phyint_flags;
8568 8566 boolean_t changed = B_FALSE;
8569 8567 boolean_t went_up;
8570 8568
8571 8569 went_up = notify->dl_notification == DL_NOTE_LINK_UP;
8572 8570 mutex_enter(&phyint->phyint_lock);
8573 8571
8574 8572 new_phyint_flags = went_up ?
8575 8573 phyint->phyint_flags | PHYI_RUNNING :
8576 8574 phyint->phyint_flags & ~PHYI_RUNNING;
8577 8575
8578 8576 if (IS_IPMP(ill)) {
8579 8577 new_phyint_flags = went_up ?
8580 8578 new_phyint_flags & ~PHYI_FAILED :
8581 8579 new_phyint_flags | PHYI_FAILED;
8582 8580 }
8583 8581
8584 8582 if (new_phyint_flags != phyint->phyint_flags) {
8585 8583 phyint->phyint_flags = new_phyint_flags;
8586 8584 changed = B_TRUE;
8587 8585 }
8588 8586 mutex_exit(&phyint->phyint_lock);
8589 8587 /*
8590 8588 * ill_restart_dad handles the DAD restart and routing
8591 8589 * socket notification logic.
8592 8590 */
8593 8591 if (changed) {
8594 8592 ill_restart_dad(phyint->phyint_illv4, went_up);
8595 8593 ill_restart_dad(phyint->phyint_illv6, went_up);
8596 8594 }
8597 8595 break;
8598 8596 }
8599 8597 case DL_NOTE_PROMISC_ON_PHYS: {
8600 8598 phyint_t *phyint = ill->ill_phyint;
8601 8599
8602 8600 mutex_enter(&phyint->phyint_lock);
8603 8601 phyint->phyint_flags |= PHYI_PROMISC;
8604 8602 mutex_exit(&phyint->phyint_lock);
8605 8603 break;
8606 8604 }
8607 8605 case DL_NOTE_PROMISC_OFF_PHYS: {
8608 8606 phyint_t *phyint = ill->ill_phyint;
8609 8607
8610 8608 mutex_enter(&phyint->phyint_lock);
8611 8609 phyint->phyint_flags &= ~PHYI_PROMISC;
8612 8610 mutex_exit(&phyint->phyint_lock);
8613 8611 break;
8614 8612 }
8615 8613 case DL_NOTE_CAPAB_RENEG:
8616 8614 /*
8617 8615 * Something changed on the driver side.
8618 8616 * It wants us to renegotiate the capabilities
8619 8617 * on this ill. One possible cause is the aggregation
8620 8618 * interface under us where a port got added or
8621 8619 * went away.
8622 8620 *
8623 8621 * If the capability negotiation is already done
8624 8622 * or is in progress, reset the capabilities and
8625 8623 * mark the ill's ill_capab_reneg to be B_TRUE,
8626 8624 * so that when the ack comes back, we can start
8627 8625 * the renegotiation process.
8628 8626 *
8629 8627 * Note that if ill_capab_reneg is already B_TRUE
8630 8628 * (ill_dlpi_capab_state is IDS_UNKNOWN in this case),
8631 8629 * the capability resetting request has been sent
8632 8630 * and the renegotiation has not been started yet;
8633 8631 * nothing needs to be done in this case.
8634 8632 */
8635 8633 ipsq_current_start(ipsq, ill->ill_ipif, 0);
8636 8634 ill_capability_reset(ill, B_TRUE);
8637 8635 ipsq_current_finish(ipsq);
8638 8636 break;
8639 8637
8640 8638 case DL_NOTE_ALLOWED_IPS:
8641 8639 ill_set_allowed_ips(ill, mp);
8642 8640 break;
8643 8641 default:
8644 8642 ip0dbg(("ip_rput_dlpi_writer: unknown notification "
8645 8643 "type 0x%x for DL_NOTIFY_IND\n",
8646 8644 notify->dl_notification));
8647 8645 break;
8648 8646 }
8649 8647
8650 8648 /*
8651 8649 * As this is an asynchronous operation, we
8652 8650 * should not call ill_dlpi_done
8653 8651 */
8654 8652 break;
8655 8653 }
8656 8654 case DL_NOTIFY_ACK: {
8657 8655 dl_notify_ack_t *noteack = (dl_notify_ack_t *)mp->b_rptr;
8658 8656
8659 8657 if (noteack->dl_notifications & DL_NOTE_LINK_UP)
8660 8658 ill->ill_note_link = 1;
8661 8659 ill_dlpi_done(ill, DL_NOTIFY_REQ);
8662 8660 break;
8663 8661 }
8664 8662 case DL_PHYS_ADDR_ACK: {
8665 8663 /*
8666 8664 * As part of plumbing the interface via SIOCSLIFNAME,
8667 8665 * ill_dl_phys() will queue a series of DL_PHYS_ADDR_REQs,
8668 8666 * whose answers we receive here. As each answer is received,
8669 8667 * we call ill_dlpi_done() to dispatch the next request as
8670 8668 * we're processing the current one. Once all answers have
8671 8669 * been received, we use ipsq_pending_mp_get() to dequeue the
8672 8670 * outstanding IOCTL and reply to it. (Because ill_dl_phys()
8673 8671 * is invoked from an ill queue, conn_oper_pending_ill is not
8674 8672 * available, but we know the ioctl is pending on ill_wq.)
8675 8673 */
8676 8674 uint_t paddrlen, paddroff;
8677 8675 uint8_t *addr;
8678 8676
8679 8677 paddrreq = ill->ill_phys_addr_pend;
8680 8678 paddrlen = ((dl_phys_addr_ack_t *)mp->b_rptr)->dl_addr_length;
8681 8679 paddroff = ((dl_phys_addr_ack_t *)mp->b_rptr)->dl_addr_offset;
8682 8680 addr = mp->b_rptr + paddroff;
8683 8681
8684 8682 ill_dlpi_done(ill, DL_PHYS_ADDR_REQ);
8685 8683 if (paddrreq == DL_IPV6_TOKEN) {
8686 8684 /*
8687 8685 * bcopy to low-order bits of ill_token
8688 8686 *
8689 8687 * XXX Temporary hack - currently, all known tokens
8690 8688 * are 64 bits, so I'll cheat for the moment.
8691 8689 */
8692 8690 bcopy(addr, &ill->ill_token.s6_addr32[2], paddrlen);
8693 8691 ill->ill_token_length = paddrlen;
8694 8692 break;
8695 8693 } else if (paddrreq == DL_IPV6_LINK_LAYER_ADDR) {
8696 8694 ASSERT(ill->ill_nd_lla_mp == NULL);
8697 8695 ill_set_ndmp(ill, mp, paddroff, paddrlen);
8698 8696 mp = NULL;
8699 8697 break;
8700 8698 } else if (paddrreq == DL_CURR_DEST_ADDR) {
8701 8699 ASSERT(ill->ill_dest_addr_mp == NULL);
8702 8700 ill->ill_dest_addr_mp = mp;
8703 8701 ill->ill_dest_addr = addr;
8704 8702 mp = NULL;
8705 8703 if (ill->ill_isv6) {
8706 8704 ill_setdesttoken(ill);
8707 8705 ipif_setdestlinklocal(ill->ill_ipif);
8708 8706 }
8709 8707 break;
8710 8708 }
8711 8709
8712 8710 ASSERT(paddrreq == DL_CURR_PHYS_ADDR);
8713 8711 ASSERT(ill->ill_phys_addr_mp == NULL);
8714 8712 if (!ill->ill_ifname_pending)
8715 8713 break;
8716 8714 ill->ill_ifname_pending = 0;
8717 8715 if (!ioctl_aborted)
8718 8716 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8719 8717 if (mp1 != NULL) {
8720 8718 ASSERT(connp == NULL);
8721 8719 q = ill->ill_wq;
8722 8720 }
8723 8721 /*
8724 8722 * If any error acks received during the plumbing sequence,
8725 8723 * ill_ifname_pending_err will be set. Break out and send up
8726 8724 * the error to the pending ioctl.
8727 8725 */
8728 8726 if (ill->ill_ifname_pending_err != 0) {
8729 8727 err = ill->ill_ifname_pending_err;
8730 8728 ill->ill_ifname_pending_err = 0;
8731 8729 break;
8732 8730 }
8733 8731
8734 8732 ill->ill_phys_addr_mp = mp;
8735 8733 ill->ill_phys_addr = (paddrlen == 0 ? NULL : addr);
8736 8734 mp = NULL;
8737 8735
8738 8736 /*
8739 8737 * If paddrlen or ill_phys_addr_length is zero, the DLPI
8740 8738 * provider doesn't support physical addresses. We check both
8741 8739 * paddrlen and ill_phys_addr_length because sppp (PPP) does
8742 8740 * not have physical addresses, but historically adversises a
8743 8741 * physical address length of 0 in its DL_INFO_ACK, but 6 in
8744 8742 * its DL_PHYS_ADDR_ACK.
8745 8743 */
8746 8744 if (paddrlen == 0 || ill->ill_phys_addr_length == 0) {
8747 8745 ill->ill_phys_addr = NULL;
8748 8746 } else if (paddrlen != ill->ill_phys_addr_length) {
8749 8747 ip0dbg(("DL_PHYS_ADDR_ACK: got addrlen %d, expected %d",
8750 8748 paddrlen, ill->ill_phys_addr_length));
8751 8749 err = EINVAL;
8752 8750 break;
8753 8751 }
8754 8752
8755 8753 if (ill->ill_nd_lla_mp == NULL) {
8756 8754 if ((mp_hw = copyb(ill->ill_phys_addr_mp)) == NULL) {
8757 8755 err = ENOMEM;
8758 8756 break;
8759 8757 }
8760 8758 ill_set_ndmp(ill, mp_hw, paddroff, paddrlen);
8761 8759 }
8762 8760
8763 8761 if (ill->ill_isv6) {
8764 8762 ill_setdefaulttoken(ill);
8765 8763 ipif_setlinklocal(ill->ill_ipif);
8766 8764 }
8767 8765 break;
8768 8766 }
8769 8767 case DL_OK_ACK:
8770 8768 ip2dbg(("DL_OK_ACK %s (0x%x)\n",
8771 8769 dl_primstr((int)dloa->dl_correct_primitive),
8772 8770 dloa->dl_correct_primitive));
8773 8771 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer ok",
8774 8772 char *, dl_primstr(dloa->dl_correct_primitive),
8775 8773 ill_t *, ill);
8776 8774
8777 8775 switch (dloa->dl_correct_primitive) {
8778 8776 case DL_ENABMULTI_REQ:
8779 8777 case DL_DISABMULTI_REQ:
8780 8778 ill_dlpi_done(ill, dloa->dl_correct_primitive);
8781 8779 break;
8782 8780 case DL_PROMISCON_REQ:
8783 8781 case DL_PROMISCOFF_REQ:
8784 8782 case DL_UNBIND_REQ:
8785 8783 case DL_ATTACH_REQ:
8786 8784 ill_dlpi_done(ill, dloa->dl_correct_primitive);
8787 8785 break;
8788 8786 }
8789 8787 break;
8790 8788 default:
8791 8789 break;
8792 8790 }
8793 8791
8794 8792 freemsg(mp);
8795 8793 if (mp1 == NULL)
8796 8794 return;
8797 8795
8798 8796 /*
8799 8797 * The operation must complete without EINPROGRESS since
8800 8798 * ipsq_pending_mp_get() has removed the mblk (mp1). Otherwise,
8801 8799 * the operation will be stuck forever inside the IPSQ.
8802 8800 */
8803 8801 ASSERT(err != EINPROGRESS);
8804 8802
8805 8803 DTRACE_PROBE4(ipif__ioctl, char *, "ip_rput_dlpi_writer finish",
8806 8804 int, ipsq->ipsq_xop->ipx_current_ioctl, ill_t *, ill,
8807 8805 ipif_t *, NULL);
8808 8806
8809 8807 switch (ipsq->ipsq_xop->ipx_current_ioctl) {
8810 8808 case 0:
8811 8809 ipsq_current_finish(ipsq);
8812 8810 break;
8813 8811
8814 8812 case SIOCSLIFNAME:
8815 8813 case IF_UNITSEL: {
8816 8814 ill_t *ill_other = ILL_OTHER(ill);
8817 8815
8818 8816 /*
8819 8817 * If SIOCSLIFNAME or IF_UNITSEL is about to succeed, and the
8820 8818 * ill has a peer which is in an IPMP group, then place ill
8821 8819 * into the same group. One catch: although ifconfig plumbs
8822 8820 * the appropriate IPMP meta-interface prior to plumbing this
8823 8821 * ill, it is possible for multiple ifconfig applications to
8824 8822 * race (or for another application to adjust plumbing), in
8825 8823 * which case the IPMP meta-interface we need will be missing.
8826 8824 * If so, kick the phyint out of the group.
8827 8825 */
8828 8826 if (err == 0 && ill_other != NULL && IS_UNDER_IPMP(ill_other)) {
8829 8827 ipmp_grp_t *grp = ill->ill_phyint->phyint_grp;
8830 8828 ipmp_illgrp_t *illg;
8831 8829
8832 8830 illg = ill->ill_isv6 ? grp->gr_v6 : grp->gr_v4;
8833 8831 if (illg == NULL)
8834 8832 ipmp_phyint_leave_grp(ill->ill_phyint);
8835 8833 else
8836 8834 ipmp_ill_join_illgrp(ill, illg);
8837 8835 }
8838 8836
8839 8837 if (ipsq->ipsq_xop->ipx_current_ioctl == IF_UNITSEL)
8840 8838 ip_ioctl_finish(q, mp1, err, NO_COPYOUT, ipsq);
8841 8839 else
8842 8840 ip_ioctl_finish(q, mp1, err, COPYOUT, ipsq);
8843 8841 break;
8844 8842 }
8845 8843 case SIOCLIFADDIF:
8846 8844 ip_ioctl_finish(q, mp1, err, COPYOUT, ipsq);
8847 8845 break;
8848 8846
8849 8847 default:
8850 8848 ip_ioctl_finish(q, mp1, err, NO_COPYOUT, ipsq);
8851 8849 break;
8852 8850 }
8853 8851 }
8854 8852
8855 8853 /*
8856 8854 * ip_rput_other is called by ip_rput to handle messages modifying the global
8857 8855 * state in IP. If 'ipsq' is non-NULL, caller is writer on it.
8858 8856 */
8859 8857 /* ARGSUSED */
8860 8858 void
8861 8859 ip_rput_other(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
8862 8860 {
8863 8861 ill_t *ill = q->q_ptr;
8864 8862 struct iocblk *iocp;
8865 8863
8866 8864 ip1dbg(("ip_rput_other "));
8867 8865 if (ipsq != NULL) {
8868 8866 ASSERT(IAM_WRITER_IPSQ(ipsq));
8869 8867 ASSERT(ipsq->ipsq_xop ==
8870 8868 ill->ill_phyint->phyint_ipsq->ipsq_xop);
8871 8869 }
8872 8870
8873 8871 switch (mp->b_datap->db_type) {
8874 8872 case M_ERROR:
8875 8873 case M_HANGUP:
8876 8874 /*
8877 8875 * The device has a problem. We force the ILL down. It can
8878 8876 * be brought up again manually using SIOCSIFFLAGS (via
8879 8877 * ifconfig or equivalent).
8880 8878 */
8881 8879 ASSERT(ipsq != NULL);
8882 8880 if (mp->b_rptr < mp->b_wptr)
8883 8881 ill->ill_error = (int)(*mp->b_rptr & 0xFF);
8884 8882 if (ill->ill_error == 0)
8885 8883 ill->ill_error = ENXIO;
8886 8884 if (!ill_down_start(q, mp))
8887 8885 return;
8888 8886 ipif_all_down_tail(ipsq, q, mp, NULL);
8889 8887 break;
8890 8888 case M_IOCNAK: {
8891 8889 iocp = (struct iocblk *)mp->b_rptr;
8892 8890
8893 8891 ASSERT(iocp->ioc_cmd == DL_IOC_HDR_INFO);
8894 8892 /*
8895 8893 * If this was the first attempt, turn off the fastpath
8896 8894 * probing.
8897 8895 */
8898 8896 mutex_enter(&ill->ill_lock);
8899 8897 if (ill->ill_dlpi_fastpath_state == IDS_INPROGRESS) {
8900 8898 ill->ill_dlpi_fastpath_state = IDS_FAILED;
8901 8899 mutex_exit(&ill->ill_lock);
8902 8900 /*
8903 8901 * don't flush the nce_t entries: we use them
8904 8902 * as an index to the ncec itself.
8905 8903 */
8906 8904 ip1dbg(("ip_rput: DLPI fastpath off on interface %s\n",
8907 8905 ill->ill_name));
8908 8906 } else {
8909 8907 mutex_exit(&ill->ill_lock);
8910 8908 }
8911 8909 freemsg(mp);
8912 8910 break;
8913 8911 }
8914 8912 default:
8915 8913 ASSERT(0);
8916 8914 break;
8917 8915 }
8918 8916 }
8919 8917
8920 8918 /*
8921 8919 * Update any source route, record route or timestamp options
8922 8920 * When it fails it has consumed the message and BUMPed the MIB.
8923 8921 */
8924 8922 boolean_t
8925 8923 ip_forward_options(mblk_t *mp, ipha_t *ipha, ill_t *dst_ill,
8926 8924 ip_recv_attr_t *ira)
8927 8925 {
8928 8926 ipoptp_t opts;
8929 8927 uchar_t *opt;
8930 8928 uint8_t optval;
8931 8929 uint8_t optlen;
8932 8930 ipaddr_t dst;
8933 8931 ipaddr_t ifaddr;
8934 8932 uint32_t ts;
8935 8933 timestruc_t now;
8936 8934 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
8937 8935
8938 8936 ip2dbg(("ip_forward_options\n"));
8939 8937 dst = ipha->ipha_dst;
8940 8938 for (optval = ipoptp_first(&opts, ipha);
8941 8939 optval != IPOPT_EOL;
8942 8940 optval = ipoptp_next(&opts)) {
8943 8941 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
8944 8942 opt = opts.ipoptp_cur;
8945 8943 optlen = opts.ipoptp_len;
8946 8944 ip2dbg(("ip_forward_options: opt %d, len %d\n",
8947 8945 optval, opts.ipoptp_len));
8948 8946 switch (optval) {
8949 8947 uint32_t off;
8950 8948 case IPOPT_SSRR:
8951 8949 case IPOPT_LSRR:
8952 8950 /* Check if adminstratively disabled */
8953 8951 if (!ipst->ips_ip_forward_src_routed) {
8954 8952 BUMP_MIB(dst_ill->ill_ip_mib,
8955 8953 ipIfStatsForwProhibits);
8956 8954 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED",
8957 8955 mp, dst_ill);
8958 8956 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED,
8959 8957 ira);
8960 8958 return (B_FALSE);
8961 8959 }
8962 8960 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
8963 8961 /*
8964 8962 * Must be partial since ip_input_options
8965 8963 * checked for strict.
8966 8964 */
8967 8965 break;
8968 8966 }
8969 8967 off = opt[IPOPT_OFFSET];
8970 8968 off--;
8971 8969 redo_srr:
8972 8970 if (optlen < IP_ADDR_LEN ||
8973 8971 off > optlen - IP_ADDR_LEN) {
8974 8972 /* End of source route */
8975 8973 ip1dbg((
8976 8974 "ip_forward_options: end of SR\n"));
8977 8975 break;
8978 8976 }
8979 8977 /* Pick a reasonable address on the outbound if */
8980 8978 ASSERT(dst_ill != NULL);
8981 8979 if (ip_select_source_v4(dst_ill, INADDR_ANY, dst,
8982 8980 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
8983 8981 NULL) != 0) {
8984 8982 /* No source! Shouldn't happen */
8985 8983 ifaddr = INADDR_ANY;
8986 8984 }
8987 8985 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
8988 8986 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
8989 8987 ip1dbg(("ip_forward_options: next hop 0x%x\n",
8990 8988 ntohl(dst)));
8991 8989
8992 8990 /*
8993 8991 * Check if our address is present more than
8994 8992 * once as consecutive hops in source route.
8995 8993 */
8996 8994 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
8997 8995 off += IP_ADDR_LEN;
8998 8996 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
8999 8997 goto redo_srr;
9000 8998 }
9001 8999 ipha->ipha_dst = dst;
9002 9000 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9003 9001 break;
9004 9002 case IPOPT_RR:
9005 9003 off = opt[IPOPT_OFFSET];
9006 9004 off--;
9007 9005 if (optlen < IP_ADDR_LEN ||
9008 9006 off > optlen - IP_ADDR_LEN) {
9009 9007 /* No more room - ignore */
9010 9008 ip1dbg((
9011 9009 "ip_forward_options: end of RR\n"));
9012 9010 break;
9013 9011 }
9014 9012 /* Pick a reasonable address on the outbound if */
9015 9013 ASSERT(dst_ill != NULL);
9016 9014 if (ip_select_source_v4(dst_ill, INADDR_ANY, dst,
9017 9015 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9018 9016 NULL) != 0) {
9019 9017 /* No source! Shouldn't happen */
9020 9018 ifaddr = INADDR_ANY;
9021 9019 }
9022 9020 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9023 9021 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9024 9022 break;
9025 9023 case IPOPT_TS:
9026 9024 /* Insert timestamp if there is room */
9027 9025 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9028 9026 case IPOPT_TS_TSONLY:
9029 9027 off = IPOPT_TS_TIMELEN;
9030 9028 break;
9031 9029 case IPOPT_TS_PRESPEC:
9032 9030 case IPOPT_TS_PRESPEC_RFC791:
9033 9031 /* Verify that the address matched */
9034 9032 off = opt[IPOPT_OFFSET] - 1;
9035 9033 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9036 9034 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9037 9035 /* Not for us */
9038 9036 break;
9039 9037 }
9040 9038 /* FALLTHROUGH */
9041 9039 case IPOPT_TS_TSANDADDR:
9042 9040 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9043 9041 break;
9044 9042 default:
9045 9043 /*
9046 9044 * ip_*put_options should have already
9047 9045 * dropped this packet.
9048 9046 */
9049 9047 cmn_err(CE_PANIC, "ip_forward_options: "
9050 9048 "unknown IT - bug in ip_input_options?\n");
9051 9049 }
9052 9050 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
9053 9051 /* Increase overflow counter */
9054 9052 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
9055 9053 opt[IPOPT_POS_OV_FLG] =
9056 9054 (uint8_t)((opt[IPOPT_POS_OV_FLG] & 0x0F) |
9057 9055 (off << 4));
9058 9056 break;
9059 9057 }
9060 9058 off = opt[IPOPT_OFFSET] - 1;
9061 9059 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9062 9060 case IPOPT_TS_PRESPEC:
9063 9061 case IPOPT_TS_PRESPEC_RFC791:
9064 9062 case IPOPT_TS_TSANDADDR:
9065 9063 /* Pick a reasonable addr on the outbound if */
9066 9064 ASSERT(dst_ill != NULL);
9067 9065 if (ip_select_source_v4(dst_ill, INADDR_ANY,
9068 9066 dst, INADDR_ANY, ALL_ZONES, ipst, &ifaddr,
9069 9067 NULL, NULL) != 0) {
9070 9068 /* No source! Shouldn't happen */
9071 9069 ifaddr = INADDR_ANY;
9072 9070 }
9073 9071 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9074 9072 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9075 9073 /* FALLTHROUGH */
9076 9074 case IPOPT_TS_TSONLY:
9077 9075 off = opt[IPOPT_OFFSET] - 1;
9078 9076 /* Compute # of milliseconds since midnight */
9079 9077 gethrestime(&now);
9080 9078 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
9081 9079 NSEC2MSEC(now.tv_nsec);
9082 9080 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
9083 9081 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
9084 9082 break;
9085 9083 }
9086 9084 break;
9087 9085 }
9088 9086 }
9089 9087 return (B_TRUE);
9090 9088 }
9091 9089
9092 9090 /*
9093 9091 * Call ill_frag_timeout to do garbage collection. ill_frag_timeout
9094 9092 * returns 'true' if there are still fragments left on the queue, in
9095 9093 * which case we restart the timer.
9096 9094 */
9097 9095 void
9098 9096 ill_frag_timer(void *arg)
9099 9097 {
9100 9098 ill_t *ill = (ill_t *)arg;
9101 9099 boolean_t frag_pending;
9102 9100 ip_stack_t *ipst = ill->ill_ipst;
9103 9101 time_t timeout;
9104 9102
9105 9103 mutex_enter(&ill->ill_lock);
9106 9104 ASSERT(!ill->ill_fragtimer_executing);
9107 9105 if (ill->ill_state_flags & ILL_CONDEMNED) {
9108 9106 ill->ill_frag_timer_id = 0;
9109 9107 mutex_exit(&ill->ill_lock);
9110 9108 return;
9111 9109 }
9112 9110 ill->ill_fragtimer_executing = 1;
9113 9111 mutex_exit(&ill->ill_lock);
9114 9112
9115 9113 timeout = (ill->ill_isv6 ? ipst->ips_ipv6_reassembly_timeout :
9116 9114 ipst->ips_ip_reassembly_timeout);
9117 9115
9118 9116 frag_pending = ill_frag_timeout(ill, timeout);
9119 9117
9120 9118 /*
9121 9119 * Restart the timer, if we have fragments pending or if someone
9122 9120 * wanted us to be scheduled again.
9123 9121 */
9124 9122 mutex_enter(&ill->ill_lock);
9125 9123 ill->ill_fragtimer_executing = 0;
9126 9124 ill->ill_frag_timer_id = 0;
9127 9125 if (frag_pending || ill->ill_fragtimer_needrestart)
9128 9126 ill_frag_timer_start(ill);
9129 9127 mutex_exit(&ill->ill_lock);
9130 9128 }
9131 9129
9132 9130 void
9133 9131 ill_frag_timer_start(ill_t *ill)
9134 9132 {
9135 9133 ip_stack_t *ipst = ill->ill_ipst;
9136 9134 clock_t timeo_ms;
9137 9135
9138 9136 ASSERT(MUTEX_HELD(&ill->ill_lock));
9139 9137
9140 9138 /* If the ill is closing or opening don't proceed */
9141 9139 if (ill->ill_state_flags & ILL_CONDEMNED)
9142 9140 return;
9143 9141
9144 9142 if (ill->ill_fragtimer_executing) {
9145 9143 /*
9146 9144 * ill_frag_timer is currently executing. Just record the
9147 9145 * the fact that we want the timer to be restarted.
9148 9146 * ill_frag_timer will post a timeout before it returns,
9149 9147 * ensuring it will be called again.
9150 9148 */
9151 9149 ill->ill_fragtimer_needrestart = 1;
9152 9150 return;
9153 9151 }
9154 9152
9155 9153 if (ill->ill_frag_timer_id == 0) {
9156 9154 timeo_ms = (ill->ill_isv6 ? ipst->ips_ipv6_reassembly_timeout :
9157 9155 ipst->ips_ip_reassembly_timeout) * SECONDS;
9158 9156
9159 9157 /*
9160 9158 * The timer is neither running nor is the timeout handler
9161 9159 * executing. Post a timeout so that ill_frag_timer will be
9162 9160 * called
9163 9161 */
9164 9162 ill->ill_frag_timer_id = timeout(ill_frag_timer, ill,
9165 9163 MSEC_TO_TICK(timeo_ms >> 1));
9166 9164 ill->ill_fragtimer_needrestart = 0;
9167 9165 }
9168 9166 }
9169 9167
9170 9168 /*
9171 9169 * Update any source route, record route or timestamp options.
9172 9170 * Check that we are at end of strict source route.
9173 9171 * The options have already been checked for sanity in ip_input_options().
9174 9172 */
9175 9173 boolean_t
9176 9174 ip_input_local_options(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
9177 9175 {
9178 9176 ipoptp_t opts;
9179 9177 uchar_t *opt;
9180 9178 uint8_t optval;
9181 9179 uint8_t optlen;
9182 9180 ipaddr_t dst;
9183 9181 ipaddr_t ifaddr;
9184 9182 uint32_t ts;
9185 9183 timestruc_t now;
9186 9184 ill_t *ill = ira->ira_ill;
9187 9185 ip_stack_t *ipst = ill->ill_ipst;
9188 9186
9189 9187 ip2dbg(("ip_input_local_options\n"));
9190 9188
9191 9189 for (optval = ipoptp_first(&opts, ipha);
9192 9190 optval != IPOPT_EOL;
9193 9191 optval = ipoptp_next(&opts)) {
9194 9192 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
9195 9193 opt = opts.ipoptp_cur;
9196 9194 optlen = opts.ipoptp_len;
9197 9195 ip2dbg(("ip_input_local_options: opt %d, len %d\n",
9198 9196 optval, optlen));
9199 9197 switch (optval) {
9200 9198 uint32_t off;
9201 9199 case IPOPT_SSRR:
9202 9200 case IPOPT_LSRR:
9203 9201 off = opt[IPOPT_OFFSET];
9204 9202 off--;
9205 9203 if (optlen < IP_ADDR_LEN ||
9206 9204 off > optlen - IP_ADDR_LEN) {
9207 9205 /* End of source route */
9208 9206 ip1dbg(("ip_input_local_options: end of SR\n"));
9209 9207 break;
9210 9208 }
9211 9209 /*
9212 9210 * This will only happen if two consecutive entries
9213 9211 * in the source route contains our address or if
9214 9212 * it is a packet with a loose source route which
9215 9213 * reaches us before consuming the whole source route
9216 9214 */
9217 9215 ip1dbg(("ip_input_local_options: not end of SR\n"));
9218 9216 if (optval == IPOPT_SSRR) {
9219 9217 goto bad_src_route;
9220 9218 }
9221 9219 /*
9222 9220 * Hack: instead of dropping the packet truncate the
9223 9221 * source route to what has been used by filling the
9224 9222 * rest with IPOPT_NOP.
9225 9223 */
9226 9224 opt[IPOPT_OLEN] = (uint8_t)off;
9227 9225 while (off < optlen) {
9228 9226 opt[off++] = IPOPT_NOP;
9229 9227 }
9230 9228 break;
9231 9229 case IPOPT_RR:
9232 9230 off = opt[IPOPT_OFFSET];
9233 9231 off--;
9234 9232 if (optlen < IP_ADDR_LEN ||
9235 9233 off > optlen - IP_ADDR_LEN) {
9236 9234 /* No more room - ignore */
9237 9235 ip1dbg((
9238 9236 "ip_input_local_options: end of RR\n"));
9239 9237 break;
9240 9238 }
9241 9239 /* Pick a reasonable address on the outbound if */
9242 9240 if (ip_select_source_v4(ill, INADDR_ANY, ipha->ipha_dst,
9243 9241 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9244 9242 NULL) != 0) {
9245 9243 /* No source! Shouldn't happen */
9246 9244 ifaddr = INADDR_ANY;
9247 9245 }
9248 9246 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9249 9247 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9250 9248 break;
9251 9249 case IPOPT_TS:
9252 9250 /* Insert timestamp if there is romm */
9253 9251 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9254 9252 case IPOPT_TS_TSONLY:
9255 9253 off = IPOPT_TS_TIMELEN;
9256 9254 break;
9257 9255 case IPOPT_TS_PRESPEC:
9258 9256 case IPOPT_TS_PRESPEC_RFC791:
9259 9257 /* Verify that the address matched */
9260 9258 off = opt[IPOPT_OFFSET] - 1;
9261 9259 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9262 9260 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9263 9261 /* Not for us */
9264 9262 break;
9265 9263 }
9266 9264 /* FALLTHROUGH */
9267 9265 case IPOPT_TS_TSANDADDR:
9268 9266 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9269 9267 break;
9270 9268 default:
9271 9269 /*
9272 9270 * ip_*put_options should have already
9273 9271 * dropped this packet.
9274 9272 */
9275 9273 cmn_err(CE_PANIC, "ip_input_local_options: "
9276 9274 "unknown IT - bug in ip_input_options?\n");
9277 9275 }
9278 9276 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
9279 9277 /* Increase overflow counter */
9280 9278 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
9281 9279 opt[IPOPT_POS_OV_FLG] =
9282 9280 (uint8_t)((opt[IPOPT_POS_OV_FLG] & 0x0F) |
9283 9281 (off << 4));
9284 9282 break;
9285 9283 }
9286 9284 off = opt[IPOPT_OFFSET] - 1;
9287 9285 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9288 9286 case IPOPT_TS_PRESPEC:
9289 9287 case IPOPT_TS_PRESPEC_RFC791:
9290 9288 case IPOPT_TS_TSANDADDR:
9291 9289 /* Pick a reasonable addr on the outbound if */
9292 9290 if (ip_select_source_v4(ill, INADDR_ANY,
9293 9291 ipha->ipha_dst, INADDR_ANY, ALL_ZONES, ipst,
9294 9292 &ifaddr, NULL, NULL) != 0) {
9295 9293 /* No source! Shouldn't happen */
9296 9294 ifaddr = INADDR_ANY;
9297 9295 }
9298 9296 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9299 9297 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9300 9298 /* FALLTHROUGH */
9301 9299 case IPOPT_TS_TSONLY:
9302 9300 off = opt[IPOPT_OFFSET] - 1;
9303 9301 /* Compute # of milliseconds since midnight */
9304 9302 gethrestime(&now);
9305 9303 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
9306 9304 NSEC2MSEC(now.tv_nsec);
9307 9305 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
9308 9306 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
9309 9307 break;
9310 9308 }
9311 9309 break;
9312 9310 }
9313 9311 }
9314 9312 return (B_TRUE);
9315 9313
9316 9314 bad_src_route:
9317 9315 /* make sure we clear any indication of a hardware checksum */
9318 9316 DB_CKSUMFLAGS(mp) = 0;
9319 9317 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ill);
9320 9318 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
9321 9319 return (B_FALSE);
9322 9320
9323 9321 }
9324 9322
9325 9323 /*
9326 9324 * Process IP options in an inbound packet. Always returns the nexthop.
9327 9325 * Normally this is the passed in nexthop, but if there is an option
9328 9326 * that effects the nexthop (such as a source route) that will be returned.
9329 9327 * Sets *errorp if there is an error, in which case an ICMP error has been sent
9330 9328 * and mp freed.
9331 9329 */
9332 9330 ipaddr_t
9333 9331 ip_input_options(ipha_t *ipha, ipaddr_t dst, mblk_t *mp,
9334 9332 ip_recv_attr_t *ira, int *errorp)
9335 9333 {
9336 9334 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
9337 9335 ipoptp_t opts;
9338 9336 uchar_t *opt;
9339 9337 uint8_t optval;
9340 9338 uint8_t optlen;
9341 9339 intptr_t code = 0;
9342 9340 ire_t *ire;
9343 9341
9344 9342 ip2dbg(("ip_input_options\n"));
9345 9343 *errorp = 0;
9346 9344 for (optval = ipoptp_first(&opts, ipha);
9347 9345 optval != IPOPT_EOL;
9348 9346 optval = ipoptp_next(&opts)) {
9349 9347 opt = opts.ipoptp_cur;
9350 9348 optlen = opts.ipoptp_len;
9351 9349 ip2dbg(("ip_input_options: opt %d, len %d\n",
9352 9350 optval, optlen));
9353 9351 /*
9354 9352 * Note: we need to verify the checksum before we
9355 9353 * modify anything thus this routine only extracts the next
9356 9354 * hop dst from any source route.
9357 9355 */
9358 9356 switch (optval) {
9359 9357 uint32_t off;
9360 9358 case IPOPT_SSRR:
9361 9359 case IPOPT_LSRR:
9362 9360 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9363 9361 if (optval == IPOPT_SSRR) {
9364 9362 ip1dbg(("ip_input_options: not next"
9365 9363 " strict source route 0x%x\n",
9366 9364 ntohl(dst)));
9367 9365 code = (char *)&ipha->ipha_dst -
9368 9366 (char *)ipha;
9369 9367 goto param_prob; /* RouterReq's */
9370 9368 }
9371 9369 ip2dbg(("ip_input_options: "
9372 9370 "not next source route 0x%x\n",
9373 9371 ntohl(dst)));
9374 9372 break;
9375 9373 }
9376 9374
9377 9375 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9378 9376 ip1dbg((
9379 9377 "ip_input_options: bad option offset\n"));
9380 9378 code = (char *)&opt[IPOPT_OLEN] -
9381 9379 (char *)ipha;
9382 9380 goto param_prob;
9383 9381 }
9384 9382 off = opt[IPOPT_OFFSET];
9385 9383 off--;
9386 9384 redo_srr:
9387 9385 if (optlen < IP_ADDR_LEN ||
9388 9386 off > optlen - IP_ADDR_LEN) {
9389 9387 /* End of source route */
9390 9388 ip1dbg(("ip_input_options: end of SR\n"));
9391 9389 break;
9392 9390 }
9393 9391 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9394 9392 ip1dbg(("ip_input_options: next hop 0x%x\n",
9395 9393 ntohl(dst)));
9396 9394
9397 9395 /*
9398 9396 * Check if our address is present more than
9399 9397 * once as consecutive hops in source route.
9400 9398 * XXX verify per-interface ip_forwarding
9401 9399 * for source route?
9402 9400 */
9403 9401 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
9404 9402 off += IP_ADDR_LEN;
9405 9403 goto redo_srr;
9406 9404 }
9407 9405
9408 9406 if (dst == htonl(INADDR_LOOPBACK)) {
9409 9407 ip1dbg(("ip_input_options: loopback addr in "
9410 9408 "source route!\n"));
9411 9409 goto bad_src_route;
9412 9410 }
9413 9411 /*
9414 9412 * For strict: verify that dst is directly
9415 9413 * reachable.
9416 9414 */
9417 9415 if (optval == IPOPT_SSRR) {
9418 9416 ire = ire_ftable_lookup_v4(dst, 0, 0,
9419 9417 IRE_INTERFACE, NULL, ALL_ZONES,
9420 9418 ira->ira_tsl,
9421 9419 MATCH_IRE_TYPE | MATCH_IRE_SECATTR, 0, ipst,
9422 9420 NULL);
9423 9421 if (ire == NULL) {
9424 9422 ip1dbg(("ip_input_options: SSRR not "
9425 9423 "directly reachable: 0x%x\n",
9426 9424 ntohl(dst)));
9427 9425 goto bad_src_route;
9428 9426 }
9429 9427 ire_refrele(ire);
9430 9428 }
9431 9429 /*
9432 9430 * Defer update of the offset and the record route
9433 9431 * until the packet is forwarded.
9434 9432 */
9435 9433 break;
9436 9434 case IPOPT_RR:
9437 9435 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9438 9436 ip1dbg((
9439 9437 "ip_input_options: bad option offset\n"));
9440 9438 code = (char *)&opt[IPOPT_OLEN] -
9441 9439 (char *)ipha;
9442 9440 goto param_prob;
9443 9441 }
9444 9442 break;
9445 9443 case IPOPT_TS:
9446 9444 /*
9447 9445 * Verify that length >= 5 and that there is either
9448 9446 * room for another timestamp or that the overflow
9449 9447 * counter is not maxed out.
9450 9448 */
9451 9449 code = (char *)&opt[IPOPT_OLEN] - (char *)ipha;
9452 9450 if (optlen < IPOPT_MINLEN_IT) {
9453 9451 goto param_prob;
9454 9452 }
9455 9453 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9456 9454 ip1dbg((
9457 9455 "ip_input_options: bad option offset\n"));
9458 9456 code = (char *)&opt[IPOPT_OFFSET] -
9459 9457 (char *)ipha;
9460 9458 goto param_prob;
9461 9459 }
9462 9460 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9463 9461 case IPOPT_TS_TSONLY:
9464 9462 off = IPOPT_TS_TIMELEN;
9465 9463 break;
9466 9464 case IPOPT_TS_TSANDADDR:
9467 9465 case IPOPT_TS_PRESPEC:
9468 9466 case IPOPT_TS_PRESPEC_RFC791:
9469 9467 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9470 9468 break;
9471 9469 default:
9472 9470 code = (char *)&opt[IPOPT_POS_OV_FLG] -
9473 9471 (char *)ipha;
9474 9472 goto param_prob;
9475 9473 }
9476 9474 if (opt[IPOPT_OFFSET] - 1 + off > optlen &&
9477 9475 (opt[IPOPT_POS_OV_FLG] & 0xF0) == 0xF0) {
9478 9476 /*
9479 9477 * No room and the overflow counter is 15
9480 9478 * already.
9481 9479 */
9482 9480 goto param_prob;
9483 9481 }
9484 9482 break;
9485 9483 }
9486 9484 }
9487 9485
9488 9486 if ((opts.ipoptp_flags & IPOPTP_ERROR) == 0) {
9489 9487 return (dst);
9490 9488 }
9491 9489
9492 9490 ip1dbg(("ip_input_options: error processing IP options."));
9493 9491 code = (char *)&opt[IPOPT_OFFSET] - (char *)ipha;
9494 9492
9495 9493 param_prob:
9496 9494 /* make sure we clear any indication of a hardware checksum */
9497 9495 DB_CKSUMFLAGS(mp) = 0;
9498 9496 ip_drop_input("ICMP_PARAM_PROBLEM", mp, ira->ira_ill);
9499 9497 icmp_param_problem(mp, (uint8_t)code, ira);
9500 9498 *errorp = -1;
9501 9499 return (dst);
9502 9500
9503 9501 bad_src_route:
9504 9502 /* make sure we clear any indication of a hardware checksum */
9505 9503 DB_CKSUMFLAGS(mp) = 0;
9506 9504 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ira->ira_ill);
9507 9505 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
9508 9506 *errorp = -1;
9509 9507 return (dst);
9510 9508 }
9511 9509
9512 9510 /*
9513 9511 * IP & ICMP info in >=14 msg's ...
9514 9512 * - ip fixed part (mib2_ip_t)
9515 9513 * - icmp fixed part (mib2_icmp_t)
9516 9514 * - ipAddrEntryTable (ip 20) all IPv4 ipifs
9517 9515 * - ipRouteEntryTable (ip 21) all IPv4 IREs
9518 9516 * - ipNetToMediaEntryTable (ip 22) all IPv4 Neighbor Cache entries
9519 9517 * - ipRouteAttributeTable (ip 102) labeled routes
9520 9518 * - ip multicast membership (ip_member_t)
9521 9519 * - ip multicast source filtering (ip_grpsrc_t)
9522 9520 * - igmp fixed part (struct igmpstat)
9523 9521 * - multicast routing stats (struct mrtstat)
9524 9522 * - multicast routing vifs (array of struct vifctl)
9525 9523 * - multicast routing routes (array of struct mfcctl)
9526 9524 * - ip6 fixed part (mib2_ipv6IfStatsEntry_t)
9527 9525 * One per ill plus one generic
9528 9526 * - icmp6 fixed part (mib2_ipv6IfIcmpEntry_t)
9529 9527 * One per ill plus one generic
9530 9528 * - ipv6RouteEntry all IPv6 IREs
9531 9529 * - ipv6RouteAttributeTable (ip6 102) labeled routes
9532 9530 * - ipv6NetToMediaEntry all IPv6 Neighbor Cache entries
9533 9531 * - ipv6AddrEntry all IPv6 ipifs
9534 9532 * - ipv6 multicast membership (ipv6_member_t)
9535 9533 * - ipv6 multicast source filtering (ipv6_grpsrc_t)
9536 9534 *
9537 9535 * NOTE: original mpctl is copied for msg's 2..N, since its ctl part is
9538 9536 * already filled in by the caller.
9539 9537 * If legacy_req is true then MIB structures needs to be truncated to their
9540 9538 * legacy sizes before being returned.
9541 9539 * Return value of 0 indicates that no messages were sent and caller
9542 9540 * should free mpctl.
9543 9541 */
9544 9542 int
9545 9543 ip_snmp_get(queue_t *q, mblk_t *mpctl, int level, boolean_t legacy_req)
9546 9544 {
9547 9545 ip_stack_t *ipst;
9548 9546 sctp_stack_t *sctps;
9549 9547
9550 9548 if (q->q_next != NULL) {
9551 9549 ipst = ILLQ_TO_IPST(q);
9552 9550 } else {
9553 9551 ipst = CONNQ_TO_IPST(q);
9554 9552 }
9555 9553 ASSERT(ipst != NULL);
9556 9554 sctps = ipst->ips_netstack->netstack_sctp;
9557 9555
9558 9556 if (mpctl == NULL || mpctl->b_cont == NULL) {
9559 9557 return (0);
9560 9558 }
9561 9559
9562 9560 /*
9563 9561 * For the purposes of the (broken) packet shell use
9564 9562 * of the level we make sure MIB2_TCP/MIB2_UDP can be used
9565 9563 * to make TCP and UDP appear first in the list of mib items.
9566 9564 * TBD: We could expand this and use it in netstat so that
9567 9565 * the kernel doesn't have to produce large tables (connections,
9568 9566 * routes, etc) when netstat only wants the statistics or a particular
9569 9567 * table.
9570 9568 */
9571 9569 if (!(level == MIB2_TCP || level == MIB2_UDP)) {
9572 9570 if ((mpctl = icmp_snmp_get(q, mpctl)) == NULL) {
9573 9571 return (1);
9574 9572 }
9575 9573 }
9576 9574
9577 9575 if (level != MIB2_TCP) {
9578 9576 if ((mpctl = udp_snmp_get(q, mpctl, legacy_req)) == NULL) {
9579 9577 return (1);
9580 9578 }
9581 9579 if (level == MIB2_UDP) {
9582 9580 goto done;
9583 9581 }
9584 9582 }
9585 9583
9586 9584 if (level != MIB2_UDP) {
9587 9585 if ((mpctl = tcp_snmp_get(q, mpctl, legacy_req)) == NULL) {
9588 9586 return (1);
9589 9587 }
9590 9588 if (level == MIB2_TCP) {
9591 9589 goto done;
9592 9590 }
9593 9591 }
9594 9592
9595 9593 if ((mpctl = ip_snmp_get_mib2_ip_traffic_stats(q, mpctl,
9596 9594 ipst, legacy_req)) == NULL) {
9597 9595 return (1);
9598 9596 }
9599 9597
9600 9598 if ((mpctl = ip_snmp_get_mib2_ip6(q, mpctl, ipst,
9601 9599 legacy_req)) == NULL) {
9602 9600 return (1);
9603 9601 }
9604 9602
9605 9603 if ((mpctl = ip_snmp_get_mib2_icmp(q, mpctl, ipst)) == NULL) {
9606 9604 return (1);
9607 9605 }
9608 9606
9609 9607 if ((mpctl = ip_snmp_get_mib2_icmp6(q, mpctl, ipst)) == NULL) {
9610 9608 return (1);
9611 9609 }
9612 9610
9613 9611 if ((mpctl = ip_snmp_get_mib2_igmp(q, mpctl, ipst)) == NULL) {
9614 9612 return (1);
9615 9613 }
9616 9614
9617 9615 if ((mpctl = ip_snmp_get_mib2_multi(q, mpctl, ipst)) == NULL) {
9618 9616 return (1);
9619 9617 }
9620 9618
9621 9619 if ((mpctl = ip_snmp_get_mib2_ip_addr(q, mpctl, ipst,
9622 9620 legacy_req)) == NULL) {
9623 9621 return (1);
9624 9622 }
9625 9623
9626 9624 if ((mpctl = ip_snmp_get_mib2_ip6_addr(q, mpctl, ipst,
9627 9625 legacy_req)) == NULL) {
9628 9626 return (1);
9629 9627 }
9630 9628
9631 9629 if ((mpctl = ip_snmp_get_mib2_ip_group_mem(q, mpctl, ipst)) == NULL) {
9632 9630 return (1);
9633 9631 }
9634 9632
9635 9633 if ((mpctl = ip_snmp_get_mib2_ip6_group_mem(q, mpctl, ipst)) == NULL) {
9636 9634 return (1);
9637 9635 }
9638 9636
9639 9637 if ((mpctl = ip_snmp_get_mib2_ip_group_src(q, mpctl, ipst)) == NULL) {
9640 9638 return (1);
9641 9639 }
9642 9640
9643 9641 if ((mpctl = ip_snmp_get_mib2_ip6_group_src(q, mpctl, ipst)) == NULL) {
9644 9642 return (1);
9645 9643 }
9646 9644
9647 9645 if ((mpctl = ip_snmp_get_mib2_virt_multi(q, mpctl, ipst)) == NULL) {
9648 9646 return (1);
9649 9647 }
9650 9648
9651 9649 if ((mpctl = ip_snmp_get_mib2_multi_rtable(q, mpctl, ipst)) == NULL) {
9652 9650 return (1);
9653 9651 }
9654 9652
9655 9653 mpctl = ip_snmp_get_mib2_ip_route_media(q, mpctl, level, ipst);
9656 9654 if (mpctl == NULL)
9657 9655 return (1);
9658 9656
9659 9657 mpctl = ip_snmp_get_mib2_ip6_route_media(q, mpctl, level, ipst);
9660 9658 if (mpctl == NULL)
9661 9659 return (1);
9662 9660
9663 9661 if ((mpctl = sctp_snmp_get_mib2(q, mpctl, sctps)) == NULL) {
9664 9662 return (1);
9665 9663 }
9666 9664 if ((mpctl = ip_snmp_get_mib2_ip_dce(q, mpctl, ipst)) == NULL) {
9667 9665 return (1);
9668 9666 }
9669 9667 done:
9670 9668 freemsg(mpctl);
9671 9669 return (1);
9672 9670 }
9673 9671
9674 9672 /* Get global (legacy) IPv4 statistics */
9675 9673 static mblk_t *
9676 9674 ip_snmp_get_mib2_ip(queue_t *q, mblk_t *mpctl, mib2_ipIfStatsEntry_t *ipmib,
9677 9675 ip_stack_t *ipst, boolean_t legacy_req)
9678 9676 {
9679 9677 mib2_ip_t old_ip_mib;
9680 9678 struct opthdr *optp;
9681 9679 mblk_t *mp2ctl;
9682 9680 mib2_ipAddrEntry_t mae;
9683 9681
9684 9682 /*
9685 9683 * make a copy of the original message
9686 9684 */
9687 9685 mp2ctl = copymsg(mpctl);
9688 9686
9689 9687 /* fixed length IP structure... */
9690 9688 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9691 9689 optp->level = MIB2_IP;
9692 9690 optp->name = 0;
9693 9691 SET_MIB(old_ip_mib.ipForwarding,
9694 9692 (WE_ARE_FORWARDING(ipst) ? 1 : 2));
9695 9693 SET_MIB(old_ip_mib.ipDefaultTTL,
9696 9694 (uint32_t)ipst->ips_ip_def_ttl);
9697 9695 SET_MIB(old_ip_mib.ipReasmTimeout,
9698 9696 ipst->ips_ip_reassembly_timeout);
9699 9697 SET_MIB(old_ip_mib.ipAddrEntrySize,
9700 9698 (legacy_req) ? LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t) :
9701 9699 sizeof (mib2_ipAddrEntry_t));
9702 9700 SET_MIB(old_ip_mib.ipRouteEntrySize,
9703 9701 sizeof (mib2_ipRouteEntry_t));
9704 9702 SET_MIB(old_ip_mib.ipNetToMediaEntrySize,
9705 9703 sizeof (mib2_ipNetToMediaEntry_t));
9706 9704 SET_MIB(old_ip_mib.ipMemberEntrySize, sizeof (ip_member_t));
9707 9705 SET_MIB(old_ip_mib.ipGroupSourceEntrySize, sizeof (ip_grpsrc_t));
9708 9706 SET_MIB(old_ip_mib.ipRouteAttributeSize,
9709 9707 sizeof (mib2_ipAttributeEntry_t));
9710 9708 SET_MIB(old_ip_mib.transportMLPSize, sizeof (mib2_transportMLPEntry_t));
9711 9709 SET_MIB(old_ip_mib.ipDestEntrySize, sizeof (dest_cache_entry_t));
9712 9710
9713 9711 /*
9714 9712 * Grab the statistics from the new IP MIB
9715 9713 */
9716 9714 SET_MIB(old_ip_mib.ipInReceives,
9717 9715 (uint32_t)ipmib->ipIfStatsHCInReceives);
9718 9716 SET_MIB(old_ip_mib.ipInHdrErrors, ipmib->ipIfStatsInHdrErrors);
9719 9717 SET_MIB(old_ip_mib.ipInAddrErrors, ipmib->ipIfStatsInAddrErrors);
9720 9718 SET_MIB(old_ip_mib.ipForwDatagrams,
9721 9719 (uint32_t)ipmib->ipIfStatsHCOutForwDatagrams);
9722 9720 SET_MIB(old_ip_mib.ipInUnknownProtos,
9723 9721 ipmib->ipIfStatsInUnknownProtos);
9724 9722 SET_MIB(old_ip_mib.ipInDiscards, ipmib->ipIfStatsInDiscards);
9725 9723 SET_MIB(old_ip_mib.ipInDelivers,
9726 9724 (uint32_t)ipmib->ipIfStatsHCInDelivers);
9727 9725 SET_MIB(old_ip_mib.ipOutRequests,
9728 9726 (uint32_t)ipmib->ipIfStatsHCOutRequests);
9729 9727 SET_MIB(old_ip_mib.ipOutDiscards, ipmib->ipIfStatsOutDiscards);
9730 9728 SET_MIB(old_ip_mib.ipOutNoRoutes, ipmib->ipIfStatsOutNoRoutes);
9731 9729 SET_MIB(old_ip_mib.ipReasmReqds, ipmib->ipIfStatsReasmReqds);
9732 9730 SET_MIB(old_ip_mib.ipReasmOKs, ipmib->ipIfStatsReasmOKs);
9733 9731 SET_MIB(old_ip_mib.ipReasmFails, ipmib->ipIfStatsReasmFails);
9734 9732 SET_MIB(old_ip_mib.ipFragOKs, ipmib->ipIfStatsOutFragOKs);
9735 9733 SET_MIB(old_ip_mib.ipFragFails, ipmib->ipIfStatsOutFragFails);
9736 9734 SET_MIB(old_ip_mib.ipFragCreates, ipmib->ipIfStatsOutFragCreates);
9737 9735
9738 9736 /* ipRoutingDiscards is not being used */
9739 9737 SET_MIB(old_ip_mib.ipRoutingDiscards, 0);
9740 9738 SET_MIB(old_ip_mib.tcpInErrs, ipmib->tcpIfStatsInErrs);
9741 9739 SET_MIB(old_ip_mib.udpNoPorts, ipmib->udpIfStatsNoPorts);
9742 9740 SET_MIB(old_ip_mib.ipInCksumErrs, ipmib->ipIfStatsInCksumErrs);
9743 9741 SET_MIB(old_ip_mib.ipReasmDuplicates,
9744 9742 ipmib->ipIfStatsReasmDuplicates);
9745 9743 SET_MIB(old_ip_mib.ipReasmPartDups, ipmib->ipIfStatsReasmPartDups);
9746 9744 SET_MIB(old_ip_mib.ipForwProhibits, ipmib->ipIfStatsForwProhibits);
9747 9745 SET_MIB(old_ip_mib.udpInCksumErrs, ipmib->udpIfStatsInCksumErrs);
9748 9746 SET_MIB(old_ip_mib.udpInOverflows, ipmib->udpIfStatsInOverflows);
9749 9747 SET_MIB(old_ip_mib.rawipInOverflows,
9750 9748 ipmib->rawipIfStatsInOverflows);
9751 9749
9752 9750 SET_MIB(old_ip_mib.ipsecInSucceeded, ipmib->ipsecIfStatsInSucceeded);
9753 9751 SET_MIB(old_ip_mib.ipsecInFailed, ipmib->ipsecIfStatsInFailed);
9754 9752 SET_MIB(old_ip_mib.ipInIPv6, ipmib->ipIfStatsInWrongIPVersion);
9755 9753 SET_MIB(old_ip_mib.ipOutIPv6, ipmib->ipIfStatsOutWrongIPVersion);
9756 9754 SET_MIB(old_ip_mib.ipOutSwitchIPv6,
9757 9755 ipmib->ipIfStatsOutSwitchIPVersion);
9758 9756
9759 9757 if (!snmp_append_data(mpctl->b_cont, (char *)&old_ip_mib,
9760 9758 (int)sizeof (old_ip_mib))) {
9761 9759 ip1dbg(("ip_snmp_get_mib2_ip: failed to allocate %u bytes\n",
9762 9760 (uint_t)sizeof (old_ip_mib)));
9763 9761 }
9764 9762
9765 9763 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9766 9764 ip3dbg(("ip_snmp_get_mib2_ip: level %d, name %d, len %d\n",
9767 9765 (int)optp->level, (int)optp->name, (int)optp->len));
9768 9766 qreply(q, mpctl);
9769 9767 return (mp2ctl);
9770 9768 }
9771 9769
9772 9770 /* Per interface IPv4 statistics */
9773 9771 static mblk_t *
9774 9772 ip_snmp_get_mib2_ip_traffic_stats(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
9775 9773 boolean_t legacy_req)
9776 9774 {
9777 9775 struct opthdr *optp;
9778 9776 mblk_t *mp2ctl;
9779 9777 ill_t *ill;
9780 9778 ill_walk_context_t ctx;
9781 9779 mblk_t *mp_tail = NULL;
9782 9780 mib2_ipIfStatsEntry_t global_ip_mib;
9783 9781 mib2_ipAddrEntry_t mae;
9784 9782
9785 9783 /*
9786 9784 * Make a copy of the original message
9787 9785 */
9788 9786 mp2ctl = copymsg(mpctl);
9789 9787
9790 9788 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9791 9789 optp->level = MIB2_IP;
9792 9790 optp->name = MIB2_IP_TRAFFIC_STATS;
9793 9791 /* Include "unknown interface" ip_mib */
9794 9792 ipst->ips_ip_mib.ipIfStatsIPVersion = MIB2_INETADDRESSTYPE_ipv4;
9795 9793 ipst->ips_ip_mib.ipIfStatsIfIndex =
9796 9794 MIB2_UNKNOWN_INTERFACE; /* Flag to netstat */
9797 9795 SET_MIB(ipst->ips_ip_mib.ipIfStatsForwarding,
9798 9796 (ipst->ips_ip_forwarding ? 1 : 2));
9799 9797 SET_MIB(ipst->ips_ip_mib.ipIfStatsDefaultTTL,
9800 9798 (uint32_t)ipst->ips_ip_def_ttl);
9801 9799 SET_MIB(ipst->ips_ip_mib.ipIfStatsEntrySize,
9802 9800 sizeof (mib2_ipIfStatsEntry_t));
9803 9801 SET_MIB(ipst->ips_ip_mib.ipIfStatsAddrEntrySize,
9804 9802 sizeof (mib2_ipAddrEntry_t));
9805 9803 SET_MIB(ipst->ips_ip_mib.ipIfStatsRouteEntrySize,
9806 9804 sizeof (mib2_ipRouteEntry_t));
9807 9805 SET_MIB(ipst->ips_ip_mib.ipIfStatsNetToMediaEntrySize,
9808 9806 sizeof (mib2_ipNetToMediaEntry_t));
9809 9807 SET_MIB(ipst->ips_ip_mib.ipIfStatsMemberEntrySize,
9810 9808 sizeof (ip_member_t));
9811 9809 SET_MIB(ipst->ips_ip_mib.ipIfStatsGroupSourceEntrySize,
9812 9810 sizeof (ip_grpsrc_t));
9813 9811
9814 9812 bcopy(&ipst->ips_ip_mib, &global_ip_mib, sizeof (global_ip_mib));
9815 9813
9816 9814 if (legacy_req) {
9817 9815 SET_MIB(global_ip_mib.ipIfStatsAddrEntrySize,
9818 9816 LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t));
9819 9817 }
9820 9818
9821 9819 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
9822 9820 (char *)&global_ip_mib, (int)sizeof (global_ip_mib))) {
9823 9821 ip1dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9824 9822 "failed to allocate %u bytes\n",
9825 9823 (uint_t)sizeof (global_ip_mib)));
9826 9824 }
9827 9825
9828 9826 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
9829 9827 ill = ILL_START_WALK_V4(&ctx, ipst);
9830 9828 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
9831 9829 ill->ill_ip_mib->ipIfStatsIfIndex =
9832 9830 ill->ill_phyint->phyint_ifindex;
9833 9831 SET_MIB(ill->ill_ip_mib->ipIfStatsForwarding,
9834 9832 (ipst->ips_ip_forwarding ? 1 : 2));
9835 9833 SET_MIB(ill->ill_ip_mib->ipIfStatsDefaultTTL,
9836 9834 (uint32_t)ipst->ips_ip_def_ttl);
9837 9835
9838 9836 ip_mib2_add_ip_stats(&global_ip_mib, ill->ill_ip_mib);
9839 9837 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
9840 9838 (char *)ill->ill_ip_mib,
9841 9839 (int)sizeof (*ill->ill_ip_mib))) {
9842 9840 ip1dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9843 9841 "failed to allocate %u bytes\n",
9844 9842 (uint_t)sizeof (*ill->ill_ip_mib)));
9845 9843 }
9846 9844 }
9847 9845 rw_exit(&ipst->ips_ill_g_lock);
9848 9846
9849 9847 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9850 9848 ip3dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9851 9849 "level %d, name %d, len %d\n",
9852 9850 (int)optp->level, (int)optp->name, (int)optp->len));
9853 9851 qreply(q, mpctl);
9854 9852
9855 9853 if (mp2ctl == NULL)
9856 9854 return (NULL);
9857 9855
9858 9856 return (ip_snmp_get_mib2_ip(q, mp2ctl, &global_ip_mib, ipst,
9859 9857 legacy_req));
9860 9858 }
9861 9859
9862 9860 /* Global IPv4 ICMP statistics */
9863 9861 static mblk_t *
9864 9862 ip_snmp_get_mib2_icmp(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9865 9863 {
9866 9864 struct opthdr *optp;
9867 9865 mblk_t *mp2ctl;
9868 9866
9869 9867 /*
9870 9868 * Make a copy of the original message
9871 9869 */
9872 9870 mp2ctl = copymsg(mpctl);
9873 9871
9874 9872 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9875 9873 optp->level = MIB2_ICMP;
9876 9874 optp->name = 0;
9877 9875 if (!snmp_append_data(mpctl->b_cont, (char *)&ipst->ips_icmp_mib,
9878 9876 (int)sizeof (ipst->ips_icmp_mib))) {
9879 9877 ip1dbg(("ip_snmp_get_mib2_icmp: failed to allocate %u bytes\n",
9880 9878 (uint_t)sizeof (ipst->ips_icmp_mib)));
9881 9879 }
9882 9880 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9883 9881 ip3dbg(("ip_snmp_get_mib2_icmp: level %d, name %d, len %d\n",
9884 9882 (int)optp->level, (int)optp->name, (int)optp->len));
9885 9883 qreply(q, mpctl);
9886 9884 return (mp2ctl);
9887 9885 }
9888 9886
9889 9887 /* Global IPv4 IGMP statistics */
9890 9888 static mblk_t *
9891 9889 ip_snmp_get_mib2_igmp(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9892 9890 {
9893 9891 struct opthdr *optp;
9894 9892 mblk_t *mp2ctl;
9895 9893
9896 9894 /*
9897 9895 * make a copy of the original message
9898 9896 */
9899 9897 mp2ctl = copymsg(mpctl);
9900 9898
9901 9899 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9902 9900 optp->level = EXPER_IGMP;
9903 9901 optp->name = 0;
9904 9902 if (!snmp_append_data(mpctl->b_cont, (char *)&ipst->ips_igmpstat,
9905 9903 (int)sizeof (ipst->ips_igmpstat))) {
9906 9904 ip1dbg(("ip_snmp_get_mib2_igmp: failed to allocate %u bytes\n",
9907 9905 (uint_t)sizeof (ipst->ips_igmpstat)));
9908 9906 }
9909 9907 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9910 9908 ip3dbg(("ip_snmp_get_mib2_igmp: level %d, name %d, len %d\n",
9911 9909 (int)optp->level, (int)optp->name, (int)optp->len));
9912 9910 qreply(q, mpctl);
9913 9911 return (mp2ctl);
9914 9912 }
9915 9913
9916 9914 /* Global IPv4 Multicast Routing statistics */
9917 9915 static mblk_t *
9918 9916 ip_snmp_get_mib2_multi(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9919 9917 {
9920 9918 struct opthdr *optp;
9921 9919 mblk_t *mp2ctl;
9922 9920
9923 9921 /*
9924 9922 * make a copy of the original message
9925 9923 */
9926 9924 mp2ctl = copymsg(mpctl);
9927 9925
9928 9926 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9929 9927 optp->level = EXPER_DVMRP;
9930 9928 optp->name = 0;
9931 9929 if (!ip_mroute_stats(mpctl->b_cont, ipst)) {
9932 9930 ip0dbg(("ip_mroute_stats: failed\n"));
9933 9931 }
9934 9932 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9935 9933 ip3dbg(("ip_snmp_get_mib2_multi: level %d, name %d, len %d\n",
9936 9934 (int)optp->level, (int)optp->name, (int)optp->len));
9937 9935 qreply(q, mpctl);
9938 9936 return (mp2ctl);
9939 9937 }
9940 9938
9941 9939 /* IPv4 address information */
9942 9940 static mblk_t *
9943 9941 ip_snmp_get_mib2_ip_addr(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
9944 9942 boolean_t legacy_req)
9945 9943 {
9946 9944 struct opthdr *optp;
9947 9945 mblk_t *mp2ctl;
9948 9946 mblk_t *mp_tail = NULL;
9949 9947 ill_t *ill;
9950 9948 ipif_t *ipif;
9951 9949 uint_t bitval;
9952 9950 mib2_ipAddrEntry_t mae;
9953 9951 size_t mae_size;
9954 9952 zoneid_t zoneid;
9955 9953 ill_walk_context_t ctx;
9956 9954
9957 9955 /*
9958 9956 * make a copy of the original message
9959 9957 */
9960 9958 mp2ctl = copymsg(mpctl);
9961 9959
9962 9960 mae_size = (legacy_req) ? LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t) :
9963 9961 sizeof (mib2_ipAddrEntry_t);
9964 9962
9965 9963 /* ipAddrEntryTable */
9966 9964
9967 9965 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9968 9966 optp->level = MIB2_IP;
9969 9967 optp->name = MIB2_IP_ADDR;
9970 9968 zoneid = Q_TO_CONN(q)->conn_zoneid;
9971 9969
9972 9970 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
9973 9971 ill = ILL_START_WALK_V4(&ctx, ipst);
9974 9972 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
9975 9973 for (ipif = ill->ill_ipif; ipif != NULL;
9976 9974 ipif = ipif->ipif_next) {
9977 9975 if (ipif->ipif_zoneid != zoneid &&
9978 9976 ipif->ipif_zoneid != ALL_ZONES)
9979 9977 continue;
9980 9978 /* Sum of count from dead IRE_LO* and our current */
9981 9979 mae.ipAdEntInfo.ae_ibcnt = ipif->ipif_ib_pkt_count;
9982 9980 if (ipif->ipif_ire_local != NULL) {
9983 9981 mae.ipAdEntInfo.ae_ibcnt +=
9984 9982 ipif->ipif_ire_local->ire_ib_pkt_count;
9985 9983 }
9986 9984 mae.ipAdEntInfo.ae_obcnt = 0;
9987 9985 mae.ipAdEntInfo.ae_focnt = 0;
9988 9986
9989 9987 ipif_get_name(ipif, mae.ipAdEntIfIndex.o_bytes,
9990 9988 OCTET_LENGTH);
9991 9989 mae.ipAdEntIfIndex.o_length =
9992 9990 mi_strlen(mae.ipAdEntIfIndex.o_bytes);
9993 9991 mae.ipAdEntAddr = ipif->ipif_lcl_addr;
9994 9992 mae.ipAdEntNetMask = ipif->ipif_net_mask;
9995 9993 mae.ipAdEntInfo.ae_subnet = ipif->ipif_subnet;
9996 9994 mae.ipAdEntInfo.ae_subnet_len =
9997 9995 ip_mask_to_plen(ipif->ipif_net_mask);
9998 9996 mae.ipAdEntInfo.ae_src_addr = ipif->ipif_lcl_addr;
9999 9997 for (bitval = 1;
10000 9998 bitval &&
10001 9999 !(bitval & ipif->ipif_brd_addr);
10002 10000 bitval <<= 1)
10003 10001 noop;
10004 10002 mae.ipAdEntBcastAddr = bitval;
10005 10003 mae.ipAdEntReasmMaxSize = IP_MAXPACKET;
10006 10004 mae.ipAdEntInfo.ae_mtu = ipif->ipif_ill->ill_mtu;
10007 10005 mae.ipAdEntInfo.ae_metric = ipif->ipif_ill->ill_metric;
10008 10006 mae.ipAdEntInfo.ae_broadcast_addr =
10009 10007 ipif->ipif_brd_addr;
10010 10008 mae.ipAdEntInfo.ae_pp_dst_addr =
10011 10009 ipif->ipif_pp_dst_addr;
10012 10010 mae.ipAdEntInfo.ae_flags = ipif->ipif_flags |
10013 10011 ill->ill_flags | ill->ill_phyint->phyint_flags;
10014 10012 mae.ipAdEntRetransmitTime =
10015 10013 ill->ill_reachable_retrans_time;
10016 10014
10017 10015 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10018 10016 (char *)&mae, (int)mae_size)) {
10019 10017 ip1dbg(("ip_snmp_get_mib2_ip_addr: failed to "
10020 10018 "allocate %u bytes\n", (uint_t)mae_size));
10021 10019 }
10022 10020 }
10023 10021 }
10024 10022 rw_exit(&ipst->ips_ill_g_lock);
10025 10023
10026 10024 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10027 10025 ip3dbg(("ip_snmp_get_mib2_ip_addr: level %d, name %d, len %d\n",
10028 10026 (int)optp->level, (int)optp->name, (int)optp->len));
10029 10027 qreply(q, mpctl);
10030 10028 return (mp2ctl);
10031 10029 }
10032 10030
10033 10031 /* IPv6 address information */
10034 10032 static mblk_t *
10035 10033 ip_snmp_get_mib2_ip6_addr(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
10036 10034 boolean_t legacy_req)
10037 10035 {
10038 10036 struct opthdr *optp;
10039 10037 mblk_t *mp2ctl;
10040 10038 mblk_t *mp_tail = NULL;
10041 10039 ill_t *ill;
10042 10040 ipif_t *ipif;
10043 10041 mib2_ipv6AddrEntry_t mae6;
10044 10042 size_t mae6_size;
10045 10043 zoneid_t zoneid;
10046 10044 ill_walk_context_t ctx;
10047 10045
10048 10046 /*
10049 10047 * make a copy of the original message
10050 10048 */
10051 10049 mp2ctl = copymsg(mpctl);
10052 10050
10053 10051 mae6_size = (legacy_req) ?
10054 10052 LEGACY_MIB_SIZE(&mae6, mib2_ipv6AddrEntry_t) :
10055 10053 sizeof (mib2_ipv6AddrEntry_t);
10056 10054
10057 10055 /* ipv6AddrEntryTable */
10058 10056
10059 10057 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10060 10058 optp->level = MIB2_IP6;
10061 10059 optp->name = MIB2_IP6_ADDR;
10062 10060 zoneid = Q_TO_CONN(q)->conn_zoneid;
10063 10061
10064 10062 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10065 10063 ill = ILL_START_WALK_V6(&ctx, ipst);
10066 10064 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10067 10065 for (ipif = ill->ill_ipif; ipif != NULL;
10068 10066 ipif = ipif->ipif_next) {
10069 10067 if (ipif->ipif_zoneid != zoneid &&
10070 10068 ipif->ipif_zoneid != ALL_ZONES)
10071 10069 continue;
10072 10070 /* Sum of count from dead IRE_LO* and our current */
10073 10071 mae6.ipv6AddrInfo.ae_ibcnt = ipif->ipif_ib_pkt_count;
10074 10072 if (ipif->ipif_ire_local != NULL) {
10075 10073 mae6.ipv6AddrInfo.ae_ibcnt +=
10076 10074 ipif->ipif_ire_local->ire_ib_pkt_count;
10077 10075 }
10078 10076 mae6.ipv6AddrInfo.ae_obcnt = 0;
10079 10077 mae6.ipv6AddrInfo.ae_focnt = 0;
10080 10078
10081 10079 ipif_get_name(ipif, mae6.ipv6AddrIfIndex.o_bytes,
10082 10080 OCTET_LENGTH);
10083 10081 mae6.ipv6AddrIfIndex.o_length =
10084 10082 mi_strlen(mae6.ipv6AddrIfIndex.o_bytes);
10085 10083 mae6.ipv6AddrAddress = ipif->ipif_v6lcl_addr;
10086 10084 mae6.ipv6AddrPfxLength =
10087 10085 ip_mask_to_plen_v6(&ipif->ipif_v6net_mask);
10088 10086 mae6.ipv6AddrInfo.ae_subnet = ipif->ipif_v6subnet;
10089 10087 mae6.ipv6AddrInfo.ae_subnet_len =
10090 10088 mae6.ipv6AddrPfxLength;
10091 10089 mae6.ipv6AddrInfo.ae_src_addr = ipif->ipif_v6lcl_addr;
10092 10090
10093 10091 /* Type: stateless(1), stateful(2), unknown(3) */
10094 10092 if (ipif->ipif_flags & IPIF_ADDRCONF)
10095 10093 mae6.ipv6AddrType = 1;
10096 10094 else
10097 10095 mae6.ipv6AddrType = 2;
10098 10096 /* Anycast: true(1), false(2) */
10099 10097 if (ipif->ipif_flags & IPIF_ANYCAST)
10100 10098 mae6.ipv6AddrAnycastFlag = 1;
10101 10099 else
10102 10100 mae6.ipv6AddrAnycastFlag = 2;
10103 10101
10104 10102 /*
10105 10103 * Address status: preferred(1), deprecated(2),
10106 10104 * invalid(3), inaccessible(4), unknown(5)
10107 10105 */
10108 10106 if (ipif->ipif_flags & IPIF_NOLOCAL)
10109 10107 mae6.ipv6AddrStatus = 3;
10110 10108 else if (ipif->ipif_flags & IPIF_DEPRECATED)
10111 10109 mae6.ipv6AddrStatus = 2;
10112 10110 else
10113 10111 mae6.ipv6AddrStatus = 1;
10114 10112 mae6.ipv6AddrInfo.ae_mtu = ipif->ipif_ill->ill_mtu;
10115 10113 mae6.ipv6AddrInfo.ae_metric =
10116 10114 ipif->ipif_ill->ill_metric;
10117 10115 mae6.ipv6AddrInfo.ae_pp_dst_addr =
10118 10116 ipif->ipif_v6pp_dst_addr;
10119 10117 mae6.ipv6AddrInfo.ae_flags = ipif->ipif_flags |
10120 10118 ill->ill_flags | ill->ill_phyint->phyint_flags;
10121 10119 mae6.ipv6AddrReasmMaxSize = IP_MAXPACKET;
10122 10120 mae6.ipv6AddrIdentifier = ill->ill_token;
10123 10121 mae6.ipv6AddrIdentifierLen = ill->ill_token_length;
10124 10122 mae6.ipv6AddrReachableTime = ill->ill_reachable_time;
10125 10123 mae6.ipv6AddrRetransmitTime =
10126 10124 ill->ill_reachable_retrans_time;
10127 10125 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10128 10126 (char *)&mae6, (int)mae6_size)) {
10129 10127 ip1dbg(("ip_snmp_get_mib2_ip6_addr: failed to "
10130 10128 "allocate %u bytes\n",
10131 10129 (uint_t)mae6_size));
10132 10130 }
10133 10131 }
10134 10132 }
10135 10133 rw_exit(&ipst->ips_ill_g_lock);
10136 10134
10137 10135 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10138 10136 ip3dbg(("ip_snmp_get_mib2_ip6_addr: level %d, name %d, len %d\n",
10139 10137 (int)optp->level, (int)optp->name, (int)optp->len));
10140 10138 qreply(q, mpctl);
10141 10139 return (mp2ctl);
10142 10140 }
10143 10141
10144 10142 /* IPv4 multicast group membership. */
10145 10143 static mblk_t *
10146 10144 ip_snmp_get_mib2_ip_group_mem(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10147 10145 {
10148 10146 struct opthdr *optp;
10149 10147 mblk_t *mp2ctl;
10150 10148 ill_t *ill;
10151 10149 ipif_t *ipif;
10152 10150 ilm_t *ilm;
10153 10151 ip_member_t ipm;
10154 10152 mblk_t *mp_tail = NULL;
10155 10153 ill_walk_context_t ctx;
10156 10154 zoneid_t zoneid;
10157 10155
10158 10156 /*
10159 10157 * make a copy of the original message
10160 10158 */
10161 10159 mp2ctl = copymsg(mpctl);
10162 10160 zoneid = Q_TO_CONN(q)->conn_zoneid;
10163 10161
10164 10162 /* ipGroupMember table */
10165 10163 optp = (struct opthdr *)&mpctl->b_rptr[
10166 10164 sizeof (struct T_optmgmt_ack)];
10167 10165 optp->level = MIB2_IP;
10168 10166 optp->name = EXPER_IP_GROUP_MEMBERSHIP;
10169 10167
10170 10168 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10171 10169 ill = ILL_START_WALK_V4(&ctx, ipst);
10172 10170 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10173 10171 /* Make sure the ill isn't going away. */
10174 10172 if (!ill_check_and_refhold(ill))
10175 10173 continue;
10176 10174 rw_exit(&ipst->ips_ill_g_lock);
10177 10175 rw_enter(&ill->ill_mcast_lock, RW_READER);
10178 10176 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10179 10177 if (ilm->ilm_zoneid != zoneid &&
10180 10178 ilm->ilm_zoneid != ALL_ZONES)
10181 10179 continue;
10182 10180
10183 10181 /* Is there an ipif for ilm_ifaddr? */
10184 10182 for (ipif = ill->ill_ipif; ipif != NULL;
10185 10183 ipif = ipif->ipif_next) {
10186 10184 if (!IPIF_IS_CONDEMNED(ipif) &&
10187 10185 ipif->ipif_lcl_addr == ilm->ilm_ifaddr &&
10188 10186 ilm->ilm_ifaddr != INADDR_ANY)
10189 10187 break;
10190 10188 }
10191 10189 if (ipif != NULL) {
10192 10190 ipif_get_name(ipif,
10193 10191 ipm.ipGroupMemberIfIndex.o_bytes,
10194 10192 OCTET_LENGTH);
10195 10193 } else {
10196 10194 ill_get_name(ill,
10197 10195 ipm.ipGroupMemberIfIndex.o_bytes,
10198 10196 OCTET_LENGTH);
10199 10197 }
10200 10198 ipm.ipGroupMemberIfIndex.o_length =
10201 10199 mi_strlen(ipm.ipGroupMemberIfIndex.o_bytes);
10202 10200
10203 10201 ipm.ipGroupMemberAddress = ilm->ilm_addr;
10204 10202 ipm.ipGroupMemberRefCnt = ilm->ilm_refcnt;
10205 10203 ipm.ipGroupMemberFilterMode = ilm->ilm_fmode;
10206 10204 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10207 10205 (char *)&ipm, (int)sizeof (ipm))) {
10208 10206 ip1dbg(("ip_snmp_get_mib2_ip_group: "
10209 10207 "failed to allocate %u bytes\n",
10210 10208 (uint_t)sizeof (ipm)));
10211 10209 }
10212 10210 }
10213 10211 rw_exit(&ill->ill_mcast_lock);
10214 10212 ill_refrele(ill);
10215 10213 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10216 10214 }
10217 10215 rw_exit(&ipst->ips_ill_g_lock);
10218 10216 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10219 10217 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10220 10218 (int)optp->level, (int)optp->name, (int)optp->len));
10221 10219 qreply(q, mpctl);
10222 10220 return (mp2ctl);
10223 10221 }
10224 10222
10225 10223 /* IPv6 multicast group membership. */
10226 10224 static mblk_t *
10227 10225 ip_snmp_get_mib2_ip6_group_mem(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10228 10226 {
10229 10227 struct opthdr *optp;
10230 10228 mblk_t *mp2ctl;
10231 10229 ill_t *ill;
10232 10230 ilm_t *ilm;
10233 10231 ipv6_member_t ipm6;
10234 10232 mblk_t *mp_tail = NULL;
10235 10233 ill_walk_context_t ctx;
10236 10234 zoneid_t zoneid;
10237 10235
10238 10236 /*
10239 10237 * make a copy of the original message
10240 10238 */
10241 10239 mp2ctl = copymsg(mpctl);
10242 10240 zoneid = Q_TO_CONN(q)->conn_zoneid;
10243 10241
10244 10242 /* ip6GroupMember table */
10245 10243 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10246 10244 optp->level = MIB2_IP6;
10247 10245 optp->name = EXPER_IP6_GROUP_MEMBERSHIP;
10248 10246
10249 10247 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10250 10248 ill = ILL_START_WALK_V6(&ctx, ipst);
10251 10249 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10252 10250 /* Make sure the ill isn't going away. */
10253 10251 if (!ill_check_and_refhold(ill))
10254 10252 continue;
10255 10253 rw_exit(&ipst->ips_ill_g_lock);
10256 10254 /*
10257 10255 * Normally we don't have any members on under IPMP interfaces.
10258 10256 * We report them as a debugging aid.
10259 10257 */
10260 10258 rw_enter(&ill->ill_mcast_lock, RW_READER);
10261 10259 ipm6.ipv6GroupMemberIfIndex = ill->ill_phyint->phyint_ifindex;
10262 10260 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10263 10261 if (ilm->ilm_zoneid != zoneid &&
10264 10262 ilm->ilm_zoneid != ALL_ZONES)
10265 10263 continue; /* not this zone */
10266 10264 ipm6.ipv6GroupMemberAddress = ilm->ilm_v6addr;
10267 10265 ipm6.ipv6GroupMemberRefCnt = ilm->ilm_refcnt;
10268 10266 ipm6.ipv6GroupMemberFilterMode = ilm->ilm_fmode;
10269 10267 if (!snmp_append_data2(mpctl->b_cont,
10270 10268 &mp_tail,
10271 10269 (char *)&ipm6, (int)sizeof (ipm6))) {
10272 10270 ip1dbg(("ip_snmp_get_mib2_ip6_group: "
10273 10271 "failed to allocate %u bytes\n",
10274 10272 (uint_t)sizeof (ipm6)));
10275 10273 }
10276 10274 }
10277 10275 rw_exit(&ill->ill_mcast_lock);
10278 10276 ill_refrele(ill);
10279 10277 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10280 10278 }
10281 10279 rw_exit(&ipst->ips_ill_g_lock);
10282 10280
10283 10281 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10284 10282 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10285 10283 (int)optp->level, (int)optp->name, (int)optp->len));
10286 10284 qreply(q, mpctl);
10287 10285 return (mp2ctl);
10288 10286 }
10289 10287
10290 10288 /* IP multicast filtered sources */
10291 10289 static mblk_t *
10292 10290 ip_snmp_get_mib2_ip_group_src(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10293 10291 {
10294 10292 struct opthdr *optp;
10295 10293 mblk_t *mp2ctl;
10296 10294 ill_t *ill;
10297 10295 ipif_t *ipif;
10298 10296 ilm_t *ilm;
10299 10297 ip_grpsrc_t ips;
10300 10298 mblk_t *mp_tail = NULL;
10301 10299 ill_walk_context_t ctx;
10302 10300 zoneid_t zoneid;
10303 10301 int i;
10304 10302 slist_t *sl;
10305 10303
10306 10304 /*
10307 10305 * make a copy of the original message
10308 10306 */
10309 10307 mp2ctl = copymsg(mpctl);
10310 10308 zoneid = Q_TO_CONN(q)->conn_zoneid;
10311 10309
10312 10310 /* ipGroupSource table */
10313 10311 optp = (struct opthdr *)&mpctl->b_rptr[
10314 10312 sizeof (struct T_optmgmt_ack)];
10315 10313 optp->level = MIB2_IP;
10316 10314 optp->name = EXPER_IP_GROUP_SOURCES;
10317 10315
10318 10316 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10319 10317 ill = ILL_START_WALK_V4(&ctx, ipst);
10320 10318 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10321 10319 /* Make sure the ill isn't going away. */
10322 10320 if (!ill_check_and_refhold(ill))
10323 10321 continue;
10324 10322 rw_exit(&ipst->ips_ill_g_lock);
10325 10323 rw_enter(&ill->ill_mcast_lock, RW_READER);
10326 10324 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10327 10325 sl = ilm->ilm_filter;
10328 10326 if (ilm->ilm_zoneid != zoneid &&
10329 10327 ilm->ilm_zoneid != ALL_ZONES)
10330 10328 continue;
10331 10329 if (SLIST_IS_EMPTY(sl))
10332 10330 continue;
10333 10331
10334 10332 /* Is there an ipif for ilm_ifaddr? */
10335 10333 for (ipif = ill->ill_ipif; ipif != NULL;
10336 10334 ipif = ipif->ipif_next) {
10337 10335 if (!IPIF_IS_CONDEMNED(ipif) &&
10338 10336 ipif->ipif_lcl_addr == ilm->ilm_ifaddr &&
10339 10337 ilm->ilm_ifaddr != INADDR_ANY)
10340 10338 break;
10341 10339 }
10342 10340 if (ipif != NULL) {
10343 10341 ipif_get_name(ipif,
10344 10342 ips.ipGroupSourceIfIndex.o_bytes,
10345 10343 OCTET_LENGTH);
10346 10344 } else {
10347 10345 ill_get_name(ill,
10348 10346 ips.ipGroupSourceIfIndex.o_bytes,
10349 10347 OCTET_LENGTH);
10350 10348 }
10351 10349 ips.ipGroupSourceIfIndex.o_length =
10352 10350 mi_strlen(ips.ipGroupSourceIfIndex.o_bytes);
10353 10351
10354 10352 ips.ipGroupSourceGroup = ilm->ilm_addr;
10355 10353 for (i = 0; i < sl->sl_numsrc; i++) {
10356 10354 if (!IN6_IS_ADDR_V4MAPPED(&sl->sl_addr[i]))
10357 10355 continue;
10358 10356 IN6_V4MAPPED_TO_IPADDR(&sl->sl_addr[i],
10359 10357 ips.ipGroupSourceAddress);
10360 10358 if (snmp_append_data2(mpctl->b_cont, &mp_tail,
10361 10359 (char *)&ips, (int)sizeof (ips)) == 0) {
10362 10360 ip1dbg(("ip_snmp_get_mib2_ip_group_src:"
10363 10361 " failed to allocate %u bytes\n",
10364 10362 (uint_t)sizeof (ips)));
10365 10363 }
10366 10364 }
10367 10365 }
10368 10366 rw_exit(&ill->ill_mcast_lock);
10369 10367 ill_refrele(ill);
10370 10368 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10371 10369 }
10372 10370 rw_exit(&ipst->ips_ill_g_lock);
10373 10371 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10374 10372 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10375 10373 (int)optp->level, (int)optp->name, (int)optp->len));
10376 10374 qreply(q, mpctl);
10377 10375 return (mp2ctl);
10378 10376 }
10379 10377
10380 10378 /* IPv6 multicast filtered sources. */
10381 10379 static mblk_t *
10382 10380 ip_snmp_get_mib2_ip6_group_src(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10383 10381 {
10384 10382 struct opthdr *optp;
10385 10383 mblk_t *mp2ctl;
10386 10384 ill_t *ill;
10387 10385 ilm_t *ilm;
10388 10386 ipv6_grpsrc_t ips6;
10389 10387 mblk_t *mp_tail = NULL;
10390 10388 ill_walk_context_t ctx;
10391 10389 zoneid_t zoneid;
10392 10390 int i;
10393 10391 slist_t *sl;
10394 10392
10395 10393 /*
10396 10394 * make a copy of the original message
10397 10395 */
10398 10396 mp2ctl = copymsg(mpctl);
10399 10397 zoneid = Q_TO_CONN(q)->conn_zoneid;
10400 10398
10401 10399 /* ip6GroupMember table */
10402 10400 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10403 10401 optp->level = MIB2_IP6;
10404 10402 optp->name = EXPER_IP6_GROUP_SOURCES;
10405 10403
10406 10404 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10407 10405 ill = ILL_START_WALK_V6(&ctx, ipst);
10408 10406 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10409 10407 /* Make sure the ill isn't going away. */
10410 10408 if (!ill_check_and_refhold(ill))
10411 10409 continue;
10412 10410 rw_exit(&ipst->ips_ill_g_lock);
10413 10411 /*
10414 10412 * Normally we don't have any members on under IPMP interfaces.
10415 10413 * We report them as a debugging aid.
10416 10414 */
10417 10415 rw_enter(&ill->ill_mcast_lock, RW_READER);
10418 10416 ips6.ipv6GroupSourceIfIndex = ill->ill_phyint->phyint_ifindex;
10419 10417 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10420 10418 sl = ilm->ilm_filter;
10421 10419 if (ilm->ilm_zoneid != zoneid &&
10422 10420 ilm->ilm_zoneid != ALL_ZONES)
10423 10421 continue;
10424 10422 if (SLIST_IS_EMPTY(sl))
10425 10423 continue;
10426 10424 ips6.ipv6GroupSourceGroup = ilm->ilm_v6addr;
10427 10425 for (i = 0; i < sl->sl_numsrc; i++) {
10428 10426 ips6.ipv6GroupSourceAddress = sl->sl_addr[i];
10429 10427 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10430 10428 (char *)&ips6, (int)sizeof (ips6))) {
10431 10429 ip1dbg(("ip_snmp_get_mib2_ip6_"
10432 10430 "group_src: failed to allocate "
10433 10431 "%u bytes\n",
10434 10432 (uint_t)sizeof (ips6)));
10435 10433 }
10436 10434 }
10437 10435 }
10438 10436 rw_exit(&ill->ill_mcast_lock);
10439 10437 ill_refrele(ill);
10440 10438 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10441 10439 }
10442 10440 rw_exit(&ipst->ips_ill_g_lock);
10443 10441
10444 10442 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10445 10443 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10446 10444 (int)optp->level, (int)optp->name, (int)optp->len));
10447 10445 qreply(q, mpctl);
10448 10446 return (mp2ctl);
10449 10447 }
10450 10448
10451 10449 /* Multicast routing virtual interface table. */
10452 10450 static mblk_t *
10453 10451 ip_snmp_get_mib2_virt_multi(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10454 10452 {
10455 10453 struct opthdr *optp;
10456 10454 mblk_t *mp2ctl;
10457 10455
10458 10456 /*
10459 10457 * make a copy of the original message
10460 10458 */
10461 10459 mp2ctl = copymsg(mpctl);
10462 10460
10463 10461 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10464 10462 optp->level = EXPER_DVMRP;
10465 10463 optp->name = EXPER_DVMRP_VIF;
10466 10464 if (!ip_mroute_vif(mpctl->b_cont, ipst)) {
10467 10465 ip0dbg(("ip_mroute_vif: failed\n"));
10468 10466 }
10469 10467 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10470 10468 ip3dbg(("ip_snmp_get_mib2_virt_multi: level %d, name %d, len %d\n",
10471 10469 (int)optp->level, (int)optp->name, (int)optp->len));
10472 10470 qreply(q, mpctl);
10473 10471 return (mp2ctl);
10474 10472 }
10475 10473
10476 10474 /* Multicast routing table. */
10477 10475 static mblk_t *
10478 10476 ip_snmp_get_mib2_multi_rtable(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10479 10477 {
10480 10478 struct opthdr *optp;
10481 10479 mblk_t *mp2ctl;
10482 10480
10483 10481 /*
10484 10482 * make a copy of the original message
10485 10483 */
10486 10484 mp2ctl = copymsg(mpctl);
10487 10485
10488 10486 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10489 10487 optp->level = EXPER_DVMRP;
10490 10488 optp->name = EXPER_DVMRP_MRT;
10491 10489 if (!ip_mroute_mrt(mpctl->b_cont, ipst)) {
10492 10490 ip0dbg(("ip_mroute_mrt: failed\n"));
10493 10491 }
10494 10492 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10495 10493 ip3dbg(("ip_snmp_get_mib2_multi_rtable: level %d, name %d, len %d\n",
10496 10494 (int)optp->level, (int)optp->name, (int)optp->len));
10497 10495 qreply(q, mpctl);
10498 10496 return (mp2ctl);
10499 10497 }
10500 10498
10501 10499 /*
10502 10500 * Return ipRouteEntryTable, ipNetToMediaEntryTable, and ipRouteAttributeTable
10503 10501 * in one IRE walk.
10504 10502 */
10505 10503 static mblk_t *
10506 10504 ip_snmp_get_mib2_ip_route_media(queue_t *q, mblk_t *mpctl, int level,
10507 10505 ip_stack_t *ipst)
10508 10506 {
10509 10507 struct opthdr *optp;
10510 10508 mblk_t *mp2ctl; /* Returned */
10511 10509 mblk_t *mp3ctl; /* nettomedia */
10512 10510 mblk_t *mp4ctl; /* routeattrs */
10513 10511 iproutedata_t ird;
10514 10512 zoneid_t zoneid;
10515 10513
10516 10514 /*
10517 10515 * make copies of the original message
10518 10516 * - mp2ctl is returned unchanged to the caller for its use
10519 10517 * - mpctl is sent upstream as ipRouteEntryTable
10520 10518 * - mp3ctl is sent upstream as ipNetToMediaEntryTable
10521 10519 * - mp4ctl is sent upstream as ipRouteAttributeTable
10522 10520 */
10523 10521 mp2ctl = copymsg(mpctl);
10524 10522 mp3ctl = copymsg(mpctl);
10525 10523 mp4ctl = copymsg(mpctl);
10526 10524 if (mp3ctl == NULL || mp4ctl == NULL) {
10527 10525 freemsg(mp4ctl);
10528 10526 freemsg(mp3ctl);
10529 10527 freemsg(mp2ctl);
10530 10528 freemsg(mpctl);
10531 10529 return (NULL);
10532 10530 }
10533 10531
10534 10532 bzero(&ird, sizeof (ird));
10535 10533
10536 10534 ird.ird_route.lp_head = mpctl->b_cont;
10537 10535 ird.ird_netmedia.lp_head = mp3ctl->b_cont;
10538 10536 ird.ird_attrs.lp_head = mp4ctl->b_cont;
10539 10537 /*
10540 10538 * If the level has been set the special EXPER_IP_AND_ALL_IRES value,
10541 10539 * then also include ire_testhidden IREs and IRE_IF_CLONE. This is
10542 10540 * intended a temporary solution until a proper MIB API is provided
10543 10541 * that provides complete filtering/caller-opt-in.
10544 10542 */
10545 10543 if (level == EXPER_IP_AND_ALL_IRES)
10546 10544 ird.ird_flags |= IRD_REPORT_ALL;
10547 10545
10548 10546 zoneid = Q_TO_CONN(q)->conn_zoneid;
10549 10547 ire_walk_v4(ip_snmp_get2_v4, &ird, zoneid, ipst);
10550 10548
10551 10549 /* ipRouteEntryTable in mpctl */
10552 10550 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10553 10551 optp->level = MIB2_IP;
10554 10552 optp->name = MIB2_IP_ROUTE;
10555 10553 optp->len = msgdsize(ird.ird_route.lp_head);
10556 10554 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10557 10555 (int)optp->level, (int)optp->name, (int)optp->len));
10558 10556 qreply(q, mpctl);
10559 10557
10560 10558 /* ipNetToMediaEntryTable in mp3ctl */
10561 10559 ncec_walk(NULL, ip_snmp_get2_v4_media, &ird, ipst);
10562 10560
10563 10561 optp = (struct opthdr *)&mp3ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10564 10562 optp->level = MIB2_IP;
10565 10563 optp->name = MIB2_IP_MEDIA;
10566 10564 optp->len = msgdsize(ird.ird_netmedia.lp_head);
10567 10565 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10568 10566 (int)optp->level, (int)optp->name, (int)optp->len));
10569 10567 qreply(q, mp3ctl);
10570 10568
10571 10569 /* ipRouteAttributeTable in mp4ctl */
10572 10570 optp = (struct opthdr *)&mp4ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10573 10571 optp->level = MIB2_IP;
10574 10572 optp->name = EXPER_IP_RTATTR;
10575 10573 optp->len = msgdsize(ird.ird_attrs.lp_head);
10576 10574 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10577 10575 (int)optp->level, (int)optp->name, (int)optp->len));
10578 10576 if (optp->len == 0)
10579 10577 freemsg(mp4ctl);
10580 10578 else
10581 10579 qreply(q, mp4ctl);
10582 10580
10583 10581 return (mp2ctl);
10584 10582 }
10585 10583
10586 10584 /*
10587 10585 * Return ipv6RouteEntryTable and ipv6RouteAttributeTable in one IRE walk, and
10588 10586 * ipv6NetToMediaEntryTable in an NDP walk.
10589 10587 */
10590 10588 static mblk_t *
10591 10589 ip_snmp_get_mib2_ip6_route_media(queue_t *q, mblk_t *mpctl, int level,
10592 10590 ip_stack_t *ipst)
10593 10591 {
10594 10592 struct opthdr *optp;
10595 10593 mblk_t *mp2ctl; /* Returned */
10596 10594 mblk_t *mp3ctl; /* nettomedia */
10597 10595 mblk_t *mp4ctl; /* routeattrs */
10598 10596 iproutedata_t ird;
10599 10597 zoneid_t zoneid;
10600 10598
10601 10599 /*
10602 10600 * make copies of the original message
10603 10601 * - mp2ctl is returned unchanged to the caller for its use
10604 10602 * - mpctl is sent upstream as ipv6RouteEntryTable
10605 10603 * - mp3ctl is sent upstream as ipv6NetToMediaEntryTable
10606 10604 * - mp4ctl is sent upstream as ipv6RouteAttributeTable
10607 10605 */
10608 10606 mp2ctl = copymsg(mpctl);
10609 10607 mp3ctl = copymsg(mpctl);
10610 10608 mp4ctl = copymsg(mpctl);
10611 10609 if (mp3ctl == NULL || mp4ctl == NULL) {
10612 10610 freemsg(mp4ctl);
10613 10611 freemsg(mp3ctl);
10614 10612 freemsg(mp2ctl);
10615 10613 freemsg(mpctl);
10616 10614 return (NULL);
10617 10615 }
10618 10616
10619 10617 bzero(&ird, sizeof (ird));
10620 10618
10621 10619 ird.ird_route.lp_head = mpctl->b_cont;
10622 10620 ird.ird_netmedia.lp_head = mp3ctl->b_cont;
10623 10621 ird.ird_attrs.lp_head = mp4ctl->b_cont;
10624 10622 /*
10625 10623 * If the level has been set the special EXPER_IP_AND_ALL_IRES value,
10626 10624 * then also include ire_testhidden IREs and IRE_IF_CLONE. This is
10627 10625 * intended a temporary solution until a proper MIB API is provided
10628 10626 * that provides complete filtering/caller-opt-in.
10629 10627 */
10630 10628 if (level == EXPER_IP_AND_ALL_IRES)
10631 10629 ird.ird_flags |= IRD_REPORT_ALL;
10632 10630
10633 10631 zoneid = Q_TO_CONN(q)->conn_zoneid;
10634 10632 ire_walk_v6(ip_snmp_get2_v6_route, &ird, zoneid, ipst);
10635 10633
10636 10634 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10637 10635 optp->level = MIB2_IP6;
10638 10636 optp->name = MIB2_IP6_ROUTE;
10639 10637 optp->len = msgdsize(ird.ird_route.lp_head);
10640 10638 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10641 10639 (int)optp->level, (int)optp->name, (int)optp->len));
10642 10640 qreply(q, mpctl);
10643 10641
10644 10642 /* ipv6NetToMediaEntryTable in mp3ctl */
10645 10643 ncec_walk(NULL, ip_snmp_get2_v6_media, &ird, ipst);
10646 10644
10647 10645 optp = (struct opthdr *)&mp3ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10648 10646 optp->level = MIB2_IP6;
10649 10647 optp->name = MIB2_IP6_MEDIA;
10650 10648 optp->len = msgdsize(ird.ird_netmedia.lp_head);
10651 10649 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10652 10650 (int)optp->level, (int)optp->name, (int)optp->len));
10653 10651 qreply(q, mp3ctl);
10654 10652
10655 10653 /* ipv6RouteAttributeTable in mp4ctl */
10656 10654 optp = (struct opthdr *)&mp4ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10657 10655 optp->level = MIB2_IP6;
10658 10656 optp->name = EXPER_IP_RTATTR;
10659 10657 optp->len = msgdsize(ird.ird_attrs.lp_head);
10660 10658 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10661 10659 (int)optp->level, (int)optp->name, (int)optp->len));
10662 10660 if (optp->len == 0)
10663 10661 freemsg(mp4ctl);
10664 10662 else
10665 10663 qreply(q, mp4ctl);
10666 10664
10667 10665 return (mp2ctl);
10668 10666 }
10669 10667
10670 10668 /*
10671 10669 * IPv6 mib: One per ill
10672 10670 */
10673 10671 static mblk_t *
10674 10672 ip_snmp_get_mib2_ip6(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
10675 10673 boolean_t legacy_req)
10676 10674 {
10677 10675 struct opthdr *optp;
10678 10676 mblk_t *mp2ctl;
10679 10677 ill_t *ill;
10680 10678 ill_walk_context_t ctx;
10681 10679 mblk_t *mp_tail = NULL;
10682 10680 mib2_ipv6AddrEntry_t mae6;
10683 10681 mib2_ipIfStatsEntry_t *ise;
10684 10682 size_t ise_size, iae_size;
10685 10683
10686 10684 /*
10687 10685 * Make a copy of the original message
10688 10686 */
10689 10687 mp2ctl = copymsg(mpctl);
10690 10688
10691 10689 /* fixed length IPv6 structure ... */
10692 10690
10693 10691 if (legacy_req) {
10694 10692 ise_size = LEGACY_MIB_SIZE(&ipst->ips_ip6_mib,
10695 10693 mib2_ipIfStatsEntry_t);
10696 10694 iae_size = LEGACY_MIB_SIZE(&mae6, mib2_ipv6AddrEntry_t);
10697 10695 } else {
10698 10696 ise_size = sizeof (mib2_ipIfStatsEntry_t);
10699 10697 iae_size = sizeof (mib2_ipv6AddrEntry_t);
10700 10698 }
10701 10699
10702 10700 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10703 10701 optp->level = MIB2_IP6;
10704 10702 optp->name = 0;
10705 10703 /* Include "unknown interface" ip6_mib */
10706 10704 ipst->ips_ip6_mib.ipIfStatsIPVersion = MIB2_INETADDRESSTYPE_ipv6;
10707 10705 ipst->ips_ip6_mib.ipIfStatsIfIndex =
10708 10706 MIB2_UNKNOWN_INTERFACE; /* Flag to netstat */
10709 10707 SET_MIB(ipst->ips_ip6_mib.ipIfStatsForwarding,
10710 10708 ipst->ips_ipv6_forwarding ? 1 : 2);
10711 10709 SET_MIB(ipst->ips_ip6_mib.ipIfStatsDefaultHopLimit,
10712 10710 ipst->ips_ipv6_def_hops);
10713 10711 SET_MIB(ipst->ips_ip6_mib.ipIfStatsEntrySize,
10714 10712 sizeof (mib2_ipIfStatsEntry_t));
10715 10713 SET_MIB(ipst->ips_ip6_mib.ipIfStatsAddrEntrySize,
10716 10714 sizeof (mib2_ipv6AddrEntry_t));
10717 10715 SET_MIB(ipst->ips_ip6_mib.ipIfStatsRouteEntrySize,
10718 10716 sizeof (mib2_ipv6RouteEntry_t));
10719 10717 SET_MIB(ipst->ips_ip6_mib.ipIfStatsNetToMediaEntrySize,
10720 10718 sizeof (mib2_ipv6NetToMediaEntry_t));
10721 10719 SET_MIB(ipst->ips_ip6_mib.ipIfStatsMemberEntrySize,
10722 10720 sizeof (ipv6_member_t));
10723 10721 SET_MIB(ipst->ips_ip6_mib.ipIfStatsGroupSourceEntrySize,
10724 10722 sizeof (ipv6_grpsrc_t));
10725 10723
10726 10724 /*
10727 10725 * Synchronize 64- and 32-bit counters
10728 10726 */
10729 10727 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInReceives,
10730 10728 ipIfStatsHCInReceives);
10731 10729 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInDelivers,
10732 10730 ipIfStatsHCInDelivers);
10733 10731 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutRequests,
10734 10732 ipIfStatsHCOutRequests);
10735 10733 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutForwDatagrams,
10736 10734 ipIfStatsHCOutForwDatagrams);
10737 10735 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutMcastPkts,
10738 10736 ipIfStatsHCOutMcastPkts);
10739 10737 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInMcastPkts,
10740 10738 ipIfStatsHCInMcastPkts);
10741 10739
10742 10740 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10743 10741 (char *)&ipst->ips_ip6_mib, (int)ise_size)) {
10744 10742 ip1dbg(("ip_snmp_get_mib2_ip6: failed to allocate %u bytes\n",
10745 10743 (uint_t)ise_size));
10746 10744 } else if (legacy_req) {
10747 10745 /* Adjust the EntrySize fields for legacy requests. */
10748 10746 ise =
10749 10747 (mib2_ipIfStatsEntry_t *)(mp_tail->b_wptr - (int)ise_size);
10750 10748 SET_MIB(ise->ipIfStatsEntrySize, ise_size);
10751 10749 SET_MIB(ise->ipIfStatsAddrEntrySize, iae_size);
10752 10750 }
10753 10751
10754 10752 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10755 10753 ill = ILL_START_WALK_V6(&ctx, ipst);
10756 10754 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10757 10755 ill->ill_ip_mib->ipIfStatsIfIndex =
10758 10756 ill->ill_phyint->phyint_ifindex;
10759 10757 SET_MIB(ill->ill_ip_mib->ipIfStatsForwarding,
10760 10758 ipst->ips_ipv6_forwarding ? 1 : 2);
10761 10759 SET_MIB(ill->ill_ip_mib->ipIfStatsDefaultHopLimit,
10762 10760 ill->ill_max_hops);
10763 10761
10764 10762 /*
10765 10763 * Synchronize 64- and 32-bit counters
10766 10764 */
10767 10765 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInReceives,
10768 10766 ipIfStatsHCInReceives);
10769 10767 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInDelivers,
10770 10768 ipIfStatsHCInDelivers);
10771 10769 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutRequests,
10772 10770 ipIfStatsHCOutRequests);
10773 10771 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutForwDatagrams,
10774 10772 ipIfStatsHCOutForwDatagrams);
10775 10773 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutMcastPkts,
10776 10774 ipIfStatsHCOutMcastPkts);
10777 10775 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInMcastPkts,
10778 10776 ipIfStatsHCInMcastPkts);
10779 10777
10780 10778 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10781 10779 (char *)ill->ill_ip_mib, (int)ise_size)) {
10782 10780 ip1dbg(("ip_snmp_get_mib2_ip6: failed to allocate "
10783 10781 "%u bytes\n", (uint_t)ise_size));
10784 10782 } else if (legacy_req) {
10785 10783 /* Adjust the EntrySize fields for legacy requests. */
10786 10784 ise = (mib2_ipIfStatsEntry_t *)(mp_tail->b_wptr -
10787 10785 (int)ise_size);
10788 10786 SET_MIB(ise->ipIfStatsEntrySize, ise_size);
10789 10787 SET_MIB(ise->ipIfStatsAddrEntrySize, iae_size);
10790 10788 }
10791 10789 }
10792 10790 rw_exit(&ipst->ips_ill_g_lock);
10793 10791
10794 10792 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10795 10793 ip3dbg(("ip_snmp_get_mib2_ip6: level %d, name %d, len %d\n",
10796 10794 (int)optp->level, (int)optp->name, (int)optp->len));
10797 10795 qreply(q, mpctl);
10798 10796 return (mp2ctl);
10799 10797 }
10800 10798
10801 10799 /*
10802 10800 * ICMPv6 mib: One per ill
10803 10801 */
10804 10802 static mblk_t *
10805 10803 ip_snmp_get_mib2_icmp6(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10806 10804 {
10807 10805 struct opthdr *optp;
10808 10806 mblk_t *mp2ctl;
10809 10807 ill_t *ill;
10810 10808 ill_walk_context_t ctx;
10811 10809 mblk_t *mp_tail = NULL;
10812 10810 /*
10813 10811 * Make a copy of the original message
10814 10812 */
10815 10813 mp2ctl = copymsg(mpctl);
10816 10814
10817 10815 /* fixed length ICMPv6 structure ... */
10818 10816
10819 10817 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10820 10818 optp->level = MIB2_ICMP6;
10821 10819 optp->name = 0;
10822 10820 /* Include "unknown interface" icmp6_mib */
10823 10821 ipst->ips_icmp6_mib.ipv6IfIcmpIfIndex =
10824 10822 MIB2_UNKNOWN_INTERFACE; /* netstat flag */
10825 10823 ipst->ips_icmp6_mib.ipv6IfIcmpEntrySize =
10826 10824 sizeof (mib2_ipv6IfIcmpEntry_t);
10827 10825 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10828 10826 (char *)&ipst->ips_icmp6_mib,
10829 10827 (int)sizeof (ipst->ips_icmp6_mib))) {
10830 10828 ip1dbg(("ip_snmp_get_mib2_icmp6: failed to allocate %u bytes\n",
10831 10829 (uint_t)sizeof (ipst->ips_icmp6_mib)));
10832 10830 }
10833 10831
10834 10832 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10835 10833 ill = ILL_START_WALK_V6(&ctx, ipst);
10836 10834 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10837 10835 ill->ill_icmp6_mib->ipv6IfIcmpIfIndex =
10838 10836 ill->ill_phyint->phyint_ifindex;
10839 10837 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10840 10838 (char *)ill->ill_icmp6_mib,
10841 10839 (int)sizeof (*ill->ill_icmp6_mib))) {
10842 10840 ip1dbg(("ip_snmp_get_mib2_icmp6: failed to allocate "
10843 10841 "%u bytes\n",
10844 10842 (uint_t)sizeof (*ill->ill_icmp6_mib)));
10845 10843 }
10846 10844 }
10847 10845 rw_exit(&ipst->ips_ill_g_lock);
10848 10846
10849 10847 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10850 10848 ip3dbg(("ip_snmp_get_mib2_icmp6: level %d, name %d, len %d\n",
10851 10849 (int)optp->level, (int)optp->name, (int)optp->len));
10852 10850 qreply(q, mpctl);
10853 10851 return (mp2ctl);
10854 10852 }
10855 10853
10856 10854 /*
10857 10855 * ire_walk routine to create both ipRouteEntryTable and
10858 10856 * ipRouteAttributeTable in one IRE walk
10859 10857 */
10860 10858 static void
10861 10859 ip_snmp_get2_v4(ire_t *ire, iproutedata_t *ird)
10862 10860 {
10863 10861 ill_t *ill;
10864 10862 mib2_ipRouteEntry_t *re;
10865 10863 mib2_ipAttributeEntry_t iaes;
10866 10864 tsol_ire_gw_secattr_t *attrp;
10867 10865 tsol_gc_t *gc = NULL;
10868 10866 tsol_gcgrp_t *gcgrp = NULL;
10869 10867 ip_stack_t *ipst = ire->ire_ipst;
10870 10868
10871 10869 ASSERT(ire->ire_ipversion == IPV4_VERSION);
10872 10870
10873 10871 if (!(ird->ird_flags & IRD_REPORT_ALL)) {
10874 10872 if (ire->ire_testhidden)
10875 10873 return;
10876 10874 if (ire->ire_type & IRE_IF_CLONE)
10877 10875 return;
10878 10876 }
10879 10877
10880 10878 if ((re = kmem_zalloc(sizeof (*re), KM_NOSLEEP)) == NULL)
10881 10879 return;
10882 10880
10883 10881 if ((attrp = ire->ire_gw_secattr) != NULL) {
10884 10882 mutex_enter(&attrp->igsa_lock);
10885 10883 if ((gc = attrp->igsa_gc) != NULL) {
10886 10884 gcgrp = gc->gc_grp;
10887 10885 ASSERT(gcgrp != NULL);
10888 10886 rw_enter(&gcgrp->gcgrp_rwlock, RW_READER);
10889 10887 }
10890 10888 mutex_exit(&attrp->igsa_lock);
10891 10889 }
10892 10890 /*
10893 10891 * Return all IRE types for route table... let caller pick and choose
10894 10892 */
10895 10893 re->ipRouteDest = ire->ire_addr;
10896 10894 ill = ire->ire_ill;
10897 10895 re->ipRouteIfIndex.o_length = 0;
10898 10896 if (ill != NULL) {
10899 10897 ill_get_name(ill, re->ipRouteIfIndex.o_bytes, OCTET_LENGTH);
10900 10898 re->ipRouteIfIndex.o_length =
10901 10899 mi_strlen(re->ipRouteIfIndex.o_bytes);
10902 10900 }
10903 10901 re->ipRouteMetric1 = -1;
10904 10902 re->ipRouteMetric2 = -1;
10905 10903 re->ipRouteMetric3 = -1;
10906 10904 re->ipRouteMetric4 = -1;
10907 10905
10908 10906 re->ipRouteNextHop = ire->ire_gateway_addr;
10909 10907 /* indirect(4), direct(3), or invalid(2) */
10910 10908 if (ire->ire_flags & (RTF_REJECT | RTF_BLACKHOLE))
10911 10909 re->ipRouteType = 2;
10912 10910 else if (ire->ire_type & IRE_ONLINK)
10913 10911 re->ipRouteType = 3;
10914 10912 else
10915 10913 re->ipRouteType = 4;
10916 10914
10917 10915 re->ipRouteProto = -1;
10918 10916 re->ipRouteAge = gethrestime_sec() - ire->ire_create_time;
10919 10917 re->ipRouteMask = ire->ire_mask;
10920 10918 re->ipRouteMetric5 = -1;
10921 10919 re->ipRouteInfo.re_max_frag = ire->ire_metrics.iulp_mtu;
10922 10920 if (ire->ire_ill != NULL && re->ipRouteInfo.re_max_frag == 0)
10923 10921 re->ipRouteInfo.re_max_frag = ire->ire_ill->ill_mtu;
10924 10922
10925 10923 re->ipRouteInfo.re_frag_flag = 0;
10926 10924 re->ipRouteInfo.re_rtt = 0;
10927 10925 re->ipRouteInfo.re_src_addr = 0;
10928 10926 re->ipRouteInfo.re_ref = ire->ire_refcnt;
10929 10927 re->ipRouteInfo.re_obpkt = ire->ire_ob_pkt_count;
10930 10928 re->ipRouteInfo.re_ibpkt = ire->ire_ib_pkt_count;
10931 10929 re->ipRouteInfo.re_flags = ire->ire_flags;
10932 10930
10933 10931 /* Add the IRE_IF_CLONE's counters to their parent IRE_INTERFACE */
10934 10932 if (ire->ire_type & IRE_INTERFACE) {
10935 10933 ire_t *child;
10936 10934
10937 10935 rw_enter(&ipst->ips_ire_dep_lock, RW_READER);
10938 10936 child = ire->ire_dep_children;
10939 10937 while (child != NULL) {
10940 10938 re->ipRouteInfo.re_obpkt += child->ire_ob_pkt_count;
10941 10939 re->ipRouteInfo.re_ibpkt += child->ire_ib_pkt_count;
10942 10940 child = child->ire_dep_sib_next;
10943 10941 }
10944 10942 rw_exit(&ipst->ips_ire_dep_lock);
10945 10943 }
10946 10944
10947 10945 if (ire->ire_flags & RTF_DYNAMIC) {
10948 10946 re->ipRouteInfo.re_ire_type = IRE_HOST_REDIRECT;
10949 10947 } else {
10950 10948 re->ipRouteInfo.re_ire_type = ire->ire_type;
10951 10949 }
10952 10950
10953 10951 if (!snmp_append_data2(ird->ird_route.lp_head, &ird->ird_route.lp_tail,
10954 10952 (char *)re, (int)sizeof (*re))) {
10955 10953 ip1dbg(("ip_snmp_get2_v4: failed to allocate %u bytes\n",
10956 10954 (uint_t)sizeof (*re)));
10957 10955 }
10958 10956
10959 10957 if (gc != NULL) {
10960 10958 iaes.iae_routeidx = ird->ird_idx;
10961 10959 iaes.iae_doi = gc->gc_db->gcdb_doi;
10962 10960 iaes.iae_slrange = gc->gc_db->gcdb_slrange;
10963 10961
10964 10962 if (!snmp_append_data2(ird->ird_attrs.lp_head,
10965 10963 &ird->ird_attrs.lp_tail, (char *)&iaes, sizeof (iaes))) {
10966 10964 ip1dbg(("ip_snmp_get2_v4: failed to allocate %u "
10967 10965 "bytes\n", (uint_t)sizeof (iaes)));
10968 10966 }
10969 10967 }
10970 10968
10971 10969 /* bump route index for next pass */
10972 10970 ird->ird_idx++;
10973 10971
10974 10972 kmem_free(re, sizeof (*re));
10975 10973 if (gcgrp != NULL)
10976 10974 rw_exit(&gcgrp->gcgrp_rwlock);
10977 10975 }
10978 10976
10979 10977 /*
10980 10978 * ire_walk routine to create ipv6RouteEntryTable and ipRouteEntryTable.
10981 10979 */
10982 10980 static void
10983 10981 ip_snmp_get2_v6_route(ire_t *ire, iproutedata_t *ird)
10984 10982 {
10985 10983 ill_t *ill;
10986 10984 mib2_ipv6RouteEntry_t *re;
10987 10985 mib2_ipAttributeEntry_t iaes;
10988 10986 tsol_ire_gw_secattr_t *attrp;
10989 10987 tsol_gc_t *gc = NULL;
10990 10988 tsol_gcgrp_t *gcgrp = NULL;
10991 10989 ip_stack_t *ipst = ire->ire_ipst;
10992 10990
10993 10991 ASSERT(ire->ire_ipversion == IPV6_VERSION);
10994 10992
10995 10993 if (!(ird->ird_flags & IRD_REPORT_ALL)) {
10996 10994 if (ire->ire_testhidden)
10997 10995 return;
10998 10996 if (ire->ire_type & IRE_IF_CLONE)
10999 10997 return;
11000 10998 }
11001 10999
11002 11000 if ((re = kmem_zalloc(sizeof (*re), KM_NOSLEEP)) == NULL)
11003 11001 return;
11004 11002
11005 11003 if ((attrp = ire->ire_gw_secattr) != NULL) {
11006 11004 mutex_enter(&attrp->igsa_lock);
11007 11005 if ((gc = attrp->igsa_gc) != NULL) {
11008 11006 gcgrp = gc->gc_grp;
11009 11007 ASSERT(gcgrp != NULL);
11010 11008 rw_enter(&gcgrp->gcgrp_rwlock, RW_READER);
11011 11009 }
11012 11010 mutex_exit(&attrp->igsa_lock);
11013 11011 }
11014 11012 /*
11015 11013 * Return all IRE types for route table... let caller pick and choose
11016 11014 */
11017 11015 re->ipv6RouteDest = ire->ire_addr_v6;
11018 11016 re->ipv6RoutePfxLength = ip_mask_to_plen_v6(&ire->ire_mask_v6);
11019 11017 re->ipv6RouteIndex = 0; /* Unique when multiple with same dest/plen */
11020 11018 re->ipv6RouteIfIndex.o_length = 0;
11021 11019 ill = ire->ire_ill;
11022 11020 if (ill != NULL) {
11023 11021 ill_get_name(ill, re->ipv6RouteIfIndex.o_bytes, OCTET_LENGTH);
11024 11022 re->ipv6RouteIfIndex.o_length =
11025 11023 mi_strlen(re->ipv6RouteIfIndex.o_bytes);
11026 11024 }
11027 11025
11028 11026 ASSERT(!(ire->ire_type & IRE_BROADCAST));
11029 11027
11030 11028 mutex_enter(&ire->ire_lock);
11031 11029 re->ipv6RouteNextHop = ire->ire_gateway_addr_v6;
11032 11030 mutex_exit(&ire->ire_lock);
11033 11031
11034 11032 /* remote(4), local(3), or discard(2) */
11035 11033 if (ire->ire_flags & (RTF_REJECT | RTF_BLACKHOLE))
11036 11034 re->ipv6RouteType = 2;
11037 11035 else if (ire->ire_type & IRE_ONLINK)
11038 11036 re->ipv6RouteType = 3;
11039 11037 else
11040 11038 re->ipv6RouteType = 4;
11041 11039
11042 11040 re->ipv6RouteProtocol = -1;
11043 11041 re->ipv6RoutePolicy = 0;
11044 11042 re->ipv6RouteAge = gethrestime_sec() - ire->ire_create_time;
11045 11043 re->ipv6RouteNextHopRDI = 0;
11046 11044 re->ipv6RouteWeight = 0;
11047 11045 re->ipv6RouteMetric = 0;
11048 11046 re->ipv6RouteInfo.re_max_frag = ire->ire_metrics.iulp_mtu;
11049 11047 if (ire->ire_ill != NULL && re->ipv6RouteInfo.re_max_frag == 0)
11050 11048 re->ipv6RouteInfo.re_max_frag = ire->ire_ill->ill_mtu;
11051 11049
11052 11050 re->ipv6RouteInfo.re_frag_flag = 0;
11053 11051 re->ipv6RouteInfo.re_rtt = 0;
11054 11052 re->ipv6RouteInfo.re_src_addr = ipv6_all_zeros;
11055 11053 re->ipv6RouteInfo.re_obpkt = ire->ire_ob_pkt_count;
11056 11054 re->ipv6RouteInfo.re_ibpkt = ire->ire_ib_pkt_count;
11057 11055 re->ipv6RouteInfo.re_ref = ire->ire_refcnt;
11058 11056 re->ipv6RouteInfo.re_flags = ire->ire_flags;
11059 11057
11060 11058 /* Add the IRE_IF_CLONE's counters to their parent IRE_INTERFACE */
11061 11059 if (ire->ire_type & IRE_INTERFACE) {
11062 11060 ire_t *child;
11063 11061
11064 11062 rw_enter(&ipst->ips_ire_dep_lock, RW_READER);
11065 11063 child = ire->ire_dep_children;
11066 11064 while (child != NULL) {
11067 11065 re->ipv6RouteInfo.re_obpkt += child->ire_ob_pkt_count;
11068 11066 re->ipv6RouteInfo.re_ibpkt += child->ire_ib_pkt_count;
11069 11067 child = child->ire_dep_sib_next;
11070 11068 }
11071 11069 rw_exit(&ipst->ips_ire_dep_lock);
11072 11070 }
11073 11071 if (ire->ire_flags & RTF_DYNAMIC) {
11074 11072 re->ipv6RouteInfo.re_ire_type = IRE_HOST_REDIRECT;
11075 11073 } else {
11076 11074 re->ipv6RouteInfo.re_ire_type = ire->ire_type;
11077 11075 }
11078 11076
11079 11077 if (!snmp_append_data2(ird->ird_route.lp_head, &ird->ird_route.lp_tail,
11080 11078 (char *)re, (int)sizeof (*re))) {
11081 11079 ip1dbg(("ip_snmp_get2_v6: failed to allocate %u bytes\n",
11082 11080 (uint_t)sizeof (*re)));
11083 11081 }
11084 11082
11085 11083 if (gc != NULL) {
11086 11084 iaes.iae_routeidx = ird->ird_idx;
11087 11085 iaes.iae_doi = gc->gc_db->gcdb_doi;
11088 11086 iaes.iae_slrange = gc->gc_db->gcdb_slrange;
11089 11087
11090 11088 if (!snmp_append_data2(ird->ird_attrs.lp_head,
11091 11089 &ird->ird_attrs.lp_tail, (char *)&iaes, sizeof (iaes))) {
11092 11090 ip1dbg(("ip_snmp_get2_v6: failed to allocate %u "
11093 11091 "bytes\n", (uint_t)sizeof (iaes)));
11094 11092 }
11095 11093 }
11096 11094
11097 11095 /* bump route index for next pass */
11098 11096 ird->ird_idx++;
11099 11097
11100 11098 kmem_free(re, sizeof (*re));
11101 11099 if (gcgrp != NULL)
11102 11100 rw_exit(&gcgrp->gcgrp_rwlock);
11103 11101 }
11104 11102
11105 11103 /*
11106 11104 * ncec_walk routine to create ipv6NetToMediaEntryTable
11107 11105 */
11108 11106 static void
11109 11107 ip_snmp_get2_v6_media(ncec_t *ncec, void *ptr)
11110 11108 {
11111 11109 iproutedata_t *ird = ptr;
11112 11110 ill_t *ill;
11113 11111 mib2_ipv6NetToMediaEntry_t ntme;
11114 11112
11115 11113 ill = ncec->ncec_ill;
11116 11114 /* skip arpce entries, and loopback ncec entries */
11117 11115 if (ill->ill_isv6 == B_FALSE || ill->ill_net_type == IRE_LOOPBACK)
11118 11116 return;
11119 11117 /*
11120 11118 * Neighbor cache entry attached to IRE with on-link
11121 11119 * destination.
11122 11120 * We report all IPMP groups on ncec_ill which is normally the upper.
11123 11121 */
11124 11122 ntme.ipv6NetToMediaIfIndex = ill->ill_phyint->phyint_ifindex;
11125 11123 ntme.ipv6NetToMediaNetAddress = ncec->ncec_addr;
11126 11124 ntme.ipv6NetToMediaPhysAddress.o_length = ill->ill_phys_addr_length;
11127 11125 if (ncec->ncec_lladdr != NULL) {
11128 11126 bcopy(ncec->ncec_lladdr, ntme.ipv6NetToMediaPhysAddress.o_bytes,
11129 11127 ntme.ipv6NetToMediaPhysAddress.o_length);
11130 11128 }
11131 11129 /*
11132 11130 * Note: Returns ND_* states. Should be:
11133 11131 * reachable(1), stale(2), delay(3), probe(4),
11134 11132 * invalid(5), unknown(6)
11135 11133 */
11136 11134 ntme.ipv6NetToMediaState = ncec->ncec_state;
11137 11135 ntme.ipv6NetToMediaLastUpdated = 0;
11138 11136
11139 11137 /* other(1), dynamic(2), static(3), local(4) */
11140 11138 if (NCE_MYADDR(ncec)) {
11141 11139 ntme.ipv6NetToMediaType = 4;
11142 11140 } else if (ncec->ncec_flags & NCE_F_PUBLISH) {
11143 11141 ntme.ipv6NetToMediaType = 1; /* proxy */
11144 11142 } else if (ncec->ncec_flags & NCE_F_STATIC) {
11145 11143 ntme.ipv6NetToMediaType = 3;
11146 11144 } else if (ncec->ncec_flags & (NCE_F_MCAST|NCE_F_BCAST)) {
11147 11145 ntme.ipv6NetToMediaType = 1;
11148 11146 } else {
11149 11147 ntme.ipv6NetToMediaType = 2;
11150 11148 }
11151 11149
11152 11150 if (!snmp_append_data2(ird->ird_netmedia.lp_head,
11153 11151 &ird->ird_netmedia.lp_tail, (char *)&ntme, sizeof (ntme))) {
11154 11152 ip1dbg(("ip_snmp_get2_v6_media: failed to allocate %u bytes\n",
11155 11153 (uint_t)sizeof (ntme)));
11156 11154 }
11157 11155 }
11158 11156
11159 11157 int
11160 11158 nce2ace(ncec_t *ncec)
11161 11159 {
11162 11160 int flags = 0;
11163 11161
11164 11162 if (NCE_ISREACHABLE(ncec))
11165 11163 flags |= ACE_F_RESOLVED;
11166 11164 if (ncec->ncec_flags & NCE_F_AUTHORITY)
11167 11165 flags |= ACE_F_AUTHORITY;
11168 11166 if (ncec->ncec_flags & NCE_F_PUBLISH)
11169 11167 flags |= ACE_F_PUBLISH;
11170 11168 if ((ncec->ncec_flags & NCE_F_NONUD) != 0)
11171 11169 flags |= ACE_F_PERMANENT;
11172 11170 if (NCE_MYADDR(ncec))
11173 11171 flags |= (ACE_F_MYADDR | ACE_F_AUTHORITY);
11174 11172 if (ncec->ncec_flags & NCE_F_UNVERIFIED)
11175 11173 flags |= ACE_F_UNVERIFIED;
11176 11174 if (ncec->ncec_flags & NCE_F_AUTHORITY)
11177 11175 flags |= ACE_F_AUTHORITY;
11178 11176 if (ncec->ncec_flags & NCE_F_DELAYED)
11179 11177 flags |= ACE_F_DELAYED;
11180 11178 return (flags);
11181 11179 }
11182 11180
11183 11181 /*
11184 11182 * ncec_walk routine to create ipNetToMediaEntryTable
11185 11183 */
11186 11184 static void
11187 11185 ip_snmp_get2_v4_media(ncec_t *ncec, void *ptr)
11188 11186 {
11189 11187 iproutedata_t *ird = ptr;
11190 11188 ill_t *ill;
11191 11189 mib2_ipNetToMediaEntry_t ntme;
11192 11190 const char *name = "unknown";
11193 11191 ipaddr_t ncec_addr;
11194 11192
11195 11193 ill = ncec->ncec_ill;
11196 11194 if (ill->ill_isv6 || (ncec->ncec_flags & NCE_F_BCAST) ||
11197 11195 ill->ill_net_type == IRE_LOOPBACK)
11198 11196 return;
11199 11197
11200 11198 /* We report all IPMP groups on ncec_ill which is normally the upper. */
11201 11199 name = ill->ill_name;
11202 11200 /* Based on RFC 4293: other(1), inval(2), dyn(3), stat(4) */
11203 11201 if (NCE_MYADDR(ncec)) {
11204 11202 ntme.ipNetToMediaType = 4;
11205 11203 } else if (ncec->ncec_flags & (NCE_F_MCAST|NCE_F_BCAST|NCE_F_PUBLISH)) {
11206 11204 ntme.ipNetToMediaType = 1;
11207 11205 } else {
11208 11206 ntme.ipNetToMediaType = 3;
11209 11207 }
11210 11208 ntme.ipNetToMediaIfIndex.o_length = MIN(OCTET_LENGTH, strlen(name));
11211 11209 bcopy(name, ntme.ipNetToMediaIfIndex.o_bytes,
11212 11210 ntme.ipNetToMediaIfIndex.o_length);
11213 11211
11214 11212 IN6_V4MAPPED_TO_IPADDR(&ncec->ncec_addr, ncec_addr);
11215 11213 bcopy(&ncec_addr, &ntme.ipNetToMediaNetAddress, sizeof (ncec_addr));
11216 11214
11217 11215 ntme.ipNetToMediaInfo.ntm_mask.o_length = sizeof (ipaddr_t);
11218 11216 ncec_addr = INADDR_BROADCAST;
11219 11217 bcopy(&ncec_addr, ntme.ipNetToMediaInfo.ntm_mask.o_bytes,
11220 11218 sizeof (ncec_addr));
11221 11219 /*
11222 11220 * map all the flags to the ACE counterpart.
11223 11221 */
11224 11222 ntme.ipNetToMediaInfo.ntm_flags = nce2ace(ncec);
11225 11223
11226 11224 ntme.ipNetToMediaPhysAddress.o_length =
11227 11225 MIN(OCTET_LENGTH, ill->ill_phys_addr_length);
11228 11226
11229 11227 if (!NCE_ISREACHABLE(ncec))
11230 11228 ntme.ipNetToMediaPhysAddress.o_length = 0;
11231 11229 else {
11232 11230 if (ncec->ncec_lladdr != NULL) {
11233 11231 bcopy(ncec->ncec_lladdr,
11234 11232 ntme.ipNetToMediaPhysAddress.o_bytes,
11235 11233 ntme.ipNetToMediaPhysAddress.o_length);
11236 11234 }
11237 11235 }
11238 11236
11239 11237 if (!snmp_append_data2(ird->ird_netmedia.lp_head,
11240 11238 &ird->ird_netmedia.lp_tail, (char *)&ntme, sizeof (ntme))) {
11241 11239 ip1dbg(("ip_snmp_get2_v4_media: failed to allocate %u bytes\n",
11242 11240 (uint_t)sizeof (ntme)));
11243 11241 }
11244 11242 }
11245 11243
11246 11244 /*
11247 11245 * return (0) if invalid set request, 1 otherwise, including non-tcp requests
11248 11246 */
11249 11247 /* ARGSUSED */
11250 11248 int
11251 11249 ip_snmp_set(queue_t *q, int level, int name, uchar_t *ptr, int len)
11252 11250 {
11253 11251 switch (level) {
11254 11252 case MIB2_IP:
11255 11253 case MIB2_ICMP:
11256 11254 switch (name) {
11257 11255 default:
11258 11256 break;
11259 11257 }
11260 11258 return (1);
11261 11259 default:
11262 11260 return (1);
11263 11261 }
11264 11262 }
11265 11263
11266 11264 /*
11267 11265 * When there exists both a 64- and 32-bit counter of a particular type
11268 11266 * (i.e., InReceives), only the 64-bit counters are added.
11269 11267 */
11270 11268 void
11271 11269 ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *o1, mib2_ipIfStatsEntry_t *o2)
11272 11270 {
11273 11271 UPDATE_MIB(o1, ipIfStatsInHdrErrors, o2->ipIfStatsInHdrErrors);
11274 11272 UPDATE_MIB(o1, ipIfStatsInTooBigErrors, o2->ipIfStatsInTooBigErrors);
11275 11273 UPDATE_MIB(o1, ipIfStatsInNoRoutes, o2->ipIfStatsInNoRoutes);
11276 11274 UPDATE_MIB(o1, ipIfStatsInAddrErrors, o2->ipIfStatsInAddrErrors);
11277 11275 UPDATE_MIB(o1, ipIfStatsInUnknownProtos, o2->ipIfStatsInUnknownProtos);
11278 11276 UPDATE_MIB(o1, ipIfStatsInTruncatedPkts, o2->ipIfStatsInTruncatedPkts);
11279 11277 UPDATE_MIB(o1, ipIfStatsInDiscards, o2->ipIfStatsInDiscards);
11280 11278 UPDATE_MIB(o1, ipIfStatsOutDiscards, o2->ipIfStatsOutDiscards);
11281 11279 UPDATE_MIB(o1, ipIfStatsOutFragOKs, o2->ipIfStatsOutFragOKs);
11282 11280 UPDATE_MIB(o1, ipIfStatsOutFragFails, o2->ipIfStatsOutFragFails);
11283 11281 UPDATE_MIB(o1, ipIfStatsOutFragCreates, o2->ipIfStatsOutFragCreates);
11284 11282 UPDATE_MIB(o1, ipIfStatsReasmReqds, o2->ipIfStatsReasmReqds);
11285 11283 UPDATE_MIB(o1, ipIfStatsReasmOKs, o2->ipIfStatsReasmOKs);
11286 11284 UPDATE_MIB(o1, ipIfStatsReasmFails, o2->ipIfStatsReasmFails);
11287 11285 UPDATE_MIB(o1, ipIfStatsOutNoRoutes, o2->ipIfStatsOutNoRoutes);
11288 11286 UPDATE_MIB(o1, ipIfStatsReasmDuplicates, o2->ipIfStatsReasmDuplicates);
11289 11287 UPDATE_MIB(o1, ipIfStatsReasmPartDups, o2->ipIfStatsReasmPartDups);
11290 11288 UPDATE_MIB(o1, ipIfStatsForwProhibits, o2->ipIfStatsForwProhibits);
11291 11289 UPDATE_MIB(o1, udpInCksumErrs, o2->udpInCksumErrs);
11292 11290 UPDATE_MIB(o1, udpInOverflows, o2->udpInOverflows);
11293 11291 UPDATE_MIB(o1, rawipInOverflows, o2->rawipInOverflows);
11294 11292 UPDATE_MIB(o1, ipIfStatsInWrongIPVersion,
11295 11293 o2->ipIfStatsInWrongIPVersion);
11296 11294 UPDATE_MIB(o1, ipIfStatsOutWrongIPVersion,
11297 11295 o2->ipIfStatsInWrongIPVersion);
11298 11296 UPDATE_MIB(o1, ipIfStatsOutSwitchIPVersion,
11299 11297 o2->ipIfStatsOutSwitchIPVersion);
11300 11298 UPDATE_MIB(o1, ipIfStatsHCInReceives, o2->ipIfStatsHCInReceives);
11301 11299 UPDATE_MIB(o1, ipIfStatsHCInOctets, o2->ipIfStatsHCInOctets);
11302 11300 UPDATE_MIB(o1, ipIfStatsHCInForwDatagrams,
11303 11301 o2->ipIfStatsHCInForwDatagrams);
11304 11302 UPDATE_MIB(o1, ipIfStatsHCInDelivers, o2->ipIfStatsHCInDelivers);
11305 11303 UPDATE_MIB(o1, ipIfStatsHCOutRequests, o2->ipIfStatsHCOutRequests);
11306 11304 UPDATE_MIB(o1, ipIfStatsHCOutForwDatagrams,
11307 11305 o2->ipIfStatsHCOutForwDatagrams);
11308 11306 UPDATE_MIB(o1, ipIfStatsOutFragReqds, o2->ipIfStatsOutFragReqds);
11309 11307 UPDATE_MIB(o1, ipIfStatsHCOutTransmits, o2->ipIfStatsHCOutTransmits);
11310 11308 UPDATE_MIB(o1, ipIfStatsHCOutOctets, o2->ipIfStatsHCOutOctets);
11311 11309 UPDATE_MIB(o1, ipIfStatsHCInMcastPkts, o2->ipIfStatsHCInMcastPkts);
11312 11310 UPDATE_MIB(o1, ipIfStatsHCInMcastOctets, o2->ipIfStatsHCInMcastOctets);
11313 11311 UPDATE_MIB(o1, ipIfStatsHCOutMcastPkts, o2->ipIfStatsHCOutMcastPkts);
11314 11312 UPDATE_MIB(o1, ipIfStatsHCOutMcastOctets,
11315 11313 o2->ipIfStatsHCOutMcastOctets);
11316 11314 UPDATE_MIB(o1, ipIfStatsHCInBcastPkts, o2->ipIfStatsHCInBcastPkts);
11317 11315 UPDATE_MIB(o1, ipIfStatsHCOutBcastPkts, o2->ipIfStatsHCOutBcastPkts);
11318 11316 UPDATE_MIB(o1, ipsecInSucceeded, o2->ipsecInSucceeded);
11319 11317 UPDATE_MIB(o1, ipsecInFailed, o2->ipsecInFailed);
11320 11318 UPDATE_MIB(o1, ipInCksumErrs, o2->ipInCksumErrs);
11321 11319 UPDATE_MIB(o1, tcpInErrs, o2->tcpInErrs);
11322 11320 UPDATE_MIB(o1, udpNoPorts, o2->udpNoPorts);
11323 11321 }
11324 11322
11325 11323 void
11326 11324 ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *o1, mib2_ipv6IfIcmpEntry_t *o2)
11327 11325 {
11328 11326 UPDATE_MIB(o1, ipv6IfIcmpInMsgs, o2->ipv6IfIcmpInMsgs);
11329 11327 UPDATE_MIB(o1, ipv6IfIcmpInErrors, o2->ipv6IfIcmpInErrors);
11330 11328 UPDATE_MIB(o1, ipv6IfIcmpInDestUnreachs, o2->ipv6IfIcmpInDestUnreachs);
11331 11329 UPDATE_MIB(o1, ipv6IfIcmpInAdminProhibs, o2->ipv6IfIcmpInAdminProhibs);
11332 11330 UPDATE_MIB(o1, ipv6IfIcmpInTimeExcds, o2->ipv6IfIcmpInTimeExcds);
11333 11331 UPDATE_MIB(o1, ipv6IfIcmpInParmProblems, o2->ipv6IfIcmpInParmProblems);
11334 11332 UPDATE_MIB(o1, ipv6IfIcmpInPktTooBigs, o2->ipv6IfIcmpInPktTooBigs);
11335 11333 UPDATE_MIB(o1, ipv6IfIcmpInEchos, o2->ipv6IfIcmpInEchos);
11336 11334 UPDATE_MIB(o1, ipv6IfIcmpInEchoReplies, o2->ipv6IfIcmpInEchoReplies);
11337 11335 UPDATE_MIB(o1, ipv6IfIcmpInRouterSolicits,
11338 11336 o2->ipv6IfIcmpInRouterSolicits);
11339 11337 UPDATE_MIB(o1, ipv6IfIcmpInRouterAdvertisements,
11340 11338 o2->ipv6IfIcmpInRouterAdvertisements);
11341 11339 UPDATE_MIB(o1, ipv6IfIcmpInNeighborSolicits,
11342 11340 o2->ipv6IfIcmpInNeighborSolicits);
11343 11341 UPDATE_MIB(o1, ipv6IfIcmpInNeighborAdvertisements,
11344 11342 o2->ipv6IfIcmpInNeighborAdvertisements);
11345 11343 UPDATE_MIB(o1, ipv6IfIcmpInRedirects, o2->ipv6IfIcmpInRedirects);
11346 11344 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembQueries,
11347 11345 o2->ipv6IfIcmpInGroupMembQueries);
11348 11346 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembResponses,
11349 11347 o2->ipv6IfIcmpInGroupMembResponses);
11350 11348 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembReductions,
11351 11349 o2->ipv6IfIcmpInGroupMembReductions);
11352 11350 UPDATE_MIB(o1, ipv6IfIcmpOutMsgs, o2->ipv6IfIcmpOutMsgs);
11353 11351 UPDATE_MIB(o1, ipv6IfIcmpOutErrors, o2->ipv6IfIcmpOutErrors);
11354 11352 UPDATE_MIB(o1, ipv6IfIcmpOutDestUnreachs,
11355 11353 o2->ipv6IfIcmpOutDestUnreachs);
11356 11354 UPDATE_MIB(o1, ipv6IfIcmpOutAdminProhibs,
11357 11355 o2->ipv6IfIcmpOutAdminProhibs);
11358 11356 UPDATE_MIB(o1, ipv6IfIcmpOutTimeExcds, o2->ipv6IfIcmpOutTimeExcds);
11359 11357 UPDATE_MIB(o1, ipv6IfIcmpOutParmProblems,
11360 11358 o2->ipv6IfIcmpOutParmProblems);
11361 11359 UPDATE_MIB(o1, ipv6IfIcmpOutPktTooBigs, o2->ipv6IfIcmpOutPktTooBigs);
11362 11360 UPDATE_MIB(o1, ipv6IfIcmpOutEchos, o2->ipv6IfIcmpOutEchos);
11363 11361 UPDATE_MIB(o1, ipv6IfIcmpOutEchoReplies, o2->ipv6IfIcmpOutEchoReplies);
11364 11362 UPDATE_MIB(o1, ipv6IfIcmpOutRouterSolicits,
11365 11363 o2->ipv6IfIcmpOutRouterSolicits);
11366 11364 UPDATE_MIB(o1, ipv6IfIcmpOutRouterAdvertisements,
11367 11365 o2->ipv6IfIcmpOutRouterAdvertisements);
11368 11366 UPDATE_MIB(o1, ipv6IfIcmpOutNeighborSolicits,
11369 11367 o2->ipv6IfIcmpOutNeighborSolicits);
11370 11368 UPDATE_MIB(o1, ipv6IfIcmpOutNeighborAdvertisements,
11371 11369 o2->ipv6IfIcmpOutNeighborAdvertisements);
11372 11370 UPDATE_MIB(o1, ipv6IfIcmpOutRedirects, o2->ipv6IfIcmpOutRedirects);
11373 11371 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembQueries,
11374 11372 o2->ipv6IfIcmpOutGroupMembQueries);
11375 11373 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembResponses,
11376 11374 o2->ipv6IfIcmpOutGroupMembResponses);
11377 11375 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembReductions,
11378 11376 o2->ipv6IfIcmpOutGroupMembReductions);
11379 11377 UPDATE_MIB(o1, ipv6IfIcmpInOverflows, o2->ipv6IfIcmpInOverflows);
11380 11378 UPDATE_MIB(o1, ipv6IfIcmpBadHoplimit, o2->ipv6IfIcmpBadHoplimit);
11381 11379 UPDATE_MIB(o1, ipv6IfIcmpInBadNeighborAdvertisements,
11382 11380 o2->ipv6IfIcmpInBadNeighborAdvertisements);
11383 11381 UPDATE_MIB(o1, ipv6IfIcmpInBadNeighborSolicitations,
11384 11382 o2->ipv6IfIcmpInBadNeighborSolicitations);
11385 11383 UPDATE_MIB(o1, ipv6IfIcmpInBadRedirects, o2->ipv6IfIcmpInBadRedirects);
11386 11384 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembTotal,
11387 11385 o2->ipv6IfIcmpInGroupMembTotal);
11388 11386 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembBadQueries,
11389 11387 o2->ipv6IfIcmpInGroupMembBadQueries);
11390 11388 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembBadReports,
11391 11389 o2->ipv6IfIcmpInGroupMembBadReports);
11392 11390 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembOurReports,
11393 11391 o2->ipv6IfIcmpInGroupMembOurReports);
11394 11392 }
11395 11393
11396 11394 /*
11397 11395 * Called before the options are updated to check if this packet will
11398 11396 * be source routed from here.
11399 11397 * This routine assumes that the options are well formed i.e. that they
11400 11398 * have already been checked.
11401 11399 */
11402 11400 boolean_t
11403 11401 ip_source_routed(ipha_t *ipha, ip_stack_t *ipst)
11404 11402 {
11405 11403 ipoptp_t opts;
11406 11404 uchar_t *opt;
11407 11405 uint8_t optval;
11408 11406 uint8_t optlen;
11409 11407 ipaddr_t dst;
11410 11408
11411 11409 if (IS_SIMPLE_IPH(ipha)) {
11412 11410 ip2dbg(("not source routed\n"));
11413 11411 return (B_FALSE);
11414 11412 }
11415 11413 dst = ipha->ipha_dst;
11416 11414 for (optval = ipoptp_first(&opts, ipha);
11417 11415 optval != IPOPT_EOL;
11418 11416 optval = ipoptp_next(&opts)) {
11419 11417 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11420 11418 opt = opts.ipoptp_cur;
11421 11419 optlen = opts.ipoptp_len;
11422 11420 ip2dbg(("ip_source_routed: opt %d, len %d\n",
11423 11421 optval, optlen));
11424 11422 switch (optval) {
11425 11423 uint32_t off;
11426 11424 case IPOPT_SSRR:
11427 11425 case IPOPT_LSRR:
11428 11426 /*
11429 11427 * If dst is one of our addresses and there are some
11430 11428 * entries left in the source route return (true).
11431 11429 */
11432 11430 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
11433 11431 ip2dbg(("ip_source_routed: not next"
11434 11432 " source route 0x%x\n",
11435 11433 ntohl(dst)));
11436 11434 return (B_FALSE);
11437 11435 }
11438 11436 off = opt[IPOPT_OFFSET];
11439 11437 off--;
11440 11438 if (optlen < IP_ADDR_LEN ||
11441 11439 off > optlen - IP_ADDR_LEN) {
11442 11440 /* End of source route */
11443 11441 ip1dbg(("ip_source_routed: end of SR\n"));
11444 11442 return (B_FALSE);
11445 11443 }
11446 11444 return (B_TRUE);
11447 11445 }
11448 11446 }
11449 11447 ip2dbg(("not source routed\n"));
11450 11448 return (B_FALSE);
11451 11449 }
11452 11450
11453 11451 /*
11454 11452 * ip_unbind is called by the transports to remove a conn from
11455 11453 * the fanout table.
11456 11454 */
11457 11455 void
11458 11456 ip_unbind(conn_t *connp)
11459 11457 {
11460 11458
11461 11459 ASSERT(!MUTEX_HELD(&connp->conn_lock));
11462 11460
11463 11461 if (is_system_labeled() && connp->conn_anon_port) {
11464 11462 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
11465 11463 connp->conn_mlp_type, connp->conn_proto,
11466 11464 ntohs(connp->conn_lport), B_FALSE);
11467 11465 connp->conn_anon_port = 0;
11468 11466 }
11469 11467 connp->conn_mlp_type = mlptSingle;
11470 11468
11471 11469 ipcl_hash_remove(connp);
11472 11470 }
11473 11471
11474 11472 /*
11475 11473 * Used for deciding the MSS size for the upper layer. Thus
11476 11474 * we need to check the outbound policy values in the conn.
11477 11475 */
11478 11476 int
11479 11477 conn_ipsec_length(conn_t *connp)
11480 11478 {
11481 11479 ipsec_latch_t *ipl;
11482 11480
11483 11481 ipl = connp->conn_latch;
11484 11482 if (ipl == NULL)
11485 11483 return (0);
11486 11484
11487 11485 if (connp->conn_ixa->ixa_ipsec_policy == NULL)
11488 11486 return (0);
11489 11487
11490 11488 return (connp->conn_ixa->ixa_ipsec_policy->ipsp_act->ipa_ovhd);
11491 11489 }
11492 11490
11493 11491 /*
11494 11492 * Returns an estimate of the IPsec headers size. This is used if
11495 11493 * we don't want to call into IPsec to get the exact size.
11496 11494 */
11497 11495 int
11498 11496 ipsec_out_extra_length(ip_xmit_attr_t *ixa)
11499 11497 {
11500 11498 ipsec_action_t *a;
11501 11499
11502 11500 if (!(ixa->ixa_flags & IXAF_IPSEC_SECURE))
11503 11501 return (0);
11504 11502
11505 11503 a = ixa->ixa_ipsec_action;
11506 11504 if (a == NULL) {
11507 11505 ASSERT(ixa->ixa_ipsec_policy != NULL);
11508 11506 a = ixa->ixa_ipsec_policy->ipsp_act;
11509 11507 }
11510 11508 ASSERT(a != NULL);
11511 11509
11512 11510 return (a->ipa_ovhd);
11513 11511 }
11514 11512
11515 11513 /*
11516 11514 * If there are any source route options, return the true final
11517 11515 * destination. Otherwise, return the destination.
11518 11516 */
11519 11517 ipaddr_t
11520 11518 ip_get_dst(ipha_t *ipha)
11521 11519 {
11522 11520 ipoptp_t opts;
11523 11521 uchar_t *opt;
11524 11522 uint8_t optval;
11525 11523 uint8_t optlen;
11526 11524 ipaddr_t dst;
11527 11525 uint32_t off;
11528 11526
11529 11527 dst = ipha->ipha_dst;
11530 11528
11531 11529 if (IS_SIMPLE_IPH(ipha))
11532 11530 return (dst);
11533 11531
11534 11532 for (optval = ipoptp_first(&opts, ipha);
11535 11533 optval != IPOPT_EOL;
11536 11534 optval = ipoptp_next(&opts)) {
11537 11535 opt = opts.ipoptp_cur;
11538 11536 optlen = opts.ipoptp_len;
11539 11537 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11540 11538 switch (optval) {
11541 11539 case IPOPT_SSRR:
11542 11540 case IPOPT_LSRR:
11543 11541 off = opt[IPOPT_OFFSET];
11544 11542 /*
11545 11543 * If one of the conditions is true, it means
11546 11544 * end of options and dst already has the right
11547 11545 * value.
11548 11546 */
11549 11547 if (!(optlen < IP_ADDR_LEN || off > optlen - 3)) {
11550 11548 off = optlen - IP_ADDR_LEN;
11551 11549 bcopy(&opt[off], &dst, IP_ADDR_LEN);
11552 11550 }
11553 11551 return (dst);
11554 11552 default:
11555 11553 break;
11556 11554 }
11557 11555 }
11558 11556
11559 11557 return (dst);
11560 11558 }
11561 11559
11562 11560 /*
11563 11561 * Outbound IP fragmentation routine.
11564 11562 * Assumes the caller has checked whether or not fragmentation should
11565 11563 * be allowed. Here we copy the DF bit from the header to all the generated
11566 11564 * fragments.
11567 11565 */
11568 11566 int
11569 11567 ip_fragment_v4(mblk_t *mp_orig, nce_t *nce, iaflags_t ixaflags,
11570 11568 uint_t pkt_len, uint32_t max_frag, uint32_t xmit_hint, zoneid_t szone,
11571 11569 zoneid_t nolzid, pfirepostfrag_t postfragfn, uintptr_t *ixa_cookie)
11572 11570 {
11573 11571 int i1;
11574 11572 int hdr_len;
11575 11573 mblk_t *hdr_mp;
11576 11574 ipha_t *ipha;
11577 11575 int ip_data_end;
11578 11576 int len;
11579 11577 mblk_t *mp = mp_orig;
11580 11578 int offset;
11581 11579 ill_t *ill = nce->nce_ill;
11582 11580 ip_stack_t *ipst = ill->ill_ipst;
11583 11581 mblk_t *carve_mp;
11584 11582 uint32_t frag_flag;
11585 11583 uint_t priority = mp->b_band;
11586 11584 int error = 0;
11587 11585
11588 11586 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragReqds);
11589 11587
11590 11588 if (pkt_len != msgdsize(mp)) {
11591 11589 ip0dbg(("Packet length mismatch: %d, %ld\n",
11592 11590 pkt_len, msgdsize(mp)));
11593 11591 freemsg(mp);
11594 11592 return (EINVAL);
11595 11593 }
11596 11594
11597 11595 if (max_frag == 0) {
11598 11596 ip1dbg(("ip_fragment_v4: max_frag is zero. Dropping packet\n"));
11599 11597 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11600 11598 ip_drop_output("FragFails: zero max_frag", mp, ill);
11601 11599 freemsg(mp);
11602 11600 return (EINVAL);
11603 11601 }
11604 11602
11605 11603 ASSERT(MBLKL(mp) >= sizeof (ipha_t));
11606 11604 ipha = (ipha_t *)mp->b_rptr;
11607 11605 ASSERT(ntohs(ipha->ipha_length) == pkt_len);
11608 11606 frag_flag = ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_DF;
11609 11607
11610 11608 /*
11611 11609 * Establish the starting offset. May not be zero if we are fragging
11612 11610 * a fragment that is being forwarded.
11613 11611 */
11614 11612 offset = ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_OFFSET;
11615 11613
11616 11614 /* TODO why is this test needed? */
11617 11615 if (((max_frag - ntohs(ipha->ipha_length)) & ~7) < 8) {
11618 11616 /* TODO: notify ulp somehow */
11619 11617 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11620 11618 ip_drop_output("FragFails: bad starting offset", mp, ill);
11621 11619 freemsg(mp);
11622 11620 return (EINVAL);
11623 11621 }
11624 11622
11625 11623 hdr_len = IPH_HDR_LENGTH(ipha);
11626 11624 ipha->ipha_hdr_checksum = 0;
11627 11625
11628 11626 /*
11629 11627 * Establish the number of bytes maximum per frag, after putting
11630 11628 * in the header.
11631 11629 */
11632 11630 len = (max_frag - hdr_len) & ~7;
11633 11631
11634 11632 /* Get a copy of the header for the trailing frags */
11635 11633 hdr_mp = ip_fragment_copyhdr((uchar_t *)ipha, hdr_len, offset, ipst,
11636 11634 mp);
11637 11635 if (hdr_mp == NULL) {
11638 11636 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11639 11637 ip_drop_output("FragFails: no hdr_mp", mp, ill);
11640 11638 freemsg(mp);
11641 11639 return (ENOBUFS);
11642 11640 }
11643 11641
11644 11642 /* Store the starting offset, with the MoreFrags flag. */
11645 11643 i1 = offset | IPH_MF | frag_flag;
11646 11644 ipha->ipha_fragment_offset_and_flags = htons((uint16_t)i1);
11647 11645
11648 11646 /* Establish the ending byte offset, based on the starting offset. */
11649 11647 offset <<= 3;
11650 11648 ip_data_end = offset + ntohs(ipha->ipha_length) - hdr_len;
11651 11649
11652 11650 /* Store the length of the first fragment in the IP header. */
11653 11651 i1 = len + hdr_len;
11654 11652 ASSERT(i1 <= IP_MAXPACKET);
11655 11653 ipha->ipha_length = htons((uint16_t)i1);
11656 11654
11657 11655 /*
11658 11656 * Compute the IP header checksum for the first frag. We have to
11659 11657 * watch out that we stop at the end of the header.
11660 11658 */
11661 11659 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
11662 11660
11663 11661 /*
11664 11662 * Now carve off the first frag. Note that this will include the
11665 11663 * original IP header.
11666 11664 */
11667 11665 if (!(mp = ip_carve_mp(&mp_orig, i1))) {
11668 11666 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11669 11667 ip_drop_output("FragFails: could not carve mp", mp_orig, ill);
11670 11668 freeb(hdr_mp);
11671 11669 freemsg(mp_orig);
11672 11670 return (ENOBUFS);
11673 11671 }
11674 11672
11675 11673 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragCreates);
11676 11674
11677 11675 error = postfragfn(mp, nce, ixaflags, i1, xmit_hint, szone, nolzid,
11678 11676 ixa_cookie);
11679 11677 if (error != 0 && error != EWOULDBLOCK) {
11680 11678 /* No point in sending the other fragments */
11681 11679 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11682 11680 ip_drop_output("FragFails: postfragfn failed", mp_orig, ill);
11683 11681 freeb(hdr_mp);
11684 11682 freemsg(mp_orig);
11685 11683 return (error);
11686 11684 }
11687 11685
11688 11686 /* No need to redo state machine in loop */
11689 11687 ixaflags &= ~IXAF_REACH_CONF;
11690 11688
11691 11689 /* Advance the offset to the second frag starting point. */
11692 11690 offset += len;
11693 11691 /*
11694 11692 * Update hdr_len from the copied header - there might be less options
11695 11693 * in the later fragments.
11696 11694 */
11697 11695 hdr_len = IPH_HDR_LENGTH(hdr_mp->b_rptr);
11698 11696 /* Loop until done. */
11699 11697 for (;;) {
11700 11698 uint16_t offset_and_flags;
11701 11699 uint16_t ip_len;
11702 11700
11703 11701 if (ip_data_end - offset > len) {
11704 11702 /*
11705 11703 * Carve off the appropriate amount from the original
11706 11704 * datagram.
11707 11705 */
11708 11706 if (!(carve_mp = ip_carve_mp(&mp_orig, len))) {
11709 11707 mp = NULL;
11710 11708 break;
11711 11709 }
11712 11710 /*
11713 11711 * More frags after this one. Get another copy
11714 11712 * of the header.
11715 11713 */
11716 11714 if (carve_mp->b_datap->db_ref == 1 &&
11717 11715 hdr_mp->b_wptr - hdr_mp->b_rptr <
11718 11716 carve_mp->b_rptr - carve_mp->b_datap->db_base) {
11719 11717 /* Inline IP header */
11720 11718 carve_mp->b_rptr -= hdr_mp->b_wptr -
11721 11719 hdr_mp->b_rptr;
11722 11720 bcopy(hdr_mp->b_rptr, carve_mp->b_rptr,
11723 11721 hdr_mp->b_wptr - hdr_mp->b_rptr);
11724 11722 mp = carve_mp;
11725 11723 } else {
11726 11724 if (!(mp = copyb(hdr_mp))) {
11727 11725 freemsg(carve_mp);
11728 11726 break;
11729 11727 }
11730 11728 /* Get priority marking, if any. */
11731 11729 mp->b_band = priority;
11732 11730 mp->b_cont = carve_mp;
11733 11731 }
11734 11732 ipha = (ipha_t *)mp->b_rptr;
11735 11733 offset_and_flags = IPH_MF;
11736 11734 } else {
11737 11735 /*
11738 11736 * Last frag. Consume the header. Set len to
11739 11737 * the length of this last piece.
11740 11738 */
11741 11739 len = ip_data_end - offset;
11742 11740
11743 11741 /*
11744 11742 * Carve off the appropriate amount from the original
11745 11743 * datagram.
11746 11744 */
11747 11745 if (!(carve_mp = ip_carve_mp(&mp_orig, len))) {
11748 11746 mp = NULL;
11749 11747 break;
11750 11748 }
11751 11749 if (carve_mp->b_datap->db_ref == 1 &&
11752 11750 hdr_mp->b_wptr - hdr_mp->b_rptr <
11753 11751 carve_mp->b_rptr - carve_mp->b_datap->db_base) {
11754 11752 /* Inline IP header */
11755 11753 carve_mp->b_rptr -= hdr_mp->b_wptr -
11756 11754 hdr_mp->b_rptr;
11757 11755 bcopy(hdr_mp->b_rptr, carve_mp->b_rptr,
11758 11756 hdr_mp->b_wptr - hdr_mp->b_rptr);
11759 11757 mp = carve_mp;
11760 11758 freeb(hdr_mp);
11761 11759 hdr_mp = mp;
11762 11760 } else {
11763 11761 mp = hdr_mp;
11764 11762 /* Get priority marking, if any. */
11765 11763 mp->b_band = priority;
11766 11764 mp->b_cont = carve_mp;
11767 11765 }
11768 11766 ipha = (ipha_t *)mp->b_rptr;
11769 11767 /* A frag of a frag might have IPH_MF non-zero */
11770 11768 offset_and_flags =
11771 11769 ntohs(ipha->ipha_fragment_offset_and_flags) &
11772 11770 IPH_MF;
11773 11771 }
11774 11772 offset_and_flags |= (uint16_t)(offset >> 3);
11775 11773 offset_and_flags |= (uint16_t)frag_flag;
11776 11774 /* Store the offset and flags in the IP header. */
11777 11775 ipha->ipha_fragment_offset_and_flags = htons(offset_and_flags);
11778 11776
11779 11777 /* Store the length in the IP header. */
11780 11778 ip_len = (uint16_t)(len + hdr_len);
11781 11779 ipha->ipha_length = htons(ip_len);
11782 11780
11783 11781 /*
11784 11782 * Set the IP header checksum. Note that mp is just
11785 11783 * the header, so this is easy to pass to ip_csum.
11786 11784 */
11787 11785 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
11788 11786
11789 11787 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragCreates);
11790 11788
11791 11789 error = postfragfn(mp, nce, ixaflags, ip_len, xmit_hint, szone,
11792 11790 nolzid, ixa_cookie);
11793 11791 /* All done if we just consumed the hdr_mp. */
11794 11792 if (mp == hdr_mp) {
11795 11793 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragOKs);
11796 11794 return (error);
11797 11795 }
11798 11796 if (error != 0 && error != EWOULDBLOCK) {
11799 11797 DTRACE_PROBE2(ip__xmit__frag__fail, ill_t *, ill,
11800 11798 mblk_t *, hdr_mp);
11801 11799 /* No point in sending the other fragments */
11802 11800 break;
11803 11801 }
11804 11802
11805 11803 /* Otherwise, advance and loop. */
11806 11804 offset += len;
11807 11805 }
11808 11806 /* Clean up following allocation failure. */
11809 11807 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11810 11808 ip_drop_output("FragFails: loop ended", NULL, ill);
11811 11809 if (mp != hdr_mp)
11812 11810 freeb(hdr_mp);
11813 11811 if (mp != mp_orig)
11814 11812 freemsg(mp_orig);
11815 11813 return (error);
11816 11814 }
11817 11815
11818 11816 /*
11819 11817 * Copy the header plus those options which have the copy bit set
11820 11818 */
11821 11819 static mblk_t *
11822 11820 ip_fragment_copyhdr(uchar_t *rptr, int hdr_len, int offset, ip_stack_t *ipst,
11823 11821 mblk_t *src)
11824 11822 {
11825 11823 mblk_t *mp;
11826 11824 uchar_t *up;
11827 11825
11828 11826 /*
11829 11827 * Quick check if we need to look for options without the copy bit
11830 11828 * set
11831 11829 */
11832 11830 mp = allocb_tmpl(ipst->ips_ip_wroff_extra + hdr_len, src);
11833 11831 if (!mp)
11834 11832 return (mp);
11835 11833 mp->b_rptr += ipst->ips_ip_wroff_extra;
11836 11834 if (hdr_len == IP_SIMPLE_HDR_LENGTH || offset != 0) {
11837 11835 bcopy(rptr, mp->b_rptr, hdr_len);
11838 11836 mp->b_wptr += hdr_len + ipst->ips_ip_wroff_extra;
11839 11837 return (mp);
11840 11838 }
11841 11839 up = mp->b_rptr;
11842 11840 bcopy(rptr, up, IP_SIMPLE_HDR_LENGTH);
11843 11841 up += IP_SIMPLE_HDR_LENGTH;
11844 11842 rptr += IP_SIMPLE_HDR_LENGTH;
11845 11843 hdr_len -= IP_SIMPLE_HDR_LENGTH;
11846 11844 while (hdr_len > 0) {
11847 11845 uint32_t optval;
11848 11846 uint32_t optlen;
11849 11847
11850 11848 optval = *rptr;
11851 11849 if (optval == IPOPT_EOL)
11852 11850 break;
11853 11851 if (optval == IPOPT_NOP)
11854 11852 optlen = 1;
11855 11853 else
11856 11854 optlen = rptr[1];
11857 11855 if (optval & IPOPT_COPY) {
11858 11856 bcopy(rptr, up, optlen);
11859 11857 up += optlen;
11860 11858 }
11861 11859 rptr += optlen;
11862 11860 hdr_len -= optlen;
11863 11861 }
11864 11862 /*
11865 11863 * Make sure that we drop an even number of words by filling
11866 11864 * with EOL to the next word boundary.
11867 11865 */
11868 11866 for (hdr_len = up - (mp->b_rptr + IP_SIMPLE_HDR_LENGTH);
11869 11867 hdr_len & 0x3; hdr_len++)
11870 11868 *up++ = IPOPT_EOL;
11871 11869 mp->b_wptr = up;
11872 11870 /* Update header length */
11873 11871 mp->b_rptr[0] = (uint8_t)((IP_VERSION << 4) | ((up - mp->b_rptr) >> 2));
11874 11872 return (mp);
11875 11873 }
11876 11874
11877 11875 /*
11878 11876 * Update any source route, record route, or timestamp options when
11879 11877 * sending a packet back to ourselves.
11880 11878 * Check that we are at end of strict source route.
11881 11879 * The options have been sanity checked by ip_output_options().
11882 11880 */
11883 11881 void
11884 11882 ip_output_local_options(ipha_t *ipha, ip_stack_t *ipst)
11885 11883 {
11886 11884 ipoptp_t opts;
11887 11885 uchar_t *opt;
11888 11886 uint8_t optval;
11889 11887 uint8_t optlen;
11890 11888 ipaddr_t dst;
11891 11889 uint32_t ts;
11892 11890 timestruc_t now;
11893 11891
11894 11892 for (optval = ipoptp_first(&opts, ipha);
11895 11893 optval != IPOPT_EOL;
11896 11894 optval = ipoptp_next(&opts)) {
11897 11895 opt = opts.ipoptp_cur;
11898 11896 optlen = opts.ipoptp_len;
11899 11897 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11900 11898 switch (optval) {
11901 11899 uint32_t off;
11902 11900 case IPOPT_SSRR:
11903 11901 case IPOPT_LSRR:
11904 11902 off = opt[IPOPT_OFFSET];
11905 11903 off--;
11906 11904 if (optlen < IP_ADDR_LEN ||
11907 11905 off > optlen - IP_ADDR_LEN) {
11908 11906 /* End of source route */
11909 11907 break;
11910 11908 }
11911 11909 /*
11912 11910 * This will only happen if two consecutive entries
11913 11911 * in the source route contains our address or if
11914 11912 * it is a packet with a loose source route which
11915 11913 * reaches us before consuming the whole source route
11916 11914 */
11917 11915
11918 11916 if (optval == IPOPT_SSRR) {
11919 11917 return;
11920 11918 }
11921 11919 /*
11922 11920 * Hack: instead of dropping the packet truncate the
11923 11921 * source route to what has been used by filling the
11924 11922 * rest with IPOPT_NOP.
11925 11923 */
11926 11924 opt[IPOPT_OLEN] = (uint8_t)off;
11927 11925 while (off < optlen) {
11928 11926 opt[off++] = IPOPT_NOP;
11929 11927 }
11930 11928 break;
11931 11929 case IPOPT_RR:
11932 11930 off = opt[IPOPT_OFFSET];
11933 11931 off--;
11934 11932 if (optlen < IP_ADDR_LEN ||
11935 11933 off > optlen - IP_ADDR_LEN) {
11936 11934 /* No more room - ignore */
11937 11935 ip1dbg((
11938 11936 "ip_output_local_options: end of RR\n"));
11939 11937 break;
11940 11938 }
11941 11939 dst = htonl(INADDR_LOOPBACK);
11942 11940 bcopy(&dst, (char *)opt + off, IP_ADDR_LEN);
11943 11941 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
11944 11942 break;
11945 11943 case IPOPT_TS:
11946 11944 /* Insert timestamp if there is romm */
11947 11945 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
11948 11946 case IPOPT_TS_TSONLY:
11949 11947 off = IPOPT_TS_TIMELEN;
11950 11948 break;
11951 11949 case IPOPT_TS_PRESPEC:
11952 11950 case IPOPT_TS_PRESPEC_RFC791:
11953 11951 /* Verify that the address matched */
11954 11952 off = opt[IPOPT_OFFSET] - 1;
11955 11953 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
11956 11954 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
11957 11955 /* Not for us */
11958 11956 break;
11959 11957 }
11960 11958 /* FALLTHROUGH */
11961 11959 case IPOPT_TS_TSANDADDR:
11962 11960 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
11963 11961 break;
11964 11962 default:
11965 11963 /*
11966 11964 * ip_*put_options should have already
11967 11965 * dropped this packet.
11968 11966 */
11969 11967 cmn_err(CE_PANIC, "ip_output_local_options: "
11970 11968 "unknown IT - bug in ip_output_options?\n");
11971 11969 }
11972 11970 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
11973 11971 /* Increase overflow counter */
11974 11972 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
11975 11973 opt[IPOPT_POS_OV_FLG] = (uint8_t)
11976 11974 (opt[IPOPT_POS_OV_FLG] & 0x0F) |
11977 11975 (off << 4);
11978 11976 break;
11979 11977 }
11980 11978 off = opt[IPOPT_OFFSET] - 1;
11981 11979 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
11982 11980 case IPOPT_TS_PRESPEC:
11983 11981 case IPOPT_TS_PRESPEC_RFC791:
11984 11982 case IPOPT_TS_TSANDADDR:
11985 11983 dst = htonl(INADDR_LOOPBACK);
11986 11984 bcopy(&dst, (char *)opt + off, IP_ADDR_LEN);
11987 11985 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
11988 11986 /* FALLTHROUGH */
11989 11987 case IPOPT_TS_TSONLY:
11990 11988 off = opt[IPOPT_OFFSET] - 1;
11991 11989 /* Compute # of milliseconds since midnight */
11992 11990 gethrestime(&now);
11993 11991 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
11994 11992 NSEC2MSEC(now.tv_nsec);
11995 11993 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
11996 11994 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
11997 11995 break;
11998 11996 }
11999 11997 break;
12000 11998 }
12001 11999 }
12002 12000 }
12003 12001
12004 12002 /*
12005 12003 * Prepend an M_DATA fastpath header, and if none present prepend a
12006 12004 * DL_UNITDATA_REQ. Frees the mblk on failure.
12007 12005 *
12008 12006 * nce_dlur_mp and nce_fp_mp can not disappear once they have been set.
12009 12007 * If there is a change to them, the nce will be deleted (condemned) and
12010 12008 * a new nce_t will be created when packets are sent. Thus we need no locks
12011 12009 * to access those fields.
12012 12010 *
12013 12011 * We preserve b_band to support IPQoS. If a DL_UNITDATA_REQ is prepended
12014 12012 * we place b_band in dl_priority.dl_max.
12015 12013 */
12016 12014 static mblk_t *
12017 12015 ip_xmit_attach_llhdr(mblk_t *mp, nce_t *nce)
12018 12016 {
12019 12017 uint_t hlen;
12020 12018 mblk_t *mp1;
12021 12019 uint_t priority;
12022 12020 uchar_t *rptr;
12023 12021
12024 12022 rptr = mp->b_rptr;
12025 12023
12026 12024 ASSERT(DB_TYPE(mp) == M_DATA);
12027 12025 priority = mp->b_band;
12028 12026
12029 12027 ASSERT(nce != NULL);
12030 12028 if ((mp1 = nce->nce_fp_mp) != NULL) {
12031 12029 hlen = MBLKL(mp1);
12032 12030 /*
12033 12031 * Check if we have enough room to prepend fastpath
12034 12032 * header
12035 12033 */
12036 12034 if (hlen != 0 && (rptr - mp->b_datap->db_base) >= hlen) {
12037 12035 rptr -= hlen;
12038 12036 bcopy(mp1->b_rptr, rptr, hlen);
12039 12037 /*
12040 12038 * Set the b_rptr to the start of the link layer
12041 12039 * header
12042 12040 */
12043 12041 mp->b_rptr = rptr;
12044 12042 return (mp);
12045 12043 }
12046 12044 mp1 = copyb(mp1);
12047 12045 if (mp1 == NULL) {
12048 12046 ill_t *ill = nce->nce_ill;
12049 12047
12050 12048 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12051 12049 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12052 12050 freemsg(mp);
12053 12051 return (NULL);
12054 12052 }
12055 12053 mp1->b_band = priority;
12056 12054 mp1->b_cont = mp;
12057 12055 DB_CKSUMSTART(mp1) = DB_CKSUMSTART(mp);
12058 12056 DB_CKSUMSTUFF(mp1) = DB_CKSUMSTUFF(mp);
12059 12057 DB_CKSUMEND(mp1) = DB_CKSUMEND(mp);
12060 12058 DB_CKSUMFLAGS(mp1) = DB_CKSUMFLAGS(mp);
12061 12059 DB_LSOMSS(mp1) = DB_LSOMSS(mp);
12062 12060 DTRACE_PROBE1(ip__xmit__copyb, (mblk_t *), mp1);
12063 12061 /*
12064 12062 * XXX disable ICK_VALID and compute checksum
12065 12063 * here; can happen if nce_fp_mp changes and
12066 12064 * it can't be copied now due to insufficient
12067 12065 * space. (unlikely, fp mp can change, but it
12068 12066 * does not increase in length)
12069 12067 */
12070 12068 return (mp1);
12071 12069 }
12072 12070 mp1 = copyb(nce->nce_dlur_mp);
12073 12071
12074 12072 if (mp1 == NULL) {
12075 12073 ill_t *ill = nce->nce_ill;
12076 12074
12077 12075 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12078 12076 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12079 12077 freemsg(mp);
12080 12078 return (NULL);
12081 12079 }
12082 12080 mp1->b_cont = mp;
12083 12081 if (priority != 0) {
12084 12082 mp1->b_band = priority;
12085 12083 ((dl_unitdata_req_t *)(mp1->b_rptr))->dl_priority.dl_max =
12086 12084 priority;
12087 12085 }
12088 12086 return (mp1);
12089 12087 }
12090 12088
12091 12089 /*
12092 12090 * Finish the outbound IPsec processing. This function is called from
12093 12091 * ipsec_out_process() if the IPsec packet was processed
12094 12092 * synchronously, or from {ah,esp}_kcf_callback_outbound() if it was processed
12095 12093 * asynchronously.
12096 12094 *
12097 12095 * This is common to IPv4 and IPv6.
12098 12096 */
12099 12097 int
12100 12098 ip_output_post_ipsec(mblk_t *mp, ip_xmit_attr_t *ixa)
12101 12099 {
12102 12100 iaflags_t ixaflags = ixa->ixa_flags;
12103 12101 uint_t pktlen;
12104 12102
12105 12103
12106 12104 /* AH/ESP don't update ixa_pktlen when they modify the packet */
12107 12105 if (ixaflags & IXAF_IS_IPV4) {
12108 12106 ipha_t *ipha = (ipha_t *)mp->b_rptr;
12109 12107
12110 12108 ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
12111 12109 pktlen = ntohs(ipha->ipha_length);
12112 12110 } else {
12113 12111 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
12114 12112
12115 12113 ASSERT(IPH_HDR_VERSION(mp->b_rptr) == IPV6_VERSION);
12116 12114 pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
12117 12115 }
12118 12116
12119 12117 /*
12120 12118 * We release any hard reference on the SAs here to make
12121 12119 * sure the SAs can be garbage collected. ipsr_sa has a soft reference
12122 12120 * on the SAs.
12123 12121 * If in the future we want the hard latching of the SAs in the
12124 12122 * ip_xmit_attr_t then we should remove this.
12125 12123 */
12126 12124 if (ixa->ixa_ipsec_esp_sa != NULL) {
12127 12125 IPSA_REFRELE(ixa->ixa_ipsec_esp_sa);
12128 12126 ixa->ixa_ipsec_esp_sa = NULL;
12129 12127 }
12130 12128 if (ixa->ixa_ipsec_ah_sa != NULL) {
12131 12129 IPSA_REFRELE(ixa->ixa_ipsec_ah_sa);
12132 12130 ixa->ixa_ipsec_ah_sa = NULL;
12133 12131 }
12134 12132
12135 12133 /* Do we need to fragment? */
12136 12134 if ((ixa->ixa_flags & IXAF_IPV6_ADD_FRAGHDR) ||
12137 12135 pktlen > ixa->ixa_fragsize) {
12138 12136 if (ixaflags & IXAF_IS_IPV4) {
12139 12137 ASSERT(!(ixa->ixa_flags & IXAF_IPV6_ADD_FRAGHDR));
12140 12138 /*
12141 12139 * We check for the DF case in ipsec_out_process
12142 12140 * hence this only handles the non-DF case.
12143 12141 */
12144 12142 return (ip_fragment_v4(mp, ixa->ixa_nce, ixa->ixa_flags,
12145 12143 pktlen, ixa->ixa_fragsize,
12146 12144 ixa->ixa_xmit_hint, ixa->ixa_zoneid,
12147 12145 ixa->ixa_no_loop_zoneid, ixa->ixa_postfragfn,
12148 12146 &ixa->ixa_cookie));
12149 12147 } else {
12150 12148 mp = ip_fraghdr_add_v6(mp, ixa->ixa_ident, ixa);
12151 12149 if (mp == NULL) {
12152 12150 /* MIB and ip_drop_output already done */
12153 12151 return (ENOMEM);
12154 12152 }
12155 12153 pktlen += sizeof (ip6_frag_t);
12156 12154 if (pktlen > ixa->ixa_fragsize) {
12157 12155 return (ip_fragment_v6(mp, ixa->ixa_nce,
12158 12156 ixa->ixa_flags, pktlen,
12159 12157 ixa->ixa_fragsize, ixa->ixa_xmit_hint,
12160 12158 ixa->ixa_zoneid, ixa->ixa_no_loop_zoneid,
12161 12159 ixa->ixa_postfragfn, &ixa->ixa_cookie));
12162 12160 }
12163 12161 }
12164 12162 }
12165 12163 return ((ixa->ixa_postfragfn)(mp, ixa->ixa_nce, ixa->ixa_flags,
12166 12164 pktlen, ixa->ixa_xmit_hint, ixa->ixa_zoneid,
12167 12165 ixa->ixa_no_loop_zoneid, NULL));
12168 12166 }
12169 12167
12170 12168 /*
12171 12169 * Finish the inbound IPsec processing. This function is called from
12172 12170 * ipsec_out_process() if the IPsec packet was processed
12173 12171 * synchronously, or from {ah,esp}_kcf_callback_outbound() if it was processed
12174 12172 * asynchronously.
12175 12173 *
12176 12174 * This is common to IPv4 and IPv6.
12177 12175 */
12178 12176 void
12179 12177 ip_input_post_ipsec(mblk_t *mp, ip_recv_attr_t *ira)
12180 12178 {
12181 12179 iaflags_t iraflags = ira->ira_flags;
12182 12180
12183 12181 /* Length might have changed */
12184 12182 if (iraflags & IRAF_IS_IPV4) {
12185 12183 ipha_t *ipha = (ipha_t *)mp->b_rptr;
12186 12184
12187 12185 ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
12188 12186 ira->ira_pktlen = ntohs(ipha->ipha_length);
12189 12187 ira->ira_ip_hdr_length = IPH_HDR_LENGTH(ipha);
12190 12188 ira->ira_protocol = ipha->ipha_protocol;
12191 12189
12192 12190 ip_fanout_v4(mp, ipha, ira);
12193 12191 } else {
12194 12192 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
12195 12193 uint8_t *nexthdrp;
12196 12194
12197 12195 ASSERT(IPH_HDR_VERSION(mp->b_rptr) == IPV6_VERSION);
12198 12196 ira->ira_pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
12199 12197 if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &ira->ira_ip_hdr_length,
12200 12198 &nexthdrp)) {
12201 12199 /* Malformed packet */
12202 12200 BUMP_MIB(ira->ira_ill->ill_ip_mib, ipIfStatsInDiscards);
12203 12201 ip_drop_input("ipIfStatsInDiscards", mp, ira->ira_ill);
12204 12202 freemsg(mp);
12205 12203 return;
12206 12204 }
12207 12205 ira->ira_protocol = *nexthdrp;
12208 12206 ip_fanout_v6(mp, ip6h, ira);
12209 12207 }
12210 12208 }
12211 12209
12212 12210 /*
12213 12211 * Select which AH & ESP SA's to use (if any) for the outbound packet.
12214 12212 *
12215 12213 * If this function returns B_TRUE, the requested SA's have been filled
12216 12214 * into the ixa_ipsec_*_sa pointers.
12217 12215 *
12218 12216 * If the function returns B_FALSE, the packet has been "consumed", most
12219 12217 * likely by an ACQUIRE sent up via PF_KEY to a key management daemon.
12220 12218 *
12221 12219 * The SA references created by the protocol-specific "select"
12222 12220 * function will be released in ip_output_post_ipsec.
12223 12221 */
12224 12222 static boolean_t
12225 12223 ipsec_out_select_sa(mblk_t *mp, ip_xmit_attr_t *ixa)
12226 12224 {
12227 12225 boolean_t need_ah_acquire = B_FALSE, need_esp_acquire = B_FALSE;
12228 12226 ipsec_policy_t *pp;
12229 12227 ipsec_action_t *ap;
12230 12228
12231 12229 ASSERT(ixa->ixa_flags & IXAF_IPSEC_SECURE);
12232 12230 ASSERT((ixa->ixa_ipsec_policy != NULL) ||
12233 12231 (ixa->ixa_ipsec_action != NULL));
12234 12232
12235 12233 ap = ixa->ixa_ipsec_action;
12236 12234 if (ap == NULL) {
12237 12235 pp = ixa->ixa_ipsec_policy;
12238 12236 ASSERT(pp != NULL);
12239 12237 ap = pp->ipsp_act;
12240 12238 ASSERT(ap != NULL);
12241 12239 }
12242 12240
12243 12241 /*
12244 12242 * We have an action. now, let's select SA's.
12245 12243 * A side effect of setting ixa_ipsec_*_sa is that it will
12246 12244 * be cached in the conn_t.
12247 12245 */
12248 12246 if (ap->ipa_want_esp) {
12249 12247 if (ixa->ixa_ipsec_esp_sa == NULL) {
12250 12248 need_esp_acquire = !ipsec_outbound_sa(mp, ixa,
12251 12249 IPPROTO_ESP);
12252 12250 }
12253 12251 ASSERT(need_esp_acquire || ixa->ixa_ipsec_esp_sa != NULL);
12254 12252 }
12255 12253
12256 12254 if (ap->ipa_want_ah) {
12257 12255 if (ixa->ixa_ipsec_ah_sa == NULL) {
12258 12256 need_ah_acquire = !ipsec_outbound_sa(mp, ixa,
12259 12257 IPPROTO_AH);
12260 12258 }
12261 12259 ASSERT(need_ah_acquire || ixa->ixa_ipsec_ah_sa != NULL);
12262 12260 /*
12263 12261 * The ESP and AH processing order needs to be preserved
12264 12262 * when both protocols are required (ESP should be applied
12265 12263 * before AH for an outbound packet). Force an ESP ACQUIRE
12266 12264 * when both ESP and AH are required, and an AH ACQUIRE
12267 12265 * is needed.
12268 12266 */
12269 12267 if (ap->ipa_want_esp && need_ah_acquire)
12270 12268 need_esp_acquire = B_TRUE;
12271 12269 }
12272 12270
12273 12271 /*
12274 12272 * Send an ACQUIRE (extended, regular, or both) if we need one.
12275 12273 * Release SAs that got referenced, but will not be used until we
12276 12274 * acquire _all_ of the SAs we need.
12277 12275 */
12278 12276 if (need_ah_acquire || need_esp_acquire) {
12279 12277 if (ixa->ixa_ipsec_ah_sa != NULL) {
12280 12278 IPSA_REFRELE(ixa->ixa_ipsec_ah_sa);
12281 12279 ixa->ixa_ipsec_ah_sa = NULL;
12282 12280 }
12283 12281 if (ixa->ixa_ipsec_esp_sa != NULL) {
12284 12282 IPSA_REFRELE(ixa->ixa_ipsec_esp_sa);
12285 12283 ixa->ixa_ipsec_esp_sa = NULL;
12286 12284 }
12287 12285
12288 12286 sadb_acquire(mp, ixa, need_ah_acquire, need_esp_acquire);
12289 12287 return (B_FALSE);
12290 12288 }
12291 12289
12292 12290 return (B_TRUE);
12293 12291 }
12294 12292
12295 12293 /*
12296 12294 * Handle IPsec output processing.
12297 12295 * This function is only entered once for a given packet.
12298 12296 * We try to do things synchronously, but if we need to have user-level
12299 12297 * set up SAs, or ESP or AH uses asynchronous kEF, then the operation
12300 12298 * will be completed
12301 12299 * - when the SAs are added in esp_add_sa_finish/ah_add_sa_finish
12302 12300 * - when asynchronous ESP is done it will do AH
12303 12301 *
12304 12302 * In all cases we come back in ip_output_post_ipsec() to fragment and
12305 12303 * send out the packet.
12306 12304 */
12307 12305 int
12308 12306 ipsec_out_process(mblk_t *mp, ip_xmit_attr_t *ixa)
12309 12307 {
12310 12308 ill_t *ill = ixa->ixa_nce->nce_ill;
12311 12309 ip_stack_t *ipst = ixa->ixa_ipst;
12312 12310 ipsec_stack_t *ipss;
12313 12311 ipsec_policy_t *pp;
12314 12312 ipsec_action_t *ap;
12315 12313
12316 12314 ASSERT(ixa->ixa_flags & IXAF_IPSEC_SECURE);
12317 12315
12318 12316 ASSERT((ixa->ixa_ipsec_policy != NULL) ||
12319 12317 (ixa->ixa_ipsec_action != NULL));
12320 12318
12321 12319 ipss = ipst->ips_netstack->netstack_ipsec;
12322 12320 if (!ipsec_loaded(ipss)) {
12323 12321 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12324 12322 ip_drop_packet(mp, B_TRUE, ill,
12325 12323 DROPPER(ipss, ipds_ip_ipsec_not_loaded),
12326 12324 &ipss->ipsec_dropper);
12327 12325 return (ENOTSUP);
12328 12326 }
12329 12327
12330 12328 ap = ixa->ixa_ipsec_action;
12331 12329 if (ap == NULL) {
12332 12330 pp = ixa->ixa_ipsec_policy;
12333 12331 ASSERT(pp != NULL);
12334 12332 ap = pp->ipsp_act;
12335 12333 ASSERT(ap != NULL);
12336 12334 }
12337 12335
12338 12336 /* Handle explicit drop action and bypass. */
12339 12337 switch (ap->ipa_act.ipa_type) {
12340 12338 case IPSEC_ACT_DISCARD:
12341 12339 case IPSEC_ACT_REJECT:
12342 12340 ip_drop_packet(mp, B_FALSE, ill,
12343 12341 DROPPER(ipss, ipds_spd_explicit), &ipss->ipsec_spd_dropper);
12344 12342 return (EHOSTUNREACH); /* IPsec policy failure */
12345 12343 case IPSEC_ACT_BYPASS:
12346 12344 return (ip_output_post_ipsec(mp, ixa));
12347 12345 }
12348 12346
12349 12347 /*
12350 12348 * The order of processing is first insert a IP header if needed.
12351 12349 * Then insert the ESP header and then the AH header.
12352 12350 */
12353 12351 if ((ixa->ixa_flags & IXAF_IS_IPV4) && ap->ipa_want_se) {
12354 12352 /*
12355 12353 * First get the outer IP header before sending
12356 12354 * it to ESP.
12357 12355 */
12358 12356 ipha_t *oipha, *iipha;
12359 12357 mblk_t *outer_mp, *inner_mp;
12360 12358
12361 12359 if ((outer_mp = allocb(sizeof (ipha_t), BPRI_HI)) == NULL) {
12362 12360 (void) mi_strlog(ill->ill_rq, 0,
12363 12361 SL_ERROR|SL_TRACE|SL_CONSOLE,
12364 12362 "ipsec_out_process: "
12365 12363 "Self-Encapsulation failed: Out of memory\n");
12366 12364 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12367 12365 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12368 12366 freemsg(mp);
12369 12367 return (ENOBUFS);
12370 12368 }
12371 12369 inner_mp = mp;
12372 12370 ASSERT(inner_mp->b_datap->db_type == M_DATA);
12373 12371 oipha = (ipha_t *)outer_mp->b_rptr;
12374 12372 iipha = (ipha_t *)inner_mp->b_rptr;
12375 12373 *oipha = *iipha;
12376 12374 outer_mp->b_wptr += sizeof (ipha_t);
12377 12375 oipha->ipha_length = htons(ntohs(iipha->ipha_length) +
12378 12376 sizeof (ipha_t));
12379 12377 oipha->ipha_protocol = IPPROTO_ENCAP;
12380 12378 oipha->ipha_version_and_hdr_length =
12381 12379 IP_SIMPLE_HDR_VERSION;
12382 12380 oipha->ipha_hdr_checksum = 0;
12383 12381 oipha->ipha_hdr_checksum = ip_csum_hdr(oipha);
12384 12382 outer_mp->b_cont = inner_mp;
12385 12383 mp = outer_mp;
12386 12384
12387 12385 ixa->ixa_flags |= IXAF_IPSEC_TUNNEL;
12388 12386 }
12389 12387
12390 12388 /* If we need to wait for a SA then we can't return any errno */
12391 12389 if (((ap->ipa_want_ah && (ixa->ixa_ipsec_ah_sa == NULL)) ||
12392 12390 (ap->ipa_want_esp && (ixa->ixa_ipsec_esp_sa == NULL))) &&
12393 12391 !ipsec_out_select_sa(mp, ixa))
12394 12392 return (0);
12395 12393
12396 12394 /*
12397 12395 * By now, we know what SA's to use. Toss over to ESP & AH
12398 12396 * to do the heavy lifting.
12399 12397 */
12400 12398 if (ap->ipa_want_esp) {
12401 12399 ASSERT(ixa->ixa_ipsec_esp_sa != NULL);
12402 12400
12403 12401 mp = ixa->ixa_ipsec_esp_sa->ipsa_output_func(mp, ixa);
12404 12402 if (mp == NULL) {
12405 12403 /*
12406 12404 * Either it failed or is pending. In the former case
12407 12405 * ipIfStatsInDiscards was increased.
12408 12406 */
12409 12407 return (0);
12410 12408 }
12411 12409 }
12412 12410
12413 12411 if (ap->ipa_want_ah) {
12414 12412 ASSERT(ixa->ixa_ipsec_ah_sa != NULL);
12415 12413
12416 12414 mp = ixa->ixa_ipsec_ah_sa->ipsa_output_func(mp, ixa);
12417 12415 if (mp == NULL) {
12418 12416 /*
12419 12417 * Either it failed or is pending. In the former case
12420 12418 * ipIfStatsInDiscards was increased.
12421 12419 */
12422 12420 return (0);
12423 12421 }
12424 12422 }
12425 12423 /*
12426 12424 * We are done with IPsec processing. Send it over
12427 12425 * the wire.
12428 12426 */
12429 12427 return (ip_output_post_ipsec(mp, ixa));
12430 12428 }
12431 12429
12432 12430 /*
12433 12431 * ioctls that go through a down/up sequence may need to wait for the down
12434 12432 * to complete. This involves waiting for the ire and ipif refcnts to go down
12435 12433 * to zero. Subsequently the ioctl is restarted from ipif_ill_refrele_tail.
12436 12434 */
12437 12435 /* ARGSUSED */
12438 12436 void
12439 12437 ip_reprocess_ioctl(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
12440 12438 {
12441 12439 struct iocblk *iocp;
12442 12440 mblk_t *mp1;
12443 12441 ip_ioctl_cmd_t *ipip;
12444 12442 int err;
12445 12443 sin_t *sin;
12446 12444 struct lifreq *lifr;
12447 12445 struct ifreq *ifr;
12448 12446
12449 12447 iocp = (struct iocblk *)mp->b_rptr;
12450 12448 ASSERT(ipsq != NULL);
12451 12449 /* Existence of mp1 verified in ip_wput_nondata */
12452 12450 mp1 = mp->b_cont->b_cont;
12453 12451 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12454 12452 if (ipip->ipi_cmd == SIOCSLIFNAME || ipip->ipi_cmd == IF_UNITSEL) {
12455 12453 /*
12456 12454 * Special case where ipx_current_ipif is not set:
12457 12455 * ill_phyint_reinit merged the v4 and v6 into a single ipsq.
12458 12456 * We are here as were not able to complete the operation in
12459 12457 * ipif_set_values because we could not become exclusive on
12460 12458 * the new ipsq.
12461 12459 */
12462 12460 ill_t *ill = q->q_ptr;
12463 12461 ipsq_current_start(ipsq, ill->ill_ipif, ipip->ipi_cmd);
12464 12462 }
12465 12463 ASSERT(ipsq->ipsq_xop->ipx_current_ipif != NULL);
12466 12464
12467 12465 if (ipip->ipi_cmd_type == IF_CMD) {
12468 12466 /* This a old style SIOC[GS]IF* command */
12469 12467 ifr = (struct ifreq *)mp1->b_rptr;
12470 12468 sin = (sin_t *)&ifr->ifr_addr;
12471 12469 } else if (ipip->ipi_cmd_type == LIF_CMD) {
12472 12470 /* This a new style SIOC[GS]LIF* command */
12473 12471 lifr = (struct lifreq *)mp1->b_rptr;
12474 12472 sin = (sin_t *)&lifr->lifr_addr;
12475 12473 } else {
12476 12474 sin = NULL;
12477 12475 }
12478 12476
12479 12477 err = (*ipip->ipi_func_restart)(ipsq->ipsq_xop->ipx_current_ipif, sin,
12480 12478 q, mp, ipip, mp1->b_rptr);
12481 12479
12482 12480 DTRACE_PROBE4(ipif__ioctl, char *, "ip_reprocess_ioctl finish",
12483 12481 int, ipip->ipi_cmd,
12484 12482 ill_t *, ipsq->ipsq_xop->ipx_current_ipif->ipif_ill,
12485 12483 ipif_t *, ipsq->ipsq_xop->ipx_current_ipif);
12486 12484
12487 12485 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), ipsq);
12488 12486 }
12489 12487
12490 12488 /*
12491 12489 * ioctl processing
12492 12490 *
12493 12491 * ioctl processing starts with ip_sioctl_copyin_setup(), which looks up
12494 12492 * the ioctl command in the ioctl tables, determines the copyin data size
12495 12493 * from the ipi_copyin_size field, and does an mi_copyin() of that size.
12496 12494 *
12497 12495 * ioctl processing then continues when the M_IOCDATA makes its way down to
12498 12496 * ip_wput_nondata(). The ioctl is looked up again in the ioctl table, its
12499 12497 * associated 'conn' is refheld till the end of the ioctl and the general
12500 12498 * ioctl processing function ip_process_ioctl() is called to extract the
12501 12499 * arguments and process the ioctl. To simplify extraction, ioctl commands
12502 12500 * are "typed" based on the arguments they take (e.g., LIF_CMD which takes a
12503 12501 * `struct lifreq'), and a common extract function (e.g., ip_extract_lifreq())
12504 12502 * is used to extract the ioctl's arguments.
12505 12503 *
12506 12504 * ip_process_ioctl determines if the ioctl needs to be serialized, and if
12507 12505 * so goes thru the serialization primitive ipsq_try_enter. Then the
12508 12506 * appropriate function to handle the ioctl is called based on the entry in
12509 12507 * the ioctl table. ioctl completion is encapsulated in ip_ioctl_finish
12510 12508 * which also refreleases the 'conn' that was refheld at the start of the
12511 12509 * ioctl. Finally ipsq_exit is called if needed to exit the ipsq.
12512 12510 *
12513 12511 * Many exclusive ioctls go thru an internal down up sequence as part of
12514 12512 * the operation. For example an attempt to change the IP address of an
12515 12513 * ipif entails ipif_down, set address, ipif_up. Bringing down the interface
12516 12514 * does all the cleanup such as deleting all ires that use this address.
12517 12515 * Then we need to wait till all references to the interface go away.
12518 12516 */
12519 12517 void
12520 12518 ip_process_ioctl(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *arg)
12521 12519 {
12522 12520 struct iocblk *iocp = (struct iocblk *)mp->b_rptr;
12523 12521 ip_ioctl_cmd_t *ipip = arg;
12524 12522 ip_extract_func_t *extract_funcp;
12525 12523 ill_t *ill;
12526 12524 cmd_info_t ci;
12527 12525 int err;
12528 12526 boolean_t entered_ipsq = B_FALSE;
12529 12527
12530 12528 ip3dbg(("ip_process_ioctl: ioctl %X\n", iocp->ioc_cmd));
12531 12529
12532 12530 if (ipip == NULL)
12533 12531 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12534 12532
12535 12533 /*
12536 12534 * SIOCLIFADDIF needs to go thru a special path since the
12537 12535 * ill may not exist yet. This happens in the case of lo0
12538 12536 * which is created using this ioctl.
12539 12537 */
12540 12538 if (ipip->ipi_cmd == SIOCLIFADDIF) {
12541 12539 err = ip_sioctl_addif(NULL, NULL, q, mp, NULL, NULL);
12542 12540 DTRACE_PROBE4(ipif__ioctl, char *, "ip_process_ioctl finish",
12543 12541 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12544 12542 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12545 12543 return;
12546 12544 }
12547 12545
12548 12546 ci.ci_ipif = NULL;
12549 12547 switch (ipip->ipi_cmd_type) {
12550 12548 case MISC_CMD:
12551 12549 case MSFILT_CMD:
12552 12550 /*
12553 12551 * All MISC_CMD ioctls come in here -- e.g. SIOCGLIFCONF.
12554 12552 */
12555 12553 if (ipip->ipi_cmd == IF_UNITSEL) {
12556 12554 /* ioctl comes down the ill */
12557 12555 ci.ci_ipif = ((ill_t *)q->q_ptr)->ill_ipif;
12558 12556 ipif_refhold(ci.ci_ipif);
12559 12557 }
12560 12558 err = 0;
12561 12559 ci.ci_sin = NULL;
12562 12560 ci.ci_sin6 = NULL;
12563 12561 ci.ci_lifr = NULL;
12564 12562 extract_funcp = NULL;
12565 12563 break;
12566 12564
12567 12565 case IF_CMD:
12568 12566 case LIF_CMD:
12569 12567 extract_funcp = ip_extract_lifreq;
12570 12568 break;
12571 12569
12572 12570 case ARP_CMD:
12573 12571 case XARP_CMD:
12574 12572 extract_funcp = ip_extract_arpreq;
12575 12573 break;
12576 12574
12577 12575 default:
12578 12576 ASSERT(0);
12579 12577 }
12580 12578
12581 12579 if (extract_funcp != NULL) {
12582 12580 err = (*extract_funcp)(q, mp, ipip, &ci);
12583 12581 if (err != 0) {
12584 12582 DTRACE_PROBE4(ipif__ioctl,
12585 12583 char *, "ip_process_ioctl finish err",
12586 12584 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12587 12585 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12588 12586 return;
12589 12587 }
12590 12588
12591 12589 /*
12592 12590 * All of the extraction functions return a refheld ipif.
12593 12591 */
12594 12592 ASSERT(ci.ci_ipif != NULL);
12595 12593 }
12596 12594
12597 12595 if (!(ipip->ipi_flags & IPI_WR)) {
12598 12596 /*
12599 12597 * A return value of EINPROGRESS means the ioctl is
12600 12598 * either queued and waiting for some reason or has
12601 12599 * already completed.
12602 12600 */
12603 12601 err = (*ipip->ipi_func)(ci.ci_ipif, ci.ci_sin, q, mp, ipip,
12604 12602 ci.ci_lifr);
12605 12603 if (ci.ci_ipif != NULL) {
12606 12604 DTRACE_PROBE4(ipif__ioctl,
12607 12605 char *, "ip_process_ioctl finish RD",
12608 12606 int, ipip->ipi_cmd, ill_t *, ci.ci_ipif->ipif_ill,
12609 12607 ipif_t *, ci.ci_ipif);
12610 12608 ipif_refrele(ci.ci_ipif);
12611 12609 } else {
12612 12610 DTRACE_PROBE4(ipif__ioctl,
12613 12611 char *, "ip_process_ioctl finish RD",
12614 12612 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12615 12613 }
12616 12614 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12617 12615 return;
12618 12616 }
12619 12617
12620 12618 ASSERT(ci.ci_ipif != NULL);
12621 12619
12622 12620 /*
12623 12621 * If ipsq is non-NULL, we are already being called exclusively
12624 12622 */
12625 12623 ASSERT(ipsq == NULL || IAM_WRITER_IPSQ(ipsq));
12626 12624 if (ipsq == NULL) {
12627 12625 ipsq = ipsq_try_enter(ci.ci_ipif, NULL, q, mp, ip_process_ioctl,
12628 12626 NEW_OP, B_TRUE);
12629 12627 if (ipsq == NULL) {
12630 12628 ipif_refrele(ci.ci_ipif);
12631 12629 return;
12632 12630 }
12633 12631 entered_ipsq = B_TRUE;
12634 12632 }
12635 12633 /*
12636 12634 * Release the ipif so that ipif_down and friends that wait for
12637 12635 * references to go away are not misled about the current ipif_refcnt
12638 12636 * values. We are writer so we can access the ipif even after releasing
12639 12637 * the ipif.
12640 12638 */
12641 12639 ipif_refrele(ci.ci_ipif);
12642 12640
12643 12641 ipsq_current_start(ipsq, ci.ci_ipif, ipip->ipi_cmd);
12644 12642
12645 12643 /*
12646 12644 * We need to cache the ill_t that we're going to use as the argument
12647 12645 * to the ipif-ioctl DTrace probe (below) because the ci_ipif can be
12648 12646 * blown away by calling ipi_func.
12649 12647 */
12650 12648 ill = ci.ci_ipif == NULL ? NULL : ci.ci_ipif->ipif_ill;
12651 12649
12652 12650 /*
12653 12651 * A return value of EINPROGRESS means the ioctl is
12654 12652 * either queued and waiting for some reason or has
12655 12653 * already completed.
12656 12654 */
12657 12655 err = (*ipip->ipi_func)(ci.ci_ipif, ci.ci_sin, q, mp, ipip, ci.ci_lifr);
12658 12656
12659 12657 DTRACE_PROBE4(ipif__ioctl, char *, "ip_process_ioctl finish WR",
12660 12658 int, ipip->ipi_cmd, ill_t *, ill, ipif_t *, ci.ci_ipif);
12661 12659 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), ipsq);
12662 12660
12663 12661 if (entered_ipsq)
12664 12662 ipsq_exit(ipsq);
12665 12663 }
12666 12664
12667 12665 /*
12668 12666 * Complete the ioctl. Typically ioctls use the mi package and need to
12669 12667 * do mi_copyout/mi_copy_done.
12670 12668 */
12671 12669 void
12672 12670 ip_ioctl_finish(queue_t *q, mblk_t *mp, int err, int mode, ipsq_t *ipsq)
12673 12671 {
12674 12672 conn_t *connp = NULL;
12675 12673
12676 12674 if (err == EINPROGRESS)
12677 12675 return;
12678 12676
12679 12677 if (CONN_Q(q)) {
12680 12678 connp = Q_TO_CONN(q);
12681 12679 ASSERT(connp->conn_ref >= 2);
12682 12680 }
12683 12681
12684 12682 switch (mode) {
12685 12683 case COPYOUT:
12686 12684 if (err == 0)
12687 12685 mi_copyout(q, mp);
12688 12686 else
12689 12687 mi_copy_done(q, mp, err);
12690 12688 break;
12691 12689
12692 12690 case NO_COPYOUT:
12693 12691 mi_copy_done(q, mp, err);
12694 12692 break;
12695 12693
12696 12694 default:
12697 12695 ASSERT(mode == CONN_CLOSE); /* aborted through CONN_CLOSE */
12698 12696 break;
12699 12697 }
12700 12698
12701 12699 /*
12702 12700 * The conn refhold and ioctlref placed on the conn at the start of the
12703 12701 * ioctl are released here.
12704 12702 */
12705 12703 if (connp != NULL) {
12706 12704 CONN_DEC_IOCTLREF(connp);
12707 12705 CONN_OPER_PENDING_DONE(connp);
12708 12706 }
12709 12707
12710 12708 if (ipsq != NULL)
12711 12709 ipsq_current_finish(ipsq);
12712 12710 }
12713 12711
12714 12712 /* Handles all non data messages */
12715 12713 int
12716 12714 ip_wput_nondata(queue_t *q, mblk_t *mp)
12717 12715 {
12718 12716 mblk_t *mp1;
12719 12717 struct iocblk *iocp;
12720 12718 ip_ioctl_cmd_t *ipip;
12721 12719 conn_t *connp;
12722 12720 cred_t *cr;
12723 12721 char *proto_str;
12724 12722
12725 12723 if (CONN_Q(q))
12726 12724 connp = Q_TO_CONN(q);
12727 12725 else
12728 12726 connp = NULL;
12729 12727
12730 12728 switch (DB_TYPE(mp)) {
12731 12729 case M_IOCTL:
12732 12730 /*
12733 12731 * IOCTL processing begins in ip_sioctl_copyin_setup which
12734 12732 * will arrange to copy in associated control structures.
12735 12733 */
12736 12734 ip_sioctl_copyin_setup(q, mp);
12737 12735 return (0);
12738 12736 case M_IOCDATA:
12739 12737 /*
12740 12738 * Ensure that this is associated with one of our trans-
12741 12739 * parent ioctls. If it's not ours, discard it if we're
12742 12740 * running as a driver, or pass it on if we're a module.
12743 12741 */
12744 12742 iocp = (struct iocblk *)mp->b_rptr;
12745 12743 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12746 12744 if (ipip == NULL) {
12747 12745 if (q->q_next == NULL) {
12748 12746 goto nak;
12749 12747 } else {
12750 12748 putnext(q, mp);
12751 12749 }
12752 12750 return (0);
12753 12751 }
12754 12752 if ((q->q_next != NULL) && !(ipip->ipi_flags & IPI_MODOK)) {
12755 12753 /*
12756 12754 * The ioctl is one we recognise, but is not consumed
12757 12755 * by IP as a module and we are a module, so we drop
12758 12756 */
12759 12757 goto nak;
12760 12758 }
12761 12759
12762 12760 /* IOCTL continuation following copyin or copyout. */
12763 12761 if (mi_copy_state(q, mp, NULL) == -1) {
12764 12762 /*
12765 12763 * The copy operation failed. mi_copy_state already
12766 12764 * cleaned up, so we're out of here.
12767 12765 */
12768 12766 return (0);
12769 12767 }
12770 12768 /*
12771 12769 * If we just completed a copy in, we become writer and
12772 12770 * continue processing in ip_sioctl_copyin_done. If it
12773 12771 * was a copy out, we call mi_copyout again. If there is
12774 12772 * nothing more to copy out, it will complete the IOCTL.
12775 12773 */
12776 12774 if (MI_COPY_DIRECTION(mp) == MI_COPY_IN) {
12777 12775 if (!(mp1 = mp->b_cont) || !(mp1 = mp1->b_cont)) {
12778 12776 mi_copy_done(q, mp, EPROTO);
12779 12777 return (0);
12780 12778 }
12781 12779 /*
12782 12780 * Check for cases that need more copying. A return
12783 12781 * value of 0 means a second copyin has been started,
12784 12782 * so we return; a return value of 1 means no more
12785 12783 * copying is needed, so we continue.
12786 12784 */
12787 12785 if (ipip->ipi_cmd_type == MSFILT_CMD &&
12788 12786 MI_COPY_COUNT(mp) == 1) {
12789 12787 if (ip_copyin_msfilter(q, mp) == 0)
12790 12788 return (0);
12791 12789 }
12792 12790 /*
12793 12791 * Refhold the conn, till the ioctl completes. This is
12794 12792 * needed in case the ioctl ends up in the pending mp
12795 12793 * list. Every mp in the ipx_pending_mp list must have
12796 12794 * a refhold on the conn to resume processing. The
12797 12795 * refhold is released when the ioctl completes
12798 12796 * (whether normally or abnormally). An ioctlref is also
12799 12797 * placed on the conn to prevent TCP from removing the
12800 12798 * queue needed to send the ioctl reply back.
12801 12799 * In all cases ip_ioctl_finish is called to finish
12802 12800 * the ioctl and release the refholds.
12803 12801 */
12804 12802 if (connp != NULL) {
12805 12803 /* This is not a reentry */
12806 12804 CONN_INC_REF(connp);
12807 12805 CONN_INC_IOCTLREF(connp);
12808 12806 } else {
12809 12807 if (!(ipip->ipi_flags & IPI_MODOK)) {
12810 12808 mi_copy_done(q, mp, EINVAL);
12811 12809 return (0);
12812 12810 }
12813 12811 }
12814 12812
12815 12813 ip_process_ioctl(NULL, q, mp, ipip);
12816 12814
12817 12815 } else {
12818 12816 mi_copyout(q, mp);
12819 12817 }
12820 12818 return (0);
12821 12819
12822 12820 case M_IOCNAK:
12823 12821 /*
12824 12822 * The only way we could get here is if a resolver didn't like
12825 12823 * an IOCTL we sent it. This shouldn't happen.
12826 12824 */
12827 12825 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
12828 12826 "ip_wput_nondata: unexpected M_IOCNAK, ioc_cmd 0x%x",
12829 12827 ((struct iocblk *)mp->b_rptr)->ioc_cmd);
12830 12828 freemsg(mp);
12831 12829 return (0);
12832 12830 case M_IOCACK:
12833 12831 /* /dev/ip shouldn't see this */
12834 12832 goto nak;
12835 12833 case M_FLUSH:
12836 12834 if (*mp->b_rptr & FLUSHW)
12837 12835 flushq(q, FLUSHALL);
12838 12836 if (q->q_next) {
12839 12837 putnext(q, mp);
12840 12838 return (0);
12841 12839 }
12842 12840 if (*mp->b_rptr & FLUSHR) {
12843 12841 *mp->b_rptr &= ~FLUSHW;
12844 12842 qreply(q, mp);
12845 12843 return (0);
12846 12844 }
12847 12845 freemsg(mp);
12848 12846 return (0);
12849 12847 case M_CTL:
12850 12848 break;
12851 12849 case M_PROTO:
12852 12850 case M_PCPROTO:
12853 12851 /*
12854 12852 * The only PROTO messages we expect are SNMP-related.
12855 12853 */
12856 12854 switch (((union T_primitives *)mp->b_rptr)->type) {
12857 12855 case T_SVR4_OPTMGMT_REQ:
12858 12856 ip2dbg(("ip_wput_nondata: T_SVR4_OPTMGMT_REQ "
12859 12857 "flags %x\n",
12860 12858 ((struct T_optmgmt_req *)mp->b_rptr)->MGMT_flags));
12861 12859
12862 12860 if (connp == NULL) {
12863 12861 proto_str = "T_SVR4_OPTMGMT_REQ";
12864 12862 goto protonak;
12865 12863 }
12866 12864
12867 12865 /*
12868 12866 * All Solaris components should pass a db_credp
12869 12867 * for this TPI message, hence we ASSERT.
12870 12868 * But in case there is some other M_PROTO that looks
12871 12869 * like a TPI message sent by some other kernel
12872 12870 * component, we check and return an error.
12873 12871 */
12874 12872 cr = msg_getcred(mp, NULL);
12875 12873 ASSERT(cr != NULL);
12876 12874 if (cr == NULL) {
12877 12875 mp = mi_tpi_err_ack_alloc(mp, TSYSERR, EINVAL);
12878 12876 if (mp != NULL)
12879 12877 qreply(q, mp);
12880 12878 return (0);
12881 12879 }
12882 12880
12883 12881 if (!snmpcom_req(q, mp, ip_snmp_set, ip_snmp_get, cr)) {
12884 12882 proto_str = "Bad SNMPCOM request?";
12885 12883 goto protonak;
12886 12884 }
12887 12885 return (0);
12888 12886 default:
12889 12887 ip1dbg(("ip_wput_nondata: dropping M_PROTO prim %u\n",
12890 12888 (int)*(uint_t *)mp->b_rptr));
12891 12889 freemsg(mp);
12892 12890 return (0);
12893 12891 }
12894 12892 default:
12895 12893 break;
12896 12894 }
12897 12895 if (q->q_next) {
12898 12896 putnext(q, mp);
12899 12897 } else
12900 12898 freemsg(mp);
12901 12899 return (0);
12902 12900
12903 12901 nak:
12904 12902 iocp->ioc_error = EINVAL;
12905 12903 mp->b_datap->db_type = M_IOCNAK;
12906 12904 iocp->ioc_count = 0;
12907 12905 qreply(q, mp);
12908 12906 return (0);
12909 12907
12910 12908 protonak:
12911 12909 cmn_err(CE_NOTE, "IP doesn't process %s as a module", proto_str);
12912 12910 if ((mp = mi_tpi_err_ack_alloc(mp, TPROTO, EINVAL)) != NULL)
12913 12911 qreply(q, mp);
12914 12912 return (0);
12915 12913 }
12916 12914
12917 12915 /*
12918 12916 * Process IP options in an outbound packet. Verify that the nexthop in a
12919 12917 * strict source route is onlink.
12920 12918 * Returns non-zero if something fails in which case an ICMP error has been
12921 12919 * sent and mp freed.
12922 12920 *
12923 12921 * Assumes the ULP has called ip_massage_options to move nexthop into ipha_dst.
12924 12922 */
12925 12923 int
12926 12924 ip_output_options(mblk_t *mp, ipha_t *ipha, ip_xmit_attr_t *ixa, ill_t *ill)
12927 12925 {
12928 12926 ipoptp_t opts;
12929 12927 uchar_t *opt;
12930 12928 uint8_t optval;
12931 12929 uint8_t optlen;
12932 12930 ipaddr_t dst;
12933 12931 intptr_t code = 0;
12934 12932 ire_t *ire;
12935 12933 ip_stack_t *ipst = ixa->ixa_ipst;
12936 12934 ip_recv_attr_t iras;
12937 12935
12938 12936 ip2dbg(("ip_output_options\n"));
12939 12937
12940 12938 dst = ipha->ipha_dst;
12941 12939 for (optval = ipoptp_first(&opts, ipha);
12942 12940 optval != IPOPT_EOL;
12943 12941 optval = ipoptp_next(&opts)) {
12944 12942 opt = opts.ipoptp_cur;
12945 12943 optlen = opts.ipoptp_len;
12946 12944 ip2dbg(("ip_output_options: opt %d, len %d\n",
12947 12945 optval, optlen));
12948 12946 switch (optval) {
12949 12947 uint32_t off;
12950 12948 case IPOPT_SSRR:
12951 12949 case IPOPT_LSRR:
12952 12950 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
12953 12951 ip1dbg((
12954 12952 "ip_output_options: bad option offset\n"));
12955 12953 code = (char *)&opt[IPOPT_OLEN] -
12956 12954 (char *)ipha;
12957 12955 goto param_prob;
12958 12956 }
12959 12957 off = opt[IPOPT_OFFSET];
12960 12958 ip1dbg(("ip_output_options: next hop 0x%x\n",
12961 12959 ntohl(dst)));
12962 12960 /*
12963 12961 * For strict: verify that dst is directly
12964 12962 * reachable.
12965 12963 */
12966 12964 if (optval == IPOPT_SSRR) {
12967 12965 ire = ire_ftable_lookup_v4(dst, 0, 0,
12968 12966 IRE_INTERFACE, NULL, ALL_ZONES,
12969 12967 ixa->ixa_tsl,
12970 12968 MATCH_IRE_TYPE | MATCH_IRE_SECATTR, 0, ipst,
12971 12969 NULL);
12972 12970 if (ire == NULL) {
12973 12971 ip1dbg(("ip_output_options: SSRR not"
12974 12972 " directly reachable: 0x%x\n",
12975 12973 ntohl(dst)));
12976 12974 goto bad_src_route;
12977 12975 }
12978 12976 ire_refrele(ire);
12979 12977 }
12980 12978 break;
12981 12979 case IPOPT_RR:
12982 12980 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
12983 12981 ip1dbg((
12984 12982 "ip_output_options: bad option offset\n"));
12985 12983 code = (char *)&opt[IPOPT_OLEN] -
12986 12984 (char *)ipha;
12987 12985 goto param_prob;
12988 12986 }
12989 12987 break;
12990 12988 case IPOPT_TS:
12991 12989 /*
12992 12990 * Verify that length >=5 and that there is either
12993 12991 * room for another timestamp or that the overflow
12994 12992 * counter is not maxed out.
12995 12993 */
12996 12994 code = (char *)&opt[IPOPT_OLEN] - (char *)ipha;
12997 12995 if (optlen < IPOPT_MINLEN_IT) {
12998 12996 goto param_prob;
12999 12997 }
13000 12998 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
13001 12999 ip1dbg((
13002 13000 "ip_output_options: bad option offset\n"));
13003 13001 code = (char *)&opt[IPOPT_OFFSET] -
13004 13002 (char *)ipha;
13005 13003 goto param_prob;
13006 13004 }
13007 13005 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
13008 13006 case IPOPT_TS_TSONLY:
13009 13007 off = IPOPT_TS_TIMELEN;
13010 13008 break;
13011 13009 case IPOPT_TS_TSANDADDR:
13012 13010 case IPOPT_TS_PRESPEC:
13013 13011 case IPOPT_TS_PRESPEC_RFC791:
13014 13012 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
13015 13013 break;
13016 13014 default:
13017 13015 code = (char *)&opt[IPOPT_POS_OV_FLG] -
13018 13016 (char *)ipha;
13019 13017 goto param_prob;
13020 13018 }
13021 13019 if (opt[IPOPT_OFFSET] - 1 + off > optlen &&
13022 13020 (opt[IPOPT_POS_OV_FLG] & 0xF0) == 0xF0) {
13023 13021 /*
13024 13022 * No room and the overflow counter is 15
13025 13023 * already.
13026 13024 */
13027 13025 goto param_prob;
13028 13026 }
13029 13027 break;
13030 13028 }
13031 13029 }
13032 13030
13033 13031 if ((opts.ipoptp_flags & IPOPTP_ERROR) == 0)
13034 13032 return (0);
13035 13033
13036 13034 ip1dbg(("ip_output_options: error processing IP options."));
13037 13035 code = (char *)&opt[IPOPT_OFFSET] - (char *)ipha;
13038 13036
13039 13037 param_prob:
13040 13038 bzero(&iras, sizeof (iras));
13041 13039 iras.ira_ill = iras.ira_rill = ill;
13042 13040 iras.ira_ruifindex = ill->ill_phyint->phyint_ifindex;
13043 13041 iras.ira_rifindex = iras.ira_ruifindex;
13044 13042 iras.ira_flags = IRAF_IS_IPV4;
13045 13043
13046 13044 ip_drop_output("ip_output_options", mp, ill);
13047 13045 icmp_param_problem(mp, (uint8_t)code, &iras);
13048 13046 ASSERT(!(iras.ira_flags & IRAF_IPSEC_SECURE));
13049 13047 return (-1);
13050 13048
13051 13049 bad_src_route:
13052 13050 bzero(&iras, sizeof (iras));
13053 13051 iras.ira_ill = iras.ira_rill = ill;
13054 13052 iras.ira_ruifindex = ill->ill_phyint->phyint_ifindex;
13055 13053 iras.ira_rifindex = iras.ira_ruifindex;
13056 13054 iras.ira_flags = IRAF_IS_IPV4;
13057 13055
13058 13056 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ill);
13059 13057 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, &iras);
13060 13058 ASSERT(!(iras.ira_flags & IRAF_IPSEC_SECURE));
13061 13059 return (-1);
13062 13060 }
13063 13061
13064 13062 /*
13065 13063 * The maximum value of conn_drain_list_cnt is CONN_MAXDRAINCNT.
13066 13064 * conn_drain_list_cnt can be changed by setting conn_drain_nthreads
13067 13065 * thru /etc/system.
13068 13066 */
13069 13067 #define CONN_MAXDRAINCNT 64
13070 13068
13071 13069 static void
13072 13070 conn_drain_init(ip_stack_t *ipst)
13073 13071 {
13074 13072 int i, j;
13075 13073 idl_tx_list_t *itl_tx;
13076 13074
13077 13075 ipst->ips_conn_drain_list_cnt = conn_drain_nthreads;
13078 13076
13079 13077 if ((ipst->ips_conn_drain_list_cnt == 0) ||
13080 13078 (ipst->ips_conn_drain_list_cnt > CONN_MAXDRAINCNT)) {
13081 13079 /*
13082 13080 * Default value of the number of drainers is the
13083 13081 * number of cpus, subject to maximum of 8 drainers.
13084 13082 */
13085 13083 if (boot_max_ncpus != -1)
13086 13084 ipst->ips_conn_drain_list_cnt = MIN(boot_max_ncpus, 8);
13087 13085 else
13088 13086 ipst->ips_conn_drain_list_cnt = MIN(max_ncpus, 8);
13089 13087 }
13090 13088
13091 13089 ipst->ips_idl_tx_list =
13092 13090 kmem_zalloc(TX_FANOUT_SIZE * sizeof (idl_tx_list_t), KM_SLEEP);
13093 13091 for (i = 0; i < TX_FANOUT_SIZE; i++) {
13094 13092 itl_tx = &ipst->ips_idl_tx_list[i];
13095 13093 itl_tx->txl_drain_list =
13096 13094 kmem_zalloc(ipst->ips_conn_drain_list_cnt *
13097 13095 sizeof (idl_t), KM_SLEEP);
13098 13096 mutex_init(&itl_tx->txl_lock, NULL, MUTEX_DEFAULT, NULL);
13099 13097 for (j = 0; j < ipst->ips_conn_drain_list_cnt; j++) {
13100 13098 mutex_init(&itl_tx->txl_drain_list[j].idl_lock, NULL,
13101 13099 MUTEX_DEFAULT, NULL);
13102 13100 itl_tx->txl_drain_list[j].idl_itl = itl_tx;
13103 13101 }
13104 13102 }
13105 13103 }
13106 13104
13107 13105 static void
13108 13106 conn_drain_fini(ip_stack_t *ipst)
13109 13107 {
13110 13108 int i;
13111 13109 idl_tx_list_t *itl_tx;
13112 13110
13113 13111 for (i = 0; i < TX_FANOUT_SIZE; i++) {
13114 13112 itl_tx = &ipst->ips_idl_tx_list[i];
13115 13113 kmem_free(itl_tx->txl_drain_list,
13116 13114 ipst->ips_conn_drain_list_cnt * sizeof (idl_t));
13117 13115 }
13118 13116 kmem_free(ipst->ips_idl_tx_list,
13119 13117 TX_FANOUT_SIZE * sizeof (idl_tx_list_t));
13120 13118 ipst->ips_idl_tx_list = NULL;
13121 13119 }
13122 13120
13123 13121 /*
13124 13122 * Flow control has blocked us from proceeding. Insert the given conn in one
13125 13123 * of the conn drain lists. When flow control is unblocked, either ip_wsrv()
13126 13124 * (STREAMS) or ill_flow_enable() (direct) will be called back, which in turn
13127 13125 * will call conn_walk_drain(). See the flow control notes at the top of this
13128 13126 * file for more details.
13129 13127 */
13130 13128 void
13131 13129 conn_drain_insert(conn_t *connp, idl_tx_list_t *tx_list)
13132 13130 {
13133 13131 idl_t *idl = tx_list->txl_drain_list;
13134 13132 uint_t index;
13135 13133 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
13136 13134
13137 13135 mutex_enter(&connp->conn_lock);
13138 13136 if (connp->conn_state_flags & CONN_CLOSING) {
13139 13137 /*
13140 13138 * The conn is closing as a result of which CONN_CLOSING
13141 13139 * is set. Return.
13142 13140 */
13143 13141 mutex_exit(&connp->conn_lock);
13144 13142 return;
13145 13143 } else if (connp->conn_idl == NULL) {
13146 13144 /*
13147 13145 * Assign the next drain list round robin. We dont' use
13148 13146 * a lock, and thus it may not be strictly round robin.
13149 13147 * Atomicity of load/stores is enough to make sure that
13150 13148 * conn_drain_list_index is always within bounds.
13151 13149 */
13152 13150 index = tx_list->txl_drain_index;
13153 13151 ASSERT(index < ipst->ips_conn_drain_list_cnt);
13154 13152 connp->conn_idl = &tx_list->txl_drain_list[index];
13155 13153 index++;
13156 13154 if (index == ipst->ips_conn_drain_list_cnt)
13157 13155 index = 0;
13158 13156 tx_list->txl_drain_index = index;
13159 13157 } else {
13160 13158 ASSERT(connp->conn_idl->idl_itl == tx_list);
13161 13159 }
13162 13160 mutex_exit(&connp->conn_lock);
13163 13161
13164 13162 idl = connp->conn_idl;
13165 13163 mutex_enter(&idl->idl_lock);
13166 13164 if ((connp->conn_drain_prev != NULL) ||
13167 13165 (connp->conn_state_flags & CONN_CLOSING)) {
13168 13166 /*
13169 13167 * The conn is either already in the drain list or closing.
13170 13168 * (We needed to check for CONN_CLOSING again since close can
13171 13169 * sneak in between dropping conn_lock and acquiring idl_lock.)
13172 13170 */
13173 13171 mutex_exit(&idl->idl_lock);
13174 13172 return;
13175 13173 }
13176 13174
13177 13175 /*
13178 13176 * The conn is not in the drain list. Insert it at the
13179 13177 * tail of the drain list. The drain list is circular
13180 13178 * and doubly linked. idl_conn points to the 1st element
13181 13179 * in the list.
13182 13180 */
13183 13181 if (idl->idl_conn == NULL) {
13184 13182 idl->idl_conn = connp;
13185 13183 connp->conn_drain_next = connp;
13186 13184 connp->conn_drain_prev = connp;
13187 13185 } else {
13188 13186 conn_t *head = idl->idl_conn;
13189 13187
13190 13188 connp->conn_drain_next = head;
13191 13189 connp->conn_drain_prev = head->conn_drain_prev;
13192 13190 head->conn_drain_prev->conn_drain_next = connp;
13193 13191 head->conn_drain_prev = connp;
13194 13192 }
13195 13193 /*
13196 13194 * For non streams based sockets assert flow control.
13197 13195 */
13198 13196 conn_setqfull(connp, NULL);
13199 13197 mutex_exit(&idl->idl_lock);
13200 13198 }
13201 13199
13202 13200 static void
13203 13201 conn_drain_remove(conn_t *connp)
13204 13202 {
13205 13203 idl_t *idl = connp->conn_idl;
13206 13204
13207 13205 if (idl != NULL) {
13208 13206 /*
13209 13207 * Remove ourself from the drain list.
13210 13208 */
13211 13209 if (connp->conn_drain_next == connp) {
13212 13210 /* Singleton in the list */
13213 13211 ASSERT(connp->conn_drain_prev == connp);
13214 13212 idl->idl_conn = NULL;
13215 13213 } else {
13216 13214 connp->conn_drain_prev->conn_drain_next =
13217 13215 connp->conn_drain_next;
13218 13216 connp->conn_drain_next->conn_drain_prev =
13219 13217 connp->conn_drain_prev;
13220 13218 if (idl->idl_conn == connp)
13221 13219 idl->idl_conn = connp->conn_drain_next;
13222 13220 }
13223 13221
13224 13222 /*
13225 13223 * NOTE: because conn_idl is associated with a specific drain
13226 13224 * list which in turn is tied to the index the TX ring
13227 13225 * (txl_cookie) hashes to, and because the TX ring can change
13228 13226 * over the lifetime of the conn_t, we must clear conn_idl so
13229 13227 * a subsequent conn_drain_insert() will set conn_idl again
13230 13228 * based on the latest txl_cookie.
13231 13229 */
13232 13230 connp->conn_idl = NULL;
13233 13231 }
13234 13232 connp->conn_drain_next = NULL;
13235 13233 connp->conn_drain_prev = NULL;
13236 13234
13237 13235 conn_clrqfull(connp, NULL);
13238 13236 /*
13239 13237 * For streams based sockets open up flow control.
13240 13238 */
13241 13239 if (!IPCL_IS_NONSTR(connp))
13242 13240 enableok(connp->conn_wq);
13243 13241 }
13244 13242
13245 13243 /*
13246 13244 * This conn is closing, and we are called from ip_close. OR
13247 13245 * this conn is draining because flow-control on the ill has been relieved.
13248 13246 *
13249 13247 * We must also need to remove conn's on this idl from the list, and also
13250 13248 * inform the sockfs upcalls about the change in flow-control.
13251 13249 */
13252 13250 static void
13253 13251 conn_drain(conn_t *connp, boolean_t closing)
13254 13252 {
13255 13253 idl_t *idl;
13256 13254 conn_t *next_connp;
13257 13255
13258 13256 /*
13259 13257 * connp->conn_idl is stable at this point, and no lock is needed
13260 13258 * to check it. If we are called from ip_close, close has already
13261 13259 * set CONN_CLOSING, thus freezing the value of conn_idl, and
13262 13260 * called us only because conn_idl is non-null. If we are called thru
13263 13261 * service, conn_idl could be null, but it cannot change because
13264 13262 * service is single-threaded per queue, and there cannot be another
13265 13263 * instance of service trying to call conn_drain_insert on this conn
13266 13264 * now.
13267 13265 */
13268 13266 ASSERT(!closing || connp == NULL || connp->conn_idl != NULL);
13269 13267
13270 13268 /*
13271 13269 * If the conn doesn't exist or is not on a drain list, bail.
13272 13270 */
13273 13271 if (connp == NULL || connp->conn_idl == NULL ||
13274 13272 connp->conn_drain_prev == NULL) {
13275 13273 return;
13276 13274 }
13277 13275
13278 13276 idl = connp->conn_idl;
13279 13277 ASSERT(MUTEX_HELD(&idl->idl_lock));
13280 13278
13281 13279 if (!closing) {
13282 13280 next_connp = connp->conn_drain_next;
13283 13281 while (next_connp != connp) {
13284 13282 conn_t *delconnp = next_connp;
13285 13283
13286 13284 next_connp = next_connp->conn_drain_next;
13287 13285 conn_drain_remove(delconnp);
13288 13286 }
13289 13287 ASSERT(connp->conn_drain_next == idl->idl_conn);
13290 13288 }
13291 13289 conn_drain_remove(connp);
13292 13290 }
13293 13291
13294 13292 /*
13295 13293 * Write service routine. Shared perimeter entry point.
13296 13294 * The device queue's messages has fallen below the low water mark and STREAMS
13297 13295 * has backenabled the ill_wq. Send sockfs notification about flow-control on
13298 13296 * each waiting conn.
13299 13297 */
13300 13298 int
13301 13299 ip_wsrv(queue_t *q)
13302 13300 {
13303 13301 ill_t *ill;
13304 13302
13305 13303 ill = (ill_t *)q->q_ptr;
13306 13304 if (ill->ill_state_flags == 0) {
13307 13305 ip_stack_t *ipst = ill->ill_ipst;
13308 13306
13309 13307 /*
13310 13308 * The device flow control has opened up.
13311 13309 * Walk through conn drain lists and qenable the
13312 13310 * first conn in each list. This makes sense only
13313 13311 * if the stream is fully plumbed and setup.
13314 13312 * Hence the ill_state_flags check above.
13315 13313 */
13316 13314 ip1dbg(("ip_wsrv: walking\n"));
13317 13315 conn_walk_drain(ipst, &ipst->ips_idl_tx_list[0]);
13318 13316 enableok(ill->ill_wq);
13319 13317 }
13320 13318 return (0);
13321 13319 }
13322 13320
13323 13321 /*
13324 13322 * Callback to disable flow control in IP.
13325 13323 *
13326 13324 * This is a mac client callback added when the DLD_CAPAB_DIRECT capability
13327 13325 * is enabled.
13328 13326 *
13329 13327 * When MAC_TX() is not able to send any more packets, dld sets its queue
13330 13328 * to QFULL and enable the STREAMS flow control. Later, when the underlying
13331 13329 * driver is able to continue to send packets, it calls mac_tx_(ring_)update()
13332 13330 * function and wakes up corresponding mac worker threads, which in turn
13333 13331 * calls this callback function, and disables flow control.
13334 13332 */
13335 13333 void
13336 13334 ill_flow_enable(void *arg, ip_mac_tx_cookie_t cookie)
13337 13335 {
13338 13336 ill_t *ill = (ill_t *)arg;
13339 13337 ip_stack_t *ipst = ill->ill_ipst;
13340 13338 idl_tx_list_t *idl_txl;
13341 13339
13342 13340 idl_txl = &ipst->ips_idl_tx_list[IDLHASHINDEX(cookie)];
13343 13341 mutex_enter(&idl_txl->txl_lock);
13344 13342 /* add code to to set a flag to indicate idl_txl is enabled */
13345 13343 conn_walk_drain(ipst, idl_txl);
13346 13344 mutex_exit(&idl_txl->txl_lock);
13347 13345 }
13348 13346
13349 13347 /*
13350 13348 * Flow control has been relieved and STREAMS has backenabled us; drain
13351 13349 * all the conn lists on `tx_list'.
13352 13350 */
13353 13351 static void
13354 13352 conn_walk_drain(ip_stack_t *ipst, idl_tx_list_t *tx_list)
13355 13353 {
13356 13354 int i;
13357 13355 idl_t *idl;
13358 13356
13359 13357 IP_STAT(ipst, ip_conn_walk_drain);
13360 13358
13361 13359 for (i = 0; i < ipst->ips_conn_drain_list_cnt; i++) {
13362 13360 idl = &tx_list->txl_drain_list[i];
13363 13361 mutex_enter(&idl->idl_lock);
13364 13362 conn_drain(idl->idl_conn, B_FALSE);
13365 13363 mutex_exit(&idl->idl_lock);
13366 13364 }
13367 13365 }
13368 13366
13369 13367 /*
13370 13368 * Determine if the ill and multicast aspects of that packets
13371 13369 * "matches" the conn.
13372 13370 */
13373 13371 boolean_t
13374 13372 conn_wantpacket(conn_t *connp, ip_recv_attr_t *ira, ipha_t *ipha)
13375 13373 {
13376 13374 ill_t *ill = ira->ira_rill;
13377 13375 zoneid_t zoneid = ira->ira_zoneid;
13378 13376 uint_t in_ifindex;
13379 13377 ipaddr_t dst, src;
13380 13378
13381 13379 dst = ipha->ipha_dst;
13382 13380 src = ipha->ipha_src;
13383 13381
13384 13382 /*
13385 13383 * conn_incoming_ifindex is set by IP_BOUND_IF which limits
13386 13384 * unicast, broadcast and multicast reception to
13387 13385 * conn_incoming_ifindex.
13388 13386 * conn_wantpacket is called for unicast, broadcast and
13389 13387 * multicast packets.
13390 13388 */
13391 13389 in_ifindex = connp->conn_incoming_ifindex;
13392 13390
13393 13391 /* mpathd can bind to the under IPMP interface, which we allow */
13394 13392 if (in_ifindex != 0 && in_ifindex != ill->ill_phyint->phyint_ifindex) {
13395 13393 if (!IS_UNDER_IPMP(ill))
13396 13394 return (B_FALSE);
13397 13395
13398 13396 if (in_ifindex != ipmp_ill_get_ipmp_ifindex(ill))
13399 13397 return (B_FALSE);
13400 13398 }
13401 13399
13402 13400 if (!IPCL_ZONE_MATCH(connp, zoneid))
13403 13401 return (B_FALSE);
13404 13402
13405 13403 if (!(ira->ira_flags & IRAF_MULTICAST))
13406 13404 return (B_TRUE);
13407 13405
13408 13406 if (connp->conn_multi_router) {
13409 13407 /* multicast packet and multicast router socket: send up */
13410 13408 return (B_TRUE);
13411 13409 }
13412 13410
13413 13411 if (ipha->ipha_protocol == IPPROTO_PIM ||
13414 13412 ipha->ipha_protocol == IPPROTO_RSVP)
13415 13413 return (B_TRUE);
13416 13414
13417 13415 return (conn_hasmembers_ill_withsrc_v4(connp, dst, src, ira->ira_ill));
13418 13416 }
13419 13417
13420 13418 void
13421 13419 conn_setqfull(conn_t *connp, boolean_t *flow_stopped)
13422 13420 {
13423 13421 if (IPCL_IS_NONSTR(connp)) {
13424 13422 (*connp->conn_upcalls->su_txq_full)
13425 13423 (connp->conn_upper_handle, B_TRUE);
13426 13424 if (flow_stopped != NULL)
13427 13425 *flow_stopped = B_TRUE;
13428 13426 } else {
13429 13427 queue_t *q = connp->conn_wq;
13430 13428
13431 13429 ASSERT(q != NULL);
13432 13430 if (!(q->q_flag & QFULL)) {
13433 13431 mutex_enter(QLOCK(q));
13434 13432 if (!(q->q_flag & QFULL)) {
13435 13433 /* still need to set QFULL */
13436 13434 q->q_flag |= QFULL;
13437 13435 /* set flow_stopped to true under QLOCK */
13438 13436 if (flow_stopped != NULL)
13439 13437 *flow_stopped = B_TRUE;
13440 13438 mutex_exit(QLOCK(q));
13441 13439 } else {
13442 13440 /* flow_stopped is left unchanged */
13443 13441 mutex_exit(QLOCK(q));
13444 13442 }
13445 13443 }
13446 13444 }
13447 13445 }
13448 13446
13449 13447 void
13450 13448 conn_clrqfull(conn_t *connp, boolean_t *flow_stopped)
13451 13449 {
13452 13450 if (IPCL_IS_NONSTR(connp)) {
13453 13451 (*connp->conn_upcalls->su_txq_full)
13454 13452 (connp->conn_upper_handle, B_FALSE);
13455 13453 if (flow_stopped != NULL)
13456 13454 *flow_stopped = B_FALSE;
13457 13455 } else {
13458 13456 queue_t *q = connp->conn_wq;
13459 13457
13460 13458 ASSERT(q != NULL);
13461 13459 if (q->q_flag & QFULL) {
13462 13460 mutex_enter(QLOCK(q));
13463 13461 if (q->q_flag & QFULL) {
13464 13462 q->q_flag &= ~QFULL;
13465 13463 /* set flow_stopped to false under QLOCK */
13466 13464 if (flow_stopped != NULL)
13467 13465 *flow_stopped = B_FALSE;
13468 13466 mutex_exit(QLOCK(q));
13469 13467 if (q->q_flag & QWANTW)
13470 13468 qbackenable(q, 0);
13471 13469 } else {
13472 13470 /* flow_stopped is left unchanged */
13473 13471 mutex_exit(QLOCK(q));
13474 13472 }
13475 13473 }
13476 13474 }
13477 13475
13478 13476 mutex_enter(&connp->conn_lock);
13479 13477 connp->conn_blocked = B_FALSE;
13480 13478 mutex_exit(&connp->conn_lock);
13481 13479 }
13482 13480
13483 13481 /*
13484 13482 * Return the length in bytes of the IPv4 headers (base header, label, and
13485 13483 * other IP options) that will be needed based on the
13486 13484 * ip_pkt_t structure passed by the caller.
13487 13485 *
13488 13486 * The returned length does not include the length of the upper level
13489 13487 * protocol (ULP) header.
13490 13488 * The caller needs to check that the length doesn't exceed the max for IPv4.
13491 13489 */
13492 13490 int
13493 13491 ip_total_hdrs_len_v4(const ip_pkt_t *ipp)
13494 13492 {
13495 13493 int len;
13496 13494
13497 13495 len = IP_SIMPLE_HDR_LENGTH;
13498 13496 if (ipp->ipp_fields & IPPF_LABEL_V4) {
13499 13497 ASSERT(ipp->ipp_label_len_v4 != 0);
13500 13498 /* We need to round up here */
13501 13499 len += (ipp->ipp_label_len_v4 + 3) & ~3;
13502 13500 }
13503 13501
13504 13502 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
13505 13503 ASSERT(ipp->ipp_ipv4_options_len != 0);
13506 13504 ASSERT((ipp->ipp_ipv4_options_len & 3) == 0);
13507 13505 len += ipp->ipp_ipv4_options_len;
13508 13506 }
13509 13507 return (len);
13510 13508 }
13511 13509
13512 13510 /*
13513 13511 * All-purpose routine to build an IPv4 header with options based
13514 13512 * on the abstract ip_pkt_t.
13515 13513 *
13516 13514 * The caller has to set the source and destination address as well as
13517 13515 * ipha_length. The caller has to massage any source route and compensate
13518 13516 * for the ULP pseudo-header checksum due to the source route.
13519 13517 */
13520 13518 void
13521 13519 ip_build_hdrs_v4(uchar_t *buf, uint_t buf_len, const ip_pkt_t *ipp,
13522 13520 uint8_t protocol)
13523 13521 {
13524 13522 ipha_t *ipha = (ipha_t *)buf;
13525 13523 uint8_t *cp;
13526 13524
13527 13525 /* Initialize IPv4 header */
13528 13526 ipha->ipha_type_of_service = ipp->ipp_type_of_service;
13529 13527 ipha->ipha_length = 0; /* Caller will set later */
13530 13528 ipha->ipha_ident = 0;
13531 13529 ipha->ipha_fragment_offset_and_flags = 0;
13532 13530 ipha->ipha_ttl = ipp->ipp_unicast_hops;
13533 13531 ipha->ipha_protocol = protocol;
13534 13532 ipha->ipha_hdr_checksum = 0;
13535 13533
13536 13534 if ((ipp->ipp_fields & IPPF_ADDR) &&
13537 13535 IN6_IS_ADDR_V4MAPPED(&ipp->ipp_addr))
13538 13536 ipha->ipha_src = ipp->ipp_addr_v4;
13539 13537
13540 13538 cp = (uint8_t *)&ipha[1];
13541 13539 if (ipp->ipp_fields & IPPF_LABEL_V4) {
13542 13540 ASSERT(ipp->ipp_label_len_v4 != 0);
13543 13541 bcopy(ipp->ipp_label_v4, cp, ipp->ipp_label_len_v4);
13544 13542 cp += ipp->ipp_label_len_v4;
13545 13543 /* We need to round up here */
13546 13544 while ((uintptr_t)cp & 0x3) {
13547 13545 *cp++ = IPOPT_NOP;
13548 13546 }
13549 13547 }
13550 13548
13551 13549 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
13552 13550 ASSERT(ipp->ipp_ipv4_options_len != 0);
13553 13551 ASSERT((ipp->ipp_ipv4_options_len & 3) == 0);
13554 13552 bcopy(ipp->ipp_ipv4_options, cp, ipp->ipp_ipv4_options_len);
13555 13553 cp += ipp->ipp_ipv4_options_len;
13556 13554 }
13557 13555 ipha->ipha_version_and_hdr_length =
13558 13556 (uint8_t)((IP_VERSION << 4) + buf_len / 4);
13559 13557
13560 13558 ASSERT((int)(cp - buf) == buf_len);
13561 13559 }
13562 13560
13563 13561 /* Allocate the private structure */
13564 13562 static int
13565 13563 ip_priv_alloc(void **bufp)
13566 13564 {
13567 13565 void *buf;
13568 13566
13569 13567 if ((buf = kmem_alloc(sizeof (ip_priv_t), KM_NOSLEEP)) == NULL)
13570 13568 return (ENOMEM);
13571 13569
13572 13570 *bufp = buf;
13573 13571 return (0);
13574 13572 }
13575 13573
13576 13574 /* Function to delete the private structure */
13577 13575 void
13578 13576 ip_priv_free(void *buf)
13579 13577 {
13580 13578 ASSERT(buf != NULL);
13581 13579 kmem_free(buf, sizeof (ip_priv_t));
13582 13580 }
13583 13581
13584 13582 /*
13585 13583 * The entry point for IPPF processing.
13586 13584 * If the classifier (IPGPC_CLASSIFY) is not loaded and configured, the
13587 13585 * routine just returns.
13588 13586 *
13589 13587 * When called, ip_process generates an ipp_packet_t structure
13590 13588 * which holds the state information for this packet and invokes the
13591 13589 * the classifier (via ipp_packet_process). The classification, depending on
13592 13590 * configured filters, results in a list of actions for this packet. Invoking
13593 13591 * an action may cause the packet to be dropped, in which case we return NULL.
13594 13592 * proc indicates the callout position for
13595 13593 * this packet and ill is the interface this packet arrived on or will leave
13596 13594 * on (inbound and outbound resp.).
13597 13595 *
13598 13596 * We do the processing on the rill (mapped to the upper if ipmp), but MIB
13599 13597 * on the ill corrsponding to the destination IP address.
13600 13598 */
13601 13599 mblk_t *
13602 13600 ip_process(ip_proc_t proc, mblk_t *mp, ill_t *rill, ill_t *ill)
13603 13601 {
13604 13602 ip_priv_t *priv;
13605 13603 ipp_action_id_t aid;
13606 13604 int rc = 0;
13607 13605 ipp_packet_t *pp;
13608 13606
13609 13607 /* If the classifier is not loaded, return */
13610 13608 if ((aid = ipp_action_lookup(IPGPC_CLASSIFY)) == IPP_ACTION_INVAL) {
13611 13609 return (mp);
13612 13610 }
13613 13611
13614 13612 ASSERT(mp != NULL);
13615 13613
13616 13614 /* Allocate the packet structure */
13617 13615 rc = ipp_packet_alloc(&pp, "ip", aid);
13618 13616 if (rc != 0)
13619 13617 goto drop;
13620 13618
13621 13619 /* Allocate the private structure */
13622 13620 rc = ip_priv_alloc((void **)&priv);
13623 13621 if (rc != 0) {
13624 13622 ipp_packet_free(pp);
13625 13623 goto drop;
13626 13624 }
13627 13625 priv->proc = proc;
13628 13626 priv->ill_index = ill_get_upper_ifindex(rill);
13629 13627
13630 13628 ipp_packet_set_private(pp, priv, ip_priv_free);
13631 13629 ipp_packet_set_data(pp, mp);
13632 13630
13633 13631 /* Invoke the classifier */
13634 13632 rc = ipp_packet_process(&pp);
13635 13633 if (pp != NULL) {
13636 13634 mp = ipp_packet_get_data(pp);
13637 13635 ipp_packet_free(pp);
13638 13636 if (rc != 0)
13639 13637 goto drop;
13640 13638 return (mp);
13641 13639 } else {
13642 13640 /* No mp to trace in ip_drop_input/ip_drop_output */
13643 13641 mp = NULL;
13644 13642 }
13645 13643 drop:
13646 13644 if (proc == IPP_LOCAL_IN || proc == IPP_FWD_IN) {
13647 13645 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
13648 13646 ip_drop_input("ip_process", mp, ill);
13649 13647 } else {
13650 13648 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
13651 13649 ip_drop_output("ip_process", mp, ill);
13652 13650 }
13653 13651 freemsg(mp);
13654 13652 return (NULL);
13655 13653 }
13656 13654
13657 13655 /*
13658 13656 * Propagate a multicast group membership operation (add/drop) on
13659 13657 * all the interfaces crossed by the related multirt routes.
13660 13658 * The call is considered successful if the operation succeeds
13661 13659 * on at least one interface.
13662 13660 *
13663 13661 * This assumes that a set of IRE_HOST/RTF_MULTIRT has been created for the
13664 13662 * multicast addresses with the ire argument being the first one.
13665 13663 * We walk the bucket to find all the of those.
13666 13664 *
13667 13665 * Common to IPv4 and IPv6.
13668 13666 */
13669 13667 static int
13670 13668 ip_multirt_apply_membership(int (*fn)(conn_t *, boolean_t,
13671 13669 const in6_addr_t *, ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *),
13672 13670 ire_t *ire, conn_t *connp, boolean_t checkonly, const in6_addr_t *v6group,
13673 13671 mcast_record_t fmode, const in6_addr_t *v6src)
13674 13672 {
13675 13673 ire_t *ire_gw;
13676 13674 irb_t *irb;
13677 13675 int ifindex;
13678 13676 int error = 0;
13679 13677 int result;
13680 13678 ip_stack_t *ipst = ire->ire_ipst;
13681 13679 ipaddr_t group;
13682 13680 boolean_t isv6;
13683 13681 int match_flags;
13684 13682
13685 13683 if (IN6_IS_ADDR_V4MAPPED(v6group)) {
13686 13684 IN6_V4MAPPED_TO_IPADDR(v6group, group);
13687 13685 isv6 = B_FALSE;
13688 13686 } else {
13689 13687 isv6 = B_TRUE;
13690 13688 }
13691 13689
13692 13690 irb = ire->ire_bucket;
13693 13691 ASSERT(irb != NULL);
13694 13692
13695 13693 result = 0;
13696 13694 irb_refhold(irb);
13697 13695 for (; ire != NULL; ire = ire->ire_next) {
13698 13696 if ((ire->ire_flags & RTF_MULTIRT) == 0)
13699 13697 continue;
13700 13698
13701 13699 /* We handle -ifp routes by matching on the ill if set */
13702 13700 match_flags = MATCH_IRE_TYPE;
13703 13701 if (ire->ire_ill != NULL)
13704 13702 match_flags |= MATCH_IRE_ILL;
13705 13703
13706 13704 if (isv6) {
13707 13705 if (!IN6_ARE_ADDR_EQUAL(&ire->ire_addr_v6, v6group))
13708 13706 continue;
13709 13707
13710 13708 ire_gw = ire_ftable_lookup_v6(&ire->ire_gateway_addr_v6,
13711 13709 0, 0, IRE_INTERFACE, ire->ire_ill, ALL_ZONES, NULL,
13712 13710 match_flags, 0, ipst, NULL);
13713 13711 } else {
13714 13712 if (ire->ire_addr != group)
13715 13713 continue;
13716 13714
13717 13715 ire_gw = ire_ftable_lookup_v4(ire->ire_gateway_addr,
13718 13716 0, 0, IRE_INTERFACE, ire->ire_ill, ALL_ZONES, NULL,
13719 13717 match_flags, 0, ipst, NULL);
13720 13718 }
13721 13719 /* No interface route exists for the gateway; skip this ire. */
13722 13720 if (ire_gw == NULL)
13723 13721 continue;
13724 13722 if (ire_gw->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
13725 13723 ire_refrele(ire_gw);
13726 13724 continue;
13727 13725 }
13728 13726 ASSERT(ire_gw->ire_ill != NULL); /* IRE_INTERFACE */
13729 13727 ifindex = ire_gw->ire_ill->ill_phyint->phyint_ifindex;
13730 13728
13731 13729 /*
13732 13730 * The operation is considered a success if
13733 13731 * it succeeds at least once on any one interface.
13734 13732 */
13735 13733 error = fn(connp, checkonly, v6group, INADDR_ANY, ifindex,
13736 13734 fmode, v6src);
13737 13735 if (error == 0)
13738 13736 result = CGTP_MCAST_SUCCESS;
13739 13737
13740 13738 ire_refrele(ire_gw);
13741 13739 }
13742 13740 irb_refrele(irb);
13743 13741 /*
13744 13742 * Consider the call as successful if we succeeded on at least
13745 13743 * one interface. Otherwise, return the last encountered error.
13746 13744 */
13747 13745 return (result == CGTP_MCAST_SUCCESS ? 0 : error);
13748 13746 }
13749 13747
13750 13748 /*
13751 13749 * Return the expected CGTP hooks version number.
13752 13750 */
13753 13751 int
13754 13752 ip_cgtp_filter_supported(void)
13755 13753 {
13756 13754 return (ip_cgtp_filter_rev);
13757 13755 }
13758 13756
13759 13757 /*
13760 13758 * CGTP hooks can be registered by invoking this function.
13761 13759 * Checks that the version number matches.
13762 13760 */
13763 13761 int
13764 13762 ip_cgtp_filter_register(netstackid_t stackid, cgtp_filter_ops_t *ops)
13765 13763 {
13766 13764 netstack_t *ns;
13767 13765 ip_stack_t *ipst;
13768 13766
13769 13767 if (ops->cfo_filter_rev != CGTP_FILTER_REV)
13770 13768 return (ENOTSUP);
13771 13769
13772 13770 ns = netstack_find_by_stackid(stackid);
13773 13771 if (ns == NULL)
13774 13772 return (EINVAL);
13775 13773 ipst = ns->netstack_ip;
13776 13774 ASSERT(ipst != NULL);
13777 13775
13778 13776 if (ipst->ips_ip_cgtp_filter_ops != NULL) {
13779 13777 netstack_rele(ns);
13780 13778 return (EALREADY);
13781 13779 }
13782 13780
13783 13781 ipst->ips_ip_cgtp_filter_ops = ops;
13784 13782
13785 13783 ill_set_inputfn_all(ipst);
13786 13784
13787 13785 netstack_rele(ns);
13788 13786 return (0);
13789 13787 }
13790 13788
13791 13789 /*
13792 13790 * CGTP hooks can be unregistered by invoking this function.
13793 13791 * Returns ENXIO if there was no registration.
13794 13792 * Returns EBUSY if the ndd variable has not been turned off.
13795 13793 */
13796 13794 int
13797 13795 ip_cgtp_filter_unregister(netstackid_t stackid)
13798 13796 {
13799 13797 netstack_t *ns;
13800 13798 ip_stack_t *ipst;
13801 13799
13802 13800 ns = netstack_find_by_stackid(stackid);
13803 13801 if (ns == NULL)
13804 13802 return (EINVAL);
13805 13803 ipst = ns->netstack_ip;
13806 13804 ASSERT(ipst != NULL);
13807 13805
13808 13806 if (ipst->ips_ip_cgtp_filter) {
13809 13807 netstack_rele(ns);
13810 13808 return (EBUSY);
13811 13809 }
13812 13810
13813 13811 if (ipst->ips_ip_cgtp_filter_ops == NULL) {
13814 13812 netstack_rele(ns);
13815 13813 return (ENXIO);
13816 13814 }
13817 13815 ipst->ips_ip_cgtp_filter_ops = NULL;
13818 13816
13819 13817 ill_set_inputfn_all(ipst);
13820 13818
13821 13819 netstack_rele(ns);
13822 13820 return (0);
13823 13821 }
13824 13822
13825 13823 /*
13826 13824 * Check whether there is a CGTP filter registration.
13827 13825 * Returns non-zero if there is a registration, otherwise returns zero.
13828 13826 * Note: returns zero if bad stackid.
13829 13827 */
13830 13828 int
13831 13829 ip_cgtp_filter_is_registered(netstackid_t stackid)
13832 13830 {
13833 13831 netstack_t *ns;
13834 13832 ip_stack_t *ipst;
13835 13833 int ret;
13836 13834
13837 13835 ns = netstack_find_by_stackid(stackid);
13838 13836 if (ns == NULL)
13839 13837 return (0);
13840 13838 ipst = ns->netstack_ip;
13841 13839 ASSERT(ipst != NULL);
13842 13840
13843 13841 if (ipst->ips_ip_cgtp_filter_ops != NULL)
13844 13842 ret = 1;
13845 13843 else
13846 13844 ret = 0;
13847 13845
13848 13846 netstack_rele(ns);
13849 13847 return (ret);
13850 13848 }
13851 13849
13852 13850 static int
13853 13851 ip_squeue_switch(int val)
13854 13852 {
13855 13853 int rval;
13856 13854
13857 13855 switch (val) {
13858 13856 case IP_SQUEUE_ENTER_NODRAIN:
13859 13857 rval = SQ_NODRAIN;
13860 13858 break;
13861 13859 case IP_SQUEUE_ENTER:
13862 13860 rval = SQ_PROCESS;
13863 13861 break;
13864 13862 case IP_SQUEUE_FILL:
13865 13863 default:
13866 13864 rval = SQ_FILL;
13867 13865 break;
13868 13866 }
13869 13867 return (rval);
13870 13868 }
13871 13869
13872 13870 static void *
13873 13871 ip_kstat2_init(netstackid_t stackid, ip_stat_t *ip_statisticsp)
13874 13872 {
13875 13873 kstat_t *ksp;
13876 13874
13877 13875 ip_stat_t template = {
13878 13876 { "ip_udp_fannorm", KSTAT_DATA_UINT64 },
13879 13877 { "ip_udp_fanmb", KSTAT_DATA_UINT64 },
13880 13878 { "ip_recv_pullup", KSTAT_DATA_UINT64 },
13881 13879 { "ip_db_ref", KSTAT_DATA_UINT64 },
13882 13880 { "ip_notaligned", KSTAT_DATA_UINT64 },
13883 13881 { "ip_multimblk", KSTAT_DATA_UINT64 },
13884 13882 { "ip_opt", KSTAT_DATA_UINT64 },
13885 13883 { "ipsec_proto_ahesp", KSTAT_DATA_UINT64 },
13886 13884 { "ip_conn_flputbq", KSTAT_DATA_UINT64 },
13887 13885 { "ip_conn_walk_drain", KSTAT_DATA_UINT64 },
13888 13886 { "ip_out_sw_cksum", KSTAT_DATA_UINT64 },
13889 13887 { "ip_out_sw_cksum_bytes", KSTAT_DATA_UINT64 },
13890 13888 { "ip_in_sw_cksum", KSTAT_DATA_UINT64 },
13891 13889 { "ip_ire_reclaim_calls", KSTAT_DATA_UINT64 },
13892 13890 { "ip_ire_reclaim_deleted", KSTAT_DATA_UINT64 },
13893 13891 { "ip_nce_reclaim_calls", KSTAT_DATA_UINT64 },
13894 13892 { "ip_nce_reclaim_deleted", KSTAT_DATA_UINT64 },
13895 13893 { "ip_nce_mcast_reclaim_calls", KSTAT_DATA_UINT64 },
13896 13894 { "ip_nce_mcast_reclaim_deleted", KSTAT_DATA_UINT64 },
13897 13895 { "ip_nce_mcast_reclaim_tqfail", KSTAT_DATA_UINT64 },
13898 13896 { "ip_dce_reclaim_calls", KSTAT_DATA_UINT64 },
13899 13897 { "ip_dce_reclaim_deleted", KSTAT_DATA_UINT64 },
13900 13898 { "ip_tcp_in_full_hw_cksum_err", KSTAT_DATA_UINT64 },
13901 13899 { "ip_tcp_in_part_hw_cksum_err", KSTAT_DATA_UINT64 },
13902 13900 { "ip_tcp_in_sw_cksum_err", KSTAT_DATA_UINT64 },
13903 13901 { "ip_udp_in_full_hw_cksum_err", KSTAT_DATA_UINT64 },
13904 13902 { "ip_udp_in_part_hw_cksum_err", KSTAT_DATA_UINT64 },
13905 13903 { "ip_udp_in_sw_cksum_err", KSTAT_DATA_UINT64 },
13906 13904 { "conn_in_recvdstaddr", KSTAT_DATA_UINT64 },
13907 13905 { "conn_in_recvopts", KSTAT_DATA_UINT64 },
13908 13906 { "conn_in_recvif", KSTAT_DATA_UINT64 },
13909 13907 { "conn_in_recvslla", KSTAT_DATA_UINT64 },
13910 13908 { "conn_in_recvucred", KSTAT_DATA_UINT64 },
13911 13909 { "conn_in_recvttl", KSTAT_DATA_UINT64 },
13912 13910 { "conn_in_recvhopopts", KSTAT_DATA_UINT64 },
13913 13911 { "conn_in_recvhoplimit", KSTAT_DATA_UINT64 },
13914 13912 { "conn_in_recvdstopts", KSTAT_DATA_UINT64 },
13915 13913 { "conn_in_recvrthdrdstopts", KSTAT_DATA_UINT64 },
13916 13914 { "conn_in_recvrthdr", KSTAT_DATA_UINT64 },
13917 13915 { "conn_in_recvpktinfo", KSTAT_DATA_UINT64 },
13918 13916 { "conn_in_recvtclass", KSTAT_DATA_UINT64 },
13919 13917 { "conn_in_timestamp", KSTAT_DATA_UINT64 },
13920 13918 };
13921 13919
13922 13920 ksp = kstat_create_netstack("ip", 0, "ipstat", "net",
13923 13921 KSTAT_TYPE_NAMED, sizeof (template) / sizeof (kstat_named_t),
13924 13922 KSTAT_FLAG_VIRTUAL, stackid);
13925 13923
13926 13924 if (ksp == NULL)
13927 13925 return (NULL);
13928 13926
13929 13927 bcopy(&template, ip_statisticsp, sizeof (template));
13930 13928 ksp->ks_data = (void *)ip_statisticsp;
13931 13929 ksp->ks_private = (void *)(uintptr_t)stackid;
13932 13930
13933 13931 kstat_install(ksp);
13934 13932 return (ksp);
13935 13933 }
13936 13934
13937 13935 static void
13938 13936 ip_kstat2_fini(netstackid_t stackid, kstat_t *ksp)
13939 13937 {
13940 13938 if (ksp != NULL) {
13941 13939 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
13942 13940 kstat_delete_netstack(ksp, stackid);
13943 13941 }
13944 13942 }
13945 13943
13946 13944 static void *
13947 13945 ip_kstat_init(netstackid_t stackid, ip_stack_t *ipst)
13948 13946 {
13949 13947 kstat_t *ksp;
13950 13948
13951 13949 ip_named_kstat_t template = {
13952 13950 { "forwarding", KSTAT_DATA_UINT32, 0 },
13953 13951 { "defaultTTL", KSTAT_DATA_UINT32, 0 },
13954 13952 { "inReceives", KSTAT_DATA_UINT64, 0 },
13955 13953 { "inHdrErrors", KSTAT_DATA_UINT32, 0 },
13956 13954 { "inAddrErrors", KSTAT_DATA_UINT32, 0 },
13957 13955 { "forwDatagrams", KSTAT_DATA_UINT64, 0 },
13958 13956 { "inUnknownProtos", KSTAT_DATA_UINT32, 0 },
13959 13957 { "inDiscards", KSTAT_DATA_UINT32, 0 },
13960 13958 { "inDelivers", KSTAT_DATA_UINT64, 0 },
13961 13959 { "outRequests", KSTAT_DATA_UINT64, 0 },
13962 13960 { "outDiscards", KSTAT_DATA_UINT32, 0 },
13963 13961 { "outNoRoutes", KSTAT_DATA_UINT32, 0 },
13964 13962 { "reasmTimeout", KSTAT_DATA_UINT32, 0 },
13965 13963 { "reasmReqds", KSTAT_DATA_UINT32, 0 },
13966 13964 { "reasmOKs", KSTAT_DATA_UINT32, 0 },
13967 13965 { "reasmFails", KSTAT_DATA_UINT32, 0 },
13968 13966 { "fragOKs", KSTAT_DATA_UINT32, 0 },
13969 13967 { "fragFails", KSTAT_DATA_UINT32, 0 },
13970 13968 { "fragCreates", KSTAT_DATA_UINT32, 0 },
13971 13969 { "addrEntrySize", KSTAT_DATA_INT32, 0 },
13972 13970 { "routeEntrySize", KSTAT_DATA_INT32, 0 },
13973 13971 { "netToMediaEntrySize", KSTAT_DATA_INT32, 0 },
13974 13972 { "routingDiscards", KSTAT_DATA_UINT32, 0 },
13975 13973 { "inErrs", KSTAT_DATA_UINT32, 0 },
13976 13974 { "noPorts", KSTAT_DATA_UINT32, 0 },
13977 13975 { "inCksumErrs", KSTAT_DATA_UINT32, 0 },
13978 13976 { "reasmDuplicates", KSTAT_DATA_UINT32, 0 },
13979 13977 { "reasmPartDups", KSTAT_DATA_UINT32, 0 },
13980 13978 { "forwProhibits", KSTAT_DATA_UINT32, 0 },
13981 13979 { "udpInCksumErrs", KSTAT_DATA_UINT32, 0 },
13982 13980 { "udpInOverflows", KSTAT_DATA_UINT32, 0 },
13983 13981 { "rawipInOverflows", KSTAT_DATA_UINT32, 0 },
13984 13982 { "ipsecInSucceeded", KSTAT_DATA_UINT32, 0 },
13985 13983 { "ipsecInFailed", KSTAT_DATA_INT32, 0 },
13986 13984 { "memberEntrySize", KSTAT_DATA_INT32, 0 },
13987 13985 { "inIPv6", KSTAT_DATA_UINT32, 0 },
13988 13986 { "outIPv6", KSTAT_DATA_UINT32, 0 },
13989 13987 { "outSwitchIPv6", KSTAT_DATA_UINT32, 0 },
13990 13988 };
13991 13989
13992 13990 ksp = kstat_create_netstack("ip", 0, "ip", "mib2", KSTAT_TYPE_NAMED,
13993 13991 NUM_OF_FIELDS(ip_named_kstat_t), 0, stackid);
13994 13992 if (ksp == NULL || ksp->ks_data == NULL)
13995 13993 return (NULL);
13996 13994
13997 13995 template.forwarding.value.ui32 = WE_ARE_FORWARDING(ipst) ? 1:2;
13998 13996 template.defaultTTL.value.ui32 = (uint32_t)ipst->ips_ip_def_ttl;
13999 13997 template.reasmTimeout.value.ui32 = ipst->ips_ip_reassembly_timeout;
14000 13998 template.addrEntrySize.value.i32 = sizeof (mib2_ipAddrEntry_t);
14001 13999 template.routeEntrySize.value.i32 = sizeof (mib2_ipRouteEntry_t);
14002 14000
14003 14001 template.netToMediaEntrySize.value.i32 =
14004 14002 sizeof (mib2_ipNetToMediaEntry_t);
14005 14003
14006 14004 template.memberEntrySize.value.i32 = sizeof (ipv6_member_t);
14007 14005
14008 14006 bcopy(&template, ksp->ks_data, sizeof (template));
14009 14007 ksp->ks_update = ip_kstat_update;
14010 14008 ksp->ks_private = (void *)(uintptr_t)stackid;
14011 14009
14012 14010 kstat_install(ksp);
14013 14011 return (ksp);
14014 14012 }
14015 14013
14016 14014 static void
14017 14015 ip_kstat_fini(netstackid_t stackid, kstat_t *ksp)
14018 14016 {
14019 14017 if (ksp != NULL) {
14020 14018 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
14021 14019 kstat_delete_netstack(ksp, stackid);
14022 14020 }
14023 14021 }
14024 14022
14025 14023 static int
14026 14024 ip_kstat_update(kstat_t *kp, int rw)
14027 14025 {
14028 14026 ip_named_kstat_t *ipkp;
14029 14027 mib2_ipIfStatsEntry_t ipmib;
14030 14028 ill_walk_context_t ctx;
14031 14029 ill_t *ill;
14032 14030 netstackid_t stackid = (zoneid_t)(uintptr_t)kp->ks_private;
14033 14031 netstack_t *ns;
14034 14032 ip_stack_t *ipst;
14035 14033
14036 14034 if (kp->ks_data == NULL)
14037 14035 return (EIO);
14038 14036
14039 14037 if (rw == KSTAT_WRITE)
14040 14038 return (EACCES);
14041 14039
14042 14040 ns = netstack_find_by_stackid(stackid);
14043 14041 if (ns == NULL)
14044 14042 return (-1);
14045 14043 ipst = ns->netstack_ip;
14046 14044 if (ipst == NULL) {
14047 14045 netstack_rele(ns);
14048 14046 return (-1);
14049 14047 }
14050 14048 ipkp = (ip_named_kstat_t *)kp->ks_data;
14051 14049
14052 14050 bcopy(&ipst->ips_ip_mib, &ipmib, sizeof (ipmib));
14053 14051 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
14054 14052 ill = ILL_START_WALK_V4(&ctx, ipst);
14055 14053 for (; ill != NULL; ill = ill_next(&ctx, ill))
14056 14054 ip_mib2_add_ip_stats(&ipmib, ill->ill_ip_mib);
14057 14055 rw_exit(&ipst->ips_ill_g_lock);
14058 14056
14059 14057 ipkp->forwarding.value.ui32 = ipmib.ipIfStatsForwarding;
14060 14058 ipkp->defaultTTL.value.ui32 = ipmib.ipIfStatsDefaultTTL;
14061 14059 ipkp->inReceives.value.ui64 = ipmib.ipIfStatsHCInReceives;
14062 14060 ipkp->inHdrErrors.value.ui32 = ipmib.ipIfStatsInHdrErrors;
14063 14061 ipkp->inAddrErrors.value.ui32 = ipmib.ipIfStatsInAddrErrors;
14064 14062 ipkp->forwDatagrams.value.ui64 = ipmib.ipIfStatsHCOutForwDatagrams;
14065 14063 ipkp->inUnknownProtos.value.ui32 = ipmib.ipIfStatsInUnknownProtos;
14066 14064 ipkp->inDiscards.value.ui32 = ipmib.ipIfStatsInDiscards;
14067 14065 ipkp->inDelivers.value.ui64 = ipmib.ipIfStatsHCInDelivers;
14068 14066 ipkp->outRequests.value.ui64 = ipmib.ipIfStatsHCOutRequests;
14069 14067 ipkp->outDiscards.value.ui32 = ipmib.ipIfStatsOutDiscards;
14070 14068 ipkp->outNoRoutes.value.ui32 = ipmib.ipIfStatsOutNoRoutes;
14071 14069 ipkp->reasmTimeout.value.ui32 = ipst->ips_ip_reassembly_timeout;
14072 14070 ipkp->reasmReqds.value.ui32 = ipmib.ipIfStatsReasmReqds;
14073 14071 ipkp->reasmOKs.value.ui32 = ipmib.ipIfStatsReasmOKs;
14074 14072 ipkp->reasmFails.value.ui32 = ipmib.ipIfStatsReasmFails;
14075 14073 ipkp->fragOKs.value.ui32 = ipmib.ipIfStatsOutFragOKs;
14076 14074 ipkp->fragFails.value.ui32 = ipmib.ipIfStatsOutFragFails;
14077 14075 ipkp->fragCreates.value.ui32 = ipmib.ipIfStatsOutFragCreates;
14078 14076
14079 14077 ipkp->routingDiscards.value.ui32 = 0;
14080 14078 ipkp->inErrs.value.ui32 = ipmib.tcpIfStatsInErrs;
14081 14079 ipkp->noPorts.value.ui32 = ipmib.udpIfStatsNoPorts;
14082 14080 ipkp->inCksumErrs.value.ui32 = ipmib.ipIfStatsInCksumErrs;
14083 14081 ipkp->reasmDuplicates.value.ui32 = ipmib.ipIfStatsReasmDuplicates;
14084 14082 ipkp->reasmPartDups.value.ui32 = ipmib.ipIfStatsReasmPartDups;
14085 14083 ipkp->forwProhibits.value.ui32 = ipmib.ipIfStatsForwProhibits;
14086 14084 ipkp->udpInCksumErrs.value.ui32 = ipmib.udpIfStatsInCksumErrs;
14087 14085 ipkp->udpInOverflows.value.ui32 = ipmib.udpIfStatsInOverflows;
14088 14086 ipkp->rawipInOverflows.value.ui32 = ipmib.rawipIfStatsInOverflows;
14089 14087 ipkp->ipsecInSucceeded.value.ui32 = ipmib.ipsecIfStatsInSucceeded;
14090 14088 ipkp->ipsecInFailed.value.i32 = ipmib.ipsecIfStatsInFailed;
14091 14089
14092 14090 ipkp->inIPv6.value.ui32 = ipmib.ipIfStatsInWrongIPVersion;
14093 14091 ipkp->outIPv6.value.ui32 = ipmib.ipIfStatsOutWrongIPVersion;
14094 14092 ipkp->outSwitchIPv6.value.ui32 = ipmib.ipIfStatsOutSwitchIPVersion;
14095 14093
14096 14094 netstack_rele(ns);
14097 14095
14098 14096 return (0);
14099 14097 }
14100 14098
14101 14099 static void *
14102 14100 icmp_kstat_init(netstackid_t stackid)
14103 14101 {
14104 14102 kstat_t *ksp;
14105 14103
14106 14104 icmp_named_kstat_t template = {
14107 14105 { "inMsgs", KSTAT_DATA_UINT32 },
14108 14106 { "inErrors", KSTAT_DATA_UINT32 },
14109 14107 { "inDestUnreachs", KSTAT_DATA_UINT32 },
14110 14108 { "inTimeExcds", KSTAT_DATA_UINT32 },
14111 14109 { "inParmProbs", KSTAT_DATA_UINT32 },
14112 14110 { "inSrcQuenchs", KSTAT_DATA_UINT32 },
14113 14111 { "inRedirects", KSTAT_DATA_UINT32 },
14114 14112 { "inEchos", KSTAT_DATA_UINT32 },
14115 14113 { "inEchoReps", KSTAT_DATA_UINT32 },
14116 14114 { "inTimestamps", KSTAT_DATA_UINT32 },
14117 14115 { "inTimestampReps", KSTAT_DATA_UINT32 },
14118 14116 { "inAddrMasks", KSTAT_DATA_UINT32 },
14119 14117 { "inAddrMaskReps", KSTAT_DATA_UINT32 },
14120 14118 { "outMsgs", KSTAT_DATA_UINT32 },
14121 14119 { "outErrors", KSTAT_DATA_UINT32 },
14122 14120 { "outDestUnreachs", KSTAT_DATA_UINT32 },
14123 14121 { "outTimeExcds", KSTAT_DATA_UINT32 },
14124 14122 { "outParmProbs", KSTAT_DATA_UINT32 },
14125 14123 { "outSrcQuenchs", KSTAT_DATA_UINT32 },
14126 14124 { "outRedirects", KSTAT_DATA_UINT32 },
14127 14125 { "outEchos", KSTAT_DATA_UINT32 },
14128 14126 { "outEchoReps", KSTAT_DATA_UINT32 },
14129 14127 { "outTimestamps", KSTAT_DATA_UINT32 },
14130 14128 { "outTimestampReps", KSTAT_DATA_UINT32 },
14131 14129 { "outAddrMasks", KSTAT_DATA_UINT32 },
14132 14130 { "outAddrMaskReps", KSTAT_DATA_UINT32 },
14133 14131 { "inChksumErrs", KSTAT_DATA_UINT32 },
14134 14132 { "inUnknowns", KSTAT_DATA_UINT32 },
14135 14133 { "inFragNeeded", KSTAT_DATA_UINT32 },
14136 14134 { "outFragNeeded", KSTAT_DATA_UINT32 },
14137 14135 { "outDrops", KSTAT_DATA_UINT32 },
14138 14136 { "inOverFlows", KSTAT_DATA_UINT32 },
14139 14137 { "inBadRedirects", KSTAT_DATA_UINT32 },
14140 14138 };
14141 14139
14142 14140 ksp = kstat_create_netstack("ip", 0, "icmp", "mib2", KSTAT_TYPE_NAMED,
14143 14141 NUM_OF_FIELDS(icmp_named_kstat_t), 0, stackid);
14144 14142 if (ksp == NULL || ksp->ks_data == NULL)
14145 14143 return (NULL);
14146 14144
14147 14145 bcopy(&template, ksp->ks_data, sizeof (template));
14148 14146
14149 14147 ksp->ks_update = icmp_kstat_update;
14150 14148 ksp->ks_private = (void *)(uintptr_t)stackid;
14151 14149
14152 14150 kstat_install(ksp);
14153 14151 return (ksp);
14154 14152 }
14155 14153
14156 14154 static void
14157 14155 icmp_kstat_fini(netstackid_t stackid, kstat_t *ksp)
14158 14156 {
14159 14157 if (ksp != NULL) {
14160 14158 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
14161 14159 kstat_delete_netstack(ksp, stackid);
14162 14160 }
14163 14161 }
14164 14162
14165 14163 static int
14166 14164 icmp_kstat_update(kstat_t *kp, int rw)
14167 14165 {
14168 14166 icmp_named_kstat_t *icmpkp;
14169 14167 netstackid_t stackid = (zoneid_t)(uintptr_t)kp->ks_private;
14170 14168 netstack_t *ns;
14171 14169 ip_stack_t *ipst;
14172 14170
14173 14171 if (kp->ks_data == NULL)
14174 14172 return (EIO);
14175 14173
14176 14174 if (rw == KSTAT_WRITE)
14177 14175 return (EACCES);
14178 14176
14179 14177 ns = netstack_find_by_stackid(stackid);
14180 14178 if (ns == NULL)
14181 14179 return (-1);
14182 14180 ipst = ns->netstack_ip;
14183 14181 if (ipst == NULL) {
14184 14182 netstack_rele(ns);
14185 14183 return (-1);
14186 14184 }
14187 14185 icmpkp = (icmp_named_kstat_t *)kp->ks_data;
14188 14186
14189 14187 icmpkp->inMsgs.value.ui32 = ipst->ips_icmp_mib.icmpInMsgs;
14190 14188 icmpkp->inErrors.value.ui32 = ipst->ips_icmp_mib.icmpInErrors;
14191 14189 icmpkp->inDestUnreachs.value.ui32 =
14192 14190 ipst->ips_icmp_mib.icmpInDestUnreachs;
14193 14191 icmpkp->inTimeExcds.value.ui32 = ipst->ips_icmp_mib.icmpInTimeExcds;
14194 14192 icmpkp->inParmProbs.value.ui32 = ipst->ips_icmp_mib.icmpInParmProbs;
14195 14193 icmpkp->inSrcQuenchs.value.ui32 = ipst->ips_icmp_mib.icmpInSrcQuenchs;
14196 14194 icmpkp->inRedirects.value.ui32 = ipst->ips_icmp_mib.icmpInRedirects;
14197 14195 icmpkp->inEchos.value.ui32 = ipst->ips_icmp_mib.icmpInEchos;
14198 14196 icmpkp->inEchoReps.value.ui32 = ipst->ips_icmp_mib.icmpInEchoReps;
14199 14197 icmpkp->inTimestamps.value.ui32 = ipst->ips_icmp_mib.icmpInTimestamps;
14200 14198 icmpkp->inTimestampReps.value.ui32 =
14201 14199 ipst->ips_icmp_mib.icmpInTimestampReps;
14202 14200 icmpkp->inAddrMasks.value.ui32 = ipst->ips_icmp_mib.icmpInAddrMasks;
14203 14201 icmpkp->inAddrMaskReps.value.ui32 =
14204 14202 ipst->ips_icmp_mib.icmpInAddrMaskReps;
14205 14203 icmpkp->outMsgs.value.ui32 = ipst->ips_icmp_mib.icmpOutMsgs;
14206 14204 icmpkp->outErrors.value.ui32 = ipst->ips_icmp_mib.icmpOutErrors;
14207 14205 icmpkp->outDestUnreachs.value.ui32 =
14208 14206 ipst->ips_icmp_mib.icmpOutDestUnreachs;
14209 14207 icmpkp->outTimeExcds.value.ui32 = ipst->ips_icmp_mib.icmpOutTimeExcds;
14210 14208 icmpkp->outParmProbs.value.ui32 = ipst->ips_icmp_mib.icmpOutParmProbs;
14211 14209 icmpkp->outSrcQuenchs.value.ui32 =
14212 14210 ipst->ips_icmp_mib.icmpOutSrcQuenchs;
14213 14211 icmpkp->outRedirects.value.ui32 = ipst->ips_icmp_mib.icmpOutRedirects;
14214 14212 icmpkp->outEchos.value.ui32 = ipst->ips_icmp_mib.icmpOutEchos;
14215 14213 icmpkp->outEchoReps.value.ui32 = ipst->ips_icmp_mib.icmpOutEchoReps;
14216 14214 icmpkp->outTimestamps.value.ui32 =
14217 14215 ipst->ips_icmp_mib.icmpOutTimestamps;
14218 14216 icmpkp->outTimestampReps.value.ui32 =
14219 14217 ipst->ips_icmp_mib.icmpOutTimestampReps;
14220 14218 icmpkp->outAddrMasks.value.ui32 =
14221 14219 ipst->ips_icmp_mib.icmpOutAddrMasks;
14222 14220 icmpkp->outAddrMaskReps.value.ui32 =
14223 14221 ipst->ips_icmp_mib.icmpOutAddrMaskReps;
14224 14222 icmpkp->inCksumErrs.value.ui32 = ipst->ips_icmp_mib.icmpInCksumErrs;
14225 14223 icmpkp->inUnknowns.value.ui32 = ipst->ips_icmp_mib.icmpInUnknowns;
14226 14224 icmpkp->inFragNeeded.value.ui32 = ipst->ips_icmp_mib.icmpInFragNeeded;
14227 14225 icmpkp->outFragNeeded.value.ui32 =
14228 14226 ipst->ips_icmp_mib.icmpOutFragNeeded;
14229 14227 icmpkp->outDrops.value.ui32 = ipst->ips_icmp_mib.icmpOutDrops;
14230 14228 icmpkp->inOverflows.value.ui32 = ipst->ips_icmp_mib.icmpInOverflows;
14231 14229 icmpkp->inBadRedirects.value.ui32 =
14232 14230 ipst->ips_icmp_mib.icmpInBadRedirects;
14233 14231
14234 14232 netstack_rele(ns);
14235 14233 return (0);
14236 14234 }
14237 14235
14238 14236 /*
14239 14237 * This is the fanout function for raw socket opened for SCTP. Note
14240 14238 * that it is called after SCTP checks that there is no socket which
14241 14239 * wants a packet. Then before SCTP handles this out of the blue packet,
14242 14240 * this function is called to see if there is any raw socket for SCTP.
14243 14241 * If there is and it is bound to the correct address, the packet will
14244 14242 * be sent to that socket. Note that only one raw socket can be bound to
14245 14243 * a port. This is assured in ipcl_sctp_hash_insert();
14246 14244 */
14247 14245 void
14248 14246 ip_fanout_sctp_raw(mblk_t *mp, ipha_t *ipha, ip6_t *ip6h, uint32_t ports,
14249 14247 ip_recv_attr_t *ira)
14250 14248 {
14251 14249 conn_t *connp;
14252 14250 queue_t *rq;
14253 14251 boolean_t secure;
14254 14252 ill_t *ill = ira->ira_ill;
14255 14253 ip_stack_t *ipst = ill->ill_ipst;
14256 14254 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
14257 14255 sctp_stack_t *sctps = ipst->ips_netstack->netstack_sctp;
14258 14256 iaflags_t iraflags = ira->ira_flags;
14259 14257 ill_t *rill = ira->ira_rill;
14260 14258
14261 14259 secure = iraflags & IRAF_IPSEC_SECURE;
14262 14260
14263 14261 connp = ipcl_classify_raw(mp, IPPROTO_SCTP, ports, ipha, ip6h,
14264 14262 ira, ipst);
14265 14263 if (connp == NULL) {
14266 14264 /*
14267 14265 * Although raw sctp is not summed, OOB chunks must be.
14268 14266 * Drop the packet here if the sctp checksum failed.
14269 14267 */
14270 14268 if (iraflags & IRAF_SCTP_CSUM_ERR) {
14271 14269 SCTPS_BUMP_MIB(sctps, sctpChecksumError);
14272 14270 freemsg(mp);
14273 14271 return;
14274 14272 }
14275 14273 ira->ira_ill = ira->ira_rill = NULL;
14276 14274 sctp_ootb_input(mp, ira, ipst);
14277 14275 ira->ira_ill = ill;
14278 14276 ira->ira_rill = rill;
14279 14277 return;
14280 14278 }
14281 14279 rq = connp->conn_rq;
14282 14280 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld : !canputnext(rq)) {
14283 14281 CONN_DEC_REF(connp);
14284 14282 BUMP_MIB(ill->ill_ip_mib, rawipIfStatsInOverflows);
14285 14283 freemsg(mp);
14286 14284 return;
14287 14285 }
14288 14286 if (((iraflags & IRAF_IS_IPV4) ?
14289 14287 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
14290 14288 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
14291 14289 secure) {
14292 14290 mp = ipsec_check_inbound_policy(mp, connp, ipha,
14293 14291 ip6h, ira);
14294 14292 if (mp == NULL) {
14295 14293 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
14296 14294 /* Note that mp is NULL */
14297 14295 ip_drop_input("ipIfStatsInDiscards", mp, ill);
14298 14296 CONN_DEC_REF(connp);
14299 14297 return;
14300 14298 }
14301 14299 }
14302 14300
14303 14301 if (iraflags & IRAF_ICMP_ERROR) {
14304 14302 (connp->conn_recvicmp)(connp, mp, NULL, ira);
14305 14303 } else {
14306 14304 ill_t *rill = ira->ira_rill;
14307 14305
14308 14306 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
14309 14307 /* This is the SOCK_RAW, IPPROTO_SCTP case. */
14310 14308 ira->ira_ill = ira->ira_rill = NULL;
14311 14309 (connp->conn_recv)(connp, mp, NULL, ira);
14312 14310 ira->ira_ill = ill;
14313 14311 ira->ira_rill = rill;
14314 14312 }
14315 14313 CONN_DEC_REF(connp);
14316 14314 }
14317 14315
14318 14316 /*
14319 14317 * Free a packet that has the link-layer dl_unitdata_req_t or fast-path
14320 14318 * header before the ip payload.
14321 14319 */
14322 14320 static void
14323 14321 ip_xmit_flowctl_drop(ill_t *ill, mblk_t *mp, boolean_t is_fp_mp, int fp_mp_len)
14324 14322 {
14325 14323 int len = (mp->b_wptr - mp->b_rptr);
14326 14324 mblk_t *ip_mp;
14327 14325
14328 14326 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14329 14327 if (is_fp_mp || len != fp_mp_len) {
14330 14328 if (len > fp_mp_len) {
14331 14329 /*
14332 14330 * fastpath header and ip header in the first mblk
14333 14331 */
14334 14332 mp->b_rptr += fp_mp_len;
14335 14333 } else {
14336 14334 /*
14337 14335 * ip_xmit_attach_llhdr had to prepend an mblk to
14338 14336 * attach the fastpath header before ip header.
14339 14337 */
14340 14338 ip_mp = mp->b_cont;
14341 14339 freeb(mp);
14342 14340 mp = ip_mp;
14343 14341 mp->b_rptr += (fp_mp_len - len);
14344 14342 }
14345 14343 } else {
14346 14344 ip_mp = mp->b_cont;
14347 14345 freeb(mp);
14348 14346 mp = ip_mp;
14349 14347 }
14350 14348 ip_drop_output("ipIfStatsOutDiscards - flow ctl", mp, ill);
14351 14349 freemsg(mp);
14352 14350 }
14353 14351
14354 14352 /*
14355 14353 * Normal post fragmentation function.
14356 14354 *
14357 14355 * Send a packet using the passed in nce. This handles both IPv4 and IPv6
14358 14356 * using the same state machine.
14359 14357 *
14360 14358 * We return an error on failure. In particular we return EWOULDBLOCK
14361 14359 * when the driver flow controls. In that case this ensures that ip_wsrv runs
14362 14360 * (currently by canputnext failure resulting in backenabling from GLD.)
14363 14361 * This allows the callers of conn_ip_output() to use EWOULDBLOCK as an
14364 14362 * indication that they can flow control until ip_wsrv() tells then to restart.
14365 14363 *
14366 14364 * If the nce passed by caller is incomplete, this function
14367 14365 * queues the packet and if necessary, sends ARP request and bails.
14368 14366 * If the Neighbor Cache passed is fully resolved, we simply prepend
14369 14367 * the link-layer header to the packet, do ipsec hw acceleration
14370 14368 * work if necessary, and send the packet out on the wire.
14371 14369 */
14372 14370 /* ARGSUSED6 */
14373 14371 int
14374 14372 ip_xmit(mblk_t *mp, nce_t *nce, iaflags_t ixaflags, uint_t pkt_len,
14375 14373 uint32_t xmit_hint, zoneid_t szone, zoneid_t nolzid, uintptr_t *ixacookie)
14376 14374 {
14377 14375 queue_t *wq;
14378 14376 ill_t *ill = nce->nce_ill;
14379 14377 ip_stack_t *ipst = ill->ill_ipst;
14380 14378 uint64_t delta;
14381 14379 boolean_t isv6 = ill->ill_isv6;
14382 14380 boolean_t fp_mp;
14383 14381 ncec_t *ncec = nce->nce_common;
14384 14382 int64_t now = LBOLT_FASTPATH64;
14385 14383 boolean_t is_probe;
14386 14384
14387 14385 DTRACE_PROBE1(ip__xmit, nce_t *, nce);
14388 14386
14389 14387 ASSERT(mp != NULL);
14390 14388 ASSERT(mp->b_datap->db_type == M_DATA);
14391 14389 ASSERT(pkt_len == msgdsize(mp));
14392 14390
14393 14391 /*
14394 14392 * If we have already been here and are coming back after ARP/ND.
14395 14393 * the IXAF_NO_TRACE flag is set. We skip FW_HOOKS, DTRACE and ipobs
14396 14394 * in that case since they have seen the packet when it came here
14397 14395 * the first time.
14398 14396 */
14399 14397 if (ixaflags & IXAF_NO_TRACE)
14400 14398 goto sendit;
14401 14399
14402 14400 if (ixaflags & IXAF_IS_IPV4) {
14403 14401 ipha_t *ipha = (ipha_t *)mp->b_rptr;
14404 14402
14405 14403 ASSERT(!isv6);
14406 14404 ASSERT(pkt_len == ntohs(((ipha_t *)mp->b_rptr)->ipha_length));
14407 14405 if (HOOKS4_INTERESTED_PHYSICAL_OUT(ipst) &&
14408 14406 !(ixaflags & IXAF_NO_PFHOOK)) {
14409 14407 int error;
14410 14408
14411 14409 FW_HOOKS(ipst->ips_ip4_physical_out_event,
14412 14410 ipst->ips_ipv4firewall_physical_out,
14413 14411 NULL, ill, ipha, mp, mp, 0, ipst, error);
14414 14412 DTRACE_PROBE1(ip4__physical__out__end,
14415 14413 mblk_t *, mp);
14416 14414 if (mp == NULL)
14417 14415 return (error);
14418 14416
14419 14417 /* The length could have changed */
14420 14418 pkt_len = msgdsize(mp);
14421 14419 }
14422 14420 if (ipst->ips_ip4_observe.he_interested) {
14423 14421 /*
14424 14422 * Note that for TX the zoneid is the sending
14425 14423 * zone, whether or not MLP is in play.
14426 14424 * Since the szone argument is the IP zoneid (i.e.,
14427 14425 * zero for exclusive-IP zones) and ipobs wants
14428 14426 * the system zoneid, we map it here.
14429 14427 */
14430 14428 szone = IP_REAL_ZONEID(szone, ipst);
14431 14429
14432 14430 /*
14433 14431 * On the outbound path the destination zone will be
14434 14432 * unknown as we're sending this packet out on the
14435 14433 * wire.
14436 14434 */
14437 14435 ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone, ALL_ZONES,
14438 14436 ill, ipst);
14439 14437 }
14440 14438 DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
14441 14439 void_ip_t *, ipha, __dtrace_ipsr_ill_t *, ill,
14442 14440 ipha_t *, ipha, ip6_t *, NULL, int, 0);
14443 14441 } else {
14444 14442 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
14445 14443
14446 14444 ASSERT(isv6);
14447 14445 ASSERT(pkt_len ==
14448 14446 ntohs(((ip6_t *)mp->b_rptr)->ip6_plen) + IPV6_HDR_LEN);
14449 14447 if (HOOKS6_INTERESTED_PHYSICAL_OUT(ipst) &&
14450 14448 !(ixaflags & IXAF_NO_PFHOOK)) {
14451 14449 int error;
14452 14450
14453 14451 FW_HOOKS6(ipst->ips_ip6_physical_out_event,
14454 14452 ipst->ips_ipv6firewall_physical_out,
14455 14453 NULL, ill, ip6h, mp, mp, 0, ipst, error);
14456 14454 DTRACE_PROBE1(ip6__physical__out__end,
14457 14455 mblk_t *, mp);
14458 14456 if (mp == NULL)
14459 14457 return (error);
14460 14458
14461 14459 /* The length could have changed */
14462 14460 pkt_len = msgdsize(mp);
14463 14461 }
14464 14462 if (ipst->ips_ip6_observe.he_interested) {
14465 14463 /* See above */
14466 14464 szone = IP_REAL_ZONEID(szone, ipst);
14467 14465
14468 14466 ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone, ALL_ZONES,
14469 14467 ill, ipst);
14470 14468 }
14471 14469 DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
14472 14470 void_ip_t *, ip6h, __dtrace_ipsr_ill_t *, ill,
14473 14471 ipha_t *, NULL, ip6_t *, ip6h, int, 0);
14474 14472 }
14475 14473
14476 14474 sendit:
14477 14475 /*
14478 14476 * We check the state without a lock because the state can never
14479 14477 * move "backwards" to initial or incomplete.
14480 14478 */
14481 14479 switch (ncec->ncec_state) {
14482 14480 case ND_REACHABLE:
14483 14481 case ND_STALE:
14484 14482 case ND_DELAY:
14485 14483 case ND_PROBE:
14486 14484 mp = ip_xmit_attach_llhdr(mp, nce);
14487 14485 if (mp == NULL) {
14488 14486 /*
14489 14487 * ip_xmit_attach_llhdr has increased
14490 14488 * ipIfStatsOutDiscards and called ip_drop_output()
14491 14489 */
14492 14490 return (ENOBUFS);
14493 14491 }
14494 14492 /*
14495 14493 * check if nce_fastpath completed and we tagged on a
14496 14494 * copy of nce_fp_mp in ip_xmit_attach_llhdr().
14497 14495 */
14498 14496 fp_mp = (mp->b_datap->db_type == M_DATA);
14499 14497
14500 14498 if (fp_mp &&
14501 14499 (ill->ill_capabilities & ILL_CAPAB_DLD_DIRECT)) {
14502 14500 ill_dld_direct_t *idd;
14503 14501
14504 14502 idd = &ill->ill_dld_capab->idc_direct;
14505 14503 /*
14506 14504 * Send the packet directly to DLD, where it
14507 14505 * may be queued depending on the availability
14508 14506 * of transmit resources at the media layer.
14509 14507 * Return value should be taken into
14510 14508 * account and flow control the TCP.
14511 14509 */
14512 14510 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
14513 14511 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
14514 14512 pkt_len);
14515 14513
14516 14514 if (ixaflags & IXAF_NO_DEV_FLOW_CTL) {
14517 14515 (void) idd->idd_tx_df(idd->idd_tx_dh, mp,
14518 14516 (uintptr_t)xmit_hint, IP_DROP_ON_NO_DESC);
14519 14517 } else {
14520 14518 uintptr_t cookie;
14521 14519
14522 14520 if ((cookie = idd->idd_tx_df(idd->idd_tx_dh,
14523 14521 mp, (uintptr_t)xmit_hint, 0)) != 0) {
14524 14522 if (ixacookie != NULL)
14525 14523 *ixacookie = cookie;
14526 14524 return (EWOULDBLOCK);
14527 14525 }
14528 14526 }
14529 14527 } else {
14530 14528 wq = ill->ill_wq;
14531 14529
14532 14530 if (!(ixaflags & IXAF_NO_DEV_FLOW_CTL) &&
14533 14531 !canputnext(wq)) {
14534 14532 if (ixacookie != NULL)
14535 14533 *ixacookie = 0;
14536 14534 ip_xmit_flowctl_drop(ill, mp, fp_mp,
14537 14535 nce->nce_fp_mp != NULL ?
14538 14536 MBLKL(nce->nce_fp_mp) : 0);
14539 14537 return (EWOULDBLOCK);
14540 14538 }
14541 14539 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
14542 14540 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
14543 14541 pkt_len);
14544 14542 putnext(wq, mp);
14545 14543 }
14546 14544
14547 14545 /*
14548 14546 * The rest of this function implements Neighbor Unreachability
14549 14547 * detection. Determine if the ncec is eligible for NUD.
14550 14548 */
14551 14549 if (ncec->ncec_flags & NCE_F_NONUD)
14552 14550 return (0);
14553 14551
14554 14552 ASSERT(ncec->ncec_state != ND_INCOMPLETE);
14555 14553
14556 14554 /*
14557 14555 * Check for upper layer advice
14558 14556 */
14559 14557 if (ixaflags & IXAF_REACH_CONF) {
14560 14558 timeout_id_t tid;
14561 14559
14562 14560 /*
14563 14561 * It should be o.k. to check the state without
14564 14562 * a lock here, at most we lose an advice.
14565 14563 */
14566 14564 ncec->ncec_last = TICK_TO_MSEC(now);
14567 14565 if (ncec->ncec_state != ND_REACHABLE) {
14568 14566 mutex_enter(&ncec->ncec_lock);
14569 14567 ncec->ncec_state = ND_REACHABLE;
14570 14568 tid = ncec->ncec_timeout_id;
14571 14569 ncec->ncec_timeout_id = 0;
14572 14570 mutex_exit(&ncec->ncec_lock);
14573 14571 (void) untimeout(tid);
14574 14572 if (ip_debug > 2) {
14575 14573 /* ip1dbg */
14576 14574 pr_addr_dbg("ip_xmit: state"
14577 14575 " for %s changed to"
14578 14576 " REACHABLE\n", AF_INET6,
14579 14577 &ncec->ncec_addr);
14580 14578 }
14581 14579 }
14582 14580 return (0);
14583 14581 }
14584 14582
14585 14583 delta = TICK_TO_MSEC(now) - ncec->ncec_last;
14586 14584 ip1dbg(("ip_xmit: delta = %" PRId64
14587 14585 " ill_reachable_time = %d \n", delta,
14588 14586 ill->ill_reachable_time));
14589 14587 if (delta > (uint64_t)ill->ill_reachable_time) {
14590 14588 mutex_enter(&ncec->ncec_lock);
14591 14589 switch (ncec->ncec_state) {
14592 14590 case ND_REACHABLE:
14593 14591 ASSERT((ncec->ncec_flags & NCE_F_NONUD) == 0);
14594 14592 /* FALLTHROUGH */
14595 14593 case ND_STALE:
14596 14594 /*
14597 14595 * ND_REACHABLE is identical to
14598 14596 * ND_STALE in this specific case. If
14599 14597 * reachable time has expired for this
14600 14598 * neighbor (delta is greater than
14601 14599 * reachable time), conceptually, the
14602 14600 * neighbor cache is no longer in
14603 14601 * REACHABLE state, but already in
14604 14602 * STALE state. So the correct
14605 14603 * transition here is to ND_DELAY.
14606 14604 */
14607 14605 ncec->ncec_state = ND_DELAY;
14608 14606 mutex_exit(&ncec->ncec_lock);
14609 14607 nce_restart_timer(ncec,
14610 14608 ipst->ips_delay_first_probe_time);
14611 14609 if (ip_debug > 3) {
14612 14610 /* ip2dbg */
14613 14611 pr_addr_dbg("ip_xmit: state"
14614 14612 " for %s changed to"
14615 14613 " DELAY\n", AF_INET6,
14616 14614 &ncec->ncec_addr);
14617 14615 }
14618 14616 break;
14619 14617 case ND_DELAY:
14620 14618 case ND_PROBE:
14621 14619 mutex_exit(&ncec->ncec_lock);
14622 14620 /* Timers have already started */
14623 14621 break;
14624 14622 case ND_UNREACHABLE:
14625 14623 /*
14626 14624 * nce_timer has detected that this ncec
14627 14625 * is unreachable and initiated deleting
14628 14626 * this ncec.
14629 14627 * This is a harmless race where we found the
14630 14628 * ncec before it was deleted and have
14631 14629 * just sent out a packet using this
14632 14630 * unreachable ncec.
14633 14631 */
14634 14632 mutex_exit(&ncec->ncec_lock);
14635 14633 break;
14636 14634 default:
14637 14635 ASSERT(0);
14638 14636 mutex_exit(&ncec->ncec_lock);
14639 14637 }
14640 14638 }
14641 14639 return (0);
14642 14640
14643 14641 case ND_INCOMPLETE:
14644 14642 /*
14645 14643 * the state could have changed since we didn't hold the lock.
14646 14644 * Re-verify state under lock.
14647 14645 */
14648 14646 is_probe = ipmp_packet_is_probe(mp, nce->nce_ill);
14649 14647 mutex_enter(&ncec->ncec_lock);
14650 14648 if (NCE_ISREACHABLE(ncec)) {
14651 14649 mutex_exit(&ncec->ncec_lock);
14652 14650 goto sendit;
14653 14651 }
14654 14652 /* queue the packet */
14655 14653 nce_queue_mp(ncec, mp, is_probe);
14656 14654 mutex_exit(&ncec->ncec_lock);
14657 14655 DTRACE_PROBE2(ip__xmit__incomplete,
14658 14656 (ncec_t *), ncec, (mblk_t *), mp);
14659 14657 return (0);
14660 14658
14661 14659 case ND_INITIAL:
14662 14660 /*
14663 14661 * State could have changed since we didn't hold the lock, so
14664 14662 * re-verify state.
14665 14663 */
14666 14664 is_probe = ipmp_packet_is_probe(mp, nce->nce_ill);
14667 14665 mutex_enter(&ncec->ncec_lock);
14668 14666 if (NCE_ISREACHABLE(ncec)) {
14669 14667 mutex_exit(&ncec->ncec_lock);
14670 14668 goto sendit;
14671 14669 }
14672 14670 nce_queue_mp(ncec, mp, is_probe);
14673 14671 if (ncec->ncec_state == ND_INITIAL) {
14674 14672 ncec->ncec_state = ND_INCOMPLETE;
14675 14673 mutex_exit(&ncec->ncec_lock);
14676 14674 /*
14677 14675 * figure out the source we want to use
14678 14676 * and resolve it.
14679 14677 */
14680 14678 ip_ndp_resolve(ncec);
14681 14679 } else {
14682 14680 mutex_exit(&ncec->ncec_lock);
14683 14681 }
14684 14682 return (0);
14685 14683
14686 14684 case ND_UNREACHABLE:
14687 14685 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14688 14686 ip_drop_output("ipIfStatsOutDiscards - ND_UNREACHABLE",
14689 14687 mp, ill);
14690 14688 freemsg(mp);
14691 14689 return (0);
14692 14690
14693 14691 default:
14694 14692 ASSERT(0);
14695 14693 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14696 14694 ip_drop_output("ipIfStatsOutDiscards - ND_other",
14697 14695 mp, ill);
14698 14696 freemsg(mp);
14699 14697 return (ENETUNREACH);
14700 14698 }
14701 14699 }
14702 14700
14703 14701 /*
14704 14702 * Return B_TRUE if the buffers differ in length or content.
14705 14703 * This is used for comparing extension header buffers.
14706 14704 * Note that an extension header would be declared different
14707 14705 * even if all that changed was the next header value in that header i.e.
14708 14706 * what really changed is the next extension header.
14709 14707 */
14710 14708 boolean_t
14711 14709 ip_cmpbuf(const void *abuf, uint_t alen, boolean_t b_valid, const void *bbuf,
14712 14710 uint_t blen)
14713 14711 {
14714 14712 if (!b_valid)
14715 14713 blen = 0;
14716 14714
14717 14715 if (alen != blen)
14718 14716 return (B_TRUE);
14719 14717 if (alen == 0)
14720 14718 return (B_FALSE); /* Both zero length */
14721 14719 return (bcmp(abuf, bbuf, alen));
14722 14720 }
14723 14721
14724 14722 /*
14725 14723 * Preallocate memory for ip_savebuf(). Returns B_TRUE if ok.
14726 14724 * Return B_FALSE if memory allocation fails - don't change any state!
14727 14725 */
14728 14726 boolean_t
14729 14727 ip_allocbuf(void **dstp, uint_t *dstlenp, boolean_t src_valid,
14730 14728 const void *src, uint_t srclen)
14731 14729 {
14732 14730 void *dst;
14733 14731
14734 14732 if (!src_valid)
14735 14733 srclen = 0;
14736 14734
14737 14735 ASSERT(*dstlenp == 0);
14738 14736 if (src != NULL && srclen != 0) {
14739 14737 dst = mi_alloc(srclen, BPRI_MED);
14740 14738 if (dst == NULL)
14741 14739 return (B_FALSE);
14742 14740 } else {
14743 14741 dst = NULL;
14744 14742 }
14745 14743 if (*dstp != NULL)
14746 14744 mi_free(*dstp);
14747 14745 *dstp = dst;
14748 14746 *dstlenp = dst == NULL ? 0 : srclen;
14749 14747 return (B_TRUE);
14750 14748 }
14751 14749
14752 14750 /*
14753 14751 * Replace what is in *dst, *dstlen with the source.
14754 14752 * Assumes ip_allocbuf has already been called.
14755 14753 */
14756 14754 void
14757 14755 ip_savebuf(void **dstp, uint_t *dstlenp, boolean_t src_valid,
14758 14756 const void *src, uint_t srclen)
14759 14757 {
14760 14758 if (!src_valid)
14761 14759 srclen = 0;
14762 14760
14763 14761 ASSERT(*dstlenp == srclen);
14764 14762 if (src != NULL && srclen != 0)
14765 14763 bcopy(src, *dstp, srclen);
14766 14764 }
14767 14765
14768 14766 /*
14769 14767 * Free the storage pointed to by the members of an ip_pkt_t.
14770 14768 */
14771 14769 void
14772 14770 ip_pkt_free(ip_pkt_t *ipp)
14773 14771 {
14774 14772 uint_t fields = ipp->ipp_fields;
14775 14773
14776 14774 if (fields & IPPF_HOPOPTS) {
14777 14775 kmem_free(ipp->ipp_hopopts, ipp->ipp_hopoptslen);
14778 14776 ipp->ipp_hopopts = NULL;
14779 14777 ipp->ipp_hopoptslen = 0;
14780 14778 }
14781 14779 if (fields & IPPF_RTHDRDSTOPTS) {
14782 14780 kmem_free(ipp->ipp_rthdrdstopts, ipp->ipp_rthdrdstoptslen);
14783 14781 ipp->ipp_rthdrdstopts = NULL;
14784 14782 ipp->ipp_rthdrdstoptslen = 0;
14785 14783 }
14786 14784 if (fields & IPPF_DSTOPTS) {
14787 14785 kmem_free(ipp->ipp_dstopts, ipp->ipp_dstoptslen);
14788 14786 ipp->ipp_dstopts = NULL;
14789 14787 ipp->ipp_dstoptslen = 0;
14790 14788 }
14791 14789 if (fields & IPPF_RTHDR) {
14792 14790 kmem_free(ipp->ipp_rthdr, ipp->ipp_rthdrlen);
14793 14791 ipp->ipp_rthdr = NULL;
14794 14792 ipp->ipp_rthdrlen = 0;
14795 14793 }
14796 14794 if (fields & IPPF_IPV4_OPTIONS) {
14797 14795 kmem_free(ipp->ipp_ipv4_options, ipp->ipp_ipv4_options_len);
14798 14796 ipp->ipp_ipv4_options = NULL;
14799 14797 ipp->ipp_ipv4_options_len = 0;
14800 14798 }
14801 14799 if (fields & IPPF_LABEL_V4) {
14802 14800 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
14803 14801 ipp->ipp_label_v4 = NULL;
14804 14802 ipp->ipp_label_len_v4 = 0;
14805 14803 }
14806 14804 if (fields & IPPF_LABEL_V6) {
14807 14805 kmem_free(ipp->ipp_label_v6, ipp->ipp_label_len_v6);
14808 14806 ipp->ipp_label_v6 = NULL;
14809 14807 ipp->ipp_label_len_v6 = 0;
14810 14808 }
14811 14809 ipp->ipp_fields &= ~(IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14812 14810 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6);
14813 14811 }
14814 14812
14815 14813 /*
14816 14814 * Copy from src to dst and allocate as needed.
14817 14815 * Returns zero or ENOMEM.
14818 14816 *
14819 14817 * The caller must initialize dst to zero.
14820 14818 */
14821 14819 int
14822 14820 ip_pkt_copy(ip_pkt_t *src, ip_pkt_t *dst, int kmflag)
14823 14821 {
14824 14822 uint_t fields = src->ipp_fields;
14825 14823
14826 14824 /* Start with fields that don't require memory allocation */
14827 14825 dst->ipp_fields = fields &
14828 14826 ~(IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14829 14827 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6);
14830 14828
14831 14829 dst->ipp_addr = src->ipp_addr;
14832 14830 dst->ipp_unicast_hops = src->ipp_unicast_hops;
14833 14831 dst->ipp_hoplimit = src->ipp_hoplimit;
14834 14832 dst->ipp_tclass = src->ipp_tclass;
14835 14833 dst->ipp_type_of_service = src->ipp_type_of_service;
14836 14834
14837 14835 if (!(fields & (IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14838 14836 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6)))
14839 14837 return (0);
14840 14838
14841 14839 if (fields & IPPF_HOPOPTS) {
14842 14840 dst->ipp_hopopts = kmem_alloc(src->ipp_hopoptslen, kmflag);
14843 14841 if (dst->ipp_hopopts == NULL) {
14844 14842 ip_pkt_free(dst);
14845 14843 return (ENOMEM);
14846 14844 }
14847 14845 dst->ipp_fields |= IPPF_HOPOPTS;
14848 14846 bcopy(src->ipp_hopopts, dst->ipp_hopopts,
14849 14847 src->ipp_hopoptslen);
14850 14848 dst->ipp_hopoptslen = src->ipp_hopoptslen;
14851 14849 }
14852 14850 if (fields & IPPF_RTHDRDSTOPTS) {
14853 14851 dst->ipp_rthdrdstopts = kmem_alloc(src->ipp_rthdrdstoptslen,
14854 14852 kmflag);
14855 14853 if (dst->ipp_rthdrdstopts == NULL) {
14856 14854 ip_pkt_free(dst);
14857 14855 return (ENOMEM);
14858 14856 }
14859 14857 dst->ipp_fields |= IPPF_RTHDRDSTOPTS;
14860 14858 bcopy(src->ipp_rthdrdstopts, dst->ipp_rthdrdstopts,
14861 14859 src->ipp_rthdrdstoptslen);
14862 14860 dst->ipp_rthdrdstoptslen = src->ipp_rthdrdstoptslen;
14863 14861 }
14864 14862 if (fields & IPPF_DSTOPTS) {
14865 14863 dst->ipp_dstopts = kmem_alloc(src->ipp_dstoptslen, kmflag);
14866 14864 if (dst->ipp_dstopts == NULL) {
14867 14865 ip_pkt_free(dst);
14868 14866 return (ENOMEM);
14869 14867 }
14870 14868 dst->ipp_fields |= IPPF_DSTOPTS;
14871 14869 bcopy(src->ipp_dstopts, dst->ipp_dstopts,
14872 14870 src->ipp_dstoptslen);
14873 14871 dst->ipp_dstoptslen = src->ipp_dstoptslen;
14874 14872 }
14875 14873 if (fields & IPPF_RTHDR) {
14876 14874 dst->ipp_rthdr = kmem_alloc(src->ipp_rthdrlen, kmflag);
14877 14875 if (dst->ipp_rthdr == NULL) {
14878 14876 ip_pkt_free(dst);
14879 14877 return (ENOMEM);
14880 14878 }
14881 14879 dst->ipp_fields |= IPPF_RTHDR;
14882 14880 bcopy(src->ipp_rthdr, dst->ipp_rthdr,
14883 14881 src->ipp_rthdrlen);
14884 14882 dst->ipp_rthdrlen = src->ipp_rthdrlen;
14885 14883 }
14886 14884 if (fields & IPPF_IPV4_OPTIONS) {
14887 14885 dst->ipp_ipv4_options = kmem_alloc(src->ipp_ipv4_options_len,
14888 14886 kmflag);
14889 14887 if (dst->ipp_ipv4_options == NULL) {
14890 14888 ip_pkt_free(dst);
14891 14889 return (ENOMEM);
14892 14890 }
14893 14891 dst->ipp_fields |= IPPF_IPV4_OPTIONS;
14894 14892 bcopy(src->ipp_ipv4_options, dst->ipp_ipv4_options,
14895 14893 src->ipp_ipv4_options_len);
14896 14894 dst->ipp_ipv4_options_len = src->ipp_ipv4_options_len;
14897 14895 }
14898 14896 if (fields & IPPF_LABEL_V4) {
14899 14897 dst->ipp_label_v4 = kmem_alloc(src->ipp_label_len_v4, kmflag);
14900 14898 if (dst->ipp_label_v4 == NULL) {
14901 14899 ip_pkt_free(dst);
14902 14900 return (ENOMEM);
14903 14901 }
14904 14902 dst->ipp_fields |= IPPF_LABEL_V4;
14905 14903 bcopy(src->ipp_label_v4, dst->ipp_label_v4,
14906 14904 src->ipp_label_len_v4);
14907 14905 dst->ipp_label_len_v4 = src->ipp_label_len_v4;
14908 14906 }
14909 14907 if (fields & IPPF_LABEL_V6) {
14910 14908 dst->ipp_label_v6 = kmem_alloc(src->ipp_label_len_v6, kmflag);
14911 14909 if (dst->ipp_label_v6 == NULL) {
14912 14910 ip_pkt_free(dst);
14913 14911 return (ENOMEM);
14914 14912 }
14915 14913 dst->ipp_fields |= IPPF_LABEL_V6;
14916 14914 bcopy(src->ipp_label_v6, dst->ipp_label_v6,
14917 14915 src->ipp_label_len_v6);
14918 14916 dst->ipp_label_len_v6 = src->ipp_label_len_v6;
14919 14917 }
14920 14918 if (fields & IPPF_FRAGHDR) {
14921 14919 dst->ipp_fraghdr = kmem_alloc(src->ipp_fraghdrlen, kmflag);
14922 14920 if (dst->ipp_fraghdr == NULL) {
14923 14921 ip_pkt_free(dst);
14924 14922 return (ENOMEM);
14925 14923 }
14926 14924 dst->ipp_fields |= IPPF_FRAGHDR;
14927 14925 bcopy(src->ipp_fraghdr, dst->ipp_fraghdr,
14928 14926 src->ipp_fraghdrlen);
14929 14927 dst->ipp_fraghdrlen = src->ipp_fraghdrlen;
14930 14928 }
14931 14929 return (0);
14932 14930 }
14933 14931
14934 14932 /*
14935 14933 * Returns INADDR_ANY if no source route
14936 14934 */
14937 14935 ipaddr_t
14938 14936 ip_pkt_source_route_v4(const ip_pkt_t *ipp)
14939 14937 {
14940 14938 ipaddr_t nexthop = INADDR_ANY;
14941 14939 ipoptp_t opts;
14942 14940 uchar_t *opt;
14943 14941 uint8_t optval;
14944 14942 uint8_t optlen;
14945 14943 uint32_t totallen;
14946 14944
14947 14945 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
14948 14946 return (INADDR_ANY);
14949 14947
14950 14948 totallen = ipp->ipp_ipv4_options_len;
14951 14949 if (totallen & 0x3)
14952 14950 return (INADDR_ANY);
14953 14951
14954 14952 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
14955 14953 optval != IPOPT_EOL;
14956 14954 optval = ipoptp_next(&opts)) {
14957 14955 opt = opts.ipoptp_cur;
14958 14956 switch (optval) {
14959 14957 uint8_t off;
14960 14958 case IPOPT_SSRR:
14961 14959 case IPOPT_LSRR:
14962 14960 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
14963 14961 break;
14964 14962 }
14965 14963 optlen = opts.ipoptp_len;
14966 14964 off = opt[IPOPT_OFFSET];
14967 14965 off--;
14968 14966 if (optlen < IP_ADDR_LEN ||
14969 14967 off > optlen - IP_ADDR_LEN) {
14970 14968 /* End of source route */
14971 14969 break;
14972 14970 }
14973 14971 bcopy((char *)opt + off, &nexthop, IP_ADDR_LEN);
14974 14972 if (nexthop == htonl(INADDR_LOOPBACK)) {
14975 14973 /* Ignore */
14976 14974 nexthop = INADDR_ANY;
14977 14975 break;
14978 14976 }
14979 14977 break;
14980 14978 }
14981 14979 }
14982 14980 return (nexthop);
14983 14981 }
14984 14982
14985 14983 /*
14986 14984 * Reverse a source route.
14987 14985 */
14988 14986 void
14989 14987 ip_pkt_source_route_reverse_v4(ip_pkt_t *ipp)
14990 14988 {
14991 14989 ipaddr_t tmp;
14992 14990 ipoptp_t opts;
14993 14991 uchar_t *opt;
14994 14992 uint8_t optval;
14995 14993 uint32_t totallen;
14996 14994
14997 14995 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
14998 14996 return;
14999 14997
15000 14998 totallen = ipp->ipp_ipv4_options_len;
15001 14999 if (totallen & 0x3)
15002 15000 return;
15003 15001
15004 15002 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
15005 15003 optval != IPOPT_EOL;
15006 15004 optval = ipoptp_next(&opts)) {
15007 15005 uint8_t off1, off2;
15008 15006
15009 15007 opt = opts.ipoptp_cur;
15010 15008 switch (optval) {
15011 15009 case IPOPT_SSRR:
15012 15010 case IPOPT_LSRR:
15013 15011 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
15014 15012 break;
15015 15013 }
15016 15014 off1 = IPOPT_MINOFF_SR - 1;
15017 15015 off2 = opt[IPOPT_OFFSET] - IP_ADDR_LEN - 1;
15018 15016 while (off2 > off1) {
15019 15017 bcopy(opt + off2, &tmp, IP_ADDR_LEN);
15020 15018 bcopy(opt + off1, opt + off2, IP_ADDR_LEN);
15021 15019 bcopy(&tmp, opt + off2, IP_ADDR_LEN);
15022 15020 off2 -= IP_ADDR_LEN;
15023 15021 off1 += IP_ADDR_LEN;
15024 15022 }
15025 15023 opt[IPOPT_OFFSET] = IPOPT_MINOFF_SR;
15026 15024 break;
15027 15025 }
15028 15026 }
15029 15027 }
15030 15028
15031 15029 /*
15032 15030 * Returns NULL if no routing header
15033 15031 */
15034 15032 in6_addr_t *
15035 15033 ip_pkt_source_route_v6(const ip_pkt_t *ipp)
15036 15034 {
15037 15035 in6_addr_t *nexthop = NULL;
15038 15036 ip6_rthdr0_t *rthdr;
15039 15037
15040 15038 if (!(ipp->ipp_fields & IPPF_RTHDR))
15041 15039 return (NULL);
15042 15040
15043 15041 rthdr = (ip6_rthdr0_t *)ipp->ipp_rthdr;
15044 15042 if (rthdr->ip6r0_segleft == 0)
15045 15043 return (NULL);
15046 15044
15047 15045 nexthop = (in6_addr_t *)((char *)rthdr + sizeof (*rthdr));
15048 15046 return (nexthop);
15049 15047 }
15050 15048
15051 15049 zoneid_t
15052 15050 ip_get_zoneid_v4(ipaddr_t addr, mblk_t *mp, ip_recv_attr_t *ira,
15053 15051 zoneid_t lookup_zoneid)
15054 15052 {
15055 15053 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
15056 15054 ire_t *ire;
15057 15055 int ire_flags = MATCH_IRE_TYPE;
15058 15056 zoneid_t zoneid = ALL_ZONES;
15059 15057
15060 15058 if (is_system_labeled() && !tsol_can_accept_raw(mp, ira, B_FALSE))
15061 15059 return (ALL_ZONES);
15062 15060
15063 15061 if (lookup_zoneid != ALL_ZONES)
15064 15062 ire_flags |= MATCH_IRE_ZONEONLY;
15065 15063 ire = ire_ftable_lookup_v4(addr, 0, 0, IRE_LOCAL | IRE_LOOPBACK,
15066 15064 NULL, lookup_zoneid, NULL, ire_flags, 0, ipst, NULL);
15067 15065 if (ire != NULL) {
15068 15066 zoneid = IP_REAL_ZONEID(ire->ire_zoneid, ipst);
15069 15067 ire_refrele(ire);
15070 15068 }
15071 15069 return (zoneid);
15072 15070 }
15073 15071
15074 15072 zoneid_t
15075 15073 ip_get_zoneid_v6(in6_addr_t *addr, mblk_t *mp, const ill_t *ill,
15076 15074 ip_recv_attr_t *ira, zoneid_t lookup_zoneid)
15077 15075 {
15078 15076 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
15079 15077 ire_t *ire;
15080 15078 int ire_flags = MATCH_IRE_TYPE;
15081 15079 zoneid_t zoneid = ALL_ZONES;
15082 15080
15083 15081 if (is_system_labeled() && !tsol_can_accept_raw(mp, ira, B_FALSE))
15084 15082 return (ALL_ZONES);
15085 15083
15086 15084 if (IN6_IS_ADDR_LINKLOCAL(addr))
15087 15085 ire_flags |= MATCH_IRE_ILL;
15088 15086
15089 15087 if (lookup_zoneid != ALL_ZONES)
15090 15088 ire_flags |= MATCH_IRE_ZONEONLY;
15091 15089 ire = ire_ftable_lookup_v6(addr, NULL, NULL, IRE_LOCAL | IRE_LOOPBACK,
15092 15090 ill, lookup_zoneid, NULL, ire_flags, 0, ipst, NULL);
15093 15091 if (ire != NULL) {
15094 15092 zoneid = IP_REAL_ZONEID(ire->ire_zoneid, ipst);
15095 15093 ire_refrele(ire);
15096 15094 }
15097 15095 return (zoneid);
15098 15096 }
15099 15097
15100 15098 /*
15101 15099 * IP obserability hook support functions.
15102 15100 */
15103 15101 static void
15104 15102 ipobs_init(ip_stack_t *ipst)
15105 15103 {
15106 15104 netid_t id;
15107 15105
15108 15106 id = net_getnetidbynetstackid(ipst->ips_netstack->netstack_stackid);
15109 15107
15110 15108 ipst->ips_ip4_observe_pr = net_protocol_lookup(id, NHF_INET);
15111 15109 VERIFY(ipst->ips_ip4_observe_pr != NULL);
15112 15110
15113 15111 ipst->ips_ip6_observe_pr = net_protocol_lookup(id, NHF_INET6);
15114 15112 VERIFY(ipst->ips_ip6_observe_pr != NULL);
15115 15113 }
15116 15114
15117 15115 static void
15118 15116 ipobs_fini(ip_stack_t *ipst)
15119 15117 {
15120 15118
15121 15119 VERIFY(net_protocol_release(ipst->ips_ip4_observe_pr) == 0);
15122 15120 VERIFY(net_protocol_release(ipst->ips_ip6_observe_pr) == 0);
15123 15121 }
15124 15122
15125 15123 /*
15126 15124 * hook_pkt_observe_t is composed in network byte order so that the
15127 15125 * entire mblk_t chain handed into hook_run can be used as-is.
15128 15126 * The caveat is that use of the fields, such as the zone fields,
15129 15127 * requires conversion into host byte order first.
15130 15128 */
15131 15129 void
15132 15130 ipobs_hook(mblk_t *mp, int htype, zoneid_t zsrc, zoneid_t zdst,
15133 15131 const ill_t *ill, ip_stack_t *ipst)
15134 15132 {
15135 15133 hook_pkt_observe_t *hdr;
15136 15134 uint64_t grifindex;
15137 15135 mblk_t *imp;
15138 15136
15139 15137 imp = allocb(sizeof (*hdr), BPRI_HI);
15140 15138 if (imp == NULL)
15141 15139 return;
15142 15140
15143 15141 hdr = (hook_pkt_observe_t *)imp->b_rptr;
15144 15142 /*
15145 15143 * b_wptr is set to make the apparent size of the data in the mblk_t
15146 15144 * to exclude the pointers at the end of hook_pkt_observer_t.
15147 15145 */
15148 15146 imp->b_wptr = imp->b_rptr + sizeof (dl_ipnetinfo_t);
15149 15147 imp->b_cont = mp;
15150 15148
15151 15149 ASSERT(DB_TYPE(mp) == M_DATA);
15152 15150
15153 15151 if (IS_UNDER_IPMP(ill))
15154 15152 grifindex = ipmp_ill_get_ipmp_ifindex(ill);
15155 15153 else
15156 15154 grifindex = 0;
15157 15155
15158 15156 hdr->hpo_version = 1;
15159 15157 hdr->hpo_htype = htons(htype);
15160 15158 hdr->hpo_pktlen = htonl((ulong_t)msgdsize(mp));
15161 15159 hdr->hpo_ifindex = htonl(ill->ill_phyint->phyint_ifindex);
15162 15160 hdr->hpo_grifindex = htonl(grifindex);
15163 15161 hdr->hpo_zsrc = htonl(zsrc);
15164 15162 hdr->hpo_zdst = htonl(zdst);
15165 15163 hdr->hpo_pkt = imp;
15166 15164 hdr->hpo_ctx = ipst->ips_netstack;
15167 15165
15168 15166 if (ill->ill_isv6) {
15169 15167 hdr->hpo_family = AF_INET6;
15170 15168 (void) hook_run(ipst->ips_ipv6_net_data->netd_hooks,
15171 15169 ipst->ips_ipv6observing, (hook_data_t)hdr);
15172 15170 } else {
15173 15171 hdr->hpo_family = AF_INET;
15174 15172 (void) hook_run(ipst->ips_ipv4_net_data->netd_hooks,
15175 15173 ipst->ips_ipv4observing, (hook_data_t)hdr);
15176 15174 }
15177 15175
15178 15176 imp->b_cont = NULL;
15179 15177 freemsg(imp);
15180 15178 }
15181 15179
15182 15180 /*
15183 15181 * Utility routine that checks if `v4srcp' is a valid address on underlying
15184 15182 * interface `ill'. If `ipifp' is non-NULL, it's set to a held ipif
15185 15183 * associated with `v4srcp' on success. NOTE: if this is not called from
15186 15184 * inside the IPSQ (ill_g_lock is not held), `ill' may be removed from the
15187 15185 * group during or after this lookup.
15188 15186 */
15189 15187 boolean_t
15190 15188 ipif_lookup_testaddr_v4(ill_t *ill, const in_addr_t *v4srcp, ipif_t **ipifp)
15191 15189 {
15192 15190 ipif_t *ipif;
15193 15191
15194 15192 ipif = ipif_lookup_addr_exact(*v4srcp, ill, ill->ill_ipst);
15195 15193 if (ipif != NULL) {
15196 15194 if (ipifp != NULL)
15197 15195 *ipifp = ipif;
15198 15196 else
15199 15197 ipif_refrele(ipif);
15200 15198 return (B_TRUE);
15201 15199 }
15202 15200
15203 15201 ip1dbg(("ipif_lookup_testaddr_v4: cannot find ipif for src %x\n",
15204 15202 *v4srcp));
15205 15203 return (B_FALSE);
15206 15204 }
15207 15205
15208 15206 /*
15209 15207 * Transport protocol call back function for CPU state change.
15210 15208 */
15211 15209 /* ARGSUSED */
15212 15210 static int
15213 15211 ip_tp_cpu_update(cpu_setup_t what, int id, void *arg)
15214 15212 {
15215 15213 processorid_t cpu_seqid;
15216 15214 netstack_handle_t nh;
15217 15215 netstack_t *ns;
15218 15216
15219 15217 ASSERT(MUTEX_HELD(&cpu_lock));
15220 15218
15221 15219 switch (what) {
15222 15220 case CPU_CONFIG:
15223 15221 case CPU_ON:
15224 15222 case CPU_INIT:
15225 15223 case CPU_CPUPART_IN:
15226 15224 cpu_seqid = cpu[id]->cpu_seqid;
15227 15225 netstack_next_init(&nh);
15228 15226 while ((ns = netstack_next(&nh)) != NULL) {
15229 15227 tcp_stack_cpu_add(ns->netstack_tcp, cpu_seqid);
15230 15228 sctp_stack_cpu_add(ns->netstack_sctp, cpu_seqid);
15231 15229 udp_stack_cpu_add(ns->netstack_udp, cpu_seqid);
15232 15230 netstack_rele(ns);
15233 15231 }
15234 15232 netstack_next_fini(&nh);
15235 15233 break;
15236 15234 case CPU_UNCONFIG:
15237 15235 case CPU_OFF:
15238 15236 case CPU_CPUPART_OUT:
15239 15237 /*
15240 15238 * Nothing to do. We don't remove the per CPU stats from
15241 15239 * the IP stack even when the CPU goes offline.
15242 15240 */
15243 15241 break;
15244 15242 default:
15245 15243 break;
15246 15244 }
15247 15245 return (0);
15248 15246 }
|
↓ open down ↓ |
11110 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX