Print this page
915 ill_taskq_dispatch() race condition
(originally Joyent OS-119 panic from mutex_vector_exit)
| 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.
|
↓ open down ↓ |
14 lines elided |
↑ open up ↑ |
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 + * Copyright (c) 2011 Joyent, Inc. All rights reserved.
25 26 */
26 27
27 28 #include <sys/types.h>
28 29 #include <sys/stream.h>
29 30 #include <sys/dlpi.h>
30 31 #include <sys/stropts.h>
31 32 #include <sys/sysmacros.h>
32 33 #include <sys/strsubr.h>
33 34 #include <sys/strlog.h>
34 35 #include <sys/strsun.h>
35 36 #include <sys/zone.h>
36 37 #define _SUN_TPI_VERSION 2
37 38 #include <sys/tihdr.h>
38 39 #include <sys/xti_inet.h>
39 40 #include <sys/ddi.h>
40 41 #include <sys/suntpi.h>
41 42 #include <sys/cmn_err.h>
42 43 #include <sys/debug.h>
43 44 #include <sys/kobj.h>
44 45 #include <sys/modctl.h>
45 46 #include <sys/atomic.h>
46 47 #include <sys/policy.h>
47 48 #include <sys/priv.h>
48 49 #include <sys/taskq.h>
49 50
50 51 #include <sys/systm.h>
51 52 #include <sys/param.h>
52 53 #include <sys/kmem.h>
53 54 #include <sys/sdt.h>
54 55 #include <sys/socket.h>
55 56 #include <sys/vtrace.h>
56 57 #include <sys/isa_defs.h>
57 58 #include <sys/mac.h>
58 59 #include <net/if.h>
59 60 #include <net/if_arp.h>
60 61 #include <net/route.h>
61 62 #include <sys/sockio.h>
62 63 #include <netinet/in.h>
63 64 #include <net/if_dl.h>
64 65
65 66 #include <inet/common.h>
66 67 #include <inet/mi.h>
67 68 #include <inet/mib2.h>
68 69 #include <inet/nd.h>
69 70 #include <inet/arp.h>
70 71 #include <inet/snmpcom.h>
71 72 #include <inet/optcom.h>
72 73 #include <inet/kstatcom.h>
73 74
74 75 #include <netinet/igmp_var.h>
75 76 #include <netinet/ip6.h>
76 77 #include <netinet/icmp6.h>
77 78 #include <netinet/sctp.h>
78 79
79 80 #include <inet/ip.h>
80 81 #include <inet/ip_impl.h>
81 82 #include <inet/ip6.h>
82 83 #include <inet/ip6_asp.h>
83 84 #include <inet/tcp.h>
84 85 #include <inet/tcp_impl.h>
85 86 #include <inet/ip_multi.h>
86 87 #include <inet/ip_if.h>
87 88 #include <inet/ip_ire.h>
88 89 #include <inet/ip_ftable.h>
89 90 #include <inet/ip_rts.h>
90 91 #include <inet/ip_ndp.h>
91 92 #include <inet/ip_listutils.h>
92 93 #include <netinet/igmp.h>
93 94 #include <netinet/ip_mroute.h>
94 95 #include <inet/ipp_common.h>
95 96
96 97 #include <net/pfkeyv2.h>
97 98 #include <inet/sadb.h>
98 99 #include <inet/ipsec_impl.h>
99 100 #include <inet/iptun/iptun_impl.h>
100 101 #include <inet/ipdrop.h>
101 102 #include <inet/ip_netinfo.h>
102 103 #include <inet/ilb_ip.h>
103 104
104 105 #include <sys/ethernet.h>
105 106 #include <net/if_types.h>
106 107 #include <sys/cpuvar.h>
107 108
108 109 #include <ipp/ipp.h>
109 110 #include <ipp/ipp_impl.h>
110 111 #include <ipp/ipgpc/ipgpc.h>
111 112
112 113 #include <sys/pattr.h>
113 114 #include <inet/ipclassifier.h>
114 115 #include <inet/sctp_ip.h>
115 116 #include <inet/sctp/sctp_impl.h>
116 117 #include <inet/udp_impl.h>
117 118 #include <inet/rawip_impl.h>
118 119 #include <inet/rts_impl.h>
119 120
120 121 #include <sys/tsol/label.h>
121 122 #include <sys/tsol/tnet.h>
122 123
123 124 #include <sys/squeue_impl.h>
124 125 #include <inet/ip_arp.h>
125 126
126 127 #include <sys/clock_impl.h> /* For LBOLT_FASTPATH{,64} */
127 128
128 129 /*
129 130 * Values for squeue switch:
130 131 * IP_SQUEUE_ENTER_NODRAIN: SQ_NODRAIN
131 132 * IP_SQUEUE_ENTER: SQ_PROCESS
132 133 * IP_SQUEUE_FILL: SQ_FILL
133 134 */
134 135 int ip_squeue_enter = IP_SQUEUE_ENTER; /* Setable in /etc/system */
135 136
136 137 int ip_squeue_flag;
137 138
138 139 /*
139 140 * Setable in /etc/system
140 141 */
141 142 int ip_poll_normal_ms = 100;
142 143 int ip_poll_normal_ticks = 0;
143 144 int ip_modclose_ackwait_ms = 3000;
144 145
145 146 /*
146 147 * It would be nice to have these present only in DEBUG systems, but the
147 148 * current design of the global symbol checking logic requires them to be
148 149 * unconditionally present.
149 150 */
150 151 uint_t ip_thread_data; /* TSD key for debug support */
151 152 krwlock_t ip_thread_rwlock;
152 153 list_t ip_thread_list;
153 154
154 155 /*
155 156 * Structure to represent a linked list of msgblks. Used by ip_snmp_ functions.
156 157 */
157 158
158 159 struct listptr_s {
159 160 mblk_t *lp_head; /* pointer to the head of the list */
160 161 mblk_t *lp_tail; /* pointer to the tail of the list */
161 162 };
162 163
163 164 typedef struct listptr_s listptr_t;
164 165
165 166 /*
166 167 * This is used by ip_snmp_get_mib2_ip_route_media and
167 168 * ip_snmp_get_mib2_ip6_route_media to carry the lists of return data.
168 169 */
169 170 typedef struct iproutedata_s {
170 171 uint_t ird_idx;
171 172 uint_t ird_flags; /* see below */
172 173 listptr_t ird_route; /* ipRouteEntryTable */
173 174 listptr_t ird_netmedia; /* ipNetToMediaEntryTable */
174 175 listptr_t ird_attrs; /* ipRouteAttributeTable */
175 176 } iproutedata_t;
176 177
177 178 /* Include ire_testhidden and IRE_IF_CLONE routes */
178 179 #define IRD_REPORT_ALL 0x01
179 180
180 181 /*
181 182 * Cluster specific hooks. These should be NULL when booted as a non-cluster
182 183 */
183 184
184 185 /*
185 186 * Hook functions to enable cluster networking
186 187 * On non-clustered systems these vectors must always be NULL.
187 188 *
188 189 * Hook function to Check ip specified ip address is a shared ip address
189 190 * in the cluster
190 191 *
191 192 */
192 193 int (*cl_inet_isclusterwide)(netstackid_t stack_id, uint8_t protocol,
193 194 sa_family_t addr_family, uint8_t *laddrp, void *args) = NULL;
194 195
195 196 /*
196 197 * Hook function to generate cluster wide ip fragment identifier
197 198 */
198 199 uint32_t (*cl_inet_ipident)(netstackid_t stack_id, uint8_t protocol,
199 200 sa_family_t addr_family, uint8_t *laddrp, uint8_t *faddrp,
200 201 void *args) = NULL;
201 202
202 203 /*
203 204 * Hook function to generate cluster wide SPI.
204 205 */
205 206 void (*cl_inet_getspi)(netstackid_t, uint8_t, uint8_t *, size_t,
206 207 void *) = NULL;
207 208
208 209 /*
209 210 * Hook function to verify if the SPI is already utlized.
210 211 */
211 212
212 213 int (*cl_inet_checkspi)(netstackid_t, uint8_t, uint32_t, void *) = NULL;
213 214
214 215 /*
215 216 * Hook function to delete the SPI from the cluster wide repository.
216 217 */
217 218
218 219 void (*cl_inet_deletespi)(netstackid_t, uint8_t, uint32_t, void *) = NULL;
219 220
220 221 /*
221 222 * Hook function to inform the cluster when packet received on an IDLE SA
222 223 */
223 224
224 225 void (*cl_inet_idlesa)(netstackid_t, uint8_t, uint32_t, sa_family_t,
225 226 in6_addr_t, in6_addr_t, void *) = NULL;
226 227
227 228 /*
228 229 * Synchronization notes:
229 230 *
230 231 * IP is a fully D_MP STREAMS module/driver. Thus it does not depend on any
231 232 * MT level protection given by STREAMS. IP uses a combination of its own
232 233 * internal serialization mechanism and standard Solaris locking techniques.
233 234 * The internal serialization is per phyint. This is used to serialize
234 235 * plumbing operations, IPMP operations, most set ioctls, etc.
235 236 *
236 237 * Plumbing is a long sequence of operations involving message
237 238 * exchanges between IP, ARP and device drivers. Many set ioctls are typically
238 239 * involved in plumbing operations. A natural model is to serialize these
239 240 * ioctls one per ill. For example plumbing of hme0 and qfe0 can go on in
240 241 * parallel without any interference. But various set ioctls on hme0 are best
241 242 * serialized, along with IPMP operations and processing of DLPI control
242 243 * messages received from drivers on a per phyint basis. This serialization is
243 244 * provided by the ipsq_t and primitives operating on this. Details can
244 245 * be found in ip_if.c above the core primitives operating on ipsq_t.
245 246 *
246 247 * Lookups of an ipif or ill by a thread return a refheld ipif / ill.
247 248 * Simiarly lookup of an ire by a thread also returns a refheld ire.
248 249 * In addition ipif's and ill's referenced by the ire are also indirectly
249 250 * refheld. Thus no ipif or ill can vanish as long as an ipif is refheld
250 251 * directly or indirectly. For example an SIOCSLIFADDR ioctl that changes the
251 252 * address of an ipif has to go through the ipsq_t. This ensures that only
252 253 * one such exclusive operation proceeds at any time on the ipif. It then
253 254 * waits for all refcnts
254 255 * associated with this ipif to come down to zero. The address is changed
255 256 * only after the ipif has been quiesced. Then the ipif is brought up again.
256 257 * More details are described above the comment in ip_sioctl_flags.
257 258 *
258 259 * Packet processing is based mostly on IREs and are fully multi-threaded
259 260 * using standard Solaris MT techniques.
260 261 *
261 262 * There are explicit locks in IP to handle:
262 263 * - The ip_g_head list maintained by mi_open_link() and friends.
263 264 *
264 265 * - The reassembly data structures (one lock per hash bucket)
265 266 *
266 267 * - conn_lock is meant to protect conn_t fields. The fields actually
267 268 * protected by conn_lock are documented in the conn_t definition.
268 269 *
269 270 * - ire_lock to protect some of the fields of the ire, IRE tables
270 271 * (one lock per hash bucket). Refer to ip_ire.c for details.
271 272 *
272 273 * - ndp_g_lock and ncec_lock for protecting NCEs.
273 274 *
274 275 * - ill_lock protects fields of the ill and ipif. Details in ip.h
275 276 *
276 277 * - ill_g_lock: This is a global reader/writer lock. Protects the following
277 278 * * The AVL tree based global multi list of all ills.
278 279 * * The linked list of all ipifs of an ill
279 280 * * The <ipsq-xop> mapping
280 281 * * <ill-phyint> association
281 282 * Insertion/deletion of an ill in the system, insertion/deletion of an ipif
282 283 * into an ill, changing the <ipsq-xop> mapping of an ill, changing the
283 284 * <ill-phyint> assoc of an ill will all have to hold the ill_g_lock as
284 285 * writer for the actual duration of the insertion/deletion/change.
285 286 *
286 287 * - ill_lock: This is a per ill mutex.
287 288 * It protects some members of the ill_t struct; see ip.h for details.
288 289 * It also protects the <ill-phyint> assoc.
289 290 * It also protects the list of ipifs hanging off the ill.
290 291 *
291 292 * - ipsq_lock: This is a per ipsq_t mutex lock.
292 293 * This protects some members of the ipsq_t struct; see ip.h for details.
293 294 * It also protects the <ipsq-ipxop> mapping
294 295 *
295 296 * - ipx_lock: This is a per ipxop_t mutex lock.
296 297 * This protects some members of the ipxop_t struct; see ip.h for details.
297 298 *
298 299 * - phyint_lock: This is a per phyint mutex lock. Protects just the
299 300 * phyint_flags
300 301 *
301 302 * - ip_addr_avail_lock: This is used to ensure the uniqueness of IP addresses.
302 303 * This lock is held in ipif_up_done and the ipif is marked IPIF_UP and the
303 304 * uniqueness check also done atomically.
304 305 *
305 306 * - ill_g_usesrc_lock: This readers/writer lock protects the usesrc
306 307 * group list linked by ill_usesrc_grp_next. It also protects the
307 308 * ill_usesrc_ifindex field. It is taken as a writer when a member of the
308 309 * group is being added or deleted. This lock is taken as a reader when
309 310 * walking the list/group(eg: to get the number of members in a usesrc group).
310 311 * Note, it is only necessary to take this lock if the ill_usesrc_grp_next
311 312 * field is changing state i.e from NULL to non-NULL or vice-versa. For
312 313 * example, it is not necessary to take this lock in the initial portion
313 314 * of ip_sioctl_slifusesrc or at all in ip_sioctl_flags since these
314 315 * operations are executed exclusively and that ensures that the "usesrc
315 316 * group state" cannot change. The "usesrc group state" change can happen
316 317 * only in the latter part of ip_sioctl_slifusesrc and in ill_delete.
317 318 *
318 319 * Changing <ill-phyint>, <ipsq-xop> assocications:
319 320 *
320 321 * To change the <ill-phyint> association, the ill_g_lock must be held
321 322 * as writer, and the ill_locks of both the v4 and v6 instance of the ill
322 323 * must be held.
323 324 *
324 325 * To change the <ipsq-xop> association, the ill_g_lock must be held as
325 326 * writer, the ipsq_lock must be held, and one must be writer on the ipsq.
326 327 * This is only done when ills are added or removed from IPMP groups.
327 328 *
328 329 * To add or delete an ipif from the list of ipifs hanging off the ill,
329 330 * ill_g_lock (writer) and ill_lock must be held and the thread must be
330 331 * a writer on the associated ipsq.
331 332 *
332 333 * To add or delete an ill to the system, the ill_g_lock must be held as
333 334 * writer and the thread must be a writer on the associated ipsq.
334 335 *
335 336 * To add or delete an ilm to an ill, the ill_lock must be held and the thread
336 337 * must be a writer on the associated ipsq.
337 338 *
338 339 * Lock hierarchy
339 340 *
340 341 * Some lock hierarchy scenarios are listed below.
341 342 *
342 343 * ill_g_lock -> conn_lock -> ill_lock -> ipsq_lock -> ipx_lock
343 344 * ill_g_lock -> ill_lock(s) -> phyint_lock
344 345 * ill_g_lock -> ndp_g_lock -> ill_lock -> ncec_lock
345 346 * ill_g_lock -> ip_addr_avail_lock
346 347 * conn_lock -> irb_lock -> ill_lock -> ire_lock
347 348 * ill_g_lock -> ip_g_nd_lock
348 349 * ill_g_lock -> ips_ipmp_lock -> ill_lock -> nce_lock
349 350 * ill_g_lock -> ndp_g_lock -> ill_lock -> ncec_lock -> nce_lock
350 351 * arl_lock -> ill_lock
351 352 * ips_ire_dep_lock -> irb_lock
352 353 *
353 354 * When more than 1 ill lock is needed to be held, all ill lock addresses
354 355 * are sorted on address and locked starting from highest addressed lock
355 356 * downward.
356 357 *
357 358 * Multicast scenarios
358 359 * ips_ill_g_lock -> ill_mcast_lock
359 360 * conn_ilg_lock -> ips_ill_g_lock -> ill_lock
360 361 * ill_mcast_serializer -> ill_mcast_lock -> ips_ipmp_lock -> ill_lock
361 362 * ill_mcast_serializer -> ill_mcast_lock -> connf_lock -> conn_lock
362 363 * ill_mcast_serializer -> ill_mcast_lock -> conn_ilg_lock
363 364 * ill_mcast_serializer -> ill_mcast_lock -> ips_igmp_timer_lock
364 365 *
365 366 * IPsec scenarios
366 367 *
367 368 * ipsa_lock -> ill_g_lock -> ill_lock
368 369 * ill_g_usesrc_lock -> ill_g_lock -> ill_lock
369 370 *
370 371 * Trusted Solaris scenarios
371 372 *
372 373 * igsa_lock -> gcgrp_rwlock -> gcgrp_lock
373 374 * igsa_lock -> gcdb_lock
374 375 * gcgrp_rwlock -> ire_lock
375 376 * gcgrp_rwlock -> gcdb_lock
376 377 *
377 378 * squeue(sq_lock), flow related (ft_lock, fe_lock) locking
378 379 *
379 380 * cpu_lock --> ill_lock --> sqset_lock --> sq_lock
380 381 * sq_lock -> conn_lock -> QLOCK(q)
381 382 * ill_lock -> ft_lock -> fe_lock
382 383 *
383 384 * Routing/forwarding table locking notes:
384 385 *
385 386 * Lock acquisition order: Radix tree lock, irb_lock.
386 387 * Requirements:
387 388 * i. Walker must not hold any locks during the walker callback.
388 389 * ii Walker must not see a truncated tree during the walk because of any node
389 390 * deletion.
390 391 * iii Existing code assumes ire_bucket is valid if it is non-null and is used
391 392 * in many places in the code to walk the irb list. Thus even if all the
392 393 * ires in a bucket have been deleted, we still can't free the radix node
393 394 * until the ires have actually been inactive'd (freed).
394 395 *
395 396 * Tree traversal - Need to hold the global tree lock in read mode.
396 397 * Before dropping the global tree lock, need to either increment the ire_refcnt
397 398 * to ensure that the radix node can't be deleted.
398 399 *
399 400 * Tree add - Need to hold the global tree lock in write mode to add a
400 401 * radix node. To prevent the node from being deleted, increment the
401 402 * irb_refcnt, after the node is added to the tree. The ire itself is
402 403 * added later while holding the irb_lock, but not the tree lock.
403 404 *
404 405 * Tree delete - Need to hold the global tree lock and irb_lock in write mode.
405 406 * All associated ires must be inactive (i.e. freed), and irb_refcnt
406 407 * must be zero.
407 408 *
408 409 * Walker - Increment irb_refcnt before calling the walker callback. Hold the
409 410 * global tree lock (read mode) for traversal.
410 411 *
411 412 * IRE dependencies - In some cases we hold ips_ire_dep_lock across ire_refrele
412 413 * hence we will acquire irb_lock while holding ips_ire_dep_lock.
413 414 *
414 415 * IPsec notes :
415 416 *
416 417 * IP interacts with the IPsec code (AH/ESP) by storing IPsec attributes
417 418 * in the ip_xmit_attr_t ip_recv_attr_t. For outbound datagrams, the
418 419 * ip_xmit_attr_t has the
419 420 * information used by the IPsec code for applying the right level of
420 421 * protection. The information initialized by IP in the ip_xmit_attr_t
421 422 * is determined by the per-socket policy or global policy in the system.
422 423 * For inbound datagrams, the ip_recv_attr_t
423 424 * starts out with nothing in it. It gets filled
424 425 * with the right information if it goes through the AH/ESP code, which
425 426 * happens if the incoming packet is secure. The information initialized
426 427 * by AH/ESP, is later used by IP (during fanouts to ULP) to see whether
427 428 * the policy requirements needed by per-socket policy or global policy
428 429 * is met or not.
429 430 *
430 431 * For fully connected sockets i.e dst, src [addr, port] is known,
431 432 * conn_policy_cached is set indicating that policy has been cached.
432 433 * conn_in_enforce_policy may or may not be set depending on whether
433 434 * there is a global policy match or per-socket policy match.
434 435 * Policy inheriting happpens in ip_policy_set once the destination is known.
435 436 * Once the right policy is set on the conn_t, policy cannot change for
436 437 * this socket. This makes life simpler for TCP (UDP ?) where
437 438 * re-transmissions go out with the same policy. For symmetry, policy
438 439 * is cached for fully connected UDP sockets also. Thus if policy is cached,
439 440 * it also implies that policy is latched i.e policy cannot change
440 441 * on these sockets. As we have the right policy on the conn, we don't
441 442 * have to lookup global policy for every outbound and inbound datagram
442 443 * and thus serving as an optimization. Note that a global policy change
443 444 * does not affect fully connected sockets if they have policy. If fully
444 445 * connected sockets did not have any policy associated with it, global
445 446 * policy change may affect them.
446 447 *
447 448 * IP Flow control notes:
448 449 * ---------------------
449 450 * Non-TCP streams are flow controlled by IP. The way this is accomplished
450 451 * differs when ILL_CAPAB_DLD_DIRECT is enabled for that IP instance. When
451 452 * ILL_DIRECT_CAPABLE(ill) is TRUE, IP can do direct function calls into
452 453 * GLDv3. Otherwise packets are sent down to lower layers using STREAMS
453 454 * functions.
454 455 *
455 456 * Per Tx ring udp flow control:
456 457 * This is applicable only when ILL_CAPAB_DLD_DIRECT capability is set in
457 458 * the ill (i.e. ILL_DIRECT_CAPABLE(ill) is true).
458 459 *
459 460 * The underlying link can expose multiple Tx rings to the GLDv3 mac layer.
460 461 * To achieve best performance, outgoing traffic need to be fanned out among
461 462 * these Tx ring. mac_tx() is called (via str_mdata_fastpath_put()) to send
462 463 * traffic out of the NIC and it takes a fanout hint. UDP connections pass
463 464 * the address of connp as fanout hint to mac_tx(). Under flow controlled
464 465 * condition, mac_tx() returns a non-NULL cookie (ip_mac_tx_cookie_t). This
465 466 * cookie points to a specific Tx ring that is blocked. The cookie is used to
466 467 * hash into an idl_tx_list[] entry in idl_tx_list[] array. Each idl_tx_list_t
467 468 * point to drain_lists (idl_t's). These drain list will store the blocked UDP
468 469 * connp's. The drain list is not a single list but a configurable number of
469 470 * lists.
470 471 *
471 472 * The diagram below shows idl_tx_list_t's and their drain_lists. ip_stack_t
472 473 * has an array of idl_tx_list_t. The size of the array is TX_FANOUT_SIZE
473 474 * which is equal to 128. This array in turn contains a pointer to idl_t[],
474 475 * the ip drain list. The idl_t[] array size is MIN(max_ncpus, 8). The drain
475 476 * list will point to the list of connp's that are flow controlled.
476 477 *
477 478 * --------------- ------- ------- -------
478 479 * |->|drain_list[0]|-->|connp|-->|connp|-->|connp|-->
479 480 * | --------------- ------- ------- -------
480 481 * | --------------- ------- ------- -------
481 482 * |->|drain_list[1]|-->|connp|-->|connp|-->|connp|-->
482 483 * ---------------- | --------------- ------- ------- -------
483 484 * |idl_tx_list[0]|->| --------------- ------- ------- -------
484 485 * ---------------- |->|drain_list[2]|-->|connp|-->|connp|-->|connp|-->
485 486 * | --------------- ------- ------- -------
486 487 * . . . . .
487 488 * | --------------- ------- ------- -------
488 489 * |->|drain_list[n]|-->|connp|-->|connp|-->|connp|-->
489 490 * --------------- ------- ------- -------
490 491 * --------------- ------- ------- -------
491 492 * |->|drain_list[0]|-->|connp|-->|connp|-->|connp|-->
492 493 * | --------------- ------- ------- -------
493 494 * | --------------- ------- ------- -------
494 495 * ---------------- |->|drain_list[1]|-->|connp|-->|connp|-->|connp|-->
495 496 * |idl_tx_list[1]|->| --------------- ------- ------- -------
496 497 * ---------------- | . . . .
497 498 * | --------------- ------- ------- -------
498 499 * |->|drain_list[n]|-->|connp|-->|connp|-->|connp|-->
499 500 * --------------- ------- ------- -------
500 501 * .....
501 502 * ----------------
502 503 * |idl_tx_list[n]|-> ...
503 504 * ----------------
504 505 *
505 506 * When mac_tx() returns a cookie, the cookie is hashed into an index into
506 507 * ips_idl_tx_list[], and conn_drain_insert() is called with the idl_tx_list
507 508 * to insert the conn onto. conn_drain_insert() asserts flow control for the
508 509 * sockets via su_txq_full() (non-STREAMS) or QFULL on conn_wq (STREAMS).
509 510 * Further, conn_blocked is set to indicate that the conn is blocked.
510 511 *
511 512 * GLDv3 calls ill_flow_enable() when flow control is relieved. The cookie
512 513 * passed in the call to ill_flow_enable() identifies the blocked Tx ring and
513 514 * is again hashed to locate the appropriate idl_tx_list, which is then
514 515 * drained via conn_walk_drain(). conn_walk_drain() goes through each conn in
515 516 * the drain list and calls conn_drain_remove() to clear flow control (via
516 517 * calling su_txq_full() or clearing QFULL), and remove the conn from the
517 518 * drain list.
518 519 *
519 520 * Note that the drain list is not a single list but a (configurable) array of
520 521 * lists (8 elements by default). Synchronization between drain insertion and
521 522 * flow control wakeup is handled by using idl_txl->txl_lock, and only
522 523 * conn_drain_insert() and conn_drain_remove() manipulate the drain list.
523 524 *
524 525 * Flow control via STREAMS is used when ILL_DIRECT_CAPABLE() returns FALSE.
525 526 * On the send side, if the packet cannot be sent down to the driver by IP
526 527 * (canput() fails), ip_xmit() drops the packet and returns EWOULDBLOCK to the
527 528 * caller, who may then invoke ixa_check_drain_insert() to insert the conn on
528 529 * the 0'th drain list. When ip_wsrv() runs on the ill_wq because flow
529 530 * control has been relieved, the blocked conns in the 0'th drain list are
530 531 * drained as in the non-STREAMS case.
531 532 *
532 533 * In both the STREAMS and non-STREAMS cases, the sockfs upcall to set QFULL
533 534 * is done when the conn is inserted into the drain list (conn_drain_insert())
534 535 * and cleared when the conn is removed from the it (conn_drain_remove()).
535 536 *
536 537 * IPQOS notes:
537 538 *
538 539 * IPQoS Policies are applied to packets using IPPF (IP Policy framework)
539 540 * and IPQoS modules. IPPF includes hooks in IP at different control points
540 541 * (callout positions) which direct packets to IPQoS modules for policy
541 542 * processing. Policies, if present, are global.
542 543 *
543 544 * The callout positions are located in the following paths:
544 545 * o local_in (packets destined for this host)
545 546 * o local_out (packets orginating from this host )
546 547 * o fwd_in (packets forwarded by this m/c - inbound)
547 548 * o fwd_out (packets forwarded by this m/c - outbound)
548 549 * Hooks at these callout points can be enabled/disabled using the ndd variable
549 550 * ip_policy_mask (a bit mask with the 4 LSB indicating the callout positions).
550 551 * By default all the callout positions are enabled.
551 552 *
552 553 * Outbound (local_out)
553 554 * Hooks are placed in ire_send_wire_v4 and ire_send_wire_v6.
554 555 *
555 556 * Inbound (local_in)
556 557 * Hooks are placed in ip_fanout_v4 and ip_fanout_v6.
557 558 *
558 559 * Forwarding (in and out)
559 560 * Hooks are placed in ire_recv_forward_v4/v6.
560 561 *
561 562 * IP Policy Framework processing (IPPF processing)
562 563 * Policy processing for a packet is initiated by ip_process, which ascertains
563 564 * that the classifier (ipgpc) is loaded and configured, failing which the
564 565 * packet resumes normal processing in IP. If the clasifier is present, the
565 566 * packet is acted upon by one or more IPQoS modules (action instances), per
566 567 * filters configured in ipgpc and resumes normal IP processing thereafter.
567 568 * An action instance can drop a packet in course of its processing.
568 569 *
569 570 * Zones notes:
570 571 *
571 572 * The partitioning rules for networking are as follows:
572 573 * 1) Packets coming from a zone must have a source address belonging to that
573 574 * zone.
574 575 * 2) Packets coming from a zone can only be sent on a physical interface on
575 576 * which the zone has an IP address.
576 577 * 3) Between two zones on the same machine, packet delivery is only allowed if
577 578 * there's a matching route for the destination and zone in the forwarding
578 579 * table.
579 580 * 4) The TCP and UDP port spaces are per-zone; that is, two processes in
580 581 * different zones can bind to the same port with the wildcard address
581 582 * (INADDR_ANY).
582 583 *
583 584 * The granularity of interface partitioning is at the logical interface level.
584 585 * Therefore, every zone has its own IP addresses, and incoming packets can be
585 586 * attributed to a zone unambiguously. A logical interface is placed into a zone
586 587 * using the SIOCSLIFZONE ioctl; this sets the ipif_zoneid field in the ipif_t
587 588 * structure. Rule (1) is implemented by modifying the source address selection
588 589 * algorithm so that the list of eligible addresses is filtered based on the
589 590 * sending process zone.
590 591 *
591 592 * The Internet Routing Entries (IREs) are either exclusive to a zone or shared
592 593 * across all zones, depending on their type. Here is the break-up:
593 594 *
594 595 * IRE type Shared/exclusive
595 596 * -------- ----------------
596 597 * IRE_BROADCAST Exclusive
597 598 * IRE_DEFAULT (default routes) Shared (*)
598 599 * IRE_LOCAL Exclusive (x)
599 600 * IRE_LOOPBACK Exclusive
600 601 * IRE_PREFIX (net routes) Shared (*)
601 602 * IRE_IF_NORESOLVER (interface routes) Exclusive
602 603 * IRE_IF_RESOLVER (interface routes) Exclusive
603 604 * IRE_IF_CLONE (interface routes) Exclusive
604 605 * IRE_HOST (host routes) Shared (*)
605 606 *
606 607 * (*) A zone can only use a default or off-subnet route if the gateway is
607 608 * directly reachable from the zone, that is, if the gateway's address matches
608 609 * one of the zone's logical interfaces.
609 610 *
610 611 * (x) IRE_LOCAL are handled a bit differently.
611 612 * When ip_restrict_interzone_loopback is set (the default),
612 613 * ire_route_recursive restricts loopback using an IRE_LOCAL
613 614 * between zone to the case when L2 would have conceptually looped the packet
614 615 * back, i.e. the loopback which is required since neither Ethernet drivers
615 616 * nor Ethernet hardware loops them back. This is the case when the normal
616 617 * routes (ignoring IREs with different zoneids) would send out the packet on
617 618 * the same ill as the ill with which is IRE_LOCAL is associated.
618 619 *
619 620 * Multiple zones can share a common broadcast address; typically all zones
620 621 * share the 255.255.255.255 address. Incoming as well as locally originated
621 622 * broadcast packets must be dispatched to all the zones on the broadcast
622 623 * network. For directed broadcasts (e.g. 10.16.72.255) this is not trivial
623 624 * since some zones may not be on the 10.16.72/24 network. To handle this, each
624 625 * zone has its own set of IRE_BROADCAST entries; then, broadcast packets are
625 626 * sent to every zone that has an IRE_BROADCAST entry for the destination
626 627 * address on the input ill, see ip_input_broadcast().
627 628 *
628 629 * Applications in different zones can join the same multicast group address.
629 630 * The same logic applies for multicast as for broadcast. ip_input_multicast
630 631 * dispatches packets to all zones that have members on the physical interface.
631 632 */
632 633
633 634 /*
634 635 * Squeue Fanout flags:
635 636 * 0: No fanout.
636 637 * 1: Fanout across all squeues
637 638 */
638 639 boolean_t ip_squeue_fanout = 0;
639 640
640 641 /*
641 642 * Maximum dups allowed per packet.
642 643 */
643 644 uint_t ip_max_frag_dups = 10;
644 645
645 646 static int ip_open(queue_t *q, dev_t *devp, int flag, int sflag,
646 647 cred_t *credp, boolean_t isv6);
647 648 static mblk_t *ip_xmit_attach_llhdr(mblk_t *, nce_t *);
648 649
649 650 static boolean_t icmp_inbound_verify_v4(mblk_t *, icmph_t *, ip_recv_attr_t *);
650 651 static void icmp_inbound_too_big_v4(icmph_t *, ip_recv_attr_t *);
651 652 static void icmp_inbound_error_fanout_v4(mblk_t *, icmph_t *,
652 653 ip_recv_attr_t *);
653 654 static void icmp_options_update(ipha_t *);
654 655 static void icmp_param_problem(mblk_t *, uint8_t, ip_recv_attr_t *);
655 656 static void icmp_pkt(mblk_t *, void *, size_t, ip_recv_attr_t *);
656 657 static mblk_t *icmp_pkt_err_ok(mblk_t *, ip_recv_attr_t *);
657 658 static void icmp_redirect_v4(mblk_t *mp, ipha_t *, icmph_t *,
658 659 ip_recv_attr_t *);
659 660 static void icmp_send_redirect(mblk_t *, ipaddr_t, ip_recv_attr_t *);
660 661 static void icmp_send_reply_v4(mblk_t *, ipha_t *, icmph_t *,
661 662 ip_recv_attr_t *);
662 663
663 664 mblk_t *ip_dlpi_alloc(size_t, t_uscalar_t);
664 665 char *ip_dot_addr(ipaddr_t, char *);
665 666 mblk_t *ip_carve_mp(mblk_t **, ssize_t);
666 667 int ip_close(queue_t *, int);
667 668 static char *ip_dot_saddr(uchar_t *, char *);
668 669 static void ip_lrput(queue_t *, mblk_t *);
669 670 ipaddr_t ip_net_mask(ipaddr_t);
670 671 char *ip_nv_lookup(nv_t *, int);
671 672 void ip_rput(queue_t *, mblk_t *);
672 673 static void ip_rput_dlpi_writer(ipsq_t *dummy_sq, queue_t *q, mblk_t *mp,
673 674 void *dummy_arg);
674 675 int ip_snmp_get(queue_t *, mblk_t *, int, boolean_t);
675 676 static mblk_t *ip_snmp_get_mib2_ip(queue_t *, mblk_t *,
676 677 mib2_ipIfStatsEntry_t *, ip_stack_t *, boolean_t);
677 678 static mblk_t *ip_snmp_get_mib2_ip_traffic_stats(queue_t *, mblk_t *,
678 679 ip_stack_t *, boolean_t);
679 680 static mblk_t *ip_snmp_get_mib2_ip6(queue_t *, mblk_t *, ip_stack_t *,
680 681 boolean_t);
681 682 static mblk_t *ip_snmp_get_mib2_icmp(queue_t *, mblk_t *, ip_stack_t *ipst);
682 683 static mblk_t *ip_snmp_get_mib2_icmp6(queue_t *, mblk_t *, ip_stack_t *ipst);
683 684 static mblk_t *ip_snmp_get_mib2_igmp(queue_t *, mblk_t *, ip_stack_t *ipst);
684 685 static mblk_t *ip_snmp_get_mib2_multi(queue_t *, mblk_t *, ip_stack_t *ipst);
685 686 static mblk_t *ip_snmp_get_mib2_ip_addr(queue_t *, mblk_t *,
686 687 ip_stack_t *ipst, boolean_t);
687 688 static mblk_t *ip_snmp_get_mib2_ip6_addr(queue_t *, mblk_t *,
688 689 ip_stack_t *ipst, boolean_t);
689 690 static mblk_t *ip_snmp_get_mib2_ip_group_src(queue_t *, mblk_t *,
690 691 ip_stack_t *ipst);
691 692 static mblk_t *ip_snmp_get_mib2_ip6_group_src(queue_t *, mblk_t *,
692 693 ip_stack_t *ipst);
693 694 static mblk_t *ip_snmp_get_mib2_ip_group_mem(queue_t *, mblk_t *,
694 695 ip_stack_t *ipst);
695 696 static mblk_t *ip_snmp_get_mib2_ip6_group_mem(queue_t *, mblk_t *,
696 697 ip_stack_t *ipst);
697 698 static mblk_t *ip_snmp_get_mib2_virt_multi(queue_t *, mblk_t *,
698 699 ip_stack_t *ipst);
699 700 static mblk_t *ip_snmp_get_mib2_multi_rtable(queue_t *, mblk_t *,
700 701 ip_stack_t *ipst);
701 702 static mblk_t *ip_snmp_get_mib2_ip_route_media(queue_t *, mblk_t *, int,
702 703 ip_stack_t *ipst);
703 704 static mblk_t *ip_snmp_get_mib2_ip6_route_media(queue_t *, mblk_t *, int,
704 705 ip_stack_t *ipst);
705 706 static void ip_snmp_get2_v4(ire_t *, iproutedata_t *);
706 707 static void ip_snmp_get2_v6_route(ire_t *, iproutedata_t *);
707 708 static int ip_snmp_get2_v4_media(ncec_t *, iproutedata_t *);
708 709 static int ip_snmp_get2_v6_media(ncec_t *, iproutedata_t *);
709 710 int ip_snmp_set(queue_t *, int, int, uchar_t *, int);
710 711
711 712 static mblk_t *ip_fragment_copyhdr(uchar_t *, int, int, ip_stack_t *,
712 713 mblk_t *);
713 714
714 715 static void conn_drain_init(ip_stack_t *);
715 716 static void conn_drain_fini(ip_stack_t *);
716 717 static void conn_drain(conn_t *connp, boolean_t closing);
717 718
718 719 static void conn_walk_drain(ip_stack_t *, idl_tx_list_t *);
719 720 static void conn_walk_sctp(pfv_t, void *, zoneid_t, netstack_t *);
720 721
721 722 static void *ip_stack_init(netstackid_t stackid, netstack_t *ns);
722 723 static void ip_stack_shutdown(netstackid_t stackid, void *arg);
723 724 static void ip_stack_fini(netstackid_t stackid, void *arg);
724 725
725 726 static int ip_multirt_apply_membership(int (*fn)(conn_t *, boolean_t,
726 727 const in6_addr_t *, ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *),
727 728 ire_t *, conn_t *, boolean_t, const in6_addr_t *, mcast_record_t,
728 729 const in6_addr_t *);
729 730
730 731 static int ip_squeue_switch(int);
731 732
732 733 static void *ip_kstat_init(netstackid_t, ip_stack_t *);
733 734 static void ip_kstat_fini(netstackid_t, kstat_t *);
734 735 static int ip_kstat_update(kstat_t *kp, int rw);
735 736 static void *icmp_kstat_init(netstackid_t);
736 737 static void icmp_kstat_fini(netstackid_t, kstat_t *);
737 738 static int icmp_kstat_update(kstat_t *kp, int rw);
738 739 static void *ip_kstat2_init(netstackid_t, ip_stat_t *);
739 740 static void ip_kstat2_fini(netstackid_t, kstat_t *);
740 741
741 742 static void ipobs_init(ip_stack_t *);
742 743 static void ipobs_fini(ip_stack_t *);
743 744
744 745 static int ip_tp_cpu_update(cpu_setup_t, int, void *);
745 746
746 747 ipaddr_t ip_g_all_ones = IP_HOST_MASK;
747 748
748 749 static long ip_rput_pullups;
749 750 int dohwcksum = 1; /* use h/w cksum if supported by the hardware */
750 751
751 752 vmem_t *ip_minor_arena_sa; /* for minor nos. from INET_MIN_DEV+2 thru 2^^18-1 */
752 753 vmem_t *ip_minor_arena_la; /* for minor nos. from 2^^18 thru 2^^32-1 */
753 754
754 755 int ip_debug;
755 756
756 757 /*
757 758 * Multirouting/CGTP stuff
758 759 */
759 760 int ip_cgtp_filter_rev = CGTP_FILTER_REV; /* CGTP hooks version */
760 761
761 762 /*
762 763 * IP tunables related declarations. Definitions are in ip_tunables.c
763 764 */
764 765 extern mod_prop_info_t ip_propinfo_tbl[];
765 766 extern int ip_propinfo_count;
766 767
767 768 /*
768 769 * Table of IP ioctls encoding the various properties of the ioctl and
769 770 * indexed based on the last byte of the ioctl command. Occasionally there
770 771 * is a clash, and there is more than 1 ioctl with the same last byte.
771 772 * In such a case 1 ioctl is encoded in the ndx table and the remaining
772 773 * ioctls are encoded in the misc table. An entry in the ndx table is
773 774 * retrieved by indexing on the last byte of the ioctl command and comparing
774 775 * the ioctl command with the value in the ndx table. In the event of a
775 776 * mismatch the misc table is then searched sequentially for the desired
776 777 * ioctl command.
777 778 *
778 779 * Entry: <command> <copyin_size> <flags> <cmd_type> <function> <restart_func>
779 780 */
780 781 ip_ioctl_cmd_t ip_ndx_ioctl_table[] = {
781 782 /* 000 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
782 783 /* 001 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
783 784 /* 002 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
784 785 /* 003 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
785 786 /* 004 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
786 787 /* 005 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
787 788 /* 006 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
788 789 /* 007 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
789 790 /* 008 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
790 791 /* 009 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
791 792
792 793 /* 010 */ { SIOCADDRT, sizeof (struct rtentry), IPI_PRIV,
793 794 MISC_CMD, ip_siocaddrt, NULL },
794 795 /* 011 */ { SIOCDELRT, sizeof (struct rtentry), IPI_PRIV,
795 796 MISC_CMD, ip_siocdelrt, NULL },
796 797
797 798 /* 012 */ { SIOCSIFADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
798 799 IF_CMD, ip_sioctl_addr, ip_sioctl_addr_restart },
799 800 /* 013 */ { SIOCGIFADDR, sizeof (struct ifreq), IPI_GET_CMD,
800 801 IF_CMD, ip_sioctl_get_addr, NULL },
801 802
802 803 /* 014 */ { SIOCSIFDSTADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
803 804 IF_CMD, ip_sioctl_dstaddr, ip_sioctl_dstaddr_restart },
804 805 /* 015 */ { SIOCGIFDSTADDR, sizeof (struct ifreq),
805 806 IPI_GET_CMD, IF_CMD, ip_sioctl_get_dstaddr, NULL },
806 807
807 808 /* 016 */ { SIOCSIFFLAGS, sizeof (struct ifreq),
808 809 IPI_PRIV | IPI_WR,
809 810 IF_CMD, ip_sioctl_flags, ip_sioctl_flags_restart },
810 811 /* 017 */ { SIOCGIFFLAGS, sizeof (struct ifreq),
811 812 IPI_MODOK | IPI_GET_CMD,
812 813 IF_CMD, ip_sioctl_get_flags, NULL },
813 814
814 815 /* 018 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
815 816 /* 019 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
816 817
817 818 /* copyin size cannot be coded for SIOCGIFCONF */
818 819 /* 020 */ { O_SIOCGIFCONF, 0, IPI_GET_CMD,
819 820 MISC_CMD, ip_sioctl_get_ifconf, NULL },
820 821
821 822 /* 021 */ { SIOCSIFMTU, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
822 823 IF_CMD, ip_sioctl_mtu, NULL },
823 824 /* 022 */ { SIOCGIFMTU, sizeof (struct ifreq), IPI_GET_CMD,
824 825 IF_CMD, ip_sioctl_get_mtu, NULL },
825 826 /* 023 */ { SIOCGIFBRDADDR, sizeof (struct ifreq),
826 827 IPI_GET_CMD, IF_CMD, ip_sioctl_get_brdaddr, NULL },
827 828 /* 024 */ { SIOCSIFBRDADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
828 829 IF_CMD, ip_sioctl_brdaddr, NULL },
829 830 /* 025 */ { SIOCGIFNETMASK, sizeof (struct ifreq),
830 831 IPI_GET_CMD, IF_CMD, ip_sioctl_get_netmask, NULL },
831 832 /* 026 */ { SIOCSIFNETMASK, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
832 833 IF_CMD, ip_sioctl_netmask, ip_sioctl_netmask_restart },
833 834 /* 027 */ { SIOCGIFMETRIC, sizeof (struct ifreq),
834 835 IPI_GET_CMD, IF_CMD, ip_sioctl_get_metric, NULL },
835 836 /* 028 */ { SIOCSIFMETRIC, sizeof (struct ifreq), IPI_PRIV,
836 837 IF_CMD, ip_sioctl_metric, NULL },
837 838 /* 029 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
838 839
839 840 /* See 166-168 below for extended SIOC*XARP ioctls */
840 841 /* 030 */ { SIOCSARP, sizeof (struct arpreq), IPI_PRIV | IPI_WR,
841 842 ARP_CMD, ip_sioctl_arp, NULL },
842 843 /* 031 */ { SIOCGARP, sizeof (struct arpreq), IPI_GET_CMD,
843 844 ARP_CMD, ip_sioctl_arp, NULL },
844 845 /* 032 */ { SIOCDARP, sizeof (struct arpreq), IPI_PRIV | IPI_WR,
845 846 ARP_CMD, ip_sioctl_arp, NULL },
846 847
847 848 /* 033 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
848 849 /* 034 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
849 850 /* 035 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
850 851 /* 036 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
851 852 /* 037 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
852 853 /* 038 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
853 854 /* 039 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
854 855 /* 040 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
855 856 /* 041 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
856 857 /* 042 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
857 858 /* 043 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
858 859 /* 044 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
859 860 /* 045 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
860 861 /* 046 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
861 862 /* 047 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
862 863 /* 048 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
863 864 /* 049 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
864 865 /* 050 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
865 866 /* 051 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
866 867 /* 052 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
867 868 /* 053 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
868 869
869 870 /* 054 */ { IF_UNITSEL, sizeof (int), IPI_PRIV | IPI_WR | IPI_MODOK,
870 871 MISC_CMD, if_unitsel, if_unitsel_restart },
871 872
872 873 /* 055 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
873 874 /* 056 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
874 875 /* 057 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
875 876 /* 058 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
876 877 /* 059 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
877 878 /* 060 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
878 879 /* 061 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
879 880 /* 062 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
880 881 /* 063 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
881 882 /* 064 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
882 883 /* 065 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
883 884 /* 066 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
884 885 /* 067 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
885 886 /* 068 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
886 887 /* 069 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
887 888 /* 070 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
888 889 /* 071 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
889 890 /* 072 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
890 891
891 892 /* 073 */ { SIOCSIFNAME, sizeof (struct ifreq),
892 893 IPI_PRIV | IPI_WR | IPI_MODOK,
893 894 IF_CMD, ip_sioctl_sifname, NULL },
894 895
895 896 /* 074 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
896 897 /* 075 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
897 898 /* 076 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
898 899 /* 077 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
899 900 /* 078 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
900 901 /* 079 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
901 902 /* 080 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
902 903 /* 081 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
903 904 /* 082 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
904 905 /* 083 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
905 906 /* 084 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
906 907 /* 085 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
907 908 /* 086 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
908 909
909 910 /* 087 */ { SIOCGIFNUM, sizeof (int), IPI_GET_CMD,
910 911 MISC_CMD, ip_sioctl_get_ifnum, NULL },
911 912 /* 088 */ { SIOCGIFMUXID, sizeof (struct ifreq), IPI_GET_CMD,
912 913 IF_CMD, ip_sioctl_get_muxid, NULL },
913 914 /* 089 */ { SIOCSIFMUXID, sizeof (struct ifreq),
914 915 IPI_PRIV | IPI_WR, IF_CMD, ip_sioctl_muxid, NULL },
915 916
916 917 /* Both if and lif variants share same func */
917 918 /* 090 */ { SIOCGIFINDEX, sizeof (struct ifreq), IPI_GET_CMD,
918 919 IF_CMD, ip_sioctl_get_lifindex, NULL },
919 920 /* Both if and lif variants share same func */
920 921 /* 091 */ { SIOCSIFINDEX, sizeof (struct ifreq),
921 922 IPI_PRIV | IPI_WR, IF_CMD, ip_sioctl_slifindex, NULL },
922 923
923 924 /* copyin size cannot be coded for SIOCGIFCONF */
924 925 /* 092 */ { SIOCGIFCONF, 0, IPI_GET_CMD,
925 926 MISC_CMD, ip_sioctl_get_ifconf, NULL },
926 927 /* 093 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
927 928 /* 094 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
928 929 /* 095 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
929 930 /* 096 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
930 931 /* 097 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
931 932 /* 098 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
932 933 /* 099 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
933 934 /* 100 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
934 935 /* 101 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
935 936 /* 102 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
936 937 /* 103 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
937 938 /* 104 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
938 939 /* 105 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
939 940 /* 106 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
940 941 /* 107 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
941 942 /* 108 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
942 943 /* 109 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
943 944
944 945 /* 110 */ { SIOCLIFREMOVEIF, sizeof (struct lifreq),
945 946 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_removeif,
946 947 ip_sioctl_removeif_restart },
947 948 /* 111 */ { SIOCLIFADDIF, sizeof (struct lifreq),
948 949 IPI_GET_CMD | IPI_PRIV | IPI_WR,
949 950 LIF_CMD, ip_sioctl_addif, NULL },
950 951 #define SIOCLIFADDR_NDX 112
951 952 /* 112 */ { SIOCSLIFADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
952 953 LIF_CMD, ip_sioctl_addr, ip_sioctl_addr_restart },
953 954 /* 113 */ { SIOCGLIFADDR, sizeof (struct lifreq),
954 955 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_addr, NULL },
955 956 /* 114 */ { SIOCSLIFDSTADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
956 957 LIF_CMD, ip_sioctl_dstaddr, ip_sioctl_dstaddr_restart },
957 958 /* 115 */ { SIOCGLIFDSTADDR, sizeof (struct lifreq),
958 959 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_dstaddr, NULL },
959 960 /* 116 */ { SIOCSLIFFLAGS, sizeof (struct lifreq),
960 961 IPI_PRIV | IPI_WR,
961 962 LIF_CMD, ip_sioctl_flags, ip_sioctl_flags_restart },
962 963 /* 117 */ { SIOCGLIFFLAGS, sizeof (struct lifreq),
963 964 IPI_GET_CMD | IPI_MODOK,
964 965 LIF_CMD, ip_sioctl_get_flags, NULL },
965 966
966 967 /* 118 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
967 968 /* 119 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
968 969
969 970 /* 120 */ { O_SIOCGLIFCONF, 0, IPI_GET_CMD, MISC_CMD,
970 971 ip_sioctl_get_lifconf, NULL },
971 972 /* 121 */ { SIOCSLIFMTU, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
972 973 LIF_CMD, ip_sioctl_mtu, NULL },
973 974 /* 122 */ { SIOCGLIFMTU, sizeof (struct lifreq), IPI_GET_CMD,
974 975 LIF_CMD, ip_sioctl_get_mtu, NULL },
975 976 /* 123 */ { SIOCGLIFBRDADDR, sizeof (struct lifreq),
976 977 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_brdaddr, NULL },
977 978 /* 124 */ { SIOCSLIFBRDADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
978 979 LIF_CMD, ip_sioctl_brdaddr, NULL },
979 980 /* 125 */ { SIOCGLIFNETMASK, sizeof (struct lifreq),
980 981 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_netmask, NULL },
981 982 /* 126 */ { SIOCSLIFNETMASK, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
982 983 LIF_CMD, ip_sioctl_netmask, ip_sioctl_netmask_restart },
983 984 /* 127 */ { SIOCGLIFMETRIC, sizeof (struct lifreq),
984 985 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_metric, NULL },
985 986 /* 128 */ { SIOCSLIFMETRIC, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
986 987 LIF_CMD, ip_sioctl_metric, NULL },
987 988 /* 129 */ { SIOCSLIFNAME, sizeof (struct lifreq),
988 989 IPI_PRIV | IPI_WR | IPI_MODOK,
989 990 LIF_CMD, ip_sioctl_slifname,
990 991 ip_sioctl_slifname_restart },
991 992
992 993 /* 130 */ { SIOCGLIFNUM, sizeof (struct lifnum), IPI_GET_CMD,
993 994 MISC_CMD, ip_sioctl_get_lifnum, NULL },
994 995 /* 131 */ { SIOCGLIFMUXID, sizeof (struct lifreq),
995 996 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_muxid, NULL },
996 997 /* 132 */ { SIOCSLIFMUXID, sizeof (struct lifreq),
997 998 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_muxid, NULL },
998 999 /* 133 */ { SIOCGLIFINDEX, sizeof (struct lifreq),
999 1000 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lifindex, 0 },
1000 1001 /* 134 */ { SIOCSLIFINDEX, sizeof (struct lifreq),
1001 1002 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_slifindex, 0 },
1002 1003 /* 135 */ { SIOCSLIFTOKEN, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1003 1004 LIF_CMD, ip_sioctl_token, NULL },
1004 1005 /* 136 */ { SIOCGLIFTOKEN, sizeof (struct lifreq),
1005 1006 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_token, NULL },
1006 1007 /* 137 */ { SIOCSLIFSUBNET, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1007 1008 LIF_CMD, ip_sioctl_subnet, ip_sioctl_subnet_restart },
1008 1009 /* 138 */ { SIOCGLIFSUBNET, sizeof (struct lifreq),
1009 1010 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_subnet, NULL },
1010 1011 /* 139 */ { SIOCSLIFLNKINFO, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1011 1012 LIF_CMD, ip_sioctl_lnkinfo, NULL },
1012 1013
1013 1014 /* 140 */ { SIOCGLIFLNKINFO, sizeof (struct lifreq),
1014 1015 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lnkinfo, NULL },
1015 1016 /* 141 */ { SIOCLIFDELND, sizeof (struct lifreq), IPI_PRIV,
1016 1017 LIF_CMD, ip_siocdelndp_v6, NULL },
1017 1018 /* 142 */ { SIOCLIFGETND, sizeof (struct lifreq), IPI_GET_CMD,
1018 1019 LIF_CMD, ip_siocqueryndp_v6, NULL },
1019 1020 /* 143 */ { SIOCLIFSETND, sizeof (struct lifreq), IPI_PRIV,
1020 1021 LIF_CMD, ip_siocsetndp_v6, NULL },
1021 1022 /* 144 */ { SIOCTMYADDR, sizeof (struct sioc_addrreq), IPI_GET_CMD,
1022 1023 MISC_CMD, ip_sioctl_tmyaddr, NULL },
1023 1024 /* 145 */ { SIOCTONLINK, sizeof (struct sioc_addrreq), IPI_GET_CMD,
1024 1025 MISC_CMD, ip_sioctl_tonlink, NULL },
1025 1026 /* 146 */ { SIOCTMYSITE, sizeof (struct sioc_addrreq), 0,
1026 1027 MISC_CMD, ip_sioctl_tmysite, NULL },
1027 1028 /* 147 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1028 1029 /* 148 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1029 1030 /* IPSECioctls handled in ip_sioctl_copyin_setup itself */
1030 1031 /* 149 */ { SIOCFIPSECONFIG, 0, IPI_PRIV, MISC_CMD, NULL, NULL },
1031 1032 /* 150 */ { SIOCSIPSECONFIG, 0, IPI_PRIV, MISC_CMD, NULL, NULL },
1032 1033 /* 151 */ { SIOCDIPSECONFIG, 0, IPI_PRIV, MISC_CMD, NULL, NULL },
1033 1034 /* 152 */ { SIOCLIPSECONFIG, 0, IPI_PRIV, MISC_CMD, NULL, NULL },
1034 1035
1035 1036 /* 153 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1036 1037
1037 1038 /* 154 */ { SIOCGLIFBINDING, sizeof (struct lifreq), IPI_GET_CMD,
1038 1039 LIF_CMD, ip_sioctl_get_binding, NULL },
1039 1040 /* 155 */ { SIOCSLIFGROUPNAME, sizeof (struct lifreq),
1040 1041 IPI_PRIV | IPI_WR,
1041 1042 LIF_CMD, ip_sioctl_groupname, ip_sioctl_groupname },
1042 1043 /* 156 */ { SIOCGLIFGROUPNAME, sizeof (struct lifreq),
1043 1044 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_groupname, NULL },
1044 1045 /* 157 */ { SIOCGLIFGROUPINFO, sizeof (lifgroupinfo_t),
1045 1046 IPI_GET_CMD, MISC_CMD, ip_sioctl_groupinfo, NULL },
1046 1047
1047 1048 /* Leave 158-160 unused; used to be SIOC*IFARP ioctls */
1048 1049 /* 158 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1049 1050 /* 159 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1050 1051 /* 160 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1051 1052
1052 1053 /* 161 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1053 1054
1054 1055 /* These are handled in ip_sioctl_copyin_setup itself */
1055 1056 /* 162 */ { SIOCGIP6ADDRPOLICY, 0, IPI_NULL_BCONT,
1056 1057 MISC_CMD, NULL, NULL },
1057 1058 /* 163 */ { SIOCSIP6ADDRPOLICY, 0, IPI_PRIV | IPI_NULL_BCONT,
1058 1059 MISC_CMD, NULL, NULL },
1059 1060 /* 164 */ { SIOCGDSTINFO, 0, IPI_GET_CMD, MISC_CMD, NULL, NULL },
1060 1061
1061 1062 /* 165 */ { SIOCGLIFCONF, 0, IPI_GET_CMD, MISC_CMD,
1062 1063 ip_sioctl_get_lifconf, NULL },
1063 1064
1064 1065 /* 166 */ { SIOCSXARP, sizeof (struct xarpreq), IPI_PRIV | IPI_WR,
1065 1066 XARP_CMD, ip_sioctl_arp, NULL },
1066 1067 /* 167 */ { SIOCGXARP, sizeof (struct xarpreq), IPI_GET_CMD,
1067 1068 XARP_CMD, ip_sioctl_arp, NULL },
1068 1069 /* 168 */ { SIOCDXARP, sizeof (struct xarpreq), IPI_PRIV | IPI_WR,
1069 1070 XARP_CMD, ip_sioctl_arp, NULL },
1070 1071
1071 1072 /* SIOCPOPSOCKFS is not handled by IP */
1072 1073 /* 169 */ { IPI_DONTCARE /* SIOCPOPSOCKFS */, 0, 0, 0, NULL, NULL },
1073 1074
1074 1075 /* 170 */ { SIOCGLIFZONE, sizeof (struct lifreq),
1075 1076 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lifzone, NULL },
1076 1077 /* 171 */ { SIOCSLIFZONE, sizeof (struct lifreq),
1077 1078 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_slifzone,
1078 1079 ip_sioctl_slifzone_restart },
1079 1080 /* 172-174 are SCTP ioctls and not handled by IP */
1080 1081 /* 172 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1081 1082 /* 173 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1082 1083 /* 174 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1083 1084 /* 175 */ { SIOCGLIFUSESRC, sizeof (struct lifreq),
1084 1085 IPI_GET_CMD, LIF_CMD,
1085 1086 ip_sioctl_get_lifusesrc, 0 },
1086 1087 /* 176 */ { SIOCSLIFUSESRC, sizeof (struct lifreq),
1087 1088 IPI_PRIV | IPI_WR,
1088 1089 LIF_CMD, ip_sioctl_slifusesrc,
1089 1090 NULL },
1090 1091 /* 177 */ { SIOCGLIFSRCOF, 0, IPI_GET_CMD, MISC_CMD,
1091 1092 ip_sioctl_get_lifsrcof, NULL },
1092 1093 /* 178 */ { SIOCGMSFILTER, sizeof (struct group_filter), IPI_GET_CMD,
1093 1094 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1094 1095 /* 179 */ { SIOCSMSFILTER, sizeof (struct group_filter), 0,
1095 1096 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1096 1097 /* 180 */ { SIOCGIPMSFILTER, sizeof (struct ip_msfilter), IPI_GET_CMD,
1097 1098 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1098 1099 /* 181 */ { SIOCSIPMSFILTER, sizeof (struct ip_msfilter), 0,
1099 1100 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1100 1101 /* 182 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1101 1102 /* SIOCSENABLESDP is handled by SDP */
1102 1103 /* 183 */ { IPI_DONTCARE /* SIOCSENABLESDP */, 0, 0, 0, NULL, NULL },
1103 1104 /* 184 */ { IPI_DONTCARE /* SIOCSQPTR */, 0, 0, 0, NULL, NULL },
1104 1105 /* 185 */ { SIOCGIFHWADDR, sizeof (struct ifreq), IPI_GET_CMD,
1105 1106 IF_CMD, ip_sioctl_get_ifhwaddr, NULL },
1106 1107 /* 186 */ { IPI_DONTCARE /* SIOCGSTAMP */, 0, 0, 0, NULL, NULL },
1107 1108 /* 187 */ { SIOCILB, 0, IPI_PRIV | IPI_GET_CMD, MISC_CMD,
1108 1109 ip_sioctl_ilb_cmd, NULL },
1109 1110 /* 188 */ { SIOCGETPROP, 0, IPI_GET_CMD, 0, NULL, NULL },
1110 1111 /* 189 */ { SIOCSETPROP, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL},
1111 1112 /* 190 */ { SIOCGLIFDADSTATE, sizeof (struct lifreq),
1112 1113 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_dadstate, NULL },
1113 1114 /* 191 */ { SIOCSLIFPREFIX, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1114 1115 LIF_CMD, ip_sioctl_prefix, ip_sioctl_prefix_restart },
1115 1116 /* 192 */ { SIOCGLIFHWADDR, sizeof (struct lifreq), IPI_GET_CMD,
1116 1117 LIF_CMD, ip_sioctl_get_lifhwaddr, NULL }
1117 1118 };
1118 1119
1119 1120 int ip_ndx_ioctl_count = sizeof (ip_ndx_ioctl_table) / sizeof (ip_ioctl_cmd_t);
1120 1121
1121 1122 ip_ioctl_cmd_t ip_misc_ioctl_table[] = {
1122 1123 { I_LINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1123 1124 { I_UNLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1124 1125 { I_PLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1125 1126 { I_PUNLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1126 1127 { ND_GET, 0, 0, 0, NULL, NULL },
1127 1128 { ND_SET, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1128 1129 { IP_IOCTL, 0, 0, 0, NULL, NULL },
1129 1130 { SIOCGETVIFCNT, sizeof (struct sioc_vif_req), IPI_GET_CMD,
1130 1131 MISC_CMD, mrt_ioctl},
1131 1132 { SIOCGETSGCNT, sizeof (struct sioc_sg_req), IPI_GET_CMD,
1132 1133 MISC_CMD, mrt_ioctl},
1133 1134 { SIOCGETLSGCNT, sizeof (struct sioc_lsg_req), IPI_GET_CMD,
1134 1135 MISC_CMD, mrt_ioctl}
1135 1136 };
1136 1137
1137 1138 int ip_misc_ioctl_count =
1138 1139 sizeof (ip_misc_ioctl_table) / sizeof (ip_ioctl_cmd_t);
1139 1140
1140 1141 int conn_drain_nthreads; /* Number of drainers reqd. */
1141 1142 /* Settable in /etc/system */
1142 1143 /* Defined in ip_ire.c */
1143 1144 extern uint32_t ip_ire_max_bucket_cnt, ip6_ire_max_bucket_cnt;
1144 1145 extern uint32_t ip_ire_min_bucket_cnt, ip6_ire_min_bucket_cnt;
1145 1146 extern uint32_t ip_ire_mem_ratio, ip_ire_cpu_ratio;
1146 1147
1147 1148 static nv_t ire_nv_arr[] = {
1148 1149 { IRE_BROADCAST, "BROADCAST" },
1149 1150 { IRE_LOCAL, "LOCAL" },
1150 1151 { IRE_LOOPBACK, "LOOPBACK" },
1151 1152 { IRE_DEFAULT, "DEFAULT" },
1152 1153 { IRE_PREFIX, "PREFIX" },
1153 1154 { IRE_IF_NORESOLVER, "IF_NORESOL" },
1154 1155 { IRE_IF_RESOLVER, "IF_RESOLV" },
1155 1156 { IRE_IF_CLONE, "IF_CLONE" },
1156 1157 { IRE_HOST, "HOST" },
1157 1158 { IRE_MULTICAST, "MULTICAST" },
1158 1159 { IRE_NOROUTE, "NOROUTE" },
1159 1160 { 0 }
1160 1161 };
1161 1162
1162 1163 nv_t *ire_nv_tbl = ire_nv_arr;
1163 1164
1164 1165 /* Simple ICMP IP Header Template */
1165 1166 static ipha_t icmp_ipha = {
1166 1167 IP_SIMPLE_HDR_VERSION, 0, 0, 0, 0, 0, IPPROTO_ICMP
1167 1168 };
1168 1169
1169 1170 struct module_info ip_mod_info = {
1170 1171 IP_MOD_ID, IP_MOD_NAME, IP_MOD_MINPSZ, IP_MOD_MAXPSZ, IP_MOD_HIWAT,
1171 1172 IP_MOD_LOWAT
1172 1173 };
1173 1174
1174 1175 /*
1175 1176 * Duplicate static symbols within a module confuses mdb; so we avoid the
1176 1177 * problem by making the symbols here distinct from those in udp.c.
1177 1178 */
1178 1179
1179 1180 /*
1180 1181 * Entry points for IP as a device and as a module.
1181 1182 * We have separate open functions for the /dev/ip and /dev/ip6 devices.
1182 1183 */
1183 1184 static struct qinit iprinitv4 = {
1184 1185 (pfi_t)ip_rput, NULL, ip_openv4, ip_close, NULL,
1185 1186 &ip_mod_info
1186 1187 };
1187 1188
1188 1189 struct qinit iprinitv6 = {
1189 1190 (pfi_t)ip_rput_v6, NULL, ip_openv6, ip_close, NULL,
1190 1191 &ip_mod_info
1191 1192 };
1192 1193
1193 1194 static struct qinit ipwinit = {
1194 1195 (pfi_t)ip_wput_nondata, (pfi_t)ip_wsrv, NULL, NULL, NULL,
1195 1196 &ip_mod_info
1196 1197 };
1197 1198
1198 1199 static struct qinit iplrinit = {
1199 1200 (pfi_t)ip_lrput, NULL, ip_openv4, ip_close, NULL,
1200 1201 &ip_mod_info
1201 1202 };
1202 1203
1203 1204 static struct qinit iplwinit = {
1204 1205 (pfi_t)ip_lwput, NULL, NULL, NULL, NULL,
1205 1206 &ip_mod_info
1206 1207 };
1207 1208
1208 1209 /* For AF_INET aka /dev/ip */
1209 1210 struct streamtab ipinfov4 = {
1210 1211 &iprinitv4, &ipwinit, &iplrinit, &iplwinit
1211 1212 };
1212 1213
1213 1214 /* For AF_INET6 aka /dev/ip6 */
1214 1215 struct streamtab ipinfov6 = {
1215 1216 &iprinitv6, &ipwinit, &iplrinit, &iplwinit
1216 1217 };
1217 1218
1218 1219 #ifdef DEBUG
1219 1220 boolean_t skip_sctp_cksum = B_FALSE;
1220 1221 #endif
1221 1222
1222 1223 /*
1223 1224 * Generate an ICMP fragmentation needed message.
1224 1225 * When called from ip_output side a minimal ip_recv_attr_t needs to be
1225 1226 * constructed by the caller.
1226 1227 */
1227 1228 void
1228 1229 icmp_frag_needed(mblk_t *mp, int mtu, ip_recv_attr_t *ira)
1229 1230 {
1230 1231 icmph_t icmph;
1231 1232 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
1232 1233
1233 1234 mp = icmp_pkt_err_ok(mp, ira);
1234 1235 if (mp == NULL)
1235 1236 return;
1236 1237
1237 1238 bzero(&icmph, sizeof (icmph_t));
1238 1239 icmph.icmph_type = ICMP_DEST_UNREACHABLE;
1239 1240 icmph.icmph_code = ICMP_FRAGMENTATION_NEEDED;
1240 1241 icmph.icmph_du_mtu = htons((uint16_t)mtu);
1241 1242 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutFragNeeded);
1242 1243 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDestUnreachs);
1243 1244
1244 1245 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
1245 1246 }
1246 1247
1247 1248 /*
1248 1249 * icmp_inbound_v4 deals with ICMP messages that are handled by IP.
1249 1250 * If the ICMP message is consumed by IP, i.e., it should not be delivered
1250 1251 * to any IPPROTO_ICMP raw sockets, then it returns NULL.
1251 1252 * Likewise, if the ICMP error is misformed (too short, etc), then it
1252 1253 * returns NULL. The caller uses this to determine whether or not to send
1253 1254 * to raw sockets.
1254 1255 *
1255 1256 * All error messages are passed to the matching transport stream.
1256 1257 *
1257 1258 * The following cases are handled by icmp_inbound:
1258 1259 * 1) It needs to send a reply back and possibly delivering it
1259 1260 * to the "interested" upper clients.
1260 1261 * 2) Return the mblk so that the caller can pass it to the RAW socket clients.
1261 1262 * 3) It needs to change some values in IP only.
1262 1263 * 4) It needs to change some values in IP and upper layers e.g TCP
1263 1264 * by delivering an error to the upper layers.
1264 1265 *
1265 1266 * We handle the above three cases in the context of IPsec in the
1266 1267 * following way :
1267 1268 *
1268 1269 * 1) Send the reply back in the same way as the request came in.
1269 1270 * If it came in encrypted, it goes out encrypted. If it came in
1270 1271 * clear, it goes out in clear. Thus, this will prevent chosen
1271 1272 * plain text attack.
1272 1273 * 2) The client may or may not expect things to come in secure.
1273 1274 * If it comes in secure, the policy constraints are checked
1274 1275 * before delivering it to the upper layers. If it comes in
1275 1276 * clear, ipsec_inbound_accept_clear will decide whether to
1276 1277 * accept this in clear or not. In both the cases, if the returned
1277 1278 * message (IP header + 8 bytes) that caused the icmp message has
1278 1279 * AH/ESP headers, it is sent up to AH/ESP for validation before
1279 1280 * sending up. If there are only 8 bytes of returned message, then
1280 1281 * upper client will not be notified.
1281 1282 * 3) Check with global policy to see whether it matches the constaints.
1282 1283 * But this will be done only if icmp_accept_messages_in_clear is
1283 1284 * zero.
1284 1285 * 4) If we need to change both in IP and ULP, then the decision taken
1285 1286 * while affecting the values in IP and while delivering up to TCP
1286 1287 * should be the same.
1287 1288 *
1288 1289 * There are two cases.
1289 1290 *
1290 1291 * a) If we reject data at the IP layer (ipsec_check_global_policy()
1291 1292 * failed), we will not deliver it to the ULP, even though they
1292 1293 * are *willing* to accept in *clear*. This is fine as our global
1293 1294 * disposition to icmp messages asks us reject the datagram.
1294 1295 *
1295 1296 * b) If we accept data at the IP layer (ipsec_check_global_policy()
1296 1297 * succeeded or icmp_accept_messages_in_clear is 1), and not able
1297 1298 * to deliver it to ULP (policy failed), it can lead to
1298 1299 * consistency problems. The cases known at this time are
1299 1300 * ICMP_DESTINATION_UNREACHABLE messages with following code
1300 1301 * values :
1301 1302 *
1302 1303 * - ICMP_FRAGMENTATION_NEEDED : IP adapts to the new value
1303 1304 * and Upper layer rejects. Then the communication will
1304 1305 * come to a stop. This is solved by making similar decisions
1305 1306 * at both levels. Currently, when we are unable to deliver
1306 1307 * to the Upper Layer (due to policy failures) while IP has
1307 1308 * adjusted dce_pmtu, the next outbound datagram would
1308 1309 * generate a local ICMP_FRAGMENTATION_NEEDED message - which
1309 1310 * will be with the right level of protection. Thus the right
1310 1311 * value will be communicated even if we are not able to
1311 1312 * communicate when we get from the wire initially. But this
1312 1313 * assumes there would be at least one outbound datagram after
1313 1314 * IP has adjusted its dce_pmtu value. To make things
1314 1315 * simpler, we accept in clear after the validation of
1315 1316 * AH/ESP headers.
1316 1317 *
1317 1318 * - Other ICMP ERRORS : We may not be able to deliver it to the
1318 1319 * upper layer depending on the level of protection the upper
1319 1320 * layer expects and the disposition in ipsec_inbound_accept_clear().
1320 1321 * ipsec_inbound_accept_clear() decides whether a given ICMP error
1321 1322 * should be accepted in clear when the Upper layer expects secure.
1322 1323 * Thus the communication may get aborted by some bad ICMP
1323 1324 * packets.
1324 1325 */
1325 1326 mblk_t *
1326 1327 icmp_inbound_v4(mblk_t *mp, ip_recv_attr_t *ira)
1327 1328 {
1328 1329 icmph_t *icmph;
1329 1330 ipha_t *ipha; /* Outer header */
1330 1331 int ip_hdr_length; /* Outer header length */
1331 1332 boolean_t interested;
1332 1333 ipif_t *ipif;
1333 1334 uint32_t ts;
1334 1335 uint32_t *tsp;
1335 1336 timestruc_t now;
1336 1337 ill_t *ill = ira->ira_ill;
1337 1338 ip_stack_t *ipst = ill->ill_ipst;
1338 1339 zoneid_t zoneid = ira->ira_zoneid;
1339 1340 int len_needed;
1340 1341 mblk_t *mp_ret = NULL;
1341 1342
1342 1343 ipha = (ipha_t *)mp->b_rptr;
1343 1344
1344 1345 BUMP_MIB(&ipst->ips_icmp_mib, icmpInMsgs);
1345 1346
1346 1347 ip_hdr_length = ira->ira_ip_hdr_length;
1347 1348 if ((mp->b_wptr - mp->b_rptr) < (ip_hdr_length + ICMPH_SIZE)) {
1348 1349 if (ira->ira_pktlen < (ip_hdr_length + ICMPH_SIZE)) {
1349 1350 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
1350 1351 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
1351 1352 freemsg(mp);
1352 1353 return (NULL);
1353 1354 }
1354 1355 /* Last chance to get real. */
1355 1356 ipha = ip_pullup(mp, ip_hdr_length + ICMPH_SIZE, ira);
1356 1357 if (ipha == NULL) {
1357 1358 BUMP_MIB(&ipst->ips_icmp_mib, icmpInErrors);
1358 1359 freemsg(mp);
1359 1360 return (NULL);
1360 1361 }
1361 1362 }
1362 1363
1363 1364 /* The IP header will always be a multiple of four bytes */
1364 1365 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1365 1366 ip2dbg(("icmp_inbound_v4: type %d code %d\n", icmph->icmph_type,
1366 1367 icmph->icmph_code));
1367 1368
1368 1369 /*
1369 1370 * We will set "interested" to "true" if we should pass a copy to
1370 1371 * the transport or if we handle the packet locally.
1371 1372 */
1372 1373 interested = B_FALSE;
1373 1374 switch (icmph->icmph_type) {
1374 1375 case ICMP_ECHO_REPLY:
1375 1376 BUMP_MIB(&ipst->ips_icmp_mib, icmpInEchoReps);
1376 1377 break;
1377 1378 case ICMP_DEST_UNREACHABLE:
1378 1379 if (icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED)
1379 1380 BUMP_MIB(&ipst->ips_icmp_mib, icmpInFragNeeded);
1380 1381 interested = B_TRUE; /* Pass up to transport */
1381 1382 BUMP_MIB(&ipst->ips_icmp_mib, icmpInDestUnreachs);
1382 1383 break;
1383 1384 case ICMP_SOURCE_QUENCH:
1384 1385 interested = B_TRUE; /* Pass up to transport */
1385 1386 BUMP_MIB(&ipst->ips_icmp_mib, icmpInSrcQuenchs);
1386 1387 break;
1387 1388 case ICMP_REDIRECT:
1388 1389 if (!ipst->ips_ip_ignore_redirect)
1389 1390 interested = B_TRUE;
1390 1391 BUMP_MIB(&ipst->ips_icmp_mib, icmpInRedirects);
1391 1392 break;
1392 1393 case ICMP_ECHO_REQUEST:
1393 1394 /*
1394 1395 * Whether to respond to echo requests that come in as IP
1395 1396 * broadcasts or as IP multicast is subject to debate
1396 1397 * (what isn't?). We aim to please, you pick it.
1397 1398 * Default is do it.
1398 1399 */
1399 1400 if (ira->ira_flags & IRAF_MULTICAST) {
1400 1401 /* multicast: respond based on tunable */
1401 1402 interested = ipst->ips_ip_g_resp_to_echo_mcast;
1402 1403 } else if (ira->ira_flags & IRAF_BROADCAST) {
1403 1404 /* broadcast: respond based on tunable */
1404 1405 interested = ipst->ips_ip_g_resp_to_echo_bcast;
1405 1406 } else {
1406 1407 /* unicast: always respond */
1407 1408 interested = B_TRUE;
1408 1409 }
1409 1410 BUMP_MIB(&ipst->ips_icmp_mib, icmpInEchos);
1410 1411 if (!interested) {
1411 1412 /* We never pass these to RAW sockets */
1412 1413 freemsg(mp);
1413 1414 return (NULL);
1414 1415 }
1415 1416
1416 1417 /* Check db_ref to make sure we can modify the packet. */
1417 1418 if (mp->b_datap->db_ref > 1) {
1418 1419 mblk_t *mp1;
1419 1420
1420 1421 mp1 = copymsg(mp);
1421 1422 freemsg(mp);
1422 1423 if (!mp1) {
1423 1424 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1424 1425 return (NULL);
1425 1426 }
1426 1427 mp = mp1;
1427 1428 ipha = (ipha_t *)mp->b_rptr;
1428 1429 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1429 1430 }
1430 1431 icmph->icmph_type = ICMP_ECHO_REPLY;
1431 1432 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutEchoReps);
1432 1433 icmp_send_reply_v4(mp, ipha, icmph, ira);
1433 1434 return (NULL);
1434 1435
1435 1436 case ICMP_ROUTER_ADVERTISEMENT:
1436 1437 case ICMP_ROUTER_SOLICITATION:
1437 1438 break;
1438 1439 case ICMP_TIME_EXCEEDED:
1439 1440 interested = B_TRUE; /* Pass up to transport */
1440 1441 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimeExcds);
1441 1442 break;
1442 1443 case ICMP_PARAM_PROBLEM:
1443 1444 interested = B_TRUE; /* Pass up to transport */
1444 1445 BUMP_MIB(&ipst->ips_icmp_mib, icmpInParmProbs);
1445 1446 break;
1446 1447 case ICMP_TIME_STAMP_REQUEST:
1447 1448 /* Response to Time Stamp Requests is local policy. */
1448 1449 if (ipst->ips_ip_g_resp_to_timestamp) {
1449 1450 if (ira->ira_flags & IRAF_MULTIBROADCAST)
1450 1451 interested =
1451 1452 ipst->ips_ip_g_resp_to_timestamp_bcast;
1452 1453 else
1453 1454 interested = B_TRUE;
1454 1455 }
1455 1456 if (!interested) {
1456 1457 /* We never pass these to RAW sockets */
1457 1458 freemsg(mp);
1458 1459 return (NULL);
1459 1460 }
1460 1461
1461 1462 /* Make sure we have enough of the packet */
1462 1463 len_needed = ip_hdr_length + ICMPH_SIZE +
1463 1464 3 * sizeof (uint32_t);
1464 1465
1465 1466 if (mp->b_wptr - mp->b_rptr < len_needed) {
1466 1467 ipha = ip_pullup(mp, len_needed, ira);
1467 1468 if (ipha == NULL) {
1468 1469 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1469 1470 ip_drop_input("ipIfStatsInDiscards - ip_pullup",
1470 1471 mp, ill);
1471 1472 freemsg(mp);
1472 1473 return (NULL);
1473 1474 }
1474 1475 /* Refresh following the pullup. */
1475 1476 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1476 1477 }
1477 1478 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimestamps);
1478 1479 /* Check db_ref to make sure we can modify the packet. */
1479 1480 if (mp->b_datap->db_ref > 1) {
1480 1481 mblk_t *mp1;
1481 1482
1482 1483 mp1 = copymsg(mp);
1483 1484 freemsg(mp);
1484 1485 if (!mp1) {
1485 1486 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1486 1487 return (NULL);
1487 1488 }
1488 1489 mp = mp1;
1489 1490 ipha = (ipha_t *)mp->b_rptr;
1490 1491 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1491 1492 }
1492 1493 icmph->icmph_type = ICMP_TIME_STAMP_REPLY;
1493 1494 tsp = (uint32_t *)&icmph[1];
1494 1495 tsp++; /* Skip past 'originate time' */
1495 1496 /* Compute # of milliseconds since midnight */
1496 1497 gethrestime(&now);
1497 1498 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
1498 1499 now.tv_nsec / (NANOSEC / MILLISEC);
1499 1500 *tsp++ = htonl(ts); /* Lay in 'receive time' */
1500 1501 *tsp++ = htonl(ts); /* Lay in 'send time' */
1501 1502 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutTimestampReps);
1502 1503 icmp_send_reply_v4(mp, ipha, icmph, ira);
1503 1504 return (NULL);
1504 1505
1505 1506 case ICMP_TIME_STAMP_REPLY:
1506 1507 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimestampReps);
1507 1508 break;
1508 1509 case ICMP_INFO_REQUEST:
1509 1510 /* Per RFC 1122 3.2.2.7, ignore this. */
1510 1511 case ICMP_INFO_REPLY:
1511 1512 break;
1512 1513 case ICMP_ADDRESS_MASK_REQUEST:
1513 1514 if (ira->ira_flags & IRAF_MULTIBROADCAST) {
1514 1515 interested =
1515 1516 ipst->ips_ip_respond_to_address_mask_broadcast;
1516 1517 } else {
1517 1518 interested = B_TRUE;
1518 1519 }
1519 1520 if (!interested) {
1520 1521 /* We never pass these to RAW sockets */
1521 1522 freemsg(mp);
1522 1523 return (NULL);
1523 1524 }
1524 1525 len_needed = ip_hdr_length + ICMPH_SIZE + IP_ADDR_LEN;
1525 1526 if (mp->b_wptr - mp->b_rptr < len_needed) {
1526 1527 ipha = ip_pullup(mp, len_needed, ira);
1527 1528 if (ipha == NULL) {
1528 1529 BUMP_MIB(ill->ill_ip_mib,
1529 1530 ipIfStatsInTruncatedPkts);
1530 1531 ip_drop_input("ipIfStatsInTruncatedPkts", mp,
1531 1532 ill);
1532 1533 freemsg(mp);
1533 1534 return (NULL);
1534 1535 }
1535 1536 /* Refresh following the pullup. */
1536 1537 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1537 1538 }
1538 1539 BUMP_MIB(&ipst->ips_icmp_mib, icmpInAddrMasks);
1539 1540 /* Check db_ref to make sure we can modify the packet. */
1540 1541 if (mp->b_datap->db_ref > 1) {
1541 1542 mblk_t *mp1;
1542 1543
1543 1544 mp1 = copymsg(mp);
1544 1545 freemsg(mp);
1545 1546 if (!mp1) {
1546 1547 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1547 1548 return (NULL);
1548 1549 }
1549 1550 mp = mp1;
1550 1551 ipha = (ipha_t *)mp->b_rptr;
1551 1552 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1552 1553 }
1553 1554 /*
1554 1555 * Need the ipif with the mask be the same as the source
1555 1556 * address of the mask reply. For unicast we have a specific
1556 1557 * ipif. For multicast/broadcast we only handle onlink
1557 1558 * senders, and use the source address to pick an ipif.
1558 1559 */
1559 1560 ipif = ipif_lookup_addr(ipha->ipha_dst, ill, zoneid, ipst);
1560 1561 if (ipif == NULL) {
1561 1562 /* Broadcast or multicast */
1562 1563 ipif = ipif_lookup_remote(ill, ipha->ipha_src, zoneid);
1563 1564 if (ipif == NULL) {
1564 1565 freemsg(mp);
1565 1566 return (NULL);
1566 1567 }
1567 1568 }
1568 1569 icmph->icmph_type = ICMP_ADDRESS_MASK_REPLY;
1569 1570 bcopy(&ipif->ipif_net_mask, &icmph[1], IP_ADDR_LEN);
1570 1571 ipif_refrele(ipif);
1571 1572 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutAddrMaskReps);
1572 1573 icmp_send_reply_v4(mp, ipha, icmph, ira);
1573 1574 return (NULL);
1574 1575
1575 1576 case ICMP_ADDRESS_MASK_REPLY:
1576 1577 BUMP_MIB(&ipst->ips_icmp_mib, icmpInAddrMaskReps);
1577 1578 break;
1578 1579 default:
1579 1580 interested = B_TRUE; /* Pass up to transport */
1580 1581 BUMP_MIB(&ipst->ips_icmp_mib, icmpInUnknowns);
1581 1582 break;
1582 1583 }
1583 1584 /*
1584 1585 * See if there is an ICMP client to avoid an extra copymsg/freemsg
1585 1586 * if there isn't one.
1586 1587 */
1587 1588 if (ipst->ips_ipcl_proto_fanout_v4[IPPROTO_ICMP].connf_head != NULL) {
1588 1589 /* If there is an ICMP client and we want one too, copy it. */
1589 1590
1590 1591 if (!interested) {
1591 1592 /* Caller will deliver to RAW sockets */
1592 1593 return (mp);
1593 1594 }
1594 1595 mp_ret = copymsg(mp);
1595 1596 if (mp_ret == NULL) {
1596 1597 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1597 1598 ip_drop_input("ipIfStatsInDiscards - copymsg", mp, ill);
1598 1599 }
1599 1600 } else if (!interested) {
1600 1601 /* Neither we nor raw sockets are interested. Drop packet now */
1601 1602 freemsg(mp);
1602 1603 return (NULL);
1603 1604 }
1604 1605
1605 1606 /*
1606 1607 * ICMP error or redirect packet. Make sure we have enough of
1607 1608 * the header and that db_ref == 1 since we might end up modifying
1608 1609 * the packet.
1609 1610 */
1610 1611 if (mp->b_cont != NULL) {
1611 1612 if (ip_pullup(mp, -1, ira) == NULL) {
1612 1613 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1613 1614 ip_drop_input("ipIfStatsInDiscards - ip_pullup",
1614 1615 mp, ill);
1615 1616 freemsg(mp);
1616 1617 return (mp_ret);
1617 1618 }
1618 1619 }
1619 1620
1620 1621 if (mp->b_datap->db_ref > 1) {
1621 1622 mblk_t *mp1;
1622 1623
1623 1624 mp1 = copymsg(mp);
1624 1625 if (mp1 == NULL) {
1625 1626 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1626 1627 ip_drop_input("ipIfStatsInDiscards - copymsg", mp, ill);
1627 1628 freemsg(mp);
1628 1629 return (mp_ret);
1629 1630 }
1630 1631 freemsg(mp);
1631 1632 mp = mp1;
1632 1633 }
1633 1634
1634 1635 /*
1635 1636 * In case mp has changed, verify the message before any further
1636 1637 * processes.
1637 1638 */
1638 1639 ipha = (ipha_t *)mp->b_rptr;
1639 1640 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1640 1641 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
1641 1642 freemsg(mp);
1642 1643 return (mp_ret);
1643 1644 }
1644 1645
1645 1646 switch (icmph->icmph_type) {
1646 1647 case ICMP_REDIRECT:
1647 1648 icmp_redirect_v4(mp, ipha, icmph, ira);
1648 1649 break;
1649 1650 case ICMP_DEST_UNREACHABLE:
1650 1651 if (icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED) {
1651 1652 /* Update DCE and adjust MTU is icmp header if needed */
1652 1653 icmp_inbound_too_big_v4(icmph, ira);
1653 1654 }
1654 1655 /* FALLTHRU */
1655 1656 default:
1656 1657 icmp_inbound_error_fanout_v4(mp, icmph, ira);
1657 1658 break;
1658 1659 }
1659 1660 return (mp_ret);
1660 1661 }
1661 1662
1662 1663 /*
1663 1664 * Send an ICMP echo, timestamp or address mask reply.
1664 1665 * The caller has already updated the payload part of the packet.
1665 1666 * We handle the ICMP checksum, IP source address selection and feed
1666 1667 * the packet into ip_output_simple.
1667 1668 */
1668 1669 static void
1669 1670 icmp_send_reply_v4(mblk_t *mp, ipha_t *ipha, icmph_t *icmph,
1670 1671 ip_recv_attr_t *ira)
1671 1672 {
1672 1673 uint_t ip_hdr_length = ira->ira_ip_hdr_length;
1673 1674 ill_t *ill = ira->ira_ill;
1674 1675 ip_stack_t *ipst = ill->ill_ipst;
1675 1676 ip_xmit_attr_t ixas;
1676 1677
1677 1678 /* Send out an ICMP packet */
1678 1679 icmph->icmph_checksum = 0;
1679 1680 icmph->icmph_checksum = IP_CSUM(mp, ip_hdr_length, 0);
1680 1681 /* Reset time to live. */
1681 1682 ipha->ipha_ttl = ipst->ips_ip_def_ttl;
1682 1683 {
1683 1684 /* Swap source and destination addresses */
1684 1685 ipaddr_t tmp;
1685 1686
1686 1687 tmp = ipha->ipha_src;
1687 1688 ipha->ipha_src = ipha->ipha_dst;
1688 1689 ipha->ipha_dst = tmp;
1689 1690 }
1690 1691 ipha->ipha_ident = 0;
1691 1692 if (!IS_SIMPLE_IPH(ipha))
1692 1693 icmp_options_update(ipha);
1693 1694
1694 1695 bzero(&ixas, sizeof (ixas));
1695 1696 ixas.ixa_flags = IXAF_BASIC_SIMPLE_V4;
1696 1697 ixas.ixa_zoneid = ira->ira_zoneid;
1697 1698 ixas.ixa_cred = kcred;
1698 1699 ixas.ixa_cpid = NOPID;
1699 1700 ixas.ixa_tsl = ira->ira_tsl; /* Behave as a multi-level responder */
1700 1701 ixas.ixa_ifindex = 0;
1701 1702 ixas.ixa_ipst = ipst;
1702 1703 ixas.ixa_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1703 1704
1704 1705 if (!(ira->ira_flags & IRAF_IPSEC_SECURE)) {
1705 1706 /*
1706 1707 * This packet should go out the same way as it
1707 1708 * came in i.e in clear, independent of the IPsec policy
1708 1709 * for transmitting packets.
1709 1710 */
1710 1711 ixas.ixa_flags |= IXAF_NO_IPSEC;
1711 1712 } else {
1712 1713 if (!ipsec_in_to_out(ira, &ixas, mp, ipha, NULL)) {
1713 1714 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1714 1715 /* Note: mp already consumed and ip_drop_packet done */
1715 1716 return;
1716 1717 }
1717 1718 }
1718 1719 if (ira->ira_flags & IRAF_MULTIBROADCAST) {
1719 1720 /*
1720 1721 * Not one or our addresses (IRE_LOCALs), thus we let
1721 1722 * ip_output_simple pick the source.
1722 1723 */
1723 1724 ipha->ipha_src = INADDR_ANY;
1724 1725 ixas.ixa_flags |= IXAF_SET_SOURCE;
1725 1726 }
1726 1727 /* Should we send with DF and use dce_pmtu? */
1727 1728 if (ipst->ips_ipv4_icmp_return_pmtu) {
1728 1729 ixas.ixa_flags |= IXAF_PMTU_DISCOVERY;
1729 1730 ipha->ipha_fragment_offset_and_flags |= IPH_DF_HTONS;
1730 1731 }
1731 1732
1732 1733 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutMsgs);
1733 1734
1734 1735 (void) ip_output_simple(mp, &ixas);
1735 1736 ixa_cleanup(&ixas);
1736 1737 }
1737 1738
1738 1739 /*
1739 1740 * Verify the ICMP messages for either for ICMP error or redirect packet.
1740 1741 * The caller should have fully pulled up the message. If it's a redirect
1741 1742 * packet, only basic checks on IP header will be done; otherwise, verify
1742 1743 * the packet by looking at the included ULP header.
1743 1744 *
1744 1745 * Called before icmp_inbound_error_fanout_v4 is called.
1745 1746 */
1746 1747 static boolean_t
1747 1748 icmp_inbound_verify_v4(mblk_t *mp, icmph_t *icmph, ip_recv_attr_t *ira)
1748 1749 {
1749 1750 ill_t *ill = ira->ira_ill;
1750 1751 int hdr_length;
1751 1752 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
1752 1753 conn_t *connp;
1753 1754 ipha_t *ipha; /* Inner IP header */
1754 1755
1755 1756 ipha = (ipha_t *)&icmph[1];
1756 1757 if ((uchar_t *)ipha + IP_SIMPLE_HDR_LENGTH > mp->b_wptr)
1757 1758 goto truncated;
1758 1759
1759 1760 hdr_length = IPH_HDR_LENGTH(ipha);
1760 1761
1761 1762 if ((IPH_HDR_VERSION(ipha) != IPV4_VERSION))
1762 1763 goto discard_pkt;
1763 1764
1764 1765 if (hdr_length < sizeof (ipha_t))
1765 1766 goto truncated;
1766 1767
1767 1768 if ((uchar_t *)ipha + hdr_length > mp->b_wptr)
1768 1769 goto truncated;
1769 1770
1770 1771 /*
1771 1772 * Stop here for ICMP_REDIRECT.
1772 1773 */
1773 1774 if (icmph->icmph_type == ICMP_REDIRECT)
1774 1775 return (B_TRUE);
1775 1776
1776 1777 /*
1777 1778 * ICMP errors only.
1778 1779 */
1779 1780 switch (ipha->ipha_protocol) {
1780 1781 case IPPROTO_UDP:
1781 1782 /*
1782 1783 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1783 1784 * transport header.
1784 1785 */
1785 1786 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1786 1787 mp->b_wptr)
1787 1788 goto truncated;
1788 1789 break;
1789 1790 case IPPROTO_TCP: {
1790 1791 tcpha_t *tcpha;
1791 1792
1792 1793 /*
1793 1794 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1794 1795 * transport header.
1795 1796 */
1796 1797 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1797 1798 mp->b_wptr)
1798 1799 goto truncated;
1799 1800
1800 1801 tcpha = (tcpha_t *)((uchar_t *)ipha + hdr_length);
1801 1802 connp = ipcl_tcp_lookup_reversed_ipv4(ipha, tcpha, TCPS_LISTEN,
1802 1803 ipst);
1803 1804 if (connp == NULL)
1804 1805 goto discard_pkt;
1805 1806
1806 1807 if ((connp->conn_verifyicmp != NULL) &&
1807 1808 !connp->conn_verifyicmp(connp, tcpha, icmph, NULL, ira)) {
1808 1809 CONN_DEC_REF(connp);
1809 1810 goto discard_pkt;
1810 1811 }
1811 1812 CONN_DEC_REF(connp);
1812 1813 break;
1813 1814 }
1814 1815 case IPPROTO_SCTP:
1815 1816 /*
1816 1817 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1817 1818 * transport header.
1818 1819 */
1819 1820 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1820 1821 mp->b_wptr)
1821 1822 goto truncated;
1822 1823 break;
1823 1824 case IPPROTO_ESP:
1824 1825 case IPPROTO_AH:
1825 1826 break;
1826 1827 case IPPROTO_ENCAP:
1827 1828 if ((uchar_t *)ipha + hdr_length + sizeof (ipha_t) >
1828 1829 mp->b_wptr)
1829 1830 goto truncated;
1830 1831 break;
1831 1832 default:
1832 1833 break;
1833 1834 }
1834 1835
1835 1836 return (B_TRUE);
1836 1837
1837 1838 discard_pkt:
1838 1839 /* Bogus ICMP error. */
1839 1840 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1840 1841 return (B_FALSE);
1841 1842
1842 1843 truncated:
1843 1844 /* We pulled up everthing already. Must be truncated */
1844 1845 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
1845 1846 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
1846 1847 return (B_FALSE);
1847 1848 }
1848 1849
1849 1850 /* Table from RFC 1191 */
1850 1851 static int icmp_frag_size_table[] =
1851 1852 { 32000, 17914, 8166, 4352, 2002, 1496, 1006, 508, 296, 68 };
1852 1853
1853 1854 /*
1854 1855 * Process received ICMP Packet too big.
1855 1856 * Just handles the DCE create/update, including using the above table of
1856 1857 * PMTU guesses. The caller is responsible for validating the packet before
1857 1858 * passing it in and also to fanout the ICMP error to any matching transport
1858 1859 * conns. Assumes the message has been fully pulled up and verified.
1859 1860 *
1860 1861 * Before getting here, the caller has called icmp_inbound_verify_v4()
1861 1862 * that should have verified with ULP to prevent undoing the changes we're
1862 1863 * going to make to DCE. For example, TCP might have verified that the packet
1863 1864 * which generated error is in the send window.
1864 1865 *
1865 1866 * In some cases modified this MTU in the ICMP header packet; the caller
1866 1867 * should pass to the matching ULP after this returns.
1867 1868 */
1868 1869 static void
1869 1870 icmp_inbound_too_big_v4(icmph_t *icmph, ip_recv_attr_t *ira)
1870 1871 {
1871 1872 dce_t *dce;
1872 1873 int old_mtu;
1873 1874 int mtu, orig_mtu;
1874 1875 ipaddr_t dst;
1875 1876 boolean_t disable_pmtud;
1876 1877 ill_t *ill = ira->ira_ill;
1877 1878 ip_stack_t *ipst = ill->ill_ipst;
1878 1879 uint_t hdr_length;
1879 1880 ipha_t *ipha;
1880 1881
1881 1882 /* Caller already pulled up everything. */
1882 1883 ipha = (ipha_t *)&icmph[1];
1883 1884 ASSERT(icmph->icmph_type == ICMP_DEST_UNREACHABLE &&
1884 1885 icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED);
1885 1886 ASSERT(ill != NULL);
1886 1887
1887 1888 hdr_length = IPH_HDR_LENGTH(ipha);
1888 1889
1889 1890 /*
1890 1891 * We handle path MTU for source routed packets since the DCE
1891 1892 * is looked up using the final destination.
1892 1893 */
1893 1894 dst = ip_get_dst(ipha);
1894 1895
1895 1896 dce = dce_lookup_and_add_v4(dst, ipst);
1896 1897 if (dce == NULL) {
1897 1898 /* Couldn't add a unique one - ENOMEM */
1898 1899 ip1dbg(("icmp_inbound_too_big_v4: no dce for 0x%x\n",
1899 1900 ntohl(dst)));
1900 1901 return;
1901 1902 }
1902 1903
1903 1904 /* Check for MTU discovery advice as described in RFC 1191 */
1904 1905 mtu = ntohs(icmph->icmph_du_mtu);
1905 1906 orig_mtu = mtu;
1906 1907 disable_pmtud = B_FALSE;
1907 1908
1908 1909 mutex_enter(&dce->dce_lock);
1909 1910 if (dce->dce_flags & DCEF_PMTU)
1910 1911 old_mtu = dce->dce_pmtu;
1911 1912 else
1912 1913 old_mtu = ill->ill_mtu;
1913 1914
1914 1915 if (icmph->icmph_du_zero != 0 || mtu < ipst->ips_ip_pmtu_min) {
1915 1916 uint32_t length;
1916 1917 int i;
1917 1918
1918 1919 /*
1919 1920 * Use the table from RFC 1191 to figure out
1920 1921 * the next "plateau" based on the length in
1921 1922 * the original IP packet.
1922 1923 */
1923 1924 length = ntohs(ipha->ipha_length);
1924 1925 DTRACE_PROBE2(ip4__pmtu__guess, dce_t *, dce,
1925 1926 uint32_t, length);
1926 1927 if (old_mtu <= length &&
1927 1928 old_mtu >= length - hdr_length) {
1928 1929 /*
1929 1930 * Handle broken BSD 4.2 systems that
1930 1931 * return the wrong ipha_length in ICMP
1931 1932 * errors.
1932 1933 */
1933 1934 ip1dbg(("Wrong mtu: sent %d, dce %d\n",
1934 1935 length, old_mtu));
1935 1936 length -= hdr_length;
1936 1937 }
1937 1938 for (i = 0; i < A_CNT(icmp_frag_size_table); i++) {
1938 1939 if (length > icmp_frag_size_table[i])
1939 1940 break;
1940 1941 }
1941 1942 if (i == A_CNT(icmp_frag_size_table)) {
1942 1943 /* Smaller than IP_MIN_MTU! */
1943 1944 ip1dbg(("Too big for packet size %d\n",
1944 1945 length));
1945 1946 disable_pmtud = B_TRUE;
1946 1947 mtu = ipst->ips_ip_pmtu_min;
1947 1948 } else {
1948 1949 mtu = icmp_frag_size_table[i];
1949 1950 ip1dbg(("Calculated mtu %d, packet size %d, "
1950 1951 "before %d\n", mtu, length, old_mtu));
1951 1952 if (mtu < ipst->ips_ip_pmtu_min) {
1952 1953 mtu = ipst->ips_ip_pmtu_min;
1953 1954 disable_pmtud = B_TRUE;
1954 1955 }
1955 1956 }
1956 1957 }
1957 1958 if (disable_pmtud)
1958 1959 dce->dce_flags |= DCEF_TOO_SMALL_PMTU;
1959 1960 else
1960 1961 dce->dce_flags &= ~DCEF_TOO_SMALL_PMTU;
1961 1962
1962 1963 dce->dce_pmtu = MIN(old_mtu, mtu);
1963 1964 /* Prepare to send the new max frag size for the ULP. */
1964 1965 icmph->icmph_du_zero = 0;
1965 1966 icmph->icmph_du_mtu = htons((uint16_t)dce->dce_pmtu);
1966 1967 DTRACE_PROBE4(ip4__pmtu__change, icmph_t *, icmph, dce_t *,
1967 1968 dce, int, orig_mtu, int, mtu);
1968 1969
1969 1970 /* We now have a PMTU for sure */
1970 1971 dce->dce_flags |= DCEF_PMTU;
1971 1972 dce->dce_last_change_time = TICK_TO_SEC(ddi_get_lbolt64());
1972 1973 mutex_exit(&dce->dce_lock);
1973 1974 /*
1974 1975 * After dropping the lock the new value is visible to everyone.
1975 1976 * Then we bump the generation number so any cached values reinspect
1976 1977 * the dce_t.
1977 1978 */
1978 1979 dce_increment_generation(dce);
1979 1980 dce_refrele(dce);
1980 1981 }
1981 1982
1982 1983 /*
1983 1984 * If the packet in error is Self-Encapsulated, icmp_inbound_error_fanout_v4
1984 1985 * calls this function.
1985 1986 */
1986 1987 static mblk_t *
1987 1988 icmp_inbound_self_encap_error_v4(mblk_t *mp, ipha_t *ipha, ipha_t *in_ipha)
1988 1989 {
1989 1990 int length;
1990 1991
1991 1992 ASSERT(mp->b_datap->db_type == M_DATA);
1992 1993
1993 1994 /* icmp_inbound_v4 has already pulled up the whole error packet */
1994 1995 ASSERT(mp->b_cont == NULL);
1995 1996
1996 1997 /*
1997 1998 * The length that we want to overlay is the inner header
1998 1999 * and what follows it.
1999 2000 */
2000 2001 length = msgdsize(mp) - ((uchar_t *)in_ipha - mp->b_rptr);
2001 2002
2002 2003 /*
2003 2004 * Overlay the inner header and whatever follows it over the
2004 2005 * outer header.
2005 2006 */
2006 2007 bcopy((uchar_t *)in_ipha, (uchar_t *)ipha, length);
2007 2008
2008 2009 /* Adjust for what we removed */
2009 2010 mp->b_wptr -= (uchar_t *)in_ipha - (uchar_t *)ipha;
2010 2011 return (mp);
2011 2012 }
2012 2013
2013 2014 /*
2014 2015 * Try to pass the ICMP message upstream in case the ULP cares.
2015 2016 *
2016 2017 * If the packet that caused the ICMP error is secure, we send
2017 2018 * it to AH/ESP to make sure that the attached packet has a
2018 2019 * valid association. ipha in the code below points to the
2019 2020 * IP header of the packet that caused the error.
2020 2021 *
2021 2022 * For IPsec cases, we let the next-layer-up (which has access to
2022 2023 * cached policy on the conn_t, or can query the SPD directly)
2023 2024 * subtract out any IPsec overhead if they must. We therefore make no
2024 2025 * adjustments here for IPsec overhead.
2025 2026 *
2026 2027 * IFN could have been generated locally or by some router.
2027 2028 *
2028 2029 * LOCAL : ire_send_wire (before calling ipsec_out_process) can call
2029 2030 * icmp_frag_needed/icmp_pkt2big_v6 to generated a local IFN.
2030 2031 * This happens because IP adjusted its value of MTU on an
2031 2032 * earlier IFN message and could not tell the upper layer,
2032 2033 * the new adjusted value of MTU e.g. Packet was encrypted
2033 2034 * or there was not enough information to fanout to upper
2034 2035 * layers. Thus on the next outbound datagram, ire_send_wire
2035 2036 * generates the IFN, where IPsec processing has *not* been
2036 2037 * done.
2037 2038 *
2038 2039 * Note that we retain ixa_fragsize across IPsec thus once
2039 2040 * we have picking ixa_fragsize and entered ipsec_out_process we do
2040 2041 * no change the fragsize even if the path MTU changes before
2041 2042 * we reach ip_output_post_ipsec.
2042 2043 *
2043 2044 * In the local case, IRAF_LOOPBACK will be set indicating
2044 2045 * that IFN was generated locally.
2045 2046 *
2046 2047 * ROUTER : IFN could be secure or non-secure.
2047 2048 *
2048 2049 * * SECURE : We use the IPSEC_IN to fanout to AH/ESP if the
2049 2050 * packet in error has AH/ESP headers to validate the AH/ESP
2050 2051 * headers. AH/ESP will verify whether there is a valid SA or
2051 2052 * not and send it back. We will fanout again if we have more
2052 2053 * data in the packet.
2053 2054 *
2054 2055 * If the packet in error does not have AH/ESP, we handle it
2055 2056 * like any other case.
2056 2057 *
2057 2058 * * NON_SECURE : If the packet in error has AH/ESP headers, we send it
2058 2059 * up to AH/ESP for validation. AH/ESP will verify whether there is a
2059 2060 * valid SA or not and send it back. We will fanout again if
2060 2061 * we have more data in the packet.
2061 2062 *
2062 2063 * If the packet in error does not have AH/ESP, we handle it
2063 2064 * like any other case.
2064 2065 *
2065 2066 * The caller must have called icmp_inbound_verify_v4.
2066 2067 */
2067 2068 static void
2068 2069 icmp_inbound_error_fanout_v4(mblk_t *mp, icmph_t *icmph, ip_recv_attr_t *ira)
2069 2070 {
2070 2071 uint16_t *up; /* Pointer to ports in ULP header */
2071 2072 uint32_t ports; /* reversed ports for fanout */
2072 2073 ipha_t ripha; /* With reversed addresses */
2073 2074 ipha_t *ipha; /* Inner IP header */
2074 2075 uint_t hdr_length; /* Inner IP header length */
2075 2076 tcpha_t *tcpha;
2076 2077 conn_t *connp;
2077 2078 ill_t *ill = ira->ira_ill;
2078 2079 ip_stack_t *ipst = ill->ill_ipst;
2079 2080 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
2080 2081 ill_t *rill = ira->ira_rill;
2081 2082
2082 2083 /* Caller already pulled up everything. */
2083 2084 ipha = (ipha_t *)&icmph[1];
2084 2085 ASSERT((uchar_t *)&ipha[1] <= mp->b_wptr);
2085 2086 ASSERT(mp->b_cont == NULL);
2086 2087
2087 2088 hdr_length = IPH_HDR_LENGTH(ipha);
2088 2089 ira->ira_protocol = ipha->ipha_protocol;
2089 2090
2090 2091 /*
2091 2092 * We need a separate IP header with the source and destination
2092 2093 * addresses reversed to do fanout/classification because the ipha in
2093 2094 * the ICMP error is in the form we sent it out.
2094 2095 */
2095 2096 ripha.ipha_src = ipha->ipha_dst;
2096 2097 ripha.ipha_dst = ipha->ipha_src;
2097 2098 ripha.ipha_protocol = ipha->ipha_protocol;
2098 2099 ripha.ipha_version_and_hdr_length = ipha->ipha_version_and_hdr_length;
2099 2100
2100 2101 ip2dbg(("icmp_inbound_error_v4: proto %d %x to %x: %d/%d\n",
2101 2102 ripha.ipha_protocol, ntohl(ipha->ipha_src),
2102 2103 ntohl(ipha->ipha_dst),
2103 2104 icmph->icmph_type, icmph->icmph_code));
2104 2105
2105 2106 switch (ipha->ipha_protocol) {
2106 2107 case IPPROTO_UDP:
2107 2108 up = (uint16_t *)((uchar_t *)ipha + hdr_length);
2108 2109
2109 2110 /* Attempt to find a client stream based on port. */
2110 2111 ip2dbg(("icmp_inbound_error_v4: UDP ports %d to %d\n",
2111 2112 ntohs(up[0]), ntohs(up[1])));
2112 2113
2113 2114 /* Note that we send error to all matches. */
2114 2115 ira->ira_flags |= IRAF_ICMP_ERROR;
2115 2116 ip_fanout_udp_multi_v4(mp, &ripha, up[0], up[1], ira);
2116 2117 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2117 2118 return;
2118 2119
2119 2120 case IPPROTO_TCP:
2120 2121 /*
2121 2122 * Find a TCP client stream for this packet.
2122 2123 * Note that we do a reverse lookup since the header is
2123 2124 * in the form we sent it out.
2124 2125 */
2125 2126 tcpha = (tcpha_t *)((uchar_t *)ipha + hdr_length);
2126 2127 connp = ipcl_tcp_lookup_reversed_ipv4(ipha, tcpha, TCPS_LISTEN,
2127 2128 ipst);
2128 2129 if (connp == NULL)
2129 2130 goto discard_pkt;
2130 2131
2131 2132 if (CONN_INBOUND_POLICY_PRESENT(connp, ipss) ||
2132 2133 (ira->ira_flags & IRAF_IPSEC_SECURE)) {
2133 2134 mp = ipsec_check_inbound_policy(mp, connp,
2134 2135 ipha, NULL, ira);
2135 2136 if (mp == NULL) {
2136 2137 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2137 2138 /* Note that mp is NULL */
2138 2139 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2139 2140 CONN_DEC_REF(connp);
2140 2141 return;
2141 2142 }
2142 2143 }
2143 2144
2144 2145 ira->ira_flags |= IRAF_ICMP_ERROR;
2145 2146 ira->ira_ill = ira->ira_rill = NULL;
2146 2147 if (IPCL_IS_TCP(connp)) {
2147 2148 SQUEUE_ENTER_ONE(connp->conn_sqp, mp,
2148 2149 connp->conn_recvicmp, connp, ira, SQ_FILL,
2149 2150 SQTAG_TCP_INPUT_ICMP_ERR);
2150 2151 } else {
2151 2152 /* Not TCP; must be SOCK_RAW, IPPROTO_TCP */
2152 2153 (connp->conn_recv)(connp, mp, NULL, ira);
2153 2154 CONN_DEC_REF(connp);
2154 2155 }
2155 2156 ira->ira_ill = ill;
2156 2157 ira->ira_rill = rill;
2157 2158 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2158 2159 return;
2159 2160
2160 2161 case IPPROTO_SCTP:
2161 2162 up = (uint16_t *)((uchar_t *)ipha + hdr_length);
2162 2163 /* Find a SCTP client stream for this packet. */
2163 2164 ((uint16_t *)&ports)[0] = up[1];
2164 2165 ((uint16_t *)&ports)[1] = up[0];
2165 2166
2166 2167 ira->ira_flags |= IRAF_ICMP_ERROR;
2167 2168 ip_fanout_sctp(mp, &ripha, NULL, ports, ira);
2168 2169 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2169 2170 return;
2170 2171
2171 2172 case IPPROTO_ESP:
2172 2173 case IPPROTO_AH:
2173 2174 if (!ipsec_loaded(ipss)) {
2174 2175 ip_proto_not_sup(mp, ira);
2175 2176 return;
2176 2177 }
2177 2178
2178 2179 if (ipha->ipha_protocol == IPPROTO_ESP)
2179 2180 mp = ipsecesp_icmp_error(mp, ira);
2180 2181 else
2181 2182 mp = ipsecah_icmp_error(mp, ira);
2182 2183 if (mp == NULL)
2183 2184 return;
2184 2185
2185 2186 /* Just in case ipsec didn't preserve the NULL b_cont */
2186 2187 if (mp->b_cont != NULL) {
2187 2188 if (!pullupmsg(mp, -1))
2188 2189 goto discard_pkt;
2189 2190 }
2190 2191
2191 2192 /*
2192 2193 * Note that ira_pktlen and ira_ip_hdr_length are no longer
2193 2194 * correct, but we don't use them any more here.
2194 2195 *
2195 2196 * If succesful, the mp has been modified to not include
2196 2197 * the ESP/AH header so we can fanout to the ULP's icmp
2197 2198 * error handler.
2198 2199 */
2199 2200 if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH)
2200 2201 goto truncated;
2201 2202
2202 2203 /* Verify the modified message before any further processes. */
2203 2204 ipha = (ipha_t *)mp->b_rptr;
2204 2205 hdr_length = IPH_HDR_LENGTH(ipha);
2205 2206 icmph = (icmph_t *)&mp->b_rptr[hdr_length];
2206 2207 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
2207 2208 freemsg(mp);
2208 2209 return;
2209 2210 }
2210 2211
2211 2212 icmp_inbound_error_fanout_v4(mp, icmph, ira);
2212 2213 return;
2213 2214
2214 2215 case IPPROTO_ENCAP: {
2215 2216 /* Look for self-encapsulated packets that caused an error */
2216 2217 ipha_t *in_ipha;
2217 2218
2218 2219 /*
2219 2220 * Caller has verified that length has to be
2220 2221 * at least the size of IP header.
2221 2222 */
2222 2223 ASSERT(hdr_length >= sizeof (ipha_t));
2223 2224 /*
2224 2225 * Check the sanity of the inner IP header like
2225 2226 * we did for the outer header.
2226 2227 */
2227 2228 in_ipha = (ipha_t *)((uchar_t *)ipha + hdr_length);
2228 2229 if ((IPH_HDR_VERSION(in_ipha) != IPV4_VERSION)) {
2229 2230 goto discard_pkt;
2230 2231 }
2231 2232 if (IPH_HDR_LENGTH(in_ipha) < sizeof (ipha_t)) {
2232 2233 goto discard_pkt;
2233 2234 }
2234 2235 /* Check for Self-encapsulated tunnels */
2235 2236 if (in_ipha->ipha_src == ipha->ipha_src &&
2236 2237 in_ipha->ipha_dst == ipha->ipha_dst) {
2237 2238
2238 2239 mp = icmp_inbound_self_encap_error_v4(mp, ipha,
2239 2240 in_ipha);
2240 2241 if (mp == NULL)
2241 2242 goto discard_pkt;
2242 2243
2243 2244 /*
2244 2245 * Just in case self_encap didn't preserve the NULL
2245 2246 * b_cont
2246 2247 */
2247 2248 if (mp->b_cont != NULL) {
2248 2249 if (!pullupmsg(mp, -1))
2249 2250 goto discard_pkt;
2250 2251 }
2251 2252 /*
2252 2253 * Note that ira_pktlen and ira_ip_hdr_length are no
2253 2254 * longer correct, but we don't use them any more here.
2254 2255 */
2255 2256 if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH)
2256 2257 goto truncated;
2257 2258
2258 2259 /*
2259 2260 * Verify the modified message before any further
2260 2261 * processes.
2261 2262 */
2262 2263 ipha = (ipha_t *)mp->b_rptr;
2263 2264 hdr_length = IPH_HDR_LENGTH(ipha);
2264 2265 icmph = (icmph_t *)&mp->b_rptr[hdr_length];
2265 2266 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
2266 2267 freemsg(mp);
2267 2268 return;
2268 2269 }
2269 2270
2270 2271 /*
2271 2272 * The packet in error is self-encapsualted.
2272 2273 * And we are finding it further encapsulated
2273 2274 * which we could not have possibly generated.
2274 2275 */
2275 2276 if (ipha->ipha_protocol == IPPROTO_ENCAP) {
2276 2277 goto discard_pkt;
2277 2278 }
2278 2279 icmp_inbound_error_fanout_v4(mp, icmph, ira);
2279 2280 return;
2280 2281 }
2281 2282 /* No self-encapsulated */
2282 2283 /* FALLTHRU */
2283 2284 }
2284 2285 case IPPROTO_IPV6:
2285 2286 if ((connp = ipcl_iptun_classify_v4(&ripha.ipha_src,
2286 2287 &ripha.ipha_dst, ipst)) != NULL) {
2287 2288 ira->ira_flags |= IRAF_ICMP_ERROR;
2288 2289 connp->conn_recvicmp(connp, mp, NULL, ira);
2289 2290 CONN_DEC_REF(connp);
2290 2291 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2291 2292 return;
2292 2293 }
2293 2294 /*
2294 2295 * No IP tunnel is interested, fallthrough and see
2295 2296 * if a raw socket will want it.
2296 2297 */
2297 2298 /* FALLTHRU */
2298 2299 default:
2299 2300 ira->ira_flags |= IRAF_ICMP_ERROR;
2300 2301 ip_fanout_proto_v4(mp, &ripha, ira);
2301 2302 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2302 2303 return;
2303 2304 }
2304 2305 /* NOTREACHED */
2305 2306 discard_pkt:
2306 2307 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2307 2308 ip1dbg(("icmp_inbound_error_fanout_v4: drop pkt\n"));
2308 2309 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2309 2310 freemsg(mp);
2310 2311 return;
2311 2312
2312 2313 truncated:
2313 2314 /* We pulled up everthing already. Must be truncated */
2314 2315 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
2315 2316 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
2316 2317 freemsg(mp);
2317 2318 }
2318 2319
2319 2320 /*
2320 2321 * Common IP options parser.
2321 2322 *
2322 2323 * Setup routine: fill in *optp with options-parsing state, then
2323 2324 * tail-call ipoptp_next to return the first option.
2324 2325 */
2325 2326 uint8_t
2326 2327 ipoptp_first(ipoptp_t *optp, ipha_t *ipha)
2327 2328 {
2328 2329 uint32_t totallen; /* total length of all options */
2329 2330
2330 2331 totallen = ipha->ipha_version_and_hdr_length -
2331 2332 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
2332 2333 totallen <<= 2;
2333 2334 optp->ipoptp_next = (uint8_t *)(&ipha[1]);
2334 2335 optp->ipoptp_end = optp->ipoptp_next + totallen;
2335 2336 optp->ipoptp_flags = 0;
2336 2337 return (ipoptp_next(optp));
2337 2338 }
2338 2339
2339 2340 /* Like above but without an ipha_t */
2340 2341 uint8_t
2341 2342 ipoptp_first2(ipoptp_t *optp, uint32_t totallen, uint8_t *opt)
2342 2343 {
2343 2344 optp->ipoptp_next = opt;
2344 2345 optp->ipoptp_end = optp->ipoptp_next + totallen;
2345 2346 optp->ipoptp_flags = 0;
2346 2347 return (ipoptp_next(optp));
2347 2348 }
2348 2349
2349 2350 /*
2350 2351 * Common IP options parser: extract next option.
2351 2352 */
2352 2353 uint8_t
2353 2354 ipoptp_next(ipoptp_t *optp)
2354 2355 {
2355 2356 uint8_t *end = optp->ipoptp_end;
2356 2357 uint8_t *cur = optp->ipoptp_next;
2357 2358 uint8_t opt, len, pointer;
2358 2359
2359 2360 /*
2360 2361 * If cur > end already, then the ipoptp_end or ipoptp_next pointer
2361 2362 * has been corrupted.
2362 2363 */
2363 2364 ASSERT(cur <= end);
2364 2365
2365 2366 if (cur == end)
2366 2367 return (IPOPT_EOL);
2367 2368
2368 2369 opt = cur[IPOPT_OPTVAL];
2369 2370
2370 2371 /*
2371 2372 * Skip any NOP options.
2372 2373 */
2373 2374 while (opt == IPOPT_NOP) {
2374 2375 cur++;
2375 2376 if (cur == end)
2376 2377 return (IPOPT_EOL);
2377 2378 opt = cur[IPOPT_OPTVAL];
2378 2379 }
2379 2380
2380 2381 if (opt == IPOPT_EOL)
2381 2382 return (IPOPT_EOL);
2382 2383
2383 2384 /*
2384 2385 * Option requiring a length.
2385 2386 */
2386 2387 if ((cur + 1) >= end) {
2387 2388 optp->ipoptp_flags |= IPOPTP_ERROR;
2388 2389 return (IPOPT_EOL);
2389 2390 }
2390 2391 len = cur[IPOPT_OLEN];
2391 2392 if (len < 2) {
2392 2393 optp->ipoptp_flags |= IPOPTP_ERROR;
2393 2394 return (IPOPT_EOL);
2394 2395 }
2395 2396 optp->ipoptp_cur = cur;
2396 2397 optp->ipoptp_len = len;
2397 2398 optp->ipoptp_next = cur + len;
2398 2399 if (cur + len > end) {
2399 2400 optp->ipoptp_flags |= IPOPTP_ERROR;
2400 2401 return (IPOPT_EOL);
2401 2402 }
2402 2403
2403 2404 /*
2404 2405 * For the options which require a pointer field, make sure
2405 2406 * its there, and make sure it points to either something
2406 2407 * inside this option, or the end of the option.
2407 2408 */
2408 2409 switch (opt) {
2409 2410 case IPOPT_RR:
2410 2411 case IPOPT_TS:
2411 2412 case IPOPT_LSRR:
2412 2413 case IPOPT_SSRR:
2413 2414 if (len <= IPOPT_OFFSET) {
2414 2415 optp->ipoptp_flags |= IPOPTP_ERROR;
2415 2416 return (opt);
2416 2417 }
2417 2418 pointer = cur[IPOPT_OFFSET];
2418 2419 if (pointer - 1 > len) {
2419 2420 optp->ipoptp_flags |= IPOPTP_ERROR;
2420 2421 return (opt);
2421 2422 }
2422 2423 break;
2423 2424 }
2424 2425
2425 2426 /*
2426 2427 * Sanity check the pointer field based on the type of the
2427 2428 * option.
2428 2429 */
2429 2430 switch (opt) {
2430 2431 case IPOPT_RR:
2431 2432 case IPOPT_SSRR:
2432 2433 case IPOPT_LSRR:
2433 2434 if (pointer < IPOPT_MINOFF_SR)
2434 2435 optp->ipoptp_flags |= IPOPTP_ERROR;
2435 2436 break;
2436 2437 case IPOPT_TS:
2437 2438 if (pointer < IPOPT_MINOFF_IT)
2438 2439 optp->ipoptp_flags |= IPOPTP_ERROR;
2439 2440 /*
2440 2441 * Note that the Internet Timestamp option also
2441 2442 * contains two four bit fields (the Overflow field,
2442 2443 * and the Flag field), which follow the pointer
2443 2444 * field. We don't need to check that these fields
2444 2445 * fall within the length of the option because this
2445 2446 * was implicitely done above. We've checked that the
2446 2447 * pointer value is at least IPOPT_MINOFF_IT, and that
2447 2448 * it falls within the option. Since IPOPT_MINOFF_IT >
2448 2449 * IPOPT_POS_OV_FLG, we don't need the explicit check.
2449 2450 */
2450 2451 ASSERT(len > IPOPT_POS_OV_FLG);
2451 2452 break;
2452 2453 }
2453 2454
2454 2455 return (opt);
2455 2456 }
2456 2457
2457 2458 /*
2458 2459 * Use the outgoing IP header to create an IP_OPTIONS option the way
2459 2460 * it was passed down from the application.
2460 2461 *
2461 2462 * This is compatible with BSD in that it returns
2462 2463 * the reverse source route with the final destination
2463 2464 * as the last entry. The first 4 bytes of the option
2464 2465 * will contain the final destination.
2465 2466 */
2466 2467 int
2467 2468 ip_opt_get_user(conn_t *connp, uchar_t *buf)
2468 2469 {
2469 2470 ipoptp_t opts;
2470 2471 uchar_t *opt;
2471 2472 uint8_t optval;
2472 2473 uint8_t optlen;
2473 2474 uint32_t len = 0;
2474 2475 uchar_t *buf1 = buf;
2475 2476 uint32_t totallen;
2476 2477 ipaddr_t dst;
2477 2478 ip_pkt_t *ipp = &connp->conn_xmit_ipp;
2478 2479
2479 2480 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
2480 2481 return (0);
2481 2482
2482 2483 totallen = ipp->ipp_ipv4_options_len;
2483 2484 if (totallen & 0x3)
2484 2485 return (0);
2485 2486
2486 2487 buf += IP_ADDR_LEN; /* Leave room for final destination */
2487 2488 len += IP_ADDR_LEN;
2488 2489 bzero(buf1, IP_ADDR_LEN);
2489 2490
2490 2491 dst = connp->conn_faddr_v4;
2491 2492
2492 2493 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
2493 2494 optval != IPOPT_EOL;
2494 2495 optval = ipoptp_next(&opts)) {
2495 2496 int off;
2496 2497
2497 2498 opt = opts.ipoptp_cur;
2498 2499 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
2499 2500 break;
2500 2501 }
2501 2502 optlen = opts.ipoptp_len;
2502 2503
2503 2504 switch (optval) {
2504 2505 case IPOPT_SSRR:
2505 2506 case IPOPT_LSRR:
2506 2507
2507 2508 /*
2508 2509 * Insert destination as the first entry in the source
2509 2510 * route and move down the entries on step.
2510 2511 * The last entry gets placed at buf1.
2511 2512 */
2512 2513 buf[IPOPT_OPTVAL] = optval;
2513 2514 buf[IPOPT_OLEN] = optlen;
2514 2515 buf[IPOPT_OFFSET] = optlen;
2515 2516
2516 2517 off = optlen - IP_ADDR_LEN;
2517 2518 if (off < 0) {
2518 2519 /* No entries in source route */
2519 2520 break;
2520 2521 }
2521 2522 /* Last entry in source route if not already set */
2522 2523 if (dst == INADDR_ANY)
2523 2524 bcopy(opt + off, buf1, IP_ADDR_LEN);
2524 2525 off -= IP_ADDR_LEN;
2525 2526
2526 2527 while (off > 0) {
2527 2528 bcopy(opt + off,
2528 2529 buf + off + IP_ADDR_LEN,
2529 2530 IP_ADDR_LEN);
2530 2531 off -= IP_ADDR_LEN;
2531 2532 }
2532 2533 /* ipha_dst into first slot */
2533 2534 bcopy(&dst, buf + off + IP_ADDR_LEN,
2534 2535 IP_ADDR_LEN);
2535 2536 buf += optlen;
2536 2537 len += optlen;
2537 2538 break;
2538 2539
2539 2540 default:
2540 2541 bcopy(opt, buf, optlen);
2541 2542 buf += optlen;
2542 2543 len += optlen;
2543 2544 break;
2544 2545 }
2545 2546 }
2546 2547 done:
2547 2548 /* Pad the resulting options */
2548 2549 while (len & 0x3) {
2549 2550 *buf++ = IPOPT_EOL;
2550 2551 len++;
2551 2552 }
2552 2553 return (len);
2553 2554 }
2554 2555
2555 2556 /*
2556 2557 * Update any record route or timestamp options to include this host.
2557 2558 * Reverse any source route option.
2558 2559 * This routine assumes that the options are well formed i.e. that they
2559 2560 * have already been checked.
2560 2561 */
2561 2562 static void
2562 2563 icmp_options_update(ipha_t *ipha)
2563 2564 {
2564 2565 ipoptp_t opts;
2565 2566 uchar_t *opt;
2566 2567 uint8_t optval;
2567 2568 ipaddr_t src; /* Our local address */
2568 2569 ipaddr_t dst;
2569 2570
2570 2571 ip2dbg(("icmp_options_update\n"));
2571 2572 src = ipha->ipha_src;
2572 2573 dst = ipha->ipha_dst;
2573 2574
2574 2575 for (optval = ipoptp_first(&opts, ipha);
2575 2576 optval != IPOPT_EOL;
2576 2577 optval = ipoptp_next(&opts)) {
2577 2578 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
2578 2579 opt = opts.ipoptp_cur;
2579 2580 ip2dbg(("icmp_options_update: opt %d, len %d\n",
2580 2581 optval, opts.ipoptp_len));
2581 2582 switch (optval) {
2582 2583 int off1, off2;
2583 2584 case IPOPT_SSRR:
2584 2585 case IPOPT_LSRR:
2585 2586 /*
2586 2587 * Reverse the source route. The first entry
2587 2588 * should be the next to last one in the current
2588 2589 * source route (the last entry is our address).
2589 2590 * The last entry should be the final destination.
2590 2591 */
2591 2592 off1 = IPOPT_MINOFF_SR - 1;
2592 2593 off2 = opt[IPOPT_OFFSET] - IP_ADDR_LEN - 1;
2593 2594 if (off2 < 0) {
2594 2595 /* No entries in source route */
2595 2596 ip1dbg((
2596 2597 "icmp_options_update: bad src route\n"));
2597 2598 break;
2598 2599 }
2599 2600 bcopy((char *)opt + off2, &dst, IP_ADDR_LEN);
2600 2601 bcopy(&ipha->ipha_dst, (char *)opt + off2, IP_ADDR_LEN);
2601 2602 bcopy(&dst, &ipha->ipha_dst, IP_ADDR_LEN);
2602 2603 off2 -= IP_ADDR_LEN;
2603 2604
2604 2605 while (off1 < off2) {
2605 2606 bcopy((char *)opt + off1, &src, IP_ADDR_LEN);
2606 2607 bcopy((char *)opt + off2, (char *)opt + off1,
2607 2608 IP_ADDR_LEN);
2608 2609 bcopy(&src, (char *)opt + off2, IP_ADDR_LEN);
2609 2610 off1 += IP_ADDR_LEN;
2610 2611 off2 -= IP_ADDR_LEN;
2611 2612 }
2612 2613 opt[IPOPT_OFFSET] = IPOPT_MINOFF_SR;
2613 2614 break;
2614 2615 }
2615 2616 }
2616 2617 }
2617 2618
2618 2619 /*
2619 2620 * Process received ICMP Redirect messages.
2620 2621 * Assumes the caller has verified that the headers are in the pulled up mblk.
2621 2622 * Consumes mp.
2622 2623 */
2623 2624 static void
2624 2625 icmp_redirect_v4(mblk_t *mp, ipha_t *ipha, icmph_t *icmph, ip_recv_attr_t *ira)
2625 2626 {
2626 2627 ire_t *ire, *nire;
2627 2628 ire_t *prev_ire;
2628 2629 ipaddr_t src, dst, gateway;
2629 2630 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2630 2631 ipha_t *inner_ipha; /* Inner IP header */
2631 2632
2632 2633 /* Caller already pulled up everything. */
2633 2634 inner_ipha = (ipha_t *)&icmph[1];
2634 2635 src = ipha->ipha_src;
2635 2636 dst = inner_ipha->ipha_dst;
2636 2637 gateway = icmph->icmph_rd_gateway;
2637 2638 /* Make sure the new gateway is reachable somehow. */
2638 2639 ire = ire_ftable_lookup_v4(gateway, 0, 0, IRE_ONLINK, NULL,
2639 2640 ALL_ZONES, NULL, MATCH_IRE_TYPE, 0, ipst, NULL);
2640 2641 /*
2641 2642 * Make sure we had a route for the dest in question and that
2642 2643 * that route was pointing to the old gateway (the source of the
2643 2644 * redirect packet.)
2644 2645 * We do longest match and then compare ire_gateway_addr below.
2645 2646 */
2646 2647 prev_ire = ire_ftable_lookup_v4(dst, 0, 0, 0, NULL, ALL_ZONES,
2647 2648 NULL, MATCH_IRE_DSTONLY, 0, ipst, NULL);
2648 2649 /*
2649 2650 * Check that
2650 2651 * the redirect was not from ourselves
2651 2652 * the new gateway and the old gateway are directly reachable
2652 2653 */
2653 2654 if (prev_ire == NULL || ire == NULL ||
2654 2655 (prev_ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK)) ||
2655 2656 (prev_ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) ||
2656 2657 !(ire->ire_type & IRE_IF_ALL) ||
2657 2658 prev_ire->ire_gateway_addr != src) {
2658 2659 BUMP_MIB(&ipst->ips_icmp_mib, icmpInBadRedirects);
2659 2660 ip_drop_input("icmpInBadRedirects - ire", mp, ira->ira_ill);
2660 2661 freemsg(mp);
2661 2662 if (ire != NULL)
2662 2663 ire_refrele(ire);
2663 2664 if (prev_ire != NULL)
2664 2665 ire_refrele(prev_ire);
2665 2666 return;
2666 2667 }
2667 2668
2668 2669 ire_refrele(prev_ire);
2669 2670 ire_refrele(ire);
2670 2671
2671 2672 /*
2672 2673 * TODO: more precise handling for cases 0, 2, 3, the latter two
2673 2674 * require TOS routing
2674 2675 */
2675 2676 switch (icmph->icmph_code) {
2676 2677 case 0:
2677 2678 case 1:
2678 2679 /* TODO: TOS specificity for cases 2 and 3 */
2679 2680 case 2:
2680 2681 case 3:
2681 2682 break;
2682 2683 default:
2683 2684 BUMP_MIB(&ipst->ips_icmp_mib, icmpInBadRedirects);
2684 2685 ip_drop_input("icmpInBadRedirects - code", mp, ira->ira_ill);
2685 2686 freemsg(mp);
2686 2687 return;
2687 2688 }
2688 2689 /*
2689 2690 * Create a Route Association. This will allow us to remember that
2690 2691 * someone we believe told us to use the particular gateway.
2691 2692 */
2692 2693 ire = ire_create(
2693 2694 (uchar_t *)&dst, /* dest addr */
2694 2695 (uchar_t *)&ip_g_all_ones, /* mask */
2695 2696 (uchar_t *)&gateway, /* gateway addr */
2696 2697 IRE_HOST,
2697 2698 NULL, /* ill */
2698 2699 ALL_ZONES,
2699 2700 (RTF_DYNAMIC | RTF_GATEWAY | RTF_HOST),
2700 2701 NULL, /* tsol_gc_t */
2701 2702 ipst);
2702 2703
2703 2704 if (ire == NULL) {
2704 2705 freemsg(mp);
2705 2706 return;
2706 2707 }
2707 2708 nire = ire_add(ire);
2708 2709 /* Check if it was a duplicate entry */
2709 2710 if (nire != NULL && nire != ire) {
2710 2711 ASSERT(nire->ire_identical_ref > 1);
2711 2712 ire_delete(nire);
2712 2713 ire_refrele(nire);
2713 2714 nire = NULL;
2714 2715 }
2715 2716 ire = nire;
2716 2717 if (ire != NULL) {
2717 2718 ire_refrele(ire); /* Held in ire_add */
2718 2719
2719 2720 /* tell routing sockets that we received a redirect */
2720 2721 ip_rts_change(RTM_REDIRECT, dst, gateway, IP_HOST_MASK, 0, src,
2721 2722 (RTF_DYNAMIC | RTF_GATEWAY | RTF_HOST), 0,
2722 2723 (RTA_DST | RTA_GATEWAY | RTA_NETMASK | RTA_AUTHOR), ipst);
2723 2724 }
2724 2725
2725 2726 /*
2726 2727 * Delete any existing IRE_HOST type redirect ires for this destination.
2727 2728 * This together with the added IRE has the effect of
2728 2729 * modifying an existing redirect.
2729 2730 */
2730 2731 prev_ire = ire_ftable_lookup_v4(dst, 0, src, IRE_HOST, NULL,
2731 2732 ALL_ZONES, NULL, (MATCH_IRE_GW | MATCH_IRE_TYPE), 0, ipst, NULL);
2732 2733 if (prev_ire != NULL) {
2733 2734 if (prev_ire ->ire_flags & RTF_DYNAMIC)
2734 2735 ire_delete(prev_ire);
2735 2736 ire_refrele(prev_ire);
2736 2737 }
2737 2738
2738 2739 freemsg(mp);
2739 2740 }
2740 2741
2741 2742 /*
2742 2743 * Generate an ICMP parameter problem message.
2743 2744 * When called from ip_output side a minimal ip_recv_attr_t needs to be
2744 2745 * constructed by the caller.
2745 2746 */
2746 2747 static void
2747 2748 icmp_param_problem(mblk_t *mp, uint8_t ptr, ip_recv_attr_t *ira)
2748 2749 {
2749 2750 icmph_t icmph;
2750 2751 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2751 2752
2752 2753 mp = icmp_pkt_err_ok(mp, ira);
2753 2754 if (mp == NULL)
2754 2755 return;
2755 2756
2756 2757 bzero(&icmph, sizeof (icmph_t));
2757 2758 icmph.icmph_type = ICMP_PARAM_PROBLEM;
2758 2759 icmph.icmph_pp_ptr = ptr;
2759 2760 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutParmProbs);
2760 2761 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
2761 2762 }
2762 2763
2763 2764 /*
2764 2765 * Build and ship an IPv4 ICMP message using the packet data in mp, and
2765 2766 * the ICMP header pointed to by "stuff". (May be called as writer.)
2766 2767 * Note: assumes that icmp_pkt_err_ok has been called to verify that
2767 2768 * an icmp error packet can be sent.
2768 2769 * Assigns an appropriate source address to the packet. If ipha_dst is
2769 2770 * one of our addresses use it for source. Otherwise let ip_output_simple
2770 2771 * pick the source address.
2771 2772 */
2772 2773 static void
2773 2774 icmp_pkt(mblk_t *mp, void *stuff, size_t len, ip_recv_attr_t *ira)
2774 2775 {
2775 2776 ipaddr_t dst;
2776 2777 icmph_t *icmph;
2777 2778 ipha_t *ipha;
2778 2779 uint_t len_needed;
2779 2780 size_t msg_len;
2780 2781 mblk_t *mp1;
2781 2782 ipaddr_t src;
2782 2783 ire_t *ire;
2783 2784 ip_xmit_attr_t ixas;
2784 2785 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2785 2786
2786 2787 ipha = (ipha_t *)mp->b_rptr;
2787 2788
2788 2789 bzero(&ixas, sizeof (ixas));
2789 2790 ixas.ixa_flags = IXAF_BASIC_SIMPLE_V4;
2790 2791 ixas.ixa_zoneid = ira->ira_zoneid;
2791 2792 ixas.ixa_ifindex = 0;
2792 2793 ixas.ixa_ipst = ipst;
2793 2794 ixas.ixa_cred = kcred;
2794 2795 ixas.ixa_cpid = NOPID;
2795 2796 ixas.ixa_tsl = ira->ira_tsl; /* Behave as a multi-level responder */
2796 2797 ixas.ixa_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
2797 2798
2798 2799 if (ira->ira_flags & IRAF_IPSEC_SECURE) {
2799 2800 /*
2800 2801 * Apply IPsec based on how IPsec was applied to
2801 2802 * the packet that had the error.
2802 2803 *
2803 2804 * If it was an outbound packet that caused the ICMP
2804 2805 * error, then the caller will have setup the IRA
2805 2806 * appropriately.
2806 2807 */
2807 2808 if (!ipsec_in_to_out(ira, &ixas, mp, ipha, NULL)) {
2808 2809 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
2809 2810 /* Note: mp already consumed and ip_drop_packet done */
2810 2811 return;
2811 2812 }
2812 2813 } else {
2813 2814 /*
2814 2815 * This is in clear. The icmp message we are building
2815 2816 * here should go out in clear, independent of our policy.
2816 2817 */
2817 2818 ixas.ixa_flags |= IXAF_NO_IPSEC;
2818 2819 }
2819 2820
2820 2821 /* Remember our eventual destination */
2821 2822 dst = ipha->ipha_src;
2822 2823
2823 2824 /*
2824 2825 * If the packet was for one of our unicast addresses, make
2825 2826 * sure we respond with that as the source. Otherwise
2826 2827 * have ip_output_simple pick the source address.
2827 2828 */
2828 2829 ire = ire_ftable_lookup_v4(ipha->ipha_dst, 0, 0,
2829 2830 (IRE_LOCAL|IRE_LOOPBACK), NULL, ira->ira_zoneid, NULL,
2830 2831 MATCH_IRE_TYPE|MATCH_IRE_ZONEONLY, 0, ipst, NULL);
2831 2832 if (ire != NULL) {
2832 2833 ire_refrele(ire);
2833 2834 src = ipha->ipha_dst;
2834 2835 } else {
2835 2836 src = INADDR_ANY;
2836 2837 ixas.ixa_flags |= IXAF_SET_SOURCE;
2837 2838 }
2838 2839
2839 2840 /*
2840 2841 * Check if we can send back more then 8 bytes in addition to
2841 2842 * the IP header. We try to send 64 bytes of data and the internal
2842 2843 * header in the special cases of ipv4 encapsulated ipv4 or ipv6.
2843 2844 */
2844 2845 len_needed = IPH_HDR_LENGTH(ipha);
2845 2846 if (ipha->ipha_protocol == IPPROTO_ENCAP ||
2846 2847 ipha->ipha_protocol == IPPROTO_IPV6) {
2847 2848 if (!pullupmsg(mp, -1)) {
2848 2849 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
2849 2850 ip_drop_output("ipIfStatsOutDiscards", mp, NULL);
2850 2851 freemsg(mp);
2851 2852 return;
2852 2853 }
2853 2854 ipha = (ipha_t *)mp->b_rptr;
2854 2855
2855 2856 if (ipha->ipha_protocol == IPPROTO_ENCAP) {
2856 2857 len_needed += IPH_HDR_LENGTH(((uchar_t *)ipha +
2857 2858 len_needed));
2858 2859 } else {
2859 2860 ip6_t *ip6h = (ip6_t *)((uchar_t *)ipha + len_needed);
2860 2861
2861 2862 ASSERT(ipha->ipha_protocol == IPPROTO_IPV6);
2862 2863 len_needed += ip_hdr_length_v6(mp, ip6h);
2863 2864 }
2864 2865 }
2865 2866 len_needed += ipst->ips_ip_icmp_return;
2866 2867 msg_len = msgdsize(mp);
2867 2868 if (msg_len > len_needed) {
2868 2869 (void) adjmsg(mp, len_needed - msg_len);
2869 2870 msg_len = len_needed;
2870 2871 }
2871 2872 mp1 = allocb(sizeof (icmp_ipha) + len, BPRI_MED);
2872 2873 if (mp1 == NULL) {
2873 2874 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutErrors);
2874 2875 freemsg(mp);
2875 2876 return;
2876 2877 }
2877 2878 mp1->b_cont = mp;
2878 2879 mp = mp1;
2879 2880
2880 2881 /*
2881 2882 * Set IXAF_TRUSTED_ICMP so we can let the ICMP messages this
2882 2883 * node generates be accepted in peace by all on-host destinations.
2883 2884 * If we do NOT assume that all on-host destinations trust
2884 2885 * self-generated ICMP messages, then rework here, ip6.c, and spd.c.
2885 2886 * (Look for IXAF_TRUSTED_ICMP).
2886 2887 */
2887 2888 ixas.ixa_flags |= IXAF_TRUSTED_ICMP;
2888 2889
2889 2890 ipha = (ipha_t *)mp->b_rptr;
2890 2891 mp1->b_wptr = (uchar_t *)ipha + (sizeof (icmp_ipha) + len);
2891 2892 *ipha = icmp_ipha;
2892 2893 ipha->ipha_src = src;
2893 2894 ipha->ipha_dst = dst;
2894 2895 ipha->ipha_ttl = ipst->ips_ip_def_ttl;
2895 2896 msg_len += sizeof (icmp_ipha) + len;
2896 2897 if (msg_len > IP_MAXPACKET) {
2897 2898 (void) adjmsg(mp, IP_MAXPACKET - msg_len);
2898 2899 msg_len = IP_MAXPACKET;
2899 2900 }
2900 2901 ipha->ipha_length = htons((uint16_t)msg_len);
2901 2902 icmph = (icmph_t *)&ipha[1];
2902 2903 bcopy(stuff, icmph, len);
2903 2904 icmph->icmph_checksum = 0;
2904 2905 icmph->icmph_checksum = IP_CSUM(mp, (int32_t)sizeof (ipha_t), 0);
2905 2906 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutMsgs);
2906 2907
2907 2908 (void) ip_output_simple(mp, &ixas);
2908 2909 ixa_cleanup(&ixas);
2909 2910 }
2910 2911
2911 2912 /*
2912 2913 * Determine if an ICMP error packet can be sent given the rate limit.
2913 2914 * The limit consists of an average frequency (icmp_pkt_err_interval measured
2914 2915 * in milliseconds) and a burst size. Burst size number of packets can
2915 2916 * be sent arbitrarely closely spaced.
2916 2917 * The state is tracked using two variables to implement an approximate
2917 2918 * token bucket filter:
2918 2919 * icmp_pkt_err_last - lbolt value when the last burst started
2919 2920 * icmp_pkt_err_sent - number of packets sent in current burst
2920 2921 */
2921 2922 boolean_t
2922 2923 icmp_err_rate_limit(ip_stack_t *ipst)
2923 2924 {
2924 2925 clock_t now = TICK_TO_MSEC(ddi_get_lbolt());
2925 2926 uint_t refilled; /* Number of packets refilled in tbf since last */
2926 2927 /* Guard against changes by loading into local variable */
2927 2928 uint_t err_interval = ipst->ips_ip_icmp_err_interval;
2928 2929
2929 2930 if (err_interval == 0)
2930 2931 return (B_FALSE);
2931 2932
2932 2933 if (ipst->ips_icmp_pkt_err_last > now) {
2933 2934 /* 100HZ lbolt in ms for 32bit arch wraps every 49.7 days */
2934 2935 ipst->ips_icmp_pkt_err_last = 0;
2935 2936 ipst->ips_icmp_pkt_err_sent = 0;
2936 2937 }
2937 2938 /*
2938 2939 * If we are in a burst update the token bucket filter.
2939 2940 * Update the "last" time to be close to "now" but make sure
2940 2941 * we don't loose precision.
2941 2942 */
2942 2943 if (ipst->ips_icmp_pkt_err_sent != 0) {
2943 2944 refilled = (now - ipst->ips_icmp_pkt_err_last)/err_interval;
2944 2945 if (refilled > ipst->ips_icmp_pkt_err_sent) {
2945 2946 ipst->ips_icmp_pkt_err_sent = 0;
2946 2947 } else {
2947 2948 ipst->ips_icmp_pkt_err_sent -= refilled;
2948 2949 ipst->ips_icmp_pkt_err_last += refilled * err_interval;
2949 2950 }
2950 2951 }
2951 2952 if (ipst->ips_icmp_pkt_err_sent == 0) {
2952 2953 /* Start of new burst */
2953 2954 ipst->ips_icmp_pkt_err_last = now;
2954 2955 }
2955 2956 if (ipst->ips_icmp_pkt_err_sent < ipst->ips_ip_icmp_err_burst) {
2956 2957 ipst->ips_icmp_pkt_err_sent++;
2957 2958 ip1dbg(("icmp_err_rate_limit: %d sent in burst\n",
2958 2959 ipst->ips_icmp_pkt_err_sent));
2959 2960 return (B_FALSE);
2960 2961 }
2961 2962 ip1dbg(("icmp_err_rate_limit: dropped\n"));
2962 2963 return (B_TRUE);
2963 2964 }
2964 2965
2965 2966 /*
2966 2967 * Check if it is ok to send an IPv4 ICMP error packet in
2967 2968 * response to the IPv4 packet in mp.
2968 2969 * Free the message and return null if no
2969 2970 * ICMP error packet should be sent.
2970 2971 */
2971 2972 static mblk_t *
2972 2973 icmp_pkt_err_ok(mblk_t *mp, ip_recv_attr_t *ira)
2973 2974 {
2974 2975 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2975 2976 icmph_t *icmph;
2976 2977 ipha_t *ipha;
2977 2978 uint_t len_needed;
2978 2979
2979 2980 if (!mp)
2980 2981 return (NULL);
2981 2982 ipha = (ipha_t *)mp->b_rptr;
2982 2983 if (ip_csum_hdr(ipha)) {
2983 2984 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsInCksumErrs);
2984 2985 ip_drop_input("ipIfStatsInCksumErrs", mp, NULL);
2985 2986 freemsg(mp);
2986 2987 return (NULL);
2987 2988 }
2988 2989 if (ip_type_v4(ipha->ipha_dst, ipst) == IRE_BROADCAST ||
2989 2990 ip_type_v4(ipha->ipha_src, ipst) == IRE_BROADCAST ||
2990 2991 CLASSD(ipha->ipha_dst) ||
2991 2992 CLASSD(ipha->ipha_src) ||
2992 2993 (ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_OFFSET)) {
2993 2994 /* Note: only errors to the fragment with offset 0 */
2994 2995 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
2995 2996 freemsg(mp);
2996 2997 return (NULL);
2997 2998 }
2998 2999 if (ipha->ipha_protocol == IPPROTO_ICMP) {
2999 3000 /*
3000 3001 * Check the ICMP type. RFC 1122 sez: don't send ICMP
3001 3002 * errors in response to any ICMP errors.
3002 3003 */
3003 3004 len_needed = IPH_HDR_LENGTH(ipha) + ICMPH_SIZE;
3004 3005 if (mp->b_wptr - mp->b_rptr < len_needed) {
3005 3006 if (!pullupmsg(mp, len_needed)) {
3006 3007 BUMP_MIB(&ipst->ips_icmp_mib, icmpInErrors);
3007 3008 freemsg(mp);
3008 3009 return (NULL);
3009 3010 }
3010 3011 ipha = (ipha_t *)mp->b_rptr;
3011 3012 }
3012 3013 icmph = (icmph_t *)
3013 3014 (&((char *)ipha)[IPH_HDR_LENGTH(ipha)]);
3014 3015 switch (icmph->icmph_type) {
3015 3016 case ICMP_DEST_UNREACHABLE:
3016 3017 case ICMP_SOURCE_QUENCH:
3017 3018 case ICMP_TIME_EXCEEDED:
3018 3019 case ICMP_PARAM_PROBLEM:
3019 3020 case ICMP_REDIRECT:
3020 3021 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
3021 3022 freemsg(mp);
3022 3023 return (NULL);
3023 3024 default:
3024 3025 break;
3025 3026 }
3026 3027 }
3027 3028 /*
3028 3029 * If this is a labeled system, then check to see if we're allowed to
3029 3030 * send a response to this particular sender. If not, then just drop.
3030 3031 */
3031 3032 if (is_system_labeled() && !tsol_can_reply_error(mp, ira)) {
3032 3033 ip2dbg(("icmp_pkt_err_ok: can't respond to packet\n"));
3033 3034 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
3034 3035 freemsg(mp);
3035 3036 return (NULL);
3036 3037 }
3037 3038 if (icmp_err_rate_limit(ipst)) {
3038 3039 /*
3039 3040 * Only send ICMP error packets every so often.
3040 3041 * This should be done on a per port/source basis,
3041 3042 * but for now this will suffice.
3042 3043 */
3043 3044 freemsg(mp);
3044 3045 return (NULL);
3045 3046 }
3046 3047 return (mp);
3047 3048 }
3048 3049
3049 3050 /*
3050 3051 * Called when a packet was sent out the same link that it arrived on.
3051 3052 * Check if it is ok to send a redirect and then send it.
3052 3053 */
3053 3054 void
3054 3055 ip_send_potential_redirect_v4(mblk_t *mp, ipha_t *ipha, ire_t *ire,
3055 3056 ip_recv_attr_t *ira)
3056 3057 {
3057 3058 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3058 3059 ipaddr_t src, nhop;
3059 3060 mblk_t *mp1;
3060 3061 ire_t *nhop_ire;
3061 3062
3062 3063 /*
3063 3064 * Check the source address to see if it originated
3064 3065 * on the same logical subnet it is going back out on.
3065 3066 * If so, we should be able to send it a redirect.
3066 3067 * Avoid sending a redirect if the destination
3067 3068 * is directly connected (i.e., we matched an IRE_ONLINK),
3068 3069 * or if the packet was source routed out this interface.
3069 3070 *
3070 3071 * We avoid sending a redirect if the
3071 3072 * destination is directly connected
3072 3073 * because it is possible that multiple
3073 3074 * IP subnets may have been configured on
3074 3075 * the link, and the source may not
3075 3076 * be on the same subnet as ip destination,
3076 3077 * even though they are on the same
3077 3078 * physical link.
3078 3079 */
3079 3080 if ((ire->ire_type & IRE_ONLINK) ||
3080 3081 ip_source_routed(ipha, ipst))
3081 3082 return;
3082 3083
3083 3084 nhop_ire = ire_nexthop(ire);
3084 3085 if (nhop_ire == NULL)
3085 3086 return;
3086 3087
3087 3088 nhop = nhop_ire->ire_addr;
3088 3089
3089 3090 if (nhop_ire->ire_type & IRE_IF_CLONE) {
3090 3091 ire_t *ire2;
3091 3092
3092 3093 /* Follow ire_dep_parent to find non-clone IRE_INTERFACE */
3093 3094 mutex_enter(&nhop_ire->ire_lock);
3094 3095 ire2 = nhop_ire->ire_dep_parent;
3095 3096 if (ire2 != NULL)
3096 3097 ire_refhold(ire2);
3097 3098 mutex_exit(&nhop_ire->ire_lock);
3098 3099 ire_refrele(nhop_ire);
3099 3100 nhop_ire = ire2;
3100 3101 }
3101 3102 if (nhop_ire == NULL)
3102 3103 return;
3103 3104
3104 3105 ASSERT(!(nhop_ire->ire_type & IRE_IF_CLONE));
3105 3106
3106 3107 src = ipha->ipha_src;
3107 3108
3108 3109 /*
3109 3110 * We look at the interface ire for the nexthop,
3110 3111 * to see if ipha_src is in the same subnet
3111 3112 * as the nexthop.
3112 3113 */
3113 3114 if ((src & nhop_ire->ire_mask) == (nhop & nhop_ire->ire_mask)) {
3114 3115 /*
3115 3116 * The source is directly connected.
3116 3117 */
3117 3118 mp1 = copymsg(mp);
3118 3119 if (mp1 != NULL) {
3119 3120 icmp_send_redirect(mp1, nhop, ira);
3120 3121 }
3121 3122 }
3122 3123 ire_refrele(nhop_ire);
3123 3124 }
3124 3125
3125 3126 /*
3126 3127 * Generate an ICMP redirect message.
3127 3128 */
3128 3129 static void
3129 3130 icmp_send_redirect(mblk_t *mp, ipaddr_t gateway, ip_recv_attr_t *ira)
3130 3131 {
3131 3132 icmph_t icmph;
3132 3133 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3133 3134
3134 3135 mp = icmp_pkt_err_ok(mp, ira);
3135 3136 if (mp == NULL)
3136 3137 return;
3137 3138
3138 3139 bzero(&icmph, sizeof (icmph_t));
3139 3140 icmph.icmph_type = ICMP_REDIRECT;
3140 3141 icmph.icmph_code = 1;
3141 3142 icmph.icmph_rd_gateway = gateway;
3142 3143 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutRedirects);
3143 3144 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3144 3145 }
3145 3146
3146 3147 /*
3147 3148 * Generate an ICMP time exceeded message.
3148 3149 */
3149 3150 void
3150 3151 icmp_time_exceeded(mblk_t *mp, uint8_t code, ip_recv_attr_t *ira)
3151 3152 {
3152 3153 icmph_t icmph;
3153 3154 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3154 3155
3155 3156 mp = icmp_pkt_err_ok(mp, ira);
3156 3157 if (mp == NULL)
3157 3158 return;
3158 3159
3159 3160 bzero(&icmph, sizeof (icmph_t));
3160 3161 icmph.icmph_type = ICMP_TIME_EXCEEDED;
3161 3162 icmph.icmph_code = code;
3162 3163 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutTimeExcds);
3163 3164 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3164 3165 }
3165 3166
3166 3167 /*
3167 3168 * Generate an ICMP unreachable message.
3168 3169 * When called from ip_output side a minimal ip_recv_attr_t needs to be
3169 3170 * constructed by the caller.
3170 3171 */
3171 3172 void
3172 3173 icmp_unreachable(mblk_t *mp, uint8_t code, ip_recv_attr_t *ira)
3173 3174 {
3174 3175 icmph_t icmph;
3175 3176 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3176 3177
3177 3178 mp = icmp_pkt_err_ok(mp, ira);
3178 3179 if (mp == NULL)
3179 3180 return;
3180 3181
3181 3182 bzero(&icmph, sizeof (icmph_t));
3182 3183 icmph.icmph_type = ICMP_DEST_UNREACHABLE;
3183 3184 icmph.icmph_code = code;
3184 3185 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDestUnreachs);
3185 3186 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3186 3187 }
3187 3188
3188 3189 /*
3189 3190 * Latch in the IPsec state for a stream based the policy in the listener
3190 3191 * and the actions in the ip_recv_attr_t.
3191 3192 * Called directly from TCP and SCTP.
3192 3193 */
3193 3194 boolean_t
3194 3195 ip_ipsec_policy_inherit(conn_t *connp, conn_t *lconnp, ip_recv_attr_t *ira)
3195 3196 {
3196 3197 ASSERT(lconnp->conn_policy != NULL);
3197 3198 ASSERT(connp->conn_policy == NULL);
3198 3199
3199 3200 IPPH_REFHOLD(lconnp->conn_policy);
3200 3201 connp->conn_policy = lconnp->conn_policy;
3201 3202
3202 3203 if (ira->ira_ipsec_action != NULL) {
3203 3204 if (connp->conn_latch == NULL) {
3204 3205 connp->conn_latch = iplatch_create();
3205 3206 if (connp->conn_latch == NULL)
3206 3207 return (B_FALSE);
3207 3208 }
3208 3209 ipsec_latch_inbound(connp, ira);
3209 3210 }
3210 3211 return (B_TRUE);
3211 3212 }
3212 3213
3213 3214 /*
3214 3215 * Verify whether or not the IP address is a valid local address.
3215 3216 * Could be a unicast, including one for a down interface.
3216 3217 * If allow_mcbc then a multicast or broadcast address is also
3217 3218 * acceptable.
3218 3219 *
3219 3220 * In the case of a broadcast/multicast address, however, the
3220 3221 * upper protocol is expected to reset the src address
3221 3222 * to zero when we return IPVL_MCAST/IPVL_BCAST so that
3222 3223 * no packets are emitted with broadcast/multicast address as
3223 3224 * source address (that violates hosts requirements RFC 1122)
3224 3225 * The addresses valid for bind are:
3225 3226 * (1) - INADDR_ANY (0)
3226 3227 * (2) - IP address of an UP interface
3227 3228 * (3) - IP address of a DOWN interface
3228 3229 * (4) - valid local IP broadcast addresses. In this case
3229 3230 * the conn will only receive packets destined to
3230 3231 * the specified broadcast address.
3231 3232 * (5) - a multicast address. In this case
3232 3233 * the conn will only receive packets destined to
3233 3234 * the specified multicast address. Note: the
3234 3235 * application still has to issue an
3235 3236 * IP_ADD_MEMBERSHIP socket option.
3236 3237 *
3237 3238 * In all the above cases, the bound address must be valid in the current zone.
3238 3239 * When the address is loopback, multicast or broadcast, there might be many
3239 3240 * matching IREs so bind has to look up based on the zone.
3240 3241 */
3241 3242 ip_laddr_t
3242 3243 ip_laddr_verify_v4(ipaddr_t src_addr, zoneid_t zoneid,
3243 3244 ip_stack_t *ipst, boolean_t allow_mcbc)
3244 3245 {
3245 3246 ire_t *src_ire;
3246 3247
3247 3248 ASSERT(src_addr != INADDR_ANY);
3248 3249
3249 3250 src_ire = ire_ftable_lookup_v4(src_addr, 0, 0, 0,
3250 3251 NULL, zoneid, NULL, MATCH_IRE_ZONEONLY, 0, ipst, NULL);
3251 3252
3252 3253 /*
3253 3254 * If an address other than in6addr_any is requested,
3254 3255 * we verify that it is a valid address for bind
3255 3256 * Note: Following code is in if-else-if form for
3256 3257 * readability compared to a condition check.
3257 3258 */
3258 3259 if (src_ire != NULL && (src_ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK))) {
3259 3260 /*
3260 3261 * (2) Bind to address of local UP interface
3261 3262 */
3262 3263 ire_refrele(src_ire);
3263 3264 return (IPVL_UNICAST_UP);
3264 3265 } else if (src_ire != NULL && src_ire->ire_type & IRE_BROADCAST) {
3265 3266 /*
3266 3267 * (4) Bind to broadcast address
3267 3268 */
3268 3269 ire_refrele(src_ire);
3269 3270 if (allow_mcbc)
3270 3271 return (IPVL_BCAST);
3271 3272 else
3272 3273 return (IPVL_BAD);
3273 3274 } else if (CLASSD(src_addr)) {
3274 3275 /* (5) bind to multicast address. */
3275 3276 if (src_ire != NULL)
3276 3277 ire_refrele(src_ire);
3277 3278
3278 3279 if (allow_mcbc)
3279 3280 return (IPVL_MCAST);
3280 3281 else
3281 3282 return (IPVL_BAD);
3282 3283 } else {
3283 3284 ipif_t *ipif;
3284 3285
3285 3286 /*
3286 3287 * (3) Bind to address of local DOWN interface?
3287 3288 * (ipif_lookup_addr() looks up all interfaces
3288 3289 * but we do not get here for UP interfaces
3289 3290 * - case (2) above)
3290 3291 */
3291 3292 if (src_ire != NULL)
3292 3293 ire_refrele(src_ire);
3293 3294
3294 3295 ipif = ipif_lookup_addr(src_addr, NULL, zoneid, ipst);
3295 3296 if (ipif == NULL)
3296 3297 return (IPVL_BAD);
3297 3298
3298 3299 /* Not a useful source? */
3299 3300 if (ipif->ipif_flags & (IPIF_NOLOCAL | IPIF_ANYCAST)) {
3300 3301 ipif_refrele(ipif);
3301 3302 return (IPVL_BAD);
3302 3303 }
3303 3304 ipif_refrele(ipif);
3304 3305 return (IPVL_UNICAST_DOWN);
3305 3306 }
3306 3307 }
3307 3308
3308 3309 /*
3309 3310 * Insert in the bind fanout for IPv4 and IPv6.
3310 3311 * The caller should already have used ip_laddr_verify_v*() before calling
3311 3312 * this.
3312 3313 */
3313 3314 int
3314 3315 ip_laddr_fanout_insert(conn_t *connp)
3315 3316 {
3316 3317 int error;
3317 3318
3318 3319 /*
3319 3320 * Allow setting new policies. For example, disconnects result
3320 3321 * in us being called. As we would have set conn_policy_cached
3321 3322 * to B_TRUE before, we should set it to B_FALSE, so that policy
3322 3323 * can change after the disconnect.
3323 3324 */
3324 3325 connp->conn_policy_cached = B_FALSE;
3325 3326
3326 3327 error = ipcl_bind_insert(connp);
3327 3328 if (error != 0) {
3328 3329 if (connp->conn_anon_port) {
3329 3330 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
3330 3331 connp->conn_mlp_type, connp->conn_proto,
3331 3332 ntohs(connp->conn_lport), B_FALSE);
3332 3333 }
3333 3334 connp->conn_mlp_type = mlptSingle;
3334 3335 }
3335 3336 return (error);
3336 3337 }
3337 3338
3338 3339 /*
3339 3340 * Verify that both the source and destination addresses are valid. If
3340 3341 * IPDF_VERIFY_DST is not set, then the destination address may be unreachable,
3341 3342 * i.e. have no route to it. Protocols like TCP want to verify destination
3342 3343 * reachability, while tunnels do not.
3343 3344 *
3344 3345 * Determine the route, the interface, and (optionally) the source address
3345 3346 * to use to reach a given destination.
3346 3347 * Note that we allow connect to broadcast and multicast addresses when
3347 3348 * IPDF_ALLOW_MCBC is set.
3348 3349 * first_hop and dst_addr are normally the same, but if source routing
3349 3350 * they will differ; in that case the first_hop is what we'll use for the
3350 3351 * routing lookup but the dce and label checks will be done on dst_addr,
3351 3352 *
3352 3353 * If uinfo is set, then we fill in the best available information
3353 3354 * we have for the destination. This is based on (in priority order) any
3354 3355 * metrics and path MTU stored in a dce_t, route metrics, and finally the
3355 3356 * ill_mtu/ill_mc_mtu.
3356 3357 *
3357 3358 * Tsol note: If we have a source route then dst_addr != firsthop. But we
3358 3359 * always do the label check on dst_addr.
3359 3360 */
3360 3361 int
3361 3362 ip_set_destination_v4(ipaddr_t *src_addrp, ipaddr_t dst_addr, ipaddr_t firsthop,
3362 3363 ip_xmit_attr_t *ixa, iulp_t *uinfo, uint32_t flags, uint_t mac_mode)
3363 3364 {
3364 3365 ire_t *ire = NULL;
3365 3366 int error = 0;
3366 3367 ipaddr_t setsrc; /* RTF_SETSRC */
3367 3368 zoneid_t zoneid = ixa->ixa_zoneid; /* Honors SO_ALLZONES */
3368 3369 ip_stack_t *ipst = ixa->ixa_ipst;
3369 3370 dce_t *dce;
3370 3371 uint_t pmtu;
3371 3372 uint_t generation;
3372 3373 nce_t *nce;
3373 3374 ill_t *ill = NULL;
3374 3375 boolean_t multirt = B_FALSE;
3375 3376
3376 3377 ASSERT(ixa->ixa_flags & IXAF_IS_IPV4);
3377 3378
3378 3379 /*
3379 3380 * We never send to zero; the ULPs map it to the loopback address.
3380 3381 * We can't allow it since we use zero to mean unitialized in some
3381 3382 * places.
3382 3383 */
3383 3384 ASSERT(dst_addr != INADDR_ANY);
3384 3385
3385 3386 if (is_system_labeled()) {
3386 3387 ts_label_t *tsl = NULL;
3387 3388
3388 3389 error = tsol_check_dest(ixa->ixa_tsl, &dst_addr, IPV4_VERSION,
3389 3390 mac_mode, (flags & IPDF_ZONE_IS_GLOBAL) != 0, &tsl);
3390 3391 if (error != 0)
3391 3392 return (error);
3392 3393 if (tsl != NULL) {
3393 3394 /* Update the label */
3394 3395 ip_xmit_attr_replace_tsl(ixa, tsl);
3395 3396 }
3396 3397 }
3397 3398
3398 3399 setsrc = INADDR_ANY;
3399 3400 /*
3400 3401 * Select a route; For IPMP interfaces, we would only select
3401 3402 * a "hidden" route (i.e., going through a specific under_ill)
3402 3403 * if ixa_ifindex has been specified.
3403 3404 */
3404 3405 ire = ip_select_route_v4(firsthop, *src_addrp, ixa,
3405 3406 &generation, &setsrc, &error, &multirt);
3406 3407 ASSERT(ire != NULL); /* IRE_NOROUTE if none found */
3407 3408 if (error != 0)
3408 3409 goto bad_addr;
3409 3410
3410 3411 /*
3411 3412 * ire can't be a broadcast or multicast unless IPDF_ALLOW_MCBC is set.
3412 3413 * If IPDF_VERIFY_DST is set, the destination must be reachable;
3413 3414 * Otherwise the destination needn't be reachable.
3414 3415 *
3415 3416 * If we match on a reject or black hole, then we've got a
3416 3417 * local failure. May as well fail out the connect() attempt,
3417 3418 * since it's never going to succeed.
3418 3419 */
3419 3420 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
3420 3421 /*
3421 3422 * If we're verifying destination reachability, we always want
3422 3423 * to complain here.
3423 3424 *
3424 3425 * If we're not verifying destination reachability but the
3425 3426 * destination has a route, we still want to fail on the
3426 3427 * temporary address and broadcast address tests.
3427 3428 *
3428 3429 * In both cases do we let the code continue so some reasonable
3429 3430 * information is returned to the caller. That enables the
3430 3431 * caller to use (and even cache) the IRE. conn_ip_ouput will
3431 3432 * use the generation mismatch path to check for the unreachable
3432 3433 * case thereby avoiding any specific check in the main path.
3433 3434 */
3434 3435 ASSERT(generation == IRE_GENERATION_VERIFY);
3435 3436 if (flags & IPDF_VERIFY_DST) {
3436 3437 /*
3437 3438 * Set errno but continue to set up ixa_ire to be
3438 3439 * the RTF_REJECT|RTF_BLACKHOLE IRE.
3439 3440 * That allows callers to use ip_output to get an
3440 3441 * ICMP error back.
3441 3442 */
3442 3443 if (!(ire->ire_type & IRE_HOST))
3443 3444 error = ENETUNREACH;
3444 3445 else
3445 3446 error = EHOSTUNREACH;
3446 3447 }
3447 3448 }
3448 3449
3449 3450 if ((ire->ire_type & (IRE_BROADCAST|IRE_MULTICAST)) &&
3450 3451 !(flags & IPDF_ALLOW_MCBC)) {
3451 3452 ire_refrele(ire);
3452 3453 ire = ire_reject(ipst, B_FALSE);
3453 3454 generation = IRE_GENERATION_VERIFY;
3454 3455 error = ENETUNREACH;
3455 3456 }
3456 3457
3457 3458 /* Cache things */
3458 3459 if (ixa->ixa_ire != NULL)
3459 3460 ire_refrele_notr(ixa->ixa_ire);
3460 3461 #ifdef DEBUG
3461 3462 ire_refhold_notr(ire);
3462 3463 ire_refrele(ire);
3463 3464 #endif
3464 3465 ixa->ixa_ire = ire;
3465 3466 ixa->ixa_ire_generation = generation;
3466 3467
3467 3468 /*
3468 3469 * Ensure that ixa_dce is always set any time that ixa_ire is set,
3469 3470 * since some callers will send a packet to conn_ip_output() even if
3470 3471 * there's an error.
3471 3472 */
3472 3473 if (flags & IPDF_UNIQUE_DCE) {
3473 3474 /* Fallback to the default dce if allocation fails */
3474 3475 dce = dce_lookup_and_add_v4(dst_addr, ipst);
3475 3476 if (dce != NULL)
3476 3477 generation = dce->dce_generation;
3477 3478 else
3478 3479 dce = dce_lookup_v4(dst_addr, ipst, &generation);
3479 3480 } else {
3480 3481 dce = dce_lookup_v4(dst_addr, ipst, &generation);
3481 3482 }
3482 3483 ASSERT(dce != NULL);
3483 3484 if (ixa->ixa_dce != NULL)
3484 3485 dce_refrele_notr(ixa->ixa_dce);
3485 3486 #ifdef DEBUG
3486 3487 dce_refhold_notr(dce);
3487 3488 dce_refrele(dce);
3488 3489 #endif
3489 3490 ixa->ixa_dce = dce;
3490 3491 ixa->ixa_dce_generation = generation;
3491 3492
3492 3493 /*
3493 3494 * For multicast with multirt we have a flag passed back from
3494 3495 * ire_lookup_multi_ill_v4 since we don't have an IRE for each
3495 3496 * possible multicast address.
3496 3497 * We also need a flag for multicast since we can't check
3497 3498 * whether RTF_MULTIRT is set in ixa_ire for multicast.
3498 3499 */
3499 3500 if (multirt) {
3500 3501 ixa->ixa_postfragfn = ip_postfrag_multirt_v4;
3501 3502 ixa->ixa_flags |= IXAF_MULTIRT_MULTICAST;
3502 3503 } else {
3503 3504 ixa->ixa_postfragfn = ire->ire_postfragfn;
3504 3505 ixa->ixa_flags &= ~IXAF_MULTIRT_MULTICAST;
3505 3506 }
3506 3507 if (!(ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))) {
3507 3508 /* Get an nce to cache. */
3508 3509 nce = ire_to_nce(ire, firsthop, NULL);
3509 3510 if (nce == NULL) {
3510 3511 /* Allocation failure? */
3511 3512 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3512 3513 } else {
3513 3514 if (ixa->ixa_nce != NULL)
3514 3515 nce_refrele(ixa->ixa_nce);
3515 3516 ixa->ixa_nce = nce;
3516 3517 }
3517 3518 }
3518 3519
3519 3520 /*
3520 3521 * If the source address is a loopback address, the
3521 3522 * destination had best be local or multicast.
3522 3523 * If we are sending to an IRE_LOCAL using a loopback source then
3523 3524 * it had better be the same zoneid.
3524 3525 */
3525 3526 if (*src_addrp == htonl(INADDR_LOOPBACK)) {
3526 3527 if ((ire->ire_type & IRE_LOCAL) && ire->ire_zoneid != zoneid) {
3527 3528 ire = NULL; /* Stored in ixa_ire */
3528 3529 error = EADDRNOTAVAIL;
3529 3530 goto bad_addr;
3530 3531 }
3531 3532 if (!(ire->ire_type & (IRE_LOOPBACK|IRE_LOCAL|IRE_MULTICAST))) {
3532 3533 ire = NULL; /* Stored in ixa_ire */
3533 3534 error = EADDRNOTAVAIL;
3534 3535 goto bad_addr;
3535 3536 }
3536 3537 }
3537 3538 if (ire->ire_type & IRE_BROADCAST) {
3538 3539 /*
3539 3540 * If the ULP didn't have a specified source, then we
3540 3541 * make sure we reselect the source when sending
3541 3542 * broadcasts out different interfaces.
3542 3543 */
3543 3544 if (flags & IPDF_SELECT_SRC)
3544 3545 ixa->ixa_flags |= IXAF_SET_SOURCE;
3545 3546 else
3546 3547 ixa->ixa_flags &= ~IXAF_SET_SOURCE;
3547 3548 }
3548 3549
3549 3550 /*
3550 3551 * Does the caller want us to pick a source address?
3551 3552 */
3552 3553 if (flags & IPDF_SELECT_SRC) {
3553 3554 ipaddr_t src_addr;
3554 3555
3555 3556 /*
3556 3557 * We use use ire_nexthop_ill to avoid the under ipmp
3557 3558 * interface for source address selection. Note that for ipmp
3558 3559 * probe packets, ixa_ifindex would have been specified, and
3559 3560 * the ip_select_route() invocation would have picked an ire
3560 3561 * will ire_ill pointing at an under interface.
3561 3562 */
3562 3563 ill = ire_nexthop_ill(ire);
3563 3564
3564 3565 /* If unreachable we have no ill but need some source */
3565 3566 if (ill == NULL) {
3566 3567 src_addr = htonl(INADDR_LOOPBACK);
3567 3568 /* Make sure we look for a better source address */
3568 3569 generation = SRC_GENERATION_VERIFY;
3569 3570 } else {
3570 3571 error = ip_select_source_v4(ill, setsrc, dst_addr,
3571 3572 ixa->ixa_multicast_ifaddr, zoneid,
3572 3573 ipst, &src_addr, &generation, NULL);
3573 3574 if (error != 0) {
3574 3575 ire = NULL; /* Stored in ixa_ire */
3575 3576 goto bad_addr;
3576 3577 }
3577 3578 }
3578 3579
3579 3580 /*
3580 3581 * We allow the source address to to down.
3581 3582 * However, we check that we don't use the loopback address
3582 3583 * as a source when sending out on the wire.
3583 3584 */
3584 3585 if ((src_addr == htonl(INADDR_LOOPBACK)) &&
3585 3586 !(ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK|IRE_MULTICAST)) &&
3586 3587 !(ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))) {
3587 3588 ire = NULL; /* Stored in ixa_ire */
3588 3589 error = EADDRNOTAVAIL;
3589 3590 goto bad_addr;
3590 3591 }
3591 3592
3592 3593 *src_addrp = src_addr;
3593 3594 ixa->ixa_src_generation = generation;
3594 3595 }
3595 3596
3596 3597 /*
3597 3598 * Make sure we don't leave an unreachable ixa_nce in place
3598 3599 * since ip_select_route is used when we unplumb i.e., remove
3599 3600 * references on ixa_ire, ixa_nce, and ixa_dce.
3600 3601 */
3601 3602 nce = ixa->ixa_nce;
3602 3603 if (nce != NULL && nce->nce_is_condemned) {
3603 3604 nce_refrele(nce);
3604 3605 ixa->ixa_nce = NULL;
3605 3606 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3606 3607 }
3607 3608
3608 3609 /*
3609 3610 * The caller has set IXAF_PMTU_DISCOVERY if path MTU is desired.
3610 3611 * However, we can't do it for IPv4 multicast or broadcast.
3611 3612 */
3612 3613 if (ire->ire_type & (IRE_BROADCAST|IRE_MULTICAST))
3613 3614 ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
3614 3615
3615 3616 /*
3616 3617 * Set initial value for fragmentation limit. Either conn_ip_output
3617 3618 * or ULP might updates it when there are routing changes.
3618 3619 * Handles a NULL ixa_ire->ire_ill or a NULL ixa_nce for RTF_REJECT.
3619 3620 */
3620 3621 pmtu = ip_get_pmtu(ixa);
3621 3622 ixa->ixa_fragsize = pmtu;
3622 3623 /* Make sure ixa_fragsize and ixa_pmtu remain identical */
3623 3624 if (ixa->ixa_flags & IXAF_VERIFY_PMTU)
3624 3625 ixa->ixa_pmtu = pmtu;
3625 3626
3626 3627 /*
3627 3628 * Extract information useful for some transports.
3628 3629 * First we look for DCE metrics. Then we take what we have in
3629 3630 * the metrics in the route, where the offlink is used if we have
3630 3631 * one.
3631 3632 */
3632 3633 if (uinfo != NULL) {
3633 3634 bzero(uinfo, sizeof (*uinfo));
3634 3635
3635 3636 if (dce->dce_flags & DCEF_UINFO)
3636 3637 *uinfo = dce->dce_uinfo;
3637 3638
3638 3639 rts_merge_metrics(uinfo, &ire->ire_metrics);
3639 3640
3640 3641 /* Allow ire_metrics to decrease the path MTU from above */
3641 3642 if (uinfo->iulp_mtu == 0 || uinfo->iulp_mtu > pmtu)
3642 3643 uinfo->iulp_mtu = pmtu;
3643 3644
3644 3645 uinfo->iulp_localnet = (ire->ire_type & IRE_ONLINK) != 0;
3645 3646 uinfo->iulp_loopback = (ire->ire_type & IRE_LOOPBACK) != 0;
3646 3647 uinfo->iulp_local = (ire->ire_type & IRE_LOCAL) != 0;
3647 3648 }
3648 3649
3649 3650 if (ill != NULL)
3650 3651 ill_refrele(ill);
3651 3652
3652 3653 return (error);
3653 3654
3654 3655 bad_addr:
3655 3656 if (ire != NULL)
3656 3657 ire_refrele(ire);
3657 3658
3658 3659 if (ill != NULL)
3659 3660 ill_refrele(ill);
3660 3661
3661 3662 /*
3662 3663 * Make sure we don't leave an unreachable ixa_nce in place
3663 3664 * since ip_select_route is used when we unplumb i.e., remove
3664 3665 * references on ixa_ire, ixa_nce, and ixa_dce.
3665 3666 */
3666 3667 nce = ixa->ixa_nce;
3667 3668 if (nce != NULL && nce->nce_is_condemned) {
3668 3669 nce_refrele(nce);
3669 3670 ixa->ixa_nce = NULL;
3670 3671 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3671 3672 }
3672 3673
3673 3674 return (error);
3674 3675 }
3675 3676
3676 3677
3677 3678 /*
3678 3679 * Get the base MTU for the case when path MTU discovery is not used.
3679 3680 * Takes the MTU of the IRE into account.
3680 3681 */
3681 3682 uint_t
3682 3683 ip_get_base_mtu(ill_t *ill, ire_t *ire)
3683 3684 {
3684 3685 uint_t mtu;
3685 3686 uint_t iremtu = ire->ire_metrics.iulp_mtu;
3686 3687
3687 3688 if (ire->ire_type & (IRE_MULTICAST|IRE_BROADCAST))
3688 3689 mtu = ill->ill_mc_mtu;
3689 3690 else
3690 3691 mtu = ill->ill_mtu;
3691 3692
3692 3693 if (iremtu != 0 && iremtu < mtu)
3693 3694 mtu = iremtu;
3694 3695
3695 3696 return (mtu);
3696 3697 }
3697 3698
3698 3699 /*
3699 3700 * Get the PMTU for the attributes. Handles both IPv4 and IPv6.
3700 3701 * Assumes that ixa_ire, dce, and nce have already been set up.
3701 3702 *
3702 3703 * The caller has set IXAF_PMTU_DISCOVERY if path MTU discovery is desired.
3703 3704 * We avoid path MTU discovery if it is disabled with ndd.
3704 3705 * Furtermore, if the path MTU is too small, then we don't set DF for IPv4.
3705 3706 *
3706 3707 * NOTE: We also used to turn it off for source routed packets. That
3707 3708 * is no longer required since the dce is per final destination.
3708 3709 */
3709 3710 uint_t
3710 3711 ip_get_pmtu(ip_xmit_attr_t *ixa)
3711 3712 {
3712 3713 ip_stack_t *ipst = ixa->ixa_ipst;
3713 3714 dce_t *dce;
3714 3715 nce_t *nce;
3715 3716 ire_t *ire;
3716 3717 uint_t pmtu;
3717 3718
3718 3719 ire = ixa->ixa_ire;
3719 3720 dce = ixa->ixa_dce;
3720 3721 nce = ixa->ixa_nce;
3721 3722
3722 3723 /*
3723 3724 * If path MTU discovery has been turned off by ndd, then we ignore
3724 3725 * any dce_pmtu and for IPv4 we will not set DF.
3725 3726 */
3726 3727 if (!ipst->ips_ip_path_mtu_discovery)
3727 3728 ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
3728 3729
3729 3730 pmtu = IP_MAXPACKET;
3730 3731 /*
3731 3732 * Decide whether whether IPv4 sets DF
3732 3733 * For IPv6 "no DF" means to use the 1280 mtu
3733 3734 */
3734 3735 if (ixa->ixa_flags & IXAF_PMTU_DISCOVERY) {
3735 3736 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3736 3737 } else {
3737 3738 ixa->ixa_flags &= ~IXAF_PMTU_IPV4_DF;
3738 3739 if (!(ixa->ixa_flags & IXAF_IS_IPV4))
3739 3740 pmtu = IPV6_MIN_MTU;
3740 3741 }
3741 3742
3742 3743 /* Check if the PMTU is to old before we use it */
3743 3744 if ((dce->dce_flags & DCEF_PMTU) &&
3744 3745 TICK_TO_SEC(ddi_get_lbolt64()) - dce->dce_last_change_time >
3745 3746 ipst->ips_ip_pathmtu_interval) {
3746 3747 /*
3747 3748 * Older than 20 minutes. Drop the path MTU information.
3748 3749 */
3749 3750 mutex_enter(&dce->dce_lock);
3750 3751 dce->dce_flags &= ~(DCEF_PMTU|DCEF_TOO_SMALL_PMTU);
3751 3752 dce->dce_last_change_time = TICK_TO_SEC(ddi_get_lbolt64());
3752 3753 mutex_exit(&dce->dce_lock);
3753 3754 dce_increment_generation(dce);
3754 3755 }
3755 3756
3756 3757 /* The metrics on the route can lower the path MTU */
3757 3758 if (ire->ire_metrics.iulp_mtu != 0 &&
3758 3759 ire->ire_metrics.iulp_mtu < pmtu)
3759 3760 pmtu = ire->ire_metrics.iulp_mtu;
3760 3761
3761 3762 /*
3762 3763 * If the path MTU is smaller than some minimum, we still use dce_pmtu
3763 3764 * above (would be 576 for IPv4 and 1280 for IPv6), but we clear
3764 3765 * IXAF_PMTU_IPV4_DF so that we avoid setting DF for IPv4.
3765 3766 */
3766 3767 if (ixa->ixa_flags & IXAF_PMTU_DISCOVERY) {
3767 3768 if (dce->dce_flags & DCEF_PMTU) {
3768 3769 if (dce->dce_pmtu < pmtu)
3769 3770 pmtu = dce->dce_pmtu;
3770 3771
3771 3772 if (dce->dce_flags & DCEF_TOO_SMALL_PMTU) {
3772 3773 ixa->ixa_flags |= IXAF_PMTU_TOO_SMALL;
3773 3774 ixa->ixa_flags &= ~IXAF_PMTU_IPV4_DF;
3774 3775 } else {
3775 3776 ixa->ixa_flags &= ~IXAF_PMTU_TOO_SMALL;
3776 3777 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3777 3778 }
3778 3779 } else {
3779 3780 ixa->ixa_flags &= ~IXAF_PMTU_TOO_SMALL;
3780 3781 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3781 3782 }
3782 3783 }
3783 3784
3784 3785 /*
3785 3786 * If we have an IRE_LOCAL we use the loopback mtu instead of
3786 3787 * the ill for going out the wire i.e., IRE_LOCAL gets the same
3787 3788 * mtu as IRE_LOOPBACK.
3788 3789 */
3789 3790 if (ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK)) {
3790 3791 uint_t loopback_mtu;
3791 3792
3792 3793 loopback_mtu = (ire->ire_ipversion == IPV6_VERSION) ?
3793 3794 ip_loopback_mtu_v6plus : ip_loopback_mtuplus;
3794 3795
3795 3796 if (loopback_mtu < pmtu)
3796 3797 pmtu = loopback_mtu;
3797 3798 } else if (nce != NULL) {
3798 3799 /*
3799 3800 * Make sure we don't exceed the interface MTU.
3800 3801 * In the case of RTF_REJECT or RTF_BLACKHOLE we might not have
3801 3802 * an ill. We'd use the above IP_MAXPACKET in that case just
3802 3803 * to tell the transport something larger than zero.
3803 3804 */
3804 3805 if (ire->ire_type & (IRE_MULTICAST|IRE_BROADCAST)) {
3805 3806 if (nce->nce_common->ncec_ill->ill_mc_mtu < pmtu)
3806 3807 pmtu = nce->nce_common->ncec_ill->ill_mc_mtu;
3807 3808 if (nce->nce_common->ncec_ill != nce->nce_ill &&
3808 3809 nce->nce_ill->ill_mc_mtu < pmtu) {
3809 3810 /*
3810 3811 * for interfaces in an IPMP group, the mtu of
3811 3812 * the nce_ill (under_ill) could be different
3812 3813 * from the mtu of the ncec_ill, so we take the
3813 3814 * min of the two.
3814 3815 */
3815 3816 pmtu = nce->nce_ill->ill_mc_mtu;
3816 3817 }
3817 3818 } else {
3818 3819 if (nce->nce_common->ncec_ill->ill_mtu < pmtu)
3819 3820 pmtu = nce->nce_common->ncec_ill->ill_mtu;
3820 3821 if (nce->nce_common->ncec_ill != nce->nce_ill &&
3821 3822 nce->nce_ill->ill_mtu < pmtu) {
3822 3823 /*
3823 3824 * for interfaces in an IPMP group, the mtu of
3824 3825 * the nce_ill (under_ill) could be different
3825 3826 * from the mtu of the ncec_ill, so we take the
3826 3827 * min of the two.
3827 3828 */
3828 3829 pmtu = nce->nce_ill->ill_mtu;
3829 3830 }
3830 3831 }
3831 3832 }
3832 3833
3833 3834 /*
3834 3835 * Handle the IPV6_USE_MIN_MTU socket option or ancillary data.
3835 3836 * Only applies to IPv6.
3836 3837 */
3837 3838 if (!(ixa->ixa_flags & IXAF_IS_IPV4)) {
3838 3839 if (ixa->ixa_flags & IXAF_USE_MIN_MTU) {
3839 3840 switch (ixa->ixa_use_min_mtu) {
3840 3841 case IPV6_USE_MIN_MTU_MULTICAST:
3841 3842 if (ire->ire_type & IRE_MULTICAST)
3842 3843 pmtu = IPV6_MIN_MTU;
3843 3844 break;
3844 3845 case IPV6_USE_MIN_MTU_ALWAYS:
3845 3846 pmtu = IPV6_MIN_MTU;
3846 3847 break;
3847 3848 case IPV6_USE_MIN_MTU_NEVER:
3848 3849 break;
3849 3850 }
3850 3851 } else {
3851 3852 /* Default is IPV6_USE_MIN_MTU_MULTICAST */
3852 3853 if (ire->ire_type & IRE_MULTICAST)
3853 3854 pmtu = IPV6_MIN_MTU;
3854 3855 }
3855 3856 }
3856 3857
3857 3858 /*
3858 3859 * After receiving an ICMPv6 "packet too big" message with a
3859 3860 * MTU < 1280, and for multirouted IPv6 packets, the IP layer
3860 3861 * will insert a 8-byte fragment header in every packet. We compensate
3861 3862 * for those cases by returning a smaller path MTU to the ULP.
3862 3863 *
3863 3864 * In the case of CGTP then ip_output will add a fragment header.
3864 3865 * Make sure there is room for it by telling a smaller number
3865 3866 * to the transport.
3866 3867 *
3867 3868 * When IXAF_IPV6_ADDR_FRAGHDR we subtract the frag hdr here
3868 3869 * so the ULPs consistently see a iulp_pmtu and ip_get_pmtu()
3869 3870 * which is the size of the packets it can send.
3870 3871 */
3871 3872 if (!(ixa->ixa_flags & IXAF_IS_IPV4)) {
3872 3873 if ((dce->dce_flags & DCEF_TOO_SMALL_PMTU) ||
3873 3874 (ire->ire_flags & RTF_MULTIRT) ||
3874 3875 (ixa->ixa_flags & IXAF_MULTIRT_MULTICAST)) {
3875 3876 pmtu -= sizeof (ip6_frag_t);
3876 3877 ixa->ixa_flags |= IXAF_IPV6_ADD_FRAGHDR;
3877 3878 }
3878 3879 }
3879 3880
3880 3881 return (pmtu);
3881 3882 }
3882 3883
3883 3884 /*
3884 3885 * Carve "len" bytes out of an mblk chain, consuming any we empty, and duping
3885 3886 * the final piece where we don't. Return a pointer to the first mblk in the
3886 3887 * result, and update the pointer to the next mblk to chew on. If anything
3887 3888 * goes wrong (i.e., dupb fails), we waste everything in sight and return a
3888 3889 * NULL pointer.
3889 3890 */
3890 3891 mblk_t *
3891 3892 ip_carve_mp(mblk_t **mpp, ssize_t len)
3892 3893 {
3893 3894 mblk_t *mp0;
3894 3895 mblk_t *mp1;
3895 3896 mblk_t *mp2;
3896 3897
3897 3898 if (!len || !mpp || !(mp0 = *mpp))
3898 3899 return (NULL);
3899 3900 /* If we aren't going to consume the first mblk, we need a dup. */
3900 3901 if (mp0->b_wptr - mp0->b_rptr > len) {
3901 3902 mp1 = dupb(mp0);
3902 3903 if (mp1) {
3903 3904 /* Partition the data between the two mblks. */
3904 3905 mp1->b_wptr = mp1->b_rptr + len;
3905 3906 mp0->b_rptr = mp1->b_wptr;
3906 3907 /*
3907 3908 * after adjustments if mblk not consumed is now
3908 3909 * unaligned, try to align it. If this fails free
3909 3910 * all messages and let upper layer recover.
3910 3911 */
3911 3912 if (!OK_32PTR(mp0->b_rptr)) {
3912 3913 if (!pullupmsg(mp0, -1)) {
3913 3914 freemsg(mp0);
3914 3915 freemsg(mp1);
3915 3916 *mpp = NULL;
3916 3917 return (NULL);
3917 3918 }
3918 3919 }
3919 3920 }
3920 3921 return (mp1);
3921 3922 }
3922 3923 /* Eat through as many mblks as we need to get len bytes. */
3923 3924 len -= mp0->b_wptr - mp0->b_rptr;
3924 3925 for (mp2 = mp1 = mp0; (mp2 = mp2->b_cont) != 0 && len; mp1 = mp2) {
3925 3926 if (mp2->b_wptr - mp2->b_rptr > len) {
3926 3927 /*
3927 3928 * We won't consume the entire last mblk. Like
3928 3929 * above, dup and partition it.
3929 3930 */
3930 3931 mp1->b_cont = dupb(mp2);
3931 3932 mp1 = mp1->b_cont;
3932 3933 if (!mp1) {
3933 3934 /*
3934 3935 * Trouble. Rather than go to a lot of
3935 3936 * trouble to clean up, we free the messages.
3936 3937 * This won't be any worse than losing it on
3937 3938 * the wire.
3938 3939 */
3939 3940 freemsg(mp0);
3940 3941 freemsg(mp2);
3941 3942 *mpp = NULL;
3942 3943 return (NULL);
3943 3944 }
3944 3945 mp1->b_wptr = mp1->b_rptr + len;
3945 3946 mp2->b_rptr = mp1->b_wptr;
3946 3947 /*
3947 3948 * after adjustments if mblk not consumed is now
3948 3949 * unaligned, try to align it. If this fails free
3949 3950 * all messages and let upper layer recover.
3950 3951 */
3951 3952 if (!OK_32PTR(mp2->b_rptr)) {
3952 3953 if (!pullupmsg(mp2, -1)) {
3953 3954 freemsg(mp0);
3954 3955 freemsg(mp2);
3955 3956 *mpp = NULL;
3956 3957 return (NULL);
3957 3958 }
3958 3959 }
3959 3960 *mpp = mp2;
3960 3961 return (mp0);
3961 3962 }
3962 3963 /* Decrement len by the amount we just got. */
3963 3964 len -= mp2->b_wptr - mp2->b_rptr;
3964 3965 }
3965 3966 /*
3966 3967 * len should be reduced to zero now. If not our caller has
3967 3968 * screwed up.
3968 3969 */
3969 3970 if (len) {
3970 3971 /* Shouldn't happen! */
3971 3972 freemsg(mp0);
3972 3973 *mpp = NULL;
3973 3974 return (NULL);
3974 3975 }
3975 3976 /*
3976 3977 * We consumed up to exactly the end of an mblk. Detach the part
3977 3978 * we are returning from the rest of the chain.
3978 3979 */
3979 3980 mp1->b_cont = NULL;
3980 3981 *mpp = mp2;
3981 3982 return (mp0);
3982 3983 }
3983 3984
3984 3985 /* The ill stream is being unplumbed. Called from ip_close */
3985 3986 int
3986 3987 ip_modclose(ill_t *ill)
3987 3988 {
3988 3989 boolean_t success;
3989 3990 ipsq_t *ipsq;
3990 3991 ipif_t *ipif;
3991 3992 queue_t *q = ill->ill_rq;
3992 3993 ip_stack_t *ipst = ill->ill_ipst;
3993 3994 int i;
3994 3995 arl_ill_common_t *ai = ill->ill_common;
3995 3996
3996 3997 /*
3997 3998 * The punlink prior to this may have initiated a capability
3998 3999 * negotiation. But ipsq_enter will block until that finishes or
3999 4000 * times out.
4000 4001 */
4001 4002 success = ipsq_enter(ill, B_FALSE, NEW_OP);
4002 4003
4003 4004 /*
4004 4005 * Open/close/push/pop is guaranteed to be single threaded
4005 4006 * per stream by STREAMS. FS guarantees that all references
4006 4007 * from top are gone before close is called. So there can't
4007 4008 * be another close thread that has set CONDEMNED on this ill.
4008 4009 * and cause ipsq_enter to return failure.
4009 4010 */
4010 4011 ASSERT(success);
4011 4012 ipsq = ill->ill_phyint->phyint_ipsq;
4012 4013
4013 4014 /*
4014 4015 * Mark it condemned. No new reference will be made to this ill.
4015 4016 * Lookup functions will return an error. Threads that try to
4016 4017 * increment the refcnt must check for ILL_CAN_LOOKUP. This ensures
4017 4018 * that the refcnt will drop down to zero.
4018 4019 */
4019 4020 mutex_enter(&ill->ill_lock);
4020 4021 ill->ill_state_flags |= ILL_CONDEMNED;
4021 4022 for (ipif = ill->ill_ipif; ipif != NULL;
4022 4023 ipif = ipif->ipif_next) {
4023 4024 ipif->ipif_state_flags |= IPIF_CONDEMNED;
4024 4025 }
4025 4026 /*
4026 4027 * Wake up anybody waiting to enter the ipsq. ipsq_enter
4027 4028 * returns error if ILL_CONDEMNED is set
4028 4029 */
4029 4030 cv_broadcast(&ill->ill_cv);
4030 4031 mutex_exit(&ill->ill_lock);
4031 4032
4032 4033 /*
4033 4034 * Send all the deferred DLPI messages downstream which came in
4034 4035 * during the small window right before ipsq_enter(). We do this
4035 4036 * without waiting for the ACKs because all the ACKs for M_PROTO
4036 4037 * messages are ignored in ip_rput() when ILL_CONDEMNED is set.
4037 4038 */
4038 4039 ill_dlpi_send_deferred(ill);
4039 4040
4040 4041 /*
4041 4042 * Shut down fragmentation reassembly.
4042 4043 * ill_frag_timer won't start a timer again.
4043 4044 * Now cancel any existing timer
4044 4045 */
4045 4046 (void) untimeout(ill->ill_frag_timer_id);
4046 4047 (void) ill_frag_timeout(ill, 0);
4047 4048
4048 4049 /*
4049 4050 * Call ill_delete to bring down the ipifs, ilms and ill on
4050 4051 * this ill. Then wait for the refcnts to drop to zero.
4051 4052 * ill_is_freeable checks whether the ill is really quiescent.
4052 4053 * Then make sure that threads that are waiting to enter the
4053 4054 * ipsq have seen the error returned by ipsq_enter and have
4054 4055 * gone away. Then we call ill_delete_tail which does the
4055 4056 * DL_UNBIND_REQ with the driver and then qprocsoff.
4056 4057 */
4057 4058 ill_delete(ill);
4058 4059 mutex_enter(&ill->ill_lock);
4059 4060 while (!ill_is_freeable(ill))
4060 4061 cv_wait(&ill->ill_cv, &ill->ill_lock);
4061 4062
4062 4063 while (ill->ill_waiters)
4063 4064 cv_wait(&ill->ill_cv, &ill->ill_lock);
4064 4065
4065 4066 mutex_exit(&ill->ill_lock);
4066 4067
4067 4068 /*
4068 4069 * ill_delete_tail drops reference on ill_ipst, but we need to keep
4069 4070 * it held until the end of the function since the cleanup
4070 4071 * below needs to be able to use the ip_stack_t.
4071 4072 */
4072 4073 netstack_hold(ipst->ips_netstack);
4073 4074
4074 4075 /* qprocsoff is done via ill_delete_tail */
4075 4076 ill_delete_tail(ill);
4076 4077 /*
4077 4078 * synchronously wait for arp stream to unbind. After this, we
4078 4079 * cannot get any data packets up from the driver.
4079 4080 */
4080 4081 arp_unbind_complete(ill);
4081 4082 ASSERT(ill->ill_ipst == NULL);
4082 4083
4083 4084 /*
4084 4085 * Walk through all conns and qenable those that have queued data.
4085 4086 * Close synchronization needs this to
4086 4087 * be done to ensure that all upper layers blocked
4087 4088 * due to flow control to the closing device
4088 4089 * get unblocked.
4089 4090 */
4090 4091 ip1dbg(("ip_wsrv: walking\n"));
4091 4092 for (i = 0; i < TX_FANOUT_SIZE; i++) {
4092 4093 conn_walk_drain(ipst, &ipst->ips_idl_tx_list[i]);
4093 4094 }
4094 4095
4095 4096 /*
4096 4097 * ai can be null if this is an IPv6 ill, or if the IPv4
4097 4098 * stream is being torn down before ARP was plumbed (e.g.,
4098 4099 * /sbin/ifconfig plumbing a stream twice, and encountering
4099 4100 * an error
4100 4101 */
4101 4102 if (ai != NULL) {
4102 4103 ASSERT(!ill->ill_isv6);
4103 4104 mutex_enter(&ai->ai_lock);
4104 4105 ai->ai_ill = NULL;
4105 4106 if (ai->ai_arl == NULL) {
4106 4107 mutex_destroy(&ai->ai_lock);
4107 4108 kmem_free(ai, sizeof (*ai));
4108 4109 } else {
4109 4110 cv_signal(&ai->ai_ill_unplumb_done);
4110 4111 mutex_exit(&ai->ai_lock);
4111 4112 }
4112 4113 }
4113 4114
4114 4115 mutex_enter(&ipst->ips_ip_mi_lock);
4115 4116 mi_close_unlink(&ipst->ips_ip_g_head, (IDP)ill);
4116 4117 mutex_exit(&ipst->ips_ip_mi_lock);
4117 4118
4118 4119 /*
4119 4120 * credp could be null if the open didn't succeed and ip_modopen
4120 4121 * itself calls ip_close.
4121 4122 */
4122 4123 if (ill->ill_credp != NULL)
4123 4124 crfree(ill->ill_credp);
4124 4125
4125 4126 mutex_destroy(&ill->ill_saved_ire_lock);
4126 4127 mutex_destroy(&ill->ill_lock);
4127 4128 rw_destroy(&ill->ill_mcast_lock);
4128 4129 mutex_destroy(&ill->ill_mcast_serializer);
4129 4130 list_destroy(&ill->ill_nce);
4130 4131
4131 4132 /*
4132 4133 * Now we are done with the module close pieces that
4133 4134 * need the netstack_t.
4134 4135 */
4135 4136 netstack_rele(ipst->ips_netstack);
4136 4137
4137 4138 mi_close_free((IDP)ill);
4138 4139 q->q_ptr = WR(q)->q_ptr = NULL;
4139 4140
4140 4141 ipsq_exit(ipsq);
4141 4142
4142 4143 return (0);
4143 4144 }
4144 4145
4145 4146 /*
4146 4147 * This is called as part of close() for IP, UDP, ICMP, and RTS
4147 4148 * in order to quiesce the conn.
4148 4149 */
4149 4150 void
4150 4151 ip_quiesce_conn(conn_t *connp)
4151 4152 {
4152 4153 boolean_t drain_cleanup_reqd = B_FALSE;
4153 4154 boolean_t conn_ioctl_cleanup_reqd = B_FALSE;
4154 4155 boolean_t ilg_cleanup_reqd = B_FALSE;
4155 4156 ip_stack_t *ipst;
4156 4157
4157 4158 ASSERT(!IPCL_IS_TCP(connp));
4158 4159 ipst = connp->conn_netstack->netstack_ip;
4159 4160
4160 4161 /*
4161 4162 * Mark the conn as closing, and this conn must not be
4162 4163 * inserted in future into any list. Eg. conn_drain_insert(),
4163 4164 * won't insert this conn into the conn_drain_list.
4164 4165 *
4165 4166 * conn_idl, and conn_ilg cannot get set henceforth.
4166 4167 */
4167 4168 mutex_enter(&connp->conn_lock);
4168 4169 ASSERT(!(connp->conn_state_flags & CONN_QUIESCED));
4169 4170 connp->conn_state_flags |= CONN_CLOSING;
4170 4171 if (connp->conn_idl != NULL)
4171 4172 drain_cleanup_reqd = B_TRUE;
4172 4173 if (connp->conn_oper_pending_ill != NULL)
4173 4174 conn_ioctl_cleanup_reqd = B_TRUE;
4174 4175 if (connp->conn_dhcpinit_ill != NULL) {
4175 4176 ASSERT(connp->conn_dhcpinit_ill->ill_dhcpinit != 0);
4176 4177 atomic_dec_32(&connp->conn_dhcpinit_ill->ill_dhcpinit);
4177 4178 ill_set_inputfn(connp->conn_dhcpinit_ill);
4178 4179 connp->conn_dhcpinit_ill = NULL;
4179 4180 }
4180 4181 if (connp->conn_ilg != NULL)
4181 4182 ilg_cleanup_reqd = B_TRUE;
4182 4183 mutex_exit(&connp->conn_lock);
4183 4184
4184 4185 if (conn_ioctl_cleanup_reqd)
4185 4186 conn_ioctl_cleanup(connp);
4186 4187
4187 4188 if (is_system_labeled() && connp->conn_anon_port) {
4188 4189 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
4189 4190 connp->conn_mlp_type, connp->conn_proto,
4190 4191 ntohs(connp->conn_lport), B_FALSE);
4191 4192 connp->conn_anon_port = 0;
4192 4193 }
4193 4194 connp->conn_mlp_type = mlptSingle;
4194 4195
4195 4196 /*
4196 4197 * Remove this conn from any fanout list it is on.
4197 4198 * and then wait for any threads currently operating
4198 4199 * on this endpoint to finish
4199 4200 */
4200 4201 ipcl_hash_remove(connp);
4201 4202
4202 4203 /*
4203 4204 * Remove this conn from the drain list, and do any other cleanup that
4204 4205 * may be required. (TCP conns are never flow controlled, and
4205 4206 * conn_idl will be NULL.)
4206 4207 */
4207 4208 if (drain_cleanup_reqd && connp->conn_idl != NULL) {
4208 4209 idl_t *idl = connp->conn_idl;
4209 4210
4210 4211 mutex_enter(&idl->idl_lock);
4211 4212 conn_drain(connp, B_TRUE);
4212 4213 mutex_exit(&idl->idl_lock);
4213 4214 }
4214 4215
4215 4216 if (connp == ipst->ips_ip_g_mrouter)
4216 4217 (void) ip_mrouter_done(ipst);
4217 4218
4218 4219 if (ilg_cleanup_reqd)
4219 4220 ilg_delete_all(connp);
4220 4221
4221 4222 /*
4222 4223 * Now conn refcnt can increase only thru CONN_INC_REF_LOCKED.
4223 4224 * callers from write side can't be there now because close
4224 4225 * is in progress. The only other caller is ipcl_walk
4225 4226 * which checks for the condemned flag.
4226 4227 */
4227 4228 mutex_enter(&connp->conn_lock);
4228 4229 connp->conn_state_flags |= CONN_CONDEMNED;
4229 4230 while (connp->conn_ref != 1)
4230 4231 cv_wait(&connp->conn_cv, &connp->conn_lock);
4231 4232 connp->conn_state_flags |= CONN_QUIESCED;
4232 4233 mutex_exit(&connp->conn_lock);
4233 4234 }
4234 4235
4235 4236 /* ARGSUSED */
4236 4237 int
4237 4238 ip_close(queue_t *q, int flags)
4238 4239 {
4239 4240 conn_t *connp;
4240 4241
4241 4242 /*
4242 4243 * Call the appropriate delete routine depending on whether this is
4243 4244 * a module or device.
4244 4245 */
4245 4246 if (WR(q)->q_next != NULL) {
4246 4247 /* This is a module close */
4247 4248 return (ip_modclose((ill_t *)q->q_ptr));
4248 4249 }
4249 4250
4250 4251 connp = q->q_ptr;
4251 4252 ip_quiesce_conn(connp);
4252 4253
4253 4254 qprocsoff(q);
4254 4255
4255 4256 /*
4256 4257 * Now we are truly single threaded on this stream, and can
4257 4258 * delete the things hanging off the connp, and finally the connp.
4258 4259 * We removed this connp from the fanout list, it cannot be
4259 4260 * accessed thru the fanouts, and we already waited for the
4260 4261 * conn_ref to drop to 0. We are already in close, so
4261 4262 * there cannot be any other thread from the top. qprocsoff
4262 4263 * has completed, and service has completed or won't run in
4263 4264 * future.
4264 4265 */
4265 4266 ASSERT(connp->conn_ref == 1);
4266 4267
4267 4268 inet_minor_free(connp->conn_minor_arena, connp->conn_dev);
4268 4269
4269 4270 connp->conn_ref--;
4270 4271 ipcl_conn_destroy(connp);
4271 4272
4272 4273 q->q_ptr = WR(q)->q_ptr = NULL;
4273 4274 return (0);
4274 4275 }
4275 4276
4276 4277 /*
4277 4278 * Wapper around putnext() so that ip_rts_request can merely use
4278 4279 * conn_recv.
4279 4280 */
4280 4281 /*ARGSUSED2*/
4281 4282 static void
4282 4283 ip_conn_input(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira)
4283 4284 {
4284 4285 conn_t *connp = (conn_t *)arg1;
4285 4286
4286 4287 putnext(connp->conn_rq, mp);
4287 4288 }
4288 4289
4289 4290 /* Dummy in case ICMP error delivery is attempted to a /dev/ip instance */
4290 4291 /* ARGSUSED */
4291 4292 static void
4292 4293 ip_conn_input_icmp(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira)
4293 4294 {
4294 4295 freemsg(mp);
4295 4296 }
4296 4297
4297 4298 /*
4298 4299 * Called when the module is about to be unloaded
4299 4300 */
4300 4301 void
4301 4302 ip_ddi_destroy(void)
4302 4303 {
4303 4304 /* This needs to be called before destroying any transports. */
4304 4305 mutex_enter(&cpu_lock);
4305 4306 unregister_cpu_setup_func(ip_tp_cpu_update, NULL);
4306 4307 mutex_exit(&cpu_lock);
4307 4308
4308 4309 tnet_fini();
4309 4310
4310 4311 icmp_ddi_g_destroy();
4311 4312 rts_ddi_g_destroy();
4312 4313 udp_ddi_g_destroy();
4313 4314 sctp_ddi_g_destroy();
4314 4315 tcp_ddi_g_destroy();
4315 4316 ilb_ddi_g_destroy();
4316 4317 dce_g_destroy();
4317 4318 ipsec_policy_g_destroy();
4318 4319 ipcl_g_destroy();
4319 4320 ip_net_g_destroy();
4320 4321 ip_ire_g_fini();
4321 4322 inet_minor_destroy(ip_minor_arena_sa);
4322 4323 #if defined(_LP64)
4323 4324 inet_minor_destroy(ip_minor_arena_la);
4324 4325 #endif
4325 4326
4326 4327 #ifdef DEBUG
4327 4328 list_destroy(&ip_thread_list);
4328 4329 rw_destroy(&ip_thread_rwlock);
4329 4330 tsd_destroy(&ip_thread_data);
4330 4331 #endif
4331 4332
4332 4333 netstack_unregister(NS_IP);
4333 4334 }
4334 4335
4335 4336 /*
4336 4337 * First step in cleanup.
4337 4338 */
4338 4339 /* ARGSUSED */
4339 4340 static void
4340 4341 ip_stack_shutdown(netstackid_t stackid, void *arg)
4341 4342 {
4342 4343 ip_stack_t *ipst = (ip_stack_t *)arg;
4343 4344
4344 4345 #ifdef NS_DEBUG
4345 4346 printf("ip_stack_shutdown(%p, stack %d)\n", (void *)ipst, stackid);
4346 4347 #endif
4347 4348
4348 4349 /*
4349 4350 * Perform cleanup for special interfaces (loopback and IPMP).
4350 4351 */
4351 4352 ip_interface_cleanup(ipst);
4352 4353
4353 4354 /*
4354 4355 * The *_hook_shutdown()s start the process of notifying any
4355 4356 * consumers that things are going away.... nothing is destroyed.
4356 4357 */
4357 4358 ipv4_hook_shutdown(ipst);
4358 4359 ipv6_hook_shutdown(ipst);
4359 4360 arp_hook_shutdown(ipst);
4360 4361
4361 4362 mutex_enter(&ipst->ips_capab_taskq_lock);
4362 4363 ipst->ips_capab_taskq_quit = B_TRUE;
4363 4364 cv_signal(&ipst->ips_capab_taskq_cv);
4364 4365 mutex_exit(&ipst->ips_capab_taskq_lock);
4365 4366 }
4366 4367
4367 4368 /*
4368 4369 * Free the IP stack instance.
4369 4370 */
4370 4371 static void
4371 4372 ip_stack_fini(netstackid_t stackid, void *arg)
4372 4373 {
4373 4374 ip_stack_t *ipst = (ip_stack_t *)arg;
4374 4375 int ret;
4375 4376
4376 4377 #ifdef NS_DEBUG
4377 4378 printf("ip_stack_fini(%p, stack %d)\n", (void *)ipst, stackid);
4378 4379 #endif
4379 4380 /*
4380 4381 * At this point, all of the notifications that the events and
4381 4382 * protocols are going away have been run, meaning that we can
4382 4383 * now set about starting to clean things up.
4383 4384 */
4384 4385 ipobs_fini(ipst);
4385 4386 ipv4_hook_destroy(ipst);
4386 4387 ipv6_hook_destroy(ipst);
4387 4388 arp_hook_destroy(ipst);
4388 4389 ip_net_destroy(ipst);
4389 4390
4390 4391 ipmp_destroy(ipst);
4391 4392
4392 4393 ip_kstat_fini(stackid, ipst->ips_ip_mibkp);
4393 4394 ipst->ips_ip_mibkp = NULL;
4394 4395 icmp_kstat_fini(stackid, ipst->ips_icmp_mibkp);
4395 4396 ipst->ips_icmp_mibkp = NULL;
4396 4397 ip_kstat2_fini(stackid, ipst->ips_ip_kstat);
4397 4398 ipst->ips_ip_kstat = NULL;
4398 4399 bzero(&ipst->ips_ip_statistics, sizeof (ipst->ips_ip_statistics));
4399 4400 ip6_kstat_fini(stackid, ipst->ips_ip6_kstat);
4400 4401 ipst->ips_ip6_kstat = NULL;
4401 4402 bzero(&ipst->ips_ip6_statistics, sizeof (ipst->ips_ip6_statistics));
4402 4403
4403 4404 kmem_free(ipst->ips_propinfo_tbl,
4404 4405 ip_propinfo_count * sizeof (mod_prop_info_t));
4405 4406 ipst->ips_propinfo_tbl = NULL;
4406 4407
4407 4408 dce_stack_destroy(ipst);
4408 4409 ip_mrouter_stack_destroy(ipst);
4409 4410
4410 4411 ret = untimeout(ipst->ips_igmp_timeout_id);
4411 4412 if (ret == -1) {
4412 4413 ASSERT(ipst->ips_igmp_timeout_id == 0);
4413 4414 } else {
4414 4415 ASSERT(ipst->ips_igmp_timeout_id != 0);
4415 4416 ipst->ips_igmp_timeout_id = 0;
4416 4417 }
4417 4418 ret = untimeout(ipst->ips_igmp_slowtimeout_id);
4418 4419 if (ret == -1) {
4419 4420 ASSERT(ipst->ips_igmp_slowtimeout_id == 0);
4420 4421 } else {
4421 4422 ASSERT(ipst->ips_igmp_slowtimeout_id != 0);
4422 4423 ipst->ips_igmp_slowtimeout_id = 0;
4423 4424 }
4424 4425 ret = untimeout(ipst->ips_mld_timeout_id);
4425 4426 if (ret == -1) {
4426 4427 ASSERT(ipst->ips_mld_timeout_id == 0);
4427 4428 } else {
4428 4429 ASSERT(ipst->ips_mld_timeout_id != 0);
4429 4430 ipst->ips_mld_timeout_id = 0;
4430 4431 }
4431 4432 ret = untimeout(ipst->ips_mld_slowtimeout_id);
4432 4433 if (ret == -1) {
4433 4434 ASSERT(ipst->ips_mld_slowtimeout_id == 0);
4434 4435 } else {
4435 4436 ASSERT(ipst->ips_mld_slowtimeout_id != 0);
4436 4437 ipst->ips_mld_slowtimeout_id = 0;
4437 4438 }
4438 4439
4439 4440 ip_ire_fini(ipst);
4440 4441 ip6_asp_free(ipst);
4441 4442 conn_drain_fini(ipst);
4442 4443 ipcl_destroy(ipst);
4443 4444
4444 4445 mutex_destroy(&ipst->ips_ndp4->ndp_g_lock);
4445 4446 mutex_destroy(&ipst->ips_ndp6->ndp_g_lock);
4446 4447 kmem_free(ipst->ips_ndp4, sizeof (ndp_g_t));
4447 4448 ipst->ips_ndp4 = NULL;
4448 4449 kmem_free(ipst->ips_ndp6, sizeof (ndp_g_t));
4449 4450 ipst->ips_ndp6 = NULL;
4450 4451
4451 4452 if (ipst->ips_loopback_ksp != NULL) {
4452 4453 kstat_delete_netstack(ipst->ips_loopback_ksp, stackid);
4453 4454 ipst->ips_loopback_ksp = NULL;
4454 4455 }
4455 4456
4456 4457 mutex_destroy(&ipst->ips_capab_taskq_lock);
4457 4458 cv_destroy(&ipst->ips_capab_taskq_cv);
4458 4459
4459 4460 rw_destroy(&ipst->ips_srcid_lock);
4460 4461
4461 4462 mutex_destroy(&ipst->ips_ip_mi_lock);
4462 4463 rw_destroy(&ipst->ips_ill_g_usesrc_lock);
4463 4464
4464 4465 mutex_destroy(&ipst->ips_igmp_timer_lock);
4465 4466 mutex_destroy(&ipst->ips_mld_timer_lock);
4466 4467 mutex_destroy(&ipst->ips_igmp_slowtimeout_lock);
4467 4468 mutex_destroy(&ipst->ips_mld_slowtimeout_lock);
4468 4469 mutex_destroy(&ipst->ips_ip_addr_avail_lock);
4469 4470 rw_destroy(&ipst->ips_ill_g_lock);
4470 4471
4471 4472 kmem_free(ipst->ips_phyint_g_list, sizeof (phyint_list_t));
4472 4473 ipst->ips_phyint_g_list = NULL;
4473 4474 kmem_free(ipst->ips_ill_g_heads, sizeof (ill_g_head_t) * MAX_G_HEADS);
4474 4475 ipst->ips_ill_g_heads = NULL;
4475 4476
4476 4477 ldi_ident_release(ipst->ips_ldi_ident);
4477 4478 kmem_free(ipst, sizeof (*ipst));
4478 4479 }
4479 4480
4480 4481 /*
4481 4482 * This function is called from the TSD destructor, and is used to debug
4482 4483 * reference count issues in IP. See block comment in <inet/ip_if.h> for
4483 4484 * details.
4484 4485 */
4485 4486 static void
4486 4487 ip_thread_exit(void *phash)
4487 4488 {
4488 4489 th_hash_t *thh = phash;
4489 4490
4490 4491 rw_enter(&ip_thread_rwlock, RW_WRITER);
4491 4492 list_remove(&ip_thread_list, thh);
4492 4493 rw_exit(&ip_thread_rwlock);
4493 4494 mod_hash_destroy_hash(thh->thh_hash);
4494 4495 kmem_free(thh, sizeof (*thh));
4495 4496 }
4496 4497
4497 4498 /*
4498 4499 * Called when the IP kernel module is loaded into the kernel
4499 4500 */
4500 4501 void
4501 4502 ip_ddi_init(void)
4502 4503 {
4503 4504 ip_squeue_flag = ip_squeue_switch(ip_squeue_enter);
4504 4505
4505 4506 /*
4506 4507 * For IP and TCP the minor numbers should start from 2 since we have 4
4507 4508 * initial devices: ip, ip6, tcp, tcp6.
4508 4509 */
4509 4510 /*
4510 4511 * If this is a 64-bit kernel, then create two separate arenas -
4511 4512 * one for TLIs in the range of INET_MIN_DEV+2 through 2^^18-1, and the
4512 4513 * other for socket apps in the range 2^^18 through 2^^32-1.
4513 4514 */
4514 4515 ip_minor_arena_la = NULL;
4515 4516 ip_minor_arena_sa = NULL;
4516 4517 #if defined(_LP64)
4517 4518 if ((ip_minor_arena_sa = inet_minor_create("ip_minor_arena_sa",
4518 4519 INET_MIN_DEV + 2, MAXMIN32, KM_SLEEP)) == NULL) {
4519 4520 cmn_err(CE_PANIC,
4520 4521 "ip_ddi_init: ip_minor_arena_sa creation failed\n");
4521 4522 }
4522 4523 if ((ip_minor_arena_la = inet_minor_create("ip_minor_arena_la",
4523 4524 MAXMIN32 + 1, MAXMIN64, KM_SLEEP)) == NULL) {
4524 4525 cmn_err(CE_PANIC,
4525 4526 "ip_ddi_init: ip_minor_arena_la creation failed\n");
4526 4527 }
4527 4528 #else
4528 4529 if ((ip_minor_arena_sa = inet_minor_create("ip_minor_arena_sa",
4529 4530 INET_MIN_DEV + 2, MAXMIN, KM_SLEEP)) == NULL) {
4530 4531 cmn_err(CE_PANIC,
4531 4532 "ip_ddi_init: ip_minor_arena_sa creation failed\n");
4532 4533 }
4533 4534 #endif
4534 4535 ip_poll_normal_ticks = MSEC_TO_TICK_ROUNDUP(ip_poll_normal_ms);
4535 4536
4536 4537 ipcl_g_init();
4537 4538 ip_ire_g_init();
4538 4539 ip_net_g_init();
4539 4540
4540 4541 #ifdef DEBUG
4541 4542 tsd_create(&ip_thread_data, ip_thread_exit);
4542 4543 rw_init(&ip_thread_rwlock, NULL, RW_DEFAULT, NULL);
4543 4544 list_create(&ip_thread_list, sizeof (th_hash_t),
4544 4545 offsetof(th_hash_t, thh_link));
4545 4546 #endif
4546 4547 ipsec_policy_g_init();
4547 4548 tcp_ddi_g_init();
4548 4549 sctp_ddi_g_init();
4549 4550 dce_g_init();
4550 4551
4551 4552 /*
4552 4553 * We want to be informed each time a stack is created or
4553 4554 * destroyed in the kernel, so we can maintain the
4554 4555 * set of udp_stack_t's.
4555 4556 */
4556 4557 netstack_register(NS_IP, ip_stack_init, ip_stack_shutdown,
4557 4558 ip_stack_fini);
4558 4559
4559 4560 tnet_init();
4560 4561
4561 4562 udp_ddi_g_init();
4562 4563 rts_ddi_g_init();
4563 4564 icmp_ddi_g_init();
4564 4565 ilb_ddi_g_init();
4565 4566
4566 4567 /* This needs to be called after all transports are initialized. */
4567 4568 mutex_enter(&cpu_lock);
4568 4569 register_cpu_setup_func(ip_tp_cpu_update, NULL);
4569 4570 mutex_exit(&cpu_lock);
4570 4571 }
4571 4572
4572 4573 /*
4573 4574 * Initialize the IP stack instance.
4574 4575 */
4575 4576 static void *
4576 4577 ip_stack_init(netstackid_t stackid, netstack_t *ns)
4577 4578 {
4578 4579 ip_stack_t *ipst;
4579 4580 size_t arrsz;
4580 4581 major_t major;
4581 4582
4582 4583 #ifdef NS_DEBUG
4583 4584 printf("ip_stack_init(stack %d)\n", stackid);
4584 4585 #endif
4585 4586
4586 4587 ipst = (ip_stack_t *)kmem_zalloc(sizeof (*ipst), KM_SLEEP);
4587 4588 ipst->ips_netstack = ns;
4588 4589
4589 4590 ipst->ips_ill_g_heads = kmem_zalloc(sizeof (ill_g_head_t) * MAX_G_HEADS,
4590 4591 KM_SLEEP);
4591 4592 ipst->ips_phyint_g_list = kmem_zalloc(sizeof (phyint_list_t),
4592 4593 KM_SLEEP);
4593 4594 ipst->ips_ndp4 = kmem_zalloc(sizeof (ndp_g_t), KM_SLEEP);
4594 4595 ipst->ips_ndp6 = kmem_zalloc(sizeof (ndp_g_t), KM_SLEEP);
4595 4596 mutex_init(&ipst->ips_ndp4->ndp_g_lock, NULL, MUTEX_DEFAULT, NULL);
4596 4597 mutex_init(&ipst->ips_ndp6->ndp_g_lock, NULL, MUTEX_DEFAULT, NULL);
4597 4598
4598 4599 mutex_init(&ipst->ips_igmp_timer_lock, NULL, MUTEX_DEFAULT, NULL);
4599 4600 ipst->ips_igmp_deferred_next = INFINITY;
4600 4601 mutex_init(&ipst->ips_mld_timer_lock, NULL, MUTEX_DEFAULT, NULL);
4601 4602 ipst->ips_mld_deferred_next = INFINITY;
4602 4603 mutex_init(&ipst->ips_igmp_slowtimeout_lock, NULL, MUTEX_DEFAULT, NULL);
4603 4604 mutex_init(&ipst->ips_mld_slowtimeout_lock, NULL, MUTEX_DEFAULT, NULL);
4604 4605 mutex_init(&ipst->ips_ip_mi_lock, NULL, MUTEX_DEFAULT, NULL);
4605 4606 mutex_init(&ipst->ips_ip_addr_avail_lock, NULL, MUTEX_DEFAULT, NULL);
4606 4607 rw_init(&ipst->ips_ill_g_lock, NULL, RW_DEFAULT, NULL);
4607 4608 rw_init(&ipst->ips_ill_g_usesrc_lock, NULL, RW_DEFAULT, NULL);
4608 4609
4609 4610 ipcl_init(ipst);
4610 4611 ip_ire_init(ipst);
4611 4612 ip6_asp_init(ipst);
4612 4613 ipif_init(ipst);
4613 4614 conn_drain_init(ipst);
4614 4615 ip_mrouter_stack_init(ipst);
4615 4616 dce_stack_init(ipst);
4616 4617
4617 4618 ipst->ips_ip_multirt_log_interval = 1000;
4618 4619
4619 4620 ipst->ips_ill_index = 1;
4620 4621
4621 4622 ipst->ips_saved_ip_forwarding = -1;
4622 4623 ipst->ips_reg_vif_num = ALL_VIFS; /* Index to Register vif */
4623 4624
4624 4625 arrsz = ip_propinfo_count * sizeof (mod_prop_info_t);
4625 4626 ipst->ips_propinfo_tbl = (mod_prop_info_t *)kmem_alloc(arrsz, KM_SLEEP);
4626 4627 bcopy(ip_propinfo_tbl, ipst->ips_propinfo_tbl, arrsz);
4627 4628
4628 4629 ipst->ips_ip_mibkp = ip_kstat_init(stackid, ipst);
4629 4630 ipst->ips_icmp_mibkp = icmp_kstat_init(stackid);
4630 4631 ipst->ips_ip_kstat = ip_kstat2_init(stackid, &ipst->ips_ip_statistics);
4631 4632 ipst->ips_ip6_kstat =
4632 4633 ip6_kstat_init(stackid, &ipst->ips_ip6_statistics);
4633 4634
4634 4635 ipst->ips_ip_src_id = 1;
4635 4636 rw_init(&ipst->ips_srcid_lock, NULL, RW_DEFAULT, NULL);
4636 4637
4637 4638 ipst->ips_src_generation = SRC_GENERATION_INITIAL;
4638 4639
|
↓ open down ↓ |
4604 lines elided |
↑ open up ↑ |
4639 4640 ip_net_init(ipst, ns);
4640 4641 ipv4_hook_init(ipst);
4641 4642 ipv6_hook_init(ipst);
4642 4643 arp_hook_init(ipst);
4643 4644 ipmp_init(ipst);
4644 4645 ipobs_init(ipst);
4645 4646
4646 4647 /*
4647 4648 * Create the taskq dispatcher thread and initialize related stuff.
4648 4649 */
4649 - ipst->ips_capab_taskq_thread = thread_create(NULL, 0,
4650 - ill_taskq_dispatch, ipst, 0, &p0, TS_RUN, minclsyspri);
4651 4650 mutex_init(&ipst->ips_capab_taskq_lock, NULL, MUTEX_DEFAULT, NULL);
4652 4651 cv_init(&ipst->ips_capab_taskq_cv, NULL, CV_DEFAULT, NULL);
4652 + ipst->ips_capab_taskq_thread = thread_create(NULL, 0,
4653 + ill_taskq_dispatch, ipst, 0, &p0, TS_RUN, minclsyspri);
4653 4654
4654 4655 major = mod_name_to_major(INET_NAME);
4655 4656 (void) ldi_ident_from_major(major, &ipst->ips_ldi_ident);
4656 4657 return (ipst);
4657 4658 }
4658 4659
4659 4660 /*
4660 4661 * Allocate and initialize a DLPI template of the specified length. (May be
4661 4662 * called as writer.)
4662 4663 */
4663 4664 mblk_t *
4664 4665 ip_dlpi_alloc(size_t len, t_uscalar_t prim)
4665 4666 {
4666 4667 mblk_t *mp;
4667 4668
4668 4669 mp = allocb(len, BPRI_MED);
4669 4670 if (!mp)
4670 4671 return (NULL);
4671 4672
4672 4673 /*
4673 4674 * DLPIv2 says that DL_INFO_REQ and DL_TOKEN_REQ (the latter
4674 4675 * of which we don't seem to use) are sent with M_PCPROTO, and
4675 4676 * that other DLPI are M_PROTO.
4676 4677 */
4677 4678 if (prim == DL_INFO_REQ) {
4678 4679 mp->b_datap->db_type = M_PCPROTO;
4679 4680 } else {
4680 4681 mp->b_datap->db_type = M_PROTO;
4681 4682 }
4682 4683
4683 4684 mp->b_wptr = mp->b_rptr + len;
4684 4685 bzero(mp->b_rptr, len);
4685 4686 ((dl_unitdata_req_t *)mp->b_rptr)->dl_primitive = prim;
4686 4687 return (mp);
4687 4688 }
4688 4689
4689 4690 /*
4690 4691 * Allocate and initialize a DLPI notification. (May be called as writer.)
4691 4692 */
4692 4693 mblk_t *
4693 4694 ip_dlnotify_alloc(uint_t notification, uint_t data)
4694 4695 {
4695 4696 dl_notify_ind_t *notifyp;
4696 4697 mblk_t *mp;
4697 4698
4698 4699 if ((mp = ip_dlpi_alloc(DL_NOTIFY_IND_SIZE, DL_NOTIFY_IND)) == NULL)
4699 4700 return (NULL);
4700 4701
4701 4702 notifyp = (dl_notify_ind_t *)mp->b_rptr;
4702 4703 notifyp->dl_notification = notification;
4703 4704 notifyp->dl_data = data;
4704 4705 return (mp);
4705 4706 }
4706 4707
4707 4708 mblk_t *
4708 4709 ip_dlnotify_alloc2(uint_t notification, uint_t data1, uint_t data2)
4709 4710 {
4710 4711 dl_notify_ind_t *notifyp;
4711 4712 mblk_t *mp;
4712 4713
4713 4714 if ((mp = ip_dlpi_alloc(DL_NOTIFY_IND_SIZE, DL_NOTIFY_IND)) == NULL)
4714 4715 return (NULL);
4715 4716
4716 4717 notifyp = (dl_notify_ind_t *)mp->b_rptr;
4717 4718 notifyp->dl_notification = notification;
4718 4719 notifyp->dl_data1 = data1;
4719 4720 notifyp->dl_data2 = data2;
4720 4721 return (mp);
4721 4722 }
4722 4723
4723 4724 /*
4724 4725 * Debug formatting routine. Returns a character string representation of the
4725 4726 * addr in buf, of the form xxx.xxx.xxx.xxx. This routine takes the address
4726 4727 * in the form of a ipaddr_t and calls ip_dot_saddr with a pointer.
4727 4728 *
4728 4729 * Once the ndd table-printing interfaces are removed, this can be changed to
4729 4730 * standard dotted-decimal form.
4730 4731 */
4731 4732 char *
4732 4733 ip_dot_addr(ipaddr_t addr, char *buf)
4733 4734 {
4734 4735 uint8_t *ap = (uint8_t *)&addr;
4735 4736
4736 4737 (void) mi_sprintf(buf, "%03d.%03d.%03d.%03d",
4737 4738 ap[0] & 0xFF, ap[1] & 0xFF, ap[2] & 0xFF, ap[3] & 0xFF);
4738 4739 return (buf);
4739 4740 }
4740 4741
4741 4742 /*
4742 4743 * Write the given MAC address as a printable string in the usual colon-
4743 4744 * separated format.
4744 4745 */
4745 4746 const char *
4746 4747 mac_colon_addr(const uint8_t *addr, size_t alen, char *buf, size_t buflen)
4747 4748 {
4748 4749 char *bp;
4749 4750
4750 4751 if (alen == 0 || buflen < 4)
4751 4752 return ("?");
4752 4753 bp = buf;
4753 4754 for (;;) {
4754 4755 /*
4755 4756 * If there are more MAC address bytes available, but we won't
4756 4757 * have any room to print them, then add "..." to the string
4757 4758 * instead. See below for the 'magic number' explanation.
4758 4759 */
4759 4760 if ((alen == 2 && buflen < 6) || (alen > 2 && buflen < 7)) {
4760 4761 (void) strcpy(bp, "...");
4761 4762 break;
4762 4763 }
4763 4764 (void) sprintf(bp, "%02x", *addr++);
4764 4765 bp += 2;
4765 4766 if (--alen == 0)
4766 4767 break;
4767 4768 *bp++ = ':';
4768 4769 buflen -= 3;
4769 4770 /*
4770 4771 * At this point, based on the first 'if' statement above,
4771 4772 * either alen == 1 and buflen >= 3, or alen > 1 and
4772 4773 * buflen >= 4. The first case leaves room for the final "xx"
4773 4774 * number and trailing NUL byte. The second leaves room for at
4774 4775 * least "...". Thus the apparently 'magic' numbers chosen for
4775 4776 * that statement.
4776 4777 */
4777 4778 }
4778 4779 return (buf);
4779 4780 }
4780 4781
4781 4782 /*
4782 4783 * Called when it is conceptually a ULP that would sent the packet
4783 4784 * e.g., port unreachable and protocol unreachable. Check that the packet
4784 4785 * would have passed the IPsec global policy before sending the error.
4785 4786 *
4786 4787 * Send an ICMP error after patching up the packet appropriately.
4787 4788 * Uses ip_drop_input and bumps the appropriate MIB.
4788 4789 */
4789 4790 void
4790 4791 ip_fanout_send_icmp_v4(mblk_t *mp, uint_t icmp_type, uint_t icmp_code,
4791 4792 ip_recv_attr_t *ira)
4792 4793 {
4793 4794 ipha_t *ipha;
4794 4795 boolean_t secure;
4795 4796 ill_t *ill = ira->ira_ill;
4796 4797 ip_stack_t *ipst = ill->ill_ipst;
4797 4798 netstack_t *ns = ipst->ips_netstack;
4798 4799 ipsec_stack_t *ipss = ns->netstack_ipsec;
4799 4800
4800 4801 secure = ira->ira_flags & IRAF_IPSEC_SECURE;
4801 4802
4802 4803 /*
4803 4804 * We are generating an icmp error for some inbound packet.
4804 4805 * Called from all ip_fanout_(udp, tcp, proto) functions.
4805 4806 * Before we generate an error, check with global policy
4806 4807 * to see whether this is allowed to enter the system. As
4807 4808 * there is no "conn", we are checking with global policy.
4808 4809 */
4809 4810 ipha = (ipha_t *)mp->b_rptr;
4810 4811 if (secure || ipss->ipsec_inbound_v4_policy_present) {
4811 4812 mp = ipsec_check_global_policy(mp, NULL, ipha, NULL, ira, ns);
4812 4813 if (mp == NULL)
4813 4814 return;
4814 4815 }
4815 4816
4816 4817 /* We never send errors for protocols that we do implement */
4817 4818 if (ira->ira_protocol == IPPROTO_ICMP ||
4818 4819 ira->ira_protocol == IPPROTO_IGMP) {
4819 4820 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
4820 4821 ip_drop_input("ip_fanout_send_icmp_v4", mp, ill);
4821 4822 freemsg(mp);
4822 4823 return;
4823 4824 }
4824 4825 /*
4825 4826 * Have to correct checksum since
4826 4827 * the packet might have been
4827 4828 * fragmented and the reassembly code in ip_rput
4828 4829 * does not restore the IP checksum.
4829 4830 */
4830 4831 ipha->ipha_hdr_checksum = 0;
4831 4832 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
4832 4833
4833 4834 switch (icmp_type) {
4834 4835 case ICMP_DEST_UNREACHABLE:
4835 4836 switch (icmp_code) {
4836 4837 case ICMP_PROTOCOL_UNREACHABLE:
4837 4838 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInUnknownProtos);
4838 4839 ip_drop_input("ipIfStatsInUnknownProtos", mp, ill);
4839 4840 break;
4840 4841 case ICMP_PORT_UNREACHABLE:
4841 4842 BUMP_MIB(ill->ill_ip_mib, udpIfStatsNoPorts);
4842 4843 ip_drop_input("ipIfStatsNoPorts", mp, ill);
4843 4844 break;
4844 4845 }
4845 4846
4846 4847 icmp_unreachable(mp, icmp_code, ira);
4847 4848 break;
4848 4849 default:
4849 4850 #ifdef DEBUG
4850 4851 panic("ip_fanout_send_icmp_v4: wrong type");
4851 4852 /*NOTREACHED*/
4852 4853 #else
4853 4854 freemsg(mp);
4854 4855 break;
4855 4856 #endif
4856 4857 }
4857 4858 }
4858 4859
4859 4860 /*
4860 4861 * Used to send an ICMP error message when a packet is received for
4861 4862 * a protocol that is not supported. The mblk passed as argument
4862 4863 * is consumed by this function.
4863 4864 */
4864 4865 void
4865 4866 ip_proto_not_sup(mblk_t *mp, ip_recv_attr_t *ira)
4866 4867 {
4867 4868 ipha_t *ipha;
4868 4869
4869 4870 ipha = (ipha_t *)mp->b_rptr;
4870 4871 if (ira->ira_flags & IRAF_IS_IPV4) {
4871 4872 ASSERT(IPH_HDR_VERSION(ipha) == IP_VERSION);
4872 4873 ip_fanout_send_icmp_v4(mp, ICMP_DEST_UNREACHABLE,
4873 4874 ICMP_PROTOCOL_UNREACHABLE, ira);
4874 4875 } else {
4875 4876 ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
4876 4877 ip_fanout_send_icmp_v6(mp, ICMP6_PARAM_PROB,
4877 4878 ICMP6_PARAMPROB_NEXTHEADER, ira);
4878 4879 }
4879 4880 }
4880 4881
4881 4882 /*
4882 4883 * Deliver a rawip packet to the given conn, possibly applying ipsec policy.
4883 4884 * Handles IPv4 and IPv6.
4884 4885 * We are responsible for disposing of mp, such as by freemsg() or putnext()
4885 4886 * Caller is responsible for dropping references to the conn.
4886 4887 */
4887 4888 void
4888 4889 ip_fanout_proto_conn(conn_t *connp, mblk_t *mp, ipha_t *ipha, ip6_t *ip6h,
4889 4890 ip_recv_attr_t *ira)
4890 4891 {
4891 4892 ill_t *ill = ira->ira_ill;
4892 4893 ip_stack_t *ipst = ill->ill_ipst;
4893 4894 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
4894 4895 boolean_t secure;
4895 4896 uint_t protocol = ira->ira_protocol;
4896 4897 iaflags_t iraflags = ira->ira_flags;
4897 4898 queue_t *rq;
4898 4899
4899 4900 secure = iraflags & IRAF_IPSEC_SECURE;
4900 4901
4901 4902 rq = connp->conn_rq;
4902 4903 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld : !canputnext(rq)) {
4903 4904 switch (protocol) {
4904 4905 case IPPROTO_ICMPV6:
4905 4906 BUMP_MIB(ill->ill_icmp6_mib, ipv6IfIcmpInOverflows);
4906 4907 break;
4907 4908 case IPPROTO_ICMP:
4908 4909 BUMP_MIB(&ipst->ips_icmp_mib, icmpInOverflows);
4909 4910 break;
4910 4911 default:
4911 4912 BUMP_MIB(ill->ill_ip_mib, rawipIfStatsInOverflows);
4912 4913 break;
4913 4914 }
4914 4915 freemsg(mp);
4915 4916 return;
4916 4917 }
4917 4918
4918 4919 ASSERT(!(IPCL_IS_IPTUN(connp)));
4919 4920
4920 4921 if (((iraflags & IRAF_IS_IPV4) ?
4921 4922 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
4922 4923 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
4923 4924 secure) {
4924 4925 mp = ipsec_check_inbound_policy(mp, connp, ipha,
4925 4926 ip6h, ira);
4926 4927 if (mp == NULL) {
4927 4928 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
4928 4929 /* Note that mp is NULL */
4929 4930 ip_drop_input("ipIfStatsInDiscards", mp, ill);
4930 4931 return;
4931 4932 }
4932 4933 }
4933 4934
4934 4935 if (iraflags & IRAF_ICMP_ERROR) {
4935 4936 (connp->conn_recvicmp)(connp, mp, NULL, ira);
4936 4937 } else {
4937 4938 ill_t *rill = ira->ira_rill;
4938 4939
4939 4940 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
4940 4941 ira->ira_ill = ira->ira_rill = NULL;
4941 4942 /* Send it upstream */
4942 4943 (connp->conn_recv)(connp, mp, NULL, ira);
4943 4944 ira->ira_ill = ill;
4944 4945 ira->ira_rill = rill;
4945 4946 }
4946 4947 }
4947 4948
4948 4949 /*
4949 4950 * Handle protocols with which IP is less intimate. There
4950 4951 * can be more than one stream bound to a particular
4951 4952 * protocol. When this is the case, normally each one gets a copy
4952 4953 * of any incoming packets.
4953 4954 *
4954 4955 * IPsec NOTE :
4955 4956 *
4956 4957 * Don't allow a secure packet going up a non-secure connection.
4957 4958 * We don't allow this because
4958 4959 *
4959 4960 * 1) Reply might go out in clear which will be dropped at
4960 4961 * the sending side.
4961 4962 * 2) If the reply goes out in clear it will give the
4962 4963 * adversary enough information for getting the key in
4963 4964 * most of the cases.
4964 4965 *
4965 4966 * Moreover getting a secure packet when we expect clear
4966 4967 * implies that SA's were added without checking for
4967 4968 * policy on both ends. This should not happen once ISAKMP
4968 4969 * is used to negotiate SAs as SAs will be added only after
4969 4970 * verifying the policy.
4970 4971 *
4971 4972 * Zones notes:
4972 4973 * Earlier in ip_input on a system with multiple shared-IP zones we
4973 4974 * duplicate the multicast and broadcast packets and send them up
4974 4975 * with each explicit zoneid that exists on that ill.
4975 4976 * This means that here we can match the zoneid with SO_ALLZONES being special.
4976 4977 */
4977 4978 void
4978 4979 ip_fanout_proto_v4(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
4979 4980 {
4980 4981 mblk_t *mp1;
4981 4982 ipaddr_t laddr;
4982 4983 conn_t *connp, *first_connp, *next_connp;
4983 4984 connf_t *connfp;
4984 4985 ill_t *ill = ira->ira_ill;
4985 4986 ip_stack_t *ipst = ill->ill_ipst;
4986 4987
4987 4988 laddr = ipha->ipha_dst;
4988 4989
4989 4990 connfp = &ipst->ips_ipcl_proto_fanout_v4[ira->ira_protocol];
4990 4991 mutex_enter(&connfp->connf_lock);
4991 4992 connp = connfp->connf_head;
4992 4993 for (connp = connfp->connf_head; connp != NULL;
4993 4994 connp = connp->conn_next) {
4994 4995 /* Note: IPCL_PROTO_MATCH includes conn_wantpacket */
4995 4996 if (IPCL_PROTO_MATCH(connp, ira, ipha) &&
4996 4997 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
4997 4998 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp))) {
4998 4999 break;
4999 5000 }
5000 5001 }
5001 5002
5002 5003 if (connp == NULL) {
5003 5004 /*
5004 5005 * No one bound to these addresses. Is
5005 5006 * there a client that wants all
5006 5007 * unclaimed datagrams?
5007 5008 */
5008 5009 mutex_exit(&connfp->connf_lock);
5009 5010 ip_fanout_send_icmp_v4(mp, ICMP_DEST_UNREACHABLE,
5010 5011 ICMP_PROTOCOL_UNREACHABLE, ira);
5011 5012 return;
5012 5013 }
5013 5014
5014 5015 ASSERT(IPCL_IS_NONSTR(connp) || connp->conn_rq != NULL);
5015 5016
5016 5017 CONN_INC_REF(connp);
5017 5018 first_connp = connp;
5018 5019 connp = connp->conn_next;
5019 5020
5020 5021 for (;;) {
5021 5022 while (connp != NULL) {
5022 5023 /* Note: IPCL_PROTO_MATCH includes conn_wantpacket */
5023 5024 if (IPCL_PROTO_MATCH(connp, ira, ipha) &&
5024 5025 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5025 5026 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5026 5027 ira, connp)))
5027 5028 break;
5028 5029 connp = connp->conn_next;
5029 5030 }
5030 5031
5031 5032 if (connp == NULL) {
5032 5033 /* No more interested clients */
5033 5034 connp = first_connp;
5034 5035 break;
5035 5036 }
5036 5037 if (((mp1 = dupmsg(mp)) == NULL) &&
5037 5038 ((mp1 = copymsg(mp)) == NULL)) {
5038 5039 /* Memory allocation failed */
5039 5040 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5040 5041 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5041 5042 connp = first_connp;
5042 5043 break;
5043 5044 }
5044 5045
5045 5046 CONN_INC_REF(connp);
5046 5047 mutex_exit(&connfp->connf_lock);
5047 5048
5048 5049 ip_fanout_proto_conn(connp, mp1, (ipha_t *)mp1->b_rptr, NULL,
5049 5050 ira);
5050 5051
5051 5052 mutex_enter(&connfp->connf_lock);
5052 5053 /* Follow the next pointer before releasing the conn. */
5053 5054 next_connp = connp->conn_next;
5054 5055 CONN_DEC_REF(connp);
5055 5056 connp = next_connp;
5056 5057 }
5057 5058
5058 5059 /* Last one. Send it upstream. */
5059 5060 mutex_exit(&connfp->connf_lock);
5060 5061
5061 5062 ip_fanout_proto_conn(connp, mp, ipha, NULL, ira);
5062 5063
5063 5064 CONN_DEC_REF(connp);
5064 5065 }
5065 5066
5066 5067 /*
5067 5068 * If we have a IPsec NAT-Traversal packet, strip the zero-SPI or
5068 5069 * pass it along to ESP if the SPI is non-zero. Returns the mblk if the mblk
5069 5070 * is not consumed.
5070 5071 *
5071 5072 * One of three things can happen, all of which affect the passed-in mblk:
5072 5073 *
5073 5074 * 1.) The packet is stock UDP and gets its zero-SPI stripped. Return mblk..
5074 5075 *
5075 5076 * 2.) The packet is ESP-in-UDP, gets transformed into an equivalent
5076 5077 * ESP packet, and is passed along to ESP for consumption. Return NULL.
5077 5078 *
5078 5079 * 3.) The packet is an ESP-in-UDP Keepalive. Drop it and return NULL.
5079 5080 */
5080 5081 mblk_t *
5081 5082 zero_spi_check(mblk_t *mp, ip_recv_attr_t *ira)
5082 5083 {
5083 5084 int shift, plen, iph_len;
5084 5085 ipha_t *ipha;
5085 5086 udpha_t *udpha;
5086 5087 uint32_t *spi;
5087 5088 uint32_t esp_ports;
5088 5089 uint8_t *orptr;
5089 5090 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
5090 5091 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
5091 5092
5092 5093 ipha = (ipha_t *)mp->b_rptr;
5093 5094 iph_len = ira->ira_ip_hdr_length;
5094 5095 plen = ira->ira_pktlen;
5095 5096
5096 5097 if (plen - iph_len - sizeof (udpha_t) < sizeof (uint32_t)) {
5097 5098 /*
5098 5099 * Most likely a keepalive for the benefit of an intervening
5099 5100 * NAT. These aren't for us, per se, so drop it.
5100 5101 *
5101 5102 * RFC 3947/8 doesn't say for sure what to do for 2-3
5102 5103 * byte packets (keepalives are 1-byte), but we'll drop them
5103 5104 * also.
5104 5105 */
5105 5106 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
5106 5107 DROPPER(ipss, ipds_esp_nat_t_ka), &ipss->ipsec_dropper);
5107 5108 return (NULL);
5108 5109 }
5109 5110
5110 5111 if (MBLKL(mp) < iph_len + sizeof (udpha_t) + sizeof (*spi)) {
5111 5112 /* might as well pull it all up - it might be ESP. */
5112 5113 if (!pullupmsg(mp, -1)) {
5113 5114 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
5114 5115 DROPPER(ipss, ipds_esp_nomem),
5115 5116 &ipss->ipsec_dropper);
5116 5117 return (NULL);
5117 5118 }
5118 5119
5119 5120 ipha = (ipha_t *)mp->b_rptr;
5120 5121 }
5121 5122 spi = (uint32_t *)(mp->b_rptr + iph_len + sizeof (udpha_t));
5122 5123 if (*spi == 0) {
5123 5124 /* UDP packet - remove 0-spi. */
5124 5125 shift = sizeof (uint32_t);
5125 5126 } else {
5126 5127 /* ESP-in-UDP packet - reduce to ESP. */
5127 5128 ipha->ipha_protocol = IPPROTO_ESP;
5128 5129 shift = sizeof (udpha_t);
5129 5130 }
5130 5131
5131 5132 /* Fix IP header */
5132 5133 ira->ira_pktlen = (plen - shift);
5133 5134 ipha->ipha_length = htons(ira->ira_pktlen);
5134 5135 ipha->ipha_hdr_checksum = 0;
5135 5136
5136 5137 orptr = mp->b_rptr;
5137 5138 mp->b_rptr += shift;
5138 5139
5139 5140 udpha = (udpha_t *)(orptr + iph_len);
5140 5141 if (*spi == 0) {
5141 5142 ASSERT((uint8_t *)ipha == orptr);
5142 5143 udpha->uha_length = htons(plen - shift - iph_len);
5143 5144 iph_len += sizeof (udpha_t); /* For the call to ovbcopy(). */
5144 5145 esp_ports = 0;
5145 5146 } else {
5146 5147 esp_ports = *((uint32_t *)udpha);
5147 5148 ASSERT(esp_ports != 0);
5148 5149 }
5149 5150 ovbcopy(orptr, orptr + shift, iph_len);
5150 5151 if (esp_ports != 0) /* Punt up for ESP processing. */ {
5151 5152 ipha = (ipha_t *)(orptr + shift);
5152 5153
5153 5154 ira->ira_flags |= IRAF_ESP_UDP_PORTS;
5154 5155 ira->ira_esp_udp_ports = esp_ports;
5155 5156 ip_fanout_v4(mp, ipha, ira);
5156 5157 return (NULL);
5157 5158 }
5158 5159 return (mp);
5159 5160 }
5160 5161
5161 5162 /*
5162 5163 * Deliver a udp packet to the given conn, possibly applying ipsec policy.
5163 5164 * Handles IPv4 and IPv6.
5164 5165 * We are responsible for disposing of mp, such as by freemsg() or putnext()
5165 5166 * Caller is responsible for dropping references to the conn.
5166 5167 */
5167 5168 void
5168 5169 ip_fanout_udp_conn(conn_t *connp, mblk_t *mp, ipha_t *ipha, ip6_t *ip6h,
5169 5170 ip_recv_attr_t *ira)
5170 5171 {
5171 5172 ill_t *ill = ira->ira_ill;
5172 5173 ip_stack_t *ipst = ill->ill_ipst;
5173 5174 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
5174 5175 boolean_t secure;
5175 5176 iaflags_t iraflags = ira->ira_flags;
5176 5177
5177 5178 secure = iraflags & IRAF_IPSEC_SECURE;
5178 5179
5179 5180 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld :
5180 5181 !canputnext(connp->conn_rq)) {
5181 5182 BUMP_MIB(ill->ill_ip_mib, udpIfStatsInOverflows);
5182 5183 freemsg(mp);
5183 5184 return;
5184 5185 }
5185 5186
5186 5187 if (((iraflags & IRAF_IS_IPV4) ?
5187 5188 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
5188 5189 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
5189 5190 secure) {
5190 5191 mp = ipsec_check_inbound_policy(mp, connp, ipha,
5191 5192 ip6h, ira);
5192 5193 if (mp == NULL) {
5193 5194 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5194 5195 /* Note that mp is NULL */
5195 5196 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5196 5197 return;
5197 5198 }
5198 5199 }
5199 5200
5200 5201 /*
5201 5202 * Since this code is not used for UDP unicast we don't need a NAT_T
5202 5203 * check. Only ip_fanout_v4 has that check.
5203 5204 */
5204 5205 if (ira->ira_flags & IRAF_ICMP_ERROR) {
5205 5206 (connp->conn_recvicmp)(connp, mp, NULL, ira);
5206 5207 } else {
5207 5208 ill_t *rill = ira->ira_rill;
5208 5209
5209 5210 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
5210 5211 ira->ira_ill = ira->ira_rill = NULL;
5211 5212 /* Send it upstream */
5212 5213 (connp->conn_recv)(connp, mp, NULL, ira);
5213 5214 ira->ira_ill = ill;
5214 5215 ira->ira_rill = rill;
5215 5216 }
5216 5217 }
5217 5218
5218 5219 /*
5219 5220 * Fanout for UDP packets that are multicast or broadcast, and ICMP errors.
5220 5221 * (Unicast fanout is handled in ip_input_v4.)
5221 5222 *
5222 5223 * If SO_REUSEADDR is set all multicast and broadcast packets
5223 5224 * will be delivered to all conns bound to the same port.
5224 5225 *
5225 5226 * If there is at least one matching AF_INET receiver, then we will
5226 5227 * ignore any AF_INET6 receivers.
5227 5228 * In the special case where an AF_INET socket binds to 0.0.0.0/<port> and an
5228 5229 * AF_INET6 socket binds to ::/<port>, only the AF_INET socket receives the IPv4
5229 5230 * packets.
5230 5231 *
5231 5232 * Zones notes:
5232 5233 * Earlier in ip_input on a system with multiple shared-IP zones we
5233 5234 * duplicate the multicast and broadcast packets and send them up
5234 5235 * with each explicit zoneid that exists on that ill.
5235 5236 * This means that here we can match the zoneid with SO_ALLZONES being special.
5236 5237 */
5237 5238 void
5238 5239 ip_fanout_udp_multi_v4(mblk_t *mp, ipha_t *ipha, uint16_t lport, uint16_t fport,
5239 5240 ip_recv_attr_t *ira)
5240 5241 {
5241 5242 ipaddr_t laddr;
5242 5243 in6_addr_t v6faddr;
5243 5244 conn_t *connp;
5244 5245 connf_t *connfp;
5245 5246 ipaddr_t faddr;
5246 5247 ill_t *ill = ira->ira_ill;
5247 5248 ip_stack_t *ipst = ill->ill_ipst;
5248 5249
5249 5250 ASSERT(ira->ira_flags & (IRAF_MULTIBROADCAST|IRAF_ICMP_ERROR));
5250 5251
5251 5252 laddr = ipha->ipha_dst;
5252 5253 faddr = ipha->ipha_src;
5253 5254
5254 5255 connfp = &ipst->ips_ipcl_udp_fanout[IPCL_UDP_HASH(lport, ipst)];
5255 5256 mutex_enter(&connfp->connf_lock);
5256 5257 connp = connfp->connf_head;
5257 5258
5258 5259 /*
5259 5260 * If SO_REUSEADDR has been set on the first we send the
5260 5261 * packet to all clients that have joined the group and
5261 5262 * match the port.
5262 5263 */
5263 5264 while (connp != NULL) {
5264 5265 if ((IPCL_UDP_MATCH(connp, lport, laddr, fport, faddr)) &&
5265 5266 conn_wantpacket(connp, ira, ipha) &&
5266 5267 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5267 5268 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp)))
5268 5269 break;
5269 5270 connp = connp->conn_next;
5270 5271 }
5271 5272
5272 5273 if (connp == NULL)
5273 5274 goto notfound;
5274 5275
5275 5276 CONN_INC_REF(connp);
5276 5277
5277 5278 if (connp->conn_reuseaddr) {
5278 5279 conn_t *first_connp = connp;
5279 5280 conn_t *next_connp;
5280 5281 mblk_t *mp1;
5281 5282
5282 5283 connp = connp->conn_next;
5283 5284 for (;;) {
5284 5285 while (connp != NULL) {
5285 5286 if (IPCL_UDP_MATCH(connp, lport, laddr,
5286 5287 fport, faddr) &&
5287 5288 conn_wantpacket(connp, ira, ipha) &&
5288 5289 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5289 5290 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5290 5291 ira, connp)))
5291 5292 break;
5292 5293 connp = connp->conn_next;
5293 5294 }
5294 5295 if (connp == NULL) {
5295 5296 /* No more interested clients */
5296 5297 connp = first_connp;
5297 5298 break;
5298 5299 }
5299 5300 if (((mp1 = dupmsg(mp)) == NULL) &&
5300 5301 ((mp1 = copymsg(mp)) == NULL)) {
5301 5302 /* Memory allocation failed */
5302 5303 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5303 5304 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5304 5305 connp = first_connp;
5305 5306 break;
5306 5307 }
5307 5308 CONN_INC_REF(connp);
5308 5309 mutex_exit(&connfp->connf_lock);
5309 5310
5310 5311 IP_STAT(ipst, ip_udp_fanmb);
5311 5312 ip_fanout_udp_conn(connp, mp1, (ipha_t *)mp1->b_rptr,
5312 5313 NULL, ira);
5313 5314 mutex_enter(&connfp->connf_lock);
5314 5315 /* Follow the next pointer before releasing the conn */
5315 5316 next_connp = connp->conn_next;
5316 5317 CONN_DEC_REF(connp);
5317 5318 connp = next_connp;
5318 5319 }
5319 5320 }
5320 5321
5321 5322 /* Last one. Send it upstream. */
5322 5323 mutex_exit(&connfp->connf_lock);
5323 5324 IP_STAT(ipst, ip_udp_fanmb);
5324 5325 ip_fanout_udp_conn(connp, mp, ipha, NULL, ira);
5325 5326 CONN_DEC_REF(connp);
5326 5327 return;
5327 5328
5328 5329 notfound:
5329 5330 mutex_exit(&connfp->connf_lock);
5330 5331 /*
5331 5332 * IPv6 endpoints bound to multicast IPv4-mapped addresses
5332 5333 * have already been matched above, since they live in the IPv4
5333 5334 * fanout tables. This implies we only need to
5334 5335 * check for IPv6 in6addr_any endpoints here.
5335 5336 * Thus we compare using ipv6_all_zeros instead of the destination
5336 5337 * address, except for the multicast group membership lookup which
5337 5338 * uses the IPv4 destination.
5338 5339 */
5339 5340 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_src, &v6faddr);
5340 5341 connfp = &ipst->ips_ipcl_udp_fanout[IPCL_UDP_HASH(lport, ipst)];
5341 5342 mutex_enter(&connfp->connf_lock);
5342 5343 connp = connfp->connf_head;
5343 5344 /*
5344 5345 * IPv4 multicast packet being delivered to an AF_INET6
5345 5346 * in6addr_any endpoint.
5346 5347 * Need to check conn_wantpacket(). Note that we use conn_wantpacket()
5347 5348 * and not conn_wantpacket_v6() since any multicast membership is
5348 5349 * for an IPv4-mapped multicast address.
5349 5350 */
5350 5351 while (connp != NULL) {
5351 5352 if (IPCL_UDP_MATCH_V6(connp, lport, ipv6_all_zeros,
5352 5353 fport, v6faddr) &&
5353 5354 conn_wantpacket(connp, ira, ipha) &&
5354 5355 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5355 5356 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp)))
5356 5357 break;
5357 5358 connp = connp->conn_next;
5358 5359 }
5359 5360
5360 5361 if (connp == NULL) {
5361 5362 /*
5362 5363 * No one bound to this port. Is
5363 5364 * there a client that wants all
5364 5365 * unclaimed datagrams?
5365 5366 */
5366 5367 mutex_exit(&connfp->connf_lock);
5367 5368
5368 5369 if (ipst->ips_ipcl_proto_fanout_v4[IPPROTO_UDP].connf_head !=
5369 5370 NULL) {
5370 5371 ASSERT(ira->ira_protocol == IPPROTO_UDP);
5371 5372 ip_fanout_proto_v4(mp, ipha, ira);
5372 5373 } else {
5373 5374 /*
5374 5375 * We used to attempt to send an icmp error here, but
5375 5376 * since this is known to be a multicast packet
5376 5377 * and we don't send icmp errors in response to
5377 5378 * multicast, just drop the packet and give up sooner.
5378 5379 */
5379 5380 BUMP_MIB(ill->ill_ip_mib, udpIfStatsNoPorts);
5380 5381 freemsg(mp);
5381 5382 }
5382 5383 return;
5383 5384 }
5384 5385 ASSERT(IPCL_IS_NONSTR(connp) || connp->conn_rq != NULL);
5385 5386
5386 5387 /*
5387 5388 * If SO_REUSEADDR has been set on the first we send the
5388 5389 * packet to all clients that have joined the group and
5389 5390 * match the port.
5390 5391 */
5391 5392 if (connp->conn_reuseaddr) {
5392 5393 conn_t *first_connp = connp;
5393 5394 conn_t *next_connp;
5394 5395 mblk_t *mp1;
5395 5396
5396 5397 CONN_INC_REF(connp);
5397 5398 connp = connp->conn_next;
5398 5399 for (;;) {
5399 5400 while (connp != NULL) {
5400 5401 if (IPCL_UDP_MATCH_V6(connp, lport,
5401 5402 ipv6_all_zeros, fport, v6faddr) &&
5402 5403 conn_wantpacket(connp, ira, ipha) &&
5403 5404 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5404 5405 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5405 5406 ira, connp)))
5406 5407 break;
5407 5408 connp = connp->conn_next;
5408 5409 }
5409 5410 if (connp == NULL) {
5410 5411 /* No more interested clients */
5411 5412 connp = first_connp;
5412 5413 break;
5413 5414 }
5414 5415 if (((mp1 = dupmsg(mp)) == NULL) &&
5415 5416 ((mp1 = copymsg(mp)) == NULL)) {
5416 5417 /* Memory allocation failed */
5417 5418 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5418 5419 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5419 5420 connp = first_connp;
5420 5421 break;
5421 5422 }
5422 5423 CONN_INC_REF(connp);
5423 5424 mutex_exit(&connfp->connf_lock);
5424 5425
5425 5426 IP_STAT(ipst, ip_udp_fanmb);
5426 5427 ip_fanout_udp_conn(connp, mp1, (ipha_t *)mp1->b_rptr,
5427 5428 NULL, ira);
5428 5429 mutex_enter(&connfp->connf_lock);
5429 5430 /* Follow the next pointer before releasing the conn */
5430 5431 next_connp = connp->conn_next;
5431 5432 CONN_DEC_REF(connp);
5432 5433 connp = next_connp;
5433 5434 }
5434 5435 }
5435 5436
5436 5437 /* Last one. Send it upstream. */
5437 5438 mutex_exit(&connfp->connf_lock);
5438 5439 IP_STAT(ipst, ip_udp_fanmb);
5439 5440 ip_fanout_udp_conn(connp, mp, ipha, NULL, ira);
5440 5441 CONN_DEC_REF(connp);
5441 5442 }
5442 5443
5443 5444 /*
5444 5445 * Split an incoming packet's IPv4 options into the label and the other options.
5445 5446 * If 'allocate' is set it does memory allocation for the ip_pkt_t, including
5446 5447 * clearing out any leftover label or options.
5447 5448 * Otherwise it just makes ipp point into the packet.
5448 5449 *
5449 5450 * Returns zero if ok; ENOMEM if the buffer couldn't be allocated.
5450 5451 */
5451 5452 int
5452 5453 ip_find_hdr_v4(ipha_t *ipha, ip_pkt_t *ipp, boolean_t allocate)
5453 5454 {
5454 5455 uchar_t *opt;
5455 5456 uint32_t totallen;
5456 5457 uint32_t optval;
5457 5458 uint32_t optlen;
5458 5459
5459 5460 ipp->ipp_fields |= IPPF_HOPLIMIT | IPPF_TCLASS | IPPF_ADDR;
5460 5461 ipp->ipp_hoplimit = ipha->ipha_ttl;
5461 5462 ipp->ipp_type_of_service = ipha->ipha_type_of_service;
5462 5463 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &ipp->ipp_addr);
5463 5464
5464 5465 /*
5465 5466 * Get length (in 4 byte octets) of IP header options.
5466 5467 */
5467 5468 totallen = ipha->ipha_version_and_hdr_length -
5468 5469 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
5469 5470
5470 5471 if (totallen == 0) {
5471 5472 if (!allocate)
5472 5473 return (0);
5473 5474
5474 5475 /* Clear out anything from a previous packet */
5475 5476 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
5476 5477 kmem_free(ipp->ipp_ipv4_options,
5477 5478 ipp->ipp_ipv4_options_len);
5478 5479 ipp->ipp_ipv4_options = NULL;
5479 5480 ipp->ipp_ipv4_options_len = 0;
5480 5481 ipp->ipp_fields &= ~IPPF_IPV4_OPTIONS;
5481 5482 }
5482 5483 if (ipp->ipp_fields & IPPF_LABEL_V4) {
5483 5484 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
5484 5485 ipp->ipp_label_v4 = NULL;
5485 5486 ipp->ipp_label_len_v4 = 0;
5486 5487 ipp->ipp_fields &= ~IPPF_LABEL_V4;
5487 5488 }
5488 5489 return (0);
5489 5490 }
5490 5491
5491 5492 totallen <<= 2;
5492 5493 opt = (uchar_t *)&ipha[1];
5493 5494 if (!is_system_labeled()) {
5494 5495
5495 5496 copyall:
5496 5497 if (!allocate) {
5497 5498 if (totallen != 0) {
5498 5499 ipp->ipp_ipv4_options = opt;
5499 5500 ipp->ipp_ipv4_options_len = totallen;
5500 5501 ipp->ipp_fields |= IPPF_IPV4_OPTIONS;
5501 5502 }
5502 5503 return (0);
5503 5504 }
5504 5505 /* Just copy all of options */
5505 5506 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
5506 5507 if (totallen == ipp->ipp_ipv4_options_len) {
5507 5508 bcopy(opt, ipp->ipp_ipv4_options, totallen);
5508 5509 return (0);
5509 5510 }
5510 5511 kmem_free(ipp->ipp_ipv4_options,
5511 5512 ipp->ipp_ipv4_options_len);
5512 5513 ipp->ipp_ipv4_options = NULL;
5513 5514 ipp->ipp_ipv4_options_len = 0;
5514 5515 ipp->ipp_fields &= ~IPPF_IPV4_OPTIONS;
5515 5516 }
5516 5517 if (totallen == 0)
5517 5518 return (0);
5518 5519
5519 5520 ipp->ipp_ipv4_options = kmem_alloc(totallen, KM_NOSLEEP);
5520 5521 if (ipp->ipp_ipv4_options == NULL)
5521 5522 return (ENOMEM);
5522 5523 ipp->ipp_ipv4_options_len = totallen;
5523 5524 ipp->ipp_fields |= IPPF_IPV4_OPTIONS;
5524 5525 bcopy(opt, ipp->ipp_ipv4_options, totallen);
5525 5526 return (0);
5526 5527 }
5527 5528
5528 5529 if (allocate && (ipp->ipp_fields & IPPF_LABEL_V4)) {
5529 5530 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
5530 5531 ipp->ipp_label_v4 = NULL;
5531 5532 ipp->ipp_label_len_v4 = 0;
5532 5533 ipp->ipp_fields &= ~IPPF_LABEL_V4;
5533 5534 }
5534 5535
5535 5536 /*
5536 5537 * Search for CIPSO option.
5537 5538 * We assume CIPSO is first in options if it is present.
5538 5539 * If it isn't, then ipp_opt_ipv4_options will not include the options
5539 5540 * prior to the CIPSO option.
5540 5541 */
5541 5542 while (totallen != 0) {
5542 5543 switch (optval = opt[IPOPT_OPTVAL]) {
5543 5544 case IPOPT_EOL:
5544 5545 return (0);
5545 5546 case IPOPT_NOP:
5546 5547 optlen = 1;
5547 5548 break;
5548 5549 default:
5549 5550 if (totallen <= IPOPT_OLEN)
5550 5551 return (EINVAL);
5551 5552 optlen = opt[IPOPT_OLEN];
5552 5553 if (optlen < 2)
5553 5554 return (EINVAL);
5554 5555 }
5555 5556 if (optlen > totallen)
5556 5557 return (EINVAL);
5557 5558
5558 5559 switch (optval) {
5559 5560 case IPOPT_COMSEC:
5560 5561 if (!allocate) {
5561 5562 ipp->ipp_label_v4 = opt;
5562 5563 ipp->ipp_label_len_v4 = optlen;
5563 5564 ipp->ipp_fields |= IPPF_LABEL_V4;
5564 5565 } else {
5565 5566 ipp->ipp_label_v4 = kmem_alloc(optlen,
5566 5567 KM_NOSLEEP);
5567 5568 if (ipp->ipp_label_v4 == NULL)
5568 5569 return (ENOMEM);
5569 5570 ipp->ipp_label_len_v4 = optlen;
5570 5571 ipp->ipp_fields |= IPPF_LABEL_V4;
5571 5572 bcopy(opt, ipp->ipp_label_v4, optlen);
5572 5573 }
5573 5574 totallen -= optlen;
5574 5575 opt += optlen;
5575 5576
5576 5577 /* Skip padding bytes until we get to a multiple of 4 */
5577 5578 while ((totallen & 3) != 0 && opt[0] == IPOPT_NOP) {
5578 5579 totallen--;
5579 5580 opt++;
5580 5581 }
5581 5582 /* Remaining as ipp_ipv4_options */
5582 5583 goto copyall;
5583 5584 }
5584 5585 totallen -= optlen;
5585 5586 opt += optlen;
5586 5587 }
5587 5588 /* No CIPSO found; return everything as ipp_ipv4_options */
5588 5589 totallen = ipha->ipha_version_and_hdr_length -
5589 5590 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
5590 5591 totallen <<= 2;
5591 5592 opt = (uchar_t *)&ipha[1];
5592 5593 goto copyall;
5593 5594 }
5594 5595
5595 5596 /*
5596 5597 * Efficient versions of lookup for an IRE when we only
5597 5598 * match the address.
5598 5599 * For RTF_REJECT or BLACKHOLE we return IRE_NOROUTE.
5599 5600 * Does not handle multicast addresses.
5600 5601 */
5601 5602 uint_t
5602 5603 ip_type_v4(ipaddr_t addr, ip_stack_t *ipst)
5603 5604 {
5604 5605 ire_t *ire;
5605 5606 uint_t result;
5606 5607
5607 5608 ire = ire_ftable_lookup_simple_v4(addr, 0, ipst, NULL);
5608 5609 ASSERT(ire != NULL);
5609 5610 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))
5610 5611 result = IRE_NOROUTE;
5611 5612 else
5612 5613 result = ire->ire_type;
5613 5614 ire_refrele(ire);
5614 5615 return (result);
5615 5616 }
5616 5617
5617 5618 /*
5618 5619 * Efficient versions of lookup for an IRE when we only
5619 5620 * match the address.
5620 5621 * For RTF_REJECT or BLACKHOLE we return IRE_NOROUTE.
5621 5622 * Does not handle multicast addresses.
5622 5623 */
5623 5624 uint_t
5624 5625 ip_type_v6(const in6_addr_t *addr, ip_stack_t *ipst)
5625 5626 {
5626 5627 ire_t *ire;
5627 5628 uint_t result;
5628 5629
5629 5630 ire = ire_ftable_lookup_simple_v6(addr, 0, ipst, NULL);
5630 5631 ASSERT(ire != NULL);
5631 5632 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))
5632 5633 result = IRE_NOROUTE;
5633 5634 else
5634 5635 result = ire->ire_type;
5635 5636 ire_refrele(ire);
5636 5637 return (result);
5637 5638 }
5638 5639
5639 5640 /*
5640 5641 * Nobody should be sending
5641 5642 * packets up this stream
5642 5643 */
5643 5644 static void
5644 5645 ip_lrput(queue_t *q, mblk_t *mp)
5645 5646 {
5646 5647 switch (mp->b_datap->db_type) {
5647 5648 case M_FLUSH:
5648 5649 /* Turn around */
5649 5650 if (*mp->b_rptr & FLUSHW) {
5650 5651 *mp->b_rptr &= ~FLUSHR;
5651 5652 qreply(q, mp);
5652 5653 return;
5653 5654 }
5654 5655 break;
5655 5656 }
5656 5657 freemsg(mp);
5657 5658 }
5658 5659
5659 5660 /* Nobody should be sending packets down this stream */
5660 5661 /* ARGSUSED */
5661 5662 void
5662 5663 ip_lwput(queue_t *q, mblk_t *mp)
5663 5664 {
5664 5665 freemsg(mp);
5665 5666 }
5666 5667
5667 5668 /*
5668 5669 * Move the first hop in any source route to ipha_dst and remove that part of
5669 5670 * the source route. Called by other protocols. Errors in option formatting
5670 5671 * are ignored - will be handled by ip_output_options. Return the final
5671 5672 * destination (either ipha_dst or the last entry in a source route.)
5672 5673 */
5673 5674 ipaddr_t
5674 5675 ip_massage_options(ipha_t *ipha, netstack_t *ns)
5675 5676 {
5676 5677 ipoptp_t opts;
5677 5678 uchar_t *opt;
5678 5679 uint8_t optval;
5679 5680 uint8_t optlen;
5680 5681 ipaddr_t dst;
5681 5682 int i;
5682 5683 ip_stack_t *ipst = ns->netstack_ip;
5683 5684
5684 5685 ip2dbg(("ip_massage_options\n"));
5685 5686 dst = ipha->ipha_dst;
5686 5687 for (optval = ipoptp_first(&opts, ipha);
5687 5688 optval != IPOPT_EOL;
5688 5689 optval = ipoptp_next(&opts)) {
5689 5690 opt = opts.ipoptp_cur;
5690 5691 switch (optval) {
5691 5692 uint8_t off;
5692 5693 case IPOPT_SSRR:
5693 5694 case IPOPT_LSRR:
5694 5695 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
5695 5696 ip1dbg(("ip_massage_options: bad src route\n"));
5696 5697 break;
5697 5698 }
5698 5699 optlen = opts.ipoptp_len;
5699 5700 off = opt[IPOPT_OFFSET];
5700 5701 off--;
5701 5702 redo_srr:
5702 5703 if (optlen < IP_ADDR_LEN ||
5703 5704 off > optlen - IP_ADDR_LEN) {
5704 5705 /* End of source route */
5705 5706 ip1dbg(("ip_massage_options: end of SR\n"));
5706 5707 break;
5707 5708 }
5708 5709 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
5709 5710 ip1dbg(("ip_massage_options: next hop 0x%x\n",
5710 5711 ntohl(dst)));
5711 5712 /*
5712 5713 * Check if our address is present more than
5713 5714 * once as consecutive hops in source route.
5714 5715 * XXX verify per-interface ip_forwarding
5715 5716 * for source route?
5716 5717 */
5717 5718 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
5718 5719 off += IP_ADDR_LEN;
5719 5720 goto redo_srr;
5720 5721 }
5721 5722 if (dst == htonl(INADDR_LOOPBACK)) {
5722 5723 ip1dbg(("ip_massage_options: loopback addr in "
5723 5724 "source route!\n"));
5724 5725 break;
5725 5726 }
5726 5727 /*
5727 5728 * Update ipha_dst to be the first hop and remove the
5728 5729 * first hop from the source route (by overwriting
5729 5730 * part of the option with NOP options).
5730 5731 */
5731 5732 ipha->ipha_dst = dst;
5732 5733 /* Put the last entry in dst */
5733 5734 off = ((optlen - IP_ADDR_LEN - 3) & ~(IP_ADDR_LEN-1)) +
5734 5735 3;
5735 5736 bcopy(&opt[off], &dst, IP_ADDR_LEN);
5736 5737
5737 5738 ip1dbg(("ip_massage_options: last hop 0x%x\n",
5738 5739 ntohl(dst)));
5739 5740 /* Move down and overwrite */
5740 5741 opt[IP_ADDR_LEN] = opt[0];
5741 5742 opt[IP_ADDR_LEN+1] = opt[IPOPT_OLEN] - IP_ADDR_LEN;
5742 5743 opt[IP_ADDR_LEN+2] = opt[IPOPT_OFFSET];
5743 5744 for (i = 0; i < IP_ADDR_LEN; i++)
5744 5745 opt[i] = IPOPT_NOP;
5745 5746 break;
5746 5747 }
5747 5748 }
5748 5749 return (dst);
5749 5750 }
5750 5751
5751 5752 /*
5752 5753 * Return the network mask
5753 5754 * associated with the specified address.
5754 5755 */
5755 5756 ipaddr_t
5756 5757 ip_net_mask(ipaddr_t addr)
5757 5758 {
5758 5759 uchar_t *up = (uchar_t *)&addr;
5759 5760 ipaddr_t mask = 0;
5760 5761 uchar_t *maskp = (uchar_t *)&mask;
5761 5762
5762 5763 #if defined(__i386) || defined(__amd64)
5763 5764 #define TOTALLY_BRAIN_DAMAGED_C_COMPILER
5764 5765 #endif
5765 5766 #ifdef TOTALLY_BRAIN_DAMAGED_C_COMPILER
5766 5767 maskp[0] = maskp[1] = maskp[2] = maskp[3] = 0;
5767 5768 #endif
5768 5769 if (CLASSD(addr)) {
5769 5770 maskp[0] = 0xF0;
5770 5771 return (mask);
5771 5772 }
5772 5773
5773 5774 /* We assume Class E default netmask to be 32 */
5774 5775 if (CLASSE(addr))
5775 5776 return (0xffffffffU);
5776 5777
5777 5778 if (addr == 0)
5778 5779 return (0);
5779 5780 maskp[0] = 0xFF;
5780 5781 if ((up[0] & 0x80) == 0)
5781 5782 return (mask);
5782 5783
5783 5784 maskp[1] = 0xFF;
5784 5785 if ((up[0] & 0xC0) == 0x80)
5785 5786 return (mask);
5786 5787
5787 5788 maskp[2] = 0xFF;
5788 5789 if ((up[0] & 0xE0) == 0xC0)
5789 5790 return (mask);
5790 5791
5791 5792 /* Otherwise return no mask */
5792 5793 return ((ipaddr_t)0);
5793 5794 }
5794 5795
5795 5796 /* Name/Value Table Lookup Routine */
5796 5797 char *
5797 5798 ip_nv_lookup(nv_t *nv, int value)
5798 5799 {
5799 5800 if (!nv)
5800 5801 return (NULL);
5801 5802 for (; nv->nv_name; nv++) {
5802 5803 if (nv->nv_value == value)
5803 5804 return (nv->nv_name);
5804 5805 }
5805 5806 return ("unknown");
5806 5807 }
5807 5808
5808 5809 static int
5809 5810 ip_wait_for_info_ack(ill_t *ill)
5810 5811 {
5811 5812 int err;
5812 5813
5813 5814 mutex_enter(&ill->ill_lock);
5814 5815 while (ill->ill_state_flags & ILL_LL_SUBNET_PENDING) {
5815 5816 /*
5816 5817 * Return value of 0 indicates a pending signal.
5817 5818 */
5818 5819 err = cv_wait_sig(&ill->ill_cv, &ill->ill_lock);
5819 5820 if (err == 0) {
5820 5821 mutex_exit(&ill->ill_lock);
5821 5822 return (EINTR);
5822 5823 }
5823 5824 }
5824 5825 mutex_exit(&ill->ill_lock);
5825 5826 /*
5826 5827 * ip_rput_other could have set an error in ill_error on
5827 5828 * receipt of M_ERROR.
5828 5829 */
5829 5830 return (ill->ill_error);
5830 5831 }
5831 5832
5832 5833 /*
5833 5834 * This is a module open, i.e. this is a control stream for access
5834 5835 * to a DLPI device. We allocate an ill_t as the instance data in
5835 5836 * this case.
5836 5837 */
5837 5838 static int
5838 5839 ip_modopen(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5839 5840 {
5840 5841 ill_t *ill;
5841 5842 int err;
5842 5843 zoneid_t zoneid;
5843 5844 netstack_t *ns;
5844 5845 ip_stack_t *ipst;
5845 5846
5846 5847 /*
5847 5848 * Prevent unprivileged processes from pushing IP so that
5848 5849 * they can't send raw IP.
5849 5850 */
5850 5851 if (secpolicy_net_rawaccess(credp) != 0)
5851 5852 return (EPERM);
5852 5853
5853 5854 ns = netstack_find_by_cred(credp);
5854 5855 ASSERT(ns != NULL);
5855 5856 ipst = ns->netstack_ip;
5856 5857 ASSERT(ipst != NULL);
5857 5858
5858 5859 /*
5859 5860 * For exclusive stacks we set the zoneid to zero
5860 5861 * to make IP operate as if in the global zone.
5861 5862 */
5862 5863 if (ipst->ips_netstack->netstack_stackid != GLOBAL_NETSTACKID)
5863 5864 zoneid = GLOBAL_ZONEID;
5864 5865 else
5865 5866 zoneid = crgetzoneid(credp);
5866 5867
5867 5868 ill = (ill_t *)mi_open_alloc_sleep(sizeof (ill_t));
5868 5869 q->q_ptr = WR(q)->q_ptr = ill;
5869 5870 ill->ill_ipst = ipst;
5870 5871 ill->ill_zoneid = zoneid;
5871 5872
5872 5873 /*
5873 5874 * ill_init initializes the ill fields and then sends down
5874 5875 * down a DL_INFO_REQ after calling qprocson.
5875 5876 */
5876 5877 err = ill_init(q, ill);
5877 5878
5878 5879 if (err != 0) {
5879 5880 mi_free(ill);
5880 5881 netstack_rele(ipst->ips_netstack);
5881 5882 q->q_ptr = NULL;
5882 5883 WR(q)->q_ptr = NULL;
5883 5884 return (err);
5884 5885 }
5885 5886
5886 5887 /*
5887 5888 * Wait for the DL_INFO_ACK if a DL_INFO_REQ was sent.
5888 5889 *
5889 5890 * ill_init initializes the ipsq marking this thread as
5890 5891 * writer
5891 5892 */
5892 5893 ipsq_exit(ill->ill_phyint->phyint_ipsq);
5893 5894 err = ip_wait_for_info_ack(ill);
5894 5895 if (err == 0)
5895 5896 ill->ill_credp = credp;
5896 5897 else
5897 5898 goto fail;
5898 5899
5899 5900 crhold(credp);
5900 5901
5901 5902 mutex_enter(&ipst->ips_ip_mi_lock);
5902 5903 err = mi_open_link(&ipst->ips_ip_g_head, (IDP)q->q_ptr, devp, flag,
5903 5904 sflag, credp);
5904 5905 mutex_exit(&ipst->ips_ip_mi_lock);
5905 5906 fail:
5906 5907 if (err) {
5907 5908 (void) ip_close(q, 0);
5908 5909 return (err);
5909 5910 }
5910 5911 return (0);
5911 5912 }
5912 5913
5913 5914 /* For /dev/ip aka AF_INET open */
5914 5915 int
5915 5916 ip_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5916 5917 {
5917 5918 return (ip_open(q, devp, flag, sflag, credp, B_FALSE));
5918 5919 }
5919 5920
5920 5921 /* For /dev/ip6 aka AF_INET6 open */
5921 5922 int
5922 5923 ip_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5923 5924 {
5924 5925 return (ip_open(q, devp, flag, sflag, credp, B_TRUE));
5925 5926 }
5926 5927
5927 5928 /* IP open routine. */
5928 5929 int
5929 5930 ip_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp,
5930 5931 boolean_t isv6)
5931 5932 {
5932 5933 conn_t *connp;
5933 5934 major_t maj;
5934 5935 zoneid_t zoneid;
5935 5936 netstack_t *ns;
5936 5937 ip_stack_t *ipst;
5937 5938
5938 5939 /* Allow reopen. */
5939 5940 if (q->q_ptr != NULL)
5940 5941 return (0);
5941 5942
5942 5943 if (sflag & MODOPEN) {
5943 5944 /* This is a module open */
5944 5945 return (ip_modopen(q, devp, flag, sflag, credp));
5945 5946 }
5946 5947
5947 5948 if ((flag & ~(FKLYR)) == IP_HELPER_STR) {
5948 5949 /*
5949 5950 * Non streams based socket looking for a stream
5950 5951 * to access IP
5951 5952 */
5952 5953 return (ip_helper_stream_setup(q, devp, flag, sflag,
5953 5954 credp, isv6));
5954 5955 }
5955 5956
5956 5957 ns = netstack_find_by_cred(credp);
5957 5958 ASSERT(ns != NULL);
5958 5959 ipst = ns->netstack_ip;
5959 5960 ASSERT(ipst != NULL);
5960 5961
5961 5962 /*
5962 5963 * For exclusive stacks we set the zoneid to zero
5963 5964 * to make IP operate as if in the global zone.
5964 5965 */
5965 5966 if (ipst->ips_netstack->netstack_stackid != GLOBAL_NETSTACKID)
5966 5967 zoneid = GLOBAL_ZONEID;
5967 5968 else
5968 5969 zoneid = crgetzoneid(credp);
5969 5970
5970 5971 /*
5971 5972 * We are opening as a device. This is an IP client stream, and we
5972 5973 * allocate an conn_t as the instance data.
5973 5974 */
5974 5975 connp = ipcl_conn_create(IPCL_IPCCONN, KM_SLEEP, ipst->ips_netstack);
5975 5976
5976 5977 /*
5977 5978 * ipcl_conn_create did a netstack_hold. Undo the hold that was
5978 5979 * done by netstack_find_by_cred()
5979 5980 */
5980 5981 netstack_rele(ipst->ips_netstack);
5981 5982
5982 5983 connp->conn_ixa->ixa_flags |= IXAF_MULTICAST_LOOP | IXAF_SET_ULP_CKSUM;
5983 5984 /* conn_allzones can not be set this early, hence no IPCL_ZONEID */
5984 5985 connp->conn_ixa->ixa_zoneid = zoneid;
5985 5986 connp->conn_zoneid = zoneid;
5986 5987
5987 5988 connp->conn_rq = q;
5988 5989 q->q_ptr = WR(q)->q_ptr = connp;
5989 5990
5990 5991 /* Minor tells us which /dev entry was opened */
5991 5992 if (isv6) {
5992 5993 connp->conn_family = AF_INET6;
5993 5994 connp->conn_ipversion = IPV6_VERSION;
5994 5995 connp->conn_ixa->ixa_flags &= ~IXAF_IS_IPV4;
5995 5996 connp->conn_ixa->ixa_src_preferences = IPV6_PREFER_SRC_DEFAULT;
5996 5997 } else {
5997 5998 connp->conn_family = AF_INET;
5998 5999 connp->conn_ipversion = IPV4_VERSION;
5999 6000 connp->conn_ixa->ixa_flags |= IXAF_IS_IPV4;
6000 6001 }
6001 6002
6002 6003 if ((ip_minor_arena_la != NULL) && (flag & SO_SOCKSTR) &&
6003 6004 ((connp->conn_dev = inet_minor_alloc(ip_minor_arena_la)) != 0)) {
6004 6005 connp->conn_minor_arena = ip_minor_arena_la;
6005 6006 } else {
6006 6007 /*
6007 6008 * Either minor numbers in the large arena were exhausted
6008 6009 * or a non socket application is doing the open.
6009 6010 * Try to allocate from the small arena.
6010 6011 */
6011 6012 if ((connp->conn_dev =
6012 6013 inet_minor_alloc(ip_minor_arena_sa)) == 0) {
6013 6014 /* CONN_DEC_REF takes care of netstack_rele() */
6014 6015 q->q_ptr = WR(q)->q_ptr = NULL;
6015 6016 CONN_DEC_REF(connp);
6016 6017 return (EBUSY);
6017 6018 }
6018 6019 connp->conn_minor_arena = ip_minor_arena_sa;
6019 6020 }
6020 6021
6021 6022 maj = getemajor(*devp);
6022 6023 *devp = makedevice(maj, (minor_t)connp->conn_dev);
6023 6024
6024 6025 /*
6025 6026 * connp->conn_cred is crfree()ed in ipcl_conn_destroy()
6026 6027 */
6027 6028 connp->conn_cred = credp;
6028 6029 connp->conn_cpid = curproc->p_pid;
6029 6030 /* Cache things in ixa without an extra refhold */
6030 6031 ASSERT(!(connp->conn_ixa->ixa_free_flags & IXA_FREE_CRED));
6031 6032 connp->conn_ixa->ixa_cred = connp->conn_cred;
6032 6033 connp->conn_ixa->ixa_cpid = connp->conn_cpid;
6033 6034 if (is_system_labeled())
6034 6035 connp->conn_ixa->ixa_tsl = crgetlabel(connp->conn_cred);
6035 6036
6036 6037 /*
6037 6038 * Handle IP_IOC_RTS_REQUEST and other ioctls which use conn_recv
6038 6039 */
6039 6040 connp->conn_recv = ip_conn_input;
6040 6041 connp->conn_recvicmp = ip_conn_input_icmp;
6041 6042
6042 6043 crhold(connp->conn_cred);
6043 6044
6044 6045 /*
6045 6046 * If the caller has the process-wide flag set, then default to MAC
6046 6047 * exempt mode. This allows read-down to unlabeled hosts.
6047 6048 */
6048 6049 if (getpflags(NET_MAC_AWARE, credp) != 0)
6049 6050 connp->conn_mac_mode = CONN_MAC_AWARE;
6050 6051
6051 6052 connp->conn_zone_is_global = (crgetzoneid(credp) == GLOBAL_ZONEID);
6052 6053
6053 6054 connp->conn_rq = q;
6054 6055 connp->conn_wq = WR(q);
6055 6056
6056 6057 /* Non-zero default values */
6057 6058 connp->conn_ixa->ixa_flags |= IXAF_MULTICAST_LOOP;
6058 6059
6059 6060 /*
6060 6061 * Make the conn globally visible to walkers
6061 6062 */
6062 6063 ASSERT(connp->conn_ref == 1);
6063 6064 mutex_enter(&connp->conn_lock);
6064 6065 connp->conn_state_flags &= ~CONN_INCIPIENT;
6065 6066 mutex_exit(&connp->conn_lock);
6066 6067
6067 6068 qprocson(q);
6068 6069
6069 6070 return (0);
6070 6071 }
6071 6072
6072 6073 /*
6073 6074 * Set IPsec policy from an ipsec_req_t. If the req is not "zero" and valid,
6074 6075 * all of them are copied to the conn_t. If the req is "zero", the policy is
6075 6076 * zeroed out. A "zero" policy has zero ipsr_{ah,req,self_encap}_req
6076 6077 * fields.
6077 6078 * We keep only the latest setting of the policy and thus policy setting
6078 6079 * is not incremental/cumulative.
6079 6080 *
6080 6081 * Requests to set policies with multiple alternative actions will
6081 6082 * go through a different API.
6082 6083 */
6083 6084 int
6084 6085 ipsec_set_req(cred_t *cr, conn_t *connp, ipsec_req_t *req)
6085 6086 {
6086 6087 uint_t ah_req = 0;
6087 6088 uint_t esp_req = 0;
6088 6089 uint_t se_req = 0;
6089 6090 ipsec_act_t *actp = NULL;
6090 6091 uint_t nact;
6091 6092 ipsec_policy_head_t *ph;
6092 6093 boolean_t is_pol_reset, is_pol_inserted = B_FALSE;
6093 6094 int error = 0;
6094 6095 netstack_t *ns = connp->conn_netstack;
6095 6096 ip_stack_t *ipst = ns->netstack_ip;
6096 6097 ipsec_stack_t *ipss = ns->netstack_ipsec;
6097 6098
6098 6099 #define REQ_MASK (IPSEC_PREF_REQUIRED|IPSEC_PREF_NEVER)
6099 6100
6100 6101 /*
6101 6102 * The IP_SEC_OPT option does not allow variable length parameters,
6102 6103 * hence a request cannot be NULL.
6103 6104 */
6104 6105 if (req == NULL)
6105 6106 return (EINVAL);
6106 6107
6107 6108 ah_req = req->ipsr_ah_req;
6108 6109 esp_req = req->ipsr_esp_req;
6109 6110 se_req = req->ipsr_self_encap_req;
6110 6111
6111 6112 /* Don't allow setting self-encap without one or more of AH/ESP. */
6112 6113 if (se_req != 0 && esp_req == 0 && ah_req == 0)
6113 6114 return (EINVAL);
6114 6115
6115 6116 /*
6116 6117 * Are we dealing with a request to reset the policy (i.e.
6117 6118 * zero requests).
6118 6119 */
6119 6120 is_pol_reset = ((ah_req & REQ_MASK) == 0 &&
6120 6121 (esp_req & REQ_MASK) == 0 &&
6121 6122 (se_req & REQ_MASK) == 0);
6122 6123
6123 6124 if (!is_pol_reset) {
6124 6125 /*
6125 6126 * If we couldn't load IPsec, fail with "protocol
6126 6127 * not supported".
6127 6128 * IPsec may not have been loaded for a request with zero
6128 6129 * policies, so we don't fail in this case.
6129 6130 */
6130 6131 mutex_enter(&ipss->ipsec_loader_lock);
6131 6132 if (ipss->ipsec_loader_state != IPSEC_LOADER_SUCCEEDED) {
6132 6133 mutex_exit(&ipss->ipsec_loader_lock);
6133 6134 return (EPROTONOSUPPORT);
6134 6135 }
6135 6136 mutex_exit(&ipss->ipsec_loader_lock);
6136 6137
6137 6138 /*
6138 6139 * Test for valid requests. Invalid algorithms
6139 6140 * need to be tested by IPsec code because new
6140 6141 * algorithms can be added dynamically.
6141 6142 */
6142 6143 if ((ah_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0 ||
6143 6144 (esp_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0 ||
6144 6145 (se_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0) {
6145 6146 return (EINVAL);
6146 6147 }
6147 6148
6148 6149 /*
6149 6150 * Only privileged users can issue these
6150 6151 * requests.
6151 6152 */
6152 6153 if (((ah_req & IPSEC_PREF_NEVER) ||
6153 6154 (esp_req & IPSEC_PREF_NEVER) ||
6154 6155 (se_req & IPSEC_PREF_NEVER)) &&
6155 6156 secpolicy_ip_config(cr, B_FALSE) != 0) {
6156 6157 return (EPERM);
6157 6158 }
6158 6159
6159 6160 /*
6160 6161 * The IPSEC_PREF_REQUIRED and IPSEC_PREF_NEVER
6161 6162 * are mutually exclusive.
6162 6163 */
6163 6164 if (((ah_req & REQ_MASK) == REQ_MASK) ||
6164 6165 ((esp_req & REQ_MASK) == REQ_MASK) ||
6165 6166 ((se_req & REQ_MASK) == REQ_MASK)) {
6166 6167 /* Both of them are set */
6167 6168 return (EINVAL);
6168 6169 }
6169 6170 }
6170 6171
6171 6172 ASSERT(MUTEX_HELD(&connp->conn_lock));
6172 6173
6173 6174 /*
6174 6175 * If we have already cached policies in conn_connect(), don't
6175 6176 * let them change now. We cache policies for connections
6176 6177 * whose src,dst [addr, port] is known.
6177 6178 */
6178 6179 if (connp->conn_policy_cached) {
6179 6180 return (EINVAL);
6180 6181 }
6181 6182
6182 6183 /*
6183 6184 * We have a zero policies, reset the connection policy if already
6184 6185 * set. This will cause the connection to inherit the
6185 6186 * global policy, if any.
6186 6187 */
6187 6188 if (is_pol_reset) {
6188 6189 if (connp->conn_policy != NULL) {
6189 6190 IPPH_REFRELE(connp->conn_policy, ipst->ips_netstack);
6190 6191 connp->conn_policy = NULL;
6191 6192 }
6192 6193 connp->conn_in_enforce_policy = B_FALSE;
6193 6194 connp->conn_out_enforce_policy = B_FALSE;
6194 6195 return (0);
6195 6196 }
6196 6197
6197 6198 ph = connp->conn_policy = ipsec_polhead_split(connp->conn_policy,
6198 6199 ipst->ips_netstack);
6199 6200 if (ph == NULL)
6200 6201 goto enomem;
6201 6202
6202 6203 ipsec_actvec_from_req(req, &actp, &nact, ipst->ips_netstack);
6203 6204 if (actp == NULL)
6204 6205 goto enomem;
6205 6206
6206 6207 /*
6207 6208 * Always insert IPv4 policy entries, since they can also apply to
6208 6209 * ipv6 sockets being used in ipv4-compat mode.
6209 6210 */
6210 6211 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V4,
6211 6212 IPSEC_TYPE_INBOUND, ns))
6212 6213 goto enomem;
6213 6214 is_pol_inserted = B_TRUE;
6214 6215 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V4,
6215 6216 IPSEC_TYPE_OUTBOUND, ns))
6216 6217 goto enomem;
6217 6218
6218 6219 /*
6219 6220 * We're looking at a v6 socket, also insert the v6-specific
6220 6221 * entries.
6221 6222 */
6222 6223 if (connp->conn_family == AF_INET6) {
6223 6224 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V6,
6224 6225 IPSEC_TYPE_INBOUND, ns))
6225 6226 goto enomem;
6226 6227 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V6,
6227 6228 IPSEC_TYPE_OUTBOUND, ns))
6228 6229 goto enomem;
6229 6230 }
6230 6231
6231 6232 ipsec_actvec_free(actp, nact);
6232 6233
6233 6234 /*
6234 6235 * If the requests need security, set enforce_policy.
6235 6236 * If the requests are IPSEC_PREF_NEVER, one should
6236 6237 * still set conn_out_enforce_policy so that ip_set_destination
6237 6238 * marks the ip_xmit_attr_t appropriatly. This is needed so that
6238 6239 * for connections that we don't cache policy in at connect time,
6239 6240 * if global policy matches in ip_output_attach_policy, we
6240 6241 * don't wrongly inherit global policy. Similarly, we need
6241 6242 * to set conn_in_enforce_policy also so that we don't verify
6242 6243 * policy wrongly.
6243 6244 */
6244 6245 if ((ah_req & REQ_MASK) != 0 ||
6245 6246 (esp_req & REQ_MASK) != 0 ||
6246 6247 (se_req & REQ_MASK) != 0) {
6247 6248 connp->conn_in_enforce_policy = B_TRUE;
6248 6249 connp->conn_out_enforce_policy = B_TRUE;
6249 6250 }
6250 6251
6251 6252 return (error);
6252 6253 #undef REQ_MASK
6253 6254
6254 6255 /*
6255 6256 * Common memory-allocation-failure exit path.
6256 6257 */
6257 6258 enomem:
6258 6259 if (actp != NULL)
6259 6260 ipsec_actvec_free(actp, nact);
6260 6261 if (is_pol_inserted)
6261 6262 ipsec_polhead_flush(ph, ns);
6262 6263 return (ENOMEM);
6263 6264 }
6264 6265
6265 6266 /*
6266 6267 * Set socket options for joining and leaving multicast groups.
6267 6268 * Common to IPv4 and IPv6; inet6 indicates the type of socket.
6268 6269 * The caller has already check that the option name is consistent with
6269 6270 * the address family of the socket.
6270 6271 */
6271 6272 int
6272 6273 ip_opt_set_multicast_group(conn_t *connp, t_scalar_t name,
6273 6274 uchar_t *invalp, boolean_t inet6, boolean_t checkonly)
6274 6275 {
6275 6276 int *i1 = (int *)invalp;
6276 6277 int error = 0;
6277 6278 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
6278 6279 struct ip_mreq *v4_mreqp;
6279 6280 struct ipv6_mreq *v6_mreqp;
6280 6281 struct group_req *greqp;
6281 6282 ire_t *ire;
6282 6283 boolean_t done = B_FALSE;
6283 6284 ipaddr_t ifaddr;
6284 6285 in6_addr_t v6group;
6285 6286 uint_t ifindex;
6286 6287 boolean_t mcast_opt = B_TRUE;
6287 6288 mcast_record_t fmode;
6288 6289 int (*optfn)(conn_t *, boolean_t, const in6_addr_t *,
6289 6290 ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *);
6290 6291
6291 6292 switch (name) {
6292 6293 case IP_ADD_MEMBERSHIP:
6293 6294 case IPV6_JOIN_GROUP:
6294 6295 mcast_opt = B_FALSE;
6295 6296 /* FALLTHRU */
6296 6297 case MCAST_JOIN_GROUP:
6297 6298 fmode = MODE_IS_EXCLUDE;
6298 6299 optfn = ip_opt_add_group;
6299 6300 break;
6300 6301
6301 6302 case IP_DROP_MEMBERSHIP:
6302 6303 case IPV6_LEAVE_GROUP:
6303 6304 mcast_opt = B_FALSE;
6304 6305 /* FALLTHRU */
6305 6306 case MCAST_LEAVE_GROUP:
6306 6307 fmode = MODE_IS_INCLUDE;
6307 6308 optfn = ip_opt_delete_group;
6308 6309 break;
6309 6310 default:
6310 6311 ASSERT(0);
6311 6312 }
6312 6313
6313 6314 if (mcast_opt) {
6314 6315 struct sockaddr_in *sin;
6315 6316 struct sockaddr_in6 *sin6;
6316 6317
6317 6318 greqp = (struct group_req *)i1;
6318 6319 if (greqp->gr_group.ss_family == AF_INET) {
6319 6320 sin = (struct sockaddr_in *)&(greqp->gr_group);
6320 6321 IN6_INADDR_TO_V4MAPPED(&sin->sin_addr, &v6group);
6321 6322 } else {
6322 6323 if (!inet6)
6323 6324 return (EINVAL); /* Not on INET socket */
6324 6325
6325 6326 sin6 = (struct sockaddr_in6 *)&(greqp->gr_group);
6326 6327 v6group = sin6->sin6_addr;
6327 6328 }
6328 6329 ifaddr = INADDR_ANY;
6329 6330 ifindex = greqp->gr_interface;
6330 6331 } else if (inet6) {
6331 6332 v6_mreqp = (struct ipv6_mreq *)i1;
6332 6333 v6group = v6_mreqp->ipv6mr_multiaddr;
6333 6334 ifaddr = INADDR_ANY;
6334 6335 ifindex = v6_mreqp->ipv6mr_interface;
6335 6336 } else {
6336 6337 v4_mreqp = (struct ip_mreq *)i1;
6337 6338 IN6_INADDR_TO_V4MAPPED(&v4_mreqp->imr_multiaddr, &v6group);
6338 6339 ifaddr = (ipaddr_t)v4_mreqp->imr_interface.s_addr;
6339 6340 ifindex = 0;
6340 6341 }
6341 6342
6342 6343 /*
6343 6344 * In the multirouting case, we need to replicate
6344 6345 * the request on all interfaces that will take part
6345 6346 * in replication. We do so because multirouting is
6346 6347 * reflective, thus we will probably receive multi-
6347 6348 * casts on those interfaces.
6348 6349 * The ip_multirt_apply_membership() succeeds if
6349 6350 * the operation succeeds on at least one interface.
6350 6351 */
6351 6352 if (IN6_IS_ADDR_V4MAPPED(&v6group)) {
6352 6353 ipaddr_t group;
6353 6354
6354 6355 IN6_V4MAPPED_TO_IPADDR(&v6group, group);
6355 6356
6356 6357 ire = ire_ftable_lookup_v4(group, IP_HOST_MASK, 0,
6357 6358 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6358 6359 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6359 6360 } else {
6360 6361 ire = ire_ftable_lookup_v6(&v6group, &ipv6_all_ones, 0,
6361 6362 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6362 6363 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6363 6364 }
6364 6365 if (ire != NULL) {
6365 6366 if (ire->ire_flags & RTF_MULTIRT) {
6366 6367 error = ip_multirt_apply_membership(optfn, ire, connp,
6367 6368 checkonly, &v6group, fmode, &ipv6_all_zeros);
6368 6369 done = B_TRUE;
6369 6370 }
6370 6371 ire_refrele(ire);
6371 6372 }
6372 6373
6373 6374 if (!done) {
6374 6375 error = optfn(connp, checkonly, &v6group, ifaddr, ifindex,
6375 6376 fmode, &ipv6_all_zeros);
6376 6377 }
6377 6378 return (error);
6378 6379 }
6379 6380
6380 6381 /*
6381 6382 * Set socket options for joining and leaving multicast groups
6382 6383 * for specific sources.
6383 6384 * Common to IPv4 and IPv6; inet6 indicates the type of socket.
6384 6385 * The caller has already check that the option name is consistent with
6385 6386 * the address family of the socket.
6386 6387 */
6387 6388 int
6388 6389 ip_opt_set_multicast_sources(conn_t *connp, t_scalar_t name,
6389 6390 uchar_t *invalp, boolean_t inet6, boolean_t checkonly)
6390 6391 {
6391 6392 int *i1 = (int *)invalp;
6392 6393 int error = 0;
6393 6394 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
6394 6395 struct ip_mreq_source *imreqp;
6395 6396 struct group_source_req *gsreqp;
6396 6397 in6_addr_t v6group, v6src;
6397 6398 uint32_t ifindex;
6398 6399 ipaddr_t ifaddr;
6399 6400 boolean_t mcast_opt = B_TRUE;
6400 6401 mcast_record_t fmode;
6401 6402 ire_t *ire;
6402 6403 boolean_t done = B_FALSE;
6403 6404 int (*optfn)(conn_t *, boolean_t, const in6_addr_t *,
6404 6405 ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *);
6405 6406
6406 6407 switch (name) {
6407 6408 case IP_BLOCK_SOURCE:
6408 6409 mcast_opt = B_FALSE;
6409 6410 /* FALLTHRU */
6410 6411 case MCAST_BLOCK_SOURCE:
6411 6412 fmode = MODE_IS_EXCLUDE;
6412 6413 optfn = ip_opt_add_group;
6413 6414 break;
6414 6415
6415 6416 case IP_UNBLOCK_SOURCE:
6416 6417 mcast_opt = B_FALSE;
6417 6418 /* FALLTHRU */
6418 6419 case MCAST_UNBLOCK_SOURCE:
6419 6420 fmode = MODE_IS_EXCLUDE;
6420 6421 optfn = ip_opt_delete_group;
6421 6422 break;
6422 6423
6423 6424 case IP_ADD_SOURCE_MEMBERSHIP:
6424 6425 mcast_opt = B_FALSE;
6425 6426 /* FALLTHRU */
6426 6427 case MCAST_JOIN_SOURCE_GROUP:
6427 6428 fmode = MODE_IS_INCLUDE;
6428 6429 optfn = ip_opt_add_group;
6429 6430 break;
6430 6431
6431 6432 case IP_DROP_SOURCE_MEMBERSHIP:
6432 6433 mcast_opt = B_FALSE;
6433 6434 /* FALLTHRU */
6434 6435 case MCAST_LEAVE_SOURCE_GROUP:
6435 6436 fmode = MODE_IS_INCLUDE;
6436 6437 optfn = ip_opt_delete_group;
6437 6438 break;
6438 6439 default:
6439 6440 ASSERT(0);
6440 6441 }
6441 6442
6442 6443 if (mcast_opt) {
6443 6444 gsreqp = (struct group_source_req *)i1;
6444 6445 ifindex = gsreqp->gsr_interface;
6445 6446 if (gsreqp->gsr_group.ss_family == AF_INET) {
6446 6447 struct sockaddr_in *s;
6447 6448 s = (struct sockaddr_in *)&gsreqp->gsr_group;
6448 6449 IN6_INADDR_TO_V4MAPPED(&s->sin_addr, &v6group);
6449 6450 s = (struct sockaddr_in *)&gsreqp->gsr_source;
6450 6451 IN6_INADDR_TO_V4MAPPED(&s->sin_addr, &v6src);
6451 6452 } else {
6452 6453 struct sockaddr_in6 *s6;
6453 6454
6454 6455 if (!inet6)
6455 6456 return (EINVAL); /* Not on INET socket */
6456 6457
6457 6458 s6 = (struct sockaddr_in6 *)&gsreqp->gsr_group;
6458 6459 v6group = s6->sin6_addr;
6459 6460 s6 = (struct sockaddr_in6 *)&gsreqp->gsr_source;
6460 6461 v6src = s6->sin6_addr;
6461 6462 }
6462 6463 ifaddr = INADDR_ANY;
6463 6464 } else {
6464 6465 imreqp = (struct ip_mreq_source *)i1;
6465 6466 IN6_INADDR_TO_V4MAPPED(&imreqp->imr_multiaddr, &v6group);
6466 6467 IN6_INADDR_TO_V4MAPPED(&imreqp->imr_sourceaddr, &v6src);
6467 6468 ifaddr = (ipaddr_t)imreqp->imr_interface.s_addr;
6468 6469 ifindex = 0;
6469 6470 }
6470 6471
6471 6472 /*
6472 6473 * Handle src being mapped INADDR_ANY by changing it to unspecified.
6473 6474 */
6474 6475 if (IN6_IS_ADDR_V4MAPPED_ANY(&v6src))
6475 6476 v6src = ipv6_all_zeros;
6476 6477
6477 6478 /*
6478 6479 * In the multirouting case, we need to replicate
6479 6480 * the request as noted in the mcast cases above.
6480 6481 */
6481 6482 if (IN6_IS_ADDR_V4MAPPED(&v6group)) {
6482 6483 ipaddr_t group;
6483 6484
6484 6485 IN6_V4MAPPED_TO_IPADDR(&v6group, group);
6485 6486
6486 6487 ire = ire_ftable_lookup_v4(group, IP_HOST_MASK, 0,
6487 6488 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6488 6489 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6489 6490 } else {
6490 6491 ire = ire_ftable_lookup_v6(&v6group, &ipv6_all_ones, 0,
6491 6492 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6492 6493 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6493 6494 }
6494 6495 if (ire != NULL) {
6495 6496 if (ire->ire_flags & RTF_MULTIRT) {
6496 6497 error = ip_multirt_apply_membership(optfn, ire, connp,
6497 6498 checkonly, &v6group, fmode, &v6src);
6498 6499 done = B_TRUE;
6499 6500 }
6500 6501 ire_refrele(ire);
6501 6502 }
6502 6503 if (!done) {
6503 6504 error = optfn(connp, checkonly, &v6group, ifaddr, ifindex,
6504 6505 fmode, &v6src);
6505 6506 }
6506 6507 return (error);
6507 6508 }
6508 6509
6509 6510 /*
6510 6511 * Given a destination address and a pointer to where to put the information
6511 6512 * this routine fills in the mtuinfo.
6512 6513 * The socket must be connected.
6513 6514 * For sctp conn_faddr is the primary address.
6514 6515 */
6515 6516 int
6516 6517 ip_fill_mtuinfo(conn_t *connp, ip_xmit_attr_t *ixa, struct ip6_mtuinfo *mtuinfo)
6517 6518 {
6518 6519 uint32_t pmtu = IP_MAXPACKET;
6519 6520 uint_t scopeid;
6520 6521
6521 6522 if (IN6_IS_ADDR_UNSPECIFIED(&connp->conn_faddr_v6))
6522 6523 return (-1);
6523 6524
6524 6525 /* In case we never sent or called ip_set_destination_v4/v6 */
6525 6526 if (ixa->ixa_ire != NULL)
6526 6527 pmtu = ip_get_pmtu(ixa);
6527 6528
6528 6529 if (ixa->ixa_flags & IXAF_SCOPEID_SET)
6529 6530 scopeid = ixa->ixa_scopeid;
6530 6531 else
6531 6532 scopeid = 0;
6532 6533
6533 6534 bzero(mtuinfo, sizeof (*mtuinfo));
6534 6535 mtuinfo->ip6m_addr.sin6_family = AF_INET6;
6535 6536 mtuinfo->ip6m_addr.sin6_port = connp->conn_fport;
6536 6537 mtuinfo->ip6m_addr.sin6_addr = connp->conn_faddr_v6;
6537 6538 mtuinfo->ip6m_addr.sin6_scope_id = scopeid;
6538 6539 mtuinfo->ip6m_mtu = pmtu;
6539 6540
6540 6541 return (sizeof (struct ip6_mtuinfo));
6541 6542 }
6542 6543
6543 6544 /*
6544 6545 * When the src multihoming is changed from weak to [strong, preferred]
6545 6546 * ip_ire_rebind_walker is called to walk the list of all ire_t entries
6546 6547 * and identify routes that were created by user-applications in the
6547 6548 * unbound state (i.e., without RTA_IFP), and for which an ire_ill is not
6548 6549 * currently defined. These routes are then 'rebound', i.e., their ire_ill
6549 6550 * is selected by finding an interface route for the gateway.
6550 6551 */
6551 6552 /* ARGSUSED */
6552 6553 void
6553 6554 ip_ire_rebind_walker(ire_t *ire, void *notused)
6554 6555 {
6555 6556 if (!ire->ire_unbound || ire->ire_ill != NULL)
6556 6557 return;
6557 6558 ire_rebind(ire);
6558 6559 ire_delete(ire);
6559 6560 }
6560 6561
6561 6562 /*
6562 6563 * When the src multihoming is changed from [strong, preferred] to weak,
6563 6564 * ip_ire_unbind_walker is called to walk the list of all ire_t entries, and
6564 6565 * set any entries that were created by user-applications in the unbound state
6565 6566 * (i.e., without RTA_IFP) back to having a NULL ire_ill.
6566 6567 */
6567 6568 /* ARGSUSED */
6568 6569 void
6569 6570 ip_ire_unbind_walker(ire_t *ire, void *notused)
6570 6571 {
6571 6572 ire_t *new_ire;
6572 6573
6573 6574 if (!ire->ire_unbound || ire->ire_ill == NULL)
6574 6575 return;
6575 6576 if (ire->ire_ipversion == IPV6_VERSION) {
6576 6577 new_ire = ire_create_v6(&ire->ire_addr_v6, &ire->ire_mask_v6,
6577 6578 &ire->ire_gateway_addr_v6, ire->ire_type, NULL,
6578 6579 ire->ire_zoneid, ire->ire_flags, NULL, ire->ire_ipst);
6579 6580 } else {
6580 6581 new_ire = ire_create((uchar_t *)&ire->ire_addr,
6581 6582 (uchar_t *)&ire->ire_mask,
6582 6583 (uchar_t *)&ire->ire_gateway_addr, ire->ire_type, NULL,
6583 6584 ire->ire_zoneid, ire->ire_flags, NULL, ire->ire_ipst);
6584 6585 }
6585 6586 if (new_ire == NULL)
6586 6587 return;
6587 6588 new_ire->ire_unbound = B_TRUE;
6588 6589 /*
6589 6590 * The bound ire must first be deleted so that we don't return
6590 6591 * the existing one on the attempt to add the unbound new_ire.
6591 6592 */
6592 6593 ire_delete(ire);
6593 6594 new_ire = ire_add(new_ire);
6594 6595 if (new_ire != NULL)
6595 6596 ire_refrele(new_ire);
6596 6597 }
6597 6598
6598 6599 /*
6599 6600 * When the settings of ip*_strict_src_multihoming tunables are changed,
6600 6601 * all cached routes need to be recomputed. This recomputation needs to be
6601 6602 * done when going from weaker to stronger modes so that the cached ire
6602 6603 * for the connection does not violate the current ip*_strict_src_multihoming
6603 6604 * setting. It also needs to be done when going from stronger to weaker modes,
6604 6605 * so that we fall back to matching on the longest-matching-route (as opposed
6605 6606 * to a shorter match that may have been selected in the strong mode
6606 6607 * to satisfy src_multihoming settings).
6607 6608 *
6608 6609 * The cached ixa_ire entires for all conn_t entries are marked as
6609 6610 * "verify" so that they will be recomputed for the next packet.
6610 6611 */
6611 6612 void
6612 6613 conn_ire_revalidate(conn_t *connp, void *arg)
6613 6614 {
6614 6615 boolean_t isv6 = (boolean_t)arg;
6615 6616
6616 6617 if ((isv6 && connp->conn_ipversion != IPV6_VERSION) ||
6617 6618 (!isv6 && connp->conn_ipversion != IPV4_VERSION))
6618 6619 return;
6619 6620 connp->conn_ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
6620 6621 }
6621 6622
6622 6623 /*
6623 6624 * Handles both IPv4 and IPv6 reassembly - doing the out-of-order cases,
6624 6625 * When an ipf is passed here for the first time, if
6625 6626 * we already have in-order fragments on the queue, we convert from the fast-
6626 6627 * path reassembly scheme to the hard-case scheme. From then on, additional
6627 6628 * fragments are reassembled here. We keep track of the start and end offsets
6628 6629 * of each piece, and the number of holes in the chain. When the hole count
6629 6630 * goes to zero, we are done!
6630 6631 *
6631 6632 * The ipf_count will be updated to account for any mblk(s) added (pointed to
6632 6633 * by mp) or subtracted (freeb()ed dups), upon return the caller must update
6633 6634 * ipfb_count and ill_frag_count by the difference of ipf_count before and
6634 6635 * after the call to ip_reassemble().
6635 6636 */
6636 6637 int
6637 6638 ip_reassemble(mblk_t *mp, ipf_t *ipf, uint_t start, boolean_t more, ill_t *ill,
6638 6639 size_t msg_len)
6639 6640 {
6640 6641 uint_t end;
6641 6642 mblk_t *next_mp;
6642 6643 mblk_t *mp1;
6643 6644 uint_t offset;
6644 6645 boolean_t incr_dups = B_TRUE;
6645 6646 boolean_t offset_zero_seen = B_FALSE;
6646 6647 boolean_t pkt_boundary_checked = B_FALSE;
6647 6648
6648 6649 /* If start == 0 then ipf_nf_hdr_len has to be set. */
6649 6650 ASSERT(start != 0 || ipf->ipf_nf_hdr_len != 0);
6650 6651
6651 6652 /* Add in byte count */
6652 6653 ipf->ipf_count += msg_len;
6653 6654 if (ipf->ipf_end) {
6654 6655 /*
6655 6656 * We were part way through in-order reassembly, but now there
6656 6657 * is a hole. We walk through messages already queued, and
6657 6658 * mark them for hard case reassembly. We know that up till
6658 6659 * now they were in order starting from offset zero.
6659 6660 */
6660 6661 offset = 0;
6661 6662 for (mp1 = ipf->ipf_mp->b_cont; mp1; mp1 = mp1->b_cont) {
6662 6663 IP_REASS_SET_START(mp1, offset);
6663 6664 if (offset == 0) {
6664 6665 ASSERT(ipf->ipf_nf_hdr_len != 0);
6665 6666 offset = -ipf->ipf_nf_hdr_len;
6666 6667 }
6667 6668 offset += mp1->b_wptr - mp1->b_rptr;
6668 6669 IP_REASS_SET_END(mp1, offset);
6669 6670 }
6670 6671 /* One hole at the end. */
6671 6672 ipf->ipf_hole_cnt = 1;
6672 6673 /* Brand it as a hard case, forever. */
6673 6674 ipf->ipf_end = 0;
6674 6675 }
6675 6676 /* Walk through all the new pieces. */
6676 6677 do {
6677 6678 end = start + (mp->b_wptr - mp->b_rptr);
6678 6679 /*
6679 6680 * If start is 0, decrease 'end' only for the first mblk of
6680 6681 * the fragment. Otherwise 'end' can get wrong value in the
6681 6682 * second pass of the loop if first mblk is exactly the
6682 6683 * size of ipf_nf_hdr_len.
6683 6684 */
6684 6685 if (start == 0 && !offset_zero_seen) {
6685 6686 /* First segment */
6686 6687 ASSERT(ipf->ipf_nf_hdr_len != 0);
6687 6688 end -= ipf->ipf_nf_hdr_len;
6688 6689 offset_zero_seen = B_TRUE;
6689 6690 }
6690 6691 next_mp = mp->b_cont;
6691 6692 /*
6692 6693 * We are checking to see if there is any interesing data
6693 6694 * to process. If there isn't and the mblk isn't the
6694 6695 * one which carries the unfragmentable header then we
6695 6696 * drop it. It's possible to have just the unfragmentable
6696 6697 * header come through without any data. That needs to be
6697 6698 * saved.
6698 6699 *
6699 6700 * If the assert at the top of this function holds then the
6700 6701 * term "ipf->ipf_nf_hdr_len != 0" isn't needed. This code
6701 6702 * is infrequently traveled enough that the test is left in
6702 6703 * to protect against future code changes which break that
6703 6704 * invariant.
6704 6705 */
6705 6706 if (start == end && start != 0 && ipf->ipf_nf_hdr_len != 0) {
6706 6707 /* Empty. Blast it. */
6707 6708 IP_REASS_SET_START(mp, 0);
6708 6709 IP_REASS_SET_END(mp, 0);
6709 6710 /*
6710 6711 * If the ipf points to the mblk we are about to free,
6711 6712 * update ipf to point to the next mblk (or NULL
6712 6713 * if none).
6713 6714 */
6714 6715 if (ipf->ipf_mp->b_cont == mp)
6715 6716 ipf->ipf_mp->b_cont = next_mp;
6716 6717 freeb(mp);
6717 6718 continue;
6718 6719 }
6719 6720 mp->b_cont = NULL;
6720 6721 IP_REASS_SET_START(mp, start);
6721 6722 IP_REASS_SET_END(mp, end);
6722 6723 if (!ipf->ipf_tail_mp) {
6723 6724 ipf->ipf_tail_mp = mp;
6724 6725 ipf->ipf_mp->b_cont = mp;
6725 6726 if (start == 0 || !more) {
6726 6727 ipf->ipf_hole_cnt = 1;
6727 6728 /*
6728 6729 * if the first fragment comes in more than one
6729 6730 * mblk, this loop will be executed for each
6730 6731 * mblk. Need to adjust hole count so exiting
6731 6732 * this routine will leave hole count at 1.
6732 6733 */
6733 6734 if (next_mp)
6734 6735 ipf->ipf_hole_cnt++;
6735 6736 } else
6736 6737 ipf->ipf_hole_cnt = 2;
6737 6738 continue;
6738 6739 } else if (ipf->ipf_last_frag_seen && !more &&
6739 6740 !pkt_boundary_checked) {
6740 6741 /*
6741 6742 * We check datagram boundary only if this fragment
6742 6743 * claims to be the last fragment and we have seen a
6743 6744 * last fragment in the past too. We do this only
6744 6745 * once for a given fragment.
6745 6746 *
6746 6747 * start cannot be 0 here as fragments with start=0
6747 6748 * and MF=0 gets handled as a complete packet. These
6748 6749 * fragments should not reach here.
6749 6750 */
6750 6751
6751 6752 if (start + msgdsize(mp) !=
6752 6753 IP_REASS_END(ipf->ipf_tail_mp)) {
6753 6754 /*
6754 6755 * We have two fragments both of which claim
6755 6756 * to be the last fragment but gives conflicting
6756 6757 * information about the whole datagram size.
6757 6758 * Something fishy is going on. Drop the
6758 6759 * fragment and free up the reassembly list.
6759 6760 */
6760 6761 return (IP_REASS_FAILED);
6761 6762 }
6762 6763
6763 6764 /*
6764 6765 * We shouldn't come to this code block again for this
6765 6766 * particular fragment.
6766 6767 */
6767 6768 pkt_boundary_checked = B_TRUE;
6768 6769 }
6769 6770
6770 6771 /* New stuff at or beyond tail? */
6771 6772 offset = IP_REASS_END(ipf->ipf_tail_mp);
6772 6773 if (start >= offset) {
6773 6774 if (ipf->ipf_last_frag_seen) {
6774 6775 /* current fragment is beyond last fragment */
6775 6776 return (IP_REASS_FAILED);
6776 6777 }
6777 6778 /* Link it on end. */
6778 6779 ipf->ipf_tail_mp->b_cont = mp;
6779 6780 ipf->ipf_tail_mp = mp;
6780 6781 if (more) {
6781 6782 if (start != offset)
6782 6783 ipf->ipf_hole_cnt++;
6783 6784 } else if (start == offset && next_mp == NULL)
6784 6785 ipf->ipf_hole_cnt--;
6785 6786 continue;
6786 6787 }
6787 6788 mp1 = ipf->ipf_mp->b_cont;
6788 6789 offset = IP_REASS_START(mp1);
6789 6790 /* New stuff at the front? */
6790 6791 if (start < offset) {
6791 6792 if (start == 0) {
6792 6793 if (end >= offset) {
6793 6794 /* Nailed the hole at the begining. */
6794 6795 ipf->ipf_hole_cnt--;
6795 6796 }
6796 6797 } else if (end < offset) {
6797 6798 /*
6798 6799 * A hole, stuff, and a hole where there used
6799 6800 * to be just a hole.
6800 6801 */
6801 6802 ipf->ipf_hole_cnt++;
6802 6803 }
6803 6804 mp->b_cont = mp1;
6804 6805 /* Check for overlap. */
6805 6806 while (end > offset) {
6806 6807 if (end < IP_REASS_END(mp1)) {
6807 6808 mp->b_wptr -= end - offset;
6808 6809 IP_REASS_SET_END(mp, offset);
6809 6810 BUMP_MIB(ill->ill_ip_mib,
6810 6811 ipIfStatsReasmPartDups);
6811 6812 break;
6812 6813 }
6813 6814 /* Did we cover another hole? */
6814 6815 if ((mp1->b_cont &&
6815 6816 IP_REASS_END(mp1) !=
6816 6817 IP_REASS_START(mp1->b_cont) &&
6817 6818 end >= IP_REASS_START(mp1->b_cont)) ||
6818 6819 (!ipf->ipf_last_frag_seen && !more)) {
6819 6820 ipf->ipf_hole_cnt--;
6820 6821 }
6821 6822 /* Clip out mp1. */
6822 6823 if ((mp->b_cont = mp1->b_cont) == NULL) {
6823 6824 /*
6824 6825 * After clipping out mp1, this guy
6825 6826 * is now hanging off the end.
6826 6827 */
6827 6828 ipf->ipf_tail_mp = mp;
6828 6829 }
6829 6830 IP_REASS_SET_START(mp1, 0);
6830 6831 IP_REASS_SET_END(mp1, 0);
6831 6832 /* Subtract byte count */
6832 6833 ipf->ipf_count -= mp1->b_datap->db_lim -
6833 6834 mp1->b_datap->db_base;
6834 6835 freeb(mp1);
6835 6836 BUMP_MIB(ill->ill_ip_mib,
6836 6837 ipIfStatsReasmPartDups);
6837 6838 mp1 = mp->b_cont;
6838 6839 if (!mp1)
6839 6840 break;
6840 6841 offset = IP_REASS_START(mp1);
6841 6842 }
6842 6843 ipf->ipf_mp->b_cont = mp;
6843 6844 continue;
6844 6845 }
6845 6846 /*
6846 6847 * The new piece starts somewhere between the start of the head
6847 6848 * and before the end of the tail.
6848 6849 */
6849 6850 for (; mp1; mp1 = mp1->b_cont) {
6850 6851 offset = IP_REASS_END(mp1);
6851 6852 if (start < offset) {
6852 6853 if (end <= offset) {
6853 6854 /* Nothing new. */
6854 6855 IP_REASS_SET_START(mp, 0);
6855 6856 IP_REASS_SET_END(mp, 0);
6856 6857 /* Subtract byte count */
6857 6858 ipf->ipf_count -= mp->b_datap->db_lim -
6858 6859 mp->b_datap->db_base;
6859 6860 if (incr_dups) {
6860 6861 ipf->ipf_num_dups++;
6861 6862 incr_dups = B_FALSE;
6862 6863 }
6863 6864 freeb(mp);
6864 6865 BUMP_MIB(ill->ill_ip_mib,
6865 6866 ipIfStatsReasmDuplicates);
6866 6867 break;
6867 6868 }
6868 6869 /*
6869 6870 * Trim redundant stuff off beginning of new
6870 6871 * piece.
6871 6872 */
6872 6873 IP_REASS_SET_START(mp, offset);
6873 6874 mp->b_rptr += offset - start;
6874 6875 BUMP_MIB(ill->ill_ip_mib,
6875 6876 ipIfStatsReasmPartDups);
6876 6877 start = offset;
6877 6878 if (!mp1->b_cont) {
6878 6879 /*
6879 6880 * After trimming, this guy is now
6880 6881 * hanging off the end.
6881 6882 */
6882 6883 mp1->b_cont = mp;
6883 6884 ipf->ipf_tail_mp = mp;
6884 6885 if (!more) {
6885 6886 ipf->ipf_hole_cnt--;
6886 6887 }
6887 6888 break;
6888 6889 }
6889 6890 }
6890 6891 if (start >= IP_REASS_START(mp1->b_cont))
6891 6892 continue;
6892 6893 /* Fill a hole */
6893 6894 if (start > offset)
6894 6895 ipf->ipf_hole_cnt++;
6895 6896 mp->b_cont = mp1->b_cont;
6896 6897 mp1->b_cont = mp;
6897 6898 mp1 = mp->b_cont;
6898 6899 offset = IP_REASS_START(mp1);
6899 6900 if (end >= offset) {
6900 6901 ipf->ipf_hole_cnt--;
6901 6902 /* Check for overlap. */
6902 6903 while (end > offset) {
6903 6904 if (end < IP_REASS_END(mp1)) {
6904 6905 mp->b_wptr -= end - offset;
6905 6906 IP_REASS_SET_END(mp, offset);
6906 6907 /*
6907 6908 * TODO we might bump
6908 6909 * this up twice if there is
6909 6910 * overlap at both ends.
6910 6911 */
6911 6912 BUMP_MIB(ill->ill_ip_mib,
6912 6913 ipIfStatsReasmPartDups);
6913 6914 break;
6914 6915 }
6915 6916 /* Did we cover another hole? */
6916 6917 if ((mp1->b_cont &&
6917 6918 IP_REASS_END(mp1)
6918 6919 != IP_REASS_START(mp1->b_cont) &&
6919 6920 end >=
6920 6921 IP_REASS_START(mp1->b_cont)) ||
6921 6922 (!ipf->ipf_last_frag_seen &&
6922 6923 !more)) {
6923 6924 ipf->ipf_hole_cnt--;
6924 6925 }
6925 6926 /* Clip out mp1. */
6926 6927 if ((mp->b_cont = mp1->b_cont) ==
6927 6928 NULL) {
6928 6929 /*
6929 6930 * After clipping out mp1,
6930 6931 * this guy is now hanging
6931 6932 * off the end.
6932 6933 */
6933 6934 ipf->ipf_tail_mp = mp;
6934 6935 }
6935 6936 IP_REASS_SET_START(mp1, 0);
6936 6937 IP_REASS_SET_END(mp1, 0);
6937 6938 /* Subtract byte count */
6938 6939 ipf->ipf_count -=
6939 6940 mp1->b_datap->db_lim -
6940 6941 mp1->b_datap->db_base;
6941 6942 freeb(mp1);
6942 6943 BUMP_MIB(ill->ill_ip_mib,
6943 6944 ipIfStatsReasmPartDups);
6944 6945 mp1 = mp->b_cont;
6945 6946 if (!mp1)
6946 6947 break;
6947 6948 offset = IP_REASS_START(mp1);
6948 6949 }
6949 6950 }
6950 6951 break;
6951 6952 }
6952 6953 } while (start = end, mp = next_mp);
6953 6954
6954 6955 /* Fragment just processed could be the last one. Remember this fact */
6955 6956 if (!more)
6956 6957 ipf->ipf_last_frag_seen = B_TRUE;
6957 6958
6958 6959 /* Still got holes? */
6959 6960 if (ipf->ipf_hole_cnt)
6960 6961 return (IP_REASS_PARTIAL);
6961 6962 /* Clean up overloaded fields to avoid upstream disasters. */
6962 6963 for (mp1 = ipf->ipf_mp->b_cont; mp1; mp1 = mp1->b_cont) {
6963 6964 IP_REASS_SET_START(mp1, 0);
6964 6965 IP_REASS_SET_END(mp1, 0);
6965 6966 }
6966 6967 return (IP_REASS_COMPLETE);
6967 6968 }
6968 6969
6969 6970 /*
6970 6971 * Fragmentation reassembly. Each ILL has a hash table for
6971 6972 * queuing packets undergoing reassembly for all IPIFs
6972 6973 * associated with the ILL. The hash is based on the packet
6973 6974 * IP ident field. The ILL frag hash table was allocated
6974 6975 * as a timer block at the time the ILL was created. Whenever
6975 6976 * there is anything on the reassembly queue, the timer will
6976 6977 * be running. Returns the reassembled packet if reassembly completes.
6977 6978 */
6978 6979 mblk_t *
6979 6980 ip_input_fragment(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
6980 6981 {
6981 6982 uint32_t frag_offset_flags;
6982 6983 mblk_t *t_mp;
6983 6984 ipaddr_t dst;
6984 6985 uint8_t proto = ipha->ipha_protocol;
6985 6986 uint32_t sum_val;
6986 6987 uint16_t sum_flags;
6987 6988 ipf_t *ipf;
6988 6989 ipf_t **ipfp;
6989 6990 ipfb_t *ipfb;
6990 6991 uint16_t ident;
6991 6992 uint32_t offset;
6992 6993 ipaddr_t src;
6993 6994 uint_t hdr_length;
6994 6995 uint32_t end;
6995 6996 mblk_t *mp1;
6996 6997 mblk_t *tail_mp;
6997 6998 size_t count;
6998 6999 size_t msg_len;
6999 7000 uint8_t ecn_info = 0;
7000 7001 uint32_t packet_size;
7001 7002 boolean_t pruned = B_FALSE;
7002 7003 ill_t *ill = ira->ira_ill;
7003 7004 ip_stack_t *ipst = ill->ill_ipst;
7004 7005
7005 7006 /*
7006 7007 * Drop the fragmented as early as possible, if
7007 7008 * we don't have resource(s) to re-assemble.
7008 7009 */
7009 7010 if (ipst->ips_ip_reass_queue_bytes == 0) {
7010 7011 freemsg(mp);
7011 7012 return (NULL);
7012 7013 }
7013 7014
7014 7015 /* Check for fragmentation offset; return if there's none */
7015 7016 if ((frag_offset_flags = ntohs(ipha->ipha_fragment_offset_and_flags) &
7016 7017 (IPH_MF | IPH_OFFSET)) == 0)
7017 7018 return (mp);
7018 7019
7019 7020 /*
7020 7021 * We utilize hardware computed checksum info only for UDP since
7021 7022 * IP fragmentation is a normal occurrence for the protocol. In
7022 7023 * addition, checksum offload support for IP fragments carrying
7023 7024 * UDP payload is commonly implemented across network adapters.
7024 7025 */
7025 7026 ASSERT(ira->ira_rill != NULL);
7026 7027 if (proto == IPPROTO_UDP && dohwcksum &&
7027 7028 ILL_HCKSUM_CAPABLE(ira->ira_rill) &&
7028 7029 (DB_CKSUMFLAGS(mp) & (HCK_FULLCKSUM | HCK_PARTIALCKSUM))) {
7029 7030 mblk_t *mp1 = mp->b_cont;
7030 7031 int32_t len;
7031 7032
7032 7033 /* Record checksum information from the packet */
7033 7034 sum_val = (uint32_t)DB_CKSUM16(mp);
7034 7035 sum_flags = DB_CKSUMFLAGS(mp);
7035 7036
7036 7037 /* IP payload offset from beginning of mblk */
7037 7038 offset = ((uchar_t *)ipha + IPH_HDR_LENGTH(ipha)) - mp->b_rptr;
7038 7039
7039 7040 if ((sum_flags & HCK_PARTIALCKSUM) &&
7040 7041 (mp1 == NULL || mp1->b_cont == NULL) &&
7041 7042 offset >= DB_CKSUMSTART(mp) &&
7042 7043 ((len = offset - DB_CKSUMSTART(mp)) & 1) == 0) {
7043 7044 uint32_t adj;
7044 7045 /*
7045 7046 * Partial checksum has been calculated by hardware
7046 7047 * and attached to the packet; in addition, any
7047 7048 * prepended extraneous data is even byte aligned.
7048 7049 * If any such data exists, we adjust the checksum;
7049 7050 * this would also handle any postpended data.
7050 7051 */
7051 7052 IP_ADJCKSUM_PARTIAL(mp->b_rptr + DB_CKSUMSTART(mp),
7052 7053 mp, mp1, len, adj);
7053 7054
7054 7055 /* One's complement subtract extraneous checksum */
7055 7056 if (adj >= sum_val)
7056 7057 sum_val = ~(adj - sum_val) & 0xFFFF;
7057 7058 else
7058 7059 sum_val -= adj;
7059 7060 }
7060 7061 } else {
7061 7062 sum_val = 0;
7062 7063 sum_flags = 0;
7063 7064 }
7064 7065
7065 7066 /* Clear hardware checksumming flag */
7066 7067 DB_CKSUMFLAGS(mp) = 0;
7067 7068
7068 7069 ident = ipha->ipha_ident;
7069 7070 offset = (frag_offset_flags << 3) & 0xFFFF;
7070 7071 src = ipha->ipha_src;
7071 7072 dst = ipha->ipha_dst;
7072 7073 hdr_length = IPH_HDR_LENGTH(ipha);
7073 7074 end = ntohs(ipha->ipha_length) - hdr_length;
7074 7075
7075 7076 /* If end == 0 then we have a packet with no data, so just free it */
7076 7077 if (end == 0) {
7077 7078 freemsg(mp);
7078 7079 return (NULL);
7079 7080 }
7080 7081
7081 7082 /* Record the ECN field info. */
7082 7083 ecn_info = (ipha->ipha_type_of_service & 0x3);
7083 7084 if (offset != 0) {
7084 7085 /*
7085 7086 * If this isn't the first piece, strip the header, and
7086 7087 * add the offset to the end value.
7087 7088 */
7088 7089 mp->b_rptr += hdr_length;
7089 7090 end += offset;
7090 7091 }
7091 7092
7092 7093 /* Handle vnic loopback of fragments */
7093 7094 if (mp->b_datap->db_ref > 2)
7094 7095 msg_len = 0;
7095 7096 else
7096 7097 msg_len = MBLKSIZE(mp);
7097 7098
7098 7099 tail_mp = mp;
7099 7100 while (tail_mp->b_cont != NULL) {
7100 7101 tail_mp = tail_mp->b_cont;
7101 7102 if (tail_mp->b_datap->db_ref <= 2)
7102 7103 msg_len += MBLKSIZE(tail_mp);
7103 7104 }
7104 7105
7105 7106 /* If the reassembly list for this ILL will get too big, prune it */
7106 7107 if ((msg_len + sizeof (*ipf) + ill->ill_frag_count) >=
7107 7108 ipst->ips_ip_reass_queue_bytes) {
7108 7109 DTRACE_PROBE3(ip_reass_queue_bytes, uint_t, msg_len,
7109 7110 uint_t, ill->ill_frag_count,
7110 7111 uint_t, ipst->ips_ip_reass_queue_bytes);
7111 7112 ill_frag_prune(ill,
7112 7113 (ipst->ips_ip_reass_queue_bytes < msg_len) ? 0 :
7113 7114 (ipst->ips_ip_reass_queue_bytes - msg_len));
7114 7115 pruned = B_TRUE;
7115 7116 }
7116 7117
7117 7118 ipfb = &ill->ill_frag_hash_tbl[ILL_FRAG_HASH(src, ident)];
7118 7119 mutex_enter(&ipfb->ipfb_lock);
7119 7120
7120 7121 ipfp = &ipfb->ipfb_ipf;
7121 7122 /* Try to find an existing fragment queue for this packet. */
7122 7123 for (;;) {
7123 7124 ipf = ipfp[0];
7124 7125 if (ipf != NULL) {
7125 7126 /*
7126 7127 * It has to match on ident and src/dst address.
7127 7128 */
7128 7129 if (ipf->ipf_ident == ident &&
7129 7130 ipf->ipf_src == src &&
7130 7131 ipf->ipf_dst == dst &&
7131 7132 ipf->ipf_protocol == proto) {
7132 7133 /*
7133 7134 * If we have received too many
7134 7135 * duplicate fragments for this packet
7135 7136 * free it.
7136 7137 */
7137 7138 if (ipf->ipf_num_dups > ip_max_frag_dups) {
7138 7139 ill_frag_free_pkts(ill, ipfb, ipf, 1);
7139 7140 freemsg(mp);
7140 7141 mutex_exit(&ipfb->ipfb_lock);
7141 7142 return (NULL);
7142 7143 }
7143 7144 /* Found it. */
7144 7145 break;
7145 7146 }
7146 7147 ipfp = &ipf->ipf_hash_next;
7147 7148 continue;
7148 7149 }
7149 7150
7150 7151 /*
7151 7152 * If we pruned the list, do we want to store this new
7152 7153 * fragment?. We apply an optimization here based on the
7153 7154 * fact that most fragments will be received in order.
7154 7155 * So if the offset of this incoming fragment is zero,
7155 7156 * it is the first fragment of a new packet. We will
7156 7157 * keep it. Otherwise drop the fragment, as we have
7157 7158 * probably pruned the packet already (since the
7158 7159 * packet cannot be found).
7159 7160 */
7160 7161 if (pruned && offset != 0) {
7161 7162 mutex_exit(&ipfb->ipfb_lock);
7162 7163 freemsg(mp);
7163 7164 return (NULL);
7164 7165 }
7165 7166
7166 7167 if (ipfb->ipfb_frag_pkts >= MAX_FRAG_PKTS(ipst)) {
7167 7168 /*
7168 7169 * Too many fragmented packets in this hash
7169 7170 * bucket. Free the oldest.
7170 7171 */
7171 7172 ill_frag_free_pkts(ill, ipfb, ipfb->ipfb_ipf, 1);
7172 7173 }
7173 7174
7174 7175 /* New guy. Allocate a frag message. */
7175 7176 mp1 = allocb(sizeof (*ipf), BPRI_MED);
7176 7177 if (mp1 == NULL) {
7177 7178 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7178 7179 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7179 7180 freemsg(mp);
7180 7181 reass_done:
7181 7182 mutex_exit(&ipfb->ipfb_lock);
7182 7183 return (NULL);
7183 7184 }
7184 7185
7185 7186 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmReqds);
7186 7187 mp1->b_cont = mp;
7187 7188
7188 7189 /* Initialize the fragment header. */
7189 7190 ipf = (ipf_t *)mp1->b_rptr;
7190 7191 ipf->ipf_mp = mp1;
7191 7192 ipf->ipf_ptphn = ipfp;
7192 7193 ipfp[0] = ipf;
7193 7194 ipf->ipf_hash_next = NULL;
7194 7195 ipf->ipf_ident = ident;
7195 7196 ipf->ipf_protocol = proto;
7196 7197 ipf->ipf_src = src;
7197 7198 ipf->ipf_dst = dst;
7198 7199 ipf->ipf_nf_hdr_len = 0;
7199 7200 /* Record reassembly start time. */
7200 7201 ipf->ipf_timestamp = gethrestime_sec();
7201 7202 /* Record ipf generation and account for frag header */
7202 7203 ipf->ipf_gen = ill->ill_ipf_gen++;
7203 7204 ipf->ipf_count = MBLKSIZE(mp1);
7204 7205 ipf->ipf_last_frag_seen = B_FALSE;
7205 7206 ipf->ipf_ecn = ecn_info;
7206 7207 ipf->ipf_num_dups = 0;
7207 7208 ipfb->ipfb_frag_pkts++;
7208 7209 ipf->ipf_checksum = 0;
7209 7210 ipf->ipf_checksum_flags = 0;
7210 7211
7211 7212 /* Store checksum value in fragment header */
7212 7213 if (sum_flags != 0) {
7213 7214 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7214 7215 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7215 7216 ipf->ipf_checksum = sum_val;
7216 7217 ipf->ipf_checksum_flags = sum_flags;
7217 7218 }
7218 7219
7219 7220 /*
7220 7221 * We handle reassembly two ways. In the easy case,
7221 7222 * where all the fragments show up in order, we do
7222 7223 * minimal bookkeeping, and just clip new pieces on
7223 7224 * the end. If we ever see a hole, then we go off
7224 7225 * to ip_reassemble which has to mark the pieces and
7225 7226 * keep track of the number of holes, etc. Obviously,
7226 7227 * the point of having both mechanisms is so we can
7227 7228 * handle the easy case as efficiently as possible.
7228 7229 */
7229 7230 if (offset == 0) {
7230 7231 /* Easy case, in-order reassembly so far. */
7231 7232 ipf->ipf_count += msg_len;
7232 7233 ipf->ipf_tail_mp = tail_mp;
7233 7234 /*
7234 7235 * Keep track of next expected offset in
7235 7236 * ipf_end.
7236 7237 */
7237 7238 ipf->ipf_end = end;
7238 7239 ipf->ipf_nf_hdr_len = hdr_length;
7239 7240 } else {
7240 7241 /* Hard case, hole at the beginning. */
7241 7242 ipf->ipf_tail_mp = NULL;
7242 7243 /*
7243 7244 * ipf_end == 0 means that we have given up
7244 7245 * on easy reassembly.
7245 7246 */
7246 7247 ipf->ipf_end = 0;
7247 7248
7248 7249 /* Forget checksum offload from now on */
7249 7250 ipf->ipf_checksum_flags = 0;
7250 7251
7251 7252 /*
7252 7253 * ipf_hole_cnt is set by ip_reassemble.
7253 7254 * ipf_count is updated by ip_reassemble.
7254 7255 * No need to check for return value here
7255 7256 * as we don't expect reassembly to complete
7256 7257 * or fail for the first fragment itself.
7257 7258 */
7258 7259 (void) ip_reassemble(mp, ipf,
7259 7260 (frag_offset_flags & IPH_OFFSET) << 3,
7260 7261 (frag_offset_flags & IPH_MF), ill, msg_len);
7261 7262 }
7262 7263 /* Update per ipfb and ill byte counts */
7263 7264 ipfb->ipfb_count += ipf->ipf_count;
7264 7265 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7265 7266 atomic_add_32(&ill->ill_frag_count, ipf->ipf_count);
7266 7267 /* If the frag timer wasn't already going, start it. */
7267 7268 mutex_enter(&ill->ill_lock);
7268 7269 ill_frag_timer_start(ill);
7269 7270 mutex_exit(&ill->ill_lock);
7270 7271 goto reass_done;
7271 7272 }
7272 7273
7273 7274 /*
7274 7275 * If the packet's flag has changed (it could be coming up
7275 7276 * from an interface different than the previous, therefore
7276 7277 * possibly different checksum capability), then forget about
7277 7278 * any stored checksum states. Otherwise add the value to
7278 7279 * the existing one stored in the fragment header.
7279 7280 */
7280 7281 if (sum_flags != 0 && sum_flags == ipf->ipf_checksum_flags) {
7281 7282 sum_val += ipf->ipf_checksum;
7282 7283 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7283 7284 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7284 7285 ipf->ipf_checksum = sum_val;
7285 7286 } else if (ipf->ipf_checksum_flags != 0) {
7286 7287 /* Forget checksum offload from now on */
7287 7288 ipf->ipf_checksum_flags = 0;
7288 7289 }
7289 7290
7290 7291 /*
7291 7292 * We have a new piece of a datagram which is already being
7292 7293 * reassembled. Update the ECN info if all IP fragments
7293 7294 * are ECN capable. If there is one which is not, clear
7294 7295 * all the info. If there is at least one which has CE
7295 7296 * code point, IP needs to report that up to transport.
7296 7297 */
7297 7298 if (ecn_info != IPH_ECN_NECT && ipf->ipf_ecn != IPH_ECN_NECT) {
7298 7299 if (ecn_info == IPH_ECN_CE)
7299 7300 ipf->ipf_ecn = IPH_ECN_CE;
7300 7301 } else {
7301 7302 ipf->ipf_ecn = IPH_ECN_NECT;
7302 7303 }
7303 7304 if (offset && ipf->ipf_end == offset) {
7304 7305 /* The new fragment fits at the end */
7305 7306 ipf->ipf_tail_mp->b_cont = mp;
7306 7307 /* Update the byte count */
7307 7308 ipf->ipf_count += msg_len;
7308 7309 /* Update per ipfb and ill byte counts */
7309 7310 ipfb->ipfb_count += msg_len;
7310 7311 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7311 7312 atomic_add_32(&ill->ill_frag_count, msg_len);
7312 7313 if (frag_offset_flags & IPH_MF) {
7313 7314 /* More to come. */
7314 7315 ipf->ipf_end = end;
7315 7316 ipf->ipf_tail_mp = tail_mp;
7316 7317 goto reass_done;
7317 7318 }
7318 7319 } else {
7319 7320 /* Go do the hard cases. */
7320 7321 int ret;
7321 7322
7322 7323 if (offset == 0)
7323 7324 ipf->ipf_nf_hdr_len = hdr_length;
7324 7325
7325 7326 /* Save current byte count */
7326 7327 count = ipf->ipf_count;
7327 7328 ret = ip_reassemble(mp, ipf,
7328 7329 (frag_offset_flags & IPH_OFFSET) << 3,
7329 7330 (frag_offset_flags & IPH_MF), ill, msg_len);
7330 7331 /* Count of bytes added and subtracted (freeb()ed) */
7331 7332 count = ipf->ipf_count - count;
7332 7333 if (count) {
7333 7334 /* Update per ipfb and ill byte counts */
7334 7335 ipfb->ipfb_count += count;
7335 7336 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7336 7337 atomic_add_32(&ill->ill_frag_count, count);
7337 7338 }
7338 7339 if (ret == IP_REASS_PARTIAL) {
7339 7340 goto reass_done;
7340 7341 } else if (ret == IP_REASS_FAILED) {
7341 7342 /* Reassembly failed. Free up all resources */
7342 7343 ill_frag_free_pkts(ill, ipfb, ipf, 1);
7343 7344 for (t_mp = mp; t_mp != NULL; t_mp = t_mp->b_cont) {
7344 7345 IP_REASS_SET_START(t_mp, 0);
7345 7346 IP_REASS_SET_END(t_mp, 0);
7346 7347 }
7347 7348 freemsg(mp);
7348 7349 goto reass_done;
7349 7350 }
7350 7351 /* We will reach here iff 'ret' is IP_REASS_COMPLETE */
7351 7352 }
7352 7353 /*
7353 7354 * We have completed reassembly. Unhook the frag header from
7354 7355 * the reassembly list.
7355 7356 *
7356 7357 * Before we free the frag header, record the ECN info
7357 7358 * to report back to the transport.
7358 7359 */
7359 7360 ecn_info = ipf->ipf_ecn;
7360 7361 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmOKs);
7361 7362 ipfp = ipf->ipf_ptphn;
7362 7363
7363 7364 /* We need to supply these to caller */
7364 7365 if ((sum_flags = ipf->ipf_checksum_flags) != 0)
7365 7366 sum_val = ipf->ipf_checksum;
7366 7367 else
7367 7368 sum_val = 0;
7368 7369
7369 7370 mp1 = ipf->ipf_mp;
7370 7371 count = ipf->ipf_count;
7371 7372 ipf = ipf->ipf_hash_next;
7372 7373 if (ipf != NULL)
7373 7374 ipf->ipf_ptphn = ipfp;
7374 7375 ipfp[0] = ipf;
7375 7376 atomic_add_32(&ill->ill_frag_count, -count);
7376 7377 ASSERT(ipfb->ipfb_count >= count);
7377 7378 ipfb->ipfb_count -= count;
7378 7379 ipfb->ipfb_frag_pkts--;
7379 7380 mutex_exit(&ipfb->ipfb_lock);
7380 7381 /* Ditch the frag header. */
7381 7382 mp = mp1->b_cont;
7382 7383
7383 7384 freeb(mp1);
7384 7385
7385 7386 /* Restore original IP length in header. */
7386 7387 packet_size = (uint32_t)msgdsize(mp);
7387 7388 if (packet_size > IP_MAXPACKET) {
7388 7389 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7389 7390 ip_drop_input("Reassembled packet too large", mp, ill);
7390 7391 freemsg(mp);
7391 7392 return (NULL);
7392 7393 }
7393 7394
7394 7395 if (DB_REF(mp) > 1) {
7395 7396 mblk_t *mp2 = copymsg(mp);
7396 7397
7397 7398 if (mp2 == NULL) {
7398 7399 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7399 7400 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7400 7401 freemsg(mp);
7401 7402 return (NULL);
7402 7403 }
7403 7404 freemsg(mp);
7404 7405 mp = mp2;
7405 7406 }
7406 7407 ipha = (ipha_t *)mp->b_rptr;
7407 7408
7408 7409 ipha->ipha_length = htons((uint16_t)packet_size);
7409 7410 /* We're now complete, zip the frag state */
7410 7411 ipha->ipha_fragment_offset_and_flags = 0;
7411 7412 /* Record the ECN info. */
7412 7413 ipha->ipha_type_of_service &= 0xFC;
7413 7414 ipha->ipha_type_of_service |= ecn_info;
7414 7415
7415 7416 /* Update the receive attributes */
7416 7417 ira->ira_pktlen = packet_size;
7417 7418 ira->ira_ip_hdr_length = IPH_HDR_LENGTH(ipha);
7418 7419
7419 7420 /* Reassembly is successful; set checksum information in packet */
7420 7421 DB_CKSUM16(mp) = (uint16_t)sum_val;
7421 7422 DB_CKSUMFLAGS(mp) = sum_flags;
7422 7423 DB_CKSUMSTART(mp) = ira->ira_ip_hdr_length;
7423 7424
7424 7425 return (mp);
7425 7426 }
7426 7427
7427 7428 /*
7428 7429 * Pullup function that should be used for IP input in order to
7429 7430 * ensure we do not loose the L2 source address; we need the l2 source
7430 7431 * address for IP_RECVSLLA and for ndp_input.
7431 7432 *
7432 7433 * We return either NULL or b_rptr.
7433 7434 */
7434 7435 void *
7435 7436 ip_pullup(mblk_t *mp, ssize_t len, ip_recv_attr_t *ira)
7436 7437 {
7437 7438 ill_t *ill = ira->ira_ill;
7438 7439
7439 7440 if (ip_rput_pullups++ == 0) {
7440 7441 (void) mi_strlog(ill->ill_rq, 1, SL_ERROR|SL_TRACE,
7441 7442 "ip_pullup: %s forced us to "
7442 7443 " pullup pkt, hdr len %ld, hdr addr %p",
7443 7444 ill->ill_name, len, (void *)mp->b_rptr);
7444 7445 }
7445 7446 if (!(ira->ira_flags & IRAF_L2SRC_SET))
7446 7447 ip_setl2src(mp, ira, ira->ira_rill);
7447 7448 ASSERT(ira->ira_flags & IRAF_L2SRC_SET);
7448 7449 if (!pullupmsg(mp, len))
7449 7450 return (NULL);
7450 7451 else
7451 7452 return (mp->b_rptr);
7452 7453 }
7453 7454
7454 7455 /*
7455 7456 * Make sure ira_l2src has an address. If we don't have one fill with zeros.
7456 7457 * When called from the ULP ira_rill will be NULL hence the caller has to
7457 7458 * pass in the ill.
7458 7459 */
7459 7460 /* ARGSUSED */
7460 7461 void
7461 7462 ip_setl2src(mblk_t *mp, ip_recv_attr_t *ira, ill_t *ill)
7462 7463 {
7463 7464 const uchar_t *addr;
7464 7465 int alen;
7465 7466
7466 7467 if (ira->ira_flags & IRAF_L2SRC_SET)
7467 7468 return;
7468 7469
7469 7470 ASSERT(ill != NULL);
7470 7471 alen = ill->ill_phys_addr_length;
7471 7472 ASSERT(alen <= sizeof (ira->ira_l2src));
7472 7473 if (ira->ira_mhip != NULL &&
7473 7474 (addr = ira->ira_mhip->mhi_saddr) != NULL) {
7474 7475 bcopy(addr, ira->ira_l2src, alen);
7475 7476 } else if ((ira->ira_flags & IRAF_L2SRC_LOOPBACK) &&
7476 7477 (addr = ill->ill_phys_addr) != NULL) {
7477 7478 bcopy(addr, ira->ira_l2src, alen);
7478 7479 } else {
7479 7480 bzero(ira->ira_l2src, alen);
7480 7481 }
7481 7482 ira->ira_flags |= IRAF_L2SRC_SET;
7482 7483 }
7483 7484
7484 7485 /*
7485 7486 * check ip header length and align it.
7486 7487 */
7487 7488 mblk_t *
7488 7489 ip_check_and_align_header(mblk_t *mp, uint_t min_size, ip_recv_attr_t *ira)
7489 7490 {
7490 7491 ill_t *ill = ira->ira_ill;
7491 7492 ssize_t len;
7492 7493
7493 7494 len = MBLKL(mp);
7494 7495
7495 7496 if (!OK_32PTR(mp->b_rptr))
7496 7497 IP_STAT(ill->ill_ipst, ip_notaligned);
7497 7498 else
7498 7499 IP_STAT(ill->ill_ipst, ip_recv_pullup);
7499 7500
7500 7501 /* Guard against bogus device drivers */
7501 7502 if (len < 0) {
7502 7503 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7503 7504 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7504 7505 freemsg(mp);
7505 7506 return (NULL);
7506 7507 }
7507 7508
7508 7509 if (len == 0) {
7509 7510 /* GLD sometimes sends up mblk with b_rptr == b_wptr! */
7510 7511 mblk_t *mp1 = mp->b_cont;
7511 7512
7512 7513 if (!(ira->ira_flags & IRAF_L2SRC_SET))
7513 7514 ip_setl2src(mp, ira, ira->ira_rill);
7514 7515 ASSERT(ira->ira_flags & IRAF_L2SRC_SET);
7515 7516
7516 7517 freeb(mp);
7517 7518 mp = mp1;
7518 7519 if (mp == NULL)
7519 7520 return (NULL);
7520 7521
7521 7522 if (OK_32PTR(mp->b_rptr) && MBLKL(mp) >= min_size)
7522 7523 return (mp);
7523 7524 }
7524 7525 if (ip_pullup(mp, min_size, ira) == NULL) {
7525 7526 if (msgdsize(mp) < min_size) {
7526 7527 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7527 7528 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7528 7529 } else {
7529 7530 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7530 7531 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7531 7532 }
7532 7533 freemsg(mp);
7533 7534 return (NULL);
7534 7535 }
7535 7536 return (mp);
7536 7537 }
7537 7538
7538 7539 /*
7539 7540 * Common code for IPv4 and IPv6 to check and pullup multi-mblks
7540 7541 */
7541 7542 mblk_t *
7542 7543 ip_check_length(mblk_t *mp, uchar_t *rptr, ssize_t len, uint_t pkt_len,
7543 7544 uint_t min_size, ip_recv_attr_t *ira)
7544 7545 {
7545 7546 ill_t *ill = ira->ira_ill;
7546 7547
7547 7548 /*
7548 7549 * Make sure we have data length consistent
7549 7550 * with the IP header.
7550 7551 */
7551 7552 if (mp->b_cont == NULL) {
7552 7553 /* pkt_len is based on ipha_len, not the mblk length */
7553 7554 if (pkt_len < min_size) {
7554 7555 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7555 7556 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7556 7557 freemsg(mp);
7557 7558 return (NULL);
7558 7559 }
7559 7560 if (len < 0) {
7560 7561 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
7561 7562 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
7562 7563 freemsg(mp);
7563 7564 return (NULL);
7564 7565 }
7565 7566 /* Drop any pad */
7566 7567 mp->b_wptr = rptr + pkt_len;
7567 7568 } else if ((len += msgdsize(mp->b_cont)) != 0) {
7568 7569 ASSERT(pkt_len >= min_size);
7569 7570 if (pkt_len < min_size) {
7570 7571 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7571 7572 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7572 7573 freemsg(mp);
7573 7574 return (NULL);
7574 7575 }
7575 7576 if (len < 0) {
7576 7577 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
7577 7578 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
7578 7579 freemsg(mp);
7579 7580 return (NULL);
7580 7581 }
7581 7582 /* Drop any pad */
7582 7583 (void) adjmsg(mp, -len);
7583 7584 /*
7584 7585 * adjmsg may have freed an mblk from the chain, hence
7585 7586 * invalidate any hw checksum here. This will force IP to
7586 7587 * calculate the checksum in sw, but only for this packet.
7587 7588 */
7588 7589 DB_CKSUMFLAGS(mp) = 0;
7589 7590 IP_STAT(ill->ill_ipst, ip_multimblk);
7590 7591 }
7591 7592 return (mp);
7592 7593 }
7593 7594
7594 7595 /*
7595 7596 * Check that the IPv4 opt_len is consistent with the packet and pullup
7596 7597 * the options.
7597 7598 */
7598 7599 mblk_t *
7599 7600 ip_check_optlen(mblk_t *mp, ipha_t *ipha, uint_t opt_len, uint_t pkt_len,
7600 7601 ip_recv_attr_t *ira)
7601 7602 {
7602 7603 ill_t *ill = ira->ira_ill;
7603 7604 ssize_t len;
7604 7605
7605 7606 /* Assume no IPv6 packets arrive over the IPv4 queue */
7606 7607 if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
7607 7608 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7608 7609 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInWrongIPVersion);
7609 7610 ip_drop_input("IPvN packet on IPv4 ill", mp, ill);
7610 7611 freemsg(mp);
7611 7612 return (NULL);
7612 7613 }
7613 7614
7614 7615 if (opt_len > (15 - IP_SIMPLE_HDR_LENGTH_IN_WORDS)) {
7615 7616 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7616 7617 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7617 7618 freemsg(mp);
7618 7619 return (NULL);
7619 7620 }
7620 7621 /*
7621 7622 * Recompute complete header length and make sure we
7622 7623 * have access to all of it.
7623 7624 */
7624 7625 len = ((size_t)opt_len + IP_SIMPLE_HDR_LENGTH_IN_WORDS) << 2;
7625 7626 if (len > (mp->b_wptr - mp->b_rptr)) {
7626 7627 if (len > pkt_len) {
7627 7628 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7628 7629 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7629 7630 freemsg(mp);
7630 7631 return (NULL);
7631 7632 }
7632 7633 if (ip_pullup(mp, len, ira) == NULL) {
7633 7634 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7634 7635 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7635 7636 freemsg(mp);
7636 7637 return (NULL);
7637 7638 }
7638 7639 }
7639 7640 return (mp);
7640 7641 }
7641 7642
7642 7643 /*
7643 7644 * Returns a new ire, or the same ire, or NULL.
7644 7645 * If a different IRE is returned, then it is held; the caller
7645 7646 * needs to release it.
7646 7647 * In no case is there any hold/release on the ire argument.
7647 7648 */
7648 7649 ire_t *
7649 7650 ip_check_multihome(void *addr, ire_t *ire, ill_t *ill)
7650 7651 {
7651 7652 ire_t *new_ire;
7652 7653 ill_t *ire_ill;
7653 7654 uint_t ifindex;
7654 7655 ip_stack_t *ipst = ill->ill_ipst;
7655 7656 boolean_t strict_check = B_FALSE;
7656 7657
7657 7658 /*
7658 7659 * IPMP common case: if IRE and ILL are in the same group, there's no
7659 7660 * issue (e.g. packet received on an underlying interface matched an
7660 7661 * IRE_LOCAL on its associated group interface).
7661 7662 */
7662 7663 ASSERT(ire->ire_ill != NULL);
7663 7664 if (IS_IN_SAME_ILLGRP(ill, ire->ire_ill))
7664 7665 return (ire);
7665 7666
7666 7667 /*
7667 7668 * Do another ire lookup here, using the ingress ill, to see if the
7668 7669 * interface is in a usesrc group.
7669 7670 * As long as the ills belong to the same group, we don't consider
7670 7671 * them to be arriving on the wrong interface. Thus, if the switch
7671 7672 * is doing inbound load spreading, we won't drop packets when the
7672 7673 * ip*_strict_dst_multihoming switch is on.
7673 7674 * We also need to check for IPIF_UNNUMBERED point2point interfaces
7674 7675 * where the local address may not be unique. In this case we were
7675 7676 * at the mercy of the initial ire lookup and the IRE_LOCAL it
7676 7677 * actually returned. The new lookup, which is more specific, should
7677 7678 * only find the IRE_LOCAL associated with the ingress ill if one
7678 7679 * exists.
7679 7680 */
7680 7681 if (ire->ire_ipversion == IPV4_VERSION) {
7681 7682 if (ipst->ips_ip_strict_dst_multihoming)
7682 7683 strict_check = B_TRUE;
7683 7684 new_ire = ire_ftable_lookup_v4(*((ipaddr_t *)addr), 0, 0,
7684 7685 IRE_LOCAL, ill, ALL_ZONES, NULL,
7685 7686 (MATCH_IRE_TYPE|MATCH_IRE_ILL), 0, ipst, NULL);
7686 7687 } else {
7687 7688 ASSERT(!IN6_IS_ADDR_MULTICAST((in6_addr_t *)addr));
7688 7689 if (ipst->ips_ipv6_strict_dst_multihoming)
7689 7690 strict_check = B_TRUE;
7690 7691 new_ire = ire_ftable_lookup_v6((in6_addr_t *)addr, NULL, NULL,
7691 7692 IRE_LOCAL, ill, ALL_ZONES, NULL,
7692 7693 (MATCH_IRE_TYPE|MATCH_IRE_ILL), 0, ipst, NULL);
7693 7694 }
7694 7695 /*
7695 7696 * If the same ire that was returned in ip_input() is found then this
7696 7697 * is an indication that usesrc groups are in use. The packet
7697 7698 * arrived on a different ill in the group than the one associated with
7698 7699 * the destination address. If a different ire was found then the same
7699 7700 * IP address must be hosted on multiple ills. This is possible with
7700 7701 * unnumbered point2point interfaces. We switch to use this new ire in
7701 7702 * order to have accurate interface statistics.
7702 7703 */
7703 7704 if (new_ire != NULL) {
7704 7705 /* Note: held in one case but not the other? Caller handles */
7705 7706 if (new_ire != ire)
7706 7707 return (new_ire);
7707 7708 /* Unchanged */
7708 7709 ire_refrele(new_ire);
7709 7710 return (ire);
7710 7711 }
7711 7712
7712 7713 /*
7713 7714 * Chase pointers once and store locally.
7714 7715 */
7715 7716 ASSERT(ire->ire_ill != NULL);
7716 7717 ire_ill = ire->ire_ill;
7717 7718 ifindex = ill->ill_usesrc_ifindex;
7718 7719
7719 7720 /*
7720 7721 * Check if it's a legal address on the 'usesrc' interface.
7721 7722 * For IPMP data addresses the IRE_LOCAL is the upper, hence we
7722 7723 * can just check phyint_ifindex.
7723 7724 */
7724 7725 if (ifindex != 0 && ifindex == ire_ill->ill_phyint->phyint_ifindex) {
7725 7726 return (ire);
7726 7727 }
7727 7728
7728 7729 /*
7729 7730 * If the ip*_strict_dst_multihoming switch is on then we can
7730 7731 * only accept this packet if the interface is marked as routing.
7731 7732 */
7732 7733 if (!(strict_check))
7733 7734 return (ire);
7734 7735
7735 7736 if ((ill->ill_flags & ire->ire_ill->ill_flags & ILLF_ROUTER) != 0) {
7736 7737 return (ire);
7737 7738 }
7738 7739 return (NULL);
7739 7740 }
7740 7741
7741 7742 /*
7742 7743 * This function is used to construct a mac_header_info_s from a
7743 7744 * DL_UNITDATA_IND message.
7744 7745 * The address fields in the mhi structure points into the message,
7745 7746 * thus the caller can't use those fields after freeing the message.
7746 7747 *
7747 7748 * We determine whether the packet received is a non-unicast packet
7748 7749 * and in doing so, determine whether or not it is broadcast vs multicast.
7749 7750 * For it to be a broadcast packet, we must have the appropriate mblk_t
7750 7751 * hanging off the ill_t. If this is either not present or doesn't match
7751 7752 * the destination mac address in the DL_UNITDATA_IND, the packet is deemed
7752 7753 * to be multicast. Thus NICs that have no broadcast address (or no
7753 7754 * capability for one, such as point to point links) cannot return as
7754 7755 * the packet being broadcast.
7755 7756 */
7756 7757 void
7757 7758 ip_dlur_to_mhi(ill_t *ill, mblk_t *mb, struct mac_header_info_s *mhip)
7758 7759 {
7759 7760 dl_unitdata_ind_t *ind = (dl_unitdata_ind_t *)mb->b_rptr;
7760 7761 mblk_t *bmp;
7761 7762 uint_t extra_offset;
7762 7763
7763 7764 bzero(mhip, sizeof (struct mac_header_info_s));
7764 7765
7765 7766 mhip->mhi_dsttype = MAC_ADDRTYPE_UNICAST;
7766 7767
7767 7768 if (ill->ill_sap_length < 0)
7768 7769 extra_offset = 0;
7769 7770 else
7770 7771 extra_offset = ill->ill_sap_length;
7771 7772
7772 7773 mhip->mhi_daddr = (uchar_t *)ind + ind->dl_dest_addr_offset +
7773 7774 extra_offset;
7774 7775 mhip->mhi_saddr = (uchar_t *)ind + ind->dl_src_addr_offset +
7775 7776 extra_offset;
7776 7777
7777 7778 if (!ind->dl_group_address)
7778 7779 return;
7779 7780
7780 7781 /* Multicast or broadcast */
7781 7782 mhip->mhi_dsttype = MAC_ADDRTYPE_MULTICAST;
7782 7783
7783 7784 if (ind->dl_dest_addr_offset > sizeof (*ind) &&
7784 7785 ind->dl_dest_addr_offset + ind->dl_dest_addr_length < MBLKL(mb) &&
7785 7786 (bmp = ill->ill_bcast_mp) != NULL) {
7786 7787 dl_unitdata_req_t *dlur;
7787 7788 uint8_t *bphys_addr;
7788 7789
7789 7790 dlur = (dl_unitdata_req_t *)bmp->b_rptr;
7790 7791 bphys_addr = (uchar_t *)dlur + dlur->dl_dest_addr_offset +
7791 7792 extra_offset;
7792 7793
7793 7794 if (bcmp(mhip->mhi_daddr, bphys_addr,
7794 7795 ind->dl_dest_addr_length) == 0)
7795 7796 mhip->mhi_dsttype = MAC_ADDRTYPE_BROADCAST;
7796 7797 }
7797 7798 }
7798 7799
7799 7800 /*
7800 7801 * This function is used to construct a mac_header_info_s from a
7801 7802 * M_DATA fastpath message from a DLPI driver.
7802 7803 * The address fields in the mhi structure points into the message,
7803 7804 * thus the caller can't use those fields after freeing the message.
7804 7805 *
7805 7806 * We determine whether the packet received is a non-unicast packet
7806 7807 * and in doing so, determine whether or not it is broadcast vs multicast.
7807 7808 * For it to be a broadcast packet, we must have the appropriate mblk_t
7808 7809 * hanging off the ill_t. If this is either not present or doesn't match
7809 7810 * the destination mac address in the DL_UNITDATA_IND, the packet is deemed
7810 7811 * to be multicast. Thus NICs that have no broadcast address (or no
7811 7812 * capability for one, such as point to point links) cannot return as
7812 7813 * the packet being broadcast.
7813 7814 */
7814 7815 void
7815 7816 ip_mdata_to_mhi(ill_t *ill, mblk_t *mp, struct mac_header_info_s *mhip)
7816 7817 {
7817 7818 mblk_t *bmp;
7818 7819 struct ether_header *pether;
7819 7820
7820 7821 bzero(mhip, sizeof (struct mac_header_info_s));
7821 7822
7822 7823 mhip->mhi_dsttype = MAC_ADDRTYPE_UNICAST;
7823 7824
7824 7825 pether = (struct ether_header *)((char *)mp->b_rptr
7825 7826 - sizeof (struct ether_header));
7826 7827
7827 7828 /*
7828 7829 * Make sure the interface is an ethernet type, since we don't
7829 7830 * know the header format for anything but Ethernet. Also make
7830 7831 * sure we are pointing correctly above db_base.
7831 7832 */
7832 7833 if (ill->ill_type != IFT_ETHER)
7833 7834 return;
7834 7835
7835 7836 retry:
7836 7837 if ((uchar_t *)pether < mp->b_datap->db_base)
7837 7838 return;
7838 7839
7839 7840 /* Is there a VLAN tag? */
7840 7841 if (ill->ill_isv6) {
7841 7842 if (pether->ether_type != htons(ETHERTYPE_IPV6)) {
7842 7843 pether = (struct ether_header *)((char *)pether - 4);
7843 7844 goto retry;
7844 7845 }
7845 7846 } else {
7846 7847 if (pether->ether_type != htons(ETHERTYPE_IP)) {
7847 7848 pether = (struct ether_header *)((char *)pether - 4);
7848 7849 goto retry;
7849 7850 }
7850 7851 }
7851 7852 mhip->mhi_daddr = (uchar_t *)&pether->ether_dhost;
7852 7853 mhip->mhi_saddr = (uchar_t *)&pether->ether_shost;
7853 7854
7854 7855 if (!(mhip->mhi_daddr[0] & 0x01))
7855 7856 return;
7856 7857
7857 7858 /* Multicast or broadcast */
7858 7859 mhip->mhi_dsttype = MAC_ADDRTYPE_MULTICAST;
7859 7860
7860 7861 if ((bmp = ill->ill_bcast_mp) != NULL) {
7861 7862 dl_unitdata_req_t *dlur;
7862 7863 uint8_t *bphys_addr;
7863 7864 uint_t addrlen;
7864 7865
7865 7866 dlur = (dl_unitdata_req_t *)bmp->b_rptr;
7866 7867 addrlen = dlur->dl_dest_addr_length;
7867 7868 if (ill->ill_sap_length < 0) {
7868 7869 bphys_addr = (uchar_t *)dlur +
7869 7870 dlur->dl_dest_addr_offset;
7870 7871 addrlen += ill->ill_sap_length;
7871 7872 } else {
7872 7873 bphys_addr = (uchar_t *)dlur +
7873 7874 dlur->dl_dest_addr_offset +
7874 7875 ill->ill_sap_length;
7875 7876 addrlen -= ill->ill_sap_length;
7876 7877 }
7877 7878 if (bcmp(mhip->mhi_daddr, bphys_addr, addrlen) == 0)
7878 7879 mhip->mhi_dsttype = MAC_ADDRTYPE_BROADCAST;
7879 7880 }
7880 7881 }
7881 7882
7882 7883 /*
7883 7884 * Handle anything but M_DATA messages
7884 7885 * We see the DL_UNITDATA_IND which are part
7885 7886 * of the data path, and also the other messages from the driver.
7886 7887 */
7887 7888 void
7888 7889 ip_rput_notdata(ill_t *ill, mblk_t *mp)
7889 7890 {
7890 7891 mblk_t *first_mp;
7891 7892 struct iocblk *iocp;
7892 7893 struct mac_header_info_s mhi;
7893 7894
7894 7895 switch (DB_TYPE(mp)) {
7895 7896 case M_PROTO:
7896 7897 case M_PCPROTO: {
7897 7898 if (((dl_unitdata_ind_t *)mp->b_rptr)->dl_primitive !=
7898 7899 DL_UNITDATA_IND) {
7899 7900 /* Go handle anything other than data elsewhere. */
7900 7901 ip_rput_dlpi(ill, mp);
7901 7902 return;
7902 7903 }
7903 7904
7904 7905 first_mp = mp;
7905 7906 mp = first_mp->b_cont;
7906 7907 first_mp->b_cont = NULL;
7907 7908
7908 7909 if (mp == NULL) {
7909 7910 freeb(first_mp);
7910 7911 return;
7911 7912 }
7912 7913 ip_dlur_to_mhi(ill, first_mp, &mhi);
7913 7914 if (ill->ill_isv6)
7914 7915 ip_input_v6(ill, NULL, mp, &mhi);
7915 7916 else
7916 7917 ip_input(ill, NULL, mp, &mhi);
7917 7918
7918 7919 /* Ditch the DLPI header. */
7919 7920 freeb(first_mp);
7920 7921 return;
7921 7922 }
7922 7923 case M_IOCACK:
7923 7924 iocp = (struct iocblk *)mp->b_rptr;
7924 7925 switch (iocp->ioc_cmd) {
7925 7926 case DL_IOC_HDR_INFO:
7926 7927 ill_fastpath_ack(ill, mp);
7927 7928 return;
7928 7929 default:
7929 7930 putnext(ill->ill_rq, mp);
7930 7931 return;
7931 7932 }
7932 7933 /* FALLTHRU */
7933 7934 case M_ERROR:
7934 7935 case M_HANGUP:
7935 7936 mutex_enter(&ill->ill_lock);
7936 7937 if (ill->ill_state_flags & ILL_CONDEMNED) {
7937 7938 mutex_exit(&ill->ill_lock);
7938 7939 freemsg(mp);
7939 7940 return;
7940 7941 }
7941 7942 ill_refhold_locked(ill);
7942 7943 mutex_exit(&ill->ill_lock);
7943 7944 qwriter_ip(ill, ill->ill_rq, mp, ip_rput_other, CUR_OP,
7944 7945 B_FALSE);
7945 7946 return;
7946 7947 case M_CTL:
7947 7948 putnext(ill->ill_rq, mp);
7948 7949 return;
7949 7950 case M_IOCNAK:
7950 7951 ip1dbg(("got iocnak "));
7951 7952 iocp = (struct iocblk *)mp->b_rptr;
7952 7953 switch (iocp->ioc_cmd) {
7953 7954 case DL_IOC_HDR_INFO:
7954 7955 ip_rput_other(NULL, ill->ill_rq, mp, NULL);
7955 7956 return;
7956 7957 default:
7957 7958 break;
7958 7959 }
7959 7960 /* FALLTHRU */
7960 7961 default:
7961 7962 putnext(ill->ill_rq, mp);
7962 7963 return;
7963 7964 }
7964 7965 }
7965 7966
7966 7967 /* Read side put procedure. Packets coming from the wire arrive here. */
7967 7968 void
7968 7969 ip_rput(queue_t *q, mblk_t *mp)
7969 7970 {
7970 7971 ill_t *ill;
7971 7972 union DL_primitives *dl;
7972 7973
7973 7974 ill = (ill_t *)q->q_ptr;
7974 7975
7975 7976 if (ill->ill_state_flags & (ILL_CONDEMNED | ILL_LL_SUBNET_PENDING)) {
7976 7977 /*
7977 7978 * If things are opening or closing, only accept high-priority
7978 7979 * DLPI messages. (On open ill->ill_ipif has not yet been
7979 7980 * created; on close, things hanging off the ill may have been
7980 7981 * freed already.)
7981 7982 */
7982 7983 dl = (union DL_primitives *)mp->b_rptr;
7983 7984 if (DB_TYPE(mp) != M_PCPROTO ||
7984 7985 dl->dl_primitive == DL_UNITDATA_IND) {
7985 7986 inet_freemsg(mp);
7986 7987 return;
7987 7988 }
7988 7989 }
7989 7990 if (DB_TYPE(mp) == M_DATA) {
7990 7991 struct mac_header_info_s mhi;
7991 7992
7992 7993 ip_mdata_to_mhi(ill, mp, &mhi);
7993 7994 ip_input(ill, NULL, mp, &mhi);
7994 7995 } else {
7995 7996 ip_rput_notdata(ill, mp);
7996 7997 }
7997 7998 }
7998 7999
7999 8000 /*
8000 8001 * Move the information to a copy.
8001 8002 */
8002 8003 mblk_t *
8003 8004 ip_fix_dbref(mblk_t *mp, ip_recv_attr_t *ira)
8004 8005 {
8005 8006 mblk_t *mp1;
8006 8007 ill_t *ill = ira->ira_ill;
8007 8008 ip_stack_t *ipst = ill->ill_ipst;
8008 8009
8009 8010 IP_STAT(ipst, ip_db_ref);
8010 8011
8011 8012 /* Make sure we have ira_l2src before we loose the original mblk */
8012 8013 if (!(ira->ira_flags & IRAF_L2SRC_SET))
8013 8014 ip_setl2src(mp, ira, ira->ira_rill);
8014 8015
8015 8016 mp1 = copymsg(mp);
8016 8017 if (mp1 == NULL) {
8017 8018 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
8018 8019 ip_drop_input("ipIfStatsInDiscards", mp, ill);
8019 8020 freemsg(mp);
8020 8021 return (NULL);
8021 8022 }
8022 8023 /* preserve the hardware checksum flags and data, if present */
8023 8024 if (DB_CKSUMFLAGS(mp) != 0) {
8024 8025 DB_CKSUMFLAGS(mp1) = DB_CKSUMFLAGS(mp);
8025 8026 DB_CKSUMSTART(mp1) = DB_CKSUMSTART(mp);
8026 8027 DB_CKSUMSTUFF(mp1) = DB_CKSUMSTUFF(mp);
8027 8028 DB_CKSUMEND(mp1) = DB_CKSUMEND(mp);
8028 8029 DB_CKSUM16(mp1) = DB_CKSUM16(mp);
8029 8030 }
8030 8031 freemsg(mp);
8031 8032 return (mp1);
8032 8033 }
8033 8034
8034 8035 static void
8035 8036 ip_dlpi_error(ill_t *ill, t_uscalar_t prim, t_uscalar_t dl_err,
8036 8037 t_uscalar_t err)
8037 8038 {
8038 8039 if (dl_err == DL_SYSERR) {
8039 8040 (void) mi_strlog(ill->ill_rq, 1, SL_CONSOLE|SL_ERROR|SL_TRACE,
8040 8041 "%s: %s failed: DL_SYSERR (errno %u)\n",
8041 8042 ill->ill_name, dl_primstr(prim), err);
8042 8043 return;
8043 8044 }
8044 8045
8045 8046 (void) mi_strlog(ill->ill_rq, 1, SL_CONSOLE|SL_ERROR|SL_TRACE,
8046 8047 "%s: %s failed: %s\n", ill->ill_name, dl_primstr(prim),
8047 8048 dl_errstr(dl_err));
8048 8049 }
8049 8050
8050 8051 /*
8051 8052 * ip_rput_dlpi is called by ip_rput to handle all DLPI messages other
8052 8053 * than DL_UNITDATA_IND messages. If we need to process this message
8053 8054 * exclusively, we call qwriter_ip, in which case we also need to call
8054 8055 * ill_refhold before that, since qwriter_ip does an ill_refrele.
8055 8056 */
8056 8057 void
8057 8058 ip_rput_dlpi(ill_t *ill, mblk_t *mp)
8058 8059 {
8059 8060 dl_ok_ack_t *dloa = (dl_ok_ack_t *)mp->b_rptr;
8060 8061 dl_error_ack_t *dlea = (dl_error_ack_t *)dloa;
8061 8062 queue_t *q = ill->ill_rq;
8062 8063 t_uscalar_t prim = dloa->dl_primitive;
8063 8064 t_uscalar_t reqprim = DL_PRIM_INVAL;
8064 8065
8065 8066 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi",
8066 8067 char *, dl_primstr(prim), ill_t *, ill);
8067 8068 ip1dbg(("ip_rput_dlpi"));
8068 8069
8069 8070 /*
8070 8071 * If we received an ACK but didn't send a request for it, then it
8071 8072 * can't be part of any pending operation; discard up-front.
8072 8073 */
8073 8074 switch (prim) {
8074 8075 case DL_ERROR_ACK:
8075 8076 reqprim = dlea->dl_error_primitive;
8076 8077 ip2dbg(("ip_rput_dlpi(%s): DL_ERROR_ACK for %s (0x%x): %s "
8077 8078 "(0x%x), unix %u\n", ill->ill_name, dl_primstr(reqprim),
8078 8079 reqprim, dl_errstr(dlea->dl_errno), dlea->dl_errno,
8079 8080 dlea->dl_unix_errno));
8080 8081 break;
8081 8082 case DL_OK_ACK:
8082 8083 reqprim = dloa->dl_correct_primitive;
8083 8084 break;
8084 8085 case DL_INFO_ACK:
8085 8086 reqprim = DL_INFO_REQ;
8086 8087 break;
8087 8088 case DL_BIND_ACK:
8088 8089 reqprim = DL_BIND_REQ;
8089 8090 break;
8090 8091 case DL_PHYS_ADDR_ACK:
8091 8092 reqprim = DL_PHYS_ADDR_REQ;
8092 8093 break;
8093 8094 case DL_NOTIFY_ACK:
8094 8095 reqprim = DL_NOTIFY_REQ;
8095 8096 break;
8096 8097 case DL_CAPABILITY_ACK:
8097 8098 reqprim = DL_CAPABILITY_REQ;
8098 8099 break;
8099 8100 }
8100 8101
8101 8102 if (prim != DL_NOTIFY_IND) {
8102 8103 if (reqprim == DL_PRIM_INVAL ||
8103 8104 !ill_dlpi_pending(ill, reqprim)) {
8104 8105 /* Not a DLPI message we support or expected */
8105 8106 freemsg(mp);
8106 8107 return;
8107 8108 }
8108 8109 ip1dbg(("ip_rput: received %s for %s\n", dl_primstr(prim),
8109 8110 dl_primstr(reqprim)));
8110 8111 }
8111 8112
8112 8113 switch (reqprim) {
8113 8114 case DL_UNBIND_REQ:
8114 8115 /*
8115 8116 * NOTE: we mark the unbind as complete even if we got a
8116 8117 * DL_ERROR_ACK, since there's not much else we can do.
8117 8118 */
8118 8119 mutex_enter(&ill->ill_lock);
8119 8120 ill->ill_state_flags &= ~ILL_DL_UNBIND_IN_PROGRESS;
8120 8121 cv_signal(&ill->ill_cv);
8121 8122 mutex_exit(&ill->ill_lock);
8122 8123 break;
8123 8124
8124 8125 case DL_ENABMULTI_REQ:
8125 8126 if (prim == DL_OK_ACK) {
8126 8127 if (ill->ill_dlpi_multicast_state == IDS_INPROGRESS)
8127 8128 ill->ill_dlpi_multicast_state = IDS_OK;
8128 8129 }
8129 8130 break;
8130 8131 }
8131 8132
8132 8133 /*
8133 8134 * The message is one we're waiting for (or DL_NOTIFY_IND), but we
8134 8135 * need to become writer to continue to process it. Because an
8135 8136 * exclusive operation doesn't complete until replies to all queued
8136 8137 * DLPI messages have been received, we know we're in the middle of an
8137 8138 * exclusive operation and pass CUR_OP (except for DL_NOTIFY_IND).
8138 8139 *
8139 8140 * As required by qwriter_ip(), we refhold the ill; it will refrele.
8140 8141 * Since this is on the ill stream we unconditionally bump up the
8141 8142 * refcount without doing ILL_CAN_LOOKUP().
8142 8143 */
8143 8144 ill_refhold(ill);
8144 8145 if (prim == DL_NOTIFY_IND)
8145 8146 qwriter_ip(ill, q, mp, ip_rput_dlpi_writer, NEW_OP, B_FALSE);
8146 8147 else
8147 8148 qwriter_ip(ill, q, mp, ip_rput_dlpi_writer, CUR_OP, B_FALSE);
8148 8149 }
8149 8150
8150 8151 /*
8151 8152 * Handling of DLPI messages that require exclusive access to the ipsq.
8152 8153 *
8153 8154 * Need to do ipsq_pending_mp_get on ioctl completion, which could
8154 8155 * happen here. (along with mi_copy_done)
8155 8156 */
8156 8157 /* ARGSUSED */
8157 8158 static void
8158 8159 ip_rput_dlpi_writer(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
8159 8160 {
8160 8161 dl_ok_ack_t *dloa = (dl_ok_ack_t *)mp->b_rptr;
8161 8162 dl_error_ack_t *dlea = (dl_error_ack_t *)dloa;
8162 8163 int err = 0;
8163 8164 ill_t *ill = (ill_t *)q->q_ptr;
8164 8165 ipif_t *ipif = NULL;
8165 8166 mblk_t *mp1 = NULL;
8166 8167 conn_t *connp = NULL;
8167 8168 t_uscalar_t paddrreq;
8168 8169 mblk_t *mp_hw;
8169 8170 boolean_t success;
8170 8171 boolean_t ioctl_aborted = B_FALSE;
8171 8172 boolean_t log = B_TRUE;
8172 8173
8173 8174 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer",
8174 8175 char *, dl_primstr(dloa->dl_primitive), ill_t *, ill);
8175 8176
8176 8177 ip1dbg(("ip_rput_dlpi_writer .."));
8177 8178 ASSERT(ipsq->ipsq_xop == ill->ill_phyint->phyint_ipsq->ipsq_xop);
8178 8179 ASSERT(IAM_WRITER_ILL(ill));
8179 8180
8180 8181 ipif = ipsq->ipsq_xop->ipx_pending_ipif;
8181 8182 /*
8182 8183 * The current ioctl could have been aborted by the user and a new
8183 8184 * ioctl to bring up another ill could have started. We could still
8184 8185 * get a response from the driver later.
8185 8186 */
8186 8187 if (ipif != NULL && ipif->ipif_ill != ill)
8187 8188 ioctl_aborted = B_TRUE;
8188 8189
8189 8190 switch (dloa->dl_primitive) {
8190 8191 case DL_ERROR_ACK:
8191 8192 ip1dbg(("ip_rput_dlpi_writer: got DL_ERROR_ACK for %s\n",
8192 8193 dl_primstr(dlea->dl_error_primitive)));
8193 8194
8194 8195 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer error",
8195 8196 char *, dl_primstr(dlea->dl_error_primitive),
8196 8197 ill_t *, ill);
8197 8198
8198 8199 switch (dlea->dl_error_primitive) {
8199 8200 case DL_DISABMULTI_REQ:
8200 8201 ill_dlpi_done(ill, dlea->dl_error_primitive);
8201 8202 break;
8202 8203 case DL_PROMISCON_REQ:
8203 8204 case DL_PROMISCOFF_REQ:
8204 8205 case DL_UNBIND_REQ:
8205 8206 case DL_ATTACH_REQ:
8206 8207 case DL_INFO_REQ:
8207 8208 ill_dlpi_done(ill, dlea->dl_error_primitive);
8208 8209 break;
8209 8210 case DL_NOTIFY_REQ:
8210 8211 ill_dlpi_done(ill, DL_NOTIFY_REQ);
8211 8212 log = B_FALSE;
8212 8213 break;
8213 8214 case DL_PHYS_ADDR_REQ:
8214 8215 /*
8215 8216 * For IPv6 only, there are two additional
8216 8217 * phys_addr_req's sent to the driver to get the
8217 8218 * IPv6 token and lla. This allows IP to acquire
8218 8219 * the hardware address format for a given interface
8219 8220 * without having built in knowledge of the hardware
8220 8221 * address. ill_phys_addr_pend keeps track of the last
8221 8222 * DL_PAR sent so we know which response we are
8222 8223 * dealing with. ill_dlpi_done will update
8223 8224 * ill_phys_addr_pend when it sends the next req.
8224 8225 * We don't complete the IOCTL until all three DL_PARs
8225 8226 * have been attempted, so set *_len to 0 and break.
8226 8227 */
8227 8228 paddrreq = ill->ill_phys_addr_pend;
8228 8229 ill_dlpi_done(ill, DL_PHYS_ADDR_REQ);
8229 8230 if (paddrreq == DL_IPV6_TOKEN) {
8230 8231 ill->ill_token_length = 0;
8231 8232 log = B_FALSE;
8232 8233 break;
8233 8234 } else if (paddrreq == DL_IPV6_LINK_LAYER_ADDR) {
8234 8235 ill->ill_nd_lla_len = 0;
8235 8236 log = B_FALSE;
8236 8237 break;
8237 8238 }
8238 8239 /*
8239 8240 * Something went wrong with the DL_PHYS_ADDR_REQ.
8240 8241 * We presumably have an IOCTL hanging out waiting
8241 8242 * for completion. Find it and complete the IOCTL
8242 8243 * with the error noted.
8243 8244 * However, ill_dl_phys was called on an ill queue
8244 8245 * (from SIOCSLIFNAME), thus conn_pending_ill is not
8245 8246 * set. But the ioctl is known to be pending on ill_wq.
8246 8247 */
8247 8248 if (!ill->ill_ifname_pending)
8248 8249 break;
8249 8250 ill->ill_ifname_pending = 0;
8250 8251 if (!ioctl_aborted)
8251 8252 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8252 8253 if (mp1 != NULL) {
8253 8254 /*
8254 8255 * This operation (SIOCSLIFNAME) must have
8255 8256 * happened on the ill. Assert there is no conn
8256 8257 */
8257 8258 ASSERT(connp == NULL);
8258 8259 q = ill->ill_wq;
8259 8260 }
8260 8261 break;
8261 8262 case DL_BIND_REQ:
8262 8263 ill_dlpi_done(ill, DL_BIND_REQ);
8263 8264 if (ill->ill_ifname_pending)
8264 8265 break;
8265 8266 mutex_enter(&ill->ill_lock);
8266 8267 ill->ill_state_flags &= ~ILL_DOWN_IN_PROGRESS;
8267 8268 mutex_exit(&ill->ill_lock);
8268 8269 /*
8269 8270 * Something went wrong with the bind. We presumably
8270 8271 * have an IOCTL hanging out waiting for completion.
8271 8272 * Find it, take down the interface that was coming
8272 8273 * up, and complete the IOCTL with the error noted.
8273 8274 */
8274 8275 if (!ioctl_aborted)
8275 8276 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8276 8277 if (mp1 != NULL) {
8277 8278 /*
8278 8279 * This might be a result of a DL_NOTE_REPLUMB
8279 8280 * notification. In that case, connp is NULL.
8280 8281 */
8281 8282 if (connp != NULL)
8282 8283 q = CONNP_TO_WQ(connp);
8283 8284
8284 8285 (void) ipif_down(ipif, NULL, NULL);
8285 8286 /* error is set below the switch */
8286 8287 }
8287 8288 break;
8288 8289 case DL_ENABMULTI_REQ:
8289 8290 ill_dlpi_done(ill, DL_ENABMULTI_REQ);
8290 8291
8291 8292 if (ill->ill_dlpi_multicast_state == IDS_INPROGRESS)
8292 8293 ill->ill_dlpi_multicast_state = IDS_FAILED;
8293 8294 if (ill->ill_dlpi_multicast_state == IDS_FAILED) {
8294 8295
8295 8296 printf("ip: joining multicasts failed (%d)"
8296 8297 " on %s - will use link layer "
8297 8298 "broadcasts for multicast\n",
8298 8299 dlea->dl_errno, ill->ill_name);
8299 8300
8300 8301 /*
8301 8302 * Set up for multi_bcast; We are the
8302 8303 * writer, so ok to access ill->ill_ipif
8303 8304 * without any lock.
8304 8305 */
8305 8306 mutex_enter(&ill->ill_phyint->phyint_lock);
8306 8307 ill->ill_phyint->phyint_flags |=
8307 8308 PHYI_MULTI_BCAST;
8308 8309 mutex_exit(&ill->ill_phyint->phyint_lock);
8309 8310
8310 8311 }
8311 8312 freemsg(mp); /* Don't want to pass this up */
8312 8313 return;
8313 8314 case DL_CAPABILITY_REQ:
8314 8315 ip1dbg(("ip_rput_dlpi_writer: got DL_ERROR_ACK for "
8315 8316 "DL_CAPABILITY REQ\n"));
8316 8317 if (ill->ill_dlpi_capab_state == IDCS_PROBE_SENT)
8317 8318 ill->ill_dlpi_capab_state = IDCS_FAILED;
8318 8319 ill_capability_done(ill);
8319 8320 freemsg(mp);
8320 8321 return;
8321 8322 }
8322 8323 /*
8323 8324 * Note the error for IOCTL completion (mp1 is set when
8324 8325 * ready to complete ioctl). If ill_ifname_pending_err is
8325 8326 * set, an error occured during plumbing (ill_ifname_pending),
8326 8327 * so we want to report that error.
8327 8328 *
8328 8329 * NOTE: there are two addtional DL_PHYS_ADDR_REQ's
8329 8330 * (DL_IPV6_TOKEN and DL_IPV6_LINK_LAYER_ADDR) that are
8330 8331 * expected to get errack'd if the driver doesn't support
8331 8332 * these flags (e.g. ethernet). log will be set to B_FALSE
8332 8333 * if these error conditions are encountered.
8333 8334 */
8334 8335 if (mp1 != NULL) {
8335 8336 if (ill->ill_ifname_pending_err != 0) {
8336 8337 err = ill->ill_ifname_pending_err;
8337 8338 ill->ill_ifname_pending_err = 0;
8338 8339 } else {
8339 8340 err = dlea->dl_unix_errno ?
8340 8341 dlea->dl_unix_errno : ENXIO;
8341 8342 }
8342 8343 /*
8343 8344 * If we're plumbing an interface and an error hasn't already
8344 8345 * been saved, set ill_ifname_pending_err to the error passed
8345 8346 * up. Ignore the error if log is B_FALSE (see comment above).
8346 8347 */
8347 8348 } else if (log && ill->ill_ifname_pending &&
8348 8349 ill->ill_ifname_pending_err == 0) {
8349 8350 ill->ill_ifname_pending_err = dlea->dl_unix_errno ?
8350 8351 dlea->dl_unix_errno : ENXIO;
8351 8352 }
8352 8353
8353 8354 if (log)
8354 8355 ip_dlpi_error(ill, dlea->dl_error_primitive,
8355 8356 dlea->dl_errno, dlea->dl_unix_errno);
8356 8357 break;
8357 8358 case DL_CAPABILITY_ACK:
8358 8359 ill_capability_ack(ill, mp);
8359 8360 /*
8360 8361 * The message has been handed off to ill_capability_ack
8361 8362 * and must not be freed below
8362 8363 */
8363 8364 mp = NULL;
8364 8365 break;
8365 8366
8366 8367 case DL_INFO_ACK:
8367 8368 /* Call a routine to handle this one. */
8368 8369 ill_dlpi_done(ill, DL_INFO_REQ);
8369 8370 ip_ll_subnet_defaults(ill, mp);
8370 8371 ASSERT(!MUTEX_HELD(&ill->ill_phyint->phyint_ipsq->ipsq_lock));
8371 8372 return;
8372 8373 case DL_BIND_ACK:
8373 8374 /*
8374 8375 * We should have an IOCTL waiting on this unless
8375 8376 * sent by ill_dl_phys, in which case just return
8376 8377 */
8377 8378 ill_dlpi_done(ill, DL_BIND_REQ);
8378 8379
8379 8380 if (ill->ill_ifname_pending) {
8380 8381 DTRACE_PROBE2(ip__rput__dlpi__ifname__pending,
8381 8382 ill_t *, ill, mblk_t *, mp);
8382 8383 break;
8383 8384 }
8384 8385 mutex_enter(&ill->ill_lock);
8385 8386 ill->ill_dl_up = 1;
8386 8387 ill->ill_state_flags &= ~ILL_DOWN_IN_PROGRESS;
8387 8388 mutex_exit(&ill->ill_lock);
8388 8389
8389 8390 if (!ioctl_aborted)
8390 8391 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8391 8392 if (mp1 == NULL) {
8392 8393 DTRACE_PROBE1(ip__rput__dlpi__no__mblk, ill_t *, ill);
8393 8394 break;
8394 8395 }
8395 8396 /*
8396 8397 * mp1 was added by ill_dl_up(). if that is a result of
8397 8398 * a DL_NOTE_REPLUMB notification, connp could be NULL.
8398 8399 */
8399 8400 if (connp != NULL)
8400 8401 q = CONNP_TO_WQ(connp);
8401 8402 /*
8402 8403 * We are exclusive. So nothing can change even after
8403 8404 * we get the pending mp.
8404 8405 */
8405 8406 ip1dbg(("ip_rput_dlpi: bind_ack %s\n", ill->ill_name));
8406 8407 DTRACE_PROBE1(ip__rput__dlpi__bind__ack, ill_t *, ill);
8407 8408 ill_nic_event_dispatch(ill, 0, NE_UP, NULL, 0);
8408 8409
8409 8410 /*
8410 8411 * Now bring up the resolver; when that is complete, we'll
8411 8412 * create IREs. Note that we intentionally mirror what
8412 8413 * ipif_up() would have done, because we got here by way of
8413 8414 * ill_dl_up(), which stopped ipif_up()'s processing.
8414 8415 */
8415 8416 if (ill->ill_isv6) {
8416 8417 /*
8417 8418 * v6 interfaces.
8418 8419 * Unlike ARP which has to do another bind
8419 8420 * and attach, once we get here we are
8420 8421 * done with NDP
8421 8422 */
8422 8423 (void) ipif_resolver_up(ipif, Res_act_initial);
8423 8424 if ((err = ipif_ndp_up(ipif, B_TRUE)) == 0)
8424 8425 err = ipif_up_done_v6(ipif);
8425 8426 } else if (ill->ill_net_type == IRE_IF_RESOLVER) {
8426 8427 /*
8427 8428 * ARP and other v4 external resolvers.
8428 8429 * Leave the pending mblk intact so that
8429 8430 * the ioctl completes in ip_rput().
8430 8431 */
8431 8432 if (connp != NULL)
8432 8433 mutex_enter(&connp->conn_lock);
8433 8434 mutex_enter(&ill->ill_lock);
8434 8435 success = ipsq_pending_mp_add(connp, ipif, q, mp1, 0);
8435 8436 mutex_exit(&ill->ill_lock);
8436 8437 if (connp != NULL)
8437 8438 mutex_exit(&connp->conn_lock);
8438 8439 if (success) {
8439 8440 err = ipif_resolver_up(ipif, Res_act_initial);
8440 8441 if (err == EINPROGRESS) {
8441 8442 freemsg(mp);
8442 8443 return;
8443 8444 }
8444 8445 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8445 8446 } else {
8446 8447 /* The conn has started closing */
8447 8448 err = EINTR;
8448 8449 }
8449 8450 } else {
8450 8451 /*
8451 8452 * This one is complete. Reply to pending ioctl.
8452 8453 */
8453 8454 (void) ipif_resolver_up(ipif, Res_act_initial);
8454 8455 err = ipif_up_done(ipif);
8455 8456 }
8456 8457
8457 8458 if ((err == 0) && (ill->ill_up_ipifs)) {
8458 8459 err = ill_up_ipifs(ill, q, mp1);
8459 8460 if (err == EINPROGRESS) {
8460 8461 freemsg(mp);
8461 8462 return;
8462 8463 }
8463 8464 }
8464 8465
8465 8466 /*
8466 8467 * If we have a moved ipif to bring up, and everything has
8467 8468 * succeeded to this point, bring it up on the IPMP ill.
8468 8469 * Otherwise, leave it down -- the admin can try to bring it
8469 8470 * up by hand if need be.
8470 8471 */
8471 8472 if (ill->ill_move_ipif != NULL) {
8472 8473 if (err != 0) {
8473 8474 ill->ill_move_ipif = NULL;
8474 8475 } else {
8475 8476 ipif = ill->ill_move_ipif;
8476 8477 ill->ill_move_ipif = NULL;
8477 8478 err = ipif_up(ipif, q, mp1);
8478 8479 if (err == EINPROGRESS) {
8479 8480 freemsg(mp);
8480 8481 return;
8481 8482 }
8482 8483 }
8483 8484 }
8484 8485 break;
8485 8486
8486 8487 case DL_NOTIFY_IND: {
8487 8488 dl_notify_ind_t *notify = (dl_notify_ind_t *)mp->b_rptr;
8488 8489 uint_t orig_mtu, orig_mc_mtu;
8489 8490
8490 8491 switch (notify->dl_notification) {
8491 8492 case DL_NOTE_PHYS_ADDR:
8492 8493 err = ill_set_phys_addr(ill, mp);
8493 8494 break;
8494 8495
8495 8496 case DL_NOTE_REPLUMB:
8496 8497 /*
8497 8498 * Directly return after calling ill_replumb().
8498 8499 * Note that we should not free mp as it is reused
8499 8500 * in the ill_replumb() function.
8500 8501 */
8501 8502 err = ill_replumb(ill, mp);
8502 8503 return;
8503 8504
8504 8505 case DL_NOTE_FASTPATH_FLUSH:
8505 8506 nce_flush(ill, B_FALSE);
8506 8507 break;
8507 8508
8508 8509 case DL_NOTE_SDU_SIZE:
8509 8510 case DL_NOTE_SDU_SIZE2:
8510 8511 /*
8511 8512 * The dce and fragmentation code can cope with
8512 8513 * this changing while packets are being sent.
8513 8514 * When packets are sent ip_output will discover
8514 8515 * a change.
8515 8516 *
8516 8517 * Change the MTU size of the interface.
8517 8518 */
8518 8519 mutex_enter(&ill->ill_lock);
8519 8520 orig_mtu = ill->ill_mtu;
8520 8521 orig_mc_mtu = ill->ill_mc_mtu;
8521 8522 switch (notify->dl_notification) {
8522 8523 case DL_NOTE_SDU_SIZE:
8523 8524 ill->ill_current_frag =
8524 8525 (uint_t)notify->dl_data;
8525 8526 ill->ill_mc_mtu = (uint_t)notify->dl_data;
8526 8527 break;
8527 8528 case DL_NOTE_SDU_SIZE2:
8528 8529 ill->ill_current_frag =
8529 8530 (uint_t)notify->dl_data1;
8530 8531 ill->ill_mc_mtu = (uint_t)notify->dl_data2;
8531 8532 break;
8532 8533 }
8533 8534 if (ill->ill_current_frag > ill->ill_max_frag)
8534 8535 ill->ill_max_frag = ill->ill_current_frag;
8535 8536
8536 8537 if (!(ill->ill_flags & ILLF_FIXEDMTU)) {
8537 8538 ill->ill_mtu = ill->ill_current_frag;
8538 8539
8539 8540 /*
8540 8541 * If ill_user_mtu was set (via
8541 8542 * SIOCSLIFLNKINFO), clamp ill_mtu at it.
8542 8543 */
8543 8544 if (ill->ill_user_mtu != 0 &&
8544 8545 ill->ill_user_mtu < ill->ill_mtu)
8545 8546 ill->ill_mtu = ill->ill_user_mtu;
8546 8547
8547 8548 if (ill->ill_user_mtu != 0 &&
8548 8549 ill->ill_user_mtu < ill->ill_mc_mtu)
8549 8550 ill->ill_mc_mtu = ill->ill_user_mtu;
8550 8551
8551 8552 if (ill->ill_isv6) {
8552 8553 if (ill->ill_mtu < IPV6_MIN_MTU)
8553 8554 ill->ill_mtu = IPV6_MIN_MTU;
8554 8555 if (ill->ill_mc_mtu < IPV6_MIN_MTU)
8555 8556 ill->ill_mc_mtu = IPV6_MIN_MTU;
8556 8557 } else {
8557 8558 if (ill->ill_mtu < IP_MIN_MTU)
8558 8559 ill->ill_mtu = IP_MIN_MTU;
8559 8560 if (ill->ill_mc_mtu < IP_MIN_MTU)
8560 8561 ill->ill_mc_mtu = IP_MIN_MTU;
8561 8562 }
8562 8563 } else if (ill->ill_mc_mtu > ill->ill_mtu) {
8563 8564 ill->ill_mc_mtu = ill->ill_mtu;
8564 8565 }
8565 8566
8566 8567 mutex_exit(&ill->ill_lock);
8567 8568 /*
8568 8569 * Make sure all dce_generation checks find out
8569 8570 * that ill_mtu/ill_mc_mtu has changed.
8570 8571 */
8571 8572 if (orig_mtu != ill->ill_mtu ||
8572 8573 orig_mc_mtu != ill->ill_mc_mtu) {
8573 8574 dce_increment_all_generations(ill->ill_isv6,
8574 8575 ill->ill_ipst);
8575 8576 }
8576 8577
8577 8578 /*
8578 8579 * Refresh IPMP meta-interface MTU if necessary.
8579 8580 */
8580 8581 if (IS_UNDER_IPMP(ill))
8581 8582 ipmp_illgrp_refresh_mtu(ill->ill_grp);
8582 8583 break;
8583 8584
8584 8585 case DL_NOTE_LINK_UP:
8585 8586 case DL_NOTE_LINK_DOWN: {
8586 8587 /*
8587 8588 * We are writer. ill / phyint / ipsq assocs stable.
8588 8589 * The RUNNING flag reflects the state of the link.
8589 8590 */
8590 8591 phyint_t *phyint = ill->ill_phyint;
8591 8592 uint64_t new_phyint_flags;
8592 8593 boolean_t changed = B_FALSE;
8593 8594 boolean_t went_up;
8594 8595
8595 8596 went_up = notify->dl_notification == DL_NOTE_LINK_UP;
8596 8597 mutex_enter(&phyint->phyint_lock);
8597 8598
8598 8599 new_phyint_flags = went_up ?
8599 8600 phyint->phyint_flags | PHYI_RUNNING :
8600 8601 phyint->phyint_flags & ~PHYI_RUNNING;
8601 8602
8602 8603 if (IS_IPMP(ill)) {
8603 8604 new_phyint_flags = went_up ?
8604 8605 new_phyint_flags & ~PHYI_FAILED :
8605 8606 new_phyint_flags | PHYI_FAILED;
8606 8607 }
8607 8608
8608 8609 if (new_phyint_flags != phyint->phyint_flags) {
8609 8610 phyint->phyint_flags = new_phyint_flags;
8610 8611 changed = B_TRUE;
8611 8612 }
8612 8613 mutex_exit(&phyint->phyint_lock);
8613 8614 /*
8614 8615 * ill_restart_dad handles the DAD restart and routing
8615 8616 * socket notification logic.
8616 8617 */
8617 8618 if (changed) {
8618 8619 ill_restart_dad(phyint->phyint_illv4, went_up);
8619 8620 ill_restart_dad(phyint->phyint_illv6, went_up);
8620 8621 }
8621 8622 break;
8622 8623 }
8623 8624 case DL_NOTE_PROMISC_ON_PHYS: {
8624 8625 phyint_t *phyint = ill->ill_phyint;
8625 8626
8626 8627 mutex_enter(&phyint->phyint_lock);
8627 8628 phyint->phyint_flags |= PHYI_PROMISC;
8628 8629 mutex_exit(&phyint->phyint_lock);
8629 8630 break;
8630 8631 }
8631 8632 case DL_NOTE_PROMISC_OFF_PHYS: {
8632 8633 phyint_t *phyint = ill->ill_phyint;
8633 8634
8634 8635 mutex_enter(&phyint->phyint_lock);
8635 8636 phyint->phyint_flags &= ~PHYI_PROMISC;
8636 8637 mutex_exit(&phyint->phyint_lock);
8637 8638 break;
8638 8639 }
8639 8640 case DL_NOTE_CAPAB_RENEG:
8640 8641 /*
8641 8642 * Something changed on the driver side.
8642 8643 * It wants us to renegotiate the capabilities
8643 8644 * on this ill. One possible cause is the aggregation
8644 8645 * interface under us where a port got added or
8645 8646 * went away.
8646 8647 *
8647 8648 * If the capability negotiation is already done
8648 8649 * or is in progress, reset the capabilities and
8649 8650 * mark the ill's ill_capab_reneg to be B_TRUE,
8650 8651 * so that when the ack comes back, we can start
8651 8652 * the renegotiation process.
8652 8653 *
8653 8654 * Note that if ill_capab_reneg is already B_TRUE
8654 8655 * (ill_dlpi_capab_state is IDS_UNKNOWN in this case),
8655 8656 * the capability resetting request has been sent
8656 8657 * and the renegotiation has not been started yet;
8657 8658 * nothing needs to be done in this case.
8658 8659 */
8659 8660 ipsq_current_start(ipsq, ill->ill_ipif, 0);
8660 8661 ill_capability_reset(ill, B_TRUE);
8661 8662 ipsq_current_finish(ipsq);
8662 8663 break;
8663 8664
8664 8665 case DL_NOTE_ALLOWED_IPS:
8665 8666 ill_set_allowed_ips(ill, mp);
8666 8667 break;
8667 8668 default:
8668 8669 ip0dbg(("ip_rput_dlpi_writer: unknown notification "
8669 8670 "type 0x%x for DL_NOTIFY_IND\n",
8670 8671 notify->dl_notification));
8671 8672 break;
8672 8673 }
8673 8674
8674 8675 /*
8675 8676 * As this is an asynchronous operation, we
8676 8677 * should not call ill_dlpi_done
8677 8678 */
8678 8679 break;
8679 8680 }
8680 8681 case DL_NOTIFY_ACK: {
8681 8682 dl_notify_ack_t *noteack = (dl_notify_ack_t *)mp->b_rptr;
8682 8683
8683 8684 if (noteack->dl_notifications & DL_NOTE_LINK_UP)
8684 8685 ill->ill_note_link = 1;
8685 8686 ill_dlpi_done(ill, DL_NOTIFY_REQ);
8686 8687 break;
8687 8688 }
8688 8689 case DL_PHYS_ADDR_ACK: {
8689 8690 /*
8690 8691 * As part of plumbing the interface via SIOCSLIFNAME,
8691 8692 * ill_dl_phys() will queue a series of DL_PHYS_ADDR_REQs,
8692 8693 * whose answers we receive here. As each answer is received,
8693 8694 * we call ill_dlpi_done() to dispatch the next request as
8694 8695 * we're processing the current one. Once all answers have
8695 8696 * been received, we use ipsq_pending_mp_get() to dequeue the
8696 8697 * outstanding IOCTL and reply to it. (Because ill_dl_phys()
8697 8698 * is invoked from an ill queue, conn_oper_pending_ill is not
8698 8699 * available, but we know the ioctl is pending on ill_wq.)
8699 8700 */
8700 8701 uint_t paddrlen, paddroff;
8701 8702 uint8_t *addr;
8702 8703
8703 8704 paddrreq = ill->ill_phys_addr_pend;
8704 8705 paddrlen = ((dl_phys_addr_ack_t *)mp->b_rptr)->dl_addr_length;
8705 8706 paddroff = ((dl_phys_addr_ack_t *)mp->b_rptr)->dl_addr_offset;
8706 8707 addr = mp->b_rptr + paddroff;
8707 8708
8708 8709 ill_dlpi_done(ill, DL_PHYS_ADDR_REQ);
8709 8710 if (paddrreq == DL_IPV6_TOKEN) {
8710 8711 /*
8711 8712 * bcopy to low-order bits of ill_token
8712 8713 *
8713 8714 * XXX Temporary hack - currently, all known tokens
8714 8715 * are 64 bits, so I'll cheat for the moment.
8715 8716 */
8716 8717 bcopy(addr, &ill->ill_token.s6_addr32[2], paddrlen);
8717 8718 ill->ill_token_length = paddrlen;
8718 8719 break;
8719 8720 } else if (paddrreq == DL_IPV6_LINK_LAYER_ADDR) {
8720 8721 ASSERT(ill->ill_nd_lla_mp == NULL);
8721 8722 ill_set_ndmp(ill, mp, paddroff, paddrlen);
8722 8723 mp = NULL;
8723 8724 break;
8724 8725 } else if (paddrreq == DL_CURR_DEST_ADDR) {
8725 8726 ASSERT(ill->ill_dest_addr_mp == NULL);
8726 8727 ill->ill_dest_addr_mp = mp;
8727 8728 ill->ill_dest_addr = addr;
8728 8729 mp = NULL;
8729 8730 if (ill->ill_isv6) {
8730 8731 ill_setdesttoken(ill);
8731 8732 ipif_setdestlinklocal(ill->ill_ipif);
8732 8733 }
8733 8734 break;
8734 8735 }
8735 8736
8736 8737 ASSERT(paddrreq == DL_CURR_PHYS_ADDR);
8737 8738 ASSERT(ill->ill_phys_addr_mp == NULL);
8738 8739 if (!ill->ill_ifname_pending)
8739 8740 break;
8740 8741 ill->ill_ifname_pending = 0;
8741 8742 if (!ioctl_aborted)
8742 8743 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8743 8744 if (mp1 != NULL) {
8744 8745 ASSERT(connp == NULL);
8745 8746 q = ill->ill_wq;
8746 8747 }
8747 8748 /*
8748 8749 * If any error acks received during the plumbing sequence,
8749 8750 * ill_ifname_pending_err will be set. Break out and send up
8750 8751 * the error to the pending ioctl.
8751 8752 */
8752 8753 if (ill->ill_ifname_pending_err != 0) {
8753 8754 err = ill->ill_ifname_pending_err;
8754 8755 ill->ill_ifname_pending_err = 0;
8755 8756 break;
8756 8757 }
8757 8758
8758 8759 ill->ill_phys_addr_mp = mp;
8759 8760 ill->ill_phys_addr = (paddrlen == 0 ? NULL : addr);
8760 8761 mp = NULL;
8761 8762
8762 8763 /*
8763 8764 * If paddrlen or ill_phys_addr_length is zero, the DLPI
8764 8765 * provider doesn't support physical addresses. We check both
8765 8766 * paddrlen and ill_phys_addr_length because sppp (PPP) does
8766 8767 * not have physical addresses, but historically adversises a
8767 8768 * physical address length of 0 in its DL_INFO_ACK, but 6 in
8768 8769 * its DL_PHYS_ADDR_ACK.
8769 8770 */
8770 8771 if (paddrlen == 0 || ill->ill_phys_addr_length == 0) {
8771 8772 ill->ill_phys_addr = NULL;
8772 8773 } else if (paddrlen != ill->ill_phys_addr_length) {
8773 8774 ip0dbg(("DL_PHYS_ADDR_ACK: got addrlen %d, expected %d",
8774 8775 paddrlen, ill->ill_phys_addr_length));
8775 8776 err = EINVAL;
8776 8777 break;
8777 8778 }
8778 8779
8779 8780 if (ill->ill_nd_lla_mp == NULL) {
8780 8781 if ((mp_hw = copyb(ill->ill_phys_addr_mp)) == NULL) {
8781 8782 err = ENOMEM;
8782 8783 break;
8783 8784 }
8784 8785 ill_set_ndmp(ill, mp_hw, paddroff, paddrlen);
8785 8786 }
8786 8787
8787 8788 if (ill->ill_isv6) {
8788 8789 ill_setdefaulttoken(ill);
8789 8790 ipif_setlinklocal(ill->ill_ipif);
8790 8791 }
8791 8792 break;
8792 8793 }
8793 8794 case DL_OK_ACK:
8794 8795 ip2dbg(("DL_OK_ACK %s (0x%x)\n",
8795 8796 dl_primstr((int)dloa->dl_correct_primitive),
8796 8797 dloa->dl_correct_primitive));
8797 8798 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer ok",
8798 8799 char *, dl_primstr(dloa->dl_correct_primitive),
8799 8800 ill_t *, ill);
8800 8801
8801 8802 switch (dloa->dl_correct_primitive) {
8802 8803 case DL_ENABMULTI_REQ:
8803 8804 case DL_DISABMULTI_REQ:
8804 8805 ill_dlpi_done(ill, dloa->dl_correct_primitive);
8805 8806 break;
8806 8807 case DL_PROMISCON_REQ:
8807 8808 case DL_PROMISCOFF_REQ:
8808 8809 case DL_UNBIND_REQ:
8809 8810 case DL_ATTACH_REQ:
8810 8811 ill_dlpi_done(ill, dloa->dl_correct_primitive);
8811 8812 break;
8812 8813 }
8813 8814 break;
8814 8815 default:
8815 8816 break;
8816 8817 }
8817 8818
8818 8819 freemsg(mp);
8819 8820 if (mp1 == NULL)
8820 8821 return;
8821 8822
8822 8823 /*
8823 8824 * The operation must complete without EINPROGRESS since
8824 8825 * ipsq_pending_mp_get() has removed the mblk (mp1). Otherwise,
8825 8826 * the operation will be stuck forever inside the IPSQ.
8826 8827 */
8827 8828 ASSERT(err != EINPROGRESS);
8828 8829
8829 8830 DTRACE_PROBE4(ipif__ioctl, char *, "ip_rput_dlpi_writer finish",
8830 8831 int, ipsq->ipsq_xop->ipx_current_ioctl, ill_t *, ill,
8831 8832 ipif_t *, NULL);
8832 8833
8833 8834 switch (ipsq->ipsq_xop->ipx_current_ioctl) {
8834 8835 case 0:
8835 8836 ipsq_current_finish(ipsq);
8836 8837 break;
8837 8838
8838 8839 case SIOCSLIFNAME:
8839 8840 case IF_UNITSEL: {
8840 8841 ill_t *ill_other = ILL_OTHER(ill);
8841 8842
8842 8843 /*
8843 8844 * If SIOCSLIFNAME or IF_UNITSEL is about to succeed, and the
8844 8845 * ill has a peer which is in an IPMP group, then place ill
8845 8846 * into the same group. One catch: although ifconfig plumbs
8846 8847 * the appropriate IPMP meta-interface prior to plumbing this
8847 8848 * ill, it is possible for multiple ifconfig applications to
8848 8849 * race (or for another application to adjust plumbing), in
8849 8850 * which case the IPMP meta-interface we need will be missing.
8850 8851 * If so, kick the phyint out of the group.
8851 8852 */
8852 8853 if (err == 0 && ill_other != NULL && IS_UNDER_IPMP(ill_other)) {
8853 8854 ipmp_grp_t *grp = ill->ill_phyint->phyint_grp;
8854 8855 ipmp_illgrp_t *illg;
8855 8856
8856 8857 illg = ill->ill_isv6 ? grp->gr_v6 : grp->gr_v4;
8857 8858 if (illg == NULL)
8858 8859 ipmp_phyint_leave_grp(ill->ill_phyint);
8859 8860 else
8860 8861 ipmp_ill_join_illgrp(ill, illg);
8861 8862 }
8862 8863
8863 8864 if (ipsq->ipsq_xop->ipx_current_ioctl == IF_UNITSEL)
8864 8865 ip_ioctl_finish(q, mp1, err, NO_COPYOUT, ipsq);
8865 8866 else
8866 8867 ip_ioctl_finish(q, mp1, err, COPYOUT, ipsq);
8867 8868 break;
8868 8869 }
8869 8870 case SIOCLIFADDIF:
8870 8871 ip_ioctl_finish(q, mp1, err, COPYOUT, ipsq);
8871 8872 break;
8872 8873
8873 8874 default:
8874 8875 ip_ioctl_finish(q, mp1, err, NO_COPYOUT, ipsq);
8875 8876 break;
8876 8877 }
8877 8878 }
8878 8879
8879 8880 /*
8880 8881 * ip_rput_other is called by ip_rput to handle messages modifying the global
8881 8882 * state in IP. If 'ipsq' is non-NULL, caller is writer on it.
8882 8883 */
8883 8884 /* ARGSUSED */
8884 8885 void
8885 8886 ip_rput_other(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
8886 8887 {
8887 8888 ill_t *ill = q->q_ptr;
8888 8889 struct iocblk *iocp;
8889 8890
8890 8891 ip1dbg(("ip_rput_other "));
8891 8892 if (ipsq != NULL) {
8892 8893 ASSERT(IAM_WRITER_IPSQ(ipsq));
8893 8894 ASSERT(ipsq->ipsq_xop ==
8894 8895 ill->ill_phyint->phyint_ipsq->ipsq_xop);
8895 8896 }
8896 8897
8897 8898 switch (mp->b_datap->db_type) {
8898 8899 case M_ERROR:
8899 8900 case M_HANGUP:
8900 8901 /*
8901 8902 * The device has a problem. We force the ILL down. It can
8902 8903 * be brought up again manually using SIOCSIFFLAGS (via
8903 8904 * ifconfig or equivalent).
8904 8905 */
8905 8906 ASSERT(ipsq != NULL);
8906 8907 if (mp->b_rptr < mp->b_wptr)
8907 8908 ill->ill_error = (int)(*mp->b_rptr & 0xFF);
8908 8909 if (ill->ill_error == 0)
8909 8910 ill->ill_error = ENXIO;
8910 8911 if (!ill_down_start(q, mp))
8911 8912 return;
8912 8913 ipif_all_down_tail(ipsq, q, mp, NULL);
8913 8914 break;
8914 8915 case M_IOCNAK: {
8915 8916 iocp = (struct iocblk *)mp->b_rptr;
8916 8917
8917 8918 ASSERT(iocp->ioc_cmd == DL_IOC_HDR_INFO);
8918 8919 /*
8919 8920 * If this was the first attempt, turn off the fastpath
8920 8921 * probing.
8921 8922 */
8922 8923 mutex_enter(&ill->ill_lock);
8923 8924 if (ill->ill_dlpi_fastpath_state == IDS_INPROGRESS) {
8924 8925 ill->ill_dlpi_fastpath_state = IDS_FAILED;
8925 8926 mutex_exit(&ill->ill_lock);
8926 8927 /*
8927 8928 * don't flush the nce_t entries: we use them
8928 8929 * as an index to the ncec itself.
8929 8930 */
8930 8931 ip1dbg(("ip_rput: DLPI fastpath off on interface %s\n",
8931 8932 ill->ill_name));
8932 8933 } else {
8933 8934 mutex_exit(&ill->ill_lock);
8934 8935 }
8935 8936 freemsg(mp);
8936 8937 break;
8937 8938 }
8938 8939 default:
8939 8940 ASSERT(0);
8940 8941 break;
8941 8942 }
8942 8943 }
8943 8944
8944 8945 /*
8945 8946 * Update any source route, record route or timestamp options
8946 8947 * When it fails it has consumed the message and BUMPed the MIB.
8947 8948 */
8948 8949 boolean_t
8949 8950 ip_forward_options(mblk_t *mp, ipha_t *ipha, ill_t *dst_ill,
8950 8951 ip_recv_attr_t *ira)
8951 8952 {
8952 8953 ipoptp_t opts;
8953 8954 uchar_t *opt;
8954 8955 uint8_t optval;
8955 8956 uint8_t optlen;
8956 8957 ipaddr_t dst;
8957 8958 ipaddr_t ifaddr;
8958 8959 uint32_t ts;
8959 8960 timestruc_t now;
8960 8961 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
8961 8962
8962 8963 ip2dbg(("ip_forward_options\n"));
8963 8964 dst = ipha->ipha_dst;
8964 8965 for (optval = ipoptp_first(&opts, ipha);
8965 8966 optval != IPOPT_EOL;
8966 8967 optval = ipoptp_next(&opts)) {
8967 8968 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
8968 8969 opt = opts.ipoptp_cur;
8969 8970 optlen = opts.ipoptp_len;
8970 8971 ip2dbg(("ip_forward_options: opt %d, len %d\n",
8971 8972 optval, opts.ipoptp_len));
8972 8973 switch (optval) {
8973 8974 uint32_t off;
8974 8975 case IPOPT_SSRR:
8975 8976 case IPOPT_LSRR:
8976 8977 /* Check if adminstratively disabled */
8977 8978 if (!ipst->ips_ip_forward_src_routed) {
8978 8979 BUMP_MIB(dst_ill->ill_ip_mib,
8979 8980 ipIfStatsForwProhibits);
8980 8981 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED",
8981 8982 mp, dst_ill);
8982 8983 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED,
8983 8984 ira);
8984 8985 return (B_FALSE);
8985 8986 }
8986 8987 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
8987 8988 /*
8988 8989 * Must be partial since ip_input_options
8989 8990 * checked for strict.
8990 8991 */
8991 8992 break;
8992 8993 }
8993 8994 off = opt[IPOPT_OFFSET];
8994 8995 off--;
8995 8996 redo_srr:
8996 8997 if (optlen < IP_ADDR_LEN ||
8997 8998 off > optlen - IP_ADDR_LEN) {
8998 8999 /* End of source route */
8999 9000 ip1dbg((
9000 9001 "ip_forward_options: end of SR\n"));
9001 9002 break;
9002 9003 }
9003 9004 /* Pick a reasonable address on the outbound if */
9004 9005 ASSERT(dst_ill != NULL);
9005 9006 if (ip_select_source_v4(dst_ill, INADDR_ANY, dst,
9006 9007 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9007 9008 NULL) != 0) {
9008 9009 /* No source! Shouldn't happen */
9009 9010 ifaddr = INADDR_ANY;
9010 9011 }
9011 9012 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9012 9013 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9013 9014 ip1dbg(("ip_forward_options: next hop 0x%x\n",
9014 9015 ntohl(dst)));
9015 9016
9016 9017 /*
9017 9018 * Check if our address is present more than
9018 9019 * once as consecutive hops in source route.
9019 9020 */
9020 9021 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
9021 9022 off += IP_ADDR_LEN;
9022 9023 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9023 9024 goto redo_srr;
9024 9025 }
9025 9026 ipha->ipha_dst = dst;
9026 9027 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9027 9028 break;
9028 9029 case IPOPT_RR:
9029 9030 off = opt[IPOPT_OFFSET];
9030 9031 off--;
9031 9032 if (optlen < IP_ADDR_LEN ||
9032 9033 off > optlen - IP_ADDR_LEN) {
9033 9034 /* No more room - ignore */
9034 9035 ip1dbg((
9035 9036 "ip_forward_options: end of RR\n"));
9036 9037 break;
9037 9038 }
9038 9039 /* Pick a reasonable address on the outbound if */
9039 9040 ASSERT(dst_ill != NULL);
9040 9041 if (ip_select_source_v4(dst_ill, INADDR_ANY, dst,
9041 9042 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9042 9043 NULL) != 0) {
9043 9044 /* No source! Shouldn't happen */
9044 9045 ifaddr = INADDR_ANY;
9045 9046 }
9046 9047 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9047 9048 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9048 9049 break;
9049 9050 case IPOPT_TS:
9050 9051 /* Insert timestamp if there is room */
9051 9052 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9052 9053 case IPOPT_TS_TSONLY:
9053 9054 off = IPOPT_TS_TIMELEN;
9054 9055 break;
9055 9056 case IPOPT_TS_PRESPEC:
9056 9057 case IPOPT_TS_PRESPEC_RFC791:
9057 9058 /* Verify that the address matched */
9058 9059 off = opt[IPOPT_OFFSET] - 1;
9059 9060 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9060 9061 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9061 9062 /* Not for us */
9062 9063 break;
9063 9064 }
9064 9065 /* FALLTHRU */
9065 9066 case IPOPT_TS_TSANDADDR:
9066 9067 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9067 9068 break;
9068 9069 default:
9069 9070 /*
9070 9071 * ip_*put_options should have already
9071 9072 * dropped this packet.
9072 9073 */
9073 9074 cmn_err(CE_PANIC, "ip_forward_options: "
9074 9075 "unknown IT - bug in ip_input_options?\n");
9075 9076 return (B_TRUE); /* Keep "lint" happy */
9076 9077 }
9077 9078 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
9078 9079 /* Increase overflow counter */
9079 9080 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
9080 9081 opt[IPOPT_POS_OV_FLG] =
9081 9082 (uint8_t)((opt[IPOPT_POS_OV_FLG] & 0x0F) |
9082 9083 (off << 4));
9083 9084 break;
9084 9085 }
9085 9086 off = opt[IPOPT_OFFSET] - 1;
9086 9087 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9087 9088 case IPOPT_TS_PRESPEC:
9088 9089 case IPOPT_TS_PRESPEC_RFC791:
9089 9090 case IPOPT_TS_TSANDADDR:
9090 9091 /* Pick a reasonable addr on the outbound if */
9091 9092 ASSERT(dst_ill != NULL);
9092 9093 if (ip_select_source_v4(dst_ill, INADDR_ANY,
9093 9094 dst, INADDR_ANY, ALL_ZONES, ipst, &ifaddr,
9094 9095 NULL, NULL) != 0) {
9095 9096 /* No source! Shouldn't happen */
9096 9097 ifaddr = INADDR_ANY;
9097 9098 }
9098 9099 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9099 9100 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9100 9101 /* FALLTHRU */
9101 9102 case IPOPT_TS_TSONLY:
9102 9103 off = opt[IPOPT_OFFSET] - 1;
9103 9104 /* Compute # of milliseconds since midnight */
9104 9105 gethrestime(&now);
9105 9106 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
9106 9107 now.tv_nsec / (NANOSEC / MILLISEC);
9107 9108 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
9108 9109 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
9109 9110 break;
9110 9111 }
9111 9112 break;
9112 9113 }
9113 9114 }
9114 9115 return (B_TRUE);
9115 9116 }
9116 9117
9117 9118 /*
9118 9119 * Call ill_frag_timeout to do garbage collection. ill_frag_timeout
9119 9120 * returns 'true' if there are still fragments left on the queue, in
9120 9121 * which case we restart the timer.
9121 9122 */
9122 9123 void
9123 9124 ill_frag_timer(void *arg)
9124 9125 {
9125 9126 ill_t *ill = (ill_t *)arg;
9126 9127 boolean_t frag_pending;
9127 9128 ip_stack_t *ipst = ill->ill_ipst;
9128 9129 time_t timeout;
9129 9130
9130 9131 mutex_enter(&ill->ill_lock);
9131 9132 ASSERT(!ill->ill_fragtimer_executing);
9132 9133 if (ill->ill_state_flags & ILL_CONDEMNED) {
9133 9134 ill->ill_frag_timer_id = 0;
9134 9135 mutex_exit(&ill->ill_lock);
9135 9136 return;
9136 9137 }
9137 9138 ill->ill_fragtimer_executing = 1;
9138 9139 mutex_exit(&ill->ill_lock);
9139 9140
9140 9141 timeout = (ill->ill_isv6 ? ipst->ips_ipv6_reassembly_timeout :
9141 9142 ipst->ips_ip_reassembly_timeout);
9142 9143
9143 9144 frag_pending = ill_frag_timeout(ill, timeout);
9144 9145
9145 9146 /*
9146 9147 * Restart the timer, if we have fragments pending or if someone
9147 9148 * wanted us to be scheduled again.
9148 9149 */
9149 9150 mutex_enter(&ill->ill_lock);
9150 9151 ill->ill_fragtimer_executing = 0;
9151 9152 ill->ill_frag_timer_id = 0;
9152 9153 if (frag_pending || ill->ill_fragtimer_needrestart)
9153 9154 ill_frag_timer_start(ill);
9154 9155 mutex_exit(&ill->ill_lock);
9155 9156 }
9156 9157
9157 9158 void
9158 9159 ill_frag_timer_start(ill_t *ill)
9159 9160 {
9160 9161 ip_stack_t *ipst = ill->ill_ipst;
9161 9162 clock_t timeo_ms;
9162 9163
9163 9164 ASSERT(MUTEX_HELD(&ill->ill_lock));
9164 9165
9165 9166 /* If the ill is closing or opening don't proceed */
9166 9167 if (ill->ill_state_flags & ILL_CONDEMNED)
9167 9168 return;
9168 9169
9169 9170 if (ill->ill_fragtimer_executing) {
9170 9171 /*
9171 9172 * ill_frag_timer is currently executing. Just record the
9172 9173 * the fact that we want the timer to be restarted.
9173 9174 * ill_frag_timer will post a timeout before it returns,
9174 9175 * ensuring it will be called again.
9175 9176 */
9176 9177 ill->ill_fragtimer_needrestart = 1;
9177 9178 return;
9178 9179 }
9179 9180
9180 9181 if (ill->ill_frag_timer_id == 0) {
9181 9182 timeo_ms = (ill->ill_isv6 ? ipst->ips_ipv6_reassembly_timeout :
9182 9183 ipst->ips_ip_reassembly_timeout) * SECONDS;
9183 9184
9184 9185 /*
9185 9186 * The timer is neither running nor is the timeout handler
9186 9187 * executing. Post a timeout so that ill_frag_timer will be
9187 9188 * called
9188 9189 */
9189 9190 ill->ill_frag_timer_id = timeout(ill_frag_timer, ill,
9190 9191 MSEC_TO_TICK(timeo_ms >> 1));
9191 9192 ill->ill_fragtimer_needrestart = 0;
9192 9193 }
9193 9194 }
9194 9195
9195 9196 /*
9196 9197 * Update any source route, record route or timestamp options.
9197 9198 * Check that we are at end of strict source route.
9198 9199 * The options have already been checked for sanity in ip_input_options().
9199 9200 */
9200 9201 boolean_t
9201 9202 ip_input_local_options(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
9202 9203 {
9203 9204 ipoptp_t opts;
9204 9205 uchar_t *opt;
9205 9206 uint8_t optval;
9206 9207 uint8_t optlen;
9207 9208 ipaddr_t dst;
9208 9209 ipaddr_t ifaddr;
9209 9210 uint32_t ts;
9210 9211 timestruc_t now;
9211 9212 ill_t *ill = ira->ira_ill;
9212 9213 ip_stack_t *ipst = ill->ill_ipst;
9213 9214
9214 9215 ip2dbg(("ip_input_local_options\n"));
9215 9216
9216 9217 for (optval = ipoptp_first(&opts, ipha);
9217 9218 optval != IPOPT_EOL;
9218 9219 optval = ipoptp_next(&opts)) {
9219 9220 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
9220 9221 opt = opts.ipoptp_cur;
9221 9222 optlen = opts.ipoptp_len;
9222 9223 ip2dbg(("ip_input_local_options: opt %d, len %d\n",
9223 9224 optval, optlen));
9224 9225 switch (optval) {
9225 9226 uint32_t off;
9226 9227 case IPOPT_SSRR:
9227 9228 case IPOPT_LSRR:
9228 9229 off = opt[IPOPT_OFFSET];
9229 9230 off--;
9230 9231 if (optlen < IP_ADDR_LEN ||
9231 9232 off > optlen - IP_ADDR_LEN) {
9232 9233 /* End of source route */
9233 9234 ip1dbg(("ip_input_local_options: end of SR\n"));
9234 9235 break;
9235 9236 }
9236 9237 /*
9237 9238 * This will only happen if two consecutive entries
9238 9239 * in the source route contains our address or if
9239 9240 * it is a packet with a loose source route which
9240 9241 * reaches us before consuming the whole source route
9241 9242 */
9242 9243 ip1dbg(("ip_input_local_options: not end of SR\n"));
9243 9244 if (optval == IPOPT_SSRR) {
9244 9245 goto bad_src_route;
9245 9246 }
9246 9247 /*
9247 9248 * Hack: instead of dropping the packet truncate the
9248 9249 * source route to what has been used by filling the
9249 9250 * rest with IPOPT_NOP.
9250 9251 */
9251 9252 opt[IPOPT_OLEN] = (uint8_t)off;
9252 9253 while (off < optlen) {
9253 9254 opt[off++] = IPOPT_NOP;
9254 9255 }
9255 9256 break;
9256 9257 case IPOPT_RR:
9257 9258 off = opt[IPOPT_OFFSET];
9258 9259 off--;
9259 9260 if (optlen < IP_ADDR_LEN ||
9260 9261 off > optlen - IP_ADDR_LEN) {
9261 9262 /* No more room - ignore */
9262 9263 ip1dbg((
9263 9264 "ip_input_local_options: end of RR\n"));
9264 9265 break;
9265 9266 }
9266 9267 /* Pick a reasonable address on the outbound if */
9267 9268 if (ip_select_source_v4(ill, INADDR_ANY, ipha->ipha_dst,
9268 9269 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9269 9270 NULL) != 0) {
9270 9271 /* No source! Shouldn't happen */
9271 9272 ifaddr = INADDR_ANY;
9272 9273 }
9273 9274 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9274 9275 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9275 9276 break;
9276 9277 case IPOPT_TS:
9277 9278 /* Insert timestamp if there is romm */
9278 9279 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9279 9280 case IPOPT_TS_TSONLY:
9280 9281 off = IPOPT_TS_TIMELEN;
9281 9282 break;
9282 9283 case IPOPT_TS_PRESPEC:
9283 9284 case IPOPT_TS_PRESPEC_RFC791:
9284 9285 /* Verify that the address matched */
9285 9286 off = opt[IPOPT_OFFSET] - 1;
9286 9287 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9287 9288 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9288 9289 /* Not for us */
9289 9290 break;
9290 9291 }
9291 9292 /* FALLTHRU */
9292 9293 case IPOPT_TS_TSANDADDR:
9293 9294 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9294 9295 break;
9295 9296 default:
9296 9297 /*
9297 9298 * ip_*put_options should have already
9298 9299 * dropped this packet.
9299 9300 */
9300 9301 cmn_err(CE_PANIC, "ip_input_local_options: "
9301 9302 "unknown IT - bug in ip_input_options?\n");
9302 9303 return (B_TRUE); /* Keep "lint" happy */
9303 9304 }
9304 9305 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
9305 9306 /* Increase overflow counter */
9306 9307 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
9307 9308 opt[IPOPT_POS_OV_FLG] =
9308 9309 (uint8_t)((opt[IPOPT_POS_OV_FLG] & 0x0F) |
9309 9310 (off << 4));
9310 9311 break;
9311 9312 }
9312 9313 off = opt[IPOPT_OFFSET] - 1;
9313 9314 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9314 9315 case IPOPT_TS_PRESPEC:
9315 9316 case IPOPT_TS_PRESPEC_RFC791:
9316 9317 case IPOPT_TS_TSANDADDR:
9317 9318 /* Pick a reasonable addr on the outbound if */
9318 9319 if (ip_select_source_v4(ill, INADDR_ANY,
9319 9320 ipha->ipha_dst, INADDR_ANY, ALL_ZONES, ipst,
9320 9321 &ifaddr, NULL, NULL) != 0) {
9321 9322 /* No source! Shouldn't happen */
9322 9323 ifaddr = INADDR_ANY;
9323 9324 }
9324 9325 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9325 9326 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9326 9327 /* FALLTHRU */
9327 9328 case IPOPT_TS_TSONLY:
9328 9329 off = opt[IPOPT_OFFSET] - 1;
9329 9330 /* Compute # of milliseconds since midnight */
9330 9331 gethrestime(&now);
9331 9332 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
9332 9333 now.tv_nsec / (NANOSEC / MILLISEC);
9333 9334 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
9334 9335 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
9335 9336 break;
9336 9337 }
9337 9338 break;
9338 9339 }
9339 9340 }
9340 9341 return (B_TRUE);
9341 9342
9342 9343 bad_src_route:
9343 9344 /* make sure we clear any indication of a hardware checksum */
9344 9345 DB_CKSUMFLAGS(mp) = 0;
9345 9346 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ill);
9346 9347 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
9347 9348 return (B_FALSE);
9348 9349
9349 9350 }
9350 9351
9351 9352 /*
9352 9353 * Process IP options in an inbound packet. Always returns the nexthop.
9353 9354 * Normally this is the passed in nexthop, but if there is an option
9354 9355 * that effects the nexthop (such as a source route) that will be returned.
9355 9356 * Sets *errorp if there is an error, in which case an ICMP error has been sent
9356 9357 * and mp freed.
9357 9358 */
9358 9359 ipaddr_t
9359 9360 ip_input_options(ipha_t *ipha, ipaddr_t dst, mblk_t *mp,
9360 9361 ip_recv_attr_t *ira, int *errorp)
9361 9362 {
9362 9363 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
9363 9364 ipoptp_t opts;
9364 9365 uchar_t *opt;
9365 9366 uint8_t optval;
9366 9367 uint8_t optlen;
9367 9368 intptr_t code = 0;
9368 9369 ire_t *ire;
9369 9370
9370 9371 ip2dbg(("ip_input_options\n"));
9371 9372 *errorp = 0;
9372 9373 for (optval = ipoptp_first(&opts, ipha);
9373 9374 optval != IPOPT_EOL;
9374 9375 optval = ipoptp_next(&opts)) {
9375 9376 opt = opts.ipoptp_cur;
9376 9377 optlen = opts.ipoptp_len;
9377 9378 ip2dbg(("ip_input_options: opt %d, len %d\n",
9378 9379 optval, optlen));
9379 9380 /*
9380 9381 * Note: we need to verify the checksum before we
9381 9382 * modify anything thus this routine only extracts the next
9382 9383 * hop dst from any source route.
9383 9384 */
9384 9385 switch (optval) {
9385 9386 uint32_t off;
9386 9387 case IPOPT_SSRR:
9387 9388 case IPOPT_LSRR:
9388 9389 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9389 9390 if (optval == IPOPT_SSRR) {
9390 9391 ip1dbg(("ip_input_options: not next"
9391 9392 " strict source route 0x%x\n",
9392 9393 ntohl(dst)));
9393 9394 code = (char *)&ipha->ipha_dst -
9394 9395 (char *)ipha;
9395 9396 goto param_prob; /* RouterReq's */
9396 9397 }
9397 9398 ip2dbg(("ip_input_options: "
9398 9399 "not next source route 0x%x\n",
9399 9400 ntohl(dst)));
9400 9401 break;
9401 9402 }
9402 9403
9403 9404 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9404 9405 ip1dbg((
9405 9406 "ip_input_options: bad option offset\n"));
9406 9407 code = (char *)&opt[IPOPT_OLEN] -
9407 9408 (char *)ipha;
9408 9409 goto param_prob;
9409 9410 }
9410 9411 off = opt[IPOPT_OFFSET];
9411 9412 off--;
9412 9413 redo_srr:
9413 9414 if (optlen < IP_ADDR_LEN ||
9414 9415 off > optlen - IP_ADDR_LEN) {
9415 9416 /* End of source route */
9416 9417 ip1dbg(("ip_input_options: end of SR\n"));
9417 9418 break;
9418 9419 }
9419 9420 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9420 9421 ip1dbg(("ip_input_options: next hop 0x%x\n",
9421 9422 ntohl(dst)));
9422 9423
9423 9424 /*
9424 9425 * Check if our address is present more than
9425 9426 * once as consecutive hops in source route.
9426 9427 * XXX verify per-interface ip_forwarding
9427 9428 * for source route?
9428 9429 */
9429 9430 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
9430 9431 off += IP_ADDR_LEN;
9431 9432 goto redo_srr;
9432 9433 }
9433 9434
9434 9435 if (dst == htonl(INADDR_LOOPBACK)) {
9435 9436 ip1dbg(("ip_input_options: loopback addr in "
9436 9437 "source route!\n"));
9437 9438 goto bad_src_route;
9438 9439 }
9439 9440 /*
9440 9441 * For strict: verify that dst is directly
9441 9442 * reachable.
9442 9443 */
9443 9444 if (optval == IPOPT_SSRR) {
9444 9445 ire = ire_ftable_lookup_v4(dst, 0, 0,
9445 9446 IRE_INTERFACE, NULL, ALL_ZONES,
9446 9447 ira->ira_tsl,
9447 9448 MATCH_IRE_TYPE | MATCH_IRE_SECATTR, 0, ipst,
9448 9449 NULL);
9449 9450 if (ire == NULL) {
9450 9451 ip1dbg(("ip_input_options: SSRR not "
9451 9452 "directly reachable: 0x%x\n",
9452 9453 ntohl(dst)));
9453 9454 goto bad_src_route;
9454 9455 }
9455 9456 ire_refrele(ire);
9456 9457 }
9457 9458 /*
9458 9459 * Defer update of the offset and the record route
9459 9460 * until the packet is forwarded.
9460 9461 */
9461 9462 break;
9462 9463 case IPOPT_RR:
9463 9464 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9464 9465 ip1dbg((
9465 9466 "ip_input_options: bad option offset\n"));
9466 9467 code = (char *)&opt[IPOPT_OLEN] -
9467 9468 (char *)ipha;
9468 9469 goto param_prob;
9469 9470 }
9470 9471 break;
9471 9472 case IPOPT_TS:
9472 9473 /*
9473 9474 * Verify that length >= 5 and that there is either
9474 9475 * room for another timestamp or that the overflow
9475 9476 * counter is not maxed out.
9476 9477 */
9477 9478 code = (char *)&opt[IPOPT_OLEN] - (char *)ipha;
9478 9479 if (optlen < IPOPT_MINLEN_IT) {
9479 9480 goto param_prob;
9480 9481 }
9481 9482 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9482 9483 ip1dbg((
9483 9484 "ip_input_options: bad option offset\n"));
9484 9485 code = (char *)&opt[IPOPT_OFFSET] -
9485 9486 (char *)ipha;
9486 9487 goto param_prob;
9487 9488 }
9488 9489 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9489 9490 case IPOPT_TS_TSONLY:
9490 9491 off = IPOPT_TS_TIMELEN;
9491 9492 break;
9492 9493 case IPOPT_TS_TSANDADDR:
9493 9494 case IPOPT_TS_PRESPEC:
9494 9495 case IPOPT_TS_PRESPEC_RFC791:
9495 9496 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9496 9497 break;
9497 9498 default:
9498 9499 code = (char *)&opt[IPOPT_POS_OV_FLG] -
9499 9500 (char *)ipha;
9500 9501 goto param_prob;
9501 9502 }
9502 9503 if (opt[IPOPT_OFFSET] - 1 + off > optlen &&
9503 9504 (opt[IPOPT_POS_OV_FLG] & 0xF0) == 0xF0) {
9504 9505 /*
9505 9506 * No room and the overflow counter is 15
9506 9507 * already.
9507 9508 */
9508 9509 goto param_prob;
9509 9510 }
9510 9511 break;
9511 9512 }
9512 9513 }
9513 9514
9514 9515 if ((opts.ipoptp_flags & IPOPTP_ERROR) == 0) {
9515 9516 return (dst);
9516 9517 }
9517 9518
9518 9519 ip1dbg(("ip_input_options: error processing IP options."));
9519 9520 code = (char *)&opt[IPOPT_OFFSET] - (char *)ipha;
9520 9521
9521 9522 param_prob:
9522 9523 /* make sure we clear any indication of a hardware checksum */
9523 9524 DB_CKSUMFLAGS(mp) = 0;
9524 9525 ip_drop_input("ICMP_PARAM_PROBLEM", mp, ira->ira_ill);
9525 9526 icmp_param_problem(mp, (uint8_t)code, ira);
9526 9527 *errorp = -1;
9527 9528 return (dst);
9528 9529
9529 9530 bad_src_route:
9530 9531 /* make sure we clear any indication of a hardware checksum */
9531 9532 DB_CKSUMFLAGS(mp) = 0;
9532 9533 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ira->ira_ill);
9533 9534 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
9534 9535 *errorp = -1;
9535 9536 return (dst);
9536 9537 }
9537 9538
9538 9539 /*
9539 9540 * IP & ICMP info in >=14 msg's ...
9540 9541 * - ip fixed part (mib2_ip_t)
9541 9542 * - icmp fixed part (mib2_icmp_t)
9542 9543 * - ipAddrEntryTable (ip 20) all IPv4 ipifs
9543 9544 * - ipRouteEntryTable (ip 21) all IPv4 IREs
9544 9545 * - ipNetToMediaEntryTable (ip 22) all IPv4 Neighbor Cache entries
9545 9546 * - ipRouteAttributeTable (ip 102) labeled routes
9546 9547 * - ip multicast membership (ip_member_t)
9547 9548 * - ip multicast source filtering (ip_grpsrc_t)
9548 9549 * - igmp fixed part (struct igmpstat)
9549 9550 * - multicast routing stats (struct mrtstat)
9550 9551 * - multicast routing vifs (array of struct vifctl)
9551 9552 * - multicast routing routes (array of struct mfcctl)
9552 9553 * - ip6 fixed part (mib2_ipv6IfStatsEntry_t)
9553 9554 * One per ill plus one generic
9554 9555 * - icmp6 fixed part (mib2_ipv6IfIcmpEntry_t)
9555 9556 * One per ill plus one generic
9556 9557 * - ipv6RouteEntry all IPv6 IREs
9557 9558 * - ipv6RouteAttributeTable (ip6 102) labeled routes
9558 9559 * - ipv6NetToMediaEntry all IPv6 Neighbor Cache entries
9559 9560 * - ipv6AddrEntry all IPv6 ipifs
9560 9561 * - ipv6 multicast membership (ipv6_member_t)
9561 9562 * - ipv6 multicast source filtering (ipv6_grpsrc_t)
9562 9563 *
9563 9564 * NOTE: original mpctl is copied for msg's 2..N, since its ctl part is
9564 9565 * already filled in by the caller.
9565 9566 * If legacy_req is true then MIB structures needs to be truncated to their
9566 9567 * legacy sizes before being returned.
9567 9568 * Return value of 0 indicates that no messages were sent and caller
9568 9569 * should free mpctl.
9569 9570 */
9570 9571 int
9571 9572 ip_snmp_get(queue_t *q, mblk_t *mpctl, int level, boolean_t legacy_req)
9572 9573 {
9573 9574 ip_stack_t *ipst;
9574 9575 sctp_stack_t *sctps;
9575 9576
9576 9577 if (q->q_next != NULL) {
9577 9578 ipst = ILLQ_TO_IPST(q);
9578 9579 } else {
9579 9580 ipst = CONNQ_TO_IPST(q);
9580 9581 }
9581 9582 ASSERT(ipst != NULL);
9582 9583 sctps = ipst->ips_netstack->netstack_sctp;
9583 9584
9584 9585 if (mpctl == NULL || mpctl->b_cont == NULL) {
9585 9586 return (0);
9586 9587 }
9587 9588
9588 9589 /*
9589 9590 * For the purposes of the (broken) packet shell use
9590 9591 * of the level we make sure MIB2_TCP/MIB2_UDP can be used
9591 9592 * to make TCP and UDP appear first in the list of mib items.
9592 9593 * TBD: We could expand this and use it in netstat so that
9593 9594 * the kernel doesn't have to produce large tables (connections,
9594 9595 * routes, etc) when netstat only wants the statistics or a particular
9595 9596 * table.
9596 9597 */
9597 9598 if (!(level == MIB2_TCP || level == MIB2_UDP)) {
9598 9599 if ((mpctl = icmp_snmp_get(q, mpctl)) == NULL) {
9599 9600 return (1);
9600 9601 }
9601 9602 }
9602 9603
9603 9604 if (level != MIB2_TCP) {
9604 9605 if ((mpctl = udp_snmp_get(q, mpctl, legacy_req)) == NULL) {
9605 9606 return (1);
9606 9607 }
9607 9608 }
9608 9609
9609 9610 if (level != MIB2_UDP) {
9610 9611 if ((mpctl = tcp_snmp_get(q, mpctl, legacy_req)) == NULL) {
9611 9612 return (1);
9612 9613 }
9613 9614 }
9614 9615
9615 9616 if ((mpctl = ip_snmp_get_mib2_ip_traffic_stats(q, mpctl,
9616 9617 ipst, legacy_req)) == NULL) {
9617 9618 return (1);
9618 9619 }
9619 9620
9620 9621 if ((mpctl = ip_snmp_get_mib2_ip6(q, mpctl, ipst,
9621 9622 legacy_req)) == NULL) {
9622 9623 return (1);
9623 9624 }
9624 9625
9625 9626 if ((mpctl = ip_snmp_get_mib2_icmp(q, mpctl, ipst)) == NULL) {
9626 9627 return (1);
9627 9628 }
9628 9629
9629 9630 if ((mpctl = ip_snmp_get_mib2_icmp6(q, mpctl, ipst)) == NULL) {
9630 9631 return (1);
9631 9632 }
9632 9633
9633 9634 if ((mpctl = ip_snmp_get_mib2_igmp(q, mpctl, ipst)) == NULL) {
9634 9635 return (1);
9635 9636 }
9636 9637
9637 9638 if ((mpctl = ip_snmp_get_mib2_multi(q, mpctl, ipst)) == NULL) {
9638 9639 return (1);
9639 9640 }
9640 9641
9641 9642 if ((mpctl = ip_snmp_get_mib2_ip_addr(q, mpctl, ipst,
9642 9643 legacy_req)) == NULL) {
9643 9644 return (1);
9644 9645 }
9645 9646
9646 9647 if ((mpctl = ip_snmp_get_mib2_ip6_addr(q, mpctl, ipst,
9647 9648 legacy_req)) == NULL) {
9648 9649 return (1);
9649 9650 }
9650 9651
9651 9652 if ((mpctl = ip_snmp_get_mib2_ip_group_mem(q, mpctl, ipst)) == NULL) {
9652 9653 return (1);
9653 9654 }
9654 9655
9655 9656 if ((mpctl = ip_snmp_get_mib2_ip6_group_mem(q, mpctl, ipst)) == NULL) {
9656 9657 return (1);
9657 9658 }
9658 9659
9659 9660 if ((mpctl = ip_snmp_get_mib2_ip_group_src(q, mpctl, ipst)) == NULL) {
9660 9661 return (1);
9661 9662 }
9662 9663
9663 9664 if ((mpctl = ip_snmp_get_mib2_ip6_group_src(q, mpctl, ipst)) == NULL) {
9664 9665 return (1);
9665 9666 }
9666 9667
9667 9668 if ((mpctl = ip_snmp_get_mib2_virt_multi(q, mpctl, ipst)) == NULL) {
9668 9669 return (1);
9669 9670 }
9670 9671
9671 9672 if ((mpctl = ip_snmp_get_mib2_multi_rtable(q, mpctl, ipst)) == NULL) {
9672 9673 return (1);
9673 9674 }
9674 9675
9675 9676 mpctl = ip_snmp_get_mib2_ip_route_media(q, mpctl, level, ipst);
9676 9677 if (mpctl == NULL)
9677 9678 return (1);
9678 9679
9679 9680 mpctl = ip_snmp_get_mib2_ip6_route_media(q, mpctl, level, ipst);
9680 9681 if (mpctl == NULL)
9681 9682 return (1);
9682 9683
9683 9684 if ((mpctl = sctp_snmp_get_mib2(q, mpctl, sctps)) == NULL) {
9684 9685 return (1);
9685 9686 }
9686 9687 if ((mpctl = ip_snmp_get_mib2_ip_dce(q, mpctl, ipst)) == NULL) {
9687 9688 return (1);
9688 9689 }
9689 9690 freemsg(mpctl);
9690 9691 return (1);
9691 9692 }
9692 9693
9693 9694 /* Get global (legacy) IPv4 statistics */
9694 9695 static mblk_t *
9695 9696 ip_snmp_get_mib2_ip(queue_t *q, mblk_t *mpctl, mib2_ipIfStatsEntry_t *ipmib,
9696 9697 ip_stack_t *ipst, boolean_t legacy_req)
9697 9698 {
9698 9699 mib2_ip_t old_ip_mib;
9699 9700 struct opthdr *optp;
9700 9701 mblk_t *mp2ctl;
9701 9702 mib2_ipAddrEntry_t mae;
9702 9703
9703 9704 /*
9704 9705 * make a copy of the original message
9705 9706 */
9706 9707 mp2ctl = copymsg(mpctl);
9707 9708
9708 9709 /* fixed length IP structure... */
9709 9710 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9710 9711 optp->level = MIB2_IP;
9711 9712 optp->name = 0;
9712 9713 SET_MIB(old_ip_mib.ipForwarding,
9713 9714 (WE_ARE_FORWARDING(ipst) ? 1 : 2));
9714 9715 SET_MIB(old_ip_mib.ipDefaultTTL,
9715 9716 (uint32_t)ipst->ips_ip_def_ttl);
9716 9717 SET_MIB(old_ip_mib.ipReasmTimeout,
9717 9718 ipst->ips_ip_reassembly_timeout);
9718 9719 SET_MIB(old_ip_mib.ipAddrEntrySize,
9719 9720 (legacy_req) ? LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t) :
9720 9721 sizeof (mib2_ipAddrEntry_t));
9721 9722 SET_MIB(old_ip_mib.ipRouteEntrySize,
9722 9723 sizeof (mib2_ipRouteEntry_t));
9723 9724 SET_MIB(old_ip_mib.ipNetToMediaEntrySize,
9724 9725 sizeof (mib2_ipNetToMediaEntry_t));
9725 9726 SET_MIB(old_ip_mib.ipMemberEntrySize, sizeof (ip_member_t));
9726 9727 SET_MIB(old_ip_mib.ipGroupSourceEntrySize, sizeof (ip_grpsrc_t));
9727 9728 SET_MIB(old_ip_mib.ipRouteAttributeSize,
9728 9729 sizeof (mib2_ipAttributeEntry_t));
9729 9730 SET_MIB(old_ip_mib.transportMLPSize, sizeof (mib2_transportMLPEntry_t));
9730 9731 SET_MIB(old_ip_mib.ipDestEntrySize, sizeof (dest_cache_entry_t));
9731 9732
9732 9733 /*
9733 9734 * Grab the statistics from the new IP MIB
9734 9735 */
9735 9736 SET_MIB(old_ip_mib.ipInReceives,
9736 9737 (uint32_t)ipmib->ipIfStatsHCInReceives);
9737 9738 SET_MIB(old_ip_mib.ipInHdrErrors, ipmib->ipIfStatsInHdrErrors);
9738 9739 SET_MIB(old_ip_mib.ipInAddrErrors, ipmib->ipIfStatsInAddrErrors);
9739 9740 SET_MIB(old_ip_mib.ipForwDatagrams,
9740 9741 (uint32_t)ipmib->ipIfStatsHCOutForwDatagrams);
9741 9742 SET_MIB(old_ip_mib.ipInUnknownProtos,
9742 9743 ipmib->ipIfStatsInUnknownProtos);
9743 9744 SET_MIB(old_ip_mib.ipInDiscards, ipmib->ipIfStatsInDiscards);
9744 9745 SET_MIB(old_ip_mib.ipInDelivers,
9745 9746 (uint32_t)ipmib->ipIfStatsHCInDelivers);
9746 9747 SET_MIB(old_ip_mib.ipOutRequests,
9747 9748 (uint32_t)ipmib->ipIfStatsHCOutRequests);
9748 9749 SET_MIB(old_ip_mib.ipOutDiscards, ipmib->ipIfStatsOutDiscards);
9749 9750 SET_MIB(old_ip_mib.ipOutNoRoutes, ipmib->ipIfStatsOutNoRoutes);
9750 9751 SET_MIB(old_ip_mib.ipReasmReqds, ipmib->ipIfStatsReasmReqds);
9751 9752 SET_MIB(old_ip_mib.ipReasmOKs, ipmib->ipIfStatsReasmOKs);
9752 9753 SET_MIB(old_ip_mib.ipReasmFails, ipmib->ipIfStatsReasmFails);
9753 9754 SET_MIB(old_ip_mib.ipFragOKs, ipmib->ipIfStatsOutFragOKs);
9754 9755 SET_MIB(old_ip_mib.ipFragFails, ipmib->ipIfStatsOutFragFails);
9755 9756 SET_MIB(old_ip_mib.ipFragCreates, ipmib->ipIfStatsOutFragCreates);
9756 9757
9757 9758 /* ipRoutingDiscards is not being used */
9758 9759 SET_MIB(old_ip_mib.ipRoutingDiscards, 0);
9759 9760 SET_MIB(old_ip_mib.tcpInErrs, ipmib->tcpIfStatsInErrs);
9760 9761 SET_MIB(old_ip_mib.udpNoPorts, ipmib->udpIfStatsNoPorts);
9761 9762 SET_MIB(old_ip_mib.ipInCksumErrs, ipmib->ipIfStatsInCksumErrs);
9762 9763 SET_MIB(old_ip_mib.ipReasmDuplicates,
9763 9764 ipmib->ipIfStatsReasmDuplicates);
9764 9765 SET_MIB(old_ip_mib.ipReasmPartDups, ipmib->ipIfStatsReasmPartDups);
9765 9766 SET_MIB(old_ip_mib.ipForwProhibits, ipmib->ipIfStatsForwProhibits);
9766 9767 SET_MIB(old_ip_mib.udpInCksumErrs, ipmib->udpIfStatsInCksumErrs);
9767 9768 SET_MIB(old_ip_mib.udpInOverflows, ipmib->udpIfStatsInOverflows);
9768 9769 SET_MIB(old_ip_mib.rawipInOverflows,
9769 9770 ipmib->rawipIfStatsInOverflows);
9770 9771
9771 9772 SET_MIB(old_ip_mib.ipsecInSucceeded, ipmib->ipsecIfStatsInSucceeded);
9772 9773 SET_MIB(old_ip_mib.ipsecInFailed, ipmib->ipsecIfStatsInFailed);
9773 9774 SET_MIB(old_ip_mib.ipInIPv6, ipmib->ipIfStatsInWrongIPVersion);
9774 9775 SET_MIB(old_ip_mib.ipOutIPv6, ipmib->ipIfStatsOutWrongIPVersion);
9775 9776 SET_MIB(old_ip_mib.ipOutSwitchIPv6,
9776 9777 ipmib->ipIfStatsOutSwitchIPVersion);
9777 9778
9778 9779 if (!snmp_append_data(mpctl->b_cont, (char *)&old_ip_mib,
9779 9780 (int)sizeof (old_ip_mib))) {
9780 9781 ip1dbg(("ip_snmp_get_mib2_ip: failed to allocate %u bytes\n",
9781 9782 (uint_t)sizeof (old_ip_mib)));
9782 9783 }
9783 9784
9784 9785 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9785 9786 ip3dbg(("ip_snmp_get_mib2_ip: level %d, name %d, len %d\n",
9786 9787 (int)optp->level, (int)optp->name, (int)optp->len));
9787 9788 qreply(q, mpctl);
9788 9789 return (mp2ctl);
9789 9790 }
9790 9791
9791 9792 /* Per interface IPv4 statistics */
9792 9793 static mblk_t *
9793 9794 ip_snmp_get_mib2_ip_traffic_stats(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
9794 9795 boolean_t legacy_req)
9795 9796 {
9796 9797 struct opthdr *optp;
9797 9798 mblk_t *mp2ctl;
9798 9799 ill_t *ill;
9799 9800 ill_walk_context_t ctx;
9800 9801 mblk_t *mp_tail = NULL;
9801 9802 mib2_ipIfStatsEntry_t global_ip_mib;
9802 9803 mib2_ipAddrEntry_t mae;
9803 9804
9804 9805 /*
9805 9806 * Make a copy of the original message
9806 9807 */
9807 9808 mp2ctl = copymsg(mpctl);
9808 9809
9809 9810 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9810 9811 optp->level = MIB2_IP;
9811 9812 optp->name = MIB2_IP_TRAFFIC_STATS;
9812 9813 /* Include "unknown interface" ip_mib */
9813 9814 ipst->ips_ip_mib.ipIfStatsIPVersion = MIB2_INETADDRESSTYPE_ipv4;
9814 9815 ipst->ips_ip_mib.ipIfStatsIfIndex =
9815 9816 MIB2_UNKNOWN_INTERFACE; /* Flag to netstat */
9816 9817 SET_MIB(ipst->ips_ip_mib.ipIfStatsForwarding,
9817 9818 (ipst->ips_ip_forwarding ? 1 : 2));
9818 9819 SET_MIB(ipst->ips_ip_mib.ipIfStatsDefaultTTL,
9819 9820 (uint32_t)ipst->ips_ip_def_ttl);
9820 9821 SET_MIB(ipst->ips_ip_mib.ipIfStatsEntrySize,
9821 9822 sizeof (mib2_ipIfStatsEntry_t));
9822 9823 SET_MIB(ipst->ips_ip_mib.ipIfStatsAddrEntrySize,
9823 9824 sizeof (mib2_ipAddrEntry_t));
9824 9825 SET_MIB(ipst->ips_ip_mib.ipIfStatsRouteEntrySize,
9825 9826 sizeof (mib2_ipRouteEntry_t));
9826 9827 SET_MIB(ipst->ips_ip_mib.ipIfStatsNetToMediaEntrySize,
9827 9828 sizeof (mib2_ipNetToMediaEntry_t));
9828 9829 SET_MIB(ipst->ips_ip_mib.ipIfStatsMemberEntrySize,
9829 9830 sizeof (ip_member_t));
9830 9831 SET_MIB(ipst->ips_ip_mib.ipIfStatsGroupSourceEntrySize,
9831 9832 sizeof (ip_grpsrc_t));
9832 9833
9833 9834 bcopy(&ipst->ips_ip_mib, &global_ip_mib, sizeof (global_ip_mib));
9834 9835
9835 9836 if (legacy_req) {
9836 9837 SET_MIB(global_ip_mib.ipIfStatsAddrEntrySize,
9837 9838 LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t));
9838 9839 }
9839 9840
9840 9841 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
9841 9842 (char *)&global_ip_mib, (int)sizeof (global_ip_mib))) {
9842 9843 ip1dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9843 9844 "failed to allocate %u bytes\n",
9844 9845 (uint_t)sizeof (global_ip_mib)));
9845 9846 }
9846 9847
9847 9848 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
9848 9849 ill = ILL_START_WALK_V4(&ctx, ipst);
9849 9850 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
9850 9851 ill->ill_ip_mib->ipIfStatsIfIndex =
9851 9852 ill->ill_phyint->phyint_ifindex;
9852 9853 SET_MIB(ill->ill_ip_mib->ipIfStatsForwarding,
9853 9854 (ipst->ips_ip_forwarding ? 1 : 2));
9854 9855 SET_MIB(ill->ill_ip_mib->ipIfStatsDefaultTTL,
9855 9856 (uint32_t)ipst->ips_ip_def_ttl);
9856 9857
9857 9858 ip_mib2_add_ip_stats(&global_ip_mib, ill->ill_ip_mib);
9858 9859 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
9859 9860 (char *)ill->ill_ip_mib,
9860 9861 (int)sizeof (*ill->ill_ip_mib))) {
9861 9862 ip1dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9862 9863 "failed to allocate %u bytes\n",
9863 9864 (uint_t)sizeof (*ill->ill_ip_mib)));
9864 9865 }
9865 9866 }
9866 9867 rw_exit(&ipst->ips_ill_g_lock);
9867 9868
9868 9869 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9869 9870 ip3dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9870 9871 "level %d, name %d, len %d\n",
9871 9872 (int)optp->level, (int)optp->name, (int)optp->len));
9872 9873 qreply(q, mpctl);
9873 9874
9874 9875 if (mp2ctl == NULL)
9875 9876 return (NULL);
9876 9877
9877 9878 return (ip_snmp_get_mib2_ip(q, mp2ctl, &global_ip_mib, ipst,
9878 9879 legacy_req));
9879 9880 }
9880 9881
9881 9882 /* Global IPv4 ICMP statistics */
9882 9883 static mblk_t *
9883 9884 ip_snmp_get_mib2_icmp(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9884 9885 {
9885 9886 struct opthdr *optp;
9886 9887 mblk_t *mp2ctl;
9887 9888
9888 9889 /*
9889 9890 * Make a copy of the original message
9890 9891 */
9891 9892 mp2ctl = copymsg(mpctl);
9892 9893
9893 9894 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9894 9895 optp->level = MIB2_ICMP;
9895 9896 optp->name = 0;
9896 9897 if (!snmp_append_data(mpctl->b_cont, (char *)&ipst->ips_icmp_mib,
9897 9898 (int)sizeof (ipst->ips_icmp_mib))) {
9898 9899 ip1dbg(("ip_snmp_get_mib2_icmp: failed to allocate %u bytes\n",
9899 9900 (uint_t)sizeof (ipst->ips_icmp_mib)));
9900 9901 }
9901 9902 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9902 9903 ip3dbg(("ip_snmp_get_mib2_icmp: level %d, name %d, len %d\n",
9903 9904 (int)optp->level, (int)optp->name, (int)optp->len));
9904 9905 qreply(q, mpctl);
9905 9906 return (mp2ctl);
9906 9907 }
9907 9908
9908 9909 /* Global IPv4 IGMP statistics */
9909 9910 static mblk_t *
9910 9911 ip_snmp_get_mib2_igmp(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9911 9912 {
9912 9913 struct opthdr *optp;
9913 9914 mblk_t *mp2ctl;
9914 9915
9915 9916 /*
9916 9917 * make a copy of the original message
9917 9918 */
9918 9919 mp2ctl = copymsg(mpctl);
9919 9920
9920 9921 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9921 9922 optp->level = EXPER_IGMP;
9922 9923 optp->name = 0;
9923 9924 if (!snmp_append_data(mpctl->b_cont, (char *)&ipst->ips_igmpstat,
9924 9925 (int)sizeof (ipst->ips_igmpstat))) {
9925 9926 ip1dbg(("ip_snmp_get_mib2_igmp: failed to allocate %u bytes\n",
9926 9927 (uint_t)sizeof (ipst->ips_igmpstat)));
9927 9928 }
9928 9929 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9929 9930 ip3dbg(("ip_snmp_get_mib2_igmp: level %d, name %d, len %d\n",
9930 9931 (int)optp->level, (int)optp->name, (int)optp->len));
9931 9932 qreply(q, mpctl);
9932 9933 return (mp2ctl);
9933 9934 }
9934 9935
9935 9936 /* Global IPv4 Multicast Routing statistics */
9936 9937 static mblk_t *
9937 9938 ip_snmp_get_mib2_multi(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9938 9939 {
9939 9940 struct opthdr *optp;
9940 9941 mblk_t *mp2ctl;
9941 9942
9942 9943 /*
9943 9944 * make a copy of the original message
9944 9945 */
9945 9946 mp2ctl = copymsg(mpctl);
9946 9947
9947 9948 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9948 9949 optp->level = EXPER_DVMRP;
9949 9950 optp->name = 0;
9950 9951 if (!ip_mroute_stats(mpctl->b_cont, ipst)) {
9951 9952 ip0dbg(("ip_mroute_stats: failed\n"));
9952 9953 }
9953 9954 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9954 9955 ip3dbg(("ip_snmp_get_mib2_multi: level %d, name %d, len %d\n",
9955 9956 (int)optp->level, (int)optp->name, (int)optp->len));
9956 9957 qreply(q, mpctl);
9957 9958 return (mp2ctl);
9958 9959 }
9959 9960
9960 9961 /* IPv4 address information */
9961 9962 static mblk_t *
9962 9963 ip_snmp_get_mib2_ip_addr(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
9963 9964 boolean_t legacy_req)
9964 9965 {
9965 9966 struct opthdr *optp;
9966 9967 mblk_t *mp2ctl;
9967 9968 mblk_t *mp_tail = NULL;
9968 9969 ill_t *ill;
9969 9970 ipif_t *ipif;
9970 9971 uint_t bitval;
9971 9972 mib2_ipAddrEntry_t mae;
9972 9973 size_t mae_size;
9973 9974 zoneid_t zoneid;
9974 9975 ill_walk_context_t ctx;
9975 9976
9976 9977 /*
9977 9978 * make a copy of the original message
9978 9979 */
9979 9980 mp2ctl = copymsg(mpctl);
9980 9981
9981 9982 mae_size = (legacy_req) ? LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t) :
9982 9983 sizeof (mib2_ipAddrEntry_t);
9983 9984
9984 9985 /* ipAddrEntryTable */
9985 9986
9986 9987 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9987 9988 optp->level = MIB2_IP;
9988 9989 optp->name = MIB2_IP_ADDR;
9989 9990 zoneid = Q_TO_CONN(q)->conn_zoneid;
9990 9991
9991 9992 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
9992 9993 ill = ILL_START_WALK_V4(&ctx, ipst);
9993 9994 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
9994 9995 for (ipif = ill->ill_ipif; ipif != NULL;
9995 9996 ipif = ipif->ipif_next) {
9996 9997 if (ipif->ipif_zoneid != zoneid &&
9997 9998 ipif->ipif_zoneid != ALL_ZONES)
9998 9999 continue;
9999 10000 /* Sum of count from dead IRE_LO* and our current */
10000 10001 mae.ipAdEntInfo.ae_ibcnt = ipif->ipif_ib_pkt_count;
10001 10002 if (ipif->ipif_ire_local != NULL) {
10002 10003 mae.ipAdEntInfo.ae_ibcnt +=
10003 10004 ipif->ipif_ire_local->ire_ib_pkt_count;
10004 10005 }
10005 10006 mae.ipAdEntInfo.ae_obcnt = 0;
10006 10007 mae.ipAdEntInfo.ae_focnt = 0;
10007 10008
10008 10009 ipif_get_name(ipif, mae.ipAdEntIfIndex.o_bytes,
10009 10010 OCTET_LENGTH);
10010 10011 mae.ipAdEntIfIndex.o_length =
10011 10012 mi_strlen(mae.ipAdEntIfIndex.o_bytes);
10012 10013 mae.ipAdEntAddr = ipif->ipif_lcl_addr;
10013 10014 mae.ipAdEntNetMask = ipif->ipif_net_mask;
10014 10015 mae.ipAdEntInfo.ae_subnet = ipif->ipif_subnet;
10015 10016 mae.ipAdEntInfo.ae_subnet_len =
10016 10017 ip_mask_to_plen(ipif->ipif_net_mask);
10017 10018 mae.ipAdEntInfo.ae_src_addr = ipif->ipif_lcl_addr;
10018 10019 for (bitval = 1;
10019 10020 bitval &&
10020 10021 !(bitval & ipif->ipif_brd_addr);
10021 10022 bitval <<= 1)
10022 10023 noop;
10023 10024 mae.ipAdEntBcastAddr = bitval;
10024 10025 mae.ipAdEntReasmMaxSize = IP_MAXPACKET;
10025 10026 mae.ipAdEntInfo.ae_mtu = ipif->ipif_ill->ill_mtu;
10026 10027 mae.ipAdEntInfo.ae_metric = ipif->ipif_ill->ill_metric;
10027 10028 mae.ipAdEntInfo.ae_broadcast_addr =
10028 10029 ipif->ipif_brd_addr;
10029 10030 mae.ipAdEntInfo.ae_pp_dst_addr =
10030 10031 ipif->ipif_pp_dst_addr;
10031 10032 mae.ipAdEntInfo.ae_flags = ipif->ipif_flags |
10032 10033 ill->ill_flags | ill->ill_phyint->phyint_flags;
10033 10034 mae.ipAdEntRetransmitTime =
10034 10035 ill->ill_reachable_retrans_time;
10035 10036
10036 10037 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10037 10038 (char *)&mae, (int)mae_size)) {
10038 10039 ip1dbg(("ip_snmp_get_mib2_ip_addr: failed to "
10039 10040 "allocate %u bytes\n", (uint_t)mae_size));
10040 10041 }
10041 10042 }
10042 10043 }
10043 10044 rw_exit(&ipst->ips_ill_g_lock);
10044 10045
10045 10046 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10046 10047 ip3dbg(("ip_snmp_get_mib2_ip_addr: level %d, name %d, len %d\n",
10047 10048 (int)optp->level, (int)optp->name, (int)optp->len));
10048 10049 qreply(q, mpctl);
10049 10050 return (mp2ctl);
10050 10051 }
10051 10052
10052 10053 /* IPv6 address information */
10053 10054 static mblk_t *
10054 10055 ip_snmp_get_mib2_ip6_addr(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
10055 10056 boolean_t legacy_req)
10056 10057 {
10057 10058 struct opthdr *optp;
10058 10059 mblk_t *mp2ctl;
10059 10060 mblk_t *mp_tail = NULL;
10060 10061 ill_t *ill;
10061 10062 ipif_t *ipif;
10062 10063 mib2_ipv6AddrEntry_t mae6;
10063 10064 size_t mae6_size;
10064 10065 zoneid_t zoneid;
10065 10066 ill_walk_context_t ctx;
10066 10067
10067 10068 /*
10068 10069 * make a copy of the original message
10069 10070 */
10070 10071 mp2ctl = copymsg(mpctl);
10071 10072
10072 10073 mae6_size = (legacy_req) ?
10073 10074 LEGACY_MIB_SIZE(&mae6, mib2_ipv6AddrEntry_t) :
10074 10075 sizeof (mib2_ipv6AddrEntry_t);
10075 10076
10076 10077 /* ipv6AddrEntryTable */
10077 10078
10078 10079 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10079 10080 optp->level = MIB2_IP6;
10080 10081 optp->name = MIB2_IP6_ADDR;
10081 10082 zoneid = Q_TO_CONN(q)->conn_zoneid;
10082 10083
10083 10084 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10084 10085 ill = ILL_START_WALK_V6(&ctx, ipst);
10085 10086 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10086 10087 for (ipif = ill->ill_ipif; ipif != NULL;
10087 10088 ipif = ipif->ipif_next) {
10088 10089 if (ipif->ipif_zoneid != zoneid &&
10089 10090 ipif->ipif_zoneid != ALL_ZONES)
10090 10091 continue;
10091 10092 /* Sum of count from dead IRE_LO* and our current */
10092 10093 mae6.ipv6AddrInfo.ae_ibcnt = ipif->ipif_ib_pkt_count;
10093 10094 if (ipif->ipif_ire_local != NULL) {
10094 10095 mae6.ipv6AddrInfo.ae_ibcnt +=
10095 10096 ipif->ipif_ire_local->ire_ib_pkt_count;
10096 10097 }
10097 10098 mae6.ipv6AddrInfo.ae_obcnt = 0;
10098 10099 mae6.ipv6AddrInfo.ae_focnt = 0;
10099 10100
10100 10101 ipif_get_name(ipif, mae6.ipv6AddrIfIndex.o_bytes,
10101 10102 OCTET_LENGTH);
10102 10103 mae6.ipv6AddrIfIndex.o_length =
10103 10104 mi_strlen(mae6.ipv6AddrIfIndex.o_bytes);
10104 10105 mae6.ipv6AddrAddress = ipif->ipif_v6lcl_addr;
10105 10106 mae6.ipv6AddrPfxLength =
10106 10107 ip_mask_to_plen_v6(&ipif->ipif_v6net_mask);
10107 10108 mae6.ipv6AddrInfo.ae_subnet = ipif->ipif_v6subnet;
10108 10109 mae6.ipv6AddrInfo.ae_subnet_len =
10109 10110 mae6.ipv6AddrPfxLength;
10110 10111 mae6.ipv6AddrInfo.ae_src_addr = ipif->ipif_v6lcl_addr;
10111 10112
10112 10113 /* Type: stateless(1), stateful(2), unknown(3) */
10113 10114 if (ipif->ipif_flags & IPIF_ADDRCONF)
10114 10115 mae6.ipv6AddrType = 1;
10115 10116 else
10116 10117 mae6.ipv6AddrType = 2;
10117 10118 /* Anycast: true(1), false(2) */
10118 10119 if (ipif->ipif_flags & IPIF_ANYCAST)
10119 10120 mae6.ipv6AddrAnycastFlag = 1;
10120 10121 else
10121 10122 mae6.ipv6AddrAnycastFlag = 2;
10122 10123
10123 10124 /*
10124 10125 * Address status: preferred(1), deprecated(2),
10125 10126 * invalid(3), inaccessible(4), unknown(5)
10126 10127 */
10127 10128 if (ipif->ipif_flags & IPIF_NOLOCAL)
10128 10129 mae6.ipv6AddrStatus = 3;
10129 10130 else if (ipif->ipif_flags & IPIF_DEPRECATED)
10130 10131 mae6.ipv6AddrStatus = 2;
10131 10132 else
10132 10133 mae6.ipv6AddrStatus = 1;
10133 10134 mae6.ipv6AddrInfo.ae_mtu = ipif->ipif_ill->ill_mtu;
10134 10135 mae6.ipv6AddrInfo.ae_metric =
10135 10136 ipif->ipif_ill->ill_metric;
10136 10137 mae6.ipv6AddrInfo.ae_pp_dst_addr =
10137 10138 ipif->ipif_v6pp_dst_addr;
10138 10139 mae6.ipv6AddrInfo.ae_flags = ipif->ipif_flags |
10139 10140 ill->ill_flags | ill->ill_phyint->phyint_flags;
10140 10141 mae6.ipv6AddrReasmMaxSize = IP_MAXPACKET;
10141 10142 mae6.ipv6AddrIdentifier = ill->ill_token;
10142 10143 mae6.ipv6AddrIdentifierLen = ill->ill_token_length;
10143 10144 mae6.ipv6AddrReachableTime = ill->ill_reachable_time;
10144 10145 mae6.ipv6AddrRetransmitTime =
10145 10146 ill->ill_reachable_retrans_time;
10146 10147 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10147 10148 (char *)&mae6, (int)mae6_size)) {
10148 10149 ip1dbg(("ip_snmp_get_mib2_ip6_addr: failed to "
10149 10150 "allocate %u bytes\n",
10150 10151 (uint_t)mae6_size));
10151 10152 }
10152 10153 }
10153 10154 }
10154 10155 rw_exit(&ipst->ips_ill_g_lock);
10155 10156
10156 10157 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10157 10158 ip3dbg(("ip_snmp_get_mib2_ip6_addr: level %d, name %d, len %d\n",
10158 10159 (int)optp->level, (int)optp->name, (int)optp->len));
10159 10160 qreply(q, mpctl);
10160 10161 return (mp2ctl);
10161 10162 }
10162 10163
10163 10164 /* IPv4 multicast group membership. */
10164 10165 static mblk_t *
10165 10166 ip_snmp_get_mib2_ip_group_mem(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10166 10167 {
10167 10168 struct opthdr *optp;
10168 10169 mblk_t *mp2ctl;
10169 10170 ill_t *ill;
10170 10171 ipif_t *ipif;
10171 10172 ilm_t *ilm;
10172 10173 ip_member_t ipm;
10173 10174 mblk_t *mp_tail = NULL;
10174 10175 ill_walk_context_t ctx;
10175 10176 zoneid_t zoneid;
10176 10177
10177 10178 /*
10178 10179 * make a copy of the original message
10179 10180 */
10180 10181 mp2ctl = copymsg(mpctl);
10181 10182 zoneid = Q_TO_CONN(q)->conn_zoneid;
10182 10183
10183 10184 /* ipGroupMember table */
10184 10185 optp = (struct opthdr *)&mpctl->b_rptr[
10185 10186 sizeof (struct T_optmgmt_ack)];
10186 10187 optp->level = MIB2_IP;
10187 10188 optp->name = EXPER_IP_GROUP_MEMBERSHIP;
10188 10189
10189 10190 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10190 10191 ill = ILL_START_WALK_V4(&ctx, ipst);
10191 10192 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10192 10193 /* Make sure the ill isn't going away. */
10193 10194 if (!ill_check_and_refhold(ill))
10194 10195 continue;
10195 10196 rw_exit(&ipst->ips_ill_g_lock);
10196 10197 rw_enter(&ill->ill_mcast_lock, RW_READER);
10197 10198 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10198 10199 if (ilm->ilm_zoneid != zoneid &&
10199 10200 ilm->ilm_zoneid != ALL_ZONES)
10200 10201 continue;
10201 10202
10202 10203 /* Is there an ipif for ilm_ifaddr? */
10203 10204 for (ipif = ill->ill_ipif; ipif != NULL;
10204 10205 ipif = ipif->ipif_next) {
10205 10206 if (!IPIF_IS_CONDEMNED(ipif) &&
10206 10207 ipif->ipif_lcl_addr == ilm->ilm_ifaddr &&
10207 10208 ilm->ilm_ifaddr != INADDR_ANY)
10208 10209 break;
10209 10210 }
10210 10211 if (ipif != NULL) {
10211 10212 ipif_get_name(ipif,
10212 10213 ipm.ipGroupMemberIfIndex.o_bytes,
10213 10214 OCTET_LENGTH);
10214 10215 } else {
10215 10216 ill_get_name(ill,
10216 10217 ipm.ipGroupMemberIfIndex.o_bytes,
10217 10218 OCTET_LENGTH);
10218 10219 }
10219 10220 ipm.ipGroupMemberIfIndex.o_length =
10220 10221 mi_strlen(ipm.ipGroupMemberIfIndex.o_bytes);
10221 10222
10222 10223 ipm.ipGroupMemberAddress = ilm->ilm_addr;
10223 10224 ipm.ipGroupMemberRefCnt = ilm->ilm_refcnt;
10224 10225 ipm.ipGroupMemberFilterMode = ilm->ilm_fmode;
10225 10226 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10226 10227 (char *)&ipm, (int)sizeof (ipm))) {
10227 10228 ip1dbg(("ip_snmp_get_mib2_ip_group: "
10228 10229 "failed to allocate %u bytes\n",
10229 10230 (uint_t)sizeof (ipm)));
10230 10231 }
10231 10232 }
10232 10233 rw_exit(&ill->ill_mcast_lock);
10233 10234 ill_refrele(ill);
10234 10235 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10235 10236 }
10236 10237 rw_exit(&ipst->ips_ill_g_lock);
10237 10238 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10238 10239 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10239 10240 (int)optp->level, (int)optp->name, (int)optp->len));
10240 10241 qreply(q, mpctl);
10241 10242 return (mp2ctl);
10242 10243 }
10243 10244
10244 10245 /* IPv6 multicast group membership. */
10245 10246 static mblk_t *
10246 10247 ip_snmp_get_mib2_ip6_group_mem(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10247 10248 {
10248 10249 struct opthdr *optp;
10249 10250 mblk_t *mp2ctl;
10250 10251 ill_t *ill;
10251 10252 ilm_t *ilm;
10252 10253 ipv6_member_t ipm6;
10253 10254 mblk_t *mp_tail = NULL;
10254 10255 ill_walk_context_t ctx;
10255 10256 zoneid_t zoneid;
10256 10257
10257 10258 /*
10258 10259 * make a copy of the original message
10259 10260 */
10260 10261 mp2ctl = copymsg(mpctl);
10261 10262 zoneid = Q_TO_CONN(q)->conn_zoneid;
10262 10263
10263 10264 /* ip6GroupMember table */
10264 10265 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10265 10266 optp->level = MIB2_IP6;
10266 10267 optp->name = EXPER_IP6_GROUP_MEMBERSHIP;
10267 10268
10268 10269 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10269 10270 ill = ILL_START_WALK_V6(&ctx, ipst);
10270 10271 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10271 10272 /* Make sure the ill isn't going away. */
10272 10273 if (!ill_check_and_refhold(ill))
10273 10274 continue;
10274 10275 rw_exit(&ipst->ips_ill_g_lock);
10275 10276 /*
10276 10277 * Normally we don't have any members on under IPMP interfaces.
10277 10278 * We report them as a debugging aid.
10278 10279 */
10279 10280 rw_enter(&ill->ill_mcast_lock, RW_READER);
10280 10281 ipm6.ipv6GroupMemberIfIndex = ill->ill_phyint->phyint_ifindex;
10281 10282 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10282 10283 if (ilm->ilm_zoneid != zoneid &&
10283 10284 ilm->ilm_zoneid != ALL_ZONES)
10284 10285 continue; /* not this zone */
10285 10286 ipm6.ipv6GroupMemberAddress = ilm->ilm_v6addr;
10286 10287 ipm6.ipv6GroupMemberRefCnt = ilm->ilm_refcnt;
10287 10288 ipm6.ipv6GroupMemberFilterMode = ilm->ilm_fmode;
10288 10289 if (!snmp_append_data2(mpctl->b_cont,
10289 10290 &mp_tail,
10290 10291 (char *)&ipm6, (int)sizeof (ipm6))) {
10291 10292 ip1dbg(("ip_snmp_get_mib2_ip6_group: "
10292 10293 "failed to allocate %u bytes\n",
10293 10294 (uint_t)sizeof (ipm6)));
10294 10295 }
10295 10296 }
10296 10297 rw_exit(&ill->ill_mcast_lock);
10297 10298 ill_refrele(ill);
10298 10299 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10299 10300 }
10300 10301 rw_exit(&ipst->ips_ill_g_lock);
10301 10302
10302 10303 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10303 10304 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10304 10305 (int)optp->level, (int)optp->name, (int)optp->len));
10305 10306 qreply(q, mpctl);
10306 10307 return (mp2ctl);
10307 10308 }
10308 10309
10309 10310 /* IP multicast filtered sources */
10310 10311 static mblk_t *
10311 10312 ip_snmp_get_mib2_ip_group_src(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10312 10313 {
10313 10314 struct opthdr *optp;
10314 10315 mblk_t *mp2ctl;
10315 10316 ill_t *ill;
10316 10317 ipif_t *ipif;
10317 10318 ilm_t *ilm;
10318 10319 ip_grpsrc_t ips;
10319 10320 mblk_t *mp_tail = NULL;
10320 10321 ill_walk_context_t ctx;
10321 10322 zoneid_t zoneid;
10322 10323 int i;
10323 10324 slist_t *sl;
10324 10325
10325 10326 /*
10326 10327 * make a copy of the original message
10327 10328 */
10328 10329 mp2ctl = copymsg(mpctl);
10329 10330 zoneid = Q_TO_CONN(q)->conn_zoneid;
10330 10331
10331 10332 /* ipGroupSource table */
10332 10333 optp = (struct opthdr *)&mpctl->b_rptr[
10333 10334 sizeof (struct T_optmgmt_ack)];
10334 10335 optp->level = MIB2_IP;
10335 10336 optp->name = EXPER_IP_GROUP_SOURCES;
10336 10337
10337 10338 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10338 10339 ill = ILL_START_WALK_V4(&ctx, ipst);
10339 10340 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10340 10341 /* Make sure the ill isn't going away. */
10341 10342 if (!ill_check_and_refhold(ill))
10342 10343 continue;
10343 10344 rw_exit(&ipst->ips_ill_g_lock);
10344 10345 rw_enter(&ill->ill_mcast_lock, RW_READER);
10345 10346 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10346 10347 sl = ilm->ilm_filter;
10347 10348 if (ilm->ilm_zoneid != zoneid &&
10348 10349 ilm->ilm_zoneid != ALL_ZONES)
10349 10350 continue;
10350 10351 if (SLIST_IS_EMPTY(sl))
10351 10352 continue;
10352 10353
10353 10354 /* Is there an ipif for ilm_ifaddr? */
10354 10355 for (ipif = ill->ill_ipif; ipif != NULL;
10355 10356 ipif = ipif->ipif_next) {
10356 10357 if (!IPIF_IS_CONDEMNED(ipif) &&
10357 10358 ipif->ipif_lcl_addr == ilm->ilm_ifaddr &&
10358 10359 ilm->ilm_ifaddr != INADDR_ANY)
10359 10360 break;
10360 10361 }
10361 10362 if (ipif != NULL) {
10362 10363 ipif_get_name(ipif,
10363 10364 ips.ipGroupSourceIfIndex.o_bytes,
10364 10365 OCTET_LENGTH);
10365 10366 } else {
10366 10367 ill_get_name(ill,
10367 10368 ips.ipGroupSourceIfIndex.o_bytes,
10368 10369 OCTET_LENGTH);
10369 10370 }
10370 10371 ips.ipGroupSourceIfIndex.o_length =
10371 10372 mi_strlen(ips.ipGroupSourceIfIndex.o_bytes);
10372 10373
10373 10374 ips.ipGroupSourceGroup = ilm->ilm_addr;
10374 10375 for (i = 0; i < sl->sl_numsrc; i++) {
10375 10376 if (!IN6_IS_ADDR_V4MAPPED(&sl->sl_addr[i]))
10376 10377 continue;
10377 10378 IN6_V4MAPPED_TO_IPADDR(&sl->sl_addr[i],
10378 10379 ips.ipGroupSourceAddress);
10379 10380 if (snmp_append_data2(mpctl->b_cont, &mp_tail,
10380 10381 (char *)&ips, (int)sizeof (ips)) == 0) {
10381 10382 ip1dbg(("ip_snmp_get_mib2_ip_group_src:"
10382 10383 " failed to allocate %u bytes\n",
10383 10384 (uint_t)sizeof (ips)));
10384 10385 }
10385 10386 }
10386 10387 }
10387 10388 rw_exit(&ill->ill_mcast_lock);
10388 10389 ill_refrele(ill);
10389 10390 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10390 10391 }
10391 10392 rw_exit(&ipst->ips_ill_g_lock);
10392 10393 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10393 10394 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10394 10395 (int)optp->level, (int)optp->name, (int)optp->len));
10395 10396 qreply(q, mpctl);
10396 10397 return (mp2ctl);
10397 10398 }
10398 10399
10399 10400 /* IPv6 multicast filtered sources. */
10400 10401 static mblk_t *
10401 10402 ip_snmp_get_mib2_ip6_group_src(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10402 10403 {
10403 10404 struct opthdr *optp;
10404 10405 mblk_t *mp2ctl;
10405 10406 ill_t *ill;
10406 10407 ilm_t *ilm;
10407 10408 ipv6_grpsrc_t ips6;
10408 10409 mblk_t *mp_tail = NULL;
10409 10410 ill_walk_context_t ctx;
10410 10411 zoneid_t zoneid;
10411 10412 int i;
10412 10413 slist_t *sl;
10413 10414
10414 10415 /*
10415 10416 * make a copy of the original message
10416 10417 */
10417 10418 mp2ctl = copymsg(mpctl);
10418 10419 zoneid = Q_TO_CONN(q)->conn_zoneid;
10419 10420
10420 10421 /* ip6GroupMember table */
10421 10422 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10422 10423 optp->level = MIB2_IP6;
10423 10424 optp->name = EXPER_IP6_GROUP_SOURCES;
10424 10425
10425 10426 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10426 10427 ill = ILL_START_WALK_V6(&ctx, ipst);
10427 10428 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10428 10429 /* Make sure the ill isn't going away. */
10429 10430 if (!ill_check_and_refhold(ill))
10430 10431 continue;
10431 10432 rw_exit(&ipst->ips_ill_g_lock);
10432 10433 /*
10433 10434 * Normally we don't have any members on under IPMP interfaces.
10434 10435 * We report them as a debugging aid.
10435 10436 */
10436 10437 rw_enter(&ill->ill_mcast_lock, RW_READER);
10437 10438 ips6.ipv6GroupSourceIfIndex = ill->ill_phyint->phyint_ifindex;
10438 10439 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10439 10440 sl = ilm->ilm_filter;
10440 10441 if (ilm->ilm_zoneid != zoneid &&
10441 10442 ilm->ilm_zoneid != ALL_ZONES)
10442 10443 continue;
10443 10444 if (SLIST_IS_EMPTY(sl))
10444 10445 continue;
10445 10446 ips6.ipv6GroupSourceGroup = ilm->ilm_v6addr;
10446 10447 for (i = 0; i < sl->sl_numsrc; i++) {
10447 10448 ips6.ipv6GroupSourceAddress = sl->sl_addr[i];
10448 10449 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10449 10450 (char *)&ips6, (int)sizeof (ips6))) {
10450 10451 ip1dbg(("ip_snmp_get_mib2_ip6_"
10451 10452 "group_src: failed to allocate "
10452 10453 "%u bytes\n",
10453 10454 (uint_t)sizeof (ips6)));
10454 10455 }
10455 10456 }
10456 10457 }
10457 10458 rw_exit(&ill->ill_mcast_lock);
10458 10459 ill_refrele(ill);
10459 10460 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10460 10461 }
10461 10462 rw_exit(&ipst->ips_ill_g_lock);
10462 10463
10463 10464 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10464 10465 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10465 10466 (int)optp->level, (int)optp->name, (int)optp->len));
10466 10467 qreply(q, mpctl);
10467 10468 return (mp2ctl);
10468 10469 }
10469 10470
10470 10471 /* Multicast routing virtual interface table. */
10471 10472 static mblk_t *
10472 10473 ip_snmp_get_mib2_virt_multi(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10473 10474 {
10474 10475 struct opthdr *optp;
10475 10476 mblk_t *mp2ctl;
10476 10477
10477 10478 /*
10478 10479 * make a copy of the original message
10479 10480 */
10480 10481 mp2ctl = copymsg(mpctl);
10481 10482
10482 10483 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10483 10484 optp->level = EXPER_DVMRP;
10484 10485 optp->name = EXPER_DVMRP_VIF;
10485 10486 if (!ip_mroute_vif(mpctl->b_cont, ipst)) {
10486 10487 ip0dbg(("ip_mroute_vif: failed\n"));
10487 10488 }
10488 10489 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10489 10490 ip3dbg(("ip_snmp_get_mib2_virt_multi: level %d, name %d, len %d\n",
10490 10491 (int)optp->level, (int)optp->name, (int)optp->len));
10491 10492 qreply(q, mpctl);
10492 10493 return (mp2ctl);
10493 10494 }
10494 10495
10495 10496 /* Multicast routing table. */
10496 10497 static mblk_t *
10497 10498 ip_snmp_get_mib2_multi_rtable(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10498 10499 {
10499 10500 struct opthdr *optp;
10500 10501 mblk_t *mp2ctl;
10501 10502
10502 10503 /*
10503 10504 * make a copy of the original message
10504 10505 */
10505 10506 mp2ctl = copymsg(mpctl);
10506 10507
10507 10508 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10508 10509 optp->level = EXPER_DVMRP;
10509 10510 optp->name = EXPER_DVMRP_MRT;
10510 10511 if (!ip_mroute_mrt(mpctl->b_cont, ipst)) {
10511 10512 ip0dbg(("ip_mroute_mrt: failed\n"));
10512 10513 }
10513 10514 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10514 10515 ip3dbg(("ip_snmp_get_mib2_multi_rtable: level %d, name %d, len %d\n",
10515 10516 (int)optp->level, (int)optp->name, (int)optp->len));
10516 10517 qreply(q, mpctl);
10517 10518 return (mp2ctl);
10518 10519 }
10519 10520
10520 10521 /*
10521 10522 * Return ipRouteEntryTable, ipNetToMediaEntryTable, and ipRouteAttributeTable
10522 10523 * in one IRE walk.
10523 10524 */
10524 10525 static mblk_t *
10525 10526 ip_snmp_get_mib2_ip_route_media(queue_t *q, mblk_t *mpctl, int level,
10526 10527 ip_stack_t *ipst)
10527 10528 {
10528 10529 struct opthdr *optp;
10529 10530 mblk_t *mp2ctl; /* Returned */
10530 10531 mblk_t *mp3ctl; /* nettomedia */
10531 10532 mblk_t *mp4ctl; /* routeattrs */
10532 10533 iproutedata_t ird;
10533 10534 zoneid_t zoneid;
10534 10535
10535 10536 /*
10536 10537 * make copies of the original message
10537 10538 * - mp2ctl is returned unchanged to the caller for his use
10538 10539 * - mpctl is sent upstream as ipRouteEntryTable
10539 10540 * - mp3ctl is sent upstream as ipNetToMediaEntryTable
10540 10541 * - mp4ctl is sent upstream as ipRouteAttributeTable
10541 10542 */
10542 10543 mp2ctl = copymsg(mpctl);
10543 10544 mp3ctl = copymsg(mpctl);
10544 10545 mp4ctl = copymsg(mpctl);
10545 10546 if (mp3ctl == NULL || mp4ctl == NULL) {
10546 10547 freemsg(mp4ctl);
10547 10548 freemsg(mp3ctl);
10548 10549 freemsg(mp2ctl);
10549 10550 freemsg(mpctl);
10550 10551 return (NULL);
10551 10552 }
10552 10553
10553 10554 bzero(&ird, sizeof (ird));
10554 10555
10555 10556 ird.ird_route.lp_head = mpctl->b_cont;
10556 10557 ird.ird_netmedia.lp_head = mp3ctl->b_cont;
10557 10558 ird.ird_attrs.lp_head = mp4ctl->b_cont;
10558 10559 /*
10559 10560 * If the level has been set the special EXPER_IP_AND_ALL_IRES value,
10560 10561 * then also include ire_testhidden IREs and IRE_IF_CLONE. This is
10561 10562 * intended a temporary solution until a proper MIB API is provided
10562 10563 * that provides complete filtering/caller-opt-in.
10563 10564 */
10564 10565 if (level == EXPER_IP_AND_ALL_IRES)
10565 10566 ird.ird_flags |= IRD_REPORT_ALL;
10566 10567
10567 10568 zoneid = Q_TO_CONN(q)->conn_zoneid;
10568 10569 ire_walk_v4(ip_snmp_get2_v4, &ird, zoneid, ipst);
10569 10570
10570 10571 /* ipRouteEntryTable in mpctl */
10571 10572 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10572 10573 optp->level = MIB2_IP;
10573 10574 optp->name = MIB2_IP_ROUTE;
10574 10575 optp->len = msgdsize(ird.ird_route.lp_head);
10575 10576 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10576 10577 (int)optp->level, (int)optp->name, (int)optp->len));
10577 10578 qreply(q, mpctl);
10578 10579
10579 10580 /* ipNetToMediaEntryTable in mp3ctl */
10580 10581 ncec_walk(NULL, ip_snmp_get2_v4_media, &ird, ipst);
10581 10582
10582 10583 optp = (struct opthdr *)&mp3ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10583 10584 optp->level = MIB2_IP;
10584 10585 optp->name = MIB2_IP_MEDIA;
10585 10586 optp->len = msgdsize(ird.ird_netmedia.lp_head);
10586 10587 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10587 10588 (int)optp->level, (int)optp->name, (int)optp->len));
10588 10589 qreply(q, mp3ctl);
10589 10590
10590 10591 /* ipRouteAttributeTable in mp4ctl */
10591 10592 optp = (struct opthdr *)&mp4ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10592 10593 optp->level = MIB2_IP;
10593 10594 optp->name = EXPER_IP_RTATTR;
10594 10595 optp->len = msgdsize(ird.ird_attrs.lp_head);
10595 10596 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10596 10597 (int)optp->level, (int)optp->name, (int)optp->len));
10597 10598 if (optp->len == 0)
10598 10599 freemsg(mp4ctl);
10599 10600 else
10600 10601 qreply(q, mp4ctl);
10601 10602
10602 10603 return (mp2ctl);
10603 10604 }
10604 10605
10605 10606 /*
10606 10607 * Return ipv6RouteEntryTable and ipv6RouteAttributeTable in one IRE walk, and
10607 10608 * ipv6NetToMediaEntryTable in an NDP walk.
10608 10609 */
10609 10610 static mblk_t *
10610 10611 ip_snmp_get_mib2_ip6_route_media(queue_t *q, mblk_t *mpctl, int level,
10611 10612 ip_stack_t *ipst)
10612 10613 {
10613 10614 struct opthdr *optp;
10614 10615 mblk_t *mp2ctl; /* Returned */
10615 10616 mblk_t *mp3ctl; /* nettomedia */
10616 10617 mblk_t *mp4ctl; /* routeattrs */
10617 10618 iproutedata_t ird;
10618 10619 zoneid_t zoneid;
10619 10620
10620 10621 /*
10621 10622 * make copies of the original message
10622 10623 * - mp2ctl is returned unchanged to the caller for his use
10623 10624 * - mpctl is sent upstream as ipv6RouteEntryTable
10624 10625 * - mp3ctl is sent upstream as ipv6NetToMediaEntryTable
10625 10626 * - mp4ctl is sent upstream as ipv6RouteAttributeTable
10626 10627 */
10627 10628 mp2ctl = copymsg(mpctl);
10628 10629 mp3ctl = copymsg(mpctl);
10629 10630 mp4ctl = copymsg(mpctl);
10630 10631 if (mp3ctl == NULL || mp4ctl == NULL) {
10631 10632 freemsg(mp4ctl);
10632 10633 freemsg(mp3ctl);
10633 10634 freemsg(mp2ctl);
10634 10635 freemsg(mpctl);
10635 10636 return (NULL);
10636 10637 }
10637 10638
10638 10639 bzero(&ird, sizeof (ird));
10639 10640
10640 10641 ird.ird_route.lp_head = mpctl->b_cont;
10641 10642 ird.ird_netmedia.lp_head = mp3ctl->b_cont;
10642 10643 ird.ird_attrs.lp_head = mp4ctl->b_cont;
10643 10644 /*
10644 10645 * If the level has been set the special EXPER_IP_AND_ALL_IRES value,
10645 10646 * then also include ire_testhidden IREs and IRE_IF_CLONE. This is
10646 10647 * intended a temporary solution until a proper MIB API is provided
10647 10648 * that provides complete filtering/caller-opt-in.
10648 10649 */
10649 10650 if (level == EXPER_IP_AND_ALL_IRES)
10650 10651 ird.ird_flags |= IRD_REPORT_ALL;
10651 10652
10652 10653 zoneid = Q_TO_CONN(q)->conn_zoneid;
10653 10654 ire_walk_v6(ip_snmp_get2_v6_route, &ird, zoneid, ipst);
10654 10655
10655 10656 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10656 10657 optp->level = MIB2_IP6;
10657 10658 optp->name = MIB2_IP6_ROUTE;
10658 10659 optp->len = msgdsize(ird.ird_route.lp_head);
10659 10660 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10660 10661 (int)optp->level, (int)optp->name, (int)optp->len));
10661 10662 qreply(q, mpctl);
10662 10663
10663 10664 /* ipv6NetToMediaEntryTable in mp3ctl */
10664 10665 ncec_walk(NULL, ip_snmp_get2_v6_media, &ird, ipst);
10665 10666
10666 10667 optp = (struct opthdr *)&mp3ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10667 10668 optp->level = MIB2_IP6;
10668 10669 optp->name = MIB2_IP6_MEDIA;
10669 10670 optp->len = msgdsize(ird.ird_netmedia.lp_head);
10670 10671 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10671 10672 (int)optp->level, (int)optp->name, (int)optp->len));
10672 10673 qreply(q, mp3ctl);
10673 10674
10674 10675 /* ipv6RouteAttributeTable in mp4ctl */
10675 10676 optp = (struct opthdr *)&mp4ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10676 10677 optp->level = MIB2_IP6;
10677 10678 optp->name = EXPER_IP_RTATTR;
10678 10679 optp->len = msgdsize(ird.ird_attrs.lp_head);
10679 10680 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10680 10681 (int)optp->level, (int)optp->name, (int)optp->len));
10681 10682 if (optp->len == 0)
10682 10683 freemsg(mp4ctl);
10683 10684 else
10684 10685 qreply(q, mp4ctl);
10685 10686
10686 10687 return (mp2ctl);
10687 10688 }
10688 10689
10689 10690 /*
10690 10691 * IPv6 mib: One per ill
10691 10692 */
10692 10693 static mblk_t *
10693 10694 ip_snmp_get_mib2_ip6(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
10694 10695 boolean_t legacy_req)
10695 10696 {
10696 10697 struct opthdr *optp;
10697 10698 mblk_t *mp2ctl;
10698 10699 ill_t *ill;
10699 10700 ill_walk_context_t ctx;
10700 10701 mblk_t *mp_tail = NULL;
10701 10702 mib2_ipv6AddrEntry_t mae6;
10702 10703 mib2_ipIfStatsEntry_t *ise;
10703 10704 size_t ise_size, iae_size;
10704 10705
10705 10706 /*
10706 10707 * Make a copy of the original message
10707 10708 */
10708 10709 mp2ctl = copymsg(mpctl);
10709 10710
10710 10711 /* fixed length IPv6 structure ... */
10711 10712
10712 10713 if (legacy_req) {
10713 10714 ise_size = LEGACY_MIB_SIZE(&ipst->ips_ip6_mib,
10714 10715 mib2_ipIfStatsEntry_t);
10715 10716 iae_size = LEGACY_MIB_SIZE(&mae6, mib2_ipv6AddrEntry_t);
10716 10717 } else {
10717 10718 ise_size = sizeof (mib2_ipIfStatsEntry_t);
10718 10719 iae_size = sizeof (mib2_ipv6AddrEntry_t);
10719 10720 }
10720 10721
10721 10722 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10722 10723 optp->level = MIB2_IP6;
10723 10724 optp->name = 0;
10724 10725 /* Include "unknown interface" ip6_mib */
10725 10726 ipst->ips_ip6_mib.ipIfStatsIPVersion = MIB2_INETADDRESSTYPE_ipv6;
10726 10727 ipst->ips_ip6_mib.ipIfStatsIfIndex =
10727 10728 MIB2_UNKNOWN_INTERFACE; /* Flag to netstat */
10728 10729 SET_MIB(ipst->ips_ip6_mib.ipIfStatsForwarding,
10729 10730 ipst->ips_ipv6_forwarding ? 1 : 2);
10730 10731 SET_MIB(ipst->ips_ip6_mib.ipIfStatsDefaultHopLimit,
10731 10732 ipst->ips_ipv6_def_hops);
10732 10733 SET_MIB(ipst->ips_ip6_mib.ipIfStatsEntrySize,
10733 10734 sizeof (mib2_ipIfStatsEntry_t));
10734 10735 SET_MIB(ipst->ips_ip6_mib.ipIfStatsAddrEntrySize,
10735 10736 sizeof (mib2_ipv6AddrEntry_t));
10736 10737 SET_MIB(ipst->ips_ip6_mib.ipIfStatsRouteEntrySize,
10737 10738 sizeof (mib2_ipv6RouteEntry_t));
10738 10739 SET_MIB(ipst->ips_ip6_mib.ipIfStatsNetToMediaEntrySize,
10739 10740 sizeof (mib2_ipv6NetToMediaEntry_t));
10740 10741 SET_MIB(ipst->ips_ip6_mib.ipIfStatsMemberEntrySize,
10741 10742 sizeof (ipv6_member_t));
10742 10743 SET_MIB(ipst->ips_ip6_mib.ipIfStatsGroupSourceEntrySize,
10743 10744 sizeof (ipv6_grpsrc_t));
10744 10745
10745 10746 /*
10746 10747 * Synchronize 64- and 32-bit counters
10747 10748 */
10748 10749 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInReceives,
10749 10750 ipIfStatsHCInReceives);
10750 10751 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInDelivers,
10751 10752 ipIfStatsHCInDelivers);
10752 10753 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutRequests,
10753 10754 ipIfStatsHCOutRequests);
10754 10755 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutForwDatagrams,
10755 10756 ipIfStatsHCOutForwDatagrams);
10756 10757 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutMcastPkts,
10757 10758 ipIfStatsHCOutMcastPkts);
10758 10759 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInMcastPkts,
10759 10760 ipIfStatsHCInMcastPkts);
10760 10761
10761 10762 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10762 10763 (char *)&ipst->ips_ip6_mib, (int)ise_size)) {
10763 10764 ip1dbg(("ip_snmp_get_mib2_ip6: failed to allocate %u bytes\n",
10764 10765 (uint_t)ise_size));
10765 10766 } else if (legacy_req) {
10766 10767 /* Adjust the EntrySize fields for legacy requests. */
10767 10768 ise =
10768 10769 (mib2_ipIfStatsEntry_t *)(mp_tail->b_wptr - (int)ise_size);
10769 10770 SET_MIB(ise->ipIfStatsEntrySize, ise_size);
10770 10771 SET_MIB(ise->ipIfStatsAddrEntrySize, iae_size);
10771 10772 }
10772 10773
10773 10774 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10774 10775 ill = ILL_START_WALK_V6(&ctx, ipst);
10775 10776 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10776 10777 ill->ill_ip_mib->ipIfStatsIfIndex =
10777 10778 ill->ill_phyint->phyint_ifindex;
10778 10779 SET_MIB(ill->ill_ip_mib->ipIfStatsForwarding,
10779 10780 ipst->ips_ipv6_forwarding ? 1 : 2);
10780 10781 SET_MIB(ill->ill_ip_mib->ipIfStatsDefaultHopLimit,
10781 10782 ill->ill_max_hops);
10782 10783
10783 10784 /*
10784 10785 * Synchronize 64- and 32-bit counters
10785 10786 */
10786 10787 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInReceives,
10787 10788 ipIfStatsHCInReceives);
10788 10789 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInDelivers,
10789 10790 ipIfStatsHCInDelivers);
10790 10791 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutRequests,
10791 10792 ipIfStatsHCOutRequests);
10792 10793 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutForwDatagrams,
10793 10794 ipIfStatsHCOutForwDatagrams);
10794 10795 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutMcastPkts,
10795 10796 ipIfStatsHCOutMcastPkts);
10796 10797 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInMcastPkts,
10797 10798 ipIfStatsHCInMcastPkts);
10798 10799
10799 10800 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10800 10801 (char *)ill->ill_ip_mib, (int)ise_size)) {
10801 10802 ip1dbg(("ip_snmp_get_mib2_ip6: failed to allocate "
10802 10803 "%u bytes\n", (uint_t)ise_size));
10803 10804 } else if (legacy_req) {
10804 10805 /* Adjust the EntrySize fields for legacy requests. */
10805 10806 ise = (mib2_ipIfStatsEntry_t *)(mp_tail->b_wptr -
10806 10807 (int)ise_size);
10807 10808 SET_MIB(ise->ipIfStatsEntrySize, ise_size);
10808 10809 SET_MIB(ise->ipIfStatsAddrEntrySize, iae_size);
10809 10810 }
10810 10811 }
10811 10812 rw_exit(&ipst->ips_ill_g_lock);
10812 10813
10813 10814 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10814 10815 ip3dbg(("ip_snmp_get_mib2_ip6: level %d, name %d, len %d\n",
10815 10816 (int)optp->level, (int)optp->name, (int)optp->len));
10816 10817 qreply(q, mpctl);
10817 10818 return (mp2ctl);
10818 10819 }
10819 10820
10820 10821 /*
10821 10822 * ICMPv6 mib: One per ill
10822 10823 */
10823 10824 static mblk_t *
10824 10825 ip_snmp_get_mib2_icmp6(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10825 10826 {
10826 10827 struct opthdr *optp;
10827 10828 mblk_t *mp2ctl;
10828 10829 ill_t *ill;
10829 10830 ill_walk_context_t ctx;
10830 10831 mblk_t *mp_tail = NULL;
10831 10832 /*
10832 10833 * Make a copy of the original message
10833 10834 */
10834 10835 mp2ctl = copymsg(mpctl);
10835 10836
10836 10837 /* fixed length ICMPv6 structure ... */
10837 10838
10838 10839 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10839 10840 optp->level = MIB2_ICMP6;
10840 10841 optp->name = 0;
10841 10842 /* Include "unknown interface" icmp6_mib */
10842 10843 ipst->ips_icmp6_mib.ipv6IfIcmpIfIndex =
10843 10844 MIB2_UNKNOWN_INTERFACE; /* netstat flag */
10844 10845 ipst->ips_icmp6_mib.ipv6IfIcmpEntrySize =
10845 10846 sizeof (mib2_ipv6IfIcmpEntry_t);
10846 10847 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10847 10848 (char *)&ipst->ips_icmp6_mib,
10848 10849 (int)sizeof (ipst->ips_icmp6_mib))) {
10849 10850 ip1dbg(("ip_snmp_get_mib2_icmp6: failed to allocate %u bytes\n",
10850 10851 (uint_t)sizeof (ipst->ips_icmp6_mib)));
10851 10852 }
10852 10853
10853 10854 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10854 10855 ill = ILL_START_WALK_V6(&ctx, ipst);
10855 10856 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10856 10857 ill->ill_icmp6_mib->ipv6IfIcmpIfIndex =
10857 10858 ill->ill_phyint->phyint_ifindex;
10858 10859 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10859 10860 (char *)ill->ill_icmp6_mib,
10860 10861 (int)sizeof (*ill->ill_icmp6_mib))) {
10861 10862 ip1dbg(("ip_snmp_get_mib2_icmp6: failed to allocate "
10862 10863 "%u bytes\n",
10863 10864 (uint_t)sizeof (*ill->ill_icmp6_mib)));
10864 10865 }
10865 10866 }
10866 10867 rw_exit(&ipst->ips_ill_g_lock);
10867 10868
10868 10869 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10869 10870 ip3dbg(("ip_snmp_get_mib2_icmp6: level %d, name %d, len %d\n",
10870 10871 (int)optp->level, (int)optp->name, (int)optp->len));
10871 10872 qreply(q, mpctl);
10872 10873 return (mp2ctl);
10873 10874 }
10874 10875
10875 10876 /*
10876 10877 * ire_walk routine to create both ipRouteEntryTable and
10877 10878 * ipRouteAttributeTable in one IRE walk
10878 10879 */
10879 10880 static void
10880 10881 ip_snmp_get2_v4(ire_t *ire, iproutedata_t *ird)
10881 10882 {
10882 10883 ill_t *ill;
10883 10884 mib2_ipRouteEntry_t *re;
10884 10885 mib2_ipAttributeEntry_t iaes;
10885 10886 tsol_ire_gw_secattr_t *attrp;
10886 10887 tsol_gc_t *gc = NULL;
10887 10888 tsol_gcgrp_t *gcgrp = NULL;
10888 10889 ip_stack_t *ipst = ire->ire_ipst;
10889 10890
10890 10891 ASSERT(ire->ire_ipversion == IPV4_VERSION);
10891 10892
10892 10893 if (!(ird->ird_flags & IRD_REPORT_ALL)) {
10893 10894 if (ire->ire_testhidden)
10894 10895 return;
10895 10896 if (ire->ire_type & IRE_IF_CLONE)
10896 10897 return;
10897 10898 }
10898 10899
10899 10900 if ((re = kmem_zalloc(sizeof (*re), KM_NOSLEEP)) == NULL)
10900 10901 return;
10901 10902
10902 10903 if ((attrp = ire->ire_gw_secattr) != NULL) {
10903 10904 mutex_enter(&attrp->igsa_lock);
10904 10905 if ((gc = attrp->igsa_gc) != NULL) {
10905 10906 gcgrp = gc->gc_grp;
10906 10907 ASSERT(gcgrp != NULL);
10907 10908 rw_enter(&gcgrp->gcgrp_rwlock, RW_READER);
10908 10909 }
10909 10910 mutex_exit(&attrp->igsa_lock);
10910 10911 }
10911 10912 /*
10912 10913 * Return all IRE types for route table... let caller pick and choose
10913 10914 */
10914 10915 re->ipRouteDest = ire->ire_addr;
10915 10916 ill = ire->ire_ill;
10916 10917 re->ipRouteIfIndex.o_length = 0;
10917 10918 if (ill != NULL) {
10918 10919 ill_get_name(ill, re->ipRouteIfIndex.o_bytes, OCTET_LENGTH);
10919 10920 re->ipRouteIfIndex.o_length =
10920 10921 mi_strlen(re->ipRouteIfIndex.o_bytes);
10921 10922 }
10922 10923 re->ipRouteMetric1 = -1;
10923 10924 re->ipRouteMetric2 = -1;
10924 10925 re->ipRouteMetric3 = -1;
10925 10926 re->ipRouteMetric4 = -1;
10926 10927
10927 10928 re->ipRouteNextHop = ire->ire_gateway_addr;
10928 10929 /* indirect(4), direct(3), or invalid(2) */
10929 10930 if (ire->ire_flags & (RTF_REJECT | RTF_BLACKHOLE))
10930 10931 re->ipRouteType = 2;
10931 10932 else if (ire->ire_type & IRE_ONLINK)
10932 10933 re->ipRouteType = 3;
10933 10934 else
10934 10935 re->ipRouteType = 4;
10935 10936
10936 10937 re->ipRouteProto = -1;
10937 10938 re->ipRouteAge = gethrestime_sec() - ire->ire_create_time;
10938 10939 re->ipRouteMask = ire->ire_mask;
10939 10940 re->ipRouteMetric5 = -1;
10940 10941 re->ipRouteInfo.re_max_frag = ire->ire_metrics.iulp_mtu;
10941 10942 if (ire->ire_ill != NULL && re->ipRouteInfo.re_max_frag == 0)
10942 10943 re->ipRouteInfo.re_max_frag = ire->ire_ill->ill_mtu;
10943 10944
10944 10945 re->ipRouteInfo.re_frag_flag = 0;
10945 10946 re->ipRouteInfo.re_rtt = 0;
10946 10947 re->ipRouteInfo.re_src_addr = 0;
10947 10948 re->ipRouteInfo.re_ref = ire->ire_refcnt;
10948 10949 re->ipRouteInfo.re_obpkt = ire->ire_ob_pkt_count;
10949 10950 re->ipRouteInfo.re_ibpkt = ire->ire_ib_pkt_count;
10950 10951 re->ipRouteInfo.re_flags = ire->ire_flags;
10951 10952
10952 10953 /* Add the IRE_IF_CLONE's counters to their parent IRE_INTERFACE */
10953 10954 if (ire->ire_type & IRE_INTERFACE) {
10954 10955 ire_t *child;
10955 10956
10956 10957 rw_enter(&ipst->ips_ire_dep_lock, RW_READER);
10957 10958 child = ire->ire_dep_children;
10958 10959 while (child != NULL) {
10959 10960 re->ipRouteInfo.re_obpkt += child->ire_ob_pkt_count;
10960 10961 re->ipRouteInfo.re_ibpkt += child->ire_ib_pkt_count;
10961 10962 child = child->ire_dep_sib_next;
10962 10963 }
10963 10964 rw_exit(&ipst->ips_ire_dep_lock);
10964 10965 }
10965 10966
10966 10967 if (ire->ire_flags & RTF_DYNAMIC) {
10967 10968 re->ipRouteInfo.re_ire_type = IRE_HOST_REDIRECT;
10968 10969 } else {
10969 10970 re->ipRouteInfo.re_ire_type = ire->ire_type;
10970 10971 }
10971 10972
10972 10973 if (!snmp_append_data2(ird->ird_route.lp_head, &ird->ird_route.lp_tail,
10973 10974 (char *)re, (int)sizeof (*re))) {
10974 10975 ip1dbg(("ip_snmp_get2_v4: failed to allocate %u bytes\n",
10975 10976 (uint_t)sizeof (*re)));
10976 10977 }
10977 10978
10978 10979 if (gc != NULL) {
10979 10980 iaes.iae_routeidx = ird->ird_idx;
10980 10981 iaes.iae_doi = gc->gc_db->gcdb_doi;
10981 10982 iaes.iae_slrange = gc->gc_db->gcdb_slrange;
10982 10983
10983 10984 if (!snmp_append_data2(ird->ird_attrs.lp_head,
10984 10985 &ird->ird_attrs.lp_tail, (char *)&iaes, sizeof (iaes))) {
10985 10986 ip1dbg(("ip_snmp_get2_v4: failed to allocate %u "
10986 10987 "bytes\n", (uint_t)sizeof (iaes)));
10987 10988 }
10988 10989 }
10989 10990
10990 10991 /* bump route index for next pass */
10991 10992 ird->ird_idx++;
10992 10993
10993 10994 kmem_free(re, sizeof (*re));
10994 10995 if (gcgrp != NULL)
10995 10996 rw_exit(&gcgrp->gcgrp_rwlock);
10996 10997 }
10997 10998
10998 10999 /*
10999 11000 * ire_walk routine to create ipv6RouteEntryTable and ipRouteEntryTable.
11000 11001 */
11001 11002 static void
11002 11003 ip_snmp_get2_v6_route(ire_t *ire, iproutedata_t *ird)
11003 11004 {
11004 11005 ill_t *ill;
11005 11006 mib2_ipv6RouteEntry_t *re;
11006 11007 mib2_ipAttributeEntry_t iaes;
11007 11008 tsol_ire_gw_secattr_t *attrp;
11008 11009 tsol_gc_t *gc = NULL;
11009 11010 tsol_gcgrp_t *gcgrp = NULL;
11010 11011 ip_stack_t *ipst = ire->ire_ipst;
11011 11012
11012 11013 ASSERT(ire->ire_ipversion == IPV6_VERSION);
11013 11014
11014 11015 if (!(ird->ird_flags & IRD_REPORT_ALL)) {
11015 11016 if (ire->ire_testhidden)
11016 11017 return;
11017 11018 if (ire->ire_type & IRE_IF_CLONE)
11018 11019 return;
11019 11020 }
11020 11021
11021 11022 if ((re = kmem_zalloc(sizeof (*re), KM_NOSLEEP)) == NULL)
11022 11023 return;
11023 11024
11024 11025 if ((attrp = ire->ire_gw_secattr) != NULL) {
11025 11026 mutex_enter(&attrp->igsa_lock);
11026 11027 if ((gc = attrp->igsa_gc) != NULL) {
11027 11028 gcgrp = gc->gc_grp;
11028 11029 ASSERT(gcgrp != NULL);
11029 11030 rw_enter(&gcgrp->gcgrp_rwlock, RW_READER);
11030 11031 }
11031 11032 mutex_exit(&attrp->igsa_lock);
11032 11033 }
11033 11034 /*
11034 11035 * Return all IRE types for route table... let caller pick and choose
11035 11036 */
11036 11037 re->ipv6RouteDest = ire->ire_addr_v6;
11037 11038 re->ipv6RoutePfxLength = ip_mask_to_plen_v6(&ire->ire_mask_v6);
11038 11039 re->ipv6RouteIndex = 0; /* Unique when multiple with same dest/plen */
11039 11040 re->ipv6RouteIfIndex.o_length = 0;
11040 11041 ill = ire->ire_ill;
11041 11042 if (ill != NULL) {
11042 11043 ill_get_name(ill, re->ipv6RouteIfIndex.o_bytes, OCTET_LENGTH);
11043 11044 re->ipv6RouteIfIndex.o_length =
11044 11045 mi_strlen(re->ipv6RouteIfIndex.o_bytes);
11045 11046 }
11046 11047
11047 11048 ASSERT(!(ire->ire_type & IRE_BROADCAST));
11048 11049
11049 11050 mutex_enter(&ire->ire_lock);
11050 11051 re->ipv6RouteNextHop = ire->ire_gateway_addr_v6;
11051 11052 mutex_exit(&ire->ire_lock);
11052 11053
11053 11054 /* remote(4), local(3), or discard(2) */
11054 11055 if (ire->ire_flags & (RTF_REJECT | RTF_BLACKHOLE))
11055 11056 re->ipv6RouteType = 2;
11056 11057 else if (ire->ire_type & IRE_ONLINK)
11057 11058 re->ipv6RouteType = 3;
11058 11059 else
11059 11060 re->ipv6RouteType = 4;
11060 11061
11061 11062 re->ipv6RouteProtocol = -1;
11062 11063 re->ipv6RoutePolicy = 0;
11063 11064 re->ipv6RouteAge = gethrestime_sec() - ire->ire_create_time;
11064 11065 re->ipv6RouteNextHopRDI = 0;
11065 11066 re->ipv6RouteWeight = 0;
11066 11067 re->ipv6RouteMetric = 0;
11067 11068 re->ipv6RouteInfo.re_max_frag = ire->ire_metrics.iulp_mtu;
11068 11069 if (ire->ire_ill != NULL && re->ipv6RouteInfo.re_max_frag == 0)
11069 11070 re->ipv6RouteInfo.re_max_frag = ire->ire_ill->ill_mtu;
11070 11071
11071 11072 re->ipv6RouteInfo.re_frag_flag = 0;
11072 11073 re->ipv6RouteInfo.re_rtt = 0;
11073 11074 re->ipv6RouteInfo.re_src_addr = ipv6_all_zeros;
11074 11075 re->ipv6RouteInfo.re_obpkt = ire->ire_ob_pkt_count;
11075 11076 re->ipv6RouteInfo.re_ibpkt = ire->ire_ib_pkt_count;
11076 11077 re->ipv6RouteInfo.re_ref = ire->ire_refcnt;
11077 11078 re->ipv6RouteInfo.re_flags = ire->ire_flags;
11078 11079
11079 11080 /* Add the IRE_IF_CLONE's counters to their parent IRE_INTERFACE */
11080 11081 if (ire->ire_type & IRE_INTERFACE) {
11081 11082 ire_t *child;
11082 11083
11083 11084 rw_enter(&ipst->ips_ire_dep_lock, RW_READER);
11084 11085 child = ire->ire_dep_children;
11085 11086 while (child != NULL) {
11086 11087 re->ipv6RouteInfo.re_obpkt += child->ire_ob_pkt_count;
11087 11088 re->ipv6RouteInfo.re_ibpkt += child->ire_ib_pkt_count;
11088 11089 child = child->ire_dep_sib_next;
11089 11090 }
11090 11091 rw_exit(&ipst->ips_ire_dep_lock);
11091 11092 }
11092 11093 if (ire->ire_flags & RTF_DYNAMIC) {
11093 11094 re->ipv6RouteInfo.re_ire_type = IRE_HOST_REDIRECT;
11094 11095 } else {
11095 11096 re->ipv6RouteInfo.re_ire_type = ire->ire_type;
11096 11097 }
11097 11098
11098 11099 if (!snmp_append_data2(ird->ird_route.lp_head, &ird->ird_route.lp_tail,
11099 11100 (char *)re, (int)sizeof (*re))) {
11100 11101 ip1dbg(("ip_snmp_get2_v6: failed to allocate %u bytes\n",
11101 11102 (uint_t)sizeof (*re)));
11102 11103 }
11103 11104
11104 11105 if (gc != NULL) {
11105 11106 iaes.iae_routeidx = ird->ird_idx;
11106 11107 iaes.iae_doi = gc->gc_db->gcdb_doi;
11107 11108 iaes.iae_slrange = gc->gc_db->gcdb_slrange;
11108 11109
11109 11110 if (!snmp_append_data2(ird->ird_attrs.lp_head,
11110 11111 &ird->ird_attrs.lp_tail, (char *)&iaes, sizeof (iaes))) {
11111 11112 ip1dbg(("ip_snmp_get2_v6: failed to allocate %u "
11112 11113 "bytes\n", (uint_t)sizeof (iaes)));
11113 11114 }
11114 11115 }
11115 11116
11116 11117 /* bump route index for next pass */
11117 11118 ird->ird_idx++;
11118 11119
11119 11120 kmem_free(re, sizeof (*re));
11120 11121 if (gcgrp != NULL)
11121 11122 rw_exit(&gcgrp->gcgrp_rwlock);
11122 11123 }
11123 11124
11124 11125 /*
11125 11126 * ncec_walk routine to create ipv6NetToMediaEntryTable
11126 11127 */
11127 11128 static int
11128 11129 ip_snmp_get2_v6_media(ncec_t *ncec, iproutedata_t *ird)
11129 11130 {
11130 11131 ill_t *ill;
11131 11132 mib2_ipv6NetToMediaEntry_t ntme;
11132 11133
11133 11134 ill = ncec->ncec_ill;
11134 11135 /* skip arpce entries, and loopback ncec entries */
11135 11136 if (ill->ill_isv6 == B_FALSE || ill->ill_net_type == IRE_LOOPBACK)
11136 11137 return (0);
11137 11138 /*
11138 11139 * Neighbor cache entry attached to IRE with on-link
11139 11140 * destination.
11140 11141 * We report all IPMP groups on ncec_ill which is normally the upper.
11141 11142 */
11142 11143 ntme.ipv6NetToMediaIfIndex = ill->ill_phyint->phyint_ifindex;
11143 11144 ntme.ipv6NetToMediaNetAddress = ncec->ncec_addr;
11144 11145 ntme.ipv6NetToMediaPhysAddress.o_length = ill->ill_phys_addr_length;
11145 11146 if (ncec->ncec_lladdr != NULL) {
11146 11147 bcopy(ncec->ncec_lladdr, ntme.ipv6NetToMediaPhysAddress.o_bytes,
11147 11148 ntme.ipv6NetToMediaPhysAddress.o_length);
11148 11149 }
11149 11150 /*
11150 11151 * Note: Returns ND_* states. Should be:
11151 11152 * reachable(1), stale(2), delay(3), probe(4),
11152 11153 * invalid(5), unknown(6)
11153 11154 */
11154 11155 ntme.ipv6NetToMediaState = ncec->ncec_state;
11155 11156 ntme.ipv6NetToMediaLastUpdated = 0;
11156 11157
11157 11158 /* other(1), dynamic(2), static(3), local(4) */
11158 11159 if (NCE_MYADDR(ncec)) {
11159 11160 ntme.ipv6NetToMediaType = 4;
11160 11161 } else if (ncec->ncec_flags & NCE_F_PUBLISH) {
11161 11162 ntme.ipv6NetToMediaType = 1; /* proxy */
11162 11163 } else if (ncec->ncec_flags & NCE_F_STATIC) {
11163 11164 ntme.ipv6NetToMediaType = 3;
11164 11165 } else if (ncec->ncec_flags & (NCE_F_MCAST|NCE_F_BCAST)) {
11165 11166 ntme.ipv6NetToMediaType = 1;
11166 11167 } else {
11167 11168 ntme.ipv6NetToMediaType = 2;
11168 11169 }
11169 11170
11170 11171 if (!snmp_append_data2(ird->ird_netmedia.lp_head,
11171 11172 &ird->ird_netmedia.lp_tail, (char *)&ntme, sizeof (ntme))) {
11172 11173 ip1dbg(("ip_snmp_get2_v6_media: failed to allocate %u bytes\n",
11173 11174 (uint_t)sizeof (ntme)));
11174 11175 }
11175 11176 return (0);
11176 11177 }
11177 11178
11178 11179 int
11179 11180 nce2ace(ncec_t *ncec)
11180 11181 {
11181 11182 int flags = 0;
11182 11183
11183 11184 if (NCE_ISREACHABLE(ncec))
11184 11185 flags |= ACE_F_RESOLVED;
11185 11186 if (ncec->ncec_flags & NCE_F_AUTHORITY)
11186 11187 flags |= ACE_F_AUTHORITY;
11187 11188 if (ncec->ncec_flags & NCE_F_PUBLISH)
11188 11189 flags |= ACE_F_PUBLISH;
11189 11190 if ((ncec->ncec_flags & NCE_F_NONUD) != 0)
11190 11191 flags |= ACE_F_PERMANENT;
11191 11192 if (NCE_MYADDR(ncec))
11192 11193 flags |= (ACE_F_MYADDR | ACE_F_AUTHORITY);
11193 11194 if (ncec->ncec_flags & NCE_F_UNVERIFIED)
11194 11195 flags |= ACE_F_UNVERIFIED;
11195 11196 if (ncec->ncec_flags & NCE_F_AUTHORITY)
11196 11197 flags |= ACE_F_AUTHORITY;
11197 11198 if (ncec->ncec_flags & NCE_F_DELAYED)
11198 11199 flags |= ACE_F_DELAYED;
11199 11200 return (flags);
11200 11201 }
11201 11202
11202 11203 /*
11203 11204 * ncec_walk routine to create ipNetToMediaEntryTable
11204 11205 */
11205 11206 static int
11206 11207 ip_snmp_get2_v4_media(ncec_t *ncec, iproutedata_t *ird)
11207 11208 {
11208 11209 ill_t *ill;
11209 11210 mib2_ipNetToMediaEntry_t ntme;
11210 11211 const char *name = "unknown";
11211 11212 ipaddr_t ncec_addr;
11212 11213
11213 11214 ill = ncec->ncec_ill;
11214 11215 if (ill->ill_isv6 || (ncec->ncec_flags & NCE_F_BCAST) ||
11215 11216 ill->ill_net_type == IRE_LOOPBACK)
11216 11217 return (0);
11217 11218
11218 11219 /* We report all IPMP groups on ncec_ill which is normally the upper. */
11219 11220 name = ill->ill_name;
11220 11221 /* Based on RFC 4293: other(1), inval(2), dyn(3), stat(4) */
11221 11222 if (NCE_MYADDR(ncec)) {
11222 11223 ntme.ipNetToMediaType = 4;
11223 11224 } else if (ncec->ncec_flags & (NCE_F_MCAST|NCE_F_BCAST|NCE_F_PUBLISH)) {
11224 11225 ntme.ipNetToMediaType = 1;
11225 11226 } else {
11226 11227 ntme.ipNetToMediaType = 3;
11227 11228 }
11228 11229 ntme.ipNetToMediaIfIndex.o_length = MIN(OCTET_LENGTH, strlen(name));
11229 11230 bcopy(name, ntme.ipNetToMediaIfIndex.o_bytes,
11230 11231 ntme.ipNetToMediaIfIndex.o_length);
11231 11232
11232 11233 IN6_V4MAPPED_TO_IPADDR(&ncec->ncec_addr, ncec_addr);
11233 11234 bcopy(&ncec_addr, &ntme.ipNetToMediaNetAddress, sizeof (ncec_addr));
11234 11235
11235 11236 ntme.ipNetToMediaInfo.ntm_mask.o_length = sizeof (ipaddr_t);
11236 11237 ncec_addr = INADDR_BROADCAST;
11237 11238 bcopy(&ncec_addr, ntme.ipNetToMediaInfo.ntm_mask.o_bytes,
11238 11239 sizeof (ncec_addr));
11239 11240 /*
11240 11241 * map all the flags to the ACE counterpart.
11241 11242 */
11242 11243 ntme.ipNetToMediaInfo.ntm_flags = nce2ace(ncec);
11243 11244
11244 11245 ntme.ipNetToMediaPhysAddress.o_length =
11245 11246 MIN(OCTET_LENGTH, ill->ill_phys_addr_length);
11246 11247
11247 11248 if (!NCE_ISREACHABLE(ncec))
11248 11249 ntme.ipNetToMediaPhysAddress.o_length = 0;
11249 11250 else {
11250 11251 if (ncec->ncec_lladdr != NULL) {
11251 11252 bcopy(ncec->ncec_lladdr,
11252 11253 ntme.ipNetToMediaPhysAddress.o_bytes,
11253 11254 ntme.ipNetToMediaPhysAddress.o_length);
11254 11255 }
11255 11256 }
11256 11257
11257 11258 if (!snmp_append_data2(ird->ird_netmedia.lp_head,
11258 11259 &ird->ird_netmedia.lp_tail, (char *)&ntme, sizeof (ntme))) {
11259 11260 ip1dbg(("ip_snmp_get2_v4_media: failed to allocate %u bytes\n",
11260 11261 (uint_t)sizeof (ntme)));
11261 11262 }
11262 11263 return (0);
11263 11264 }
11264 11265
11265 11266 /*
11266 11267 * return (0) if invalid set request, 1 otherwise, including non-tcp requests
11267 11268 */
11268 11269 /* ARGSUSED */
11269 11270 int
11270 11271 ip_snmp_set(queue_t *q, int level, int name, uchar_t *ptr, int len)
11271 11272 {
11272 11273 switch (level) {
11273 11274 case MIB2_IP:
11274 11275 case MIB2_ICMP:
11275 11276 switch (name) {
11276 11277 default:
11277 11278 break;
11278 11279 }
11279 11280 return (1);
11280 11281 default:
11281 11282 return (1);
11282 11283 }
11283 11284 }
11284 11285
11285 11286 /*
11286 11287 * When there exists both a 64- and 32-bit counter of a particular type
11287 11288 * (i.e., InReceives), only the 64-bit counters are added.
11288 11289 */
11289 11290 void
11290 11291 ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *o1, mib2_ipIfStatsEntry_t *o2)
11291 11292 {
11292 11293 UPDATE_MIB(o1, ipIfStatsInHdrErrors, o2->ipIfStatsInHdrErrors);
11293 11294 UPDATE_MIB(o1, ipIfStatsInTooBigErrors, o2->ipIfStatsInTooBigErrors);
11294 11295 UPDATE_MIB(o1, ipIfStatsInNoRoutes, o2->ipIfStatsInNoRoutes);
11295 11296 UPDATE_MIB(o1, ipIfStatsInAddrErrors, o2->ipIfStatsInAddrErrors);
11296 11297 UPDATE_MIB(o1, ipIfStatsInUnknownProtos, o2->ipIfStatsInUnknownProtos);
11297 11298 UPDATE_MIB(o1, ipIfStatsInTruncatedPkts, o2->ipIfStatsInTruncatedPkts);
11298 11299 UPDATE_MIB(o1, ipIfStatsInDiscards, o2->ipIfStatsInDiscards);
11299 11300 UPDATE_MIB(o1, ipIfStatsOutDiscards, o2->ipIfStatsOutDiscards);
11300 11301 UPDATE_MIB(o1, ipIfStatsOutFragOKs, o2->ipIfStatsOutFragOKs);
11301 11302 UPDATE_MIB(o1, ipIfStatsOutFragFails, o2->ipIfStatsOutFragFails);
11302 11303 UPDATE_MIB(o1, ipIfStatsOutFragCreates, o2->ipIfStatsOutFragCreates);
11303 11304 UPDATE_MIB(o1, ipIfStatsReasmReqds, o2->ipIfStatsReasmReqds);
11304 11305 UPDATE_MIB(o1, ipIfStatsReasmOKs, o2->ipIfStatsReasmOKs);
11305 11306 UPDATE_MIB(o1, ipIfStatsReasmFails, o2->ipIfStatsReasmFails);
11306 11307 UPDATE_MIB(o1, ipIfStatsOutNoRoutes, o2->ipIfStatsOutNoRoutes);
11307 11308 UPDATE_MIB(o1, ipIfStatsReasmDuplicates, o2->ipIfStatsReasmDuplicates);
11308 11309 UPDATE_MIB(o1, ipIfStatsReasmPartDups, o2->ipIfStatsReasmPartDups);
11309 11310 UPDATE_MIB(o1, ipIfStatsForwProhibits, o2->ipIfStatsForwProhibits);
11310 11311 UPDATE_MIB(o1, udpInCksumErrs, o2->udpInCksumErrs);
11311 11312 UPDATE_MIB(o1, udpInOverflows, o2->udpInOverflows);
11312 11313 UPDATE_MIB(o1, rawipInOverflows, o2->rawipInOverflows);
11313 11314 UPDATE_MIB(o1, ipIfStatsInWrongIPVersion,
11314 11315 o2->ipIfStatsInWrongIPVersion);
11315 11316 UPDATE_MIB(o1, ipIfStatsOutWrongIPVersion,
11316 11317 o2->ipIfStatsInWrongIPVersion);
11317 11318 UPDATE_MIB(o1, ipIfStatsOutSwitchIPVersion,
11318 11319 o2->ipIfStatsOutSwitchIPVersion);
11319 11320 UPDATE_MIB(o1, ipIfStatsHCInReceives, o2->ipIfStatsHCInReceives);
11320 11321 UPDATE_MIB(o1, ipIfStatsHCInOctets, o2->ipIfStatsHCInOctets);
11321 11322 UPDATE_MIB(o1, ipIfStatsHCInForwDatagrams,
11322 11323 o2->ipIfStatsHCInForwDatagrams);
11323 11324 UPDATE_MIB(o1, ipIfStatsHCInDelivers, o2->ipIfStatsHCInDelivers);
11324 11325 UPDATE_MIB(o1, ipIfStatsHCOutRequests, o2->ipIfStatsHCOutRequests);
11325 11326 UPDATE_MIB(o1, ipIfStatsHCOutForwDatagrams,
11326 11327 o2->ipIfStatsHCOutForwDatagrams);
11327 11328 UPDATE_MIB(o1, ipIfStatsOutFragReqds, o2->ipIfStatsOutFragReqds);
11328 11329 UPDATE_MIB(o1, ipIfStatsHCOutTransmits, o2->ipIfStatsHCOutTransmits);
11329 11330 UPDATE_MIB(o1, ipIfStatsHCOutOctets, o2->ipIfStatsHCOutOctets);
11330 11331 UPDATE_MIB(o1, ipIfStatsHCInMcastPkts, o2->ipIfStatsHCInMcastPkts);
11331 11332 UPDATE_MIB(o1, ipIfStatsHCInMcastOctets, o2->ipIfStatsHCInMcastOctets);
11332 11333 UPDATE_MIB(o1, ipIfStatsHCOutMcastPkts, o2->ipIfStatsHCOutMcastPkts);
11333 11334 UPDATE_MIB(o1, ipIfStatsHCOutMcastOctets,
11334 11335 o2->ipIfStatsHCOutMcastOctets);
11335 11336 UPDATE_MIB(o1, ipIfStatsHCInBcastPkts, o2->ipIfStatsHCInBcastPkts);
11336 11337 UPDATE_MIB(o1, ipIfStatsHCOutBcastPkts, o2->ipIfStatsHCOutBcastPkts);
11337 11338 UPDATE_MIB(o1, ipsecInSucceeded, o2->ipsecInSucceeded);
11338 11339 UPDATE_MIB(o1, ipsecInFailed, o2->ipsecInFailed);
11339 11340 UPDATE_MIB(o1, ipInCksumErrs, o2->ipInCksumErrs);
11340 11341 UPDATE_MIB(o1, tcpInErrs, o2->tcpInErrs);
11341 11342 UPDATE_MIB(o1, udpNoPorts, o2->udpNoPorts);
11342 11343 }
11343 11344
11344 11345 void
11345 11346 ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *o1, mib2_ipv6IfIcmpEntry_t *o2)
11346 11347 {
11347 11348 UPDATE_MIB(o1, ipv6IfIcmpInMsgs, o2->ipv6IfIcmpInMsgs);
11348 11349 UPDATE_MIB(o1, ipv6IfIcmpInErrors, o2->ipv6IfIcmpInErrors);
11349 11350 UPDATE_MIB(o1, ipv6IfIcmpInDestUnreachs, o2->ipv6IfIcmpInDestUnreachs);
11350 11351 UPDATE_MIB(o1, ipv6IfIcmpInAdminProhibs, o2->ipv6IfIcmpInAdminProhibs);
11351 11352 UPDATE_MIB(o1, ipv6IfIcmpInTimeExcds, o2->ipv6IfIcmpInTimeExcds);
11352 11353 UPDATE_MIB(o1, ipv6IfIcmpInParmProblems, o2->ipv6IfIcmpInParmProblems);
11353 11354 UPDATE_MIB(o1, ipv6IfIcmpInPktTooBigs, o2->ipv6IfIcmpInPktTooBigs);
11354 11355 UPDATE_MIB(o1, ipv6IfIcmpInEchos, o2->ipv6IfIcmpInEchos);
11355 11356 UPDATE_MIB(o1, ipv6IfIcmpInEchoReplies, o2->ipv6IfIcmpInEchoReplies);
11356 11357 UPDATE_MIB(o1, ipv6IfIcmpInRouterSolicits,
11357 11358 o2->ipv6IfIcmpInRouterSolicits);
11358 11359 UPDATE_MIB(o1, ipv6IfIcmpInRouterAdvertisements,
11359 11360 o2->ipv6IfIcmpInRouterAdvertisements);
11360 11361 UPDATE_MIB(o1, ipv6IfIcmpInNeighborSolicits,
11361 11362 o2->ipv6IfIcmpInNeighborSolicits);
11362 11363 UPDATE_MIB(o1, ipv6IfIcmpInNeighborAdvertisements,
11363 11364 o2->ipv6IfIcmpInNeighborAdvertisements);
11364 11365 UPDATE_MIB(o1, ipv6IfIcmpInRedirects, o2->ipv6IfIcmpInRedirects);
11365 11366 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembQueries,
11366 11367 o2->ipv6IfIcmpInGroupMembQueries);
11367 11368 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembResponses,
11368 11369 o2->ipv6IfIcmpInGroupMembResponses);
11369 11370 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembReductions,
11370 11371 o2->ipv6IfIcmpInGroupMembReductions);
11371 11372 UPDATE_MIB(o1, ipv6IfIcmpOutMsgs, o2->ipv6IfIcmpOutMsgs);
11372 11373 UPDATE_MIB(o1, ipv6IfIcmpOutErrors, o2->ipv6IfIcmpOutErrors);
11373 11374 UPDATE_MIB(o1, ipv6IfIcmpOutDestUnreachs,
11374 11375 o2->ipv6IfIcmpOutDestUnreachs);
11375 11376 UPDATE_MIB(o1, ipv6IfIcmpOutAdminProhibs,
11376 11377 o2->ipv6IfIcmpOutAdminProhibs);
11377 11378 UPDATE_MIB(o1, ipv6IfIcmpOutTimeExcds, o2->ipv6IfIcmpOutTimeExcds);
11378 11379 UPDATE_MIB(o1, ipv6IfIcmpOutParmProblems,
11379 11380 o2->ipv6IfIcmpOutParmProblems);
11380 11381 UPDATE_MIB(o1, ipv6IfIcmpOutPktTooBigs, o2->ipv6IfIcmpOutPktTooBigs);
11381 11382 UPDATE_MIB(o1, ipv6IfIcmpOutEchos, o2->ipv6IfIcmpOutEchos);
11382 11383 UPDATE_MIB(o1, ipv6IfIcmpOutEchoReplies, o2->ipv6IfIcmpOutEchoReplies);
11383 11384 UPDATE_MIB(o1, ipv6IfIcmpOutRouterSolicits,
11384 11385 o2->ipv6IfIcmpOutRouterSolicits);
11385 11386 UPDATE_MIB(o1, ipv6IfIcmpOutRouterAdvertisements,
11386 11387 o2->ipv6IfIcmpOutRouterAdvertisements);
11387 11388 UPDATE_MIB(o1, ipv6IfIcmpOutNeighborSolicits,
11388 11389 o2->ipv6IfIcmpOutNeighborSolicits);
11389 11390 UPDATE_MIB(o1, ipv6IfIcmpOutNeighborAdvertisements,
11390 11391 o2->ipv6IfIcmpOutNeighborAdvertisements);
11391 11392 UPDATE_MIB(o1, ipv6IfIcmpOutRedirects, o2->ipv6IfIcmpOutRedirects);
11392 11393 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembQueries,
11393 11394 o2->ipv6IfIcmpOutGroupMembQueries);
11394 11395 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembResponses,
11395 11396 o2->ipv6IfIcmpOutGroupMembResponses);
11396 11397 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembReductions,
11397 11398 o2->ipv6IfIcmpOutGroupMembReductions);
11398 11399 UPDATE_MIB(o1, ipv6IfIcmpInOverflows, o2->ipv6IfIcmpInOverflows);
11399 11400 UPDATE_MIB(o1, ipv6IfIcmpBadHoplimit, o2->ipv6IfIcmpBadHoplimit);
11400 11401 UPDATE_MIB(o1, ipv6IfIcmpInBadNeighborAdvertisements,
11401 11402 o2->ipv6IfIcmpInBadNeighborAdvertisements);
11402 11403 UPDATE_MIB(o1, ipv6IfIcmpInBadNeighborSolicitations,
11403 11404 o2->ipv6IfIcmpInBadNeighborSolicitations);
11404 11405 UPDATE_MIB(o1, ipv6IfIcmpInBadRedirects, o2->ipv6IfIcmpInBadRedirects);
11405 11406 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembTotal,
11406 11407 o2->ipv6IfIcmpInGroupMembTotal);
11407 11408 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembBadQueries,
11408 11409 o2->ipv6IfIcmpInGroupMembBadQueries);
11409 11410 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembBadReports,
11410 11411 o2->ipv6IfIcmpInGroupMembBadReports);
11411 11412 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembOurReports,
11412 11413 o2->ipv6IfIcmpInGroupMembOurReports);
11413 11414 }
11414 11415
11415 11416 /*
11416 11417 * Called before the options are updated to check if this packet will
11417 11418 * be source routed from here.
11418 11419 * This routine assumes that the options are well formed i.e. that they
11419 11420 * have already been checked.
11420 11421 */
11421 11422 boolean_t
11422 11423 ip_source_routed(ipha_t *ipha, ip_stack_t *ipst)
11423 11424 {
11424 11425 ipoptp_t opts;
11425 11426 uchar_t *opt;
11426 11427 uint8_t optval;
11427 11428 uint8_t optlen;
11428 11429 ipaddr_t dst;
11429 11430
11430 11431 if (IS_SIMPLE_IPH(ipha)) {
11431 11432 ip2dbg(("not source routed\n"));
11432 11433 return (B_FALSE);
11433 11434 }
11434 11435 dst = ipha->ipha_dst;
11435 11436 for (optval = ipoptp_first(&opts, ipha);
11436 11437 optval != IPOPT_EOL;
11437 11438 optval = ipoptp_next(&opts)) {
11438 11439 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11439 11440 opt = opts.ipoptp_cur;
11440 11441 optlen = opts.ipoptp_len;
11441 11442 ip2dbg(("ip_source_routed: opt %d, len %d\n",
11442 11443 optval, optlen));
11443 11444 switch (optval) {
11444 11445 uint32_t off;
11445 11446 case IPOPT_SSRR:
11446 11447 case IPOPT_LSRR:
11447 11448 /*
11448 11449 * If dst is one of our addresses and there are some
11449 11450 * entries left in the source route return (true).
11450 11451 */
11451 11452 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
11452 11453 ip2dbg(("ip_source_routed: not next"
11453 11454 " source route 0x%x\n",
11454 11455 ntohl(dst)));
11455 11456 return (B_FALSE);
11456 11457 }
11457 11458 off = opt[IPOPT_OFFSET];
11458 11459 off--;
11459 11460 if (optlen < IP_ADDR_LEN ||
11460 11461 off > optlen - IP_ADDR_LEN) {
11461 11462 /* End of source route */
11462 11463 ip1dbg(("ip_source_routed: end of SR\n"));
11463 11464 return (B_FALSE);
11464 11465 }
11465 11466 return (B_TRUE);
11466 11467 }
11467 11468 }
11468 11469 ip2dbg(("not source routed\n"));
11469 11470 return (B_FALSE);
11470 11471 }
11471 11472
11472 11473 /*
11473 11474 * ip_unbind is called by the transports to remove a conn from
11474 11475 * the fanout table.
11475 11476 */
11476 11477 void
11477 11478 ip_unbind(conn_t *connp)
11478 11479 {
11479 11480
11480 11481 ASSERT(!MUTEX_HELD(&connp->conn_lock));
11481 11482
11482 11483 if (is_system_labeled() && connp->conn_anon_port) {
11483 11484 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
11484 11485 connp->conn_mlp_type, connp->conn_proto,
11485 11486 ntohs(connp->conn_lport), B_FALSE);
11486 11487 connp->conn_anon_port = 0;
11487 11488 }
11488 11489 connp->conn_mlp_type = mlptSingle;
11489 11490
11490 11491 ipcl_hash_remove(connp);
11491 11492 }
11492 11493
11493 11494 /*
11494 11495 * Used for deciding the MSS size for the upper layer. Thus
11495 11496 * we need to check the outbound policy values in the conn.
11496 11497 */
11497 11498 int
11498 11499 conn_ipsec_length(conn_t *connp)
11499 11500 {
11500 11501 ipsec_latch_t *ipl;
11501 11502
11502 11503 ipl = connp->conn_latch;
11503 11504 if (ipl == NULL)
11504 11505 return (0);
11505 11506
11506 11507 if (connp->conn_ixa->ixa_ipsec_policy == NULL)
11507 11508 return (0);
11508 11509
11509 11510 return (connp->conn_ixa->ixa_ipsec_policy->ipsp_act->ipa_ovhd);
11510 11511 }
11511 11512
11512 11513 /*
11513 11514 * Returns an estimate of the IPsec headers size. This is used if
11514 11515 * we don't want to call into IPsec to get the exact size.
11515 11516 */
11516 11517 int
11517 11518 ipsec_out_extra_length(ip_xmit_attr_t *ixa)
11518 11519 {
11519 11520 ipsec_action_t *a;
11520 11521
11521 11522 if (!(ixa->ixa_flags & IXAF_IPSEC_SECURE))
11522 11523 return (0);
11523 11524
11524 11525 a = ixa->ixa_ipsec_action;
11525 11526 if (a == NULL) {
11526 11527 ASSERT(ixa->ixa_ipsec_policy != NULL);
11527 11528 a = ixa->ixa_ipsec_policy->ipsp_act;
11528 11529 }
11529 11530 ASSERT(a != NULL);
11530 11531
11531 11532 return (a->ipa_ovhd);
11532 11533 }
11533 11534
11534 11535 /*
11535 11536 * If there are any source route options, return the true final
11536 11537 * destination. Otherwise, return the destination.
11537 11538 */
11538 11539 ipaddr_t
11539 11540 ip_get_dst(ipha_t *ipha)
11540 11541 {
11541 11542 ipoptp_t opts;
11542 11543 uchar_t *opt;
11543 11544 uint8_t optval;
11544 11545 uint8_t optlen;
11545 11546 ipaddr_t dst;
11546 11547 uint32_t off;
11547 11548
11548 11549 dst = ipha->ipha_dst;
11549 11550
11550 11551 if (IS_SIMPLE_IPH(ipha))
11551 11552 return (dst);
11552 11553
11553 11554 for (optval = ipoptp_first(&opts, ipha);
11554 11555 optval != IPOPT_EOL;
11555 11556 optval = ipoptp_next(&opts)) {
11556 11557 opt = opts.ipoptp_cur;
11557 11558 optlen = opts.ipoptp_len;
11558 11559 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11559 11560 switch (optval) {
11560 11561 case IPOPT_SSRR:
11561 11562 case IPOPT_LSRR:
11562 11563 off = opt[IPOPT_OFFSET];
11563 11564 /*
11564 11565 * If one of the conditions is true, it means
11565 11566 * end of options and dst already has the right
11566 11567 * value.
11567 11568 */
11568 11569 if (!(optlen < IP_ADDR_LEN || off > optlen - 3)) {
11569 11570 off = optlen - IP_ADDR_LEN;
11570 11571 bcopy(&opt[off], &dst, IP_ADDR_LEN);
11571 11572 }
11572 11573 return (dst);
11573 11574 default:
11574 11575 break;
11575 11576 }
11576 11577 }
11577 11578
11578 11579 return (dst);
11579 11580 }
11580 11581
11581 11582 /*
11582 11583 * Outbound IP fragmentation routine.
11583 11584 * Assumes the caller has checked whether or not fragmentation should
11584 11585 * be allowed. Here we copy the DF bit from the header to all the generated
11585 11586 * fragments.
11586 11587 */
11587 11588 int
11588 11589 ip_fragment_v4(mblk_t *mp_orig, nce_t *nce, iaflags_t ixaflags,
11589 11590 uint_t pkt_len, uint32_t max_frag, uint32_t xmit_hint, zoneid_t szone,
11590 11591 zoneid_t nolzid, pfirepostfrag_t postfragfn, uintptr_t *ixa_cookie)
11591 11592 {
11592 11593 int i1;
11593 11594 int hdr_len;
11594 11595 mblk_t *hdr_mp;
11595 11596 ipha_t *ipha;
11596 11597 int ip_data_end;
11597 11598 int len;
11598 11599 mblk_t *mp = mp_orig;
11599 11600 int offset;
11600 11601 ill_t *ill = nce->nce_ill;
11601 11602 ip_stack_t *ipst = ill->ill_ipst;
11602 11603 mblk_t *carve_mp;
11603 11604 uint32_t frag_flag;
11604 11605 uint_t priority = mp->b_band;
11605 11606 int error = 0;
11606 11607
11607 11608 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragReqds);
11608 11609
11609 11610 if (pkt_len != msgdsize(mp)) {
11610 11611 ip0dbg(("Packet length mismatch: %d, %ld\n",
11611 11612 pkt_len, msgdsize(mp)));
11612 11613 freemsg(mp);
11613 11614 return (EINVAL);
11614 11615 }
11615 11616
11616 11617 if (max_frag == 0) {
11617 11618 ip1dbg(("ip_fragment_v4: max_frag is zero. Dropping packet\n"));
11618 11619 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11619 11620 ip_drop_output("FragFails: zero max_frag", mp, ill);
11620 11621 freemsg(mp);
11621 11622 return (EINVAL);
11622 11623 }
11623 11624
11624 11625 ASSERT(MBLKL(mp) >= sizeof (ipha_t));
11625 11626 ipha = (ipha_t *)mp->b_rptr;
11626 11627 ASSERT(ntohs(ipha->ipha_length) == pkt_len);
11627 11628 frag_flag = ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_DF;
11628 11629
11629 11630 /*
11630 11631 * Establish the starting offset. May not be zero if we are fragging
11631 11632 * a fragment that is being forwarded.
11632 11633 */
11633 11634 offset = ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_OFFSET;
11634 11635
11635 11636 /* TODO why is this test needed? */
11636 11637 if (((max_frag - ntohs(ipha->ipha_length)) & ~7) < 8) {
11637 11638 /* TODO: notify ulp somehow */
11638 11639 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11639 11640 ip_drop_output("FragFails: bad starting offset", mp, ill);
11640 11641 freemsg(mp);
11641 11642 return (EINVAL);
11642 11643 }
11643 11644
11644 11645 hdr_len = IPH_HDR_LENGTH(ipha);
11645 11646 ipha->ipha_hdr_checksum = 0;
11646 11647
11647 11648 /*
11648 11649 * Establish the number of bytes maximum per frag, after putting
11649 11650 * in the header.
11650 11651 */
11651 11652 len = (max_frag - hdr_len) & ~7;
11652 11653
11653 11654 /* Get a copy of the header for the trailing frags */
11654 11655 hdr_mp = ip_fragment_copyhdr((uchar_t *)ipha, hdr_len, offset, ipst,
11655 11656 mp);
11656 11657 if (hdr_mp == NULL) {
11657 11658 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11658 11659 ip_drop_output("FragFails: no hdr_mp", mp, ill);
11659 11660 freemsg(mp);
11660 11661 return (ENOBUFS);
11661 11662 }
11662 11663
11663 11664 /* Store the starting offset, with the MoreFrags flag. */
11664 11665 i1 = offset | IPH_MF | frag_flag;
11665 11666 ipha->ipha_fragment_offset_and_flags = htons((uint16_t)i1);
11666 11667
11667 11668 /* Establish the ending byte offset, based on the starting offset. */
11668 11669 offset <<= 3;
11669 11670 ip_data_end = offset + ntohs(ipha->ipha_length) - hdr_len;
11670 11671
11671 11672 /* Store the length of the first fragment in the IP header. */
11672 11673 i1 = len + hdr_len;
11673 11674 ASSERT(i1 <= IP_MAXPACKET);
11674 11675 ipha->ipha_length = htons((uint16_t)i1);
11675 11676
11676 11677 /*
11677 11678 * Compute the IP header checksum for the first frag. We have to
11678 11679 * watch out that we stop at the end of the header.
11679 11680 */
11680 11681 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
11681 11682
11682 11683 /*
11683 11684 * Now carve off the first frag. Note that this will include the
11684 11685 * original IP header.
11685 11686 */
11686 11687 if (!(mp = ip_carve_mp(&mp_orig, i1))) {
11687 11688 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11688 11689 ip_drop_output("FragFails: could not carve mp", mp_orig, ill);
11689 11690 freeb(hdr_mp);
11690 11691 freemsg(mp_orig);
11691 11692 return (ENOBUFS);
11692 11693 }
11693 11694
11694 11695 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragCreates);
11695 11696
11696 11697 error = postfragfn(mp, nce, ixaflags, i1, xmit_hint, szone, nolzid,
11697 11698 ixa_cookie);
11698 11699 if (error != 0 && error != EWOULDBLOCK) {
11699 11700 /* No point in sending the other fragments */
11700 11701 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11701 11702 ip_drop_output("FragFails: postfragfn failed", mp_orig, ill);
11702 11703 freeb(hdr_mp);
11703 11704 freemsg(mp_orig);
11704 11705 return (error);
11705 11706 }
11706 11707
11707 11708 /* No need to redo state machine in loop */
11708 11709 ixaflags &= ~IXAF_REACH_CONF;
11709 11710
11710 11711 /* Advance the offset to the second frag starting point. */
11711 11712 offset += len;
11712 11713 /*
11713 11714 * Update hdr_len from the copied header - there might be less options
11714 11715 * in the later fragments.
11715 11716 */
11716 11717 hdr_len = IPH_HDR_LENGTH(hdr_mp->b_rptr);
11717 11718 /* Loop until done. */
11718 11719 for (;;) {
11719 11720 uint16_t offset_and_flags;
11720 11721 uint16_t ip_len;
11721 11722
11722 11723 if (ip_data_end - offset > len) {
11723 11724 /*
11724 11725 * Carve off the appropriate amount from the original
11725 11726 * datagram.
11726 11727 */
11727 11728 if (!(carve_mp = ip_carve_mp(&mp_orig, len))) {
11728 11729 mp = NULL;
11729 11730 break;
11730 11731 }
11731 11732 /*
11732 11733 * More frags after this one. Get another copy
11733 11734 * of the header.
11734 11735 */
11735 11736 if (carve_mp->b_datap->db_ref == 1 &&
11736 11737 hdr_mp->b_wptr - hdr_mp->b_rptr <
11737 11738 carve_mp->b_rptr - carve_mp->b_datap->db_base) {
11738 11739 /* Inline IP header */
11739 11740 carve_mp->b_rptr -= hdr_mp->b_wptr -
11740 11741 hdr_mp->b_rptr;
11741 11742 bcopy(hdr_mp->b_rptr, carve_mp->b_rptr,
11742 11743 hdr_mp->b_wptr - hdr_mp->b_rptr);
11743 11744 mp = carve_mp;
11744 11745 } else {
11745 11746 if (!(mp = copyb(hdr_mp))) {
11746 11747 freemsg(carve_mp);
11747 11748 break;
11748 11749 }
11749 11750 /* Get priority marking, if any. */
11750 11751 mp->b_band = priority;
11751 11752 mp->b_cont = carve_mp;
11752 11753 }
11753 11754 ipha = (ipha_t *)mp->b_rptr;
11754 11755 offset_and_flags = IPH_MF;
11755 11756 } else {
11756 11757 /*
11757 11758 * Last frag. Consume the header. Set len to
11758 11759 * the length of this last piece.
11759 11760 */
11760 11761 len = ip_data_end - offset;
11761 11762
11762 11763 /*
11763 11764 * Carve off the appropriate amount from the original
11764 11765 * datagram.
11765 11766 */
11766 11767 if (!(carve_mp = ip_carve_mp(&mp_orig, len))) {
11767 11768 mp = NULL;
11768 11769 break;
11769 11770 }
11770 11771 if (carve_mp->b_datap->db_ref == 1 &&
11771 11772 hdr_mp->b_wptr - hdr_mp->b_rptr <
11772 11773 carve_mp->b_rptr - carve_mp->b_datap->db_base) {
11773 11774 /* Inline IP header */
11774 11775 carve_mp->b_rptr -= hdr_mp->b_wptr -
11775 11776 hdr_mp->b_rptr;
11776 11777 bcopy(hdr_mp->b_rptr, carve_mp->b_rptr,
11777 11778 hdr_mp->b_wptr - hdr_mp->b_rptr);
11778 11779 mp = carve_mp;
11779 11780 freeb(hdr_mp);
11780 11781 hdr_mp = mp;
11781 11782 } else {
11782 11783 mp = hdr_mp;
11783 11784 /* Get priority marking, if any. */
11784 11785 mp->b_band = priority;
11785 11786 mp->b_cont = carve_mp;
11786 11787 }
11787 11788 ipha = (ipha_t *)mp->b_rptr;
11788 11789 /* A frag of a frag might have IPH_MF non-zero */
11789 11790 offset_and_flags =
11790 11791 ntohs(ipha->ipha_fragment_offset_and_flags) &
11791 11792 IPH_MF;
11792 11793 }
11793 11794 offset_and_flags |= (uint16_t)(offset >> 3);
11794 11795 offset_and_flags |= (uint16_t)frag_flag;
11795 11796 /* Store the offset and flags in the IP header. */
11796 11797 ipha->ipha_fragment_offset_and_flags = htons(offset_and_flags);
11797 11798
11798 11799 /* Store the length in the IP header. */
11799 11800 ip_len = (uint16_t)(len + hdr_len);
11800 11801 ipha->ipha_length = htons(ip_len);
11801 11802
11802 11803 /*
11803 11804 * Set the IP header checksum. Note that mp is just
11804 11805 * the header, so this is easy to pass to ip_csum.
11805 11806 */
11806 11807 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
11807 11808
11808 11809 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragCreates);
11809 11810
11810 11811 error = postfragfn(mp, nce, ixaflags, ip_len, xmit_hint, szone,
11811 11812 nolzid, ixa_cookie);
11812 11813 /* All done if we just consumed the hdr_mp. */
11813 11814 if (mp == hdr_mp) {
11814 11815 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragOKs);
11815 11816 return (error);
11816 11817 }
11817 11818 if (error != 0 && error != EWOULDBLOCK) {
11818 11819 DTRACE_PROBE2(ip__xmit__frag__fail, ill_t *, ill,
11819 11820 mblk_t *, hdr_mp);
11820 11821 /* No point in sending the other fragments */
11821 11822 break;
11822 11823 }
11823 11824
11824 11825 /* Otherwise, advance and loop. */
11825 11826 offset += len;
11826 11827 }
11827 11828 /* Clean up following allocation failure. */
11828 11829 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11829 11830 ip_drop_output("FragFails: loop ended", NULL, ill);
11830 11831 if (mp != hdr_mp)
11831 11832 freeb(hdr_mp);
11832 11833 if (mp != mp_orig)
11833 11834 freemsg(mp_orig);
11834 11835 return (error);
11835 11836 }
11836 11837
11837 11838 /*
11838 11839 * Copy the header plus those options which have the copy bit set
11839 11840 */
11840 11841 static mblk_t *
11841 11842 ip_fragment_copyhdr(uchar_t *rptr, int hdr_len, int offset, ip_stack_t *ipst,
11842 11843 mblk_t *src)
11843 11844 {
11844 11845 mblk_t *mp;
11845 11846 uchar_t *up;
11846 11847
11847 11848 /*
11848 11849 * Quick check if we need to look for options without the copy bit
11849 11850 * set
11850 11851 */
11851 11852 mp = allocb_tmpl(ipst->ips_ip_wroff_extra + hdr_len, src);
11852 11853 if (!mp)
11853 11854 return (mp);
11854 11855 mp->b_rptr += ipst->ips_ip_wroff_extra;
11855 11856 if (hdr_len == IP_SIMPLE_HDR_LENGTH || offset != 0) {
11856 11857 bcopy(rptr, mp->b_rptr, hdr_len);
11857 11858 mp->b_wptr += hdr_len + ipst->ips_ip_wroff_extra;
11858 11859 return (mp);
11859 11860 }
11860 11861 up = mp->b_rptr;
11861 11862 bcopy(rptr, up, IP_SIMPLE_HDR_LENGTH);
11862 11863 up += IP_SIMPLE_HDR_LENGTH;
11863 11864 rptr += IP_SIMPLE_HDR_LENGTH;
11864 11865 hdr_len -= IP_SIMPLE_HDR_LENGTH;
11865 11866 while (hdr_len > 0) {
11866 11867 uint32_t optval;
11867 11868 uint32_t optlen;
11868 11869
11869 11870 optval = *rptr;
11870 11871 if (optval == IPOPT_EOL)
11871 11872 break;
11872 11873 if (optval == IPOPT_NOP)
11873 11874 optlen = 1;
11874 11875 else
11875 11876 optlen = rptr[1];
11876 11877 if (optval & IPOPT_COPY) {
11877 11878 bcopy(rptr, up, optlen);
11878 11879 up += optlen;
11879 11880 }
11880 11881 rptr += optlen;
11881 11882 hdr_len -= optlen;
11882 11883 }
11883 11884 /*
11884 11885 * Make sure that we drop an even number of words by filling
11885 11886 * with EOL to the next word boundary.
11886 11887 */
11887 11888 for (hdr_len = up - (mp->b_rptr + IP_SIMPLE_HDR_LENGTH);
11888 11889 hdr_len & 0x3; hdr_len++)
11889 11890 *up++ = IPOPT_EOL;
11890 11891 mp->b_wptr = up;
11891 11892 /* Update header length */
11892 11893 mp->b_rptr[0] = (uint8_t)((IP_VERSION << 4) | ((up - mp->b_rptr) >> 2));
11893 11894 return (mp);
11894 11895 }
11895 11896
11896 11897 /*
11897 11898 * Update any source route, record route, or timestamp options when
11898 11899 * sending a packet back to ourselves.
11899 11900 * Check that we are at end of strict source route.
11900 11901 * The options have been sanity checked by ip_output_options().
11901 11902 */
11902 11903 void
11903 11904 ip_output_local_options(ipha_t *ipha, ip_stack_t *ipst)
11904 11905 {
11905 11906 ipoptp_t opts;
11906 11907 uchar_t *opt;
11907 11908 uint8_t optval;
11908 11909 uint8_t optlen;
11909 11910 ipaddr_t dst;
11910 11911 uint32_t ts;
11911 11912 timestruc_t now;
11912 11913
11913 11914 for (optval = ipoptp_first(&opts, ipha);
11914 11915 optval != IPOPT_EOL;
11915 11916 optval = ipoptp_next(&opts)) {
11916 11917 opt = opts.ipoptp_cur;
11917 11918 optlen = opts.ipoptp_len;
11918 11919 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11919 11920 switch (optval) {
11920 11921 uint32_t off;
11921 11922 case IPOPT_SSRR:
11922 11923 case IPOPT_LSRR:
11923 11924 off = opt[IPOPT_OFFSET];
11924 11925 off--;
11925 11926 if (optlen < IP_ADDR_LEN ||
11926 11927 off > optlen - IP_ADDR_LEN) {
11927 11928 /* End of source route */
11928 11929 break;
11929 11930 }
11930 11931 /*
11931 11932 * This will only happen if two consecutive entries
11932 11933 * in the source route contains our address or if
11933 11934 * it is a packet with a loose source route which
11934 11935 * reaches us before consuming the whole source route
11935 11936 */
11936 11937
11937 11938 if (optval == IPOPT_SSRR) {
11938 11939 return;
11939 11940 }
11940 11941 /*
11941 11942 * Hack: instead of dropping the packet truncate the
11942 11943 * source route to what has been used by filling the
11943 11944 * rest with IPOPT_NOP.
11944 11945 */
11945 11946 opt[IPOPT_OLEN] = (uint8_t)off;
11946 11947 while (off < optlen) {
11947 11948 opt[off++] = IPOPT_NOP;
11948 11949 }
11949 11950 break;
11950 11951 case IPOPT_RR:
11951 11952 off = opt[IPOPT_OFFSET];
11952 11953 off--;
11953 11954 if (optlen < IP_ADDR_LEN ||
11954 11955 off > optlen - IP_ADDR_LEN) {
11955 11956 /* No more room - ignore */
11956 11957 ip1dbg((
11957 11958 "ip_output_local_options: end of RR\n"));
11958 11959 break;
11959 11960 }
11960 11961 dst = htonl(INADDR_LOOPBACK);
11961 11962 bcopy(&dst, (char *)opt + off, IP_ADDR_LEN);
11962 11963 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
11963 11964 break;
11964 11965 case IPOPT_TS:
11965 11966 /* Insert timestamp if there is romm */
11966 11967 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
11967 11968 case IPOPT_TS_TSONLY:
11968 11969 off = IPOPT_TS_TIMELEN;
11969 11970 break;
11970 11971 case IPOPT_TS_PRESPEC:
11971 11972 case IPOPT_TS_PRESPEC_RFC791:
11972 11973 /* Verify that the address matched */
11973 11974 off = opt[IPOPT_OFFSET] - 1;
11974 11975 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
11975 11976 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
11976 11977 /* Not for us */
11977 11978 break;
11978 11979 }
11979 11980 /* FALLTHRU */
11980 11981 case IPOPT_TS_TSANDADDR:
11981 11982 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
11982 11983 break;
11983 11984 default:
11984 11985 /*
11985 11986 * ip_*put_options should have already
11986 11987 * dropped this packet.
11987 11988 */
11988 11989 cmn_err(CE_PANIC, "ip_output_local_options: "
11989 11990 "unknown IT - bug in ip_output_options?\n");
11990 11991 return; /* Keep "lint" happy */
11991 11992 }
11992 11993 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
11993 11994 /* Increase overflow counter */
11994 11995 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
11995 11996 opt[IPOPT_POS_OV_FLG] = (uint8_t)
11996 11997 (opt[IPOPT_POS_OV_FLG] & 0x0F) |
11997 11998 (off << 4);
11998 11999 break;
11999 12000 }
12000 12001 off = opt[IPOPT_OFFSET] - 1;
12001 12002 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
12002 12003 case IPOPT_TS_PRESPEC:
12003 12004 case IPOPT_TS_PRESPEC_RFC791:
12004 12005 case IPOPT_TS_TSANDADDR:
12005 12006 dst = htonl(INADDR_LOOPBACK);
12006 12007 bcopy(&dst, (char *)opt + off, IP_ADDR_LEN);
12007 12008 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
12008 12009 /* FALLTHRU */
12009 12010 case IPOPT_TS_TSONLY:
12010 12011 off = opt[IPOPT_OFFSET] - 1;
12011 12012 /* Compute # of milliseconds since midnight */
12012 12013 gethrestime(&now);
12013 12014 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
12014 12015 now.tv_nsec / (NANOSEC / MILLISEC);
12015 12016 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
12016 12017 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
12017 12018 break;
12018 12019 }
12019 12020 break;
12020 12021 }
12021 12022 }
12022 12023 }
12023 12024
12024 12025 /*
12025 12026 * Prepend an M_DATA fastpath header, and if none present prepend a
12026 12027 * DL_UNITDATA_REQ. Frees the mblk on failure.
12027 12028 *
12028 12029 * nce_dlur_mp and nce_fp_mp can not disappear once they have been set.
12029 12030 * If there is a change to them, the nce will be deleted (condemned) and
12030 12031 * a new nce_t will be created when packets are sent. Thus we need no locks
12031 12032 * to access those fields.
12032 12033 *
12033 12034 * We preserve b_band to support IPQoS. If a DL_UNITDATA_REQ is prepended
12034 12035 * we place b_band in dl_priority.dl_max.
12035 12036 */
12036 12037 static mblk_t *
12037 12038 ip_xmit_attach_llhdr(mblk_t *mp, nce_t *nce)
12038 12039 {
12039 12040 uint_t hlen;
12040 12041 mblk_t *mp1;
12041 12042 uint_t priority;
12042 12043 uchar_t *rptr;
12043 12044
12044 12045 rptr = mp->b_rptr;
12045 12046
12046 12047 ASSERT(DB_TYPE(mp) == M_DATA);
12047 12048 priority = mp->b_band;
12048 12049
12049 12050 ASSERT(nce != NULL);
12050 12051 if ((mp1 = nce->nce_fp_mp) != NULL) {
12051 12052 hlen = MBLKL(mp1);
12052 12053 /*
12053 12054 * Check if we have enough room to prepend fastpath
12054 12055 * header
12055 12056 */
12056 12057 if (hlen != 0 && (rptr - mp->b_datap->db_base) >= hlen) {
12057 12058 rptr -= hlen;
12058 12059 bcopy(mp1->b_rptr, rptr, hlen);
12059 12060 /*
12060 12061 * Set the b_rptr to the start of the link layer
12061 12062 * header
12062 12063 */
12063 12064 mp->b_rptr = rptr;
12064 12065 return (mp);
12065 12066 }
12066 12067 mp1 = copyb(mp1);
12067 12068 if (mp1 == NULL) {
12068 12069 ill_t *ill = nce->nce_ill;
12069 12070
12070 12071 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12071 12072 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12072 12073 freemsg(mp);
12073 12074 return (NULL);
12074 12075 }
12075 12076 mp1->b_band = priority;
12076 12077 mp1->b_cont = mp;
12077 12078 DB_CKSUMSTART(mp1) = DB_CKSUMSTART(mp);
12078 12079 DB_CKSUMSTUFF(mp1) = DB_CKSUMSTUFF(mp);
12079 12080 DB_CKSUMEND(mp1) = DB_CKSUMEND(mp);
12080 12081 DB_CKSUMFLAGS(mp1) = DB_CKSUMFLAGS(mp);
12081 12082 DB_LSOMSS(mp1) = DB_LSOMSS(mp);
12082 12083 DTRACE_PROBE1(ip__xmit__copyb, (mblk_t *), mp1);
12083 12084 /*
12084 12085 * XXX disable ICK_VALID and compute checksum
12085 12086 * here; can happen if nce_fp_mp changes and
12086 12087 * it can't be copied now due to insufficient
12087 12088 * space. (unlikely, fp mp can change, but it
12088 12089 * does not increase in length)
12089 12090 */
12090 12091 return (mp1);
12091 12092 }
12092 12093 mp1 = copyb(nce->nce_dlur_mp);
12093 12094
12094 12095 if (mp1 == NULL) {
12095 12096 ill_t *ill = nce->nce_ill;
12096 12097
12097 12098 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12098 12099 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12099 12100 freemsg(mp);
12100 12101 return (NULL);
12101 12102 }
12102 12103 mp1->b_cont = mp;
12103 12104 if (priority != 0) {
12104 12105 mp1->b_band = priority;
12105 12106 ((dl_unitdata_req_t *)(mp1->b_rptr))->dl_priority.dl_max =
12106 12107 priority;
12107 12108 }
12108 12109 return (mp1);
12109 12110 #undef rptr
12110 12111 }
12111 12112
12112 12113 /*
12113 12114 * Finish the outbound IPsec processing. This function is called from
12114 12115 * ipsec_out_process() if the IPsec packet was processed
12115 12116 * synchronously, or from {ah,esp}_kcf_callback_outbound() if it was processed
12116 12117 * asynchronously.
12117 12118 *
12118 12119 * This is common to IPv4 and IPv6.
12119 12120 */
12120 12121 int
12121 12122 ip_output_post_ipsec(mblk_t *mp, ip_xmit_attr_t *ixa)
12122 12123 {
12123 12124 iaflags_t ixaflags = ixa->ixa_flags;
12124 12125 uint_t pktlen;
12125 12126
12126 12127
12127 12128 /* AH/ESP don't update ixa_pktlen when they modify the packet */
12128 12129 if (ixaflags & IXAF_IS_IPV4) {
12129 12130 ipha_t *ipha = (ipha_t *)mp->b_rptr;
12130 12131
12131 12132 ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
12132 12133 pktlen = ntohs(ipha->ipha_length);
12133 12134 } else {
12134 12135 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
12135 12136
12136 12137 ASSERT(IPH_HDR_VERSION(mp->b_rptr) == IPV6_VERSION);
12137 12138 pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
12138 12139 }
12139 12140
12140 12141 /*
12141 12142 * We release any hard reference on the SAs here to make
12142 12143 * sure the SAs can be garbage collected. ipsr_sa has a soft reference
12143 12144 * on the SAs.
12144 12145 * If in the future we want the hard latching of the SAs in the
12145 12146 * ip_xmit_attr_t then we should remove this.
12146 12147 */
12147 12148 if (ixa->ixa_ipsec_esp_sa != NULL) {
12148 12149 IPSA_REFRELE(ixa->ixa_ipsec_esp_sa);
12149 12150 ixa->ixa_ipsec_esp_sa = NULL;
12150 12151 }
12151 12152 if (ixa->ixa_ipsec_ah_sa != NULL) {
12152 12153 IPSA_REFRELE(ixa->ixa_ipsec_ah_sa);
12153 12154 ixa->ixa_ipsec_ah_sa = NULL;
12154 12155 }
12155 12156
12156 12157 /* Do we need to fragment? */
12157 12158 if ((ixa->ixa_flags & IXAF_IPV6_ADD_FRAGHDR) ||
12158 12159 pktlen > ixa->ixa_fragsize) {
12159 12160 if (ixaflags & IXAF_IS_IPV4) {
12160 12161 ASSERT(!(ixa->ixa_flags & IXAF_IPV6_ADD_FRAGHDR));
12161 12162 /*
12162 12163 * We check for the DF case in ipsec_out_process
12163 12164 * hence this only handles the non-DF case.
12164 12165 */
12165 12166 return (ip_fragment_v4(mp, ixa->ixa_nce, ixa->ixa_flags,
12166 12167 pktlen, ixa->ixa_fragsize,
12167 12168 ixa->ixa_xmit_hint, ixa->ixa_zoneid,
12168 12169 ixa->ixa_no_loop_zoneid, ixa->ixa_postfragfn,
12169 12170 &ixa->ixa_cookie));
12170 12171 } else {
12171 12172 mp = ip_fraghdr_add_v6(mp, ixa->ixa_ident, ixa);
12172 12173 if (mp == NULL) {
12173 12174 /* MIB and ip_drop_output already done */
12174 12175 return (ENOMEM);
12175 12176 }
12176 12177 pktlen += sizeof (ip6_frag_t);
12177 12178 if (pktlen > ixa->ixa_fragsize) {
12178 12179 return (ip_fragment_v6(mp, ixa->ixa_nce,
12179 12180 ixa->ixa_flags, pktlen,
12180 12181 ixa->ixa_fragsize, ixa->ixa_xmit_hint,
12181 12182 ixa->ixa_zoneid, ixa->ixa_no_loop_zoneid,
12182 12183 ixa->ixa_postfragfn, &ixa->ixa_cookie));
12183 12184 }
12184 12185 }
12185 12186 }
12186 12187 return ((ixa->ixa_postfragfn)(mp, ixa->ixa_nce, ixa->ixa_flags,
12187 12188 pktlen, ixa->ixa_xmit_hint, ixa->ixa_zoneid,
12188 12189 ixa->ixa_no_loop_zoneid, NULL));
12189 12190 }
12190 12191
12191 12192 /*
12192 12193 * Finish the inbound IPsec processing. This function is called from
12193 12194 * ipsec_out_process() if the IPsec packet was processed
12194 12195 * synchronously, or from {ah,esp}_kcf_callback_outbound() if it was processed
12195 12196 * asynchronously.
12196 12197 *
12197 12198 * This is common to IPv4 and IPv6.
12198 12199 */
12199 12200 void
12200 12201 ip_input_post_ipsec(mblk_t *mp, ip_recv_attr_t *ira)
12201 12202 {
12202 12203 iaflags_t iraflags = ira->ira_flags;
12203 12204
12204 12205 /* Length might have changed */
12205 12206 if (iraflags & IRAF_IS_IPV4) {
12206 12207 ipha_t *ipha = (ipha_t *)mp->b_rptr;
12207 12208
12208 12209 ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
12209 12210 ira->ira_pktlen = ntohs(ipha->ipha_length);
12210 12211 ira->ira_ip_hdr_length = IPH_HDR_LENGTH(ipha);
12211 12212 ira->ira_protocol = ipha->ipha_protocol;
12212 12213
12213 12214 ip_fanout_v4(mp, ipha, ira);
12214 12215 } else {
12215 12216 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
12216 12217 uint8_t *nexthdrp;
12217 12218
12218 12219 ASSERT(IPH_HDR_VERSION(mp->b_rptr) == IPV6_VERSION);
12219 12220 ira->ira_pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
12220 12221 if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &ira->ira_ip_hdr_length,
12221 12222 &nexthdrp)) {
12222 12223 /* Malformed packet */
12223 12224 BUMP_MIB(ira->ira_ill->ill_ip_mib, ipIfStatsInDiscards);
12224 12225 ip_drop_input("ipIfStatsInDiscards", mp, ira->ira_ill);
12225 12226 freemsg(mp);
12226 12227 return;
12227 12228 }
12228 12229 ira->ira_protocol = *nexthdrp;
12229 12230 ip_fanout_v6(mp, ip6h, ira);
12230 12231 }
12231 12232 }
12232 12233
12233 12234 /*
12234 12235 * Select which AH & ESP SA's to use (if any) for the outbound packet.
12235 12236 *
12236 12237 * If this function returns B_TRUE, the requested SA's have been filled
12237 12238 * into the ixa_ipsec_*_sa pointers.
12238 12239 *
12239 12240 * If the function returns B_FALSE, the packet has been "consumed", most
12240 12241 * likely by an ACQUIRE sent up via PF_KEY to a key management daemon.
12241 12242 *
12242 12243 * The SA references created by the protocol-specific "select"
12243 12244 * function will be released in ip_output_post_ipsec.
12244 12245 */
12245 12246 static boolean_t
12246 12247 ipsec_out_select_sa(mblk_t *mp, ip_xmit_attr_t *ixa)
12247 12248 {
12248 12249 boolean_t need_ah_acquire = B_FALSE, need_esp_acquire = B_FALSE;
12249 12250 ipsec_policy_t *pp;
12250 12251 ipsec_action_t *ap;
12251 12252
12252 12253 ASSERT(ixa->ixa_flags & IXAF_IPSEC_SECURE);
12253 12254 ASSERT((ixa->ixa_ipsec_policy != NULL) ||
12254 12255 (ixa->ixa_ipsec_action != NULL));
12255 12256
12256 12257 ap = ixa->ixa_ipsec_action;
12257 12258 if (ap == NULL) {
12258 12259 pp = ixa->ixa_ipsec_policy;
12259 12260 ASSERT(pp != NULL);
12260 12261 ap = pp->ipsp_act;
12261 12262 ASSERT(ap != NULL);
12262 12263 }
12263 12264
12264 12265 /*
12265 12266 * We have an action. now, let's select SA's.
12266 12267 * A side effect of setting ixa_ipsec_*_sa is that it will
12267 12268 * be cached in the conn_t.
12268 12269 */
12269 12270 if (ap->ipa_want_esp) {
12270 12271 if (ixa->ixa_ipsec_esp_sa == NULL) {
12271 12272 need_esp_acquire = !ipsec_outbound_sa(mp, ixa,
12272 12273 IPPROTO_ESP);
12273 12274 }
12274 12275 ASSERT(need_esp_acquire || ixa->ixa_ipsec_esp_sa != NULL);
12275 12276 }
12276 12277
12277 12278 if (ap->ipa_want_ah) {
12278 12279 if (ixa->ixa_ipsec_ah_sa == NULL) {
12279 12280 need_ah_acquire = !ipsec_outbound_sa(mp, ixa,
12280 12281 IPPROTO_AH);
12281 12282 }
12282 12283 ASSERT(need_ah_acquire || ixa->ixa_ipsec_ah_sa != NULL);
12283 12284 /*
12284 12285 * The ESP and AH processing order needs to be preserved
12285 12286 * when both protocols are required (ESP should be applied
12286 12287 * before AH for an outbound packet). Force an ESP ACQUIRE
12287 12288 * when both ESP and AH are required, and an AH ACQUIRE
12288 12289 * is needed.
12289 12290 */
12290 12291 if (ap->ipa_want_esp && need_ah_acquire)
12291 12292 need_esp_acquire = B_TRUE;
12292 12293 }
12293 12294
12294 12295 /*
12295 12296 * Send an ACQUIRE (extended, regular, or both) if we need one.
12296 12297 * Release SAs that got referenced, but will not be used until we
12297 12298 * acquire _all_ of the SAs we need.
12298 12299 */
12299 12300 if (need_ah_acquire || need_esp_acquire) {
12300 12301 if (ixa->ixa_ipsec_ah_sa != NULL) {
12301 12302 IPSA_REFRELE(ixa->ixa_ipsec_ah_sa);
12302 12303 ixa->ixa_ipsec_ah_sa = NULL;
12303 12304 }
12304 12305 if (ixa->ixa_ipsec_esp_sa != NULL) {
12305 12306 IPSA_REFRELE(ixa->ixa_ipsec_esp_sa);
12306 12307 ixa->ixa_ipsec_esp_sa = NULL;
12307 12308 }
12308 12309
12309 12310 sadb_acquire(mp, ixa, need_ah_acquire, need_esp_acquire);
12310 12311 return (B_FALSE);
12311 12312 }
12312 12313
12313 12314 return (B_TRUE);
12314 12315 }
12315 12316
12316 12317 /*
12317 12318 * Handle IPsec output processing.
12318 12319 * This function is only entered once for a given packet.
12319 12320 * We try to do things synchronously, but if we need to have user-level
12320 12321 * set up SAs, or ESP or AH uses asynchronous kEF, then the operation
12321 12322 * will be completed
12322 12323 * - when the SAs are added in esp_add_sa_finish/ah_add_sa_finish
12323 12324 * - when asynchronous ESP is done it will do AH
12324 12325 *
12325 12326 * In all cases we come back in ip_output_post_ipsec() to fragment and
12326 12327 * send out the packet.
12327 12328 */
12328 12329 int
12329 12330 ipsec_out_process(mblk_t *mp, ip_xmit_attr_t *ixa)
12330 12331 {
12331 12332 ill_t *ill = ixa->ixa_nce->nce_ill;
12332 12333 ip_stack_t *ipst = ixa->ixa_ipst;
12333 12334 ipsec_stack_t *ipss;
12334 12335 ipsec_policy_t *pp;
12335 12336 ipsec_action_t *ap;
12336 12337
12337 12338 ASSERT(ixa->ixa_flags & IXAF_IPSEC_SECURE);
12338 12339
12339 12340 ASSERT((ixa->ixa_ipsec_policy != NULL) ||
12340 12341 (ixa->ixa_ipsec_action != NULL));
12341 12342
12342 12343 ipss = ipst->ips_netstack->netstack_ipsec;
12343 12344 if (!ipsec_loaded(ipss)) {
12344 12345 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12345 12346 ip_drop_packet(mp, B_TRUE, ill,
12346 12347 DROPPER(ipss, ipds_ip_ipsec_not_loaded),
12347 12348 &ipss->ipsec_dropper);
12348 12349 return (ENOTSUP);
12349 12350 }
12350 12351
12351 12352 ap = ixa->ixa_ipsec_action;
12352 12353 if (ap == NULL) {
12353 12354 pp = ixa->ixa_ipsec_policy;
12354 12355 ASSERT(pp != NULL);
12355 12356 ap = pp->ipsp_act;
12356 12357 ASSERT(ap != NULL);
12357 12358 }
12358 12359
12359 12360 /* Handle explicit drop action and bypass. */
12360 12361 switch (ap->ipa_act.ipa_type) {
12361 12362 case IPSEC_ACT_DISCARD:
12362 12363 case IPSEC_ACT_REJECT:
12363 12364 ip_drop_packet(mp, B_FALSE, ill,
12364 12365 DROPPER(ipss, ipds_spd_explicit), &ipss->ipsec_spd_dropper);
12365 12366 return (EHOSTUNREACH); /* IPsec policy failure */
12366 12367 case IPSEC_ACT_BYPASS:
12367 12368 return (ip_output_post_ipsec(mp, ixa));
12368 12369 }
12369 12370
12370 12371 /*
12371 12372 * The order of processing is first insert a IP header if needed.
12372 12373 * Then insert the ESP header and then the AH header.
12373 12374 */
12374 12375 if ((ixa->ixa_flags & IXAF_IS_IPV4) && ap->ipa_want_se) {
12375 12376 /*
12376 12377 * First get the outer IP header before sending
12377 12378 * it to ESP.
12378 12379 */
12379 12380 ipha_t *oipha, *iipha;
12380 12381 mblk_t *outer_mp, *inner_mp;
12381 12382
12382 12383 if ((outer_mp = allocb(sizeof (ipha_t), BPRI_HI)) == NULL) {
12383 12384 (void) mi_strlog(ill->ill_rq, 0,
12384 12385 SL_ERROR|SL_TRACE|SL_CONSOLE,
12385 12386 "ipsec_out_process: "
12386 12387 "Self-Encapsulation failed: Out of memory\n");
12387 12388 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12388 12389 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12389 12390 freemsg(mp);
12390 12391 return (ENOBUFS);
12391 12392 }
12392 12393 inner_mp = mp;
12393 12394 ASSERT(inner_mp->b_datap->db_type == M_DATA);
12394 12395 oipha = (ipha_t *)outer_mp->b_rptr;
12395 12396 iipha = (ipha_t *)inner_mp->b_rptr;
12396 12397 *oipha = *iipha;
12397 12398 outer_mp->b_wptr += sizeof (ipha_t);
12398 12399 oipha->ipha_length = htons(ntohs(iipha->ipha_length) +
12399 12400 sizeof (ipha_t));
12400 12401 oipha->ipha_protocol = IPPROTO_ENCAP;
12401 12402 oipha->ipha_version_and_hdr_length =
12402 12403 IP_SIMPLE_HDR_VERSION;
12403 12404 oipha->ipha_hdr_checksum = 0;
12404 12405 oipha->ipha_hdr_checksum = ip_csum_hdr(oipha);
12405 12406 outer_mp->b_cont = inner_mp;
12406 12407 mp = outer_mp;
12407 12408
12408 12409 ixa->ixa_flags |= IXAF_IPSEC_TUNNEL;
12409 12410 }
12410 12411
12411 12412 /* If we need to wait for a SA then we can't return any errno */
12412 12413 if (((ap->ipa_want_ah && (ixa->ixa_ipsec_ah_sa == NULL)) ||
12413 12414 (ap->ipa_want_esp && (ixa->ixa_ipsec_esp_sa == NULL))) &&
12414 12415 !ipsec_out_select_sa(mp, ixa))
12415 12416 return (0);
12416 12417
12417 12418 /*
12418 12419 * By now, we know what SA's to use. Toss over to ESP & AH
12419 12420 * to do the heavy lifting.
12420 12421 */
12421 12422 if (ap->ipa_want_esp) {
12422 12423 ASSERT(ixa->ixa_ipsec_esp_sa != NULL);
12423 12424
12424 12425 mp = ixa->ixa_ipsec_esp_sa->ipsa_output_func(mp, ixa);
12425 12426 if (mp == NULL) {
12426 12427 /*
12427 12428 * Either it failed or is pending. In the former case
12428 12429 * ipIfStatsInDiscards was increased.
12429 12430 */
12430 12431 return (0);
12431 12432 }
12432 12433 }
12433 12434
12434 12435 if (ap->ipa_want_ah) {
12435 12436 ASSERT(ixa->ixa_ipsec_ah_sa != NULL);
12436 12437
12437 12438 mp = ixa->ixa_ipsec_ah_sa->ipsa_output_func(mp, ixa);
12438 12439 if (mp == NULL) {
12439 12440 /*
12440 12441 * Either it failed or is pending. In the former case
12441 12442 * ipIfStatsInDiscards was increased.
12442 12443 */
12443 12444 return (0);
12444 12445 }
12445 12446 }
12446 12447 /*
12447 12448 * We are done with IPsec processing. Send it over
12448 12449 * the wire.
12449 12450 */
12450 12451 return (ip_output_post_ipsec(mp, ixa));
12451 12452 }
12452 12453
12453 12454 /*
12454 12455 * ioctls that go through a down/up sequence may need to wait for the down
12455 12456 * to complete. This involves waiting for the ire and ipif refcnts to go down
12456 12457 * to zero. Subsequently the ioctl is restarted from ipif_ill_refrele_tail.
12457 12458 */
12458 12459 /* ARGSUSED */
12459 12460 void
12460 12461 ip_reprocess_ioctl(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
12461 12462 {
12462 12463 struct iocblk *iocp;
12463 12464 mblk_t *mp1;
12464 12465 ip_ioctl_cmd_t *ipip;
12465 12466 int err;
12466 12467 sin_t *sin;
12467 12468 struct lifreq *lifr;
12468 12469 struct ifreq *ifr;
12469 12470
12470 12471 iocp = (struct iocblk *)mp->b_rptr;
12471 12472 ASSERT(ipsq != NULL);
12472 12473 /* Existence of mp1 verified in ip_wput_nondata */
12473 12474 mp1 = mp->b_cont->b_cont;
12474 12475 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12475 12476 if (ipip->ipi_cmd == SIOCSLIFNAME || ipip->ipi_cmd == IF_UNITSEL) {
12476 12477 /*
12477 12478 * Special case where ipx_current_ipif is not set:
12478 12479 * ill_phyint_reinit merged the v4 and v6 into a single ipsq.
12479 12480 * We are here as were not able to complete the operation in
12480 12481 * ipif_set_values because we could not become exclusive on
12481 12482 * the new ipsq.
12482 12483 */
12483 12484 ill_t *ill = q->q_ptr;
12484 12485 ipsq_current_start(ipsq, ill->ill_ipif, ipip->ipi_cmd);
12485 12486 }
12486 12487 ASSERT(ipsq->ipsq_xop->ipx_current_ipif != NULL);
12487 12488
12488 12489 if (ipip->ipi_cmd_type == IF_CMD) {
12489 12490 /* This a old style SIOC[GS]IF* command */
12490 12491 ifr = (struct ifreq *)mp1->b_rptr;
12491 12492 sin = (sin_t *)&ifr->ifr_addr;
12492 12493 } else if (ipip->ipi_cmd_type == LIF_CMD) {
12493 12494 /* This a new style SIOC[GS]LIF* command */
12494 12495 lifr = (struct lifreq *)mp1->b_rptr;
12495 12496 sin = (sin_t *)&lifr->lifr_addr;
12496 12497 } else {
12497 12498 sin = NULL;
12498 12499 }
12499 12500
12500 12501 err = (*ipip->ipi_func_restart)(ipsq->ipsq_xop->ipx_current_ipif, sin,
12501 12502 q, mp, ipip, mp1->b_rptr);
12502 12503
12503 12504 DTRACE_PROBE4(ipif__ioctl, char *, "ip_reprocess_ioctl finish",
12504 12505 int, ipip->ipi_cmd,
12505 12506 ill_t *, ipsq->ipsq_xop->ipx_current_ipif->ipif_ill,
12506 12507 ipif_t *, ipsq->ipsq_xop->ipx_current_ipif);
12507 12508
12508 12509 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), ipsq);
12509 12510 }
12510 12511
12511 12512 /*
12512 12513 * ioctl processing
12513 12514 *
12514 12515 * ioctl processing starts with ip_sioctl_copyin_setup(), which looks up
12515 12516 * the ioctl command in the ioctl tables, determines the copyin data size
12516 12517 * from the ipi_copyin_size field, and does an mi_copyin() of that size.
12517 12518 *
12518 12519 * ioctl processing then continues when the M_IOCDATA makes its way down to
12519 12520 * ip_wput_nondata(). The ioctl is looked up again in the ioctl table, its
12520 12521 * associated 'conn' is refheld till the end of the ioctl and the general
12521 12522 * ioctl processing function ip_process_ioctl() is called to extract the
12522 12523 * arguments and process the ioctl. To simplify extraction, ioctl commands
12523 12524 * are "typed" based on the arguments they take (e.g., LIF_CMD which takes a
12524 12525 * `struct lifreq'), and a common extract function (e.g., ip_extract_lifreq())
12525 12526 * is used to extract the ioctl's arguments.
12526 12527 *
12527 12528 * ip_process_ioctl determines if the ioctl needs to be serialized, and if
12528 12529 * so goes thru the serialization primitive ipsq_try_enter. Then the
12529 12530 * appropriate function to handle the ioctl is called based on the entry in
12530 12531 * the ioctl table. ioctl completion is encapsulated in ip_ioctl_finish
12531 12532 * which also refreleases the 'conn' that was refheld at the start of the
12532 12533 * ioctl. Finally ipsq_exit is called if needed to exit the ipsq.
12533 12534 *
12534 12535 * Many exclusive ioctls go thru an internal down up sequence as part of
12535 12536 * the operation. For example an attempt to change the IP address of an
12536 12537 * ipif entails ipif_down, set address, ipif_up. Bringing down the interface
12537 12538 * does all the cleanup such as deleting all ires that use this address.
12538 12539 * Then we need to wait till all references to the interface go away.
12539 12540 */
12540 12541 void
12541 12542 ip_process_ioctl(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *arg)
12542 12543 {
12543 12544 struct iocblk *iocp = (struct iocblk *)mp->b_rptr;
12544 12545 ip_ioctl_cmd_t *ipip = arg;
12545 12546 ip_extract_func_t *extract_funcp;
12546 12547 cmd_info_t ci;
12547 12548 int err;
12548 12549 boolean_t entered_ipsq = B_FALSE;
12549 12550
12550 12551 ip3dbg(("ip_process_ioctl: ioctl %X\n", iocp->ioc_cmd));
12551 12552
12552 12553 if (ipip == NULL)
12553 12554 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12554 12555
12555 12556 /*
12556 12557 * SIOCLIFADDIF needs to go thru a special path since the
12557 12558 * ill may not exist yet. This happens in the case of lo0
12558 12559 * which is created using this ioctl.
12559 12560 */
12560 12561 if (ipip->ipi_cmd == SIOCLIFADDIF) {
12561 12562 err = ip_sioctl_addif(NULL, NULL, q, mp, NULL, NULL);
12562 12563 DTRACE_PROBE4(ipif__ioctl, char *, "ip_process_ioctl finish",
12563 12564 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12564 12565 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12565 12566 return;
12566 12567 }
12567 12568
12568 12569 ci.ci_ipif = NULL;
12569 12570 switch (ipip->ipi_cmd_type) {
12570 12571 case MISC_CMD:
12571 12572 case MSFILT_CMD:
12572 12573 /*
12573 12574 * All MISC_CMD ioctls come in here -- e.g. SIOCGLIFCONF.
12574 12575 */
12575 12576 if (ipip->ipi_cmd == IF_UNITSEL) {
12576 12577 /* ioctl comes down the ill */
12577 12578 ci.ci_ipif = ((ill_t *)q->q_ptr)->ill_ipif;
12578 12579 ipif_refhold(ci.ci_ipif);
12579 12580 }
12580 12581 err = 0;
12581 12582 ci.ci_sin = NULL;
12582 12583 ci.ci_sin6 = NULL;
12583 12584 ci.ci_lifr = NULL;
12584 12585 extract_funcp = NULL;
12585 12586 break;
12586 12587
12587 12588 case IF_CMD:
12588 12589 case LIF_CMD:
12589 12590 extract_funcp = ip_extract_lifreq;
12590 12591 break;
12591 12592
12592 12593 case ARP_CMD:
12593 12594 case XARP_CMD:
12594 12595 extract_funcp = ip_extract_arpreq;
12595 12596 break;
12596 12597
12597 12598 default:
12598 12599 ASSERT(0);
12599 12600 }
12600 12601
12601 12602 if (extract_funcp != NULL) {
12602 12603 err = (*extract_funcp)(q, mp, ipip, &ci);
12603 12604 if (err != 0) {
12604 12605 DTRACE_PROBE4(ipif__ioctl,
12605 12606 char *, "ip_process_ioctl finish err",
12606 12607 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12607 12608 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12608 12609 return;
12609 12610 }
12610 12611
12611 12612 /*
12612 12613 * All of the extraction functions return a refheld ipif.
12613 12614 */
12614 12615 ASSERT(ci.ci_ipif != NULL);
12615 12616 }
12616 12617
12617 12618 if (!(ipip->ipi_flags & IPI_WR)) {
12618 12619 /*
12619 12620 * A return value of EINPROGRESS means the ioctl is
12620 12621 * either queued and waiting for some reason or has
12621 12622 * already completed.
12622 12623 */
12623 12624 err = (*ipip->ipi_func)(ci.ci_ipif, ci.ci_sin, q, mp, ipip,
12624 12625 ci.ci_lifr);
12625 12626 if (ci.ci_ipif != NULL) {
12626 12627 DTRACE_PROBE4(ipif__ioctl,
12627 12628 char *, "ip_process_ioctl finish RD",
12628 12629 int, ipip->ipi_cmd, ill_t *, ci.ci_ipif->ipif_ill,
12629 12630 ipif_t *, ci.ci_ipif);
12630 12631 ipif_refrele(ci.ci_ipif);
12631 12632 } else {
12632 12633 DTRACE_PROBE4(ipif__ioctl,
12633 12634 char *, "ip_process_ioctl finish RD",
12634 12635 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12635 12636 }
12636 12637 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12637 12638 return;
12638 12639 }
12639 12640
12640 12641 ASSERT(ci.ci_ipif != NULL);
12641 12642
12642 12643 /*
12643 12644 * If ipsq is non-NULL, we are already being called exclusively
12644 12645 */
12645 12646 ASSERT(ipsq == NULL || IAM_WRITER_IPSQ(ipsq));
12646 12647 if (ipsq == NULL) {
12647 12648 ipsq = ipsq_try_enter(ci.ci_ipif, NULL, q, mp, ip_process_ioctl,
12648 12649 NEW_OP, B_TRUE);
12649 12650 if (ipsq == NULL) {
12650 12651 ipif_refrele(ci.ci_ipif);
12651 12652 return;
12652 12653 }
12653 12654 entered_ipsq = B_TRUE;
12654 12655 }
12655 12656 /*
12656 12657 * Release the ipif so that ipif_down and friends that wait for
12657 12658 * references to go away are not misled about the current ipif_refcnt
12658 12659 * values. We are writer so we can access the ipif even after releasing
12659 12660 * the ipif.
12660 12661 */
12661 12662 ipif_refrele(ci.ci_ipif);
12662 12663
12663 12664 ipsq_current_start(ipsq, ci.ci_ipif, ipip->ipi_cmd);
12664 12665
12665 12666 /*
12666 12667 * A return value of EINPROGRESS means the ioctl is
12667 12668 * either queued and waiting for some reason or has
12668 12669 * already completed.
12669 12670 */
12670 12671 err = (*ipip->ipi_func)(ci.ci_ipif, ci.ci_sin, q, mp, ipip, ci.ci_lifr);
12671 12672
12672 12673 DTRACE_PROBE4(ipif__ioctl, char *, "ip_process_ioctl finish WR",
12673 12674 int, ipip->ipi_cmd,
12674 12675 ill_t *, ci.ci_ipif == NULL ? NULL : ci.ci_ipif->ipif_ill,
12675 12676 ipif_t *, ci.ci_ipif);
12676 12677 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), ipsq);
12677 12678
12678 12679 if (entered_ipsq)
12679 12680 ipsq_exit(ipsq);
12680 12681 }
12681 12682
12682 12683 /*
12683 12684 * Complete the ioctl. Typically ioctls use the mi package and need to
12684 12685 * do mi_copyout/mi_copy_done.
12685 12686 */
12686 12687 void
12687 12688 ip_ioctl_finish(queue_t *q, mblk_t *mp, int err, int mode, ipsq_t *ipsq)
12688 12689 {
12689 12690 conn_t *connp = NULL;
12690 12691
12691 12692 if (err == EINPROGRESS)
12692 12693 return;
12693 12694
12694 12695 if (CONN_Q(q)) {
12695 12696 connp = Q_TO_CONN(q);
12696 12697 ASSERT(connp->conn_ref >= 2);
12697 12698 }
12698 12699
12699 12700 switch (mode) {
12700 12701 case COPYOUT:
12701 12702 if (err == 0)
12702 12703 mi_copyout(q, mp);
12703 12704 else
12704 12705 mi_copy_done(q, mp, err);
12705 12706 break;
12706 12707
12707 12708 case NO_COPYOUT:
12708 12709 mi_copy_done(q, mp, err);
12709 12710 break;
12710 12711
12711 12712 default:
12712 12713 ASSERT(mode == CONN_CLOSE); /* aborted through CONN_CLOSE */
12713 12714 break;
12714 12715 }
12715 12716
12716 12717 /*
12717 12718 * The conn refhold and ioctlref placed on the conn at the start of the
12718 12719 * ioctl are released here.
12719 12720 */
12720 12721 if (connp != NULL) {
12721 12722 CONN_DEC_IOCTLREF(connp);
12722 12723 CONN_OPER_PENDING_DONE(connp);
12723 12724 }
12724 12725
12725 12726 if (ipsq != NULL)
12726 12727 ipsq_current_finish(ipsq);
12727 12728 }
12728 12729
12729 12730 /* Handles all non data messages */
12730 12731 void
12731 12732 ip_wput_nondata(queue_t *q, mblk_t *mp)
12732 12733 {
12733 12734 mblk_t *mp1;
12734 12735 struct iocblk *iocp;
12735 12736 ip_ioctl_cmd_t *ipip;
12736 12737 conn_t *connp;
12737 12738 cred_t *cr;
12738 12739 char *proto_str;
12739 12740
12740 12741 if (CONN_Q(q))
12741 12742 connp = Q_TO_CONN(q);
12742 12743 else
12743 12744 connp = NULL;
12744 12745
12745 12746 switch (DB_TYPE(mp)) {
12746 12747 case M_IOCTL:
12747 12748 /*
12748 12749 * IOCTL processing begins in ip_sioctl_copyin_setup which
12749 12750 * will arrange to copy in associated control structures.
12750 12751 */
12751 12752 ip_sioctl_copyin_setup(q, mp);
12752 12753 return;
12753 12754 case M_IOCDATA:
12754 12755 /*
12755 12756 * Ensure that this is associated with one of our trans-
12756 12757 * parent ioctls. If it's not ours, discard it if we're
12757 12758 * running as a driver, or pass it on if we're a module.
12758 12759 */
12759 12760 iocp = (struct iocblk *)mp->b_rptr;
12760 12761 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12761 12762 if (ipip == NULL) {
12762 12763 if (q->q_next == NULL) {
12763 12764 goto nak;
12764 12765 } else {
12765 12766 putnext(q, mp);
12766 12767 }
12767 12768 return;
12768 12769 }
12769 12770 if ((q->q_next != NULL) && !(ipip->ipi_flags & IPI_MODOK)) {
12770 12771 /*
12771 12772 * The ioctl is one we recognise, but is not consumed
12772 12773 * by IP as a module and we are a module, so we drop
12773 12774 */
12774 12775 goto nak;
12775 12776 }
12776 12777
12777 12778 /* IOCTL continuation following copyin or copyout. */
12778 12779 if (mi_copy_state(q, mp, NULL) == -1) {
12779 12780 /*
12780 12781 * The copy operation failed. mi_copy_state already
12781 12782 * cleaned up, so we're out of here.
12782 12783 */
12783 12784 return;
12784 12785 }
12785 12786 /*
12786 12787 * If we just completed a copy in, we become writer and
12787 12788 * continue processing in ip_sioctl_copyin_done. If it
12788 12789 * was a copy out, we call mi_copyout again. If there is
12789 12790 * nothing more to copy out, it will complete the IOCTL.
12790 12791 */
12791 12792 if (MI_COPY_DIRECTION(mp) == MI_COPY_IN) {
12792 12793 if (!(mp1 = mp->b_cont) || !(mp1 = mp1->b_cont)) {
12793 12794 mi_copy_done(q, mp, EPROTO);
12794 12795 return;
12795 12796 }
12796 12797 /*
12797 12798 * Check for cases that need more copying. A return
12798 12799 * value of 0 means a second copyin has been started,
12799 12800 * so we return; a return value of 1 means no more
12800 12801 * copying is needed, so we continue.
12801 12802 */
12802 12803 if (ipip->ipi_cmd_type == MSFILT_CMD &&
12803 12804 MI_COPY_COUNT(mp) == 1) {
12804 12805 if (ip_copyin_msfilter(q, mp) == 0)
12805 12806 return;
12806 12807 }
12807 12808 /*
12808 12809 * Refhold the conn, till the ioctl completes. This is
12809 12810 * needed in case the ioctl ends up in the pending mp
12810 12811 * list. Every mp in the ipx_pending_mp list must have
12811 12812 * a refhold on the conn to resume processing. The
12812 12813 * refhold is released when the ioctl completes
12813 12814 * (whether normally or abnormally). An ioctlref is also
12814 12815 * placed on the conn to prevent TCP from removing the
12815 12816 * queue needed to send the ioctl reply back.
12816 12817 * In all cases ip_ioctl_finish is called to finish
12817 12818 * the ioctl and release the refholds.
12818 12819 */
12819 12820 if (connp != NULL) {
12820 12821 /* This is not a reentry */
12821 12822 CONN_INC_REF(connp);
12822 12823 CONN_INC_IOCTLREF(connp);
12823 12824 } else {
12824 12825 if (!(ipip->ipi_flags & IPI_MODOK)) {
12825 12826 mi_copy_done(q, mp, EINVAL);
12826 12827 return;
12827 12828 }
12828 12829 }
12829 12830
12830 12831 ip_process_ioctl(NULL, q, mp, ipip);
12831 12832
12832 12833 } else {
12833 12834 mi_copyout(q, mp);
12834 12835 }
12835 12836 return;
12836 12837
12837 12838 case M_IOCNAK:
12838 12839 /*
12839 12840 * The only way we could get here is if a resolver didn't like
12840 12841 * an IOCTL we sent it. This shouldn't happen.
12841 12842 */
12842 12843 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
12843 12844 "ip_wput_nondata: unexpected M_IOCNAK, ioc_cmd 0x%x",
12844 12845 ((struct iocblk *)mp->b_rptr)->ioc_cmd);
12845 12846 freemsg(mp);
12846 12847 return;
12847 12848 case M_IOCACK:
12848 12849 /* /dev/ip shouldn't see this */
12849 12850 goto nak;
12850 12851 case M_FLUSH:
12851 12852 if (*mp->b_rptr & FLUSHW)
12852 12853 flushq(q, FLUSHALL);
12853 12854 if (q->q_next) {
12854 12855 putnext(q, mp);
12855 12856 return;
12856 12857 }
12857 12858 if (*mp->b_rptr & FLUSHR) {
12858 12859 *mp->b_rptr &= ~FLUSHW;
12859 12860 qreply(q, mp);
12860 12861 return;
12861 12862 }
12862 12863 freemsg(mp);
12863 12864 return;
12864 12865 case M_CTL:
12865 12866 break;
12866 12867 case M_PROTO:
12867 12868 case M_PCPROTO:
12868 12869 /*
12869 12870 * The only PROTO messages we expect are SNMP-related.
12870 12871 */
12871 12872 switch (((union T_primitives *)mp->b_rptr)->type) {
12872 12873 case T_SVR4_OPTMGMT_REQ:
12873 12874 ip2dbg(("ip_wput_nondata: T_SVR4_OPTMGMT_REQ "
12874 12875 "flags %x\n",
12875 12876 ((struct T_optmgmt_req *)mp->b_rptr)->MGMT_flags));
12876 12877
12877 12878 if (connp == NULL) {
12878 12879 proto_str = "T_SVR4_OPTMGMT_REQ";
12879 12880 goto protonak;
12880 12881 }
12881 12882
12882 12883 /*
12883 12884 * All Solaris components should pass a db_credp
12884 12885 * for this TPI message, hence we ASSERT.
12885 12886 * But in case there is some other M_PROTO that looks
12886 12887 * like a TPI message sent by some other kernel
12887 12888 * component, we check and return an error.
12888 12889 */
12889 12890 cr = msg_getcred(mp, NULL);
12890 12891 ASSERT(cr != NULL);
12891 12892 if (cr == NULL) {
12892 12893 mp = mi_tpi_err_ack_alloc(mp, TSYSERR, EINVAL);
12893 12894 if (mp != NULL)
12894 12895 qreply(q, mp);
12895 12896 return;
12896 12897 }
12897 12898
12898 12899 if (!snmpcom_req(q, mp, ip_snmp_set, ip_snmp_get, cr)) {
12899 12900 proto_str = "Bad SNMPCOM request?";
12900 12901 goto protonak;
12901 12902 }
12902 12903 return;
12903 12904 default:
12904 12905 ip1dbg(("ip_wput_nondata: dropping M_PROTO prim %u\n",
12905 12906 (int)*(uint_t *)mp->b_rptr));
12906 12907 freemsg(mp);
12907 12908 return;
12908 12909 }
12909 12910 default:
12910 12911 break;
12911 12912 }
12912 12913 if (q->q_next) {
12913 12914 putnext(q, mp);
12914 12915 } else
12915 12916 freemsg(mp);
12916 12917 return;
12917 12918
12918 12919 nak:
12919 12920 iocp->ioc_error = EINVAL;
12920 12921 mp->b_datap->db_type = M_IOCNAK;
12921 12922 iocp->ioc_count = 0;
12922 12923 qreply(q, mp);
12923 12924 return;
12924 12925
12925 12926 protonak:
12926 12927 cmn_err(CE_NOTE, "IP doesn't process %s as a module", proto_str);
12927 12928 if ((mp = mi_tpi_err_ack_alloc(mp, TPROTO, EINVAL)) != NULL)
12928 12929 qreply(q, mp);
12929 12930 }
12930 12931
12931 12932 /*
12932 12933 * Process IP options in an outbound packet. Verify that the nexthop in a
12933 12934 * strict source route is onlink.
12934 12935 * Returns non-zero if something fails in which case an ICMP error has been
12935 12936 * sent and mp freed.
12936 12937 *
12937 12938 * Assumes the ULP has called ip_massage_options to move nexthop into ipha_dst.
12938 12939 */
12939 12940 int
12940 12941 ip_output_options(mblk_t *mp, ipha_t *ipha, ip_xmit_attr_t *ixa, ill_t *ill)
12941 12942 {
12942 12943 ipoptp_t opts;
12943 12944 uchar_t *opt;
12944 12945 uint8_t optval;
12945 12946 uint8_t optlen;
12946 12947 ipaddr_t dst;
12947 12948 intptr_t code = 0;
12948 12949 ire_t *ire;
12949 12950 ip_stack_t *ipst = ixa->ixa_ipst;
12950 12951 ip_recv_attr_t iras;
12951 12952
12952 12953 ip2dbg(("ip_output_options\n"));
12953 12954
12954 12955 dst = ipha->ipha_dst;
12955 12956 for (optval = ipoptp_first(&opts, ipha);
12956 12957 optval != IPOPT_EOL;
12957 12958 optval = ipoptp_next(&opts)) {
12958 12959 opt = opts.ipoptp_cur;
12959 12960 optlen = opts.ipoptp_len;
12960 12961 ip2dbg(("ip_output_options: opt %d, len %d\n",
12961 12962 optval, optlen));
12962 12963 switch (optval) {
12963 12964 uint32_t off;
12964 12965 case IPOPT_SSRR:
12965 12966 case IPOPT_LSRR:
12966 12967 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
12967 12968 ip1dbg((
12968 12969 "ip_output_options: bad option offset\n"));
12969 12970 code = (char *)&opt[IPOPT_OLEN] -
12970 12971 (char *)ipha;
12971 12972 goto param_prob;
12972 12973 }
12973 12974 off = opt[IPOPT_OFFSET];
12974 12975 ip1dbg(("ip_output_options: next hop 0x%x\n",
12975 12976 ntohl(dst)));
12976 12977 /*
12977 12978 * For strict: verify that dst is directly
12978 12979 * reachable.
12979 12980 */
12980 12981 if (optval == IPOPT_SSRR) {
12981 12982 ire = ire_ftable_lookup_v4(dst, 0, 0,
12982 12983 IRE_INTERFACE, NULL, ALL_ZONES,
12983 12984 ixa->ixa_tsl,
12984 12985 MATCH_IRE_TYPE | MATCH_IRE_SECATTR, 0, ipst,
12985 12986 NULL);
12986 12987 if (ire == NULL) {
12987 12988 ip1dbg(("ip_output_options: SSRR not"
12988 12989 " directly reachable: 0x%x\n",
12989 12990 ntohl(dst)));
12990 12991 goto bad_src_route;
12991 12992 }
12992 12993 ire_refrele(ire);
12993 12994 }
12994 12995 break;
12995 12996 case IPOPT_RR:
12996 12997 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
12997 12998 ip1dbg((
12998 12999 "ip_output_options: bad option offset\n"));
12999 13000 code = (char *)&opt[IPOPT_OLEN] -
13000 13001 (char *)ipha;
13001 13002 goto param_prob;
13002 13003 }
13003 13004 break;
13004 13005 case IPOPT_TS:
13005 13006 /*
13006 13007 * Verify that length >=5 and that there is either
13007 13008 * room for another timestamp or that the overflow
13008 13009 * counter is not maxed out.
13009 13010 */
13010 13011 code = (char *)&opt[IPOPT_OLEN] - (char *)ipha;
13011 13012 if (optlen < IPOPT_MINLEN_IT) {
13012 13013 goto param_prob;
13013 13014 }
13014 13015 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
13015 13016 ip1dbg((
13016 13017 "ip_output_options: bad option offset\n"));
13017 13018 code = (char *)&opt[IPOPT_OFFSET] -
13018 13019 (char *)ipha;
13019 13020 goto param_prob;
13020 13021 }
13021 13022 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
13022 13023 case IPOPT_TS_TSONLY:
13023 13024 off = IPOPT_TS_TIMELEN;
13024 13025 break;
13025 13026 case IPOPT_TS_TSANDADDR:
13026 13027 case IPOPT_TS_PRESPEC:
13027 13028 case IPOPT_TS_PRESPEC_RFC791:
13028 13029 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
13029 13030 break;
13030 13031 default:
13031 13032 code = (char *)&opt[IPOPT_POS_OV_FLG] -
13032 13033 (char *)ipha;
13033 13034 goto param_prob;
13034 13035 }
13035 13036 if (opt[IPOPT_OFFSET] - 1 + off > optlen &&
13036 13037 (opt[IPOPT_POS_OV_FLG] & 0xF0) == 0xF0) {
13037 13038 /*
13038 13039 * No room and the overflow counter is 15
13039 13040 * already.
13040 13041 */
13041 13042 goto param_prob;
13042 13043 }
13043 13044 break;
13044 13045 }
13045 13046 }
13046 13047
13047 13048 if ((opts.ipoptp_flags & IPOPTP_ERROR) == 0)
13048 13049 return (0);
13049 13050
13050 13051 ip1dbg(("ip_output_options: error processing IP options."));
13051 13052 code = (char *)&opt[IPOPT_OFFSET] - (char *)ipha;
13052 13053
13053 13054 param_prob:
13054 13055 bzero(&iras, sizeof (iras));
13055 13056 iras.ira_ill = iras.ira_rill = ill;
13056 13057 iras.ira_ruifindex = ill->ill_phyint->phyint_ifindex;
13057 13058 iras.ira_rifindex = iras.ira_ruifindex;
13058 13059 iras.ira_flags = IRAF_IS_IPV4;
13059 13060
13060 13061 ip_drop_output("ip_output_options", mp, ill);
13061 13062 icmp_param_problem(mp, (uint8_t)code, &iras);
13062 13063 ASSERT(!(iras.ira_flags & IRAF_IPSEC_SECURE));
13063 13064 return (-1);
13064 13065
13065 13066 bad_src_route:
13066 13067 bzero(&iras, sizeof (iras));
13067 13068 iras.ira_ill = iras.ira_rill = ill;
13068 13069 iras.ira_ruifindex = ill->ill_phyint->phyint_ifindex;
13069 13070 iras.ira_rifindex = iras.ira_ruifindex;
13070 13071 iras.ira_flags = IRAF_IS_IPV4;
13071 13072
13072 13073 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ill);
13073 13074 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, &iras);
13074 13075 ASSERT(!(iras.ira_flags & IRAF_IPSEC_SECURE));
13075 13076 return (-1);
13076 13077 }
13077 13078
13078 13079 /*
13079 13080 * The maximum value of conn_drain_list_cnt is CONN_MAXDRAINCNT.
13080 13081 * conn_drain_list_cnt can be changed by setting conn_drain_nthreads
13081 13082 * thru /etc/system.
13082 13083 */
13083 13084 #define CONN_MAXDRAINCNT 64
13084 13085
13085 13086 static void
13086 13087 conn_drain_init(ip_stack_t *ipst)
13087 13088 {
13088 13089 int i, j;
13089 13090 idl_tx_list_t *itl_tx;
13090 13091
13091 13092 ipst->ips_conn_drain_list_cnt = conn_drain_nthreads;
13092 13093
13093 13094 if ((ipst->ips_conn_drain_list_cnt == 0) ||
13094 13095 (ipst->ips_conn_drain_list_cnt > CONN_MAXDRAINCNT)) {
13095 13096 /*
13096 13097 * Default value of the number of drainers is the
13097 13098 * number of cpus, subject to maximum of 8 drainers.
13098 13099 */
13099 13100 if (boot_max_ncpus != -1)
13100 13101 ipst->ips_conn_drain_list_cnt = MIN(boot_max_ncpus, 8);
13101 13102 else
13102 13103 ipst->ips_conn_drain_list_cnt = MIN(max_ncpus, 8);
13103 13104 }
13104 13105
13105 13106 ipst->ips_idl_tx_list =
13106 13107 kmem_zalloc(TX_FANOUT_SIZE * sizeof (idl_tx_list_t), KM_SLEEP);
13107 13108 for (i = 0; i < TX_FANOUT_SIZE; i++) {
13108 13109 itl_tx = &ipst->ips_idl_tx_list[i];
13109 13110 itl_tx->txl_drain_list =
13110 13111 kmem_zalloc(ipst->ips_conn_drain_list_cnt *
13111 13112 sizeof (idl_t), KM_SLEEP);
13112 13113 mutex_init(&itl_tx->txl_lock, NULL, MUTEX_DEFAULT, NULL);
13113 13114 for (j = 0; j < ipst->ips_conn_drain_list_cnt; j++) {
13114 13115 mutex_init(&itl_tx->txl_drain_list[j].idl_lock, NULL,
13115 13116 MUTEX_DEFAULT, NULL);
13116 13117 itl_tx->txl_drain_list[j].idl_itl = itl_tx;
13117 13118 }
13118 13119 }
13119 13120 }
13120 13121
13121 13122 static void
13122 13123 conn_drain_fini(ip_stack_t *ipst)
13123 13124 {
13124 13125 int i;
13125 13126 idl_tx_list_t *itl_tx;
13126 13127
13127 13128 for (i = 0; i < TX_FANOUT_SIZE; i++) {
13128 13129 itl_tx = &ipst->ips_idl_tx_list[i];
13129 13130 kmem_free(itl_tx->txl_drain_list,
13130 13131 ipst->ips_conn_drain_list_cnt * sizeof (idl_t));
13131 13132 }
13132 13133 kmem_free(ipst->ips_idl_tx_list,
13133 13134 TX_FANOUT_SIZE * sizeof (idl_tx_list_t));
13134 13135 ipst->ips_idl_tx_list = NULL;
13135 13136 }
13136 13137
13137 13138 /*
13138 13139 * Flow control has blocked us from proceeding. Insert the given conn in one
13139 13140 * of the conn drain lists. When flow control is unblocked, either ip_wsrv()
13140 13141 * (STREAMS) or ill_flow_enable() (direct) will be called back, which in turn
13141 13142 * will call conn_walk_drain(). See the flow control notes at the top of this
13142 13143 * file for more details.
13143 13144 */
13144 13145 void
13145 13146 conn_drain_insert(conn_t *connp, idl_tx_list_t *tx_list)
13146 13147 {
13147 13148 idl_t *idl = tx_list->txl_drain_list;
13148 13149 uint_t index;
13149 13150 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
13150 13151
13151 13152 mutex_enter(&connp->conn_lock);
13152 13153 if (connp->conn_state_flags & CONN_CLOSING) {
13153 13154 /*
13154 13155 * The conn is closing as a result of which CONN_CLOSING
13155 13156 * is set. Return.
13156 13157 */
13157 13158 mutex_exit(&connp->conn_lock);
13158 13159 return;
13159 13160 } else if (connp->conn_idl == NULL) {
13160 13161 /*
13161 13162 * Assign the next drain list round robin. We dont' use
13162 13163 * a lock, and thus it may not be strictly round robin.
13163 13164 * Atomicity of load/stores is enough to make sure that
13164 13165 * conn_drain_list_index is always within bounds.
13165 13166 */
13166 13167 index = tx_list->txl_drain_index;
13167 13168 ASSERT(index < ipst->ips_conn_drain_list_cnt);
13168 13169 connp->conn_idl = &tx_list->txl_drain_list[index];
13169 13170 index++;
13170 13171 if (index == ipst->ips_conn_drain_list_cnt)
13171 13172 index = 0;
13172 13173 tx_list->txl_drain_index = index;
13173 13174 } else {
13174 13175 ASSERT(connp->conn_idl->idl_itl == tx_list);
13175 13176 }
13176 13177 mutex_exit(&connp->conn_lock);
13177 13178
13178 13179 idl = connp->conn_idl;
13179 13180 mutex_enter(&idl->idl_lock);
13180 13181 if ((connp->conn_drain_prev != NULL) ||
13181 13182 (connp->conn_state_flags & CONN_CLOSING)) {
13182 13183 /*
13183 13184 * The conn is either already in the drain list or closing.
13184 13185 * (We needed to check for CONN_CLOSING again since close can
13185 13186 * sneak in between dropping conn_lock and acquiring idl_lock.)
13186 13187 */
13187 13188 mutex_exit(&idl->idl_lock);
13188 13189 return;
13189 13190 }
13190 13191
13191 13192 /*
13192 13193 * The conn is not in the drain list. Insert it at the
13193 13194 * tail of the drain list. The drain list is circular
13194 13195 * and doubly linked. idl_conn points to the 1st element
13195 13196 * in the list.
13196 13197 */
13197 13198 if (idl->idl_conn == NULL) {
13198 13199 idl->idl_conn = connp;
13199 13200 connp->conn_drain_next = connp;
13200 13201 connp->conn_drain_prev = connp;
13201 13202 } else {
13202 13203 conn_t *head = idl->idl_conn;
13203 13204
13204 13205 connp->conn_drain_next = head;
13205 13206 connp->conn_drain_prev = head->conn_drain_prev;
13206 13207 head->conn_drain_prev->conn_drain_next = connp;
13207 13208 head->conn_drain_prev = connp;
13208 13209 }
13209 13210 /*
13210 13211 * For non streams based sockets assert flow control.
13211 13212 */
13212 13213 conn_setqfull(connp, NULL);
13213 13214 mutex_exit(&idl->idl_lock);
13214 13215 }
13215 13216
13216 13217 static void
13217 13218 conn_drain_remove(conn_t *connp)
13218 13219 {
13219 13220 idl_t *idl = connp->conn_idl;
13220 13221
13221 13222 if (idl != NULL) {
13222 13223 /*
13223 13224 * Remove ourself from the drain list.
13224 13225 */
13225 13226 if (connp->conn_drain_next == connp) {
13226 13227 /* Singleton in the list */
13227 13228 ASSERT(connp->conn_drain_prev == connp);
13228 13229 idl->idl_conn = NULL;
13229 13230 } else {
13230 13231 connp->conn_drain_prev->conn_drain_next =
13231 13232 connp->conn_drain_next;
13232 13233 connp->conn_drain_next->conn_drain_prev =
13233 13234 connp->conn_drain_prev;
13234 13235 if (idl->idl_conn == connp)
13235 13236 idl->idl_conn = connp->conn_drain_next;
13236 13237 }
13237 13238
13238 13239 /*
13239 13240 * NOTE: because conn_idl is associated with a specific drain
13240 13241 * list which in turn is tied to the index the TX ring
13241 13242 * (txl_cookie) hashes to, and because the TX ring can change
13242 13243 * over the lifetime of the conn_t, we must clear conn_idl so
13243 13244 * a subsequent conn_drain_insert() will set conn_idl again
13244 13245 * based on the latest txl_cookie.
13245 13246 */
13246 13247 connp->conn_idl = NULL;
13247 13248 }
13248 13249 connp->conn_drain_next = NULL;
13249 13250 connp->conn_drain_prev = NULL;
13250 13251
13251 13252 conn_clrqfull(connp, NULL);
13252 13253 /*
13253 13254 * For streams based sockets open up flow control.
13254 13255 */
13255 13256 if (!IPCL_IS_NONSTR(connp))
13256 13257 enableok(connp->conn_wq);
13257 13258 }
13258 13259
13259 13260 /*
13260 13261 * This conn is closing, and we are called from ip_close. OR
13261 13262 * this conn is draining because flow-control on the ill has been relieved.
13262 13263 *
13263 13264 * We must also need to remove conn's on this idl from the list, and also
13264 13265 * inform the sockfs upcalls about the change in flow-control.
13265 13266 */
13266 13267 static void
13267 13268 conn_drain(conn_t *connp, boolean_t closing)
13268 13269 {
13269 13270 idl_t *idl;
13270 13271 conn_t *next_connp;
13271 13272
13272 13273 /*
13273 13274 * connp->conn_idl is stable at this point, and no lock is needed
13274 13275 * to check it. If we are called from ip_close, close has already
13275 13276 * set CONN_CLOSING, thus freezing the value of conn_idl, and
13276 13277 * called us only because conn_idl is non-null. If we are called thru
13277 13278 * service, conn_idl could be null, but it cannot change because
13278 13279 * service is single-threaded per queue, and there cannot be another
13279 13280 * instance of service trying to call conn_drain_insert on this conn
13280 13281 * now.
13281 13282 */
13282 13283 ASSERT(!closing || connp == NULL || connp->conn_idl != NULL);
13283 13284
13284 13285 /*
13285 13286 * If the conn doesn't exist or is not on a drain list, bail.
13286 13287 */
13287 13288 if (connp == NULL || connp->conn_idl == NULL ||
13288 13289 connp->conn_drain_prev == NULL) {
13289 13290 return;
13290 13291 }
13291 13292
13292 13293 idl = connp->conn_idl;
13293 13294 ASSERT(MUTEX_HELD(&idl->idl_lock));
13294 13295
13295 13296 if (!closing) {
13296 13297 next_connp = connp->conn_drain_next;
13297 13298 while (next_connp != connp) {
13298 13299 conn_t *delconnp = next_connp;
13299 13300
13300 13301 next_connp = next_connp->conn_drain_next;
13301 13302 conn_drain_remove(delconnp);
13302 13303 }
13303 13304 ASSERT(connp->conn_drain_next == idl->idl_conn);
13304 13305 }
13305 13306 conn_drain_remove(connp);
13306 13307 }
13307 13308
13308 13309 /*
13309 13310 * Write service routine. Shared perimeter entry point.
13310 13311 * The device queue's messages has fallen below the low water mark and STREAMS
13311 13312 * has backenabled the ill_wq. Send sockfs notification about flow-control on
13312 13313 * each waiting conn.
13313 13314 */
13314 13315 void
13315 13316 ip_wsrv(queue_t *q)
13316 13317 {
13317 13318 ill_t *ill;
13318 13319
13319 13320 ill = (ill_t *)q->q_ptr;
13320 13321 if (ill->ill_state_flags == 0) {
13321 13322 ip_stack_t *ipst = ill->ill_ipst;
13322 13323
13323 13324 /*
13324 13325 * The device flow control has opened up.
13325 13326 * Walk through conn drain lists and qenable the
13326 13327 * first conn in each list. This makes sense only
13327 13328 * if the stream is fully plumbed and setup.
13328 13329 * Hence the ill_state_flags check above.
13329 13330 */
13330 13331 ip1dbg(("ip_wsrv: walking\n"));
13331 13332 conn_walk_drain(ipst, &ipst->ips_idl_tx_list[0]);
13332 13333 enableok(ill->ill_wq);
13333 13334 }
13334 13335 }
13335 13336
13336 13337 /*
13337 13338 * Callback to disable flow control in IP.
13338 13339 *
13339 13340 * This is a mac client callback added when the DLD_CAPAB_DIRECT capability
13340 13341 * is enabled.
13341 13342 *
13342 13343 * When MAC_TX() is not able to send any more packets, dld sets its queue
13343 13344 * to QFULL and enable the STREAMS flow control. Later, when the underlying
13344 13345 * driver is able to continue to send packets, it calls mac_tx_(ring_)update()
13345 13346 * function and wakes up corresponding mac worker threads, which in turn
13346 13347 * calls this callback function, and disables flow control.
13347 13348 */
13348 13349 void
13349 13350 ill_flow_enable(void *arg, ip_mac_tx_cookie_t cookie)
13350 13351 {
13351 13352 ill_t *ill = (ill_t *)arg;
13352 13353 ip_stack_t *ipst = ill->ill_ipst;
13353 13354 idl_tx_list_t *idl_txl;
13354 13355
13355 13356 idl_txl = &ipst->ips_idl_tx_list[IDLHASHINDEX(cookie)];
13356 13357 mutex_enter(&idl_txl->txl_lock);
13357 13358 /* add code to to set a flag to indicate idl_txl is enabled */
13358 13359 conn_walk_drain(ipst, idl_txl);
13359 13360 mutex_exit(&idl_txl->txl_lock);
13360 13361 }
13361 13362
13362 13363 /*
13363 13364 * Flow control has been relieved and STREAMS has backenabled us; drain
13364 13365 * all the conn lists on `tx_list'.
13365 13366 */
13366 13367 static void
13367 13368 conn_walk_drain(ip_stack_t *ipst, idl_tx_list_t *tx_list)
13368 13369 {
13369 13370 int i;
13370 13371 idl_t *idl;
13371 13372
13372 13373 IP_STAT(ipst, ip_conn_walk_drain);
13373 13374
13374 13375 for (i = 0; i < ipst->ips_conn_drain_list_cnt; i++) {
13375 13376 idl = &tx_list->txl_drain_list[i];
13376 13377 mutex_enter(&idl->idl_lock);
13377 13378 conn_drain(idl->idl_conn, B_FALSE);
13378 13379 mutex_exit(&idl->idl_lock);
13379 13380 }
13380 13381 }
13381 13382
13382 13383 /*
13383 13384 * Determine if the ill and multicast aspects of that packets
13384 13385 * "matches" the conn.
13385 13386 */
13386 13387 boolean_t
13387 13388 conn_wantpacket(conn_t *connp, ip_recv_attr_t *ira, ipha_t *ipha)
13388 13389 {
13389 13390 ill_t *ill = ira->ira_rill;
13390 13391 zoneid_t zoneid = ira->ira_zoneid;
13391 13392 uint_t in_ifindex;
13392 13393 ipaddr_t dst, src;
13393 13394
13394 13395 dst = ipha->ipha_dst;
13395 13396 src = ipha->ipha_src;
13396 13397
13397 13398 /*
13398 13399 * conn_incoming_ifindex is set by IP_BOUND_IF which limits
13399 13400 * unicast, broadcast and multicast reception to
13400 13401 * conn_incoming_ifindex.
13401 13402 * conn_wantpacket is called for unicast, broadcast and
13402 13403 * multicast packets.
13403 13404 */
13404 13405 in_ifindex = connp->conn_incoming_ifindex;
13405 13406
13406 13407 /* mpathd can bind to the under IPMP interface, which we allow */
13407 13408 if (in_ifindex != 0 && in_ifindex != ill->ill_phyint->phyint_ifindex) {
13408 13409 if (!IS_UNDER_IPMP(ill))
13409 13410 return (B_FALSE);
13410 13411
13411 13412 if (in_ifindex != ipmp_ill_get_ipmp_ifindex(ill))
13412 13413 return (B_FALSE);
13413 13414 }
13414 13415
13415 13416 if (!IPCL_ZONE_MATCH(connp, zoneid))
13416 13417 return (B_FALSE);
13417 13418
13418 13419 if (!(ira->ira_flags & IRAF_MULTICAST))
13419 13420 return (B_TRUE);
13420 13421
13421 13422 if (connp->conn_multi_router) {
13422 13423 /* multicast packet and multicast router socket: send up */
13423 13424 return (B_TRUE);
13424 13425 }
13425 13426
13426 13427 if (ipha->ipha_protocol == IPPROTO_PIM ||
13427 13428 ipha->ipha_protocol == IPPROTO_RSVP)
13428 13429 return (B_TRUE);
13429 13430
13430 13431 return (conn_hasmembers_ill_withsrc_v4(connp, dst, src, ira->ira_ill));
13431 13432 }
13432 13433
13433 13434 void
13434 13435 conn_setqfull(conn_t *connp, boolean_t *flow_stopped)
13435 13436 {
13436 13437 if (IPCL_IS_NONSTR(connp)) {
13437 13438 (*connp->conn_upcalls->su_txq_full)
13438 13439 (connp->conn_upper_handle, B_TRUE);
13439 13440 if (flow_stopped != NULL)
13440 13441 *flow_stopped = B_TRUE;
13441 13442 } else {
13442 13443 queue_t *q = connp->conn_wq;
13443 13444
13444 13445 ASSERT(q != NULL);
13445 13446 if (!(q->q_flag & QFULL)) {
13446 13447 mutex_enter(QLOCK(q));
13447 13448 if (!(q->q_flag & QFULL)) {
13448 13449 /* still need to set QFULL */
13449 13450 q->q_flag |= QFULL;
13450 13451 /* set flow_stopped to true under QLOCK */
13451 13452 if (flow_stopped != NULL)
13452 13453 *flow_stopped = B_TRUE;
13453 13454 mutex_exit(QLOCK(q));
13454 13455 } else {
13455 13456 /* flow_stopped is left unchanged */
13456 13457 mutex_exit(QLOCK(q));
13457 13458 }
13458 13459 }
13459 13460 }
13460 13461 }
13461 13462
13462 13463 void
13463 13464 conn_clrqfull(conn_t *connp, boolean_t *flow_stopped)
13464 13465 {
13465 13466 if (IPCL_IS_NONSTR(connp)) {
13466 13467 (*connp->conn_upcalls->su_txq_full)
13467 13468 (connp->conn_upper_handle, B_FALSE);
13468 13469 if (flow_stopped != NULL)
13469 13470 *flow_stopped = B_FALSE;
13470 13471 } else {
13471 13472 queue_t *q = connp->conn_wq;
13472 13473
13473 13474 ASSERT(q != NULL);
13474 13475 if (q->q_flag & QFULL) {
13475 13476 mutex_enter(QLOCK(q));
13476 13477 if (q->q_flag & QFULL) {
13477 13478 q->q_flag &= ~QFULL;
13478 13479 /* set flow_stopped to false under QLOCK */
13479 13480 if (flow_stopped != NULL)
13480 13481 *flow_stopped = B_FALSE;
13481 13482 mutex_exit(QLOCK(q));
13482 13483 if (q->q_flag & QWANTW)
13483 13484 qbackenable(q, 0);
13484 13485 } else {
13485 13486 /* flow_stopped is left unchanged */
13486 13487 mutex_exit(QLOCK(q));
13487 13488 }
13488 13489 }
13489 13490 }
13490 13491
13491 13492 mutex_enter(&connp->conn_lock);
13492 13493 connp->conn_blocked = B_FALSE;
13493 13494 mutex_exit(&connp->conn_lock);
13494 13495 }
13495 13496
13496 13497 /*
13497 13498 * Return the length in bytes of the IPv4 headers (base header, label, and
13498 13499 * other IP options) that will be needed based on the
13499 13500 * ip_pkt_t structure passed by the caller.
13500 13501 *
13501 13502 * The returned length does not include the length of the upper level
13502 13503 * protocol (ULP) header.
13503 13504 * The caller needs to check that the length doesn't exceed the max for IPv4.
13504 13505 */
13505 13506 int
13506 13507 ip_total_hdrs_len_v4(const ip_pkt_t *ipp)
13507 13508 {
13508 13509 int len;
13509 13510
13510 13511 len = IP_SIMPLE_HDR_LENGTH;
13511 13512 if (ipp->ipp_fields & IPPF_LABEL_V4) {
13512 13513 ASSERT(ipp->ipp_label_len_v4 != 0);
13513 13514 /* We need to round up here */
13514 13515 len += (ipp->ipp_label_len_v4 + 3) & ~3;
13515 13516 }
13516 13517
13517 13518 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
13518 13519 ASSERT(ipp->ipp_ipv4_options_len != 0);
13519 13520 ASSERT((ipp->ipp_ipv4_options_len & 3) == 0);
13520 13521 len += ipp->ipp_ipv4_options_len;
13521 13522 }
13522 13523 return (len);
13523 13524 }
13524 13525
13525 13526 /*
13526 13527 * All-purpose routine to build an IPv4 header with options based
13527 13528 * on the abstract ip_pkt_t.
13528 13529 *
13529 13530 * The caller has to set the source and destination address as well as
13530 13531 * ipha_length. The caller has to massage any source route and compensate
13531 13532 * for the ULP pseudo-header checksum due to the source route.
13532 13533 */
13533 13534 void
13534 13535 ip_build_hdrs_v4(uchar_t *buf, uint_t buf_len, const ip_pkt_t *ipp,
13535 13536 uint8_t protocol)
13536 13537 {
13537 13538 ipha_t *ipha = (ipha_t *)buf;
13538 13539 uint8_t *cp;
13539 13540
13540 13541 /* Initialize IPv4 header */
13541 13542 ipha->ipha_type_of_service = ipp->ipp_type_of_service;
13542 13543 ipha->ipha_length = 0; /* Caller will set later */
13543 13544 ipha->ipha_ident = 0;
13544 13545 ipha->ipha_fragment_offset_and_flags = 0;
13545 13546 ipha->ipha_ttl = ipp->ipp_unicast_hops;
13546 13547 ipha->ipha_protocol = protocol;
13547 13548 ipha->ipha_hdr_checksum = 0;
13548 13549
13549 13550 if ((ipp->ipp_fields & IPPF_ADDR) &&
13550 13551 IN6_IS_ADDR_V4MAPPED(&ipp->ipp_addr))
13551 13552 ipha->ipha_src = ipp->ipp_addr_v4;
13552 13553
13553 13554 cp = (uint8_t *)&ipha[1];
13554 13555 if (ipp->ipp_fields & IPPF_LABEL_V4) {
13555 13556 ASSERT(ipp->ipp_label_len_v4 != 0);
13556 13557 bcopy(ipp->ipp_label_v4, cp, ipp->ipp_label_len_v4);
13557 13558 cp += ipp->ipp_label_len_v4;
13558 13559 /* We need to round up here */
13559 13560 while ((uintptr_t)cp & 0x3) {
13560 13561 *cp++ = IPOPT_NOP;
13561 13562 }
13562 13563 }
13563 13564
13564 13565 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
13565 13566 ASSERT(ipp->ipp_ipv4_options_len != 0);
13566 13567 ASSERT((ipp->ipp_ipv4_options_len & 3) == 0);
13567 13568 bcopy(ipp->ipp_ipv4_options, cp, ipp->ipp_ipv4_options_len);
13568 13569 cp += ipp->ipp_ipv4_options_len;
13569 13570 }
13570 13571 ipha->ipha_version_and_hdr_length =
13571 13572 (uint8_t)((IP_VERSION << 4) + buf_len / 4);
13572 13573
13573 13574 ASSERT((int)(cp - buf) == buf_len);
13574 13575 }
13575 13576
13576 13577 /* Allocate the private structure */
13577 13578 static int
13578 13579 ip_priv_alloc(void **bufp)
13579 13580 {
13580 13581 void *buf;
13581 13582
13582 13583 if ((buf = kmem_alloc(sizeof (ip_priv_t), KM_NOSLEEP)) == NULL)
13583 13584 return (ENOMEM);
13584 13585
13585 13586 *bufp = buf;
13586 13587 return (0);
13587 13588 }
13588 13589
13589 13590 /* Function to delete the private structure */
13590 13591 void
13591 13592 ip_priv_free(void *buf)
13592 13593 {
13593 13594 ASSERT(buf != NULL);
13594 13595 kmem_free(buf, sizeof (ip_priv_t));
13595 13596 }
13596 13597
13597 13598 /*
13598 13599 * The entry point for IPPF processing.
13599 13600 * If the classifier (IPGPC_CLASSIFY) is not loaded and configured, the
13600 13601 * routine just returns.
13601 13602 *
13602 13603 * When called, ip_process generates an ipp_packet_t structure
13603 13604 * which holds the state information for this packet and invokes the
13604 13605 * the classifier (via ipp_packet_process). The classification, depending on
13605 13606 * configured filters, results in a list of actions for this packet. Invoking
13606 13607 * an action may cause the packet to be dropped, in which case we return NULL.
13607 13608 * proc indicates the callout position for
13608 13609 * this packet and ill is the interface this packet arrived on or will leave
13609 13610 * on (inbound and outbound resp.).
13610 13611 *
13611 13612 * We do the processing on the rill (mapped to the upper if ipmp), but MIB
13612 13613 * on the ill corrsponding to the destination IP address.
13613 13614 */
13614 13615 mblk_t *
13615 13616 ip_process(ip_proc_t proc, mblk_t *mp, ill_t *rill, ill_t *ill)
13616 13617 {
13617 13618 ip_priv_t *priv;
13618 13619 ipp_action_id_t aid;
13619 13620 int rc = 0;
13620 13621 ipp_packet_t *pp;
13621 13622
13622 13623 /* If the classifier is not loaded, return */
13623 13624 if ((aid = ipp_action_lookup(IPGPC_CLASSIFY)) == IPP_ACTION_INVAL) {
13624 13625 return (mp);
13625 13626 }
13626 13627
13627 13628 ASSERT(mp != NULL);
13628 13629
13629 13630 /* Allocate the packet structure */
13630 13631 rc = ipp_packet_alloc(&pp, "ip", aid);
13631 13632 if (rc != 0)
13632 13633 goto drop;
13633 13634
13634 13635 /* Allocate the private structure */
13635 13636 rc = ip_priv_alloc((void **)&priv);
13636 13637 if (rc != 0) {
13637 13638 ipp_packet_free(pp);
13638 13639 goto drop;
13639 13640 }
13640 13641 priv->proc = proc;
13641 13642 priv->ill_index = ill_get_upper_ifindex(rill);
13642 13643
13643 13644 ipp_packet_set_private(pp, priv, ip_priv_free);
13644 13645 ipp_packet_set_data(pp, mp);
13645 13646
13646 13647 /* Invoke the classifier */
13647 13648 rc = ipp_packet_process(&pp);
13648 13649 if (pp != NULL) {
13649 13650 mp = ipp_packet_get_data(pp);
13650 13651 ipp_packet_free(pp);
13651 13652 if (rc != 0)
13652 13653 goto drop;
13653 13654 return (mp);
13654 13655 } else {
13655 13656 /* No mp to trace in ip_drop_input/ip_drop_output */
13656 13657 mp = NULL;
13657 13658 }
13658 13659 drop:
13659 13660 if (proc == IPP_LOCAL_IN || proc == IPP_FWD_IN) {
13660 13661 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
13661 13662 ip_drop_input("ip_process", mp, ill);
13662 13663 } else {
13663 13664 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
13664 13665 ip_drop_output("ip_process", mp, ill);
13665 13666 }
13666 13667 freemsg(mp);
13667 13668 return (NULL);
13668 13669 }
13669 13670
13670 13671 /*
13671 13672 * Propagate a multicast group membership operation (add/drop) on
13672 13673 * all the interfaces crossed by the related multirt routes.
13673 13674 * The call is considered successful if the operation succeeds
13674 13675 * on at least one interface.
13675 13676 *
13676 13677 * This assumes that a set of IRE_HOST/RTF_MULTIRT has been created for the
13677 13678 * multicast addresses with the ire argument being the first one.
13678 13679 * We walk the bucket to find all the of those.
13679 13680 *
13680 13681 * Common to IPv4 and IPv6.
13681 13682 */
13682 13683 static int
13683 13684 ip_multirt_apply_membership(int (*fn)(conn_t *, boolean_t,
13684 13685 const in6_addr_t *, ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *),
13685 13686 ire_t *ire, conn_t *connp, boolean_t checkonly, const in6_addr_t *v6group,
13686 13687 mcast_record_t fmode, const in6_addr_t *v6src)
13687 13688 {
13688 13689 ire_t *ire_gw;
13689 13690 irb_t *irb;
13690 13691 int ifindex;
13691 13692 int error = 0;
13692 13693 int result;
13693 13694 ip_stack_t *ipst = ire->ire_ipst;
13694 13695 ipaddr_t group;
13695 13696 boolean_t isv6;
13696 13697 int match_flags;
13697 13698
13698 13699 if (IN6_IS_ADDR_V4MAPPED(v6group)) {
13699 13700 IN6_V4MAPPED_TO_IPADDR(v6group, group);
13700 13701 isv6 = B_FALSE;
13701 13702 } else {
13702 13703 isv6 = B_TRUE;
13703 13704 }
13704 13705
13705 13706 irb = ire->ire_bucket;
13706 13707 ASSERT(irb != NULL);
13707 13708
13708 13709 result = 0;
13709 13710 irb_refhold(irb);
13710 13711 for (; ire != NULL; ire = ire->ire_next) {
13711 13712 if ((ire->ire_flags & RTF_MULTIRT) == 0)
13712 13713 continue;
13713 13714
13714 13715 /* We handle -ifp routes by matching on the ill if set */
13715 13716 match_flags = MATCH_IRE_TYPE;
13716 13717 if (ire->ire_ill != NULL)
13717 13718 match_flags |= MATCH_IRE_ILL;
13718 13719
13719 13720 if (isv6) {
13720 13721 if (!IN6_ARE_ADDR_EQUAL(&ire->ire_addr_v6, v6group))
13721 13722 continue;
13722 13723
13723 13724 ire_gw = ire_ftable_lookup_v6(&ire->ire_gateway_addr_v6,
13724 13725 0, 0, IRE_INTERFACE, ire->ire_ill, ALL_ZONES, NULL,
13725 13726 match_flags, 0, ipst, NULL);
13726 13727 } else {
13727 13728 if (ire->ire_addr != group)
13728 13729 continue;
13729 13730
13730 13731 ire_gw = ire_ftable_lookup_v4(ire->ire_gateway_addr,
13731 13732 0, 0, IRE_INTERFACE, ire->ire_ill, ALL_ZONES, NULL,
13732 13733 match_flags, 0, ipst, NULL);
13733 13734 }
13734 13735 /* No interface route exists for the gateway; skip this ire. */
13735 13736 if (ire_gw == NULL)
13736 13737 continue;
13737 13738 if (ire_gw->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
13738 13739 ire_refrele(ire_gw);
13739 13740 continue;
13740 13741 }
13741 13742 ASSERT(ire_gw->ire_ill != NULL); /* IRE_INTERFACE */
13742 13743 ifindex = ire_gw->ire_ill->ill_phyint->phyint_ifindex;
13743 13744
13744 13745 /*
13745 13746 * The operation is considered a success if
13746 13747 * it succeeds at least once on any one interface.
13747 13748 */
13748 13749 error = fn(connp, checkonly, v6group, INADDR_ANY, ifindex,
13749 13750 fmode, v6src);
13750 13751 if (error == 0)
13751 13752 result = CGTP_MCAST_SUCCESS;
13752 13753
13753 13754 ire_refrele(ire_gw);
13754 13755 }
13755 13756 irb_refrele(irb);
13756 13757 /*
13757 13758 * Consider the call as successful if we succeeded on at least
13758 13759 * one interface. Otherwise, return the last encountered error.
13759 13760 */
13760 13761 return (result == CGTP_MCAST_SUCCESS ? 0 : error);
13761 13762 }
13762 13763
13763 13764 /*
13764 13765 * Return the expected CGTP hooks version number.
13765 13766 */
13766 13767 int
13767 13768 ip_cgtp_filter_supported(void)
13768 13769 {
13769 13770 return (ip_cgtp_filter_rev);
13770 13771 }
13771 13772
13772 13773 /*
13773 13774 * CGTP hooks can be registered by invoking this function.
13774 13775 * Checks that the version number matches.
13775 13776 */
13776 13777 int
13777 13778 ip_cgtp_filter_register(netstackid_t stackid, cgtp_filter_ops_t *ops)
13778 13779 {
13779 13780 netstack_t *ns;
13780 13781 ip_stack_t *ipst;
13781 13782
13782 13783 if (ops->cfo_filter_rev != CGTP_FILTER_REV)
13783 13784 return (ENOTSUP);
13784 13785
13785 13786 ns = netstack_find_by_stackid(stackid);
13786 13787 if (ns == NULL)
13787 13788 return (EINVAL);
13788 13789 ipst = ns->netstack_ip;
13789 13790 ASSERT(ipst != NULL);
13790 13791
13791 13792 if (ipst->ips_ip_cgtp_filter_ops != NULL) {
13792 13793 netstack_rele(ns);
13793 13794 return (EALREADY);
13794 13795 }
13795 13796
13796 13797 ipst->ips_ip_cgtp_filter_ops = ops;
13797 13798
13798 13799 ill_set_inputfn_all(ipst);
13799 13800
13800 13801 netstack_rele(ns);
13801 13802 return (0);
13802 13803 }
13803 13804
13804 13805 /*
13805 13806 * CGTP hooks can be unregistered by invoking this function.
13806 13807 * Returns ENXIO if there was no registration.
13807 13808 * Returns EBUSY if the ndd variable has not been turned off.
13808 13809 */
13809 13810 int
13810 13811 ip_cgtp_filter_unregister(netstackid_t stackid)
13811 13812 {
13812 13813 netstack_t *ns;
13813 13814 ip_stack_t *ipst;
13814 13815
13815 13816 ns = netstack_find_by_stackid(stackid);
13816 13817 if (ns == NULL)
13817 13818 return (EINVAL);
13818 13819 ipst = ns->netstack_ip;
13819 13820 ASSERT(ipst != NULL);
13820 13821
13821 13822 if (ipst->ips_ip_cgtp_filter) {
13822 13823 netstack_rele(ns);
13823 13824 return (EBUSY);
13824 13825 }
13825 13826
13826 13827 if (ipst->ips_ip_cgtp_filter_ops == NULL) {
13827 13828 netstack_rele(ns);
13828 13829 return (ENXIO);
13829 13830 }
13830 13831 ipst->ips_ip_cgtp_filter_ops = NULL;
13831 13832
13832 13833 ill_set_inputfn_all(ipst);
13833 13834
13834 13835 netstack_rele(ns);
13835 13836 return (0);
13836 13837 }
13837 13838
13838 13839 /*
13839 13840 * Check whether there is a CGTP filter registration.
13840 13841 * Returns non-zero if there is a registration, otherwise returns zero.
13841 13842 * Note: returns zero if bad stackid.
13842 13843 */
13843 13844 int
13844 13845 ip_cgtp_filter_is_registered(netstackid_t stackid)
13845 13846 {
13846 13847 netstack_t *ns;
13847 13848 ip_stack_t *ipst;
13848 13849 int ret;
13849 13850
13850 13851 ns = netstack_find_by_stackid(stackid);
13851 13852 if (ns == NULL)
13852 13853 return (0);
13853 13854 ipst = ns->netstack_ip;
13854 13855 ASSERT(ipst != NULL);
13855 13856
13856 13857 if (ipst->ips_ip_cgtp_filter_ops != NULL)
13857 13858 ret = 1;
13858 13859 else
13859 13860 ret = 0;
13860 13861
13861 13862 netstack_rele(ns);
13862 13863 return (ret);
13863 13864 }
13864 13865
13865 13866 static int
13866 13867 ip_squeue_switch(int val)
13867 13868 {
13868 13869 int rval;
13869 13870
13870 13871 switch (val) {
13871 13872 case IP_SQUEUE_ENTER_NODRAIN:
13872 13873 rval = SQ_NODRAIN;
13873 13874 break;
13874 13875 case IP_SQUEUE_ENTER:
13875 13876 rval = SQ_PROCESS;
13876 13877 break;
13877 13878 case IP_SQUEUE_FILL:
13878 13879 default:
13879 13880 rval = SQ_FILL;
13880 13881 break;
13881 13882 }
13882 13883 return (rval);
13883 13884 }
13884 13885
13885 13886 static void *
13886 13887 ip_kstat2_init(netstackid_t stackid, ip_stat_t *ip_statisticsp)
13887 13888 {
13888 13889 kstat_t *ksp;
13889 13890
13890 13891 ip_stat_t template = {
13891 13892 { "ip_udp_fannorm", KSTAT_DATA_UINT64 },
13892 13893 { "ip_udp_fanmb", KSTAT_DATA_UINT64 },
13893 13894 { "ip_recv_pullup", KSTAT_DATA_UINT64 },
13894 13895 { "ip_db_ref", KSTAT_DATA_UINT64 },
13895 13896 { "ip_notaligned", KSTAT_DATA_UINT64 },
13896 13897 { "ip_multimblk", KSTAT_DATA_UINT64 },
13897 13898 { "ip_opt", KSTAT_DATA_UINT64 },
13898 13899 { "ipsec_proto_ahesp", KSTAT_DATA_UINT64 },
13899 13900 { "ip_conn_flputbq", KSTAT_DATA_UINT64 },
13900 13901 { "ip_conn_walk_drain", KSTAT_DATA_UINT64 },
13901 13902 { "ip_out_sw_cksum", KSTAT_DATA_UINT64 },
13902 13903 { "ip_out_sw_cksum_bytes", KSTAT_DATA_UINT64 },
13903 13904 { "ip_in_sw_cksum", KSTAT_DATA_UINT64 },
13904 13905 { "ip_ire_reclaim_calls", KSTAT_DATA_UINT64 },
13905 13906 { "ip_ire_reclaim_deleted", KSTAT_DATA_UINT64 },
13906 13907 { "ip_nce_reclaim_calls", KSTAT_DATA_UINT64 },
13907 13908 { "ip_nce_reclaim_deleted", KSTAT_DATA_UINT64 },
13908 13909 { "ip_dce_reclaim_calls", KSTAT_DATA_UINT64 },
13909 13910 { "ip_dce_reclaim_deleted", KSTAT_DATA_UINT64 },
13910 13911 { "ip_tcp_in_full_hw_cksum_err", KSTAT_DATA_UINT64 },
13911 13912 { "ip_tcp_in_part_hw_cksum_err", KSTAT_DATA_UINT64 },
13912 13913 { "ip_tcp_in_sw_cksum_err", KSTAT_DATA_UINT64 },
13913 13914 { "ip_udp_in_full_hw_cksum_err", KSTAT_DATA_UINT64 },
13914 13915 { "ip_udp_in_part_hw_cksum_err", KSTAT_DATA_UINT64 },
13915 13916 { "ip_udp_in_sw_cksum_err", KSTAT_DATA_UINT64 },
13916 13917 { "conn_in_recvdstaddr", KSTAT_DATA_UINT64 },
13917 13918 { "conn_in_recvopts", KSTAT_DATA_UINT64 },
13918 13919 { "conn_in_recvif", KSTAT_DATA_UINT64 },
13919 13920 { "conn_in_recvslla", KSTAT_DATA_UINT64 },
13920 13921 { "conn_in_recvucred", KSTAT_DATA_UINT64 },
13921 13922 { "conn_in_recvttl", KSTAT_DATA_UINT64 },
13922 13923 { "conn_in_recvhopopts", KSTAT_DATA_UINT64 },
13923 13924 { "conn_in_recvhoplimit", KSTAT_DATA_UINT64 },
13924 13925 { "conn_in_recvdstopts", KSTAT_DATA_UINT64 },
13925 13926 { "conn_in_recvrthdrdstopts", KSTAT_DATA_UINT64 },
13926 13927 { "conn_in_recvrthdr", KSTAT_DATA_UINT64 },
13927 13928 { "conn_in_recvpktinfo", KSTAT_DATA_UINT64 },
13928 13929 { "conn_in_recvtclass", KSTAT_DATA_UINT64 },
13929 13930 { "conn_in_timestamp", KSTAT_DATA_UINT64 },
13930 13931 };
13931 13932
13932 13933 ksp = kstat_create_netstack("ip", 0, "ipstat", "net",
13933 13934 KSTAT_TYPE_NAMED, sizeof (template) / sizeof (kstat_named_t),
13934 13935 KSTAT_FLAG_VIRTUAL, stackid);
13935 13936
13936 13937 if (ksp == NULL)
13937 13938 return (NULL);
13938 13939
13939 13940 bcopy(&template, ip_statisticsp, sizeof (template));
13940 13941 ksp->ks_data = (void *)ip_statisticsp;
13941 13942 ksp->ks_private = (void *)(uintptr_t)stackid;
13942 13943
13943 13944 kstat_install(ksp);
13944 13945 return (ksp);
13945 13946 }
13946 13947
13947 13948 static void
13948 13949 ip_kstat2_fini(netstackid_t stackid, kstat_t *ksp)
13949 13950 {
13950 13951 if (ksp != NULL) {
13951 13952 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
13952 13953 kstat_delete_netstack(ksp, stackid);
13953 13954 }
13954 13955 }
13955 13956
13956 13957 static void *
13957 13958 ip_kstat_init(netstackid_t stackid, ip_stack_t *ipst)
13958 13959 {
13959 13960 kstat_t *ksp;
13960 13961
13961 13962 ip_named_kstat_t template = {
13962 13963 { "forwarding", KSTAT_DATA_UINT32, 0 },
13963 13964 { "defaultTTL", KSTAT_DATA_UINT32, 0 },
13964 13965 { "inReceives", KSTAT_DATA_UINT64, 0 },
13965 13966 { "inHdrErrors", KSTAT_DATA_UINT32, 0 },
13966 13967 { "inAddrErrors", KSTAT_DATA_UINT32, 0 },
13967 13968 { "forwDatagrams", KSTAT_DATA_UINT64, 0 },
13968 13969 { "inUnknownProtos", KSTAT_DATA_UINT32, 0 },
13969 13970 { "inDiscards", KSTAT_DATA_UINT32, 0 },
13970 13971 { "inDelivers", KSTAT_DATA_UINT64, 0 },
13971 13972 { "outRequests", KSTAT_DATA_UINT64, 0 },
13972 13973 { "outDiscards", KSTAT_DATA_UINT32, 0 },
13973 13974 { "outNoRoutes", KSTAT_DATA_UINT32, 0 },
13974 13975 { "reasmTimeout", KSTAT_DATA_UINT32, 0 },
13975 13976 { "reasmReqds", KSTAT_DATA_UINT32, 0 },
13976 13977 { "reasmOKs", KSTAT_DATA_UINT32, 0 },
13977 13978 { "reasmFails", KSTAT_DATA_UINT32, 0 },
13978 13979 { "fragOKs", KSTAT_DATA_UINT32, 0 },
13979 13980 { "fragFails", KSTAT_DATA_UINT32, 0 },
13980 13981 { "fragCreates", KSTAT_DATA_UINT32, 0 },
13981 13982 { "addrEntrySize", KSTAT_DATA_INT32, 0 },
13982 13983 { "routeEntrySize", KSTAT_DATA_INT32, 0 },
13983 13984 { "netToMediaEntrySize", KSTAT_DATA_INT32, 0 },
13984 13985 { "routingDiscards", KSTAT_DATA_UINT32, 0 },
13985 13986 { "inErrs", KSTAT_DATA_UINT32, 0 },
13986 13987 { "noPorts", KSTAT_DATA_UINT32, 0 },
13987 13988 { "inCksumErrs", KSTAT_DATA_UINT32, 0 },
13988 13989 { "reasmDuplicates", KSTAT_DATA_UINT32, 0 },
13989 13990 { "reasmPartDups", KSTAT_DATA_UINT32, 0 },
13990 13991 { "forwProhibits", KSTAT_DATA_UINT32, 0 },
13991 13992 { "udpInCksumErrs", KSTAT_DATA_UINT32, 0 },
13992 13993 { "udpInOverflows", KSTAT_DATA_UINT32, 0 },
13993 13994 { "rawipInOverflows", KSTAT_DATA_UINT32, 0 },
13994 13995 { "ipsecInSucceeded", KSTAT_DATA_UINT32, 0 },
13995 13996 { "ipsecInFailed", KSTAT_DATA_INT32, 0 },
13996 13997 { "memberEntrySize", KSTAT_DATA_INT32, 0 },
13997 13998 { "inIPv6", KSTAT_DATA_UINT32, 0 },
13998 13999 { "outIPv6", KSTAT_DATA_UINT32, 0 },
13999 14000 { "outSwitchIPv6", KSTAT_DATA_UINT32, 0 },
14000 14001 };
14001 14002
14002 14003 ksp = kstat_create_netstack("ip", 0, "ip", "mib2", KSTAT_TYPE_NAMED,
14003 14004 NUM_OF_FIELDS(ip_named_kstat_t), 0, stackid);
14004 14005 if (ksp == NULL || ksp->ks_data == NULL)
14005 14006 return (NULL);
14006 14007
14007 14008 template.forwarding.value.ui32 = WE_ARE_FORWARDING(ipst) ? 1:2;
14008 14009 template.defaultTTL.value.ui32 = (uint32_t)ipst->ips_ip_def_ttl;
14009 14010 template.reasmTimeout.value.ui32 = ipst->ips_ip_reassembly_timeout;
14010 14011 template.addrEntrySize.value.i32 = sizeof (mib2_ipAddrEntry_t);
14011 14012 template.routeEntrySize.value.i32 = sizeof (mib2_ipRouteEntry_t);
14012 14013
14013 14014 template.netToMediaEntrySize.value.i32 =
14014 14015 sizeof (mib2_ipNetToMediaEntry_t);
14015 14016
14016 14017 template.memberEntrySize.value.i32 = sizeof (ipv6_member_t);
14017 14018
14018 14019 bcopy(&template, ksp->ks_data, sizeof (template));
14019 14020 ksp->ks_update = ip_kstat_update;
14020 14021 ksp->ks_private = (void *)(uintptr_t)stackid;
14021 14022
14022 14023 kstat_install(ksp);
14023 14024 return (ksp);
14024 14025 }
14025 14026
14026 14027 static void
14027 14028 ip_kstat_fini(netstackid_t stackid, kstat_t *ksp)
14028 14029 {
14029 14030 if (ksp != NULL) {
14030 14031 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
14031 14032 kstat_delete_netstack(ksp, stackid);
14032 14033 }
14033 14034 }
14034 14035
14035 14036 static int
14036 14037 ip_kstat_update(kstat_t *kp, int rw)
14037 14038 {
14038 14039 ip_named_kstat_t *ipkp;
14039 14040 mib2_ipIfStatsEntry_t ipmib;
14040 14041 ill_walk_context_t ctx;
14041 14042 ill_t *ill;
14042 14043 netstackid_t stackid = (zoneid_t)(uintptr_t)kp->ks_private;
14043 14044 netstack_t *ns;
14044 14045 ip_stack_t *ipst;
14045 14046
14046 14047 if (kp == NULL || kp->ks_data == NULL)
14047 14048 return (EIO);
14048 14049
14049 14050 if (rw == KSTAT_WRITE)
14050 14051 return (EACCES);
14051 14052
14052 14053 ns = netstack_find_by_stackid(stackid);
14053 14054 if (ns == NULL)
14054 14055 return (-1);
14055 14056 ipst = ns->netstack_ip;
14056 14057 if (ipst == NULL) {
14057 14058 netstack_rele(ns);
14058 14059 return (-1);
14059 14060 }
14060 14061 ipkp = (ip_named_kstat_t *)kp->ks_data;
14061 14062
14062 14063 bcopy(&ipst->ips_ip_mib, &ipmib, sizeof (ipmib));
14063 14064 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
14064 14065 ill = ILL_START_WALK_V4(&ctx, ipst);
14065 14066 for (; ill != NULL; ill = ill_next(&ctx, ill))
14066 14067 ip_mib2_add_ip_stats(&ipmib, ill->ill_ip_mib);
14067 14068 rw_exit(&ipst->ips_ill_g_lock);
14068 14069
14069 14070 ipkp->forwarding.value.ui32 = ipmib.ipIfStatsForwarding;
14070 14071 ipkp->defaultTTL.value.ui32 = ipmib.ipIfStatsDefaultTTL;
14071 14072 ipkp->inReceives.value.ui64 = ipmib.ipIfStatsHCInReceives;
14072 14073 ipkp->inHdrErrors.value.ui32 = ipmib.ipIfStatsInHdrErrors;
14073 14074 ipkp->inAddrErrors.value.ui32 = ipmib.ipIfStatsInAddrErrors;
14074 14075 ipkp->forwDatagrams.value.ui64 = ipmib.ipIfStatsHCOutForwDatagrams;
14075 14076 ipkp->inUnknownProtos.value.ui32 = ipmib.ipIfStatsInUnknownProtos;
14076 14077 ipkp->inDiscards.value.ui32 = ipmib.ipIfStatsInDiscards;
14077 14078 ipkp->inDelivers.value.ui64 = ipmib.ipIfStatsHCInDelivers;
14078 14079 ipkp->outRequests.value.ui64 = ipmib.ipIfStatsHCOutRequests;
14079 14080 ipkp->outDiscards.value.ui32 = ipmib.ipIfStatsOutDiscards;
14080 14081 ipkp->outNoRoutes.value.ui32 = ipmib.ipIfStatsOutNoRoutes;
14081 14082 ipkp->reasmTimeout.value.ui32 = ipst->ips_ip_reassembly_timeout;
14082 14083 ipkp->reasmReqds.value.ui32 = ipmib.ipIfStatsReasmReqds;
14083 14084 ipkp->reasmOKs.value.ui32 = ipmib.ipIfStatsReasmOKs;
14084 14085 ipkp->reasmFails.value.ui32 = ipmib.ipIfStatsReasmFails;
14085 14086 ipkp->fragOKs.value.ui32 = ipmib.ipIfStatsOutFragOKs;
14086 14087 ipkp->fragFails.value.ui32 = ipmib.ipIfStatsOutFragFails;
14087 14088 ipkp->fragCreates.value.ui32 = ipmib.ipIfStatsOutFragCreates;
14088 14089
14089 14090 ipkp->routingDiscards.value.ui32 = 0;
14090 14091 ipkp->inErrs.value.ui32 = ipmib.tcpIfStatsInErrs;
14091 14092 ipkp->noPorts.value.ui32 = ipmib.udpIfStatsNoPorts;
14092 14093 ipkp->inCksumErrs.value.ui32 = ipmib.ipIfStatsInCksumErrs;
14093 14094 ipkp->reasmDuplicates.value.ui32 = ipmib.ipIfStatsReasmDuplicates;
14094 14095 ipkp->reasmPartDups.value.ui32 = ipmib.ipIfStatsReasmPartDups;
14095 14096 ipkp->forwProhibits.value.ui32 = ipmib.ipIfStatsForwProhibits;
14096 14097 ipkp->udpInCksumErrs.value.ui32 = ipmib.udpIfStatsInCksumErrs;
14097 14098 ipkp->udpInOverflows.value.ui32 = ipmib.udpIfStatsInOverflows;
14098 14099 ipkp->rawipInOverflows.value.ui32 = ipmib.rawipIfStatsInOverflows;
14099 14100 ipkp->ipsecInSucceeded.value.ui32 = ipmib.ipsecIfStatsInSucceeded;
14100 14101 ipkp->ipsecInFailed.value.i32 = ipmib.ipsecIfStatsInFailed;
14101 14102
14102 14103 ipkp->inIPv6.value.ui32 = ipmib.ipIfStatsInWrongIPVersion;
14103 14104 ipkp->outIPv6.value.ui32 = ipmib.ipIfStatsOutWrongIPVersion;
14104 14105 ipkp->outSwitchIPv6.value.ui32 = ipmib.ipIfStatsOutSwitchIPVersion;
14105 14106
14106 14107 netstack_rele(ns);
14107 14108
14108 14109 return (0);
14109 14110 }
14110 14111
14111 14112 static void *
14112 14113 icmp_kstat_init(netstackid_t stackid)
14113 14114 {
14114 14115 kstat_t *ksp;
14115 14116
14116 14117 icmp_named_kstat_t template = {
14117 14118 { "inMsgs", KSTAT_DATA_UINT32 },
14118 14119 { "inErrors", KSTAT_DATA_UINT32 },
14119 14120 { "inDestUnreachs", KSTAT_DATA_UINT32 },
14120 14121 { "inTimeExcds", KSTAT_DATA_UINT32 },
14121 14122 { "inParmProbs", KSTAT_DATA_UINT32 },
14122 14123 { "inSrcQuenchs", KSTAT_DATA_UINT32 },
14123 14124 { "inRedirects", KSTAT_DATA_UINT32 },
14124 14125 { "inEchos", KSTAT_DATA_UINT32 },
14125 14126 { "inEchoReps", KSTAT_DATA_UINT32 },
14126 14127 { "inTimestamps", KSTAT_DATA_UINT32 },
14127 14128 { "inTimestampReps", KSTAT_DATA_UINT32 },
14128 14129 { "inAddrMasks", KSTAT_DATA_UINT32 },
14129 14130 { "inAddrMaskReps", KSTAT_DATA_UINT32 },
14130 14131 { "outMsgs", KSTAT_DATA_UINT32 },
14131 14132 { "outErrors", KSTAT_DATA_UINT32 },
14132 14133 { "outDestUnreachs", KSTAT_DATA_UINT32 },
14133 14134 { "outTimeExcds", KSTAT_DATA_UINT32 },
14134 14135 { "outParmProbs", KSTAT_DATA_UINT32 },
14135 14136 { "outSrcQuenchs", KSTAT_DATA_UINT32 },
14136 14137 { "outRedirects", KSTAT_DATA_UINT32 },
14137 14138 { "outEchos", KSTAT_DATA_UINT32 },
14138 14139 { "outEchoReps", KSTAT_DATA_UINT32 },
14139 14140 { "outTimestamps", KSTAT_DATA_UINT32 },
14140 14141 { "outTimestampReps", KSTAT_DATA_UINT32 },
14141 14142 { "outAddrMasks", KSTAT_DATA_UINT32 },
14142 14143 { "outAddrMaskReps", KSTAT_DATA_UINT32 },
14143 14144 { "inChksumErrs", KSTAT_DATA_UINT32 },
14144 14145 { "inUnknowns", KSTAT_DATA_UINT32 },
14145 14146 { "inFragNeeded", KSTAT_DATA_UINT32 },
14146 14147 { "outFragNeeded", KSTAT_DATA_UINT32 },
14147 14148 { "outDrops", KSTAT_DATA_UINT32 },
14148 14149 { "inOverFlows", KSTAT_DATA_UINT32 },
14149 14150 { "inBadRedirects", KSTAT_DATA_UINT32 },
14150 14151 };
14151 14152
14152 14153 ksp = kstat_create_netstack("ip", 0, "icmp", "mib2", KSTAT_TYPE_NAMED,
14153 14154 NUM_OF_FIELDS(icmp_named_kstat_t), 0, stackid);
14154 14155 if (ksp == NULL || ksp->ks_data == NULL)
14155 14156 return (NULL);
14156 14157
14157 14158 bcopy(&template, ksp->ks_data, sizeof (template));
14158 14159
14159 14160 ksp->ks_update = icmp_kstat_update;
14160 14161 ksp->ks_private = (void *)(uintptr_t)stackid;
14161 14162
14162 14163 kstat_install(ksp);
14163 14164 return (ksp);
14164 14165 }
14165 14166
14166 14167 static void
14167 14168 icmp_kstat_fini(netstackid_t stackid, kstat_t *ksp)
14168 14169 {
14169 14170 if (ksp != NULL) {
14170 14171 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
14171 14172 kstat_delete_netstack(ksp, stackid);
14172 14173 }
14173 14174 }
14174 14175
14175 14176 static int
14176 14177 icmp_kstat_update(kstat_t *kp, int rw)
14177 14178 {
14178 14179 icmp_named_kstat_t *icmpkp;
14179 14180 netstackid_t stackid = (zoneid_t)(uintptr_t)kp->ks_private;
14180 14181 netstack_t *ns;
14181 14182 ip_stack_t *ipst;
14182 14183
14183 14184 if ((kp == NULL) || (kp->ks_data == NULL))
14184 14185 return (EIO);
14185 14186
14186 14187 if (rw == KSTAT_WRITE)
14187 14188 return (EACCES);
14188 14189
14189 14190 ns = netstack_find_by_stackid(stackid);
14190 14191 if (ns == NULL)
14191 14192 return (-1);
14192 14193 ipst = ns->netstack_ip;
14193 14194 if (ipst == NULL) {
14194 14195 netstack_rele(ns);
14195 14196 return (-1);
14196 14197 }
14197 14198 icmpkp = (icmp_named_kstat_t *)kp->ks_data;
14198 14199
14199 14200 icmpkp->inMsgs.value.ui32 = ipst->ips_icmp_mib.icmpInMsgs;
14200 14201 icmpkp->inErrors.value.ui32 = ipst->ips_icmp_mib.icmpInErrors;
14201 14202 icmpkp->inDestUnreachs.value.ui32 =
14202 14203 ipst->ips_icmp_mib.icmpInDestUnreachs;
14203 14204 icmpkp->inTimeExcds.value.ui32 = ipst->ips_icmp_mib.icmpInTimeExcds;
14204 14205 icmpkp->inParmProbs.value.ui32 = ipst->ips_icmp_mib.icmpInParmProbs;
14205 14206 icmpkp->inSrcQuenchs.value.ui32 = ipst->ips_icmp_mib.icmpInSrcQuenchs;
14206 14207 icmpkp->inRedirects.value.ui32 = ipst->ips_icmp_mib.icmpInRedirects;
14207 14208 icmpkp->inEchos.value.ui32 = ipst->ips_icmp_mib.icmpInEchos;
14208 14209 icmpkp->inEchoReps.value.ui32 = ipst->ips_icmp_mib.icmpInEchoReps;
14209 14210 icmpkp->inTimestamps.value.ui32 = ipst->ips_icmp_mib.icmpInTimestamps;
14210 14211 icmpkp->inTimestampReps.value.ui32 =
14211 14212 ipst->ips_icmp_mib.icmpInTimestampReps;
14212 14213 icmpkp->inAddrMasks.value.ui32 = ipst->ips_icmp_mib.icmpInAddrMasks;
14213 14214 icmpkp->inAddrMaskReps.value.ui32 =
14214 14215 ipst->ips_icmp_mib.icmpInAddrMaskReps;
14215 14216 icmpkp->outMsgs.value.ui32 = ipst->ips_icmp_mib.icmpOutMsgs;
14216 14217 icmpkp->outErrors.value.ui32 = ipst->ips_icmp_mib.icmpOutErrors;
14217 14218 icmpkp->outDestUnreachs.value.ui32 =
14218 14219 ipst->ips_icmp_mib.icmpOutDestUnreachs;
14219 14220 icmpkp->outTimeExcds.value.ui32 = ipst->ips_icmp_mib.icmpOutTimeExcds;
14220 14221 icmpkp->outParmProbs.value.ui32 = ipst->ips_icmp_mib.icmpOutParmProbs;
14221 14222 icmpkp->outSrcQuenchs.value.ui32 =
14222 14223 ipst->ips_icmp_mib.icmpOutSrcQuenchs;
14223 14224 icmpkp->outRedirects.value.ui32 = ipst->ips_icmp_mib.icmpOutRedirects;
14224 14225 icmpkp->outEchos.value.ui32 = ipst->ips_icmp_mib.icmpOutEchos;
14225 14226 icmpkp->outEchoReps.value.ui32 = ipst->ips_icmp_mib.icmpOutEchoReps;
14226 14227 icmpkp->outTimestamps.value.ui32 =
14227 14228 ipst->ips_icmp_mib.icmpOutTimestamps;
14228 14229 icmpkp->outTimestampReps.value.ui32 =
14229 14230 ipst->ips_icmp_mib.icmpOutTimestampReps;
14230 14231 icmpkp->outAddrMasks.value.ui32 =
14231 14232 ipst->ips_icmp_mib.icmpOutAddrMasks;
14232 14233 icmpkp->outAddrMaskReps.value.ui32 =
14233 14234 ipst->ips_icmp_mib.icmpOutAddrMaskReps;
14234 14235 icmpkp->inCksumErrs.value.ui32 = ipst->ips_icmp_mib.icmpInCksumErrs;
14235 14236 icmpkp->inUnknowns.value.ui32 = ipst->ips_icmp_mib.icmpInUnknowns;
14236 14237 icmpkp->inFragNeeded.value.ui32 = ipst->ips_icmp_mib.icmpInFragNeeded;
14237 14238 icmpkp->outFragNeeded.value.ui32 =
14238 14239 ipst->ips_icmp_mib.icmpOutFragNeeded;
14239 14240 icmpkp->outDrops.value.ui32 = ipst->ips_icmp_mib.icmpOutDrops;
14240 14241 icmpkp->inOverflows.value.ui32 = ipst->ips_icmp_mib.icmpInOverflows;
14241 14242 icmpkp->inBadRedirects.value.ui32 =
14242 14243 ipst->ips_icmp_mib.icmpInBadRedirects;
14243 14244
14244 14245 netstack_rele(ns);
14245 14246 return (0);
14246 14247 }
14247 14248
14248 14249 /*
14249 14250 * This is the fanout function for raw socket opened for SCTP. Note
14250 14251 * that it is called after SCTP checks that there is no socket which
14251 14252 * wants a packet. Then before SCTP handles this out of the blue packet,
14252 14253 * this function is called to see if there is any raw socket for SCTP.
14253 14254 * If there is and it is bound to the correct address, the packet will
14254 14255 * be sent to that socket. Note that only one raw socket can be bound to
14255 14256 * a port. This is assured in ipcl_sctp_hash_insert();
14256 14257 */
14257 14258 void
14258 14259 ip_fanout_sctp_raw(mblk_t *mp, ipha_t *ipha, ip6_t *ip6h, uint32_t ports,
14259 14260 ip_recv_attr_t *ira)
14260 14261 {
14261 14262 conn_t *connp;
14262 14263 queue_t *rq;
14263 14264 boolean_t secure;
14264 14265 ill_t *ill = ira->ira_ill;
14265 14266 ip_stack_t *ipst = ill->ill_ipst;
14266 14267 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
14267 14268 sctp_stack_t *sctps = ipst->ips_netstack->netstack_sctp;
14268 14269 iaflags_t iraflags = ira->ira_flags;
14269 14270 ill_t *rill = ira->ira_rill;
14270 14271
14271 14272 secure = iraflags & IRAF_IPSEC_SECURE;
14272 14273
14273 14274 connp = ipcl_classify_raw(mp, IPPROTO_SCTP, ports, ipha, ip6h,
14274 14275 ira, ipst);
14275 14276 if (connp == NULL) {
14276 14277 /*
14277 14278 * Although raw sctp is not summed, OOB chunks must be.
14278 14279 * Drop the packet here if the sctp checksum failed.
14279 14280 */
14280 14281 if (iraflags & IRAF_SCTP_CSUM_ERR) {
14281 14282 SCTPS_BUMP_MIB(sctps, sctpChecksumError);
14282 14283 freemsg(mp);
14283 14284 return;
14284 14285 }
14285 14286 ira->ira_ill = ira->ira_rill = NULL;
14286 14287 sctp_ootb_input(mp, ira, ipst);
14287 14288 ira->ira_ill = ill;
14288 14289 ira->ira_rill = rill;
14289 14290 return;
14290 14291 }
14291 14292 rq = connp->conn_rq;
14292 14293 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld : !canputnext(rq)) {
14293 14294 CONN_DEC_REF(connp);
14294 14295 BUMP_MIB(ill->ill_ip_mib, rawipIfStatsInOverflows);
14295 14296 freemsg(mp);
14296 14297 return;
14297 14298 }
14298 14299 if (((iraflags & IRAF_IS_IPV4) ?
14299 14300 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
14300 14301 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
14301 14302 secure) {
14302 14303 mp = ipsec_check_inbound_policy(mp, connp, ipha,
14303 14304 ip6h, ira);
14304 14305 if (mp == NULL) {
14305 14306 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
14306 14307 /* Note that mp is NULL */
14307 14308 ip_drop_input("ipIfStatsInDiscards", mp, ill);
14308 14309 CONN_DEC_REF(connp);
14309 14310 return;
14310 14311 }
14311 14312 }
14312 14313
14313 14314 if (iraflags & IRAF_ICMP_ERROR) {
14314 14315 (connp->conn_recvicmp)(connp, mp, NULL, ira);
14315 14316 } else {
14316 14317 ill_t *rill = ira->ira_rill;
14317 14318
14318 14319 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
14319 14320 /* This is the SOCK_RAW, IPPROTO_SCTP case. */
14320 14321 ira->ira_ill = ira->ira_rill = NULL;
14321 14322 (connp->conn_recv)(connp, mp, NULL, ira);
14322 14323 ira->ira_ill = ill;
14323 14324 ira->ira_rill = rill;
14324 14325 }
14325 14326 CONN_DEC_REF(connp);
14326 14327 }
14327 14328
14328 14329 /*
14329 14330 * Free a packet that has the link-layer dl_unitdata_req_t or fast-path
14330 14331 * header before the ip payload.
14331 14332 */
14332 14333 static void
14333 14334 ip_xmit_flowctl_drop(ill_t *ill, mblk_t *mp, boolean_t is_fp_mp, int fp_mp_len)
14334 14335 {
14335 14336 int len = (mp->b_wptr - mp->b_rptr);
14336 14337 mblk_t *ip_mp;
14337 14338
14338 14339 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14339 14340 if (is_fp_mp || len != fp_mp_len) {
14340 14341 if (len > fp_mp_len) {
14341 14342 /*
14342 14343 * fastpath header and ip header in the first mblk
14343 14344 */
14344 14345 mp->b_rptr += fp_mp_len;
14345 14346 } else {
14346 14347 /*
14347 14348 * ip_xmit_attach_llhdr had to prepend an mblk to
14348 14349 * attach the fastpath header before ip header.
14349 14350 */
14350 14351 ip_mp = mp->b_cont;
14351 14352 freeb(mp);
14352 14353 mp = ip_mp;
14353 14354 mp->b_rptr += (fp_mp_len - len);
14354 14355 }
14355 14356 } else {
14356 14357 ip_mp = mp->b_cont;
14357 14358 freeb(mp);
14358 14359 mp = ip_mp;
14359 14360 }
14360 14361 ip_drop_output("ipIfStatsOutDiscards - flow ctl", mp, ill);
14361 14362 freemsg(mp);
14362 14363 }
14363 14364
14364 14365 /*
14365 14366 * Normal post fragmentation function.
14366 14367 *
14367 14368 * Send a packet using the passed in nce. This handles both IPv4 and IPv6
14368 14369 * using the same state machine.
14369 14370 *
14370 14371 * We return an error on failure. In particular we return EWOULDBLOCK
14371 14372 * when the driver flow controls. In that case this ensures that ip_wsrv runs
14372 14373 * (currently by canputnext failure resulting in backenabling from GLD.)
14373 14374 * This allows the callers of conn_ip_output() to use EWOULDBLOCK as an
14374 14375 * indication that they can flow control until ip_wsrv() tells then to restart.
14375 14376 *
14376 14377 * If the nce passed by caller is incomplete, this function
14377 14378 * queues the packet and if necessary, sends ARP request and bails.
14378 14379 * If the Neighbor Cache passed is fully resolved, we simply prepend
14379 14380 * the link-layer header to the packet, do ipsec hw acceleration
14380 14381 * work if necessary, and send the packet out on the wire.
14381 14382 */
14382 14383 /* ARGSUSED6 */
14383 14384 int
14384 14385 ip_xmit(mblk_t *mp, nce_t *nce, iaflags_t ixaflags, uint_t pkt_len,
14385 14386 uint32_t xmit_hint, zoneid_t szone, zoneid_t nolzid, uintptr_t *ixacookie)
14386 14387 {
14387 14388 queue_t *wq;
14388 14389 ill_t *ill = nce->nce_ill;
14389 14390 ip_stack_t *ipst = ill->ill_ipst;
14390 14391 uint64_t delta;
14391 14392 boolean_t isv6 = ill->ill_isv6;
14392 14393 boolean_t fp_mp;
14393 14394 ncec_t *ncec = nce->nce_common;
14394 14395 int64_t now = LBOLT_FASTPATH64;
14395 14396 boolean_t is_probe;
14396 14397
14397 14398 DTRACE_PROBE1(ip__xmit, nce_t *, nce);
14398 14399
14399 14400 ASSERT(mp != NULL);
14400 14401 ASSERT(mp->b_datap->db_type == M_DATA);
14401 14402 ASSERT(pkt_len == msgdsize(mp));
14402 14403
14403 14404 /*
14404 14405 * If we have already been here and are coming back after ARP/ND.
14405 14406 * the IXAF_NO_TRACE flag is set. We skip FW_HOOKS, DTRACE and ipobs
14406 14407 * in that case since they have seen the packet when it came here
14407 14408 * the first time.
14408 14409 */
14409 14410 if (ixaflags & IXAF_NO_TRACE)
14410 14411 goto sendit;
14411 14412
14412 14413 if (ixaflags & IXAF_IS_IPV4) {
14413 14414 ipha_t *ipha = (ipha_t *)mp->b_rptr;
14414 14415
14415 14416 ASSERT(!isv6);
14416 14417 ASSERT(pkt_len == ntohs(((ipha_t *)mp->b_rptr)->ipha_length));
14417 14418 if (HOOKS4_INTERESTED_PHYSICAL_OUT(ipst) &&
14418 14419 !(ixaflags & IXAF_NO_PFHOOK)) {
14419 14420 int error;
14420 14421
14421 14422 FW_HOOKS(ipst->ips_ip4_physical_out_event,
14422 14423 ipst->ips_ipv4firewall_physical_out,
14423 14424 NULL, ill, ipha, mp, mp, 0, ipst, error);
14424 14425 DTRACE_PROBE1(ip4__physical__out__end,
14425 14426 mblk_t *, mp);
14426 14427 if (mp == NULL)
14427 14428 return (error);
14428 14429
14429 14430 /* The length could have changed */
14430 14431 pkt_len = msgdsize(mp);
14431 14432 }
14432 14433 if (ipst->ips_ip4_observe.he_interested) {
14433 14434 /*
14434 14435 * Note that for TX the zoneid is the sending
14435 14436 * zone, whether or not MLP is in play.
14436 14437 * Since the szone argument is the IP zoneid (i.e.,
14437 14438 * zero for exclusive-IP zones) and ipobs wants
14438 14439 * the system zoneid, we map it here.
14439 14440 */
14440 14441 szone = IP_REAL_ZONEID(szone, ipst);
14441 14442
14442 14443 /*
14443 14444 * On the outbound path the destination zone will be
14444 14445 * unknown as we're sending this packet out on the
14445 14446 * wire.
14446 14447 */
14447 14448 ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone, ALL_ZONES,
14448 14449 ill, ipst);
14449 14450 }
14450 14451 DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
14451 14452 void_ip_t *, ipha, __dtrace_ipsr_ill_t *, ill,
14452 14453 ipha_t *, ipha, ip6_t *, NULL, int, 0);
14453 14454 } else {
14454 14455 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
14455 14456
14456 14457 ASSERT(isv6);
14457 14458 ASSERT(pkt_len ==
14458 14459 ntohs(((ip6_t *)mp->b_rptr)->ip6_plen) + IPV6_HDR_LEN);
14459 14460 if (HOOKS6_INTERESTED_PHYSICAL_OUT(ipst) &&
14460 14461 !(ixaflags & IXAF_NO_PFHOOK)) {
14461 14462 int error;
14462 14463
14463 14464 FW_HOOKS6(ipst->ips_ip6_physical_out_event,
14464 14465 ipst->ips_ipv6firewall_physical_out,
14465 14466 NULL, ill, ip6h, mp, mp, 0, ipst, error);
14466 14467 DTRACE_PROBE1(ip6__physical__out__end,
14467 14468 mblk_t *, mp);
14468 14469 if (mp == NULL)
14469 14470 return (error);
14470 14471
14471 14472 /* The length could have changed */
14472 14473 pkt_len = msgdsize(mp);
14473 14474 }
14474 14475 if (ipst->ips_ip6_observe.he_interested) {
14475 14476 /* See above */
14476 14477 szone = IP_REAL_ZONEID(szone, ipst);
14477 14478
14478 14479 ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone, ALL_ZONES,
14479 14480 ill, ipst);
14480 14481 }
14481 14482 DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
14482 14483 void_ip_t *, ip6h, __dtrace_ipsr_ill_t *, ill,
14483 14484 ipha_t *, NULL, ip6_t *, ip6h, int, 0);
14484 14485 }
14485 14486
14486 14487 sendit:
14487 14488 /*
14488 14489 * We check the state without a lock because the state can never
14489 14490 * move "backwards" to initial or incomplete.
14490 14491 */
14491 14492 switch (ncec->ncec_state) {
14492 14493 case ND_REACHABLE:
14493 14494 case ND_STALE:
14494 14495 case ND_DELAY:
14495 14496 case ND_PROBE:
14496 14497 mp = ip_xmit_attach_llhdr(mp, nce);
14497 14498 if (mp == NULL) {
14498 14499 /*
14499 14500 * ip_xmit_attach_llhdr has increased
14500 14501 * ipIfStatsOutDiscards and called ip_drop_output()
14501 14502 */
14502 14503 return (ENOBUFS);
14503 14504 }
14504 14505 /*
14505 14506 * check if nce_fastpath completed and we tagged on a
14506 14507 * copy of nce_fp_mp in ip_xmit_attach_llhdr().
14507 14508 */
14508 14509 fp_mp = (mp->b_datap->db_type == M_DATA);
14509 14510
14510 14511 if (fp_mp &&
14511 14512 (ill->ill_capabilities & ILL_CAPAB_DLD_DIRECT)) {
14512 14513 ill_dld_direct_t *idd;
14513 14514
14514 14515 idd = &ill->ill_dld_capab->idc_direct;
14515 14516 /*
14516 14517 * Send the packet directly to DLD, where it
14517 14518 * may be queued depending on the availability
14518 14519 * of transmit resources at the media layer.
14519 14520 * Return value should be taken into
14520 14521 * account and flow control the TCP.
14521 14522 */
14522 14523 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
14523 14524 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
14524 14525 pkt_len);
14525 14526
14526 14527 if (ixaflags & IXAF_NO_DEV_FLOW_CTL) {
14527 14528 (void) idd->idd_tx_df(idd->idd_tx_dh, mp,
14528 14529 (uintptr_t)xmit_hint, IP_DROP_ON_NO_DESC);
14529 14530 } else {
14530 14531 uintptr_t cookie;
14531 14532
14532 14533 if ((cookie = idd->idd_tx_df(idd->idd_tx_dh,
14533 14534 mp, (uintptr_t)xmit_hint, 0)) != 0) {
14534 14535 if (ixacookie != NULL)
14535 14536 *ixacookie = cookie;
14536 14537 return (EWOULDBLOCK);
14537 14538 }
14538 14539 }
14539 14540 } else {
14540 14541 wq = ill->ill_wq;
14541 14542
14542 14543 if (!(ixaflags & IXAF_NO_DEV_FLOW_CTL) &&
14543 14544 !canputnext(wq)) {
14544 14545 if (ixacookie != NULL)
14545 14546 *ixacookie = 0;
14546 14547 ip_xmit_flowctl_drop(ill, mp, fp_mp,
14547 14548 nce->nce_fp_mp != NULL ?
14548 14549 MBLKL(nce->nce_fp_mp) : 0);
14549 14550 return (EWOULDBLOCK);
14550 14551 }
14551 14552 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
14552 14553 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
14553 14554 pkt_len);
14554 14555 putnext(wq, mp);
14555 14556 }
14556 14557
14557 14558 /*
14558 14559 * The rest of this function implements Neighbor Unreachability
14559 14560 * detection. Determine if the ncec is eligible for NUD.
14560 14561 */
14561 14562 if (ncec->ncec_flags & NCE_F_NONUD)
14562 14563 return (0);
14563 14564
14564 14565 ASSERT(ncec->ncec_state != ND_INCOMPLETE);
14565 14566
14566 14567 /*
14567 14568 * Check for upper layer advice
14568 14569 */
14569 14570 if (ixaflags & IXAF_REACH_CONF) {
14570 14571 timeout_id_t tid;
14571 14572
14572 14573 /*
14573 14574 * It should be o.k. to check the state without
14574 14575 * a lock here, at most we lose an advice.
14575 14576 */
14576 14577 ncec->ncec_last = TICK_TO_MSEC(now);
14577 14578 if (ncec->ncec_state != ND_REACHABLE) {
14578 14579 mutex_enter(&ncec->ncec_lock);
14579 14580 ncec->ncec_state = ND_REACHABLE;
14580 14581 tid = ncec->ncec_timeout_id;
14581 14582 ncec->ncec_timeout_id = 0;
14582 14583 mutex_exit(&ncec->ncec_lock);
14583 14584 (void) untimeout(tid);
14584 14585 if (ip_debug > 2) {
14585 14586 /* ip1dbg */
14586 14587 pr_addr_dbg("ip_xmit: state"
14587 14588 " for %s changed to"
14588 14589 " REACHABLE\n", AF_INET6,
14589 14590 &ncec->ncec_addr);
14590 14591 }
14591 14592 }
14592 14593 return (0);
14593 14594 }
14594 14595
14595 14596 delta = TICK_TO_MSEC(now) - ncec->ncec_last;
14596 14597 ip1dbg(("ip_xmit: delta = %" PRId64
14597 14598 " ill_reachable_time = %d \n", delta,
14598 14599 ill->ill_reachable_time));
14599 14600 if (delta > (uint64_t)ill->ill_reachable_time) {
14600 14601 mutex_enter(&ncec->ncec_lock);
14601 14602 switch (ncec->ncec_state) {
14602 14603 case ND_REACHABLE:
14603 14604 ASSERT((ncec->ncec_flags & NCE_F_NONUD) == 0);
14604 14605 /* FALLTHROUGH */
14605 14606 case ND_STALE:
14606 14607 /*
14607 14608 * ND_REACHABLE is identical to
14608 14609 * ND_STALE in this specific case. If
14609 14610 * reachable time has expired for this
14610 14611 * neighbor (delta is greater than
14611 14612 * reachable time), conceptually, the
14612 14613 * neighbor cache is no longer in
14613 14614 * REACHABLE state, but already in
14614 14615 * STALE state. So the correct
14615 14616 * transition here is to ND_DELAY.
14616 14617 */
14617 14618 ncec->ncec_state = ND_DELAY;
14618 14619 mutex_exit(&ncec->ncec_lock);
14619 14620 nce_restart_timer(ncec,
14620 14621 ipst->ips_delay_first_probe_time);
14621 14622 if (ip_debug > 3) {
14622 14623 /* ip2dbg */
14623 14624 pr_addr_dbg("ip_xmit: state"
14624 14625 " for %s changed to"
14625 14626 " DELAY\n", AF_INET6,
14626 14627 &ncec->ncec_addr);
14627 14628 }
14628 14629 break;
14629 14630 case ND_DELAY:
14630 14631 case ND_PROBE:
14631 14632 mutex_exit(&ncec->ncec_lock);
14632 14633 /* Timers have already started */
14633 14634 break;
14634 14635 case ND_UNREACHABLE:
14635 14636 /*
14636 14637 * nce_timer has detected that this ncec
14637 14638 * is unreachable and initiated deleting
14638 14639 * this ncec.
14639 14640 * This is a harmless race where we found the
14640 14641 * ncec before it was deleted and have
14641 14642 * just sent out a packet using this
14642 14643 * unreachable ncec.
14643 14644 */
14644 14645 mutex_exit(&ncec->ncec_lock);
14645 14646 break;
14646 14647 default:
14647 14648 ASSERT(0);
14648 14649 mutex_exit(&ncec->ncec_lock);
14649 14650 }
14650 14651 }
14651 14652 return (0);
14652 14653
14653 14654 case ND_INCOMPLETE:
14654 14655 /*
14655 14656 * the state could have changed since we didn't hold the lock.
14656 14657 * Re-verify state under lock.
14657 14658 */
14658 14659 is_probe = ipmp_packet_is_probe(mp, nce->nce_ill);
14659 14660 mutex_enter(&ncec->ncec_lock);
14660 14661 if (NCE_ISREACHABLE(ncec)) {
14661 14662 mutex_exit(&ncec->ncec_lock);
14662 14663 goto sendit;
14663 14664 }
14664 14665 /* queue the packet */
14665 14666 nce_queue_mp(ncec, mp, is_probe);
14666 14667 mutex_exit(&ncec->ncec_lock);
14667 14668 DTRACE_PROBE2(ip__xmit__incomplete,
14668 14669 (ncec_t *), ncec, (mblk_t *), mp);
14669 14670 return (0);
14670 14671
14671 14672 case ND_INITIAL:
14672 14673 /*
14673 14674 * State could have changed since we didn't hold the lock, so
14674 14675 * re-verify state.
14675 14676 */
14676 14677 is_probe = ipmp_packet_is_probe(mp, nce->nce_ill);
14677 14678 mutex_enter(&ncec->ncec_lock);
14678 14679 if (NCE_ISREACHABLE(ncec)) {
14679 14680 mutex_exit(&ncec->ncec_lock);
14680 14681 goto sendit;
14681 14682 }
14682 14683 nce_queue_mp(ncec, mp, is_probe);
14683 14684 if (ncec->ncec_state == ND_INITIAL) {
14684 14685 ncec->ncec_state = ND_INCOMPLETE;
14685 14686 mutex_exit(&ncec->ncec_lock);
14686 14687 /*
14687 14688 * figure out the source we want to use
14688 14689 * and resolve it.
14689 14690 */
14690 14691 ip_ndp_resolve(ncec);
14691 14692 } else {
14692 14693 mutex_exit(&ncec->ncec_lock);
14693 14694 }
14694 14695 return (0);
14695 14696
14696 14697 case ND_UNREACHABLE:
14697 14698 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14698 14699 ip_drop_output("ipIfStatsOutDiscards - ND_UNREACHABLE",
14699 14700 mp, ill);
14700 14701 freemsg(mp);
14701 14702 return (0);
14702 14703
14703 14704 default:
14704 14705 ASSERT(0);
14705 14706 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14706 14707 ip_drop_output("ipIfStatsOutDiscards - ND_other",
14707 14708 mp, ill);
14708 14709 freemsg(mp);
14709 14710 return (ENETUNREACH);
14710 14711 }
14711 14712 }
14712 14713
14713 14714 /*
14714 14715 * Return B_TRUE if the buffers differ in length or content.
14715 14716 * This is used for comparing extension header buffers.
14716 14717 * Note that an extension header would be declared different
14717 14718 * even if all that changed was the next header value in that header i.e.
14718 14719 * what really changed is the next extension header.
14719 14720 */
14720 14721 boolean_t
14721 14722 ip_cmpbuf(const void *abuf, uint_t alen, boolean_t b_valid, const void *bbuf,
14722 14723 uint_t blen)
14723 14724 {
14724 14725 if (!b_valid)
14725 14726 blen = 0;
14726 14727
14727 14728 if (alen != blen)
14728 14729 return (B_TRUE);
14729 14730 if (alen == 0)
14730 14731 return (B_FALSE); /* Both zero length */
14731 14732 return (bcmp(abuf, bbuf, alen));
14732 14733 }
14733 14734
14734 14735 /*
14735 14736 * Preallocate memory for ip_savebuf(). Returns B_TRUE if ok.
14736 14737 * Return B_FALSE if memory allocation fails - don't change any state!
14737 14738 */
14738 14739 boolean_t
14739 14740 ip_allocbuf(void **dstp, uint_t *dstlenp, boolean_t src_valid,
14740 14741 const void *src, uint_t srclen)
14741 14742 {
14742 14743 void *dst;
14743 14744
14744 14745 if (!src_valid)
14745 14746 srclen = 0;
14746 14747
14747 14748 ASSERT(*dstlenp == 0);
14748 14749 if (src != NULL && srclen != 0) {
14749 14750 dst = mi_alloc(srclen, BPRI_MED);
14750 14751 if (dst == NULL)
14751 14752 return (B_FALSE);
14752 14753 } else {
14753 14754 dst = NULL;
14754 14755 }
14755 14756 if (*dstp != NULL)
14756 14757 mi_free(*dstp);
14757 14758 *dstp = dst;
14758 14759 *dstlenp = dst == NULL ? 0 : srclen;
14759 14760 return (B_TRUE);
14760 14761 }
14761 14762
14762 14763 /*
14763 14764 * Replace what is in *dst, *dstlen with the source.
14764 14765 * Assumes ip_allocbuf has already been called.
14765 14766 */
14766 14767 void
14767 14768 ip_savebuf(void **dstp, uint_t *dstlenp, boolean_t src_valid,
14768 14769 const void *src, uint_t srclen)
14769 14770 {
14770 14771 if (!src_valid)
14771 14772 srclen = 0;
14772 14773
14773 14774 ASSERT(*dstlenp == srclen);
14774 14775 if (src != NULL && srclen != 0)
14775 14776 bcopy(src, *dstp, srclen);
14776 14777 }
14777 14778
14778 14779 /*
14779 14780 * Free the storage pointed to by the members of an ip_pkt_t.
14780 14781 */
14781 14782 void
14782 14783 ip_pkt_free(ip_pkt_t *ipp)
14783 14784 {
14784 14785 uint_t fields = ipp->ipp_fields;
14785 14786
14786 14787 if (fields & IPPF_HOPOPTS) {
14787 14788 kmem_free(ipp->ipp_hopopts, ipp->ipp_hopoptslen);
14788 14789 ipp->ipp_hopopts = NULL;
14789 14790 ipp->ipp_hopoptslen = 0;
14790 14791 }
14791 14792 if (fields & IPPF_RTHDRDSTOPTS) {
14792 14793 kmem_free(ipp->ipp_rthdrdstopts, ipp->ipp_rthdrdstoptslen);
14793 14794 ipp->ipp_rthdrdstopts = NULL;
14794 14795 ipp->ipp_rthdrdstoptslen = 0;
14795 14796 }
14796 14797 if (fields & IPPF_DSTOPTS) {
14797 14798 kmem_free(ipp->ipp_dstopts, ipp->ipp_dstoptslen);
14798 14799 ipp->ipp_dstopts = NULL;
14799 14800 ipp->ipp_dstoptslen = 0;
14800 14801 }
14801 14802 if (fields & IPPF_RTHDR) {
14802 14803 kmem_free(ipp->ipp_rthdr, ipp->ipp_rthdrlen);
14803 14804 ipp->ipp_rthdr = NULL;
14804 14805 ipp->ipp_rthdrlen = 0;
14805 14806 }
14806 14807 if (fields & IPPF_IPV4_OPTIONS) {
14807 14808 kmem_free(ipp->ipp_ipv4_options, ipp->ipp_ipv4_options_len);
14808 14809 ipp->ipp_ipv4_options = NULL;
14809 14810 ipp->ipp_ipv4_options_len = 0;
14810 14811 }
14811 14812 if (fields & IPPF_LABEL_V4) {
14812 14813 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
14813 14814 ipp->ipp_label_v4 = NULL;
14814 14815 ipp->ipp_label_len_v4 = 0;
14815 14816 }
14816 14817 if (fields & IPPF_LABEL_V6) {
14817 14818 kmem_free(ipp->ipp_label_v6, ipp->ipp_label_len_v6);
14818 14819 ipp->ipp_label_v6 = NULL;
14819 14820 ipp->ipp_label_len_v6 = 0;
14820 14821 }
14821 14822 ipp->ipp_fields &= ~(IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14822 14823 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6);
14823 14824 }
14824 14825
14825 14826 /*
14826 14827 * Copy from src to dst and allocate as needed.
14827 14828 * Returns zero or ENOMEM.
14828 14829 *
14829 14830 * The caller must initialize dst to zero.
14830 14831 */
14831 14832 int
14832 14833 ip_pkt_copy(ip_pkt_t *src, ip_pkt_t *dst, int kmflag)
14833 14834 {
14834 14835 uint_t fields = src->ipp_fields;
14835 14836
14836 14837 /* Start with fields that don't require memory allocation */
14837 14838 dst->ipp_fields = fields &
14838 14839 ~(IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14839 14840 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6);
14840 14841
14841 14842 dst->ipp_addr = src->ipp_addr;
14842 14843 dst->ipp_unicast_hops = src->ipp_unicast_hops;
14843 14844 dst->ipp_hoplimit = src->ipp_hoplimit;
14844 14845 dst->ipp_tclass = src->ipp_tclass;
14845 14846 dst->ipp_type_of_service = src->ipp_type_of_service;
14846 14847
14847 14848 if (!(fields & (IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14848 14849 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6)))
14849 14850 return (0);
14850 14851
14851 14852 if (fields & IPPF_HOPOPTS) {
14852 14853 dst->ipp_hopopts = kmem_alloc(src->ipp_hopoptslen, kmflag);
14853 14854 if (dst->ipp_hopopts == NULL) {
14854 14855 ip_pkt_free(dst);
14855 14856 return (ENOMEM);
14856 14857 }
14857 14858 dst->ipp_fields |= IPPF_HOPOPTS;
14858 14859 bcopy(src->ipp_hopopts, dst->ipp_hopopts,
14859 14860 src->ipp_hopoptslen);
14860 14861 dst->ipp_hopoptslen = src->ipp_hopoptslen;
14861 14862 }
14862 14863 if (fields & IPPF_RTHDRDSTOPTS) {
14863 14864 dst->ipp_rthdrdstopts = kmem_alloc(src->ipp_rthdrdstoptslen,
14864 14865 kmflag);
14865 14866 if (dst->ipp_rthdrdstopts == NULL) {
14866 14867 ip_pkt_free(dst);
14867 14868 return (ENOMEM);
14868 14869 }
14869 14870 dst->ipp_fields |= IPPF_RTHDRDSTOPTS;
14870 14871 bcopy(src->ipp_rthdrdstopts, dst->ipp_rthdrdstopts,
14871 14872 src->ipp_rthdrdstoptslen);
14872 14873 dst->ipp_rthdrdstoptslen = src->ipp_rthdrdstoptslen;
14873 14874 }
14874 14875 if (fields & IPPF_DSTOPTS) {
14875 14876 dst->ipp_dstopts = kmem_alloc(src->ipp_dstoptslen, kmflag);
14876 14877 if (dst->ipp_dstopts == NULL) {
14877 14878 ip_pkt_free(dst);
14878 14879 return (ENOMEM);
14879 14880 }
14880 14881 dst->ipp_fields |= IPPF_DSTOPTS;
14881 14882 bcopy(src->ipp_dstopts, dst->ipp_dstopts,
14882 14883 src->ipp_dstoptslen);
14883 14884 dst->ipp_dstoptslen = src->ipp_dstoptslen;
14884 14885 }
14885 14886 if (fields & IPPF_RTHDR) {
14886 14887 dst->ipp_rthdr = kmem_alloc(src->ipp_rthdrlen, kmflag);
14887 14888 if (dst->ipp_rthdr == NULL) {
14888 14889 ip_pkt_free(dst);
14889 14890 return (ENOMEM);
14890 14891 }
14891 14892 dst->ipp_fields |= IPPF_RTHDR;
14892 14893 bcopy(src->ipp_rthdr, dst->ipp_rthdr,
14893 14894 src->ipp_rthdrlen);
14894 14895 dst->ipp_rthdrlen = src->ipp_rthdrlen;
14895 14896 }
14896 14897 if (fields & IPPF_IPV4_OPTIONS) {
14897 14898 dst->ipp_ipv4_options = kmem_alloc(src->ipp_ipv4_options_len,
14898 14899 kmflag);
14899 14900 if (dst->ipp_ipv4_options == NULL) {
14900 14901 ip_pkt_free(dst);
14901 14902 return (ENOMEM);
14902 14903 }
14903 14904 dst->ipp_fields |= IPPF_IPV4_OPTIONS;
14904 14905 bcopy(src->ipp_ipv4_options, dst->ipp_ipv4_options,
14905 14906 src->ipp_ipv4_options_len);
14906 14907 dst->ipp_ipv4_options_len = src->ipp_ipv4_options_len;
14907 14908 }
14908 14909 if (fields & IPPF_LABEL_V4) {
14909 14910 dst->ipp_label_v4 = kmem_alloc(src->ipp_label_len_v4, kmflag);
14910 14911 if (dst->ipp_label_v4 == NULL) {
14911 14912 ip_pkt_free(dst);
14912 14913 return (ENOMEM);
14913 14914 }
14914 14915 dst->ipp_fields |= IPPF_LABEL_V4;
14915 14916 bcopy(src->ipp_label_v4, dst->ipp_label_v4,
14916 14917 src->ipp_label_len_v4);
14917 14918 dst->ipp_label_len_v4 = src->ipp_label_len_v4;
14918 14919 }
14919 14920 if (fields & IPPF_LABEL_V6) {
14920 14921 dst->ipp_label_v6 = kmem_alloc(src->ipp_label_len_v6, kmflag);
14921 14922 if (dst->ipp_label_v6 == NULL) {
14922 14923 ip_pkt_free(dst);
14923 14924 return (ENOMEM);
14924 14925 }
14925 14926 dst->ipp_fields |= IPPF_LABEL_V6;
14926 14927 bcopy(src->ipp_label_v6, dst->ipp_label_v6,
14927 14928 src->ipp_label_len_v6);
14928 14929 dst->ipp_label_len_v6 = src->ipp_label_len_v6;
14929 14930 }
14930 14931 if (fields & IPPF_FRAGHDR) {
14931 14932 dst->ipp_fraghdr = kmem_alloc(src->ipp_fraghdrlen, kmflag);
14932 14933 if (dst->ipp_fraghdr == NULL) {
14933 14934 ip_pkt_free(dst);
14934 14935 return (ENOMEM);
14935 14936 }
14936 14937 dst->ipp_fields |= IPPF_FRAGHDR;
14937 14938 bcopy(src->ipp_fraghdr, dst->ipp_fraghdr,
14938 14939 src->ipp_fraghdrlen);
14939 14940 dst->ipp_fraghdrlen = src->ipp_fraghdrlen;
14940 14941 }
14941 14942 return (0);
14942 14943 }
14943 14944
14944 14945 /*
14945 14946 * Returns INADDR_ANY if no source route
14946 14947 */
14947 14948 ipaddr_t
14948 14949 ip_pkt_source_route_v4(const ip_pkt_t *ipp)
14949 14950 {
14950 14951 ipaddr_t nexthop = INADDR_ANY;
14951 14952 ipoptp_t opts;
14952 14953 uchar_t *opt;
14953 14954 uint8_t optval;
14954 14955 uint8_t optlen;
14955 14956 uint32_t totallen;
14956 14957
14957 14958 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
14958 14959 return (INADDR_ANY);
14959 14960
14960 14961 totallen = ipp->ipp_ipv4_options_len;
14961 14962 if (totallen & 0x3)
14962 14963 return (INADDR_ANY);
14963 14964
14964 14965 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
14965 14966 optval != IPOPT_EOL;
14966 14967 optval = ipoptp_next(&opts)) {
14967 14968 opt = opts.ipoptp_cur;
14968 14969 switch (optval) {
14969 14970 uint8_t off;
14970 14971 case IPOPT_SSRR:
14971 14972 case IPOPT_LSRR:
14972 14973 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
14973 14974 break;
14974 14975 }
14975 14976 optlen = opts.ipoptp_len;
14976 14977 off = opt[IPOPT_OFFSET];
14977 14978 off--;
14978 14979 if (optlen < IP_ADDR_LEN ||
14979 14980 off > optlen - IP_ADDR_LEN) {
14980 14981 /* End of source route */
14981 14982 break;
14982 14983 }
14983 14984 bcopy((char *)opt + off, &nexthop, IP_ADDR_LEN);
14984 14985 if (nexthop == htonl(INADDR_LOOPBACK)) {
14985 14986 /* Ignore */
14986 14987 nexthop = INADDR_ANY;
14987 14988 break;
14988 14989 }
14989 14990 break;
14990 14991 }
14991 14992 }
14992 14993 return (nexthop);
14993 14994 }
14994 14995
14995 14996 /*
14996 14997 * Reverse a source route.
14997 14998 */
14998 14999 void
14999 15000 ip_pkt_source_route_reverse_v4(ip_pkt_t *ipp)
15000 15001 {
15001 15002 ipaddr_t tmp;
15002 15003 ipoptp_t opts;
15003 15004 uchar_t *opt;
15004 15005 uint8_t optval;
15005 15006 uint32_t totallen;
15006 15007
15007 15008 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
15008 15009 return;
15009 15010
15010 15011 totallen = ipp->ipp_ipv4_options_len;
15011 15012 if (totallen & 0x3)
15012 15013 return;
15013 15014
15014 15015 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
15015 15016 optval != IPOPT_EOL;
15016 15017 optval = ipoptp_next(&opts)) {
15017 15018 uint8_t off1, off2;
15018 15019
15019 15020 opt = opts.ipoptp_cur;
15020 15021 switch (optval) {
15021 15022 case IPOPT_SSRR:
15022 15023 case IPOPT_LSRR:
15023 15024 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
15024 15025 break;
15025 15026 }
15026 15027 off1 = IPOPT_MINOFF_SR - 1;
15027 15028 off2 = opt[IPOPT_OFFSET] - IP_ADDR_LEN - 1;
15028 15029 while (off2 > off1) {
15029 15030 bcopy(opt + off2, &tmp, IP_ADDR_LEN);
15030 15031 bcopy(opt + off1, opt + off2, IP_ADDR_LEN);
15031 15032 bcopy(&tmp, opt + off2, IP_ADDR_LEN);
15032 15033 off2 -= IP_ADDR_LEN;
15033 15034 off1 += IP_ADDR_LEN;
15034 15035 }
15035 15036 opt[IPOPT_OFFSET] = IPOPT_MINOFF_SR;
15036 15037 break;
15037 15038 }
15038 15039 }
15039 15040 }
15040 15041
15041 15042 /*
15042 15043 * Returns NULL if no routing header
15043 15044 */
15044 15045 in6_addr_t *
15045 15046 ip_pkt_source_route_v6(const ip_pkt_t *ipp)
15046 15047 {
15047 15048 in6_addr_t *nexthop = NULL;
15048 15049 ip6_rthdr0_t *rthdr;
15049 15050
15050 15051 if (!(ipp->ipp_fields & IPPF_RTHDR))
15051 15052 return (NULL);
15052 15053
15053 15054 rthdr = (ip6_rthdr0_t *)ipp->ipp_rthdr;
15054 15055 if (rthdr->ip6r0_segleft == 0)
15055 15056 return (NULL);
15056 15057
15057 15058 nexthop = (in6_addr_t *)((char *)rthdr + sizeof (*rthdr));
15058 15059 return (nexthop);
15059 15060 }
15060 15061
15061 15062 zoneid_t
15062 15063 ip_get_zoneid_v4(ipaddr_t addr, mblk_t *mp, ip_recv_attr_t *ira,
15063 15064 zoneid_t lookup_zoneid)
15064 15065 {
15065 15066 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
15066 15067 ire_t *ire;
15067 15068 int ire_flags = MATCH_IRE_TYPE;
15068 15069 zoneid_t zoneid = ALL_ZONES;
15069 15070
15070 15071 if (is_system_labeled() && !tsol_can_accept_raw(mp, ira, B_FALSE))
15071 15072 return (ALL_ZONES);
15072 15073
15073 15074 if (lookup_zoneid != ALL_ZONES)
15074 15075 ire_flags |= MATCH_IRE_ZONEONLY;
15075 15076 ire = ire_ftable_lookup_v4(addr, NULL, NULL, IRE_LOCAL | IRE_LOOPBACK,
15076 15077 NULL, lookup_zoneid, NULL, ire_flags, 0, ipst, NULL);
15077 15078 if (ire != NULL) {
15078 15079 zoneid = IP_REAL_ZONEID(ire->ire_zoneid, ipst);
15079 15080 ire_refrele(ire);
15080 15081 }
15081 15082 return (zoneid);
15082 15083 }
15083 15084
15084 15085 zoneid_t
15085 15086 ip_get_zoneid_v6(in6_addr_t *addr, mblk_t *mp, const ill_t *ill,
15086 15087 ip_recv_attr_t *ira, zoneid_t lookup_zoneid)
15087 15088 {
15088 15089 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
15089 15090 ire_t *ire;
15090 15091 int ire_flags = MATCH_IRE_TYPE;
15091 15092 zoneid_t zoneid = ALL_ZONES;
15092 15093
15093 15094 if (is_system_labeled() && !tsol_can_accept_raw(mp, ira, B_FALSE))
15094 15095 return (ALL_ZONES);
15095 15096
15096 15097 if (IN6_IS_ADDR_LINKLOCAL(addr))
15097 15098 ire_flags |= MATCH_IRE_ILL;
15098 15099
15099 15100 if (lookup_zoneid != ALL_ZONES)
15100 15101 ire_flags |= MATCH_IRE_ZONEONLY;
15101 15102 ire = ire_ftable_lookup_v6(addr, NULL, NULL, IRE_LOCAL | IRE_LOOPBACK,
15102 15103 ill, lookup_zoneid, NULL, ire_flags, 0, ipst, NULL);
15103 15104 if (ire != NULL) {
15104 15105 zoneid = IP_REAL_ZONEID(ire->ire_zoneid, ipst);
15105 15106 ire_refrele(ire);
15106 15107 }
15107 15108 return (zoneid);
15108 15109 }
15109 15110
15110 15111 /*
15111 15112 * IP obserability hook support functions.
15112 15113 */
15113 15114 static void
15114 15115 ipobs_init(ip_stack_t *ipst)
15115 15116 {
15116 15117 netid_t id;
15117 15118
15118 15119 id = net_getnetidbynetstackid(ipst->ips_netstack->netstack_stackid);
15119 15120
15120 15121 ipst->ips_ip4_observe_pr = net_protocol_lookup(id, NHF_INET);
15121 15122 VERIFY(ipst->ips_ip4_observe_pr != NULL);
15122 15123
15123 15124 ipst->ips_ip6_observe_pr = net_protocol_lookup(id, NHF_INET6);
15124 15125 VERIFY(ipst->ips_ip6_observe_pr != NULL);
15125 15126 }
15126 15127
15127 15128 static void
15128 15129 ipobs_fini(ip_stack_t *ipst)
15129 15130 {
15130 15131
15131 15132 VERIFY(net_protocol_release(ipst->ips_ip4_observe_pr) == 0);
15132 15133 VERIFY(net_protocol_release(ipst->ips_ip6_observe_pr) == 0);
15133 15134 }
15134 15135
15135 15136 /*
15136 15137 * hook_pkt_observe_t is composed in network byte order so that the
15137 15138 * entire mblk_t chain handed into hook_run can be used as-is.
15138 15139 * The caveat is that use of the fields, such as the zone fields,
15139 15140 * requires conversion into host byte order first.
15140 15141 */
15141 15142 void
15142 15143 ipobs_hook(mblk_t *mp, int htype, zoneid_t zsrc, zoneid_t zdst,
15143 15144 const ill_t *ill, ip_stack_t *ipst)
15144 15145 {
15145 15146 hook_pkt_observe_t *hdr;
15146 15147 uint64_t grifindex;
15147 15148 mblk_t *imp;
15148 15149
15149 15150 imp = allocb(sizeof (*hdr), BPRI_HI);
15150 15151 if (imp == NULL)
15151 15152 return;
15152 15153
15153 15154 hdr = (hook_pkt_observe_t *)imp->b_rptr;
15154 15155 /*
15155 15156 * b_wptr is set to make the apparent size of the data in the mblk_t
15156 15157 * to exclude the pointers at the end of hook_pkt_observer_t.
15157 15158 */
15158 15159 imp->b_wptr = imp->b_rptr + sizeof (dl_ipnetinfo_t);
15159 15160 imp->b_cont = mp;
15160 15161
15161 15162 ASSERT(DB_TYPE(mp) == M_DATA);
15162 15163
15163 15164 if (IS_UNDER_IPMP(ill))
15164 15165 grifindex = ipmp_ill_get_ipmp_ifindex(ill);
15165 15166 else
15166 15167 grifindex = 0;
15167 15168
15168 15169 hdr->hpo_version = 1;
15169 15170 hdr->hpo_htype = htons(htype);
15170 15171 hdr->hpo_pktlen = htonl((ulong_t)msgdsize(mp));
15171 15172 hdr->hpo_ifindex = htonl(ill->ill_phyint->phyint_ifindex);
15172 15173 hdr->hpo_grifindex = htonl(grifindex);
15173 15174 hdr->hpo_zsrc = htonl(zsrc);
15174 15175 hdr->hpo_zdst = htonl(zdst);
15175 15176 hdr->hpo_pkt = imp;
15176 15177 hdr->hpo_ctx = ipst->ips_netstack;
15177 15178
15178 15179 if (ill->ill_isv6) {
15179 15180 hdr->hpo_family = AF_INET6;
15180 15181 (void) hook_run(ipst->ips_ipv6_net_data->netd_hooks,
15181 15182 ipst->ips_ipv6observing, (hook_data_t)hdr);
15182 15183 } else {
15183 15184 hdr->hpo_family = AF_INET;
15184 15185 (void) hook_run(ipst->ips_ipv4_net_data->netd_hooks,
15185 15186 ipst->ips_ipv4observing, (hook_data_t)hdr);
15186 15187 }
15187 15188
15188 15189 imp->b_cont = NULL;
15189 15190 freemsg(imp);
15190 15191 }
15191 15192
15192 15193 /*
15193 15194 * Utility routine that checks if `v4srcp' is a valid address on underlying
15194 15195 * interface `ill'. If `ipifp' is non-NULL, it's set to a held ipif
15195 15196 * associated with `v4srcp' on success. NOTE: if this is not called from
15196 15197 * inside the IPSQ (ill_g_lock is not held), `ill' may be removed from the
15197 15198 * group during or after this lookup.
15198 15199 */
15199 15200 boolean_t
15200 15201 ipif_lookup_testaddr_v4(ill_t *ill, const in_addr_t *v4srcp, ipif_t **ipifp)
15201 15202 {
15202 15203 ipif_t *ipif;
15203 15204
15204 15205 ipif = ipif_lookup_addr_exact(*v4srcp, ill, ill->ill_ipst);
15205 15206 if (ipif != NULL) {
15206 15207 if (ipifp != NULL)
15207 15208 *ipifp = ipif;
15208 15209 else
15209 15210 ipif_refrele(ipif);
15210 15211 return (B_TRUE);
15211 15212 }
15212 15213
15213 15214 ip1dbg(("ipif_lookup_testaddr_v4: cannot find ipif for src %x\n",
15214 15215 *v4srcp));
15215 15216 return (B_FALSE);
15216 15217 }
15217 15218
15218 15219 /*
15219 15220 * Transport protocol call back function for CPU state change.
15220 15221 */
15221 15222 /* ARGSUSED */
15222 15223 static int
15223 15224 ip_tp_cpu_update(cpu_setup_t what, int id, void *arg)
15224 15225 {
15225 15226 processorid_t cpu_seqid;
15226 15227 netstack_handle_t nh;
15227 15228 netstack_t *ns;
15228 15229
15229 15230 ASSERT(MUTEX_HELD(&cpu_lock));
15230 15231
15231 15232 switch (what) {
15232 15233 case CPU_CONFIG:
15233 15234 case CPU_ON:
15234 15235 case CPU_INIT:
15235 15236 case CPU_CPUPART_IN:
15236 15237 cpu_seqid = cpu[id]->cpu_seqid;
15237 15238 netstack_next_init(&nh);
15238 15239 while ((ns = netstack_next(&nh)) != NULL) {
15239 15240 tcp_stack_cpu_add(ns->netstack_tcp, cpu_seqid);
15240 15241 sctp_stack_cpu_add(ns->netstack_sctp, cpu_seqid);
15241 15242 udp_stack_cpu_add(ns->netstack_udp, cpu_seqid);
15242 15243 netstack_rele(ns);
15243 15244 }
15244 15245 netstack_next_fini(&nh);
15245 15246 break;
15246 15247 case CPU_UNCONFIG:
15247 15248 case CPU_OFF:
15248 15249 case CPU_CPUPART_OUT:
15249 15250 /*
15250 15251 * Nothing to do. We don't remove the per CPU stats from
15251 15252 * the IP stack even when the CPU goes offline.
15252 15253 */
15253 15254 break;
15254 15255 default:
15255 15256 break;
15256 15257 }
15257 15258 return (0);
15258 15259 }
|
↓ open down ↓ |
10596 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX