1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21
22 /*
23 * Copyright (c) 1991, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 1990 Mentat Inc.
25 * Copyright (c) 2017 OmniTI Computer Consulting, Inc. All rights reserved.
26 * Copyright (c) 2016 by Delphix. All rights reserved.
27 * Copyright (c) 2019 Joyent, Inc. All rights reserved.
28 */
29
30 #include <sys/types.h>
31 #include <sys/stream.h>
32 #include <sys/dlpi.h>
33 #include <sys/stropts.h>
34 #include <sys/sysmacros.h>
35 #include <sys/strsubr.h>
36 #include <sys/strlog.h>
37 #include <sys/strsun.h>
38 #include <sys/zone.h>
39 #define _SUN_TPI_VERSION 2
40 #include <sys/tihdr.h>
41 #include <sys/xti_inet.h>
42 #include <sys/ddi.h>
43 #include <sys/suntpi.h>
44 #include <sys/cmn_err.h>
45 #include <sys/debug.h>
46 #include <sys/kobj.h>
47 #include <sys/modctl.h>
48 #include <sys/atomic.h>
49 #include <sys/policy.h>
50 #include <sys/priv.h>
51 #include <sys/taskq.h>
52
53 #include <sys/systm.h>
54 #include <sys/param.h>
55 #include <sys/kmem.h>
56 #include <sys/sdt.h>
57 #include <sys/socket.h>
58 #include <sys/vtrace.h>
59 #include <sys/isa_defs.h>
60 #include <sys/mac.h>
61 #include <net/if.h>
62 #include <net/if_arp.h>
63 #include <net/route.h>
64 #include <sys/sockio.h>
65 #include <netinet/in.h>
66 #include <net/if_dl.h>
67
68 #include <inet/common.h>
69 #include <inet/mi.h>
70 #include <inet/mib2.h>
71 #include <inet/nd.h>
72 #include <inet/arp.h>
73 #include <inet/snmpcom.h>
74 #include <inet/optcom.h>
75 #include <inet/kstatcom.h>
76
77 #include <netinet/igmp_var.h>
78 #include <netinet/ip6.h>
79 #include <netinet/icmp6.h>
80 #include <netinet/sctp.h>
81
82 #include <inet/ip.h>
83 #include <inet/ip_impl.h>
84 #include <inet/ip6.h>
85 #include <inet/ip6_asp.h>
86 #include <inet/tcp.h>
87 #include <inet/tcp_impl.h>
88 #include <inet/ip_multi.h>
89 #include <inet/ip_if.h>
90 #include <inet/ip_ire.h>
91 #include <inet/ip_ftable.h>
92 #include <inet/ip_rts.h>
93 #include <inet/ip_ndp.h>
94 #include <inet/ip_listutils.h>
95 #include <netinet/igmp.h>
96 #include <netinet/ip_mroute.h>
97 #include <inet/ipp_common.h>
98 #include <inet/cc.h>
99
100 #include <net/pfkeyv2.h>
101 #include <inet/sadb.h>
102 #include <inet/ipsec_impl.h>
103 #include <inet/iptun/iptun_impl.h>
104 #include <inet/ipdrop.h>
105 #include <inet/ip_netinfo.h>
106 #include <inet/ilb_ip.h>
107
108 #include <sys/ethernet.h>
109 #include <net/if_types.h>
110 #include <sys/cpuvar.h>
111
112 #include <ipp/ipp.h>
113 #include <ipp/ipp_impl.h>
114 #include <ipp/ipgpc/ipgpc.h>
115
116 #include <sys/pattr.h>
117 #include <inet/ipclassifier.h>
118 #include <inet/sctp_ip.h>
119 #include <inet/sctp/sctp_impl.h>
120 #include <inet/udp_impl.h>
121 #include <inet/rawip_impl.h>
122 #include <inet/rts_impl.h>
123
124 #include <sys/tsol/label.h>
125 #include <sys/tsol/tnet.h>
126
127 #include <sys/squeue_impl.h>
128 #include <inet/ip_arp.h>
129
130 #include <sys/clock_impl.h> /* For LBOLT_FASTPATH{,64} */
131
132 /*
133 * Values for squeue switch:
134 * IP_SQUEUE_ENTER_NODRAIN: SQ_NODRAIN
135 * IP_SQUEUE_ENTER: SQ_PROCESS
136 * IP_SQUEUE_FILL: SQ_FILL
137 */
138 int ip_squeue_enter = IP_SQUEUE_ENTER; /* Setable in /etc/system */
139
140 int ip_squeue_flag;
141
142 /*
143 * Setable in /etc/system
144 */
145 int ip_poll_normal_ms = 100;
146 int ip_poll_normal_ticks = 0;
147 int ip_modclose_ackwait_ms = 3000;
148
149 /*
150 * It would be nice to have these present only in DEBUG systems, but the
151 * current design of the global symbol checking logic requires them to be
152 * unconditionally present.
153 */
154 uint_t ip_thread_data; /* TSD key for debug support */
155 krwlock_t ip_thread_rwlock;
156 list_t ip_thread_list;
157
158 /*
159 * Structure to represent a linked list of msgblks. Used by ip_snmp_ functions.
160 */
161
162 struct listptr_s {
163 mblk_t *lp_head; /* pointer to the head of the list */
164 mblk_t *lp_tail; /* pointer to the tail of the list */
165 };
166
167 typedef struct listptr_s listptr_t;
168
169 /*
170 * This is used by ip_snmp_get_mib2_ip_route_media and
171 * ip_snmp_get_mib2_ip6_route_media to carry the lists of return data.
172 */
173 typedef struct iproutedata_s {
174 uint_t ird_idx;
175 uint_t ird_flags; /* see below */
176 listptr_t ird_route; /* ipRouteEntryTable */
177 listptr_t ird_netmedia; /* ipNetToMediaEntryTable */
178 listptr_t ird_attrs; /* ipRouteAttributeTable */
179 } iproutedata_t;
180
181 /* Include ire_testhidden and IRE_IF_CLONE routes */
182 #define IRD_REPORT_ALL 0x01
183
184 /*
185 * Cluster specific hooks. These should be NULL when booted as a non-cluster
186 */
187
188 /*
189 * Hook functions to enable cluster networking
190 * On non-clustered systems these vectors must always be NULL.
191 *
192 * Hook function to Check ip specified ip address is a shared ip address
193 * in the cluster
194 *
195 */
196 int (*cl_inet_isclusterwide)(netstackid_t stack_id, uint8_t protocol,
197 sa_family_t addr_family, uint8_t *laddrp, void *args) = NULL;
198
199 /*
200 * Hook function to generate cluster wide ip fragment identifier
201 */
202 uint32_t (*cl_inet_ipident)(netstackid_t stack_id, uint8_t protocol,
203 sa_family_t addr_family, uint8_t *laddrp, uint8_t *faddrp,
204 void *args) = NULL;
205
206 /*
207 * Hook function to generate cluster wide SPI.
208 */
209 void (*cl_inet_getspi)(netstackid_t, uint8_t, uint8_t *, size_t,
210 void *) = NULL;
211
212 /*
213 * Hook function to verify if the SPI is already utlized.
214 */
215
216 int (*cl_inet_checkspi)(netstackid_t, uint8_t, uint32_t, void *) = NULL;
217
218 /*
219 * Hook function to delete the SPI from the cluster wide repository.
220 */
221
222 void (*cl_inet_deletespi)(netstackid_t, uint8_t, uint32_t, void *) = NULL;
223
224 /*
225 * Hook function to inform the cluster when packet received on an IDLE SA
226 */
227
228 void (*cl_inet_idlesa)(netstackid_t, uint8_t, uint32_t, sa_family_t,
229 in6_addr_t, in6_addr_t, void *) = NULL;
230
231 /*
232 * Synchronization notes:
233 *
234 * IP is a fully D_MP STREAMS module/driver. Thus it does not depend on any
235 * MT level protection given by STREAMS. IP uses a combination of its own
236 * internal serialization mechanism and standard Solaris locking techniques.
237 * The internal serialization is per phyint. This is used to serialize
238 * plumbing operations, IPMP operations, most set ioctls, etc.
239 *
240 * Plumbing is a long sequence of operations involving message
241 * exchanges between IP, ARP and device drivers. Many set ioctls are typically
242 * involved in plumbing operations. A natural model is to serialize these
243 * ioctls one per ill. For example plumbing of hme0 and qfe0 can go on in
244 * parallel without any interference. But various set ioctls on hme0 are best
245 * serialized, along with IPMP operations and processing of DLPI control
246 * messages received from drivers on a per phyint basis. This serialization is
247 * provided by the ipsq_t and primitives operating on this. Details can
248 * be found in ip_if.c above the core primitives operating on ipsq_t.
249 *
250 * Lookups of an ipif or ill by a thread return a refheld ipif / ill.
251 * Simiarly lookup of an ire by a thread also returns a refheld ire.
252 * In addition ipif's and ill's referenced by the ire are also indirectly
253 * refheld. Thus no ipif or ill can vanish as long as an ipif is refheld
254 * directly or indirectly. For example an SIOCSLIFADDR ioctl that changes the
255 * address of an ipif has to go through the ipsq_t. This ensures that only
256 * one such exclusive operation proceeds at any time on the ipif. It then
257 * waits for all refcnts
258 * associated with this ipif to come down to zero. The address is changed
259 * only after the ipif has been quiesced. Then the ipif is brought up again.
260 * More details are described above the comment in ip_sioctl_flags.
261 *
262 * Packet processing is based mostly on IREs and are fully multi-threaded
263 * using standard Solaris MT techniques.
264 *
265 * There are explicit locks in IP to handle:
266 * - The ip_g_head list maintained by mi_open_link() and friends.
267 *
268 * - The reassembly data structures (one lock per hash bucket)
269 *
270 * - conn_lock is meant to protect conn_t fields. The fields actually
271 * protected by conn_lock are documented in the conn_t definition.
272 *
273 * - ire_lock to protect some of the fields of the ire, IRE tables
274 * (one lock per hash bucket). Refer to ip_ire.c for details.
275 *
276 * - ndp_g_lock and ncec_lock for protecting NCEs.
277 *
278 * - ill_lock protects fields of the ill and ipif. Details in ip.h
279 *
280 * - ill_g_lock: This is a global reader/writer lock. Protects the following
281 * * The AVL tree based global multi list of all ills.
282 * * The linked list of all ipifs of an ill
283 * * The <ipsq-xop> mapping
284 * * <ill-phyint> association
285 * Insertion/deletion of an ill in the system, insertion/deletion of an ipif
286 * into an ill, changing the <ipsq-xop> mapping of an ill, changing the
287 * <ill-phyint> assoc of an ill will all have to hold the ill_g_lock as
288 * writer for the actual duration of the insertion/deletion/change.
289 *
290 * - ill_lock: This is a per ill mutex.
291 * It protects some members of the ill_t struct; see ip.h for details.
292 * It also protects the <ill-phyint> assoc.
293 * It also protects the list of ipifs hanging off the ill.
294 *
295 * - ipsq_lock: This is a per ipsq_t mutex lock.
296 * This protects some members of the ipsq_t struct; see ip.h for details.
297 * It also protects the <ipsq-ipxop> mapping
298 *
299 * - ipx_lock: This is a per ipxop_t mutex lock.
300 * This protects some members of the ipxop_t struct; see ip.h for details.
301 *
302 * - phyint_lock: This is a per phyint mutex lock. Protects just the
303 * phyint_flags
304 *
305 * - ip_addr_avail_lock: This is used to ensure the uniqueness of IP addresses.
306 * This lock is held in ipif_up_done and the ipif is marked IPIF_UP and the
307 * uniqueness check also done atomically.
308 *
309 * - ill_g_usesrc_lock: This readers/writer lock protects the usesrc
310 * group list linked by ill_usesrc_grp_next. It also protects the
311 * ill_usesrc_ifindex field. It is taken as a writer when a member of the
312 * group is being added or deleted. This lock is taken as a reader when
313 * walking the list/group(eg: to get the number of members in a usesrc group).
314 * Note, it is only necessary to take this lock if the ill_usesrc_grp_next
315 * field is changing state i.e from NULL to non-NULL or vice-versa. For
316 * example, it is not necessary to take this lock in the initial portion
317 * of ip_sioctl_slifusesrc or at all in ip_sioctl_flags since these
318 * operations are executed exclusively and that ensures that the "usesrc
319 * group state" cannot change. The "usesrc group state" change can happen
320 * only in the latter part of ip_sioctl_slifusesrc and in ill_delete.
321 *
322 * Changing <ill-phyint>, <ipsq-xop> assocications:
323 *
324 * To change the <ill-phyint> association, the ill_g_lock must be held
325 * as writer, and the ill_locks of both the v4 and v6 instance of the ill
326 * must be held.
327 *
328 * To change the <ipsq-xop> association, the ill_g_lock must be held as
329 * writer, the ipsq_lock must be held, and one must be writer on the ipsq.
330 * This is only done when ills are added or removed from IPMP groups.
331 *
332 * To add or delete an ipif from the list of ipifs hanging off the ill,
333 * ill_g_lock (writer) and ill_lock must be held and the thread must be
334 * a writer on the associated ipsq.
335 *
336 * To add or delete an ill to the system, the ill_g_lock must be held as
337 * writer and the thread must be a writer on the associated ipsq.
338 *
339 * To add or delete an ilm to an ill, the ill_lock must be held and the thread
340 * must be a writer on the associated ipsq.
341 *
342 * Lock hierarchy
343 *
344 * Some lock hierarchy scenarios are listed below.
345 *
346 * ill_g_lock -> conn_lock -> ill_lock -> ipsq_lock -> ipx_lock
347 * ill_g_lock -> ill_lock(s) -> phyint_lock
348 * ill_g_lock -> ndp_g_lock -> ill_lock -> ncec_lock
349 * ill_g_lock -> ip_addr_avail_lock
350 * conn_lock -> irb_lock -> ill_lock -> ire_lock
351 * ill_g_lock -> ip_g_nd_lock
352 * ill_g_lock -> ips_ipmp_lock -> ill_lock -> nce_lock
353 * ill_g_lock -> ndp_g_lock -> ill_lock -> ncec_lock -> nce_lock
354 * arl_lock -> ill_lock
355 * ips_ire_dep_lock -> irb_lock
356 *
357 * When more than 1 ill lock is needed to be held, all ill lock addresses
358 * are sorted on address and locked starting from highest addressed lock
359 * downward.
360 *
361 * Multicast scenarios
362 * ips_ill_g_lock -> ill_mcast_lock
363 * conn_ilg_lock -> ips_ill_g_lock -> ill_lock
364 * ill_mcast_serializer -> ill_mcast_lock -> ips_ipmp_lock -> ill_lock
365 * ill_mcast_serializer -> ill_mcast_lock -> connf_lock -> conn_lock
366 * ill_mcast_serializer -> ill_mcast_lock -> conn_ilg_lock
367 * ill_mcast_serializer -> ill_mcast_lock -> ips_igmp_timer_lock
368 *
369 * IPsec scenarios
370 *
371 * ipsa_lock -> ill_g_lock -> ill_lock
372 * ill_g_usesrc_lock -> ill_g_lock -> ill_lock
373 *
374 * Trusted Solaris scenarios
375 *
376 * igsa_lock -> gcgrp_rwlock -> gcgrp_lock
377 * igsa_lock -> gcdb_lock
378 * gcgrp_rwlock -> ire_lock
379 * gcgrp_rwlock -> gcdb_lock
380 *
381 * squeue(sq_lock), flow related (ft_lock, fe_lock) locking
382 *
383 * cpu_lock --> ill_lock --> sqset_lock --> sq_lock
384 * sq_lock -> conn_lock -> QLOCK(q)
385 * ill_lock -> ft_lock -> fe_lock
386 *
387 * Routing/forwarding table locking notes:
388 *
389 * Lock acquisition order: Radix tree lock, irb_lock.
390 * Requirements:
391 * i. Walker must not hold any locks during the walker callback.
392 * ii Walker must not see a truncated tree during the walk because of any node
393 * deletion.
394 * iii Existing code assumes ire_bucket is valid if it is non-null and is used
395 * in many places in the code to walk the irb list. Thus even if all the
396 * ires in a bucket have been deleted, we still can't free the radix node
397 * until the ires have actually been inactive'd (freed).
398 *
399 * Tree traversal - Need to hold the global tree lock in read mode.
400 * Before dropping the global tree lock, need to either increment the ire_refcnt
401 * to ensure that the radix node can't be deleted.
402 *
403 * Tree add - Need to hold the global tree lock in write mode to add a
404 * radix node. To prevent the node from being deleted, increment the
405 * irb_refcnt, after the node is added to the tree. The ire itself is
406 * added later while holding the irb_lock, but not the tree lock.
407 *
408 * Tree delete - Need to hold the global tree lock and irb_lock in write mode.
409 * All associated ires must be inactive (i.e. freed), and irb_refcnt
410 * must be zero.
411 *
412 * Walker - Increment irb_refcnt before calling the walker callback. Hold the
413 * global tree lock (read mode) for traversal.
414 *
415 * IRE dependencies - In some cases we hold ips_ire_dep_lock across ire_refrele
416 * hence we will acquire irb_lock while holding ips_ire_dep_lock.
417 *
418 * IPsec notes :
419 *
420 * IP interacts with the IPsec code (AH/ESP) by storing IPsec attributes
421 * in the ip_xmit_attr_t ip_recv_attr_t. For outbound datagrams, the
422 * ip_xmit_attr_t has the
423 * information used by the IPsec code for applying the right level of
424 * protection. The information initialized by IP in the ip_xmit_attr_t
425 * is determined by the per-socket policy or global policy in the system.
426 * For inbound datagrams, the ip_recv_attr_t
427 * starts out with nothing in it. It gets filled
428 * with the right information if it goes through the AH/ESP code, which
429 * happens if the incoming packet is secure. The information initialized
430 * by AH/ESP, is later used by IP (during fanouts to ULP) to see whether
431 * the policy requirements needed by per-socket policy or global policy
432 * is met or not.
433 *
434 * For fully connected sockets i.e dst, src [addr, port] is known,
435 * conn_policy_cached is set indicating that policy has been cached.
436 * conn_in_enforce_policy may or may not be set depending on whether
437 * there is a global policy match or per-socket policy match.
438 * Policy inheriting happpens in ip_policy_set once the destination is known.
439 * Once the right policy is set on the conn_t, policy cannot change for
440 * this socket. This makes life simpler for TCP (UDP ?) where
441 * re-transmissions go out with the same policy. For symmetry, policy
442 * is cached for fully connected UDP sockets also. Thus if policy is cached,
443 * it also implies that policy is latched i.e policy cannot change
444 * on these sockets. As we have the right policy on the conn, we don't
445 * have to lookup global policy for every outbound and inbound datagram
446 * and thus serving as an optimization. Note that a global policy change
447 * does not affect fully connected sockets if they have policy. If fully
448 * connected sockets did not have any policy associated with it, global
449 * policy change may affect them.
450 *
451 * IP Flow control notes:
452 * ---------------------
453 * Non-TCP streams are flow controlled by IP. The way this is accomplished
454 * differs when ILL_CAPAB_DLD_DIRECT is enabled for that IP instance. When
455 * ILL_DIRECT_CAPABLE(ill) is TRUE, IP can do direct function calls into
456 * GLDv3. Otherwise packets are sent down to lower layers using STREAMS
457 * functions.
458 *
459 * Per Tx ring udp flow control:
460 * This is applicable only when ILL_CAPAB_DLD_DIRECT capability is set in
461 * the ill (i.e. ILL_DIRECT_CAPABLE(ill) is true).
462 *
463 * The underlying link can expose multiple Tx rings to the GLDv3 mac layer.
464 * To achieve best performance, outgoing traffic need to be fanned out among
465 * these Tx ring. mac_tx() is called (via str_mdata_fastpath_put()) to send
466 * traffic out of the NIC and it takes a fanout hint. UDP connections pass
467 * the address of connp as fanout hint to mac_tx(). Under flow controlled
468 * condition, mac_tx() returns a non-NULL cookie (ip_mac_tx_cookie_t). This
469 * cookie points to a specific Tx ring that is blocked. The cookie is used to
470 * hash into an idl_tx_list[] entry in idl_tx_list[] array. Each idl_tx_list_t
471 * point to drain_lists (idl_t's). These drain list will store the blocked UDP
472 * connp's. The drain list is not a single list but a configurable number of
473 * lists.
474 *
475 * The diagram below shows idl_tx_list_t's and their drain_lists. ip_stack_t
476 * has an array of idl_tx_list_t. The size of the array is TX_FANOUT_SIZE
477 * which is equal to 128. This array in turn contains a pointer to idl_t[],
478 * the ip drain list. The idl_t[] array size is MIN(max_ncpus, 8). The drain
479 * list will point to the list of connp's that are flow controlled.
480 *
481 * --------------- ------- ------- -------
482 * |->|drain_list[0]|-->|connp|-->|connp|-->|connp|-->
483 * | --------------- ------- ------- -------
484 * | --------------- ------- ------- -------
485 * |->|drain_list[1]|-->|connp|-->|connp|-->|connp|-->
486 * ---------------- | --------------- ------- ------- -------
487 * |idl_tx_list[0]|->| --------------- ------- ------- -------
488 * ---------------- |->|drain_list[2]|-->|connp|-->|connp|-->|connp|-->
489 * | --------------- ------- ------- -------
490 * . . . . .
491 * | --------------- ------- ------- -------
492 * |->|drain_list[n]|-->|connp|-->|connp|-->|connp|-->
493 * --------------- ------- ------- -------
494 * --------------- ------- ------- -------
495 * |->|drain_list[0]|-->|connp|-->|connp|-->|connp|-->
496 * | --------------- ------- ------- -------
497 * | --------------- ------- ------- -------
498 * ---------------- |->|drain_list[1]|-->|connp|-->|connp|-->|connp|-->
499 * |idl_tx_list[1]|->| --------------- ------- ------- -------
500 * ---------------- | . . . .
501 * | --------------- ------- ------- -------
502 * |->|drain_list[n]|-->|connp|-->|connp|-->|connp|-->
503 * --------------- ------- ------- -------
504 * .....
505 * ----------------
506 * |idl_tx_list[n]|-> ...
507 * ----------------
508 *
509 * When mac_tx() returns a cookie, the cookie is hashed into an index into
510 * ips_idl_tx_list[], and conn_drain_insert() is called with the idl_tx_list
511 * to insert the conn onto. conn_drain_insert() asserts flow control for the
512 * sockets via su_txq_full() (non-STREAMS) or QFULL on conn_wq (STREAMS).
513 * Further, conn_blocked is set to indicate that the conn is blocked.
514 *
515 * GLDv3 calls ill_flow_enable() when flow control is relieved. The cookie
516 * passed in the call to ill_flow_enable() identifies the blocked Tx ring and
517 * is again hashed to locate the appropriate idl_tx_list, which is then
518 * drained via conn_walk_drain(). conn_walk_drain() goes through each conn in
519 * the drain list and calls conn_drain_remove() to clear flow control (via
520 * calling su_txq_full() or clearing QFULL), and remove the conn from the
521 * drain list.
522 *
523 * Note that the drain list is not a single list but a (configurable) array of
524 * lists (8 elements by default). Synchronization between drain insertion and
525 * flow control wakeup is handled by using idl_txl->txl_lock, and only
526 * conn_drain_insert() and conn_drain_remove() manipulate the drain list.
527 *
528 * Flow control via STREAMS is used when ILL_DIRECT_CAPABLE() returns FALSE.
529 * On the send side, if the packet cannot be sent down to the driver by IP
530 * (canput() fails), ip_xmit() drops the packet and returns EWOULDBLOCK to the
531 * caller, who may then invoke ixa_check_drain_insert() to insert the conn on
532 * the 0'th drain list. When ip_wsrv() runs on the ill_wq because flow
533 * control has been relieved, the blocked conns in the 0'th drain list are
534 * drained as in the non-STREAMS case.
535 *
536 * In both the STREAMS and non-STREAMS cases, the sockfs upcall to set QFULL
537 * is done when the conn is inserted into the drain list (conn_drain_insert())
538 * and cleared when the conn is removed from the it (conn_drain_remove()).
539 *
540 * IPQOS notes:
541 *
542 * IPQoS Policies are applied to packets using IPPF (IP Policy framework)
543 * and IPQoS modules. IPPF includes hooks in IP at different control points
544 * (callout positions) which direct packets to IPQoS modules for policy
545 * processing. Policies, if present, are global.
546 *
547 * The callout positions are located in the following paths:
548 * o local_in (packets destined for this host)
549 * o local_out (packets orginating from this host )
550 * o fwd_in (packets forwarded by this m/c - inbound)
551 * o fwd_out (packets forwarded by this m/c - outbound)
552 * Hooks at these callout points can be enabled/disabled using the ndd variable
553 * ip_policy_mask (a bit mask with the 4 LSB indicating the callout positions).
554 * By default all the callout positions are enabled.
555 *
556 * Outbound (local_out)
557 * Hooks are placed in ire_send_wire_v4 and ire_send_wire_v6.
558 *
559 * Inbound (local_in)
560 * Hooks are placed in ip_fanout_v4 and ip_fanout_v6.
561 *
562 * Forwarding (in and out)
563 * Hooks are placed in ire_recv_forward_v4/v6.
564 *
565 * IP Policy Framework processing (IPPF processing)
566 * Policy processing for a packet is initiated by ip_process, which ascertains
567 * that the classifier (ipgpc) is loaded and configured, failing which the
568 * packet resumes normal processing in IP. If the clasifier is present, the
569 * packet is acted upon by one or more IPQoS modules (action instances), per
570 * filters configured in ipgpc and resumes normal IP processing thereafter.
571 * An action instance can drop a packet in course of its processing.
572 *
573 * Zones notes:
574 *
575 * The partitioning rules for networking are as follows:
576 * 1) Packets coming from a zone must have a source address belonging to that
577 * zone.
578 * 2) Packets coming from a zone can only be sent on a physical interface on
579 * which the zone has an IP address.
580 * 3) Between two zones on the same machine, packet delivery is only allowed if
581 * there's a matching route for the destination and zone in the forwarding
582 * table.
583 * 4) The TCP and UDP port spaces are per-zone; that is, two processes in
584 * different zones can bind to the same port with the wildcard address
585 * (INADDR_ANY).
586 *
587 * The granularity of interface partitioning is at the logical interface level.
588 * Therefore, every zone has its own IP addresses, and incoming packets can be
589 * attributed to a zone unambiguously. A logical interface is placed into a zone
590 * using the SIOCSLIFZONE ioctl; this sets the ipif_zoneid field in the ipif_t
591 * structure. Rule (1) is implemented by modifying the source address selection
592 * algorithm so that the list of eligible addresses is filtered based on the
593 * sending process zone.
594 *
595 * The Internet Routing Entries (IREs) are either exclusive to a zone or shared
596 * across all zones, depending on their type. Here is the break-up:
597 *
598 * IRE type Shared/exclusive
599 * -------- ----------------
600 * IRE_BROADCAST Exclusive
601 * IRE_DEFAULT (default routes) Shared (*)
602 * IRE_LOCAL Exclusive (x)
603 * IRE_LOOPBACK Exclusive
604 * IRE_PREFIX (net routes) Shared (*)
605 * IRE_IF_NORESOLVER (interface routes) Exclusive
606 * IRE_IF_RESOLVER (interface routes) Exclusive
607 * IRE_IF_CLONE (interface routes) Exclusive
608 * IRE_HOST (host routes) Shared (*)
609 *
610 * (*) A zone can only use a default or off-subnet route if the gateway is
611 * directly reachable from the zone, that is, if the gateway's address matches
612 * one of the zone's logical interfaces.
613 *
614 * (x) IRE_LOCAL are handled a bit differently.
615 * When ip_restrict_interzone_loopback is set (the default),
616 * ire_route_recursive restricts loopback using an IRE_LOCAL
617 * between zone to the case when L2 would have conceptually looped the packet
618 * back, i.e. the loopback which is required since neither Ethernet drivers
619 * nor Ethernet hardware loops them back. This is the case when the normal
620 * routes (ignoring IREs with different zoneids) would send out the packet on
621 * the same ill as the ill with which is IRE_LOCAL is associated.
622 *
623 * Multiple zones can share a common broadcast address; typically all zones
624 * share the 255.255.255.255 address. Incoming as well as locally originated
625 * broadcast packets must be dispatched to all the zones on the broadcast
626 * network. For directed broadcasts (e.g. 10.16.72.255) this is not trivial
627 * since some zones may not be on the 10.16.72/24 network. To handle this, each
628 * zone has its own set of IRE_BROADCAST entries; then, broadcast packets are
629 * sent to every zone that has an IRE_BROADCAST entry for the destination
630 * address on the input ill, see ip_input_broadcast().
631 *
632 * Applications in different zones can join the same multicast group address.
633 * The same logic applies for multicast as for broadcast. ip_input_multicast
634 * dispatches packets to all zones that have members on the physical interface.
635 */
636
637 /*
638 * Squeue Fanout flags:
639 * 0: No fanout.
640 * 1: Fanout across all squeues
641 */
642 boolean_t ip_squeue_fanout = 0;
643
644 /*
645 * Maximum dups allowed per packet.
646 */
647 uint_t ip_max_frag_dups = 10;
648
649 static int ip_open(queue_t *q, dev_t *devp, int flag, int sflag,
650 cred_t *credp, boolean_t isv6);
651 static mblk_t *ip_xmit_attach_llhdr(mblk_t *, nce_t *);
652
653 static boolean_t icmp_inbound_verify_v4(mblk_t *, icmph_t *, ip_recv_attr_t *);
654 static void icmp_inbound_too_big_v4(icmph_t *, ip_recv_attr_t *);
655 static void icmp_inbound_error_fanout_v4(mblk_t *, icmph_t *,
656 ip_recv_attr_t *);
657 static void icmp_options_update(ipha_t *);
658 static void icmp_param_problem(mblk_t *, uint8_t, ip_recv_attr_t *);
659 static void icmp_pkt(mblk_t *, void *, size_t, ip_recv_attr_t *);
660 static mblk_t *icmp_pkt_err_ok(mblk_t *, ip_recv_attr_t *);
661 static void icmp_redirect_v4(mblk_t *mp, ipha_t *, icmph_t *,
662 ip_recv_attr_t *);
663 static void icmp_send_redirect(mblk_t *, ipaddr_t, ip_recv_attr_t *);
664 static void icmp_send_reply_v4(mblk_t *, ipha_t *, icmph_t *,
665 ip_recv_attr_t *);
666
667 mblk_t *ip_dlpi_alloc(size_t, t_uscalar_t);
668 char *ip_dot_addr(ipaddr_t, char *);
669 mblk_t *ip_carve_mp(mblk_t **, ssize_t);
670 static char *ip_dot_saddr(uchar_t *, char *);
671 static int ip_lrput(queue_t *, mblk_t *);
672 ipaddr_t ip_net_mask(ipaddr_t);
673 char *ip_nv_lookup(nv_t *, int);
674 int ip_rput(queue_t *, mblk_t *);
675 static void ip_rput_dlpi_writer(ipsq_t *dummy_sq, queue_t *q, mblk_t *mp,
676 void *dummy_arg);
677 int ip_snmp_get(queue_t *, mblk_t *, int, boolean_t);
678 static mblk_t *ip_snmp_get_mib2_ip(queue_t *, mblk_t *,
679 mib2_ipIfStatsEntry_t *, ip_stack_t *, boolean_t);
680 static mblk_t *ip_snmp_get_mib2_ip_traffic_stats(queue_t *, mblk_t *,
681 ip_stack_t *, boolean_t);
682 static mblk_t *ip_snmp_get_mib2_ip6(queue_t *, mblk_t *, ip_stack_t *,
683 boolean_t);
684 static mblk_t *ip_snmp_get_mib2_icmp(queue_t *, mblk_t *, ip_stack_t *ipst);
685 static mblk_t *ip_snmp_get_mib2_icmp6(queue_t *, mblk_t *, ip_stack_t *ipst);
686 static mblk_t *ip_snmp_get_mib2_igmp(queue_t *, mblk_t *, ip_stack_t *ipst);
687 static mblk_t *ip_snmp_get_mib2_multi(queue_t *, mblk_t *, ip_stack_t *ipst);
688 static mblk_t *ip_snmp_get_mib2_ip_addr(queue_t *, mblk_t *,
689 ip_stack_t *ipst, boolean_t);
690 static mblk_t *ip_snmp_get_mib2_ip6_addr(queue_t *, mblk_t *,
691 ip_stack_t *ipst, boolean_t);
692 static mblk_t *ip_snmp_get_mib2_ip_group_src(queue_t *, mblk_t *,
693 ip_stack_t *ipst);
694 static mblk_t *ip_snmp_get_mib2_ip6_group_src(queue_t *, mblk_t *,
695 ip_stack_t *ipst);
696 static mblk_t *ip_snmp_get_mib2_ip_group_mem(queue_t *, mblk_t *,
697 ip_stack_t *ipst);
698 static mblk_t *ip_snmp_get_mib2_ip6_group_mem(queue_t *, mblk_t *,
699 ip_stack_t *ipst);
700 static mblk_t *ip_snmp_get_mib2_virt_multi(queue_t *, mblk_t *,
701 ip_stack_t *ipst);
702 static mblk_t *ip_snmp_get_mib2_multi_rtable(queue_t *, mblk_t *,
703 ip_stack_t *ipst);
704 static mblk_t *ip_snmp_get_mib2_ip_route_media(queue_t *, mblk_t *, int,
705 ip_stack_t *ipst);
706 static mblk_t *ip_snmp_get_mib2_ip6_route_media(queue_t *, mblk_t *, int,
707 ip_stack_t *ipst);
708 static void ip_snmp_get2_v4(ire_t *, iproutedata_t *);
709 static void ip_snmp_get2_v6_route(ire_t *, iproutedata_t *);
710 static void ip_snmp_get2_v4_media(ncec_t *, void *);
711 static void ip_snmp_get2_v6_media(ncec_t *, void *);
712 int ip_snmp_set(queue_t *, int, int, uchar_t *, int);
713
714 static mblk_t *ip_fragment_copyhdr(uchar_t *, int, int, ip_stack_t *,
715 mblk_t *);
716
717 static void conn_drain_init(ip_stack_t *);
718 static void conn_drain_fini(ip_stack_t *);
719 static void conn_drain(conn_t *connp, boolean_t closing);
720
721 static void conn_walk_drain(ip_stack_t *, idl_tx_list_t *);
722 static void conn_walk_sctp(pfv_t, void *, zoneid_t, netstack_t *);
723
724 static void *ip_stack_init(netstackid_t stackid, netstack_t *ns);
725 static void ip_stack_shutdown(netstackid_t stackid, void *arg);
726 static void ip_stack_fini(netstackid_t stackid, void *arg);
727
728 static int ip_multirt_apply_membership(int (*fn)(conn_t *, boolean_t,
729 const in6_addr_t *, ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *),
730 ire_t *, conn_t *, boolean_t, const in6_addr_t *, mcast_record_t,
731 const in6_addr_t *);
732
733 static int ip_squeue_switch(int);
734
735 static void *ip_kstat_init(netstackid_t, ip_stack_t *);
736 static void ip_kstat_fini(netstackid_t, kstat_t *);
737 static int ip_kstat_update(kstat_t *kp, int rw);
738 static void *icmp_kstat_init(netstackid_t);
739 static void icmp_kstat_fini(netstackid_t, kstat_t *);
740 static int icmp_kstat_update(kstat_t *kp, int rw);
741 static void *ip_kstat2_init(netstackid_t, ip_stat_t *);
742 static void ip_kstat2_fini(netstackid_t, kstat_t *);
743
744 static void ipobs_init(ip_stack_t *);
745 static void ipobs_fini(ip_stack_t *);
746
747 static int ip_tp_cpu_update(cpu_setup_t, int, void *);
748
749 ipaddr_t ip_g_all_ones = IP_HOST_MASK;
750
751 static long ip_rput_pullups;
752 int dohwcksum = 1; /* use h/w cksum if supported by the hardware */
753
754 vmem_t *ip_minor_arena_sa; /* for minor nos. from INET_MIN_DEV+2 thru 2^^18-1 */
755 vmem_t *ip_minor_arena_la; /* for minor nos. from 2^^18 thru 2^^32-1 */
756
757 int ip_debug;
758
759 /*
760 * Multirouting/CGTP stuff
761 */
762 int ip_cgtp_filter_rev = CGTP_FILTER_REV; /* CGTP hooks version */
763
764 /*
765 * IP tunables related declarations. Definitions are in ip_tunables.c
766 */
767 extern mod_prop_info_t ip_propinfo_tbl[];
768 extern int ip_propinfo_count;
769
770 /*
771 * Table of IP ioctls encoding the various properties of the ioctl and
772 * indexed based on the last byte of the ioctl command. Occasionally there
773 * is a clash, and there is more than 1 ioctl with the same last byte.
774 * In such a case 1 ioctl is encoded in the ndx table and the remaining
775 * ioctls are encoded in the misc table. An entry in the ndx table is
776 * retrieved by indexing on the last byte of the ioctl command and comparing
777 * the ioctl command with the value in the ndx table. In the event of a
778 * mismatch the misc table is then searched sequentially for the desired
779 * ioctl command.
780 *
781 * Entry: <command> <copyin_size> <flags> <cmd_type> <function> <restart_func>
782 */
783 ip_ioctl_cmd_t ip_ndx_ioctl_table[] = {
784 /* 000 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
785 /* 001 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
786 /* 002 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
787 /* 003 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
788 /* 004 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
789 /* 005 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
790 /* 006 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
791 /* 007 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
792 /* 008 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
793 /* 009 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
794
795 /* 010 */ { SIOCADDRT, sizeof (struct rtentry), IPI_PRIV,
796 MISC_CMD, ip_siocaddrt, NULL },
797 /* 011 */ { SIOCDELRT, sizeof (struct rtentry), IPI_PRIV,
798 MISC_CMD, ip_siocdelrt, NULL },
799
800 /* 012 */ { SIOCSIFADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
801 IF_CMD, ip_sioctl_addr, ip_sioctl_addr_restart },
802 /* 013 */ { SIOCGIFADDR, sizeof (struct ifreq), IPI_GET_CMD,
803 IF_CMD, ip_sioctl_get_addr, NULL },
804
805 /* 014 */ { SIOCSIFDSTADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
806 IF_CMD, ip_sioctl_dstaddr, ip_sioctl_dstaddr_restart },
807 /* 015 */ { SIOCGIFDSTADDR, sizeof (struct ifreq),
808 IPI_GET_CMD, IF_CMD, ip_sioctl_get_dstaddr, NULL },
809
810 /* 016 */ { SIOCSIFFLAGS, sizeof (struct ifreq),
811 IPI_PRIV | IPI_WR,
812 IF_CMD, ip_sioctl_flags, ip_sioctl_flags_restart },
813 /* 017 */ { SIOCGIFFLAGS, sizeof (struct ifreq),
814 IPI_MODOK | IPI_GET_CMD,
815 IF_CMD, ip_sioctl_get_flags, NULL },
816
817 /* 018 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
818 /* 019 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
819
820 /* copyin size cannot be coded for SIOCGIFCONF */
821 /* 020 */ { O_SIOCGIFCONF, 0, IPI_GET_CMD,
822 MISC_CMD, ip_sioctl_get_ifconf, NULL },
823
824 /* 021 */ { SIOCSIFMTU, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
825 IF_CMD, ip_sioctl_mtu, NULL },
826 /* 022 */ { SIOCGIFMTU, sizeof (struct ifreq), IPI_GET_CMD,
827 IF_CMD, ip_sioctl_get_mtu, NULL },
828 /* 023 */ { SIOCGIFBRDADDR, sizeof (struct ifreq),
829 IPI_GET_CMD, IF_CMD, ip_sioctl_get_brdaddr, NULL },
830 /* 024 */ { SIOCSIFBRDADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
831 IF_CMD, ip_sioctl_brdaddr, NULL },
832 /* 025 */ { SIOCGIFNETMASK, sizeof (struct ifreq),
833 IPI_GET_CMD, IF_CMD, ip_sioctl_get_netmask, NULL },
834 /* 026 */ { SIOCSIFNETMASK, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
835 IF_CMD, ip_sioctl_netmask, ip_sioctl_netmask_restart },
836 /* 027 */ { SIOCGIFMETRIC, sizeof (struct ifreq),
837 IPI_GET_CMD, IF_CMD, ip_sioctl_get_metric, NULL },
838 /* 028 */ { SIOCSIFMETRIC, sizeof (struct ifreq), IPI_PRIV,
839 IF_CMD, ip_sioctl_metric, NULL },
840 /* 029 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
841
842 /* See 166-168 below for extended SIOC*XARP ioctls */
843 /* 030 */ { SIOCSARP, sizeof (struct arpreq), IPI_PRIV | IPI_WR,
844 ARP_CMD, ip_sioctl_arp, NULL },
845 /* 031 */ { SIOCGARP, sizeof (struct arpreq), IPI_GET_CMD,
846 ARP_CMD, ip_sioctl_arp, NULL },
847 /* 032 */ { SIOCDARP, sizeof (struct arpreq), IPI_PRIV | IPI_WR,
848 ARP_CMD, ip_sioctl_arp, NULL },
849
850 /* 033 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
851 /* 034 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
852 /* 035 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
853 /* 036 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
854 /* 037 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
855 /* 038 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
856 /* 039 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
857 /* 040 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
858 /* 041 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
859 /* 042 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
860 /* 043 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
861 /* 044 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
862 /* 045 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
863 /* 046 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
864 /* 047 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
865 /* 048 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
866 /* 049 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
867 /* 050 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
868 /* 051 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
869 /* 052 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
870 /* 053 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
871
872 /* 054 */ { IF_UNITSEL, sizeof (int), IPI_PRIV | IPI_WR | IPI_MODOK,
873 MISC_CMD, if_unitsel, if_unitsel_restart },
874
875 /* 055 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
876 /* 056 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
877 /* 057 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
878 /* 058 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
879 /* 059 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
880 /* 060 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
881 /* 061 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
882 /* 062 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
883 /* 063 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
884 /* 064 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
885 /* 065 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
886 /* 066 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
887 /* 067 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
888 /* 068 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
889 /* 069 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
890 /* 070 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
891 /* 071 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
892 /* 072 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
893
894 /* 073 */ { SIOCSIFNAME, sizeof (struct ifreq),
895 IPI_PRIV | IPI_WR | IPI_MODOK,
896 IF_CMD, ip_sioctl_sifname, NULL },
897
898 /* 074 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
899 /* 075 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
900 /* 076 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
901 /* 077 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
902 /* 078 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
903 /* 079 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
904 /* 080 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
905 /* 081 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
906 /* 082 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
907 /* 083 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
908 /* 084 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
909 /* 085 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
910 /* 086 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
911
912 /* 087 */ { SIOCGIFNUM, sizeof (int), IPI_GET_CMD,
913 MISC_CMD, ip_sioctl_get_ifnum, NULL },
914 /* 088 */ { SIOCGIFMUXID, sizeof (struct ifreq), IPI_GET_CMD,
915 IF_CMD, ip_sioctl_get_muxid, NULL },
916 /* 089 */ { SIOCSIFMUXID, sizeof (struct ifreq),
917 IPI_PRIV | IPI_WR, IF_CMD, ip_sioctl_muxid, NULL },
918
919 /* Both if and lif variants share same func */
920 /* 090 */ { SIOCGIFINDEX, sizeof (struct ifreq), IPI_GET_CMD,
921 IF_CMD, ip_sioctl_get_lifindex, NULL },
922 /* Both if and lif variants share same func */
923 /* 091 */ { SIOCSIFINDEX, sizeof (struct ifreq),
924 IPI_PRIV | IPI_WR, IF_CMD, ip_sioctl_slifindex, NULL },
925
926 /* copyin size cannot be coded for SIOCGIFCONF */
927 /* 092 */ { SIOCGIFCONF, 0, IPI_GET_CMD,
928 MISC_CMD, ip_sioctl_get_ifconf, NULL },
929 /* 093 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
930 /* 094 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
931 /* 095 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
932 /* 096 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
933 /* 097 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
934 /* 098 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
935 /* 099 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
936 /* 100 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
937 /* 101 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
938 /* 102 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
939 /* 103 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
940 /* 104 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
941 /* 105 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
942 /* 106 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
943 /* 107 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
944 /* 108 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
945 /* 109 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
946
947 /* 110 */ { SIOCLIFREMOVEIF, sizeof (struct lifreq),
948 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_removeif,
949 ip_sioctl_removeif_restart },
950 /* 111 */ { SIOCLIFADDIF, sizeof (struct lifreq),
951 IPI_GET_CMD | IPI_PRIV | IPI_WR,
952 LIF_CMD, ip_sioctl_addif, NULL },
953 #define SIOCLIFADDR_NDX 112
954 /* 112 */ { SIOCSLIFADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
955 LIF_CMD, ip_sioctl_addr, ip_sioctl_addr_restart },
956 /* 113 */ { SIOCGLIFADDR, sizeof (struct lifreq),
957 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_addr, NULL },
958 /* 114 */ { SIOCSLIFDSTADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
959 LIF_CMD, ip_sioctl_dstaddr, ip_sioctl_dstaddr_restart },
960 /* 115 */ { SIOCGLIFDSTADDR, sizeof (struct lifreq),
961 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_dstaddr, NULL },
962 /* 116 */ { SIOCSLIFFLAGS, sizeof (struct lifreq),
963 IPI_PRIV | IPI_WR,
964 LIF_CMD, ip_sioctl_flags, ip_sioctl_flags_restart },
965 /* 117 */ { SIOCGLIFFLAGS, sizeof (struct lifreq),
966 IPI_GET_CMD | IPI_MODOK,
967 LIF_CMD, ip_sioctl_get_flags, NULL },
968
969 /* 118 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
970 /* 119 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
971
972 /* 120 */ { O_SIOCGLIFCONF, 0, IPI_GET_CMD, MISC_CMD,
973 ip_sioctl_get_lifconf, NULL },
974 /* 121 */ { SIOCSLIFMTU, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
975 LIF_CMD, ip_sioctl_mtu, NULL },
976 /* 122 */ { SIOCGLIFMTU, sizeof (struct lifreq), IPI_GET_CMD,
977 LIF_CMD, ip_sioctl_get_mtu, NULL },
978 /* 123 */ { SIOCGLIFBRDADDR, sizeof (struct lifreq),
979 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_brdaddr, NULL },
980 /* 124 */ { SIOCSLIFBRDADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
981 LIF_CMD, ip_sioctl_brdaddr, NULL },
982 /* 125 */ { SIOCGLIFNETMASK, sizeof (struct lifreq),
983 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_netmask, NULL },
984 /* 126 */ { SIOCSLIFNETMASK, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
985 LIF_CMD, ip_sioctl_netmask, ip_sioctl_netmask_restart },
986 /* 127 */ { SIOCGLIFMETRIC, sizeof (struct lifreq),
987 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_metric, NULL },
988 /* 128 */ { SIOCSLIFMETRIC, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
989 LIF_CMD, ip_sioctl_metric, NULL },
990 /* 129 */ { SIOCSLIFNAME, sizeof (struct lifreq),
991 IPI_PRIV | IPI_WR | IPI_MODOK,
992 LIF_CMD, ip_sioctl_slifname,
993 ip_sioctl_slifname_restart },
994
995 /* 130 */ { SIOCGLIFNUM, sizeof (struct lifnum), IPI_GET_CMD,
996 MISC_CMD, ip_sioctl_get_lifnum, NULL },
997 /* 131 */ { SIOCGLIFMUXID, sizeof (struct lifreq),
998 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_muxid, NULL },
999 /* 132 */ { SIOCSLIFMUXID, sizeof (struct lifreq),
1000 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_muxid, NULL },
1001 /* 133 */ { SIOCGLIFINDEX, sizeof (struct lifreq),
1002 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lifindex, 0 },
1003 /* 134 */ { SIOCSLIFINDEX, sizeof (struct lifreq),
1004 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_slifindex, 0 },
1005 /* 135 */ { SIOCSLIFTOKEN, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1006 LIF_CMD, ip_sioctl_token, NULL },
1007 /* 136 */ { SIOCGLIFTOKEN, sizeof (struct lifreq),
1008 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_token, NULL },
1009 /* 137 */ { SIOCSLIFSUBNET, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1010 LIF_CMD, ip_sioctl_subnet, ip_sioctl_subnet_restart },
1011 /* 138 */ { SIOCGLIFSUBNET, sizeof (struct lifreq),
1012 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_subnet, NULL },
1013 /* 139 */ { SIOCSLIFLNKINFO, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1014 LIF_CMD, ip_sioctl_lnkinfo, NULL },
1015
1016 /* 140 */ { SIOCGLIFLNKINFO, sizeof (struct lifreq),
1017 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lnkinfo, NULL },
1018 /* 141 */ { SIOCLIFDELND, sizeof (struct lifreq), IPI_PRIV,
1019 LIF_CMD, ip_siocdelndp_v6, NULL },
1020 /* 142 */ { SIOCLIFGETND, sizeof (struct lifreq), IPI_GET_CMD,
1021 LIF_CMD, ip_siocqueryndp_v6, NULL },
1022 /* 143 */ { SIOCLIFSETND, sizeof (struct lifreq), IPI_PRIV,
1023 LIF_CMD, ip_siocsetndp_v6, NULL },
1024 /* 144 */ { SIOCTMYADDR, sizeof (struct sioc_addrreq), IPI_GET_CMD,
1025 MISC_CMD, ip_sioctl_tmyaddr, NULL },
1026 /* 145 */ { SIOCTONLINK, sizeof (struct sioc_addrreq), IPI_GET_CMD,
1027 MISC_CMD, ip_sioctl_tonlink, NULL },
1028 /* 146 */ { SIOCTMYSITE, sizeof (struct sioc_addrreq), 0,
1029 MISC_CMD, ip_sioctl_tmysite, NULL },
1030 /* 147 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1031 /* 148 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1032
1033 /* Old *IPSECONFIG ioctls are now deprecated, now see spdsock.c */
1034 /* 149 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1035 /* 150 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1036 /* 151 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1037 /* 152 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1038
1039 /* 153 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1040
1041 /* 154 */ { SIOCGLIFBINDING, sizeof (struct lifreq), IPI_GET_CMD,
1042 LIF_CMD, ip_sioctl_get_binding, NULL },
1043 /* 155 */ { SIOCSLIFGROUPNAME, sizeof (struct lifreq),
1044 IPI_PRIV | IPI_WR,
1045 LIF_CMD, ip_sioctl_groupname, ip_sioctl_groupname },
1046 /* 156 */ { SIOCGLIFGROUPNAME, sizeof (struct lifreq),
1047 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_groupname, NULL },
1048 /* 157 */ { SIOCGLIFGROUPINFO, sizeof (lifgroupinfo_t),
1049 IPI_GET_CMD, MISC_CMD, ip_sioctl_groupinfo, NULL },
1050
1051 /* Leave 158-160 unused; used to be SIOC*IFARP ioctls */
1052 /* 158 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1053 /* 159 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1054 /* 160 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1055
1056 /* 161 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1057
1058 /* These are handled in ip_sioctl_copyin_setup itself */
1059 /* 162 */ { SIOCGIP6ADDRPOLICY, 0, IPI_NULL_BCONT,
1060 MISC_CMD, NULL, NULL },
1061 /* 163 */ { SIOCSIP6ADDRPOLICY, 0, IPI_PRIV | IPI_NULL_BCONT,
1062 MISC_CMD, NULL, NULL },
1063 /* 164 */ { SIOCGDSTINFO, 0, IPI_GET_CMD, MISC_CMD, NULL, NULL },
1064
1065 /* 165 */ { SIOCGLIFCONF, 0, IPI_GET_CMD, MISC_CMD,
1066 ip_sioctl_get_lifconf, NULL },
1067
1068 /* 166 */ { SIOCSXARP, sizeof (struct xarpreq), IPI_PRIV | IPI_WR,
1069 XARP_CMD, ip_sioctl_arp, NULL },
1070 /* 167 */ { SIOCGXARP, sizeof (struct xarpreq), IPI_GET_CMD,
1071 XARP_CMD, ip_sioctl_arp, NULL },
1072 /* 168 */ { SIOCDXARP, sizeof (struct xarpreq), IPI_PRIV | IPI_WR,
1073 XARP_CMD, ip_sioctl_arp, NULL },
1074
1075 /* SIOCPOPSOCKFS is not handled by IP */
1076 /* 169 */ { IPI_DONTCARE /* SIOCPOPSOCKFS */, 0, 0, 0, NULL, NULL },
1077
1078 /* 170 */ { SIOCGLIFZONE, sizeof (struct lifreq),
1079 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lifzone, NULL },
1080 /* 171 */ { SIOCSLIFZONE, sizeof (struct lifreq),
1081 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_slifzone,
1082 ip_sioctl_slifzone_restart },
1083 /* 172-174 are SCTP ioctls and not handled by IP */
1084 /* 172 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1085 /* 173 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1086 /* 174 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1087 /* 175 */ { SIOCGLIFUSESRC, sizeof (struct lifreq),
1088 IPI_GET_CMD, LIF_CMD,
1089 ip_sioctl_get_lifusesrc, 0 },
1090 /* 176 */ { SIOCSLIFUSESRC, sizeof (struct lifreq),
1091 IPI_PRIV | IPI_WR,
1092 LIF_CMD, ip_sioctl_slifusesrc,
1093 NULL },
1094 /* 177 */ { SIOCGLIFSRCOF, 0, IPI_GET_CMD, MISC_CMD,
1095 ip_sioctl_get_lifsrcof, NULL },
1096 /* 178 */ { SIOCGMSFILTER, sizeof (struct group_filter), IPI_GET_CMD,
1097 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1098 /* 179 */ { SIOCSMSFILTER, sizeof (struct group_filter), 0,
1099 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1100 /* 180 */ { SIOCGIPMSFILTER, sizeof (struct ip_msfilter), IPI_GET_CMD,
1101 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1102 /* 181 */ { SIOCSIPMSFILTER, sizeof (struct ip_msfilter), 0,
1103 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1104 /* 182 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1105 /* SIOCSENABLESDP is handled by SDP */
1106 /* 183 */ { IPI_DONTCARE /* SIOCSENABLESDP */, 0, 0, 0, NULL, NULL },
1107 /* 184 */ { IPI_DONTCARE /* SIOCSQPTR */, 0, 0, 0, NULL, NULL },
1108 /* 185 */ { SIOCGIFHWADDR, sizeof (struct ifreq), IPI_GET_CMD,
1109 IF_CMD, ip_sioctl_get_ifhwaddr, NULL },
1110 /* 186 */ { IPI_DONTCARE /* SIOCGSTAMP */, 0, 0, 0, NULL, NULL },
1111 /* 187 */ { SIOCILB, 0, IPI_PRIV | IPI_GET_CMD, MISC_CMD,
1112 ip_sioctl_ilb_cmd, NULL },
1113 /* 188 */ { SIOCGETPROP, 0, IPI_GET_CMD, 0, NULL, NULL },
1114 /* 189 */ { SIOCSETPROP, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL},
1115 /* 190 */ { SIOCGLIFDADSTATE, sizeof (struct lifreq),
1116 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_dadstate, NULL },
1117 /* 191 */ { SIOCSLIFPREFIX, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1118 LIF_CMD, ip_sioctl_prefix, ip_sioctl_prefix_restart },
1119 /* 192 */ { SIOCGLIFHWADDR, sizeof (struct lifreq), IPI_GET_CMD,
1120 LIF_CMD, ip_sioctl_get_lifhwaddr, NULL }
1121 };
1122
1123 int ip_ndx_ioctl_count = sizeof (ip_ndx_ioctl_table) / sizeof (ip_ioctl_cmd_t);
1124
1125 ip_ioctl_cmd_t ip_misc_ioctl_table[] = {
1126 { I_LINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1127 { I_UNLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1128 { I_PLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1129 { I_PUNLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1130 { ND_GET, 0, 0, 0, NULL, NULL },
1131 { ND_SET, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1132 { IP_IOCTL, 0, 0, 0, NULL, NULL },
1133 { SIOCGETVIFCNT, sizeof (struct sioc_vif_req), IPI_GET_CMD,
1134 MISC_CMD, mrt_ioctl},
1135 { SIOCGETSGCNT, sizeof (struct sioc_sg_req), IPI_GET_CMD,
1136 MISC_CMD, mrt_ioctl},
1137 { SIOCGETLSGCNT, sizeof (struct sioc_lsg_req), IPI_GET_CMD,
1138 MISC_CMD, mrt_ioctl}
1139 };
1140
1141 int ip_misc_ioctl_count =
1142 sizeof (ip_misc_ioctl_table) / sizeof (ip_ioctl_cmd_t);
1143
1144 int conn_drain_nthreads; /* Number of drainers reqd. */
1145 /* Settable in /etc/system */
1146 /* Defined in ip_ire.c */
1147 extern uint32_t ip_ire_max_bucket_cnt, ip6_ire_max_bucket_cnt;
1148 extern uint32_t ip_ire_min_bucket_cnt, ip6_ire_min_bucket_cnt;
1149 extern uint32_t ip_ire_mem_ratio, ip_ire_cpu_ratio;
1150
1151 static nv_t ire_nv_arr[] = {
1152 { IRE_BROADCAST, "BROADCAST" },
1153 { IRE_LOCAL, "LOCAL" },
1154 { IRE_LOOPBACK, "LOOPBACK" },
1155 { IRE_DEFAULT, "DEFAULT" },
1156 { IRE_PREFIX, "PREFIX" },
1157 { IRE_IF_NORESOLVER, "IF_NORESOL" },
1158 { IRE_IF_RESOLVER, "IF_RESOLV" },
1159 { IRE_IF_CLONE, "IF_CLONE" },
1160 { IRE_HOST, "HOST" },
1161 { IRE_MULTICAST, "MULTICAST" },
1162 { IRE_NOROUTE, "NOROUTE" },
1163 { 0 }
1164 };
1165
1166 nv_t *ire_nv_tbl = ire_nv_arr;
1167
1168 /* Simple ICMP IP Header Template */
1169 static ipha_t icmp_ipha = {
1170 IP_SIMPLE_HDR_VERSION, 0, 0, 0, 0, 0, IPPROTO_ICMP
1171 };
1172
1173 struct module_info ip_mod_info = {
1174 IP_MOD_ID, IP_MOD_NAME, IP_MOD_MINPSZ, IP_MOD_MAXPSZ, IP_MOD_HIWAT,
1175 IP_MOD_LOWAT
1176 };
1177
1178 /*
1179 * Duplicate static symbols within a module confuses mdb; so we avoid the
1180 * problem by making the symbols here distinct from those in udp.c.
1181 */
1182
1183 /*
1184 * Entry points for IP as a device and as a module.
1185 * We have separate open functions for the /dev/ip and /dev/ip6 devices.
1186 */
1187 static struct qinit iprinitv4 = {
1188 ip_rput, NULL, ip_openv4, ip_close, NULL, &ip_mod_info
1189 };
1190
1191 struct qinit iprinitv6 = {
1192 ip_rput_v6, NULL, ip_openv6, ip_close, NULL, &ip_mod_info
1193 };
1194
1195 static struct qinit ipwinit = {
1196 ip_wput_nondata, ip_wsrv, NULL, NULL, NULL, &ip_mod_info
1197 };
1198
1199 static struct qinit iplrinit = {
1200 ip_lrput, NULL, ip_openv4, ip_close, NULL, &ip_mod_info
1201 };
1202
1203 static struct qinit iplwinit = {
1204 ip_lwput, NULL, NULL, NULL, NULL, &ip_mod_info
1205 };
1206
1207 /* For AF_INET aka /dev/ip */
1208 struct streamtab ipinfov4 = {
1209 &iprinitv4, &ipwinit, &iplrinit, &iplwinit
1210 };
1211
1212 /* For AF_INET6 aka /dev/ip6 */
1213 struct streamtab ipinfov6 = {
1214 &iprinitv6, &ipwinit, &iplrinit, &iplwinit
1215 };
1216
1217 #ifdef DEBUG
1218 boolean_t skip_sctp_cksum = B_FALSE;
1219 #endif
1220
1221 /*
1222 * Generate an ICMP fragmentation needed message.
1223 * When called from ip_output side a minimal ip_recv_attr_t needs to be
1224 * constructed by the caller.
1225 */
1226 void
1227 icmp_frag_needed(mblk_t *mp, int mtu, ip_recv_attr_t *ira)
1228 {
1229 icmph_t icmph;
1230 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
1231
1232 mp = icmp_pkt_err_ok(mp, ira);
1233 if (mp == NULL)
1234 return;
1235
1236 bzero(&icmph, sizeof (icmph_t));
1237 icmph.icmph_type = ICMP_DEST_UNREACHABLE;
1238 icmph.icmph_code = ICMP_FRAGMENTATION_NEEDED;
1239 icmph.icmph_du_mtu = htons((uint16_t)mtu);
1240 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutFragNeeded);
1241 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDestUnreachs);
1242
1243 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
1244 }
1245
1246 /*
1247 * icmp_inbound_v4 deals with ICMP messages that are handled by IP.
1248 * If the ICMP message is consumed by IP, i.e., it should not be delivered
1249 * to any IPPROTO_ICMP raw sockets, then it returns NULL.
1250 * Likewise, if the ICMP error is misformed (too short, etc), then it
1251 * returns NULL. The caller uses this to determine whether or not to send
1252 * to raw sockets.
1253 *
1254 * All error messages are passed to the matching transport stream.
1255 *
1256 * The following cases are handled by icmp_inbound:
1257 * 1) It needs to send a reply back and possibly delivering it
1258 * to the "interested" upper clients.
1259 * 2) Return the mblk so that the caller can pass it to the RAW socket clients.
1260 * 3) It needs to change some values in IP only.
1261 * 4) It needs to change some values in IP and upper layers e.g TCP
1262 * by delivering an error to the upper layers.
1263 *
1264 * We handle the above three cases in the context of IPsec in the
1265 * following way :
1266 *
1267 * 1) Send the reply back in the same way as the request came in.
1268 * If it came in encrypted, it goes out encrypted. If it came in
1269 * clear, it goes out in clear. Thus, this will prevent chosen
1270 * plain text attack.
1271 * 2) The client may or may not expect things to come in secure.
1272 * If it comes in secure, the policy constraints are checked
1273 * before delivering it to the upper layers. If it comes in
1274 * clear, ipsec_inbound_accept_clear will decide whether to
1275 * accept this in clear or not. In both the cases, if the returned
1276 * message (IP header + 8 bytes) that caused the icmp message has
1277 * AH/ESP headers, it is sent up to AH/ESP for validation before
1278 * sending up. If there are only 8 bytes of returned message, then
1279 * upper client will not be notified.
1280 * 3) Check with global policy to see whether it matches the constaints.
1281 * But this will be done only if icmp_accept_messages_in_clear is
1282 * zero.
1283 * 4) If we need to change both in IP and ULP, then the decision taken
1284 * while affecting the values in IP and while delivering up to TCP
1285 * should be the same.
1286 *
1287 * There are two cases.
1288 *
1289 * a) If we reject data at the IP layer (ipsec_check_global_policy()
1290 * failed), we will not deliver it to the ULP, even though they
1291 * are *willing* to accept in *clear*. This is fine as our global
1292 * disposition to icmp messages asks us reject the datagram.
1293 *
1294 * b) If we accept data at the IP layer (ipsec_check_global_policy()
1295 * succeeded or icmp_accept_messages_in_clear is 1), and not able
1296 * to deliver it to ULP (policy failed), it can lead to
1297 * consistency problems. The cases known at this time are
1298 * ICMP_DESTINATION_UNREACHABLE messages with following code
1299 * values :
1300 *
1301 * - ICMP_FRAGMENTATION_NEEDED : IP adapts to the new value
1302 * and Upper layer rejects. Then the communication will
1303 * come to a stop. This is solved by making similar decisions
1304 * at both levels. Currently, when we are unable to deliver
1305 * to the Upper Layer (due to policy failures) while IP has
1306 * adjusted dce_pmtu, the next outbound datagram would
1307 * generate a local ICMP_FRAGMENTATION_NEEDED message - which
1308 * will be with the right level of protection. Thus the right
1309 * value will be communicated even if we are not able to
1310 * communicate when we get from the wire initially. But this
1311 * assumes there would be at least one outbound datagram after
1312 * IP has adjusted its dce_pmtu value. To make things
1313 * simpler, we accept in clear after the validation of
1314 * AH/ESP headers.
1315 *
1316 * - Other ICMP ERRORS : We may not be able to deliver it to the
1317 * upper layer depending on the level of protection the upper
1318 * layer expects and the disposition in ipsec_inbound_accept_clear().
1319 * ipsec_inbound_accept_clear() decides whether a given ICMP error
1320 * should be accepted in clear when the Upper layer expects secure.
1321 * Thus the communication may get aborted by some bad ICMP
1322 * packets.
1323 */
1324 mblk_t *
1325 icmp_inbound_v4(mblk_t *mp, ip_recv_attr_t *ira)
1326 {
1327 icmph_t *icmph;
1328 ipha_t *ipha; /* Outer header */
1329 int ip_hdr_length; /* Outer header length */
1330 boolean_t interested;
1331 ipif_t *ipif;
1332 uint32_t ts;
1333 uint32_t *tsp;
1334 timestruc_t now;
1335 ill_t *ill = ira->ira_ill;
1336 ip_stack_t *ipst = ill->ill_ipst;
1337 zoneid_t zoneid = ira->ira_zoneid;
1338 int len_needed;
1339 mblk_t *mp_ret = NULL;
1340
1341 ipha = (ipha_t *)mp->b_rptr;
1342
1343 BUMP_MIB(&ipst->ips_icmp_mib, icmpInMsgs);
1344
1345 ip_hdr_length = ira->ira_ip_hdr_length;
1346 if ((mp->b_wptr - mp->b_rptr) < (ip_hdr_length + ICMPH_SIZE)) {
1347 if (ira->ira_pktlen < (ip_hdr_length + ICMPH_SIZE)) {
1348 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
1349 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
1350 freemsg(mp);
1351 return (NULL);
1352 }
1353 /* Last chance to get real. */
1354 ipha = ip_pullup(mp, ip_hdr_length + ICMPH_SIZE, ira);
1355 if (ipha == NULL) {
1356 BUMP_MIB(&ipst->ips_icmp_mib, icmpInErrors);
1357 freemsg(mp);
1358 return (NULL);
1359 }
1360 }
1361
1362 /* The IP header will always be a multiple of four bytes */
1363 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1364 ip2dbg(("icmp_inbound_v4: type %d code %d\n", icmph->icmph_type,
1365 icmph->icmph_code));
1366
1367 /*
1368 * We will set "interested" to "true" if we should pass a copy to
1369 * the transport or if we handle the packet locally.
1370 */
1371 interested = B_FALSE;
1372 switch (icmph->icmph_type) {
1373 case ICMP_ECHO_REPLY:
1374 BUMP_MIB(&ipst->ips_icmp_mib, icmpInEchoReps);
1375 break;
1376 case ICMP_DEST_UNREACHABLE:
1377 if (icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED)
1378 BUMP_MIB(&ipst->ips_icmp_mib, icmpInFragNeeded);
1379 interested = B_TRUE; /* Pass up to transport */
1380 BUMP_MIB(&ipst->ips_icmp_mib, icmpInDestUnreachs);
1381 break;
1382 case ICMP_SOURCE_QUENCH:
1383 interested = B_TRUE; /* Pass up to transport */
1384 BUMP_MIB(&ipst->ips_icmp_mib, icmpInSrcQuenchs);
1385 break;
1386 case ICMP_REDIRECT:
1387 if (!ipst->ips_ip_ignore_redirect)
1388 interested = B_TRUE;
1389 BUMP_MIB(&ipst->ips_icmp_mib, icmpInRedirects);
1390 break;
1391 case ICMP_ECHO_REQUEST:
1392 /*
1393 * Whether to respond to echo requests that come in as IP
1394 * broadcasts or as IP multicast is subject to debate
1395 * (what isn't?). We aim to please, you pick it.
1396 * Default is do it.
1397 */
1398 if (ira->ira_flags & IRAF_MULTICAST) {
1399 /* multicast: respond based on tunable */
1400 interested = ipst->ips_ip_g_resp_to_echo_mcast;
1401 } else if (ira->ira_flags & IRAF_BROADCAST) {
1402 /* broadcast: respond based on tunable */
1403 interested = ipst->ips_ip_g_resp_to_echo_bcast;
1404 } else {
1405 /* unicast: always respond */
1406 interested = B_TRUE;
1407 }
1408 BUMP_MIB(&ipst->ips_icmp_mib, icmpInEchos);
1409 if (!interested) {
1410 /* We never pass these to RAW sockets */
1411 freemsg(mp);
1412 return (NULL);
1413 }
1414
1415 /* Check db_ref to make sure we can modify the packet. */
1416 if (mp->b_datap->db_ref > 1) {
1417 mblk_t *mp1;
1418
1419 mp1 = copymsg(mp);
1420 freemsg(mp);
1421 if (!mp1) {
1422 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1423 return (NULL);
1424 }
1425 mp = mp1;
1426 ipha = (ipha_t *)mp->b_rptr;
1427 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1428 }
1429 icmph->icmph_type = ICMP_ECHO_REPLY;
1430 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutEchoReps);
1431 icmp_send_reply_v4(mp, ipha, icmph, ira);
1432 return (NULL);
1433
1434 case ICMP_ROUTER_ADVERTISEMENT:
1435 case ICMP_ROUTER_SOLICITATION:
1436 break;
1437 case ICMP_TIME_EXCEEDED:
1438 interested = B_TRUE; /* Pass up to transport */
1439 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimeExcds);
1440 break;
1441 case ICMP_PARAM_PROBLEM:
1442 interested = B_TRUE; /* Pass up to transport */
1443 BUMP_MIB(&ipst->ips_icmp_mib, icmpInParmProbs);
1444 break;
1445 case ICMP_TIME_STAMP_REQUEST:
1446 /* Response to Time Stamp Requests is local policy. */
1447 if (ipst->ips_ip_g_resp_to_timestamp) {
1448 if (ira->ira_flags & IRAF_MULTIBROADCAST)
1449 interested =
1450 ipst->ips_ip_g_resp_to_timestamp_bcast;
1451 else
1452 interested = B_TRUE;
1453 }
1454 if (!interested) {
1455 /* We never pass these to RAW sockets */
1456 freemsg(mp);
1457 return (NULL);
1458 }
1459
1460 /* Make sure we have enough of the packet */
1461 len_needed = ip_hdr_length + ICMPH_SIZE +
1462 3 * sizeof (uint32_t);
1463
1464 if (mp->b_wptr - mp->b_rptr < len_needed) {
1465 ipha = ip_pullup(mp, len_needed, ira);
1466 if (ipha == NULL) {
1467 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1468 ip_drop_input("ipIfStatsInDiscards - ip_pullup",
1469 mp, ill);
1470 freemsg(mp);
1471 return (NULL);
1472 }
1473 /* Refresh following the pullup. */
1474 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1475 }
1476 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimestamps);
1477 /* Check db_ref to make sure we can modify the packet. */
1478 if (mp->b_datap->db_ref > 1) {
1479 mblk_t *mp1;
1480
1481 mp1 = copymsg(mp);
1482 freemsg(mp);
1483 if (!mp1) {
1484 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1485 return (NULL);
1486 }
1487 mp = mp1;
1488 ipha = (ipha_t *)mp->b_rptr;
1489 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1490 }
1491 icmph->icmph_type = ICMP_TIME_STAMP_REPLY;
1492 tsp = (uint32_t *)&icmph[1];
1493 tsp++; /* Skip past 'originate time' */
1494 /* Compute # of milliseconds since midnight */
1495 gethrestime(&now);
1496 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
1497 NSEC2MSEC(now.tv_nsec);
1498 *tsp++ = htonl(ts); /* Lay in 'receive time' */
1499 *tsp++ = htonl(ts); /* Lay in 'send time' */
1500 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutTimestampReps);
1501 icmp_send_reply_v4(mp, ipha, icmph, ira);
1502 return (NULL);
1503
1504 case ICMP_TIME_STAMP_REPLY:
1505 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimestampReps);
1506 break;
1507 case ICMP_INFO_REQUEST:
1508 /* Per RFC 1122 3.2.2.7, ignore this. */
1509 case ICMP_INFO_REPLY:
1510 break;
1511 case ICMP_ADDRESS_MASK_REQUEST:
1512 if (ira->ira_flags & IRAF_MULTIBROADCAST) {
1513 interested =
1514 ipst->ips_ip_respond_to_address_mask_broadcast;
1515 } else {
1516 interested = B_TRUE;
1517 }
1518 if (!interested) {
1519 /* We never pass these to RAW sockets */
1520 freemsg(mp);
1521 return (NULL);
1522 }
1523 len_needed = ip_hdr_length + ICMPH_SIZE + IP_ADDR_LEN;
1524 if (mp->b_wptr - mp->b_rptr < len_needed) {
1525 ipha = ip_pullup(mp, len_needed, ira);
1526 if (ipha == NULL) {
1527 BUMP_MIB(ill->ill_ip_mib,
1528 ipIfStatsInTruncatedPkts);
1529 ip_drop_input("ipIfStatsInTruncatedPkts", mp,
1530 ill);
1531 freemsg(mp);
1532 return (NULL);
1533 }
1534 /* Refresh following the pullup. */
1535 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1536 }
1537 BUMP_MIB(&ipst->ips_icmp_mib, icmpInAddrMasks);
1538 /* Check db_ref to make sure we can modify the packet. */
1539 if (mp->b_datap->db_ref > 1) {
1540 mblk_t *mp1;
1541
1542 mp1 = copymsg(mp);
1543 freemsg(mp);
1544 if (!mp1) {
1545 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1546 return (NULL);
1547 }
1548 mp = mp1;
1549 ipha = (ipha_t *)mp->b_rptr;
1550 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1551 }
1552 /*
1553 * Need the ipif with the mask be the same as the source
1554 * address of the mask reply. For unicast we have a specific
1555 * ipif. For multicast/broadcast we only handle onlink
1556 * senders, and use the source address to pick an ipif.
1557 */
1558 ipif = ipif_lookup_addr(ipha->ipha_dst, ill, zoneid, ipst);
1559 if (ipif == NULL) {
1560 /* Broadcast or multicast */
1561 ipif = ipif_lookup_remote(ill, ipha->ipha_src, zoneid);
1562 if (ipif == NULL) {
1563 freemsg(mp);
1564 return (NULL);
1565 }
1566 }
1567 icmph->icmph_type = ICMP_ADDRESS_MASK_REPLY;
1568 bcopy(&ipif->ipif_net_mask, &icmph[1], IP_ADDR_LEN);
1569 ipif_refrele(ipif);
1570 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutAddrMaskReps);
1571 icmp_send_reply_v4(mp, ipha, icmph, ira);
1572 return (NULL);
1573
1574 case ICMP_ADDRESS_MASK_REPLY:
1575 BUMP_MIB(&ipst->ips_icmp_mib, icmpInAddrMaskReps);
1576 break;
1577 default:
1578 interested = B_TRUE; /* Pass up to transport */
1579 BUMP_MIB(&ipst->ips_icmp_mib, icmpInUnknowns);
1580 break;
1581 }
1582 /*
1583 * See if there is an ICMP client to avoid an extra copymsg/freemsg
1584 * if there isn't one.
1585 */
1586 if (ipst->ips_ipcl_proto_fanout_v4[IPPROTO_ICMP].connf_head != NULL) {
1587 /* If there is an ICMP client and we want one too, copy it. */
1588
1589 if (!interested) {
1590 /* Caller will deliver to RAW sockets */
1591 return (mp);
1592 }
1593 mp_ret = copymsg(mp);
1594 if (mp_ret == NULL) {
1595 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1596 ip_drop_input("ipIfStatsInDiscards - copymsg", mp, ill);
1597 }
1598 } else if (!interested) {
1599 /* Neither we nor raw sockets are interested. Drop packet now */
1600 freemsg(mp);
1601 return (NULL);
1602 }
1603
1604 /*
1605 * ICMP error or redirect packet. Make sure we have enough of
1606 * the header and that db_ref == 1 since we might end up modifying
1607 * the packet.
1608 */
1609 if (mp->b_cont != NULL) {
1610 if (ip_pullup(mp, -1, ira) == NULL) {
1611 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1612 ip_drop_input("ipIfStatsInDiscards - ip_pullup",
1613 mp, ill);
1614 freemsg(mp);
1615 return (mp_ret);
1616 }
1617 }
1618
1619 if (mp->b_datap->db_ref > 1) {
1620 mblk_t *mp1;
1621
1622 mp1 = copymsg(mp);
1623 if (mp1 == NULL) {
1624 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1625 ip_drop_input("ipIfStatsInDiscards - copymsg", mp, ill);
1626 freemsg(mp);
1627 return (mp_ret);
1628 }
1629 freemsg(mp);
1630 mp = mp1;
1631 }
1632
1633 /*
1634 * In case mp has changed, verify the message before any further
1635 * processes.
1636 */
1637 ipha = (ipha_t *)mp->b_rptr;
1638 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1639 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
1640 freemsg(mp);
1641 return (mp_ret);
1642 }
1643
1644 switch (icmph->icmph_type) {
1645 case ICMP_REDIRECT:
1646 icmp_redirect_v4(mp, ipha, icmph, ira);
1647 break;
1648 case ICMP_DEST_UNREACHABLE:
1649 if (icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED) {
1650 /* Update DCE and adjust MTU is icmp header if needed */
1651 icmp_inbound_too_big_v4(icmph, ira);
1652 }
1653 /* FALLTHROUGH */
1654 default:
1655 icmp_inbound_error_fanout_v4(mp, icmph, ira);
1656 break;
1657 }
1658 return (mp_ret);
1659 }
1660
1661 /*
1662 * Send an ICMP echo, timestamp or address mask reply.
1663 * The caller has already updated the payload part of the packet.
1664 * We handle the ICMP checksum, IP source address selection and feed
1665 * the packet into ip_output_simple.
1666 */
1667 static void
1668 icmp_send_reply_v4(mblk_t *mp, ipha_t *ipha, icmph_t *icmph,
1669 ip_recv_attr_t *ira)
1670 {
1671 uint_t ip_hdr_length = ira->ira_ip_hdr_length;
1672 ill_t *ill = ira->ira_ill;
1673 ip_stack_t *ipst = ill->ill_ipst;
1674 ip_xmit_attr_t ixas;
1675
1676 /* Send out an ICMP packet */
1677 icmph->icmph_checksum = 0;
1678 icmph->icmph_checksum = IP_CSUM(mp, ip_hdr_length, 0);
1679 /* Reset time to live. */
1680 ipha->ipha_ttl = ipst->ips_ip_def_ttl;
1681 {
1682 /* Swap source and destination addresses */
1683 ipaddr_t tmp;
1684
1685 tmp = ipha->ipha_src;
1686 ipha->ipha_src = ipha->ipha_dst;
1687 ipha->ipha_dst = tmp;
1688 }
1689 ipha->ipha_ident = 0;
1690 if (!IS_SIMPLE_IPH(ipha))
1691 icmp_options_update(ipha);
1692
1693 bzero(&ixas, sizeof (ixas));
1694 ixas.ixa_flags = IXAF_BASIC_SIMPLE_V4;
1695 ixas.ixa_zoneid = ira->ira_zoneid;
1696 ixas.ixa_cred = kcred;
1697 ixas.ixa_cpid = NOPID;
1698 ixas.ixa_tsl = ira->ira_tsl; /* Behave as a multi-level responder */
1699 ixas.ixa_ifindex = 0;
1700 ixas.ixa_ipst = ipst;
1701 ixas.ixa_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1702
1703 if (!(ira->ira_flags & IRAF_IPSEC_SECURE)) {
1704 /*
1705 * This packet should go out the same way as it
1706 * came in i.e in clear, independent of the IPsec policy
1707 * for transmitting packets.
1708 */
1709 ixas.ixa_flags |= IXAF_NO_IPSEC;
1710 } else {
1711 if (!ipsec_in_to_out(ira, &ixas, mp, ipha, NULL)) {
1712 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1713 /* Note: mp already consumed and ip_drop_packet done */
1714 return;
1715 }
1716 }
1717 if (ira->ira_flags & IRAF_MULTIBROADCAST) {
1718 /*
1719 * Not one or our addresses (IRE_LOCALs), thus we let
1720 * ip_output_simple pick the source.
1721 */
1722 ipha->ipha_src = INADDR_ANY;
1723 ixas.ixa_flags |= IXAF_SET_SOURCE;
1724 }
1725 /* Should we send with DF and use dce_pmtu? */
1726 if (ipst->ips_ipv4_icmp_return_pmtu) {
1727 ixas.ixa_flags |= IXAF_PMTU_DISCOVERY;
1728 ipha->ipha_fragment_offset_and_flags |= IPH_DF_HTONS;
1729 }
1730
1731 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutMsgs);
1732
1733 (void) ip_output_simple(mp, &ixas);
1734 ixa_cleanup(&ixas);
1735 }
1736
1737 /*
1738 * Verify the ICMP messages for either for ICMP error or redirect packet.
1739 * The caller should have fully pulled up the message. If it's a redirect
1740 * packet, only basic checks on IP header will be done; otherwise, verify
1741 * the packet by looking at the included ULP header.
1742 *
1743 * Called before icmp_inbound_error_fanout_v4 is called.
1744 */
1745 static boolean_t
1746 icmp_inbound_verify_v4(mblk_t *mp, icmph_t *icmph, ip_recv_attr_t *ira)
1747 {
1748 ill_t *ill = ira->ira_ill;
1749 int hdr_length;
1750 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
1751 conn_t *connp;
1752 ipha_t *ipha; /* Inner IP header */
1753
1754 ipha = (ipha_t *)&icmph[1];
1755 if ((uchar_t *)ipha + IP_SIMPLE_HDR_LENGTH > mp->b_wptr)
1756 goto truncated;
1757
1758 hdr_length = IPH_HDR_LENGTH(ipha);
1759
1760 if ((IPH_HDR_VERSION(ipha) != IPV4_VERSION))
1761 goto discard_pkt;
1762
1763 if (hdr_length < sizeof (ipha_t))
1764 goto truncated;
1765
1766 if ((uchar_t *)ipha + hdr_length > mp->b_wptr)
1767 goto truncated;
1768
1769 /*
1770 * Stop here for ICMP_REDIRECT.
1771 */
1772 if (icmph->icmph_type == ICMP_REDIRECT)
1773 return (B_TRUE);
1774
1775 /*
1776 * ICMP errors only.
1777 */
1778 switch (ipha->ipha_protocol) {
1779 case IPPROTO_UDP:
1780 /*
1781 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1782 * transport header.
1783 */
1784 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1785 mp->b_wptr)
1786 goto truncated;
1787 break;
1788 case IPPROTO_TCP: {
1789 tcpha_t *tcpha;
1790
1791 /*
1792 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1793 * transport header.
1794 */
1795 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1796 mp->b_wptr)
1797 goto truncated;
1798
1799 tcpha = (tcpha_t *)((uchar_t *)ipha + hdr_length);
1800 connp = ipcl_tcp_lookup_reversed_ipv4(ipha, tcpha, TCPS_LISTEN,
1801 ipst);
1802 if (connp == NULL)
1803 goto discard_pkt;
1804
1805 if ((connp->conn_verifyicmp != NULL) &&
1806 !connp->conn_verifyicmp(connp, tcpha, icmph, NULL, ira)) {
1807 CONN_DEC_REF(connp);
1808 goto discard_pkt;
1809 }
1810 CONN_DEC_REF(connp);
1811 break;
1812 }
1813 case IPPROTO_SCTP:
1814 /*
1815 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1816 * transport header.
1817 */
1818 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1819 mp->b_wptr)
1820 goto truncated;
1821 break;
1822 case IPPROTO_ESP:
1823 case IPPROTO_AH:
1824 break;
1825 case IPPROTO_ENCAP:
1826 if ((uchar_t *)ipha + hdr_length + sizeof (ipha_t) >
1827 mp->b_wptr)
1828 goto truncated;
1829 break;
1830 default:
1831 break;
1832 }
1833
1834 return (B_TRUE);
1835
1836 discard_pkt:
1837 /* Bogus ICMP error. */
1838 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1839 return (B_FALSE);
1840
1841 truncated:
1842 /* We pulled up everthing already. Must be truncated */
1843 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
1844 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
1845 return (B_FALSE);
1846 }
1847
1848 /* Table from RFC 1191 */
1849 static int icmp_frag_size_table[] =
1850 { 32000, 17914, 8166, 4352, 2002, 1496, 1006, 508, 296, 68 };
1851
1852 /*
1853 * Process received ICMP Packet too big.
1854 * Just handles the DCE create/update, including using the above table of
1855 * PMTU guesses. The caller is responsible for validating the packet before
1856 * passing it in and also to fanout the ICMP error to any matching transport
1857 * conns. Assumes the message has been fully pulled up and verified.
1858 *
1859 * Before getting here, the caller has called icmp_inbound_verify_v4()
1860 * that should have verified with ULP to prevent undoing the changes we're
1861 * going to make to DCE. For example, TCP might have verified that the packet
1862 * which generated error is in the send window.
1863 *
1864 * In some cases modified this MTU in the ICMP header packet; the caller
1865 * should pass to the matching ULP after this returns.
1866 */
1867 static void
1868 icmp_inbound_too_big_v4(icmph_t *icmph, ip_recv_attr_t *ira)
1869 {
1870 dce_t *dce;
1871 int old_mtu;
1872 int mtu, orig_mtu;
1873 ipaddr_t dst;
1874 boolean_t disable_pmtud;
1875 ill_t *ill = ira->ira_ill;
1876 ip_stack_t *ipst = ill->ill_ipst;
1877 uint_t hdr_length;
1878 ipha_t *ipha;
1879
1880 /* Caller already pulled up everything. */
1881 ipha = (ipha_t *)&icmph[1];
1882 ASSERT(icmph->icmph_type == ICMP_DEST_UNREACHABLE &&
1883 icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED);
1884 ASSERT(ill != NULL);
1885
1886 hdr_length = IPH_HDR_LENGTH(ipha);
1887
1888 /*
1889 * We handle path MTU for source routed packets since the DCE
1890 * is looked up using the final destination.
1891 */
1892 dst = ip_get_dst(ipha);
1893
1894 dce = dce_lookup_and_add_v4(dst, ipst);
1895 if (dce == NULL) {
1896 /* Couldn't add a unique one - ENOMEM */
1897 ip1dbg(("icmp_inbound_too_big_v4: no dce for 0x%x\n",
1898 ntohl(dst)));
1899 return;
1900 }
1901
1902 /* Check for MTU discovery advice as described in RFC 1191 */
1903 mtu = ntohs(icmph->icmph_du_mtu);
1904 orig_mtu = mtu;
1905 disable_pmtud = B_FALSE;
1906
1907 mutex_enter(&dce->dce_lock);
1908 if (dce->dce_flags & DCEF_PMTU)
1909 old_mtu = dce->dce_pmtu;
1910 else
1911 old_mtu = ill->ill_mtu;
1912
1913 if (icmph->icmph_du_zero != 0 || mtu < ipst->ips_ip_pmtu_min) {
1914 uint32_t length;
1915 int i;
1916
1917 /*
1918 * Use the table from RFC 1191 to figure out
1919 * the next "plateau" based on the length in
1920 * the original IP packet.
1921 */
1922 length = ntohs(ipha->ipha_length);
1923 DTRACE_PROBE2(ip4__pmtu__guess, dce_t *, dce,
1924 uint32_t, length);
1925 if (old_mtu <= length &&
1926 old_mtu >= length - hdr_length) {
1927 /*
1928 * Handle broken BSD 4.2 systems that
1929 * return the wrong ipha_length in ICMP
1930 * errors.
1931 */
1932 ip1dbg(("Wrong mtu: sent %d, dce %d\n",
1933 length, old_mtu));
1934 length -= hdr_length;
1935 }
1936 for (i = 0; i < A_CNT(icmp_frag_size_table); i++) {
1937 if (length > icmp_frag_size_table[i])
1938 break;
1939 }
1940 if (i == A_CNT(icmp_frag_size_table)) {
1941 /* Smaller than IP_MIN_MTU! */
1942 ip1dbg(("Too big for packet size %d\n",
1943 length));
1944 disable_pmtud = B_TRUE;
1945 mtu = ipst->ips_ip_pmtu_min;
1946 } else {
1947 mtu = icmp_frag_size_table[i];
1948 ip1dbg(("Calculated mtu %d, packet size %d, "
1949 "before %d\n", mtu, length, old_mtu));
1950 if (mtu < ipst->ips_ip_pmtu_min) {
1951 mtu = ipst->ips_ip_pmtu_min;
1952 disable_pmtud = B_TRUE;
1953 }
1954 }
1955 }
1956 if (disable_pmtud)
1957 dce->dce_flags |= DCEF_TOO_SMALL_PMTU;
1958 else
1959 dce->dce_flags &= ~DCEF_TOO_SMALL_PMTU;
1960
1961 dce->dce_pmtu = MIN(old_mtu, mtu);
1962 /* Prepare to send the new max frag size for the ULP. */
1963 icmph->icmph_du_zero = 0;
1964 icmph->icmph_du_mtu = htons((uint16_t)dce->dce_pmtu);
1965 DTRACE_PROBE4(ip4__pmtu__change, icmph_t *, icmph, dce_t *,
1966 dce, int, orig_mtu, int, mtu);
1967
1968 /* We now have a PMTU for sure */
1969 dce->dce_flags |= DCEF_PMTU;
1970 dce->dce_last_change_time = TICK_TO_SEC(ddi_get_lbolt64());
1971 mutex_exit(&dce->dce_lock);
1972 /*
1973 * After dropping the lock the new value is visible to everyone.
1974 * Then we bump the generation number so any cached values reinspect
1975 * the dce_t.
1976 */
1977 dce_increment_generation(dce);
1978 dce_refrele(dce);
1979 }
1980
1981 /*
1982 * If the packet in error is Self-Encapsulated, icmp_inbound_error_fanout_v4
1983 * calls this function.
1984 */
1985 static mblk_t *
1986 icmp_inbound_self_encap_error_v4(mblk_t *mp, ipha_t *ipha, ipha_t *in_ipha)
1987 {
1988 int length;
1989
1990 ASSERT(mp->b_datap->db_type == M_DATA);
1991
1992 /* icmp_inbound_v4 has already pulled up the whole error packet */
1993 ASSERT(mp->b_cont == NULL);
1994
1995 /*
1996 * The length that we want to overlay is the inner header
1997 * and what follows it.
1998 */
1999 length = msgdsize(mp) - ((uchar_t *)in_ipha - mp->b_rptr);
2000
2001 /*
2002 * Overlay the inner header and whatever follows it over the
2003 * outer header.
2004 */
2005 bcopy((uchar_t *)in_ipha, (uchar_t *)ipha, length);
2006
2007 /* Adjust for what we removed */
2008 mp->b_wptr -= (uchar_t *)in_ipha - (uchar_t *)ipha;
2009 return (mp);
2010 }
2011
2012 /*
2013 * Try to pass the ICMP message upstream in case the ULP cares.
2014 *
2015 * If the packet that caused the ICMP error is secure, we send
2016 * it to AH/ESP to make sure that the attached packet has a
2017 * valid association. ipha in the code below points to the
2018 * IP header of the packet that caused the error.
2019 *
2020 * For IPsec cases, we let the next-layer-up (which has access to
2021 * cached policy on the conn_t, or can query the SPD directly)
2022 * subtract out any IPsec overhead if they must. We therefore make no
2023 * adjustments here for IPsec overhead.
2024 *
2025 * IFN could have been generated locally or by some router.
2026 *
2027 * LOCAL : ire_send_wire (before calling ipsec_out_process) can call
2028 * icmp_frag_needed/icmp_pkt2big_v6 to generated a local IFN.
2029 * This happens because IP adjusted its value of MTU on an
2030 * earlier IFN message and could not tell the upper layer,
2031 * the new adjusted value of MTU e.g. Packet was encrypted
2032 * or there was not enough information to fanout to upper
2033 * layers. Thus on the next outbound datagram, ire_send_wire
2034 * generates the IFN, where IPsec processing has *not* been
2035 * done.
2036 *
2037 * Note that we retain ixa_fragsize across IPsec thus once
2038 * we have picking ixa_fragsize and entered ipsec_out_process we do
2039 * no change the fragsize even if the path MTU changes before
2040 * we reach ip_output_post_ipsec.
2041 *
2042 * In the local case, IRAF_LOOPBACK will be set indicating
2043 * that IFN was generated locally.
2044 *
2045 * ROUTER : IFN could be secure or non-secure.
2046 *
2047 * * SECURE : We use the IPSEC_IN to fanout to AH/ESP if the
2048 * packet in error has AH/ESP headers to validate the AH/ESP
2049 * headers. AH/ESP will verify whether there is a valid SA or
2050 * not and send it back. We will fanout again if we have more
2051 * data in the packet.
2052 *
2053 * If the packet in error does not have AH/ESP, we handle it
2054 * like any other case.
2055 *
2056 * * NON_SECURE : If the packet in error has AH/ESP headers, we send it
2057 * up to AH/ESP for validation. AH/ESP will verify whether there is a
2058 * valid SA or not and send it back. We will fanout again if
2059 * we have more data in the packet.
2060 *
2061 * If the packet in error does not have AH/ESP, we handle it
2062 * like any other case.
2063 *
2064 * The caller must have called icmp_inbound_verify_v4.
2065 */
2066 static void
2067 icmp_inbound_error_fanout_v4(mblk_t *mp, icmph_t *icmph, ip_recv_attr_t *ira)
2068 {
2069 uint16_t *up; /* Pointer to ports in ULP header */
2070 uint32_t ports; /* reversed ports for fanout */
2071 ipha_t ripha; /* With reversed addresses */
2072 ipha_t *ipha; /* Inner IP header */
2073 uint_t hdr_length; /* Inner IP header length */
2074 tcpha_t *tcpha;
2075 conn_t *connp;
2076 ill_t *ill = ira->ira_ill;
2077 ip_stack_t *ipst = ill->ill_ipst;
2078 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
2079 ill_t *rill = ira->ira_rill;
2080
2081 /* Caller already pulled up everything. */
2082 ipha = (ipha_t *)&icmph[1];
2083 ASSERT((uchar_t *)&ipha[1] <= mp->b_wptr);
2084 ASSERT(mp->b_cont == NULL);
2085
2086 hdr_length = IPH_HDR_LENGTH(ipha);
2087 ira->ira_protocol = ipha->ipha_protocol;
2088
2089 /*
2090 * We need a separate IP header with the source and destination
2091 * addresses reversed to do fanout/classification because the ipha in
2092 * the ICMP error is in the form we sent it out.
2093 */
2094 ripha.ipha_src = ipha->ipha_dst;
2095 ripha.ipha_dst = ipha->ipha_src;
2096 ripha.ipha_protocol = ipha->ipha_protocol;
2097 ripha.ipha_version_and_hdr_length = ipha->ipha_version_and_hdr_length;
2098
2099 ip2dbg(("icmp_inbound_error_v4: proto %d %x to %x: %d/%d\n",
2100 ripha.ipha_protocol, ntohl(ipha->ipha_src),
2101 ntohl(ipha->ipha_dst),
2102 icmph->icmph_type, icmph->icmph_code));
2103
2104 switch (ipha->ipha_protocol) {
2105 case IPPROTO_UDP:
2106 up = (uint16_t *)((uchar_t *)ipha + hdr_length);
2107
2108 /* Attempt to find a client stream based on port. */
2109 ip2dbg(("icmp_inbound_error_v4: UDP ports %d to %d\n",
2110 ntohs(up[0]), ntohs(up[1])));
2111
2112 /* Note that we send error to all matches. */
2113 ira->ira_flags |= IRAF_ICMP_ERROR;
2114 ip_fanout_udp_multi_v4(mp, &ripha, up[0], up[1], ira);
2115 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2116 return;
2117
2118 case IPPROTO_TCP:
2119 /*
2120 * Find a TCP client stream for this packet.
2121 * Note that we do a reverse lookup since the header is
2122 * in the form we sent it out.
2123 */
2124 tcpha = (tcpha_t *)((uchar_t *)ipha + hdr_length);
2125 connp = ipcl_tcp_lookup_reversed_ipv4(ipha, tcpha, TCPS_LISTEN,
2126 ipst);
2127 if (connp == NULL)
2128 goto discard_pkt;
2129
2130 if (CONN_INBOUND_POLICY_PRESENT(connp, ipss) ||
2131 (ira->ira_flags & IRAF_IPSEC_SECURE)) {
2132 mp = ipsec_check_inbound_policy(mp, connp,
2133 ipha, NULL, ira);
2134 if (mp == NULL) {
2135 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2136 /* Note that mp is NULL */
2137 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2138 CONN_DEC_REF(connp);
2139 return;
2140 }
2141 }
2142
2143 ira->ira_flags |= IRAF_ICMP_ERROR;
2144 ira->ira_ill = ira->ira_rill = NULL;
2145 if (IPCL_IS_TCP(connp)) {
2146 SQUEUE_ENTER_ONE(connp->conn_sqp, mp,
2147 connp->conn_recvicmp, connp, ira, SQ_FILL,
2148 SQTAG_TCP_INPUT_ICMP_ERR);
2149 } else {
2150 /* Not TCP; must be SOCK_RAW, IPPROTO_TCP */
2151 (connp->conn_recv)(connp, mp, NULL, ira);
2152 CONN_DEC_REF(connp);
2153 }
2154 ira->ira_ill = ill;
2155 ira->ira_rill = rill;
2156 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2157 return;
2158
2159 case IPPROTO_SCTP:
2160 up = (uint16_t *)((uchar_t *)ipha + hdr_length);
2161 /* Find a SCTP client stream for this packet. */
2162 ((uint16_t *)&ports)[0] = up[1];
2163 ((uint16_t *)&ports)[1] = up[0];
2164
2165 ira->ira_flags |= IRAF_ICMP_ERROR;
2166 ip_fanout_sctp(mp, &ripha, NULL, ports, ira);
2167 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2168 return;
2169
2170 case IPPROTO_ESP:
2171 case IPPROTO_AH:
2172 if (!ipsec_loaded(ipss)) {
2173 ip_proto_not_sup(mp, ira);
2174 return;
2175 }
2176
2177 if (ipha->ipha_protocol == IPPROTO_ESP)
2178 mp = ipsecesp_icmp_error(mp, ira);
2179 else
2180 mp = ipsecah_icmp_error(mp, ira);
2181 if (mp == NULL)
2182 return;
2183
2184 /* Just in case ipsec didn't preserve the NULL b_cont */
2185 if (mp->b_cont != NULL) {
2186 if (!pullupmsg(mp, -1))
2187 goto discard_pkt;
2188 }
2189
2190 /*
2191 * Note that ira_pktlen and ira_ip_hdr_length are no longer
2192 * correct, but we don't use them any more here.
2193 *
2194 * If succesful, the mp has been modified to not include
2195 * the ESP/AH header so we can fanout to the ULP's icmp
2196 * error handler.
2197 */
2198 if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH)
2199 goto truncated;
2200
2201 /* Verify the modified message before any further processes. */
2202 ipha = (ipha_t *)mp->b_rptr;
2203 hdr_length = IPH_HDR_LENGTH(ipha);
2204 icmph = (icmph_t *)&mp->b_rptr[hdr_length];
2205 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
2206 freemsg(mp);
2207 return;
2208 }
2209
2210 icmp_inbound_error_fanout_v4(mp, icmph, ira);
2211 return;
2212
2213 case IPPROTO_ENCAP: {
2214 /* Look for self-encapsulated packets that caused an error */
2215 ipha_t *in_ipha;
2216
2217 /*
2218 * Caller has verified that length has to be
2219 * at least the size of IP header.
2220 */
2221 ASSERT(hdr_length >= sizeof (ipha_t));
2222 /*
2223 * Check the sanity of the inner IP header like
2224 * we did for the outer header.
2225 */
2226 in_ipha = (ipha_t *)((uchar_t *)ipha + hdr_length);
2227 if ((IPH_HDR_VERSION(in_ipha) != IPV4_VERSION)) {
2228 goto discard_pkt;
2229 }
2230 if (IPH_HDR_LENGTH(in_ipha) < sizeof (ipha_t)) {
2231 goto discard_pkt;
2232 }
2233 /* Check for Self-encapsulated tunnels */
2234 if (in_ipha->ipha_src == ipha->ipha_src &&
2235 in_ipha->ipha_dst == ipha->ipha_dst) {
2236
2237 mp = icmp_inbound_self_encap_error_v4(mp, ipha,
2238 in_ipha);
2239 if (mp == NULL)
2240 goto discard_pkt;
2241
2242 /*
2243 * Just in case self_encap didn't preserve the NULL
2244 * b_cont
2245 */
2246 if (mp->b_cont != NULL) {
2247 if (!pullupmsg(mp, -1))
2248 goto discard_pkt;
2249 }
2250 /*
2251 * Note that ira_pktlen and ira_ip_hdr_length are no
2252 * longer correct, but we don't use them any more here.
2253 */
2254 if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH)
2255 goto truncated;
2256
2257 /*
2258 * Verify the modified message before any further
2259 * processes.
2260 */
2261 ipha = (ipha_t *)mp->b_rptr;
2262 hdr_length = IPH_HDR_LENGTH(ipha);
2263 icmph = (icmph_t *)&mp->b_rptr[hdr_length];
2264 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
2265 freemsg(mp);
2266 return;
2267 }
2268
2269 /*
2270 * The packet in error is self-encapsualted.
2271 * And we are finding it further encapsulated
2272 * which we could not have possibly generated.
2273 */
2274 if (ipha->ipha_protocol == IPPROTO_ENCAP) {
2275 goto discard_pkt;
2276 }
2277 icmp_inbound_error_fanout_v4(mp, icmph, ira);
2278 return;
2279 }
2280 /* No self-encapsulated */
2281 }
2282 /* FALLTHROUGH */
2283 case IPPROTO_IPV6:
2284 if ((connp = ipcl_iptun_classify_v4(&ripha.ipha_src,
2285 &ripha.ipha_dst, ipst)) != NULL) {
2286 ira->ira_flags |= IRAF_ICMP_ERROR;
2287 connp->conn_recvicmp(connp, mp, NULL, ira);
2288 CONN_DEC_REF(connp);
2289 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2290 return;
2291 }
2292 /*
2293 * No IP tunnel is interested, fallthrough and see
2294 * if a raw socket will want it.
2295 */
2296 /* FALLTHROUGH */
2297 default:
2298 ira->ira_flags |= IRAF_ICMP_ERROR;
2299 ip_fanout_proto_v4(mp, &ripha, ira);
2300 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2301 return;
2302 }
2303 /* NOTREACHED */
2304 discard_pkt:
2305 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2306 ip1dbg(("icmp_inbound_error_fanout_v4: drop pkt\n"));
2307 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2308 freemsg(mp);
2309 return;
2310
2311 truncated:
2312 /* We pulled up everthing already. Must be truncated */
2313 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
2314 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
2315 freemsg(mp);
2316 }
2317
2318 /*
2319 * Common IP options parser.
2320 *
2321 * Setup routine: fill in *optp with options-parsing state, then
2322 * tail-call ipoptp_next to return the first option.
2323 */
2324 uint8_t
2325 ipoptp_first(ipoptp_t *optp, ipha_t *ipha)
2326 {
2327 uint32_t totallen; /* total length of all options */
2328
2329 totallen = ipha->ipha_version_and_hdr_length -
2330 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
2331 totallen <<= 2;
2332 optp->ipoptp_next = (uint8_t *)(&ipha[1]);
2333 optp->ipoptp_end = optp->ipoptp_next + totallen;
2334 optp->ipoptp_flags = 0;
2335 return (ipoptp_next(optp));
2336 }
2337
2338 /* Like above but without an ipha_t */
2339 uint8_t
2340 ipoptp_first2(ipoptp_t *optp, uint32_t totallen, uint8_t *opt)
2341 {
2342 optp->ipoptp_next = opt;
2343 optp->ipoptp_end = optp->ipoptp_next + totallen;
2344 optp->ipoptp_flags = 0;
2345 return (ipoptp_next(optp));
2346 }
2347
2348 /*
2349 * Common IP options parser: extract next option.
2350 */
2351 uint8_t
2352 ipoptp_next(ipoptp_t *optp)
2353 {
2354 uint8_t *end = optp->ipoptp_end;
2355 uint8_t *cur = optp->ipoptp_next;
2356 uint8_t opt, len, pointer;
2357
2358 /*
2359 * If cur > end already, then the ipoptp_end or ipoptp_next pointer
2360 * has been corrupted.
2361 */
2362 ASSERT(cur <= end);
2363
2364 if (cur == end)
2365 return (IPOPT_EOL);
2366
2367 opt = cur[IPOPT_OPTVAL];
2368
2369 /*
2370 * Skip any NOP options.
2371 */
2372 while (opt == IPOPT_NOP) {
2373 cur++;
2374 if (cur == end)
2375 return (IPOPT_EOL);
2376 opt = cur[IPOPT_OPTVAL];
2377 }
2378
2379 if (opt == IPOPT_EOL)
2380 return (IPOPT_EOL);
2381
2382 /*
2383 * Option requiring a length.
2384 */
2385 if ((cur + 1) >= end) {
2386 optp->ipoptp_flags |= IPOPTP_ERROR;
2387 return (IPOPT_EOL);
2388 }
2389 len = cur[IPOPT_OLEN];
2390 if (len < 2) {
2391 optp->ipoptp_flags |= IPOPTP_ERROR;
2392 return (IPOPT_EOL);
2393 }
2394 optp->ipoptp_cur = cur;
2395 optp->ipoptp_len = len;
2396 optp->ipoptp_next = cur + len;
2397 if (cur + len > end) {
2398 optp->ipoptp_flags |= IPOPTP_ERROR;
2399 return (IPOPT_EOL);
2400 }
2401
2402 /*
2403 * For the options which require a pointer field, make sure
2404 * its there, and make sure it points to either something
2405 * inside this option, or the end of the option.
2406 */
2407 switch (opt) {
2408 case IPOPT_RR:
2409 case IPOPT_TS:
2410 case IPOPT_LSRR:
2411 case IPOPT_SSRR:
2412 if (len <= IPOPT_OFFSET) {
2413 optp->ipoptp_flags |= IPOPTP_ERROR;
2414 return (opt);
2415 }
2416 pointer = cur[IPOPT_OFFSET];
2417 if (pointer - 1 > len) {
2418 optp->ipoptp_flags |= IPOPTP_ERROR;
2419 return (opt);
2420 }
2421 break;
2422 }
2423
2424 /*
2425 * Sanity check the pointer field based on the type of the
2426 * option.
2427 */
2428 switch (opt) {
2429 case IPOPT_RR:
2430 case IPOPT_SSRR:
2431 case IPOPT_LSRR:
2432 if (pointer < IPOPT_MINOFF_SR)
2433 optp->ipoptp_flags |= IPOPTP_ERROR;
2434 break;
2435 case IPOPT_TS:
2436 if (pointer < IPOPT_MINOFF_IT)
2437 optp->ipoptp_flags |= IPOPTP_ERROR;
2438 /*
2439 * Note that the Internet Timestamp option also
2440 * contains two four bit fields (the Overflow field,
2441 * and the Flag field), which follow the pointer
2442 * field. We don't need to check that these fields
2443 * fall within the length of the option because this
2444 * was implicitely done above. We've checked that the
2445 * pointer value is at least IPOPT_MINOFF_IT, and that
2446 * it falls within the option. Since IPOPT_MINOFF_IT >
2447 * IPOPT_POS_OV_FLG, we don't need the explicit check.
2448 */
2449 ASSERT(len > IPOPT_POS_OV_FLG);
2450 break;
2451 }
2452
2453 return (opt);
2454 }
2455
2456 /*
2457 * Use the outgoing IP header to create an IP_OPTIONS option the way
2458 * it was passed down from the application.
2459 *
2460 * This is compatible with BSD in that it returns
2461 * the reverse source route with the final destination
2462 * as the last entry. The first 4 bytes of the option
2463 * will contain the final destination.
2464 */
2465 int
2466 ip_opt_get_user(conn_t *connp, uchar_t *buf)
2467 {
2468 ipoptp_t opts;
2469 uchar_t *opt;
2470 uint8_t optval;
2471 uint8_t optlen;
2472 uint32_t len = 0;
2473 uchar_t *buf1 = buf;
2474 uint32_t totallen;
2475 ipaddr_t dst;
2476 ip_pkt_t *ipp = &connp->conn_xmit_ipp;
2477
2478 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
2479 return (0);
2480
2481 totallen = ipp->ipp_ipv4_options_len;
2482 if (totallen & 0x3)
2483 return (0);
2484
2485 buf += IP_ADDR_LEN; /* Leave room for final destination */
2486 len += IP_ADDR_LEN;
2487 bzero(buf1, IP_ADDR_LEN);
2488
2489 dst = connp->conn_faddr_v4;
2490
2491 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
2492 optval != IPOPT_EOL;
2493 optval = ipoptp_next(&opts)) {
2494 int off;
2495
2496 opt = opts.ipoptp_cur;
2497 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
2498 break;
2499 }
2500 optlen = opts.ipoptp_len;
2501
2502 switch (optval) {
2503 case IPOPT_SSRR:
2504 case IPOPT_LSRR:
2505
2506 /*
2507 * Insert destination as the first entry in the source
2508 * route and move down the entries on step.
2509 * The last entry gets placed at buf1.
2510 */
2511 buf[IPOPT_OPTVAL] = optval;
2512 buf[IPOPT_OLEN] = optlen;
2513 buf[IPOPT_OFFSET] = optlen;
2514
2515 off = optlen - IP_ADDR_LEN;
2516 if (off < 0) {
2517 /* No entries in source route */
2518 break;
2519 }
2520 /* Last entry in source route if not already set */
2521 if (dst == INADDR_ANY)
2522 bcopy(opt + off, buf1, IP_ADDR_LEN);
2523 off -= IP_ADDR_LEN;
2524
2525 while (off > 0) {
2526 bcopy(opt + off,
2527 buf + off + IP_ADDR_LEN,
2528 IP_ADDR_LEN);
2529 off -= IP_ADDR_LEN;
2530 }
2531 /* ipha_dst into first slot */
2532 bcopy(&dst, buf + off + IP_ADDR_LEN,
2533 IP_ADDR_LEN);
2534 buf += optlen;
2535 len += optlen;
2536 break;
2537
2538 default:
2539 bcopy(opt, buf, optlen);
2540 buf += optlen;
2541 len += optlen;
2542 break;
2543 }
2544 }
2545 done:
2546 /* Pad the resulting options */
2547 while (len & 0x3) {
2548 *buf++ = IPOPT_EOL;
2549 len++;
2550 }
2551 return (len);
2552 }
2553
2554 /*
2555 * Update any record route or timestamp options to include this host.
2556 * Reverse any source route option.
2557 * This routine assumes that the options are well formed i.e. that they
2558 * have already been checked.
2559 */
2560 static void
2561 icmp_options_update(ipha_t *ipha)
2562 {
2563 ipoptp_t opts;
2564 uchar_t *opt;
2565 uint8_t optval;
2566 ipaddr_t src; /* Our local address */
2567 ipaddr_t dst;
2568
2569 ip2dbg(("icmp_options_update\n"));
2570 src = ipha->ipha_src;
2571 dst = ipha->ipha_dst;
2572
2573 for (optval = ipoptp_first(&opts, ipha);
2574 optval != IPOPT_EOL;
2575 optval = ipoptp_next(&opts)) {
2576 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
2577 opt = opts.ipoptp_cur;
2578 ip2dbg(("icmp_options_update: opt %d, len %d\n",
2579 optval, opts.ipoptp_len));
2580 switch (optval) {
2581 int off1, off2;
2582 case IPOPT_SSRR:
2583 case IPOPT_LSRR:
2584 /*
2585 * Reverse the source route. The first entry
2586 * should be the next to last one in the current
2587 * source route (the last entry is our address).
2588 * The last entry should be the final destination.
2589 */
2590 off1 = IPOPT_MINOFF_SR - 1;
2591 off2 = opt[IPOPT_OFFSET] - IP_ADDR_LEN - 1;
2592 if (off2 < 0) {
2593 /* No entries in source route */
2594 ip1dbg((
2595 "icmp_options_update: bad src route\n"));
2596 break;
2597 }
2598 bcopy((char *)opt + off2, &dst, IP_ADDR_LEN);
2599 bcopy(&ipha->ipha_dst, (char *)opt + off2, IP_ADDR_LEN);
2600 bcopy(&dst, &ipha->ipha_dst, IP_ADDR_LEN);
2601 off2 -= IP_ADDR_LEN;
2602
2603 while (off1 < off2) {
2604 bcopy((char *)opt + off1, &src, IP_ADDR_LEN);
2605 bcopy((char *)opt + off2, (char *)opt + off1,
2606 IP_ADDR_LEN);
2607 bcopy(&src, (char *)opt + off2, IP_ADDR_LEN);
2608 off1 += IP_ADDR_LEN;
2609 off2 -= IP_ADDR_LEN;
2610 }
2611 opt[IPOPT_OFFSET] = IPOPT_MINOFF_SR;
2612 break;
2613 }
2614 }
2615 }
2616
2617 /*
2618 * Process received ICMP Redirect messages.
2619 * Assumes the caller has verified that the headers are in the pulled up mblk.
2620 * Consumes mp.
2621 */
2622 static void
2623 icmp_redirect_v4(mblk_t *mp, ipha_t *ipha, icmph_t *icmph, ip_recv_attr_t *ira)
2624 {
2625 ire_t *ire, *nire;
2626 ire_t *prev_ire;
2627 ipaddr_t src, dst, gateway;
2628 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2629 ipha_t *inner_ipha; /* Inner IP header */
2630
2631 /* Caller already pulled up everything. */
2632 inner_ipha = (ipha_t *)&icmph[1];
2633 src = ipha->ipha_src;
2634 dst = inner_ipha->ipha_dst;
2635 gateway = icmph->icmph_rd_gateway;
2636 /* Make sure the new gateway is reachable somehow. */
2637 ire = ire_ftable_lookup_v4(gateway, 0, 0, IRE_ONLINK, NULL,
2638 ALL_ZONES, NULL, MATCH_IRE_TYPE, 0, ipst, NULL);
2639 /*
2640 * Make sure we had a route for the dest in question and that
2641 * that route was pointing to the old gateway (the source of the
2642 * redirect packet.)
2643 * We do longest match and then compare ire_gateway_addr below.
2644 */
2645 prev_ire = ire_ftable_lookup_v4(dst, 0, 0, 0, NULL, ALL_ZONES,
2646 NULL, MATCH_IRE_DSTONLY, 0, ipst, NULL);
2647 /*
2648 * Check that
2649 * the redirect was not from ourselves
2650 * the new gateway and the old gateway are directly reachable
2651 */
2652 if (prev_ire == NULL || ire == NULL ||
2653 (prev_ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK)) ||
2654 (prev_ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) ||
2655 !(ire->ire_type & IRE_IF_ALL) ||
2656 prev_ire->ire_gateway_addr != src) {
2657 BUMP_MIB(&ipst->ips_icmp_mib, icmpInBadRedirects);
2658 ip_drop_input("icmpInBadRedirects - ire", mp, ira->ira_ill);
2659 freemsg(mp);
2660 if (ire != NULL)
2661 ire_refrele(ire);
2662 if (prev_ire != NULL)
2663 ire_refrele(prev_ire);
2664 return;
2665 }
2666
2667 ire_refrele(prev_ire);
2668 ire_refrele(ire);
2669
2670 /*
2671 * TODO: more precise handling for cases 0, 2, 3, the latter two
2672 * require TOS routing
2673 */
2674 switch (icmph->icmph_code) {
2675 case 0:
2676 case 1:
2677 /* TODO: TOS specificity for cases 2 and 3 */
2678 case 2:
2679 case 3:
2680 break;
2681 default:
2682 BUMP_MIB(&ipst->ips_icmp_mib, icmpInBadRedirects);
2683 ip_drop_input("icmpInBadRedirects - code", mp, ira->ira_ill);
2684 freemsg(mp);
2685 return;
2686 }
2687 /*
2688 * Create a Route Association. This will allow us to remember that
2689 * someone we believe told us to use the particular gateway.
2690 */
2691 ire = ire_create(
2692 (uchar_t *)&dst, /* dest addr */
2693 (uchar_t *)&ip_g_all_ones, /* mask */
2694 (uchar_t *)&gateway, /* gateway addr */
2695 IRE_HOST,
2696 NULL, /* ill */
2697 ALL_ZONES,
2698 (RTF_DYNAMIC | RTF_GATEWAY | RTF_HOST),
2699 NULL, /* tsol_gc_t */
2700 ipst);
2701
2702 if (ire == NULL) {
2703 freemsg(mp);
2704 return;
2705 }
2706 nire = ire_add(ire);
2707 /* Check if it was a duplicate entry */
2708 if (nire != NULL && nire != ire) {
2709 ASSERT(nire->ire_identical_ref > 1);
2710 ire_delete(nire);
2711 ire_refrele(nire);
2712 nire = NULL;
2713 }
2714 ire = nire;
2715 if (ire != NULL) {
2716 ire_refrele(ire); /* Held in ire_add */
2717
2718 /* tell routing sockets that we received a redirect */
2719 ip_rts_change(RTM_REDIRECT, dst, gateway, IP_HOST_MASK, 0, src,
2720 (RTF_DYNAMIC | RTF_GATEWAY | RTF_HOST), 0,
2721 (RTA_DST | RTA_GATEWAY | RTA_NETMASK | RTA_AUTHOR), ipst);
2722 }
2723
2724 /*
2725 * Delete any existing IRE_HOST type redirect ires for this destination.
2726 * This together with the added IRE has the effect of
2727 * modifying an existing redirect.
2728 */
2729 prev_ire = ire_ftable_lookup_v4(dst, 0, src, IRE_HOST, NULL,
2730 ALL_ZONES, NULL, (MATCH_IRE_GW | MATCH_IRE_TYPE), 0, ipst, NULL);
2731 if (prev_ire != NULL) {
2732 if (prev_ire ->ire_flags & RTF_DYNAMIC)
2733 ire_delete(prev_ire);
2734 ire_refrele(prev_ire);
2735 }
2736
2737 freemsg(mp);
2738 }
2739
2740 /*
2741 * Generate an ICMP parameter problem message.
2742 * When called from ip_output side a minimal ip_recv_attr_t needs to be
2743 * constructed by the caller.
2744 */
2745 static void
2746 icmp_param_problem(mblk_t *mp, uint8_t ptr, ip_recv_attr_t *ira)
2747 {
2748 icmph_t icmph;
2749 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2750
2751 mp = icmp_pkt_err_ok(mp, ira);
2752 if (mp == NULL)
2753 return;
2754
2755 bzero(&icmph, sizeof (icmph_t));
2756 icmph.icmph_type = ICMP_PARAM_PROBLEM;
2757 icmph.icmph_pp_ptr = ptr;
2758 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutParmProbs);
2759 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
2760 }
2761
2762 /*
2763 * Build and ship an IPv4 ICMP message using the packet data in mp, and
2764 * the ICMP header pointed to by "stuff". (May be called as writer.)
2765 * Note: assumes that icmp_pkt_err_ok has been called to verify that
2766 * an icmp error packet can be sent.
2767 * Assigns an appropriate source address to the packet. If ipha_dst is
2768 * one of our addresses use it for source. Otherwise let ip_output_simple
2769 * pick the source address.
2770 */
2771 static void
2772 icmp_pkt(mblk_t *mp, void *stuff, size_t len, ip_recv_attr_t *ira)
2773 {
2774 ipaddr_t dst;
2775 icmph_t *icmph;
2776 ipha_t *ipha;
2777 uint_t len_needed;
2778 size_t msg_len;
2779 mblk_t *mp1;
2780 ipaddr_t src;
2781 ire_t *ire;
2782 ip_xmit_attr_t ixas;
2783 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2784
2785 ipha = (ipha_t *)mp->b_rptr;
2786
2787 bzero(&ixas, sizeof (ixas));
2788 ixas.ixa_flags = IXAF_BASIC_SIMPLE_V4;
2789 ixas.ixa_zoneid = ira->ira_zoneid;
2790 ixas.ixa_ifindex = 0;
2791 ixas.ixa_ipst = ipst;
2792 ixas.ixa_cred = kcred;
2793 ixas.ixa_cpid = NOPID;
2794 ixas.ixa_tsl = ira->ira_tsl; /* Behave as a multi-level responder */
2795 ixas.ixa_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
2796
2797 if (ira->ira_flags & IRAF_IPSEC_SECURE) {
2798 /*
2799 * Apply IPsec based on how IPsec was applied to
2800 * the packet that had the error.
2801 *
2802 * If it was an outbound packet that caused the ICMP
2803 * error, then the caller will have setup the IRA
2804 * appropriately.
2805 */
2806 if (!ipsec_in_to_out(ira, &ixas, mp, ipha, NULL)) {
2807 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
2808 /* Note: mp already consumed and ip_drop_packet done */
2809 return;
2810 }
2811 } else {
2812 /*
2813 * This is in clear. The icmp message we are building
2814 * here should go out in clear, independent of our policy.
2815 */
2816 ixas.ixa_flags |= IXAF_NO_IPSEC;
2817 }
2818
2819 /* Remember our eventual destination */
2820 dst = ipha->ipha_src;
2821
2822 /*
2823 * If the packet was for one of our unicast addresses, make
2824 * sure we respond with that as the source. Otherwise
2825 * have ip_output_simple pick the source address.
2826 */
2827 ire = ire_ftable_lookup_v4(ipha->ipha_dst, 0, 0,
2828 (IRE_LOCAL|IRE_LOOPBACK), NULL, ira->ira_zoneid, NULL,
2829 MATCH_IRE_TYPE|MATCH_IRE_ZONEONLY, 0, ipst, NULL);
2830 if (ire != NULL) {
2831 ire_refrele(ire);
2832 src = ipha->ipha_dst;
2833 } else {
2834 src = INADDR_ANY;
2835 ixas.ixa_flags |= IXAF_SET_SOURCE;
2836 }
2837
2838 /*
2839 * Check if we can send back more then 8 bytes in addition to
2840 * the IP header. We try to send 64 bytes of data and the internal
2841 * header in the special cases of ipv4 encapsulated ipv4 or ipv6.
2842 */
2843 len_needed = IPH_HDR_LENGTH(ipha);
2844 if (ipha->ipha_protocol == IPPROTO_ENCAP ||
2845 ipha->ipha_protocol == IPPROTO_IPV6) {
2846 if (!pullupmsg(mp, -1)) {
2847 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
2848 ip_drop_output("ipIfStatsOutDiscards", mp, NULL);
2849 freemsg(mp);
2850 return;
2851 }
2852 ipha = (ipha_t *)mp->b_rptr;
2853
2854 if (ipha->ipha_protocol == IPPROTO_ENCAP) {
2855 len_needed += IPH_HDR_LENGTH(((uchar_t *)ipha +
2856 len_needed));
2857 } else {
2858 ip6_t *ip6h = (ip6_t *)((uchar_t *)ipha + len_needed);
2859
2860 ASSERT(ipha->ipha_protocol == IPPROTO_IPV6);
2861 len_needed += ip_hdr_length_v6(mp, ip6h);
2862 }
2863 }
2864 len_needed += ipst->ips_ip_icmp_return;
2865 msg_len = msgdsize(mp);
2866 if (msg_len > len_needed) {
2867 (void) adjmsg(mp, len_needed - msg_len);
2868 msg_len = len_needed;
2869 }
2870 mp1 = allocb(sizeof (icmp_ipha) + len, BPRI_MED);
2871 if (mp1 == NULL) {
2872 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutErrors);
2873 freemsg(mp);
2874 return;
2875 }
2876 mp1->b_cont = mp;
2877 mp = mp1;
2878
2879 /*
2880 * Set IXAF_TRUSTED_ICMP so we can let the ICMP messages this
2881 * node generates be accepted in peace by all on-host destinations.
2882 * If we do NOT assume that all on-host destinations trust
2883 * self-generated ICMP messages, then rework here, ip6.c, and spd.c.
2884 * (Look for IXAF_TRUSTED_ICMP).
2885 */
2886 ixas.ixa_flags |= IXAF_TRUSTED_ICMP;
2887
2888 ipha = (ipha_t *)mp->b_rptr;
2889 mp1->b_wptr = (uchar_t *)ipha + (sizeof (icmp_ipha) + len);
2890 *ipha = icmp_ipha;
2891 ipha->ipha_src = src;
2892 ipha->ipha_dst = dst;
2893 ipha->ipha_ttl = ipst->ips_ip_def_ttl;
2894 msg_len += sizeof (icmp_ipha) + len;
2895 if (msg_len > IP_MAXPACKET) {
2896 (void) adjmsg(mp, IP_MAXPACKET - msg_len);
2897 msg_len = IP_MAXPACKET;
2898 }
2899 ipha->ipha_length = htons((uint16_t)msg_len);
2900 icmph = (icmph_t *)&ipha[1];
2901 bcopy(stuff, icmph, len);
2902 icmph->icmph_checksum = 0;
2903 icmph->icmph_checksum = IP_CSUM(mp, (int32_t)sizeof (ipha_t), 0);
2904 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutMsgs);
2905
2906 (void) ip_output_simple(mp, &ixas);
2907 ixa_cleanup(&ixas);
2908 }
2909
2910 /*
2911 * Determine if an ICMP error packet can be sent given the rate limit.
2912 * The limit consists of an average frequency (icmp_pkt_err_interval measured
2913 * in milliseconds) and a burst size. Burst size number of packets can
2914 * be sent arbitrarely closely spaced.
2915 * The state is tracked using two variables to implement an approximate
2916 * token bucket filter:
2917 * icmp_pkt_err_last - lbolt value when the last burst started
2918 * icmp_pkt_err_sent - number of packets sent in current burst
2919 */
2920 boolean_t
2921 icmp_err_rate_limit(ip_stack_t *ipst)
2922 {
2923 clock_t now = TICK_TO_MSEC(ddi_get_lbolt());
2924 uint_t refilled; /* Number of packets refilled in tbf since last */
2925 /* Guard against changes by loading into local variable */
2926 uint_t err_interval = ipst->ips_ip_icmp_err_interval;
2927
2928 if (err_interval == 0)
2929 return (B_FALSE);
2930
2931 if (ipst->ips_icmp_pkt_err_last > now) {
2932 /* 100HZ lbolt in ms for 32bit arch wraps every 49.7 days */
2933 ipst->ips_icmp_pkt_err_last = 0;
2934 ipst->ips_icmp_pkt_err_sent = 0;
2935 }
2936 /*
2937 * If we are in a burst update the token bucket filter.
2938 * Update the "last" time to be close to "now" but make sure
2939 * we don't loose precision.
2940 */
2941 if (ipst->ips_icmp_pkt_err_sent != 0) {
2942 refilled = (now - ipst->ips_icmp_pkt_err_last)/err_interval;
2943 if (refilled > ipst->ips_icmp_pkt_err_sent) {
2944 ipst->ips_icmp_pkt_err_sent = 0;
2945 } else {
2946 ipst->ips_icmp_pkt_err_sent -= refilled;
2947 ipst->ips_icmp_pkt_err_last += refilled * err_interval;
2948 }
2949 }
2950 if (ipst->ips_icmp_pkt_err_sent == 0) {
2951 /* Start of new burst */
2952 ipst->ips_icmp_pkt_err_last = now;
2953 }
2954 if (ipst->ips_icmp_pkt_err_sent < ipst->ips_ip_icmp_err_burst) {
2955 ipst->ips_icmp_pkt_err_sent++;
2956 ip1dbg(("icmp_err_rate_limit: %d sent in burst\n",
2957 ipst->ips_icmp_pkt_err_sent));
2958 return (B_FALSE);
2959 }
2960 ip1dbg(("icmp_err_rate_limit: dropped\n"));
2961 return (B_TRUE);
2962 }
2963
2964 /*
2965 * Check if it is ok to send an IPv4 ICMP error packet in
2966 * response to the IPv4 packet in mp.
2967 * Free the message and return null if no
2968 * ICMP error packet should be sent.
2969 */
2970 static mblk_t *
2971 icmp_pkt_err_ok(mblk_t *mp, ip_recv_attr_t *ira)
2972 {
2973 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2974 icmph_t *icmph;
2975 ipha_t *ipha;
2976 uint_t len_needed;
2977
2978 if (!mp)
2979 return (NULL);
2980 ipha = (ipha_t *)mp->b_rptr;
2981 if (ip_csum_hdr(ipha)) {
2982 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsInCksumErrs);
2983 ip_drop_input("ipIfStatsInCksumErrs", mp, NULL);
2984 freemsg(mp);
2985 return (NULL);
2986 }
2987 if (ip_type_v4(ipha->ipha_dst, ipst) == IRE_BROADCAST ||
2988 ip_type_v4(ipha->ipha_src, ipst) == IRE_BROADCAST ||
2989 CLASSD(ipha->ipha_dst) ||
2990 CLASSD(ipha->ipha_src) ||
2991 (ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_OFFSET)) {
2992 /* Note: only errors to the fragment with offset 0 */
2993 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
2994 freemsg(mp);
2995 return (NULL);
2996 }
2997 if (ipha->ipha_protocol == IPPROTO_ICMP) {
2998 /*
2999 * Check the ICMP type. RFC 1122 sez: don't send ICMP
3000 * errors in response to any ICMP errors.
3001 */
3002 len_needed = IPH_HDR_LENGTH(ipha) + ICMPH_SIZE;
3003 if (mp->b_wptr - mp->b_rptr < len_needed) {
3004 if (!pullupmsg(mp, len_needed)) {
3005 BUMP_MIB(&ipst->ips_icmp_mib, icmpInErrors);
3006 freemsg(mp);
3007 return (NULL);
3008 }
3009 ipha = (ipha_t *)mp->b_rptr;
3010 }
3011 icmph = (icmph_t *)
3012 (&((char *)ipha)[IPH_HDR_LENGTH(ipha)]);
3013 switch (icmph->icmph_type) {
3014 case ICMP_DEST_UNREACHABLE:
3015 case ICMP_SOURCE_QUENCH:
3016 case ICMP_TIME_EXCEEDED:
3017 case ICMP_PARAM_PROBLEM:
3018 case ICMP_REDIRECT:
3019 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
3020 freemsg(mp);
3021 return (NULL);
3022 default:
3023 break;
3024 }
3025 }
3026 /*
3027 * If this is a labeled system, then check to see if we're allowed to
3028 * send a response to this particular sender. If not, then just drop.
3029 */
3030 if (is_system_labeled() && !tsol_can_reply_error(mp, ira)) {
3031 ip2dbg(("icmp_pkt_err_ok: can't respond to packet\n"));
3032 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
3033 freemsg(mp);
3034 return (NULL);
3035 }
3036 if (icmp_err_rate_limit(ipst)) {
3037 /*
3038 * Only send ICMP error packets every so often.
3039 * This should be done on a per port/source basis,
3040 * but for now this will suffice.
3041 */
3042 freemsg(mp);
3043 return (NULL);
3044 }
3045 return (mp);
3046 }
3047
3048 /*
3049 * Called when a packet was sent out the same link that it arrived on.
3050 * Check if it is ok to send a redirect and then send it.
3051 */
3052 void
3053 ip_send_potential_redirect_v4(mblk_t *mp, ipha_t *ipha, ire_t *ire,
3054 ip_recv_attr_t *ira)
3055 {
3056 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3057 ipaddr_t src, nhop;
3058 mblk_t *mp1;
3059 ire_t *nhop_ire;
3060
3061 /*
3062 * Check the source address to see if it originated
3063 * on the same logical subnet it is going back out on.
3064 * If so, we should be able to send it a redirect.
3065 * Avoid sending a redirect if the destination
3066 * is directly connected (i.e., we matched an IRE_ONLINK),
3067 * or if the packet was source routed out this interface.
3068 *
3069 * We avoid sending a redirect if the
3070 * destination is directly connected
3071 * because it is possible that multiple
3072 * IP subnets may have been configured on
3073 * the link, and the source may not
3074 * be on the same subnet as ip destination,
3075 * even though they are on the same
3076 * physical link.
3077 */
3078 if ((ire->ire_type & IRE_ONLINK) ||
3079 ip_source_routed(ipha, ipst))
3080 return;
3081
3082 nhop_ire = ire_nexthop(ire);
3083 if (nhop_ire == NULL)
3084 return;
3085
3086 nhop = nhop_ire->ire_addr;
3087
3088 if (nhop_ire->ire_type & IRE_IF_CLONE) {
3089 ire_t *ire2;
3090
3091 /* Follow ire_dep_parent to find non-clone IRE_INTERFACE */
3092 mutex_enter(&nhop_ire->ire_lock);
3093 ire2 = nhop_ire->ire_dep_parent;
3094 if (ire2 != NULL)
3095 ire_refhold(ire2);
3096 mutex_exit(&nhop_ire->ire_lock);
3097 ire_refrele(nhop_ire);
3098 nhop_ire = ire2;
3099 }
3100 if (nhop_ire == NULL)
3101 return;
3102
3103 ASSERT(!(nhop_ire->ire_type & IRE_IF_CLONE));
3104
3105 src = ipha->ipha_src;
3106
3107 /*
3108 * We look at the interface ire for the nexthop,
3109 * to see if ipha_src is in the same subnet
3110 * as the nexthop.
3111 */
3112 if ((src & nhop_ire->ire_mask) == (nhop & nhop_ire->ire_mask)) {
3113 /*
3114 * The source is directly connected.
3115 */
3116 mp1 = copymsg(mp);
3117 if (mp1 != NULL) {
3118 icmp_send_redirect(mp1, nhop, ira);
3119 }
3120 }
3121 ire_refrele(nhop_ire);
3122 }
3123
3124 /*
3125 * Generate an ICMP redirect message.
3126 */
3127 static void
3128 icmp_send_redirect(mblk_t *mp, ipaddr_t gateway, ip_recv_attr_t *ira)
3129 {
3130 icmph_t icmph;
3131 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3132
3133 mp = icmp_pkt_err_ok(mp, ira);
3134 if (mp == NULL)
3135 return;
3136
3137 bzero(&icmph, sizeof (icmph_t));
3138 icmph.icmph_type = ICMP_REDIRECT;
3139 icmph.icmph_code = 1;
3140 icmph.icmph_rd_gateway = gateway;
3141 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutRedirects);
3142 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3143 }
3144
3145 /*
3146 * Generate an ICMP time exceeded message.
3147 */
3148 void
3149 icmp_time_exceeded(mblk_t *mp, uint8_t code, ip_recv_attr_t *ira)
3150 {
3151 icmph_t icmph;
3152 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3153
3154 mp = icmp_pkt_err_ok(mp, ira);
3155 if (mp == NULL)
3156 return;
3157
3158 bzero(&icmph, sizeof (icmph_t));
3159 icmph.icmph_type = ICMP_TIME_EXCEEDED;
3160 icmph.icmph_code = code;
3161 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutTimeExcds);
3162 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3163 }
3164
3165 /*
3166 * Generate an ICMP unreachable message.
3167 * When called from ip_output side a minimal ip_recv_attr_t needs to be
3168 * constructed by the caller.
3169 */
3170 void
3171 icmp_unreachable(mblk_t *mp, uint8_t code, ip_recv_attr_t *ira)
3172 {
3173 icmph_t icmph;
3174 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3175
3176 mp = icmp_pkt_err_ok(mp, ira);
3177 if (mp == NULL)
3178 return;
3179
3180 bzero(&icmph, sizeof (icmph_t));
3181 icmph.icmph_type = ICMP_DEST_UNREACHABLE;
3182 icmph.icmph_code = code;
3183 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDestUnreachs);
3184 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3185 }
3186
3187 /*
3188 * Latch in the IPsec state for a stream based the policy in the listener
3189 * and the actions in the ip_recv_attr_t.
3190 * Called directly from TCP and SCTP.
3191 */
3192 boolean_t
3193 ip_ipsec_policy_inherit(conn_t *connp, conn_t *lconnp, ip_recv_attr_t *ira)
3194 {
3195 ASSERT(lconnp->conn_policy != NULL);
3196 ASSERT(connp->conn_policy == NULL);
3197
3198 IPPH_REFHOLD(lconnp->conn_policy);
3199 connp->conn_policy = lconnp->conn_policy;
3200
3201 if (ira->ira_ipsec_action != NULL) {
3202 if (connp->conn_latch == NULL) {
3203 connp->conn_latch = iplatch_create();
3204 if (connp->conn_latch == NULL)
3205 return (B_FALSE);
3206 }
3207 ipsec_latch_inbound(connp, ira);
3208 }
3209 return (B_TRUE);
3210 }
3211
3212 /*
3213 * Verify whether or not the IP address is a valid local address.
3214 * Could be a unicast, including one for a down interface.
3215 * If allow_mcbc then a multicast or broadcast address is also
3216 * acceptable.
3217 *
3218 * In the case of a broadcast/multicast address, however, the
3219 * upper protocol is expected to reset the src address
3220 * to zero when we return IPVL_MCAST/IPVL_BCAST so that
3221 * no packets are emitted with broadcast/multicast address as
3222 * source address (that violates hosts requirements RFC 1122)
3223 * The addresses valid for bind are:
3224 * (1) - INADDR_ANY (0)
3225 * (2) - IP address of an UP interface
3226 * (3) - IP address of a DOWN interface
3227 * (4) - valid local IP broadcast addresses. In this case
3228 * the conn will only receive packets destined to
3229 * the specified broadcast address.
3230 * (5) - a multicast address. In this case
3231 * the conn will only receive packets destined to
3232 * the specified multicast address. Note: the
3233 * application still has to issue an
3234 * IP_ADD_MEMBERSHIP socket option.
3235 *
3236 * In all the above cases, the bound address must be valid in the current zone.
3237 * When the address is loopback, multicast or broadcast, there might be many
3238 * matching IREs so bind has to look up based on the zone.
3239 */
3240 ip_laddr_t
3241 ip_laddr_verify_v4(ipaddr_t src_addr, zoneid_t zoneid,
3242 ip_stack_t *ipst, boolean_t allow_mcbc)
3243 {
3244 ire_t *src_ire;
3245
3246 ASSERT(src_addr != INADDR_ANY);
3247
3248 src_ire = ire_ftable_lookup_v4(src_addr, 0, 0, 0,
3249 NULL, zoneid, NULL, MATCH_IRE_ZONEONLY, 0, ipst, NULL);
3250
3251 /*
3252 * If an address other than in6addr_any is requested,
3253 * we verify that it is a valid address for bind
3254 * Note: Following code is in if-else-if form for
3255 * readability compared to a condition check.
3256 */
3257 if (src_ire != NULL && (src_ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK))) {
3258 /*
3259 * (2) Bind to address of local UP interface
3260 */
3261 ire_refrele(src_ire);
3262 return (IPVL_UNICAST_UP);
3263 } else if (src_ire != NULL && src_ire->ire_type & IRE_BROADCAST) {
3264 /*
3265 * (4) Bind to broadcast address
3266 */
3267 ire_refrele(src_ire);
3268 if (allow_mcbc)
3269 return (IPVL_BCAST);
3270 else
3271 return (IPVL_BAD);
3272 } else if (CLASSD(src_addr)) {
3273 /* (5) bind to multicast address. */
3274 if (src_ire != NULL)
3275 ire_refrele(src_ire);
3276
3277 if (allow_mcbc)
3278 return (IPVL_MCAST);
3279 else
3280 return (IPVL_BAD);
3281 } else {
3282 ipif_t *ipif;
3283
3284 /*
3285 * (3) Bind to address of local DOWN interface?
3286 * (ipif_lookup_addr() looks up all interfaces
3287 * but we do not get here for UP interfaces
3288 * - case (2) above)
3289 */
3290 if (src_ire != NULL)
3291 ire_refrele(src_ire);
3292
3293 ipif = ipif_lookup_addr(src_addr, NULL, zoneid, ipst);
3294 if (ipif == NULL)
3295 return (IPVL_BAD);
3296
3297 /* Not a useful source? */
3298 if (ipif->ipif_flags & (IPIF_NOLOCAL | IPIF_ANYCAST)) {
3299 ipif_refrele(ipif);
3300 return (IPVL_BAD);
3301 }
3302 ipif_refrele(ipif);
3303 return (IPVL_UNICAST_DOWN);
3304 }
3305 }
3306
3307 /*
3308 * Insert in the bind fanout for IPv4 and IPv6.
3309 * The caller should already have used ip_laddr_verify_v*() before calling
3310 * this.
3311 */
3312 int
3313 ip_laddr_fanout_insert(conn_t *connp)
3314 {
3315 int error;
3316
3317 /*
3318 * Allow setting new policies. For example, disconnects result
3319 * in us being called. As we would have set conn_policy_cached
3320 * to B_TRUE before, we should set it to B_FALSE, so that policy
3321 * can change after the disconnect.
3322 */
3323 connp->conn_policy_cached = B_FALSE;
3324
3325 error = ipcl_bind_insert(connp);
3326 if (error != 0) {
3327 if (connp->conn_anon_port) {
3328 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
3329 connp->conn_mlp_type, connp->conn_proto,
3330 ntohs(connp->conn_lport), B_FALSE);
3331 }
3332 connp->conn_mlp_type = mlptSingle;
3333 }
3334 return (error);
3335 }
3336
3337 /*
3338 * Verify that both the source and destination addresses are valid. If
3339 * IPDF_VERIFY_DST is not set, then the destination address may be unreachable,
3340 * i.e. have no route to it. Protocols like TCP want to verify destination
3341 * reachability, while tunnels do not.
3342 *
3343 * Determine the route, the interface, and (optionally) the source address
3344 * to use to reach a given destination.
3345 * Note that we allow connect to broadcast and multicast addresses when
3346 * IPDF_ALLOW_MCBC is set.
3347 * first_hop and dst_addr are normally the same, but if source routing
3348 * they will differ; in that case the first_hop is what we'll use for the
3349 * routing lookup but the dce and label checks will be done on dst_addr,
3350 *
3351 * If uinfo is set, then we fill in the best available information
3352 * we have for the destination. This is based on (in priority order) any
3353 * metrics and path MTU stored in a dce_t, route metrics, and finally the
3354 * ill_mtu/ill_mc_mtu.
3355 *
3356 * Tsol note: If we have a source route then dst_addr != firsthop. But we
3357 * always do the label check on dst_addr.
3358 */
3359 int
3360 ip_set_destination_v4(ipaddr_t *src_addrp, ipaddr_t dst_addr, ipaddr_t firsthop,
3361 ip_xmit_attr_t *ixa, iulp_t *uinfo, uint32_t flags, uint_t mac_mode)
3362 {
3363 ire_t *ire = NULL;
3364 int error = 0;
3365 ipaddr_t setsrc; /* RTF_SETSRC */
3366 zoneid_t zoneid = ixa->ixa_zoneid; /* Honors SO_ALLZONES */
3367 ip_stack_t *ipst = ixa->ixa_ipst;
3368 dce_t *dce;
3369 uint_t pmtu;
3370 uint_t generation;
3371 nce_t *nce;
3372 ill_t *ill = NULL;
3373 boolean_t multirt = B_FALSE;
3374
3375 ASSERT(ixa->ixa_flags & IXAF_IS_IPV4);
3376
3377 /*
3378 * We never send to zero; the ULPs map it to the loopback address.
3379 * We can't allow it since we use zero to mean unitialized in some
3380 * places.
3381 */
3382 ASSERT(dst_addr != INADDR_ANY);
3383
3384 if (is_system_labeled()) {
3385 ts_label_t *tsl = NULL;
3386
3387 error = tsol_check_dest(ixa->ixa_tsl, &dst_addr, IPV4_VERSION,
3388 mac_mode, (flags & IPDF_ZONE_IS_GLOBAL) != 0, &tsl);
3389 if (error != 0)
3390 return (error);
3391 if (tsl != NULL) {
3392 /* Update the label */
3393 ip_xmit_attr_replace_tsl(ixa, tsl);
3394 }
3395 }
3396
3397 setsrc = INADDR_ANY;
3398 /*
3399 * Select a route; For IPMP interfaces, we would only select
3400 * a "hidden" route (i.e., going through a specific under_ill)
3401 * if ixa_ifindex has been specified.
3402 */
3403 ire = ip_select_route_v4(firsthop, *src_addrp, ixa,
3404 &generation, &setsrc, &error, &multirt);
3405 ASSERT(ire != NULL); /* IRE_NOROUTE if none found */
3406 if (error != 0)
3407 goto bad_addr;
3408
3409 /*
3410 * ire can't be a broadcast or multicast unless IPDF_ALLOW_MCBC is set.
3411 * If IPDF_VERIFY_DST is set, the destination must be reachable;
3412 * Otherwise the destination needn't be reachable.
3413 *
3414 * If we match on a reject or black hole, then we've got a
3415 * local failure. May as well fail out the connect() attempt,
3416 * since it's never going to succeed.
3417 */
3418 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
3419 /*
3420 * If we're verifying destination reachability, we always want
3421 * to complain here.
3422 *
3423 * If we're not verifying destination reachability but the
3424 * destination has a route, we still want to fail on the
3425 * temporary address and broadcast address tests.
3426 *
3427 * In both cases do we let the code continue so some reasonable
3428 * information is returned to the caller. That enables the
3429 * caller to use (and even cache) the IRE. conn_ip_ouput will
3430 * use the generation mismatch path to check for the unreachable
3431 * case thereby avoiding any specific check in the main path.
3432 */
3433 ASSERT(generation == IRE_GENERATION_VERIFY);
3434 if (flags & IPDF_VERIFY_DST) {
3435 /*
3436 * Set errno but continue to set up ixa_ire to be
3437 * the RTF_REJECT|RTF_BLACKHOLE IRE.
3438 * That allows callers to use ip_output to get an
3439 * ICMP error back.
3440 */
3441 if (!(ire->ire_type & IRE_HOST))
3442 error = ENETUNREACH;
3443 else
3444 error = EHOSTUNREACH;
3445 }
3446 }
3447
3448 if ((ire->ire_type & (IRE_BROADCAST|IRE_MULTICAST)) &&
3449 !(flags & IPDF_ALLOW_MCBC)) {
3450 ire_refrele(ire);
3451 ire = ire_reject(ipst, B_FALSE);
3452 generation = IRE_GENERATION_VERIFY;
3453 error = ENETUNREACH;
3454 }
3455
3456 /* Cache things */
3457 if (ixa->ixa_ire != NULL)
3458 ire_refrele_notr(ixa->ixa_ire);
3459 #ifdef DEBUG
3460 ire_refhold_notr(ire);
3461 ire_refrele(ire);
3462 #endif
3463 ixa->ixa_ire = ire;
3464 ixa->ixa_ire_generation = generation;
3465
3466 /*
3467 * Ensure that ixa_dce is always set any time that ixa_ire is set,
3468 * since some callers will send a packet to conn_ip_output() even if
3469 * there's an error.
3470 */
3471 if (flags & IPDF_UNIQUE_DCE) {
3472 /* Fallback to the default dce if allocation fails */
3473 dce = dce_lookup_and_add_v4(dst_addr, ipst);
3474 if (dce != NULL)
3475 generation = dce->dce_generation;
3476 else
3477 dce = dce_lookup_v4(dst_addr, ipst, &generation);
3478 } else {
3479 dce = dce_lookup_v4(dst_addr, ipst, &generation);
3480 }
3481 ASSERT(dce != NULL);
3482 if (ixa->ixa_dce != NULL)
3483 dce_refrele_notr(ixa->ixa_dce);
3484 #ifdef DEBUG
3485 dce_refhold_notr(dce);
3486 dce_refrele(dce);
3487 #endif
3488 ixa->ixa_dce = dce;
3489 ixa->ixa_dce_generation = generation;
3490
3491 /*
3492 * For multicast with multirt we have a flag passed back from
3493 * ire_lookup_multi_ill_v4 since we don't have an IRE for each
3494 * possible multicast address.
3495 * We also need a flag for multicast since we can't check
3496 * whether RTF_MULTIRT is set in ixa_ire for multicast.
3497 */
3498 if (multirt) {
3499 ixa->ixa_postfragfn = ip_postfrag_multirt_v4;
3500 ixa->ixa_flags |= IXAF_MULTIRT_MULTICAST;
3501 } else {
3502 ixa->ixa_postfragfn = ire->ire_postfragfn;
3503 ixa->ixa_flags &= ~IXAF_MULTIRT_MULTICAST;
3504 }
3505 if (!(ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))) {
3506 /* Get an nce to cache. */
3507 nce = ire_to_nce(ire, firsthop, NULL);
3508 if (nce == NULL) {
3509 /* Allocation failure? */
3510 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3511 } else {
3512 if (ixa->ixa_nce != NULL)
3513 nce_refrele(ixa->ixa_nce);
3514 ixa->ixa_nce = nce;
3515 }
3516 }
3517
3518 /*
3519 * If the source address is a loopback address, the
3520 * destination had best be local or multicast.
3521 * If we are sending to an IRE_LOCAL using a loopback source then
3522 * it had better be the same zoneid.
3523 */
3524 if (*src_addrp == htonl(INADDR_LOOPBACK)) {
3525 if ((ire->ire_type & IRE_LOCAL) && ire->ire_zoneid != zoneid) {
3526 ire = NULL; /* Stored in ixa_ire */
3527 error = EADDRNOTAVAIL;
3528 goto bad_addr;
3529 }
3530 if (!(ire->ire_type & (IRE_LOOPBACK|IRE_LOCAL|IRE_MULTICAST))) {
3531 ire = NULL; /* Stored in ixa_ire */
3532 error = EADDRNOTAVAIL;
3533 goto bad_addr;
3534 }
3535 }
3536 if (ire->ire_type & IRE_BROADCAST) {
3537 /*
3538 * If the ULP didn't have a specified source, then we
3539 * make sure we reselect the source when sending
3540 * broadcasts out different interfaces.
3541 */
3542 if (flags & IPDF_SELECT_SRC)
3543 ixa->ixa_flags |= IXAF_SET_SOURCE;
3544 else
3545 ixa->ixa_flags &= ~IXAF_SET_SOURCE;
3546 }
3547
3548 /*
3549 * Does the caller want us to pick a source address?
3550 */
3551 if (flags & IPDF_SELECT_SRC) {
3552 ipaddr_t src_addr;
3553
3554 /*
3555 * We use use ire_nexthop_ill to avoid the under ipmp
3556 * interface for source address selection. Note that for ipmp
3557 * probe packets, ixa_ifindex would have been specified, and
3558 * the ip_select_route() invocation would have picked an ire
3559 * will ire_ill pointing at an under interface.
3560 */
3561 ill = ire_nexthop_ill(ire);
3562
3563 /* If unreachable we have no ill but need some source */
3564 if (ill == NULL) {
3565 src_addr = htonl(INADDR_LOOPBACK);
3566 /* Make sure we look for a better source address */
3567 generation = SRC_GENERATION_VERIFY;
3568 } else {
3569 error = ip_select_source_v4(ill, setsrc, dst_addr,
3570 ixa->ixa_multicast_ifaddr, zoneid,
3571 ipst, &src_addr, &generation, NULL);
3572 if (error != 0) {
3573 ire = NULL; /* Stored in ixa_ire */
3574 goto bad_addr;
3575 }
3576 }
3577
3578 /*
3579 * We allow the source address to to down.
3580 * However, we check that we don't use the loopback address
3581 * as a source when sending out on the wire.
3582 */
3583 if ((src_addr == htonl(INADDR_LOOPBACK)) &&
3584 !(ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK|IRE_MULTICAST)) &&
3585 !(ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))) {
3586 ire = NULL; /* Stored in ixa_ire */
3587 error = EADDRNOTAVAIL;
3588 goto bad_addr;
3589 }
3590
3591 *src_addrp = src_addr;
3592 ixa->ixa_src_generation = generation;
3593 }
3594
3595 /*
3596 * Make sure we don't leave an unreachable ixa_nce in place
3597 * since ip_select_route is used when we unplumb i.e., remove
3598 * references on ixa_ire, ixa_nce, and ixa_dce.
3599 */
3600 nce = ixa->ixa_nce;
3601 if (nce != NULL && nce->nce_is_condemned) {
3602 nce_refrele(nce);
3603 ixa->ixa_nce = NULL;
3604 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3605 }
3606
3607 /*
3608 * The caller has set IXAF_PMTU_DISCOVERY if path MTU is desired.
3609 * However, we can't do it for IPv4 multicast or broadcast.
3610 */
3611 if (ire->ire_type & (IRE_BROADCAST|IRE_MULTICAST))
3612 ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
3613
3614 /*
3615 * Set initial value for fragmentation limit. Either conn_ip_output
3616 * or ULP might updates it when there are routing changes.
3617 * Handles a NULL ixa_ire->ire_ill or a NULL ixa_nce for RTF_REJECT.
3618 */
3619 pmtu = ip_get_pmtu(ixa);
3620 ixa->ixa_fragsize = pmtu;
3621 /* Make sure ixa_fragsize and ixa_pmtu remain identical */
3622 if (ixa->ixa_flags & IXAF_VERIFY_PMTU)
3623 ixa->ixa_pmtu = pmtu;
3624
3625 /*
3626 * Extract information useful for some transports.
3627 * First we look for DCE metrics. Then we take what we have in
3628 * the metrics in the route, where the offlink is used if we have
3629 * one.
3630 */
3631 if (uinfo != NULL) {
3632 bzero(uinfo, sizeof (*uinfo));
3633
3634 if (dce->dce_flags & DCEF_UINFO)
3635 *uinfo = dce->dce_uinfo;
3636
3637 rts_merge_metrics(uinfo, &ire->ire_metrics);
3638
3639 /* Allow ire_metrics to decrease the path MTU from above */
3640 if (uinfo->iulp_mtu == 0 || uinfo->iulp_mtu > pmtu)
3641 uinfo->iulp_mtu = pmtu;
3642
3643 uinfo->iulp_localnet = (ire->ire_type & IRE_ONLINK) != 0;
3644 uinfo->iulp_loopback = (ire->ire_type & IRE_LOOPBACK) != 0;
3645 uinfo->iulp_local = (ire->ire_type & IRE_LOCAL) != 0;
3646 }
3647
3648 if (ill != NULL)
3649 ill_refrele(ill);
3650
3651 return (error);
3652
3653 bad_addr:
3654 if (ire != NULL)
3655 ire_refrele(ire);
3656
3657 if (ill != NULL)
3658 ill_refrele(ill);
3659
3660 /*
3661 * Make sure we don't leave an unreachable ixa_nce in place
3662 * since ip_select_route is used when we unplumb i.e., remove
3663 * references on ixa_ire, ixa_nce, and ixa_dce.
3664 */
3665 nce = ixa->ixa_nce;
3666 if (nce != NULL && nce->nce_is_condemned) {
3667 nce_refrele(nce);
3668 ixa->ixa_nce = NULL;
3669 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3670 }
3671
3672 return (error);
3673 }
3674
3675
3676 /*
3677 * Get the base MTU for the case when path MTU discovery is not used.
3678 * Takes the MTU of the IRE into account.
3679 */
3680 uint_t
3681 ip_get_base_mtu(ill_t *ill, ire_t *ire)
3682 {
3683 uint_t mtu;
3684 uint_t iremtu = ire->ire_metrics.iulp_mtu;
3685
3686 if (ire->ire_type & (IRE_MULTICAST|IRE_BROADCAST))
3687 mtu = ill->ill_mc_mtu;
3688 else
3689 mtu = ill->ill_mtu;
3690
3691 if (iremtu != 0 && iremtu < mtu)
3692 mtu = iremtu;
3693
3694 return (mtu);
3695 }
3696
3697 /*
3698 * Get the PMTU for the attributes. Handles both IPv4 and IPv6.
3699 * Assumes that ixa_ire, dce, and nce have already been set up.
3700 *
3701 * The caller has set IXAF_PMTU_DISCOVERY if path MTU discovery is desired.
3702 * We avoid path MTU discovery if it is disabled with ndd.
3703 * Furtermore, if the path MTU is too small, then we don't set DF for IPv4.
3704 *
3705 * NOTE: We also used to turn it off for source routed packets. That
3706 * is no longer required since the dce is per final destination.
3707 */
3708 uint_t
3709 ip_get_pmtu(ip_xmit_attr_t *ixa)
3710 {
3711 ip_stack_t *ipst = ixa->ixa_ipst;
3712 dce_t *dce;
3713 nce_t *nce;
3714 ire_t *ire;
3715 uint_t pmtu;
3716
3717 ire = ixa->ixa_ire;
3718 dce = ixa->ixa_dce;
3719 nce = ixa->ixa_nce;
3720
3721 /*
3722 * If path MTU discovery has been turned off by ndd, then we ignore
3723 * any dce_pmtu and for IPv4 we will not set DF.
3724 */
3725 if (!ipst->ips_ip_path_mtu_discovery)
3726 ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
3727
3728 pmtu = IP_MAXPACKET;
3729 /*
3730 * Decide whether whether IPv4 sets DF
3731 * For IPv6 "no DF" means to use the 1280 mtu
3732 */
3733 if (ixa->ixa_flags & IXAF_PMTU_DISCOVERY) {
3734 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3735 } else {
3736 ixa->ixa_flags &= ~IXAF_PMTU_IPV4_DF;
3737 if (!(ixa->ixa_flags & IXAF_IS_IPV4))
3738 pmtu = IPV6_MIN_MTU;
3739 }
3740
3741 /* Check if the PMTU is to old before we use it */
3742 if ((dce->dce_flags & DCEF_PMTU) &&
3743 TICK_TO_SEC(ddi_get_lbolt64()) - dce->dce_last_change_time >
3744 ipst->ips_ip_pathmtu_interval) {
3745 /*
3746 * Older than 20 minutes. Drop the path MTU information.
3747 */
3748 mutex_enter(&dce->dce_lock);
3749 dce->dce_flags &= ~(DCEF_PMTU|DCEF_TOO_SMALL_PMTU);
3750 dce->dce_last_change_time = TICK_TO_SEC(ddi_get_lbolt64());
3751 mutex_exit(&dce->dce_lock);
3752 dce_increment_generation(dce);
3753 }
3754
3755 /* The metrics on the route can lower the path MTU */
3756 if (ire->ire_metrics.iulp_mtu != 0 &&
3757 ire->ire_metrics.iulp_mtu < pmtu)
3758 pmtu = ire->ire_metrics.iulp_mtu;
3759
3760 /*
3761 * If the path MTU is smaller than some minimum, we still use dce_pmtu
3762 * above (would be 576 for IPv4 and 1280 for IPv6), but we clear
3763 * IXAF_PMTU_IPV4_DF so that we avoid setting DF for IPv4.
3764 */
3765 if (ixa->ixa_flags & IXAF_PMTU_DISCOVERY) {
3766 if (dce->dce_flags & DCEF_PMTU) {
3767 if (dce->dce_pmtu < pmtu)
3768 pmtu = dce->dce_pmtu;
3769
3770 if (dce->dce_flags & DCEF_TOO_SMALL_PMTU) {
3771 ixa->ixa_flags |= IXAF_PMTU_TOO_SMALL;
3772 ixa->ixa_flags &= ~IXAF_PMTU_IPV4_DF;
3773 } else {
3774 ixa->ixa_flags &= ~IXAF_PMTU_TOO_SMALL;
3775 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3776 }
3777 } else {
3778 ixa->ixa_flags &= ~IXAF_PMTU_TOO_SMALL;
3779 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3780 }
3781 }
3782
3783 /*
3784 * If we have an IRE_LOCAL we use the loopback mtu instead of
3785 * the ill for going out the wire i.e., IRE_LOCAL gets the same
3786 * mtu as IRE_LOOPBACK.
3787 */
3788 if (ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK)) {
3789 uint_t loopback_mtu;
3790
3791 loopback_mtu = (ire->ire_ipversion == IPV6_VERSION) ?
3792 ip_loopback_mtu_v6plus : ip_loopback_mtuplus;
3793
3794 if (loopback_mtu < pmtu)
3795 pmtu = loopback_mtu;
3796 } else if (nce != NULL) {
3797 /*
3798 * Make sure we don't exceed the interface MTU.
3799 * In the case of RTF_REJECT or RTF_BLACKHOLE we might not have
3800 * an ill. We'd use the above IP_MAXPACKET in that case just
3801 * to tell the transport something larger than zero.
3802 */
3803 if (ire->ire_type & (IRE_MULTICAST|IRE_BROADCAST)) {
3804 if (nce->nce_common->ncec_ill->ill_mc_mtu < pmtu)
3805 pmtu = nce->nce_common->ncec_ill->ill_mc_mtu;
3806 if (nce->nce_common->ncec_ill != nce->nce_ill &&
3807 nce->nce_ill->ill_mc_mtu < pmtu) {
3808 /*
3809 * for interfaces in an IPMP group, the mtu of
3810 * the nce_ill (under_ill) could be different
3811 * from the mtu of the ncec_ill, so we take the
3812 * min of the two.
3813 */
3814 pmtu = nce->nce_ill->ill_mc_mtu;
3815 }
3816 } else {
3817 if (nce->nce_common->ncec_ill->ill_mtu < pmtu)
3818 pmtu = nce->nce_common->ncec_ill->ill_mtu;
3819 if (nce->nce_common->ncec_ill != nce->nce_ill &&
3820 nce->nce_ill->ill_mtu < pmtu) {
3821 /*
3822 * for interfaces in an IPMP group, the mtu of
3823 * the nce_ill (under_ill) could be different
3824 * from the mtu of the ncec_ill, so we take the
3825 * min of the two.
3826 */
3827 pmtu = nce->nce_ill->ill_mtu;
3828 }
3829 }
3830 }
3831
3832 /*
3833 * Handle the IPV6_USE_MIN_MTU socket option or ancillary data.
3834 * Only applies to IPv6.
3835 */
3836 if (!(ixa->ixa_flags & IXAF_IS_IPV4)) {
3837 if (ixa->ixa_flags & IXAF_USE_MIN_MTU) {
3838 switch (ixa->ixa_use_min_mtu) {
3839 case IPV6_USE_MIN_MTU_MULTICAST:
3840 if (ire->ire_type & IRE_MULTICAST)
3841 pmtu = IPV6_MIN_MTU;
3842 break;
3843 case IPV6_USE_MIN_MTU_ALWAYS:
3844 pmtu = IPV6_MIN_MTU;
3845 break;
3846 case IPV6_USE_MIN_MTU_NEVER:
3847 break;
3848 }
3849 } else {
3850 /* Default is IPV6_USE_MIN_MTU_MULTICAST */
3851 if (ire->ire_type & IRE_MULTICAST)
3852 pmtu = IPV6_MIN_MTU;
3853 }
3854 }
3855
3856 /*
3857 * For multirouted IPv6 packets, the IP layer will insert a 8-byte
3858 * fragment header in every packet. We compensate for those cases by
3859 * returning a smaller path MTU to the ULP.
3860 *
3861 * In the case of CGTP then ip_output will add a fragment header.
3862 * Make sure there is room for it by telling a smaller number
3863 * to the transport.
3864 *
3865 * When IXAF_IPV6_ADDR_FRAGHDR we subtract the frag hdr here
3866 * so the ULPs consistently see a iulp_pmtu and ip_get_pmtu()
3867 * which is the size of the packets it can send.
3868 */
3869 if (!(ixa->ixa_flags & IXAF_IS_IPV4)) {
3870 if ((ire->ire_flags & RTF_MULTIRT) ||
3871 (ixa->ixa_flags & IXAF_MULTIRT_MULTICAST)) {
3872 pmtu -= sizeof (ip6_frag_t);
3873 ixa->ixa_flags |= IXAF_IPV6_ADD_FRAGHDR;
3874 }
3875 }
3876
3877 return (pmtu);
3878 }
3879
3880 /*
3881 * Carve "len" bytes out of an mblk chain, consuming any we empty, and duping
3882 * the final piece where we don't. Return a pointer to the first mblk in the
3883 * result, and update the pointer to the next mblk to chew on. If anything
3884 * goes wrong (i.e., dupb fails), we waste everything in sight and return a
3885 * NULL pointer.
3886 */
3887 mblk_t *
3888 ip_carve_mp(mblk_t **mpp, ssize_t len)
3889 {
3890 mblk_t *mp0;
3891 mblk_t *mp1;
3892 mblk_t *mp2;
3893
3894 if (!len || !mpp || !(mp0 = *mpp))
3895 return (NULL);
3896 /* If we aren't going to consume the first mblk, we need a dup. */
3897 if (mp0->b_wptr - mp0->b_rptr > len) {
3898 mp1 = dupb(mp0);
3899 if (mp1) {
3900 /* Partition the data between the two mblks. */
3901 mp1->b_wptr = mp1->b_rptr + len;
3902 mp0->b_rptr = mp1->b_wptr;
3903 /*
3904 * after adjustments if mblk not consumed is now
3905 * unaligned, try to align it. If this fails free
3906 * all messages and let upper layer recover.
3907 */
3908 if (!OK_32PTR(mp0->b_rptr)) {
3909 if (!pullupmsg(mp0, -1)) {
3910 freemsg(mp0);
3911 freemsg(mp1);
3912 *mpp = NULL;
3913 return (NULL);
3914 }
3915 }
3916 }
3917 return (mp1);
3918 }
3919 /* Eat through as many mblks as we need to get len bytes. */
3920 len -= mp0->b_wptr - mp0->b_rptr;
3921 for (mp2 = mp1 = mp0; (mp2 = mp2->b_cont) != 0 && len; mp1 = mp2) {
3922 if (mp2->b_wptr - mp2->b_rptr > len) {
3923 /*
3924 * We won't consume the entire last mblk. Like
3925 * above, dup and partition it.
3926 */
3927 mp1->b_cont = dupb(mp2);
3928 mp1 = mp1->b_cont;
3929 if (!mp1) {
3930 /*
3931 * Trouble. Rather than go to a lot of
3932 * trouble to clean up, we free the messages.
3933 * This won't be any worse than losing it on
3934 * the wire.
3935 */
3936 freemsg(mp0);
3937 freemsg(mp2);
3938 *mpp = NULL;
3939 return (NULL);
3940 }
3941 mp1->b_wptr = mp1->b_rptr + len;
3942 mp2->b_rptr = mp1->b_wptr;
3943 /*
3944 * after adjustments if mblk not consumed is now
3945 * unaligned, try to align it. If this fails free
3946 * all messages and let upper layer recover.
3947 */
3948 if (!OK_32PTR(mp2->b_rptr)) {
3949 if (!pullupmsg(mp2, -1)) {
3950 freemsg(mp0);
3951 freemsg(mp2);
3952 *mpp = NULL;
3953 return (NULL);
3954 }
3955 }
3956 *mpp = mp2;
3957 return (mp0);
3958 }
3959 /* Decrement len by the amount we just got. */
3960 len -= mp2->b_wptr - mp2->b_rptr;
3961 }
3962 /*
3963 * len should be reduced to zero now. If not our caller has
3964 * screwed up.
3965 */
3966 if (len) {
3967 /* Shouldn't happen! */
3968 freemsg(mp0);
3969 *mpp = NULL;
3970 return (NULL);
3971 }
3972 /*
3973 * We consumed up to exactly the end of an mblk. Detach the part
3974 * we are returning from the rest of the chain.
3975 */
3976 mp1->b_cont = NULL;
3977 *mpp = mp2;
3978 return (mp0);
3979 }
3980
3981 /* The ill stream is being unplumbed. Called from ip_close */
3982 int
3983 ip_modclose(ill_t *ill)
3984 {
3985 boolean_t success;
3986 ipsq_t *ipsq;
3987 ipif_t *ipif;
3988 queue_t *q = ill->ill_rq;
3989 ip_stack_t *ipst = ill->ill_ipst;
3990 int i;
3991 arl_ill_common_t *ai = ill->ill_common;
3992
3993 /*
3994 * The punlink prior to this may have initiated a capability
3995 * negotiation. But ipsq_enter will block until that finishes or
3996 * times out.
3997 */
3998 success = ipsq_enter(ill, B_FALSE, NEW_OP);
3999
4000 /*
4001 * Open/close/push/pop is guaranteed to be single threaded
4002 * per stream by STREAMS. FS guarantees that all references
4003 * from top are gone before close is called. So there can't
4004 * be another close thread that has set CONDEMNED on this ill.
4005 * and cause ipsq_enter to return failure.
4006 */
4007 ASSERT(success);
4008 ipsq = ill->ill_phyint->phyint_ipsq;
4009
4010 /*
4011 * Mark it condemned. No new reference will be made to this ill.
4012 * Lookup functions will return an error. Threads that try to
4013 * increment the refcnt must check for ILL_CAN_LOOKUP. This ensures
4014 * that the refcnt will drop down to zero.
4015 */
4016 mutex_enter(&ill->ill_lock);
4017 ill->ill_state_flags |= ILL_CONDEMNED;
4018 for (ipif = ill->ill_ipif; ipif != NULL;
4019 ipif = ipif->ipif_next) {
4020 ipif->ipif_state_flags |= IPIF_CONDEMNED;
4021 }
4022 /*
4023 * Wake up anybody waiting to enter the ipsq. ipsq_enter
4024 * returns error if ILL_CONDEMNED is set
4025 */
4026 cv_broadcast(&ill->ill_cv);
4027 mutex_exit(&ill->ill_lock);
4028
4029 /*
4030 * Send all the deferred DLPI messages downstream which came in
4031 * during the small window right before ipsq_enter(). We do this
4032 * without waiting for the ACKs because all the ACKs for M_PROTO
4033 * messages are ignored in ip_rput() when ILL_CONDEMNED is set.
4034 */
4035 ill_dlpi_send_deferred(ill);
4036
4037 /*
4038 * Shut down fragmentation reassembly.
4039 * ill_frag_timer won't start a timer again.
4040 * Now cancel any existing timer
4041 */
4042 (void) untimeout(ill->ill_frag_timer_id);
4043 (void) ill_frag_timeout(ill, 0);
4044
4045 /*
4046 * Call ill_delete to bring down the ipifs, ilms and ill on
4047 * this ill. Then wait for the refcnts to drop to zero.
4048 * ill_is_freeable checks whether the ill is really quiescent.
4049 * Then make sure that threads that are waiting to enter the
4050 * ipsq have seen the error returned by ipsq_enter and have
4051 * gone away. Then we call ill_delete_tail which does the
4052 * DL_UNBIND_REQ with the driver and then qprocsoff.
4053 */
4054 ill_delete(ill);
4055 mutex_enter(&ill->ill_lock);
4056 while (!ill_is_freeable(ill))
4057 cv_wait(&ill->ill_cv, &ill->ill_lock);
4058
4059 while (ill->ill_waiters)
4060 cv_wait(&ill->ill_cv, &ill->ill_lock);
4061
4062 mutex_exit(&ill->ill_lock);
4063
4064 /*
4065 * ill_delete_tail drops reference on ill_ipst, but we need to keep
4066 * it held until the end of the function since the cleanup
4067 * below needs to be able to use the ip_stack_t.
4068 */
4069 netstack_hold(ipst->ips_netstack);
4070
4071 /* qprocsoff is done via ill_delete_tail */
4072 ill_delete_tail(ill);
4073 /*
4074 * synchronously wait for arp stream to unbind. After this, we
4075 * cannot get any data packets up from the driver.
4076 */
4077 arp_unbind_complete(ill);
4078 ASSERT(ill->ill_ipst == NULL);
4079
4080 /*
4081 * Walk through all conns and qenable those that have queued data.
4082 * Close synchronization needs this to
4083 * be done to ensure that all upper layers blocked
4084 * due to flow control to the closing device
4085 * get unblocked.
4086 */
4087 ip1dbg(("ip_wsrv: walking\n"));
4088 for (i = 0; i < TX_FANOUT_SIZE; i++) {
4089 conn_walk_drain(ipst, &ipst->ips_idl_tx_list[i]);
4090 }
4091
4092 /*
4093 * ai can be null if this is an IPv6 ill, or if the IPv4
4094 * stream is being torn down before ARP was plumbed (e.g.,
4095 * /sbin/ifconfig plumbing a stream twice, and encountering
4096 * an error
4097 */
4098 if (ai != NULL) {
4099 ASSERT(!ill->ill_isv6);
4100 mutex_enter(&ai->ai_lock);
4101 ai->ai_ill = NULL;
4102 if (ai->ai_arl == NULL) {
4103 mutex_destroy(&ai->ai_lock);
4104 kmem_free(ai, sizeof (*ai));
4105 } else {
4106 cv_signal(&ai->ai_ill_unplumb_done);
4107 mutex_exit(&ai->ai_lock);
4108 }
4109 }
4110
4111 mutex_enter(&ipst->ips_ip_mi_lock);
4112 mi_close_unlink(&ipst->ips_ip_g_head, (IDP)ill);
4113 mutex_exit(&ipst->ips_ip_mi_lock);
4114
4115 /*
4116 * credp could be null if the open didn't succeed and ip_modopen
4117 * itself calls ip_close.
4118 */
4119 if (ill->ill_credp != NULL)
4120 crfree(ill->ill_credp);
4121
4122 mutex_destroy(&ill->ill_saved_ire_lock);
4123 mutex_destroy(&ill->ill_lock);
4124 rw_destroy(&ill->ill_mcast_lock);
4125 mutex_destroy(&ill->ill_mcast_serializer);
4126 list_destroy(&ill->ill_nce);
4127
4128 /*
4129 * Now we are done with the module close pieces that
4130 * need the netstack_t.
4131 */
4132 netstack_rele(ipst->ips_netstack);
4133
4134 mi_close_free((IDP)ill);
4135 q->q_ptr = WR(q)->q_ptr = NULL;
4136
4137 ipsq_exit(ipsq);
4138
4139 return (0);
4140 }
4141
4142 /*
4143 * This is called as part of close() for IP, UDP, ICMP, and RTS
4144 * in order to quiesce the conn.
4145 */
4146 void
4147 ip_quiesce_conn(conn_t *connp)
4148 {
4149 boolean_t drain_cleanup_reqd = B_FALSE;
4150 boolean_t conn_ioctl_cleanup_reqd = B_FALSE;
4151 boolean_t ilg_cleanup_reqd = B_FALSE;
4152 ip_stack_t *ipst;
4153
4154 ASSERT(!IPCL_IS_TCP(connp));
4155 ipst = connp->conn_netstack->netstack_ip;
4156
4157 /*
4158 * Mark the conn as closing, and this conn must not be
4159 * inserted in future into any list. Eg. conn_drain_insert(),
4160 * won't insert this conn into the conn_drain_list.
4161 *
4162 * conn_idl, and conn_ilg cannot get set henceforth.
4163 */
4164 mutex_enter(&connp->conn_lock);
4165 ASSERT(!(connp->conn_state_flags & CONN_QUIESCED));
4166 connp->conn_state_flags |= CONN_CLOSING;
4167 if (connp->conn_idl != NULL)
4168 drain_cleanup_reqd = B_TRUE;
4169 if (connp->conn_oper_pending_ill != NULL)
4170 conn_ioctl_cleanup_reqd = B_TRUE;
4171 if (connp->conn_dhcpinit_ill != NULL) {
4172 ASSERT(connp->conn_dhcpinit_ill->ill_dhcpinit != 0);
4173 atomic_dec_32(&connp->conn_dhcpinit_ill->ill_dhcpinit);
4174 ill_set_inputfn(connp->conn_dhcpinit_ill);
4175 connp->conn_dhcpinit_ill = NULL;
4176 }
4177 if (connp->conn_ilg != NULL)
4178 ilg_cleanup_reqd = B_TRUE;
4179 mutex_exit(&connp->conn_lock);
4180
4181 if (conn_ioctl_cleanup_reqd)
4182 conn_ioctl_cleanup(connp);
4183
4184 if (is_system_labeled() && connp->conn_anon_port) {
4185 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
4186 connp->conn_mlp_type, connp->conn_proto,
4187 ntohs(connp->conn_lport), B_FALSE);
4188 connp->conn_anon_port = 0;
4189 }
4190 connp->conn_mlp_type = mlptSingle;
4191
4192 /*
4193 * Remove this conn from any fanout list it is on.
4194 * and then wait for any threads currently operating
4195 * on this endpoint to finish
4196 */
4197 ipcl_hash_remove(connp);
4198
4199 /*
4200 * Remove this conn from the drain list, and do any other cleanup that
4201 * may be required. (TCP conns are never flow controlled, and
4202 * conn_idl will be NULL.)
4203 */
4204 if (drain_cleanup_reqd && connp->conn_idl != NULL) {
4205 idl_t *idl = connp->conn_idl;
4206
4207 mutex_enter(&idl->idl_lock);
4208 conn_drain(connp, B_TRUE);
4209 mutex_exit(&idl->idl_lock);
4210 }
4211
4212 if (connp == ipst->ips_ip_g_mrouter)
4213 (void) ip_mrouter_done(ipst);
4214
4215 if (ilg_cleanup_reqd)
4216 ilg_delete_all(connp);
4217
4218 /*
4219 * Now conn refcnt can increase only thru CONN_INC_REF_LOCKED.
4220 * callers from write side can't be there now because close
4221 * is in progress. The only other caller is ipcl_walk
4222 * which checks for the condemned flag.
4223 */
4224 mutex_enter(&connp->conn_lock);
4225 connp->conn_state_flags |= CONN_CONDEMNED;
4226 while (connp->conn_ref != 1)
4227 cv_wait(&connp->conn_cv, &connp->conn_lock);
4228 connp->conn_state_flags |= CONN_QUIESCED;
4229 mutex_exit(&connp->conn_lock);
4230 }
4231
4232 /* ARGSUSED */
4233 int
4234 ip_close(queue_t *q, int flags, cred_t *credp __unused)
4235 {
4236 conn_t *connp;
4237
4238 /*
4239 * Call the appropriate delete routine depending on whether this is
4240 * a module or device.
4241 */
4242 if (WR(q)->q_next != NULL) {
4243 /* This is a module close */
4244 return (ip_modclose((ill_t *)q->q_ptr));
4245 }
4246
4247 connp = q->q_ptr;
4248 ip_quiesce_conn(connp);
4249
4250 qprocsoff(q);
4251
4252 /*
4253 * Now we are truly single threaded on this stream, and can
4254 * delete the things hanging off the connp, and finally the connp.
4255 * We removed this connp from the fanout list, it cannot be
4256 * accessed thru the fanouts, and we already waited for the
4257 * conn_ref to drop to 0. We are already in close, so
4258 * there cannot be any other thread from the top. qprocsoff
4259 * has completed, and service has completed or won't run in
4260 * future.
4261 */
4262 ASSERT(connp->conn_ref == 1);
4263
4264 inet_minor_free(connp->conn_minor_arena, connp->conn_dev);
4265
4266 connp->conn_ref--;
4267 ipcl_conn_destroy(connp);
4268
4269 q->q_ptr = WR(q)->q_ptr = NULL;
4270 return (0);
4271 }
4272
4273 /*
4274 * Wapper around putnext() so that ip_rts_request can merely use
4275 * conn_recv.
4276 */
4277 /*ARGSUSED2*/
4278 static void
4279 ip_conn_input(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira)
4280 {
4281 conn_t *connp = (conn_t *)arg1;
4282
4283 putnext(connp->conn_rq, mp);
4284 }
4285
4286 /* Dummy in case ICMP error delivery is attempted to a /dev/ip instance */
4287 /* ARGSUSED */
4288 static void
4289 ip_conn_input_icmp(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira)
4290 {
4291 freemsg(mp);
4292 }
4293
4294 /*
4295 * Called when the module is about to be unloaded
4296 */
4297 void
4298 ip_ddi_destroy(void)
4299 {
4300 /* This needs to be called before destroying any transports. */
4301 mutex_enter(&cpu_lock);
4302 unregister_cpu_setup_func(ip_tp_cpu_update, NULL);
4303 mutex_exit(&cpu_lock);
4304
4305 tnet_fini();
4306
4307 icmp_ddi_g_destroy();
4308 rts_ddi_g_destroy();
4309 udp_ddi_g_destroy();
4310 sctp_ddi_g_destroy();
4311 tcp_ddi_g_destroy();
4312 ilb_ddi_g_destroy();
4313 dce_g_destroy();
4314 ipsec_policy_g_destroy();
4315 ipcl_g_destroy();
4316 ip_net_g_destroy();
4317 ip_ire_g_fini();
4318 inet_minor_destroy(ip_minor_arena_sa);
4319 #if defined(_LP64)
4320 inet_minor_destroy(ip_minor_arena_la);
4321 #endif
4322
4323 #ifdef DEBUG
4324 list_destroy(&ip_thread_list);
4325 rw_destroy(&ip_thread_rwlock);
4326 tsd_destroy(&ip_thread_data);
4327 #endif
4328
4329 netstack_unregister(NS_IP);
4330 }
4331
4332 /*
4333 * First step in cleanup.
4334 */
4335 /* ARGSUSED */
4336 static void
4337 ip_stack_shutdown(netstackid_t stackid, void *arg)
4338 {
4339 ip_stack_t *ipst = (ip_stack_t *)arg;
4340 kt_did_t ktid;
4341
4342 #ifdef NS_DEBUG
4343 printf("ip_stack_shutdown(%p, stack %d)\n", (void *)ipst, stackid);
4344 #endif
4345
4346 /*
4347 * Perform cleanup for special interfaces (loopback and IPMP).
4348 */
4349 ip_interface_cleanup(ipst);
4350
4351 /*
4352 * The *_hook_shutdown()s start the process of notifying any
4353 * consumers that things are going away.... nothing is destroyed.
4354 */
4355 ipv4_hook_shutdown(ipst);
4356 ipv6_hook_shutdown(ipst);
4357 arp_hook_shutdown(ipst);
4358
4359 mutex_enter(&ipst->ips_capab_taskq_lock);
4360 ktid = ipst->ips_capab_taskq_thread->t_did;
4361 ipst->ips_capab_taskq_quit = B_TRUE;
4362 cv_signal(&ipst->ips_capab_taskq_cv);
4363 mutex_exit(&ipst->ips_capab_taskq_lock);
4364
4365 /*
4366 * In rare occurrences, particularly on virtual hardware where CPUs can
4367 * be de-scheduled, the thread that we just signaled will not run until
4368 * after we have gotten through parts of ip_stack_fini. If that happens
4369 * then we'll try to grab the ips_capab_taskq_lock as part of returning
4370 * from cv_wait which no longer exists.
4371 */
4372 thread_join(ktid);
4373 }
4374
4375 /*
4376 * Free the IP stack instance.
4377 */
4378 static void
4379 ip_stack_fini(netstackid_t stackid, void *arg)
4380 {
4381 ip_stack_t *ipst = (ip_stack_t *)arg;
4382 int ret;
4383
4384 #ifdef NS_DEBUG
4385 printf("ip_stack_fini(%p, stack %d)\n", (void *)ipst, stackid);
4386 #endif
4387 /*
4388 * At this point, all of the notifications that the events and
4389 * protocols are going away have been run, meaning that we can
4390 * now set about starting to clean things up.
4391 */
4392 ipobs_fini(ipst);
4393 ipv4_hook_destroy(ipst);
4394 ipv6_hook_destroy(ipst);
4395 arp_hook_destroy(ipst);
4396 ip_net_destroy(ipst);
4397
4398 ipmp_destroy(ipst);
4399
4400 ip_kstat_fini(stackid, ipst->ips_ip_mibkp);
4401 ipst->ips_ip_mibkp = NULL;
4402 icmp_kstat_fini(stackid, ipst->ips_icmp_mibkp);
4403 ipst->ips_icmp_mibkp = NULL;
4404 ip_kstat2_fini(stackid, ipst->ips_ip_kstat);
4405 ipst->ips_ip_kstat = NULL;
4406 bzero(&ipst->ips_ip_statistics, sizeof (ipst->ips_ip_statistics));
4407 ip6_kstat_fini(stackid, ipst->ips_ip6_kstat);
4408 ipst->ips_ip6_kstat = NULL;
4409 bzero(&ipst->ips_ip6_statistics, sizeof (ipst->ips_ip6_statistics));
4410
4411 kmem_free(ipst->ips_propinfo_tbl,
4412 ip_propinfo_count * sizeof (mod_prop_info_t));
4413 ipst->ips_propinfo_tbl = NULL;
4414
4415 dce_stack_destroy(ipst);
4416 ip_mrouter_stack_destroy(ipst);
4417
4418 /*
4419 * Quiesce all of our timers. Note we set the quiesce flags before we
4420 * call untimeout. The slowtimers may actually kick off another instance
4421 * of the non-slow timers.
4422 */
4423 mutex_enter(&ipst->ips_igmp_timer_lock);
4424 ipst->ips_igmp_timer_quiesce = B_TRUE;
4425 mutex_exit(&ipst->ips_igmp_timer_lock);
4426
4427 mutex_enter(&ipst->ips_mld_timer_lock);
4428 ipst->ips_mld_timer_quiesce = B_TRUE;
4429 mutex_exit(&ipst->ips_mld_timer_lock);
4430
4431 mutex_enter(&ipst->ips_igmp_slowtimeout_lock);
4432 ipst->ips_igmp_slowtimeout_quiesce = B_TRUE;
4433 mutex_exit(&ipst->ips_igmp_slowtimeout_lock);
4434
4435 mutex_enter(&ipst->ips_mld_slowtimeout_lock);
4436 ipst->ips_mld_slowtimeout_quiesce = B_TRUE;
4437 mutex_exit(&ipst->ips_mld_slowtimeout_lock);
4438
4439 ret = untimeout(ipst->ips_igmp_timeout_id);
4440 if (ret == -1) {
4441 ASSERT(ipst->ips_igmp_timeout_id == 0);
4442 } else {
4443 ASSERT(ipst->ips_igmp_timeout_id != 0);
4444 ipst->ips_igmp_timeout_id = 0;
4445 }
4446 ret = untimeout(ipst->ips_igmp_slowtimeout_id);
4447 if (ret == -1) {
4448 ASSERT(ipst->ips_igmp_slowtimeout_id == 0);
4449 } else {
4450 ASSERT(ipst->ips_igmp_slowtimeout_id != 0);
4451 ipst->ips_igmp_slowtimeout_id = 0;
4452 }
4453 ret = untimeout(ipst->ips_mld_timeout_id);
4454 if (ret == -1) {
4455 ASSERT(ipst->ips_mld_timeout_id == 0);
4456 } else {
4457 ASSERT(ipst->ips_mld_timeout_id != 0);
4458 ipst->ips_mld_timeout_id = 0;
4459 }
4460 ret = untimeout(ipst->ips_mld_slowtimeout_id);
4461 if (ret == -1) {
4462 ASSERT(ipst->ips_mld_slowtimeout_id == 0);
4463 } else {
4464 ASSERT(ipst->ips_mld_slowtimeout_id != 0);
4465 ipst->ips_mld_slowtimeout_id = 0;
4466 }
4467
4468 ip_ire_fini(ipst);
4469 ip6_asp_free(ipst);
4470 conn_drain_fini(ipst);
4471 ipcl_destroy(ipst);
4472
4473 mutex_destroy(&ipst->ips_ndp4->ndp_g_lock);
4474 mutex_destroy(&ipst->ips_ndp6->ndp_g_lock);
4475 kmem_free(ipst->ips_ndp4, sizeof (ndp_g_t));
4476 ipst->ips_ndp4 = NULL;
4477 kmem_free(ipst->ips_ndp6, sizeof (ndp_g_t));
4478 ipst->ips_ndp6 = NULL;
4479
4480 if (ipst->ips_loopback_ksp != NULL) {
4481 kstat_delete_netstack(ipst->ips_loopback_ksp, stackid);
4482 ipst->ips_loopback_ksp = NULL;
4483 }
4484
4485 mutex_destroy(&ipst->ips_capab_taskq_lock);
4486 cv_destroy(&ipst->ips_capab_taskq_cv);
4487
4488 rw_destroy(&ipst->ips_srcid_lock);
4489
4490 mutex_destroy(&ipst->ips_ip_mi_lock);
4491 rw_destroy(&ipst->ips_ill_g_usesrc_lock);
4492
4493 mutex_destroy(&ipst->ips_igmp_timer_lock);
4494 mutex_destroy(&ipst->ips_mld_timer_lock);
4495 mutex_destroy(&ipst->ips_igmp_slowtimeout_lock);
4496 mutex_destroy(&ipst->ips_mld_slowtimeout_lock);
4497 mutex_destroy(&ipst->ips_ip_addr_avail_lock);
4498 rw_destroy(&ipst->ips_ill_g_lock);
4499
4500 kmem_free(ipst->ips_phyint_g_list, sizeof (phyint_list_t));
4501 ipst->ips_phyint_g_list = NULL;
4502 kmem_free(ipst->ips_ill_g_heads, sizeof (ill_g_head_t) * MAX_G_HEADS);
4503 ipst->ips_ill_g_heads = NULL;
4504
4505 ldi_ident_release(ipst->ips_ldi_ident);
4506 kmem_free(ipst, sizeof (*ipst));
4507 }
4508
4509 /*
4510 * This function is called from the TSD destructor, and is used to debug
4511 * reference count issues in IP. See block comment in <inet/ip_if.h> for
4512 * details.
4513 */
4514 static void
4515 ip_thread_exit(void *phash)
4516 {
4517 th_hash_t *thh = phash;
4518
4519 rw_enter(&ip_thread_rwlock, RW_WRITER);
4520 list_remove(&ip_thread_list, thh);
4521 rw_exit(&ip_thread_rwlock);
4522 mod_hash_destroy_hash(thh->thh_hash);
4523 kmem_free(thh, sizeof (*thh));
4524 }
4525
4526 /*
4527 * Called when the IP kernel module is loaded into the kernel
4528 */
4529 void
4530 ip_ddi_init(void)
4531 {
4532 ip_squeue_flag = ip_squeue_switch(ip_squeue_enter);
4533
4534 /*
4535 * For IP and TCP the minor numbers should start from 2 since we have 4
4536 * initial devices: ip, ip6, tcp, tcp6.
4537 */
4538 /*
4539 * If this is a 64-bit kernel, then create two separate arenas -
4540 * one for TLIs in the range of INET_MIN_DEV+2 through 2^^18-1, and the
4541 * other for socket apps in the range 2^^18 through 2^^32-1.
4542 */
4543 ip_minor_arena_la = NULL;
4544 ip_minor_arena_sa = NULL;
4545 #if defined(_LP64)
4546 if ((ip_minor_arena_sa = inet_minor_create("ip_minor_arena_sa",
4547 INET_MIN_DEV + 2, MAXMIN32, KM_SLEEP)) == NULL) {
4548 cmn_err(CE_PANIC,
4549 "ip_ddi_init: ip_minor_arena_sa creation failed\n");
4550 }
4551 if ((ip_minor_arena_la = inet_minor_create("ip_minor_arena_la",
4552 MAXMIN32 + 1, MAXMIN64, KM_SLEEP)) == NULL) {
4553 cmn_err(CE_PANIC,
4554 "ip_ddi_init: ip_minor_arena_la creation failed\n");
4555 }
4556 #else
4557 if ((ip_minor_arena_sa = inet_minor_create("ip_minor_arena_sa",
4558 INET_MIN_DEV + 2, MAXMIN, KM_SLEEP)) == NULL) {
4559 cmn_err(CE_PANIC,
4560 "ip_ddi_init: ip_minor_arena_sa creation failed\n");
4561 }
4562 #endif
4563 ip_poll_normal_ticks = MSEC_TO_TICK_ROUNDUP(ip_poll_normal_ms);
4564
4565 ipcl_g_init();
4566 ip_ire_g_init();
4567 ip_net_g_init();
4568
4569 #ifdef DEBUG
4570 tsd_create(&ip_thread_data, ip_thread_exit);
4571 rw_init(&ip_thread_rwlock, NULL, RW_DEFAULT, NULL);
4572 list_create(&ip_thread_list, sizeof (th_hash_t),
4573 offsetof(th_hash_t, thh_link));
4574 #endif
4575 ipsec_policy_g_init();
4576 tcp_ddi_g_init();
4577 sctp_ddi_g_init();
4578 dce_g_init();
4579
4580 /*
4581 * We want to be informed each time a stack is created or
4582 * destroyed in the kernel, so we can maintain the
4583 * set of udp_stack_t's.
4584 */
4585 netstack_register(NS_IP, ip_stack_init, ip_stack_shutdown,
4586 ip_stack_fini);
4587
4588 tnet_init();
4589
4590 udp_ddi_g_init();
4591 rts_ddi_g_init();
4592 icmp_ddi_g_init();
4593 ilb_ddi_g_init();
4594
4595 /* This needs to be called after all transports are initialized. */
4596 mutex_enter(&cpu_lock);
4597 register_cpu_setup_func(ip_tp_cpu_update, NULL);
4598 mutex_exit(&cpu_lock);
4599 }
4600
4601 /*
4602 * Initialize the IP stack instance.
4603 */
4604 static void *
4605 ip_stack_init(netstackid_t stackid, netstack_t *ns)
4606 {
4607 ip_stack_t *ipst;
4608 size_t arrsz;
4609 major_t major;
4610
4611 #ifdef NS_DEBUG
4612 printf("ip_stack_init(stack %d)\n", stackid);
4613 #endif
4614
4615 ipst = (ip_stack_t *)kmem_zalloc(sizeof (*ipst), KM_SLEEP);
4616 ipst->ips_netstack = ns;
4617
4618 ipst->ips_ill_g_heads = kmem_zalloc(sizeof (ill_g_head_t) * MAX_G_HEADS,
4619 KM_SLEEP);
4620 ipst->ips_phyint_g_list = kmem_zalloc(sizeof (phyint_list_t),
4621 KM_SLEEP);
4622 ipst->ips_ndp4 = kmem_zalloc(sizeof (ndp_g_t), KM_SLEEP);
4623 ipst->ips_ndp6 = kmem_zalloc(sizeof (ndp_g_t), KM_SLEEP);
4624 mutex_init(&ipst->ips_ndp4->ndp_g_lock, NULL, MUTEX_DEFAULT, NULL);
4625 mutex_init(&ipst->ips_ndp6->ndp_g_lock, NULL, MUTEX_DEFAULT, NULL);
4626
4627 mutex_init(&ipst->ips_igmp_timer_lock, NULL, MUTEX_DEFAULT, NULL);
4628 ipst->ips_igmp_deferred_next = INFINITY;
4629 mutex_init(&ipst->ips_mld_timer_lock, NULL, MUTEX_DEFAULT, NULL);
4630 ipst->ips_mld_deferred_next = INFINITY;
4631 mutex_init(&ipst->ips_igmp_slowtimeout_lock, NULL, MUTEX_DEFAULT, NULL);
4632 mutex_init(&ipst->ips_mld_slowtimeout_lock, NULL, MUTEX_DEFAULT, NULL);
4633 mutex_init(&ipst->ips_ip_mi_lock, NULL, MUTEX_DEFAULT, NULL);
4634 mutex_init(&ipst->ips_ip_addr_avail_lock, NULL, MUTEX_DEFAULT, NULL);
4635 rw_init(&ipst->ips_ill_g_lock, NULL, RW_DEFAULT, NULL);
4636 rw_init(&ipst->ips_ill_g_usesrc_lock, NULL, RW_DEFAULT, NULL);
4637
4638 ipcl_init(ipst);
4639 ip_ire_init(ipst);
4640 ip6_asp_init(ipst);
4641 ipif_init(ipst);
4642 conn_drain_init(ipst);
4643 ip_mrouter_stack_init(ipst);
4644 dce_stack_init(ipst);
4645
4646 ipst->ips_ip_multirt_log_interval = 1000;
4647
4648 ipst->ips_ill_index = 1;
4649
4650 ipst->ips_saved_ip_forwarding = -1;
4651 ipst->ips_reg_vif_num = ALL_VIFS; /* Index to Register vif */
4652
4653 arrsz = ip_propinfo_count * sizeof (mod_prop_info_t);
4654 ipst->ips_propinfo_tbl = (mod_prop_info_t *)kmem_alloc(arrsz, KM_SLEEP);
4655 bcopy(ip_propinfo_tbl, ipst->ips_propinfo_tbl, arrsz);
4656
4657 ipst->ips_ip_mibkp = ip_kstat_init(stackid, ipst);
4658 ipst->ips_icmp_mibkp = icmp_kstat_init(stackid);
4659 ipst->ips_ip_kstat = ip_kstat2_init(stackid, &ipst->ips_ip_statistics);
4660 ipst->ips_ip6_kstat =
4661 ip6_kstat_init(stackid, &ipst->ips_ip6_statistics);
4662
4663 ipst->ips_ip_src_id = 1;
4664 rw_init(&ipst->ips_srcid_lock, NULL, RW_DEFAULT, NULL);
4665
4666 ipst->ips_src_generation = SRC_GENERATION_INITIAL;
4667
4668 ip_net_init(ipst, ns);
4669 ipv4_hook_init(ipst);
4670 ipv6_hook_init(ipst);
4671 arp_hook_init(ipst);
4672 ipmp_init(ipst);
4673 ipobs_init(ipst);
4674
4675 /*
4676 * Create the taskq dispatcher thread and initialize related stuff.
4677 */
4678 mutex_init(&ipst->ips_capab_taskq_lock, NULL, MUTEX_DEFAULT, NULL);
4679 cv_init(&ipst->ips_capab_taskq_cv, NULL, CV_DEFAULT, NULL);
4680 ipst->ips_capab_taskq_thread = thread_create(NULL, 0,
4681 ill_taskq_dispatch, ipst, 0, &p0, TS_RUN, minclsyspri);
4682
4683 major = mod_name_to_major(INET_NAME);
4684 (void) ldi_ident_from_major(major, &ipst->ips_ldi_ident);
4685 return (ipst);
4686 }
4687
4688 /*
4689 * Allocate and initialize a DLPI template of the specified length. (May be
4690 * called as writer.)
4691 */
4692 mblk_t *
4693 ip_dlpi_alloc(size_t len, t_uscalar_t prim)
4694 {
4695 mblk_t *mp;
4696
4697 mp = allocb(len, BPRI_MED);
4698 if (!mp)
4699 return (NULL);
4700
4701 /*
4702 * DLPIv2 says that DL_INFO_REQ and DL_TOKEN_REQ (the latter
4703 * of which we don't seem to use) are sent with M_PCPROTO, and
4704 * that other DLPI are M_PROTO.
4705 */
4706 if (prim == DL_INFO_REQ) {
4707 mp->b_datap->db_type = M_PCPROTO;
4708 } else {
4709 mp->b_datap->db_type = M_PROTO;
4710 }
4711
4712 mp->b_wptr = mp->b_rptr + len;
4713 bzero(mp->b_rptr, len);
4714 ((dl_unitdata_req_t *)mp->b_rptr)->dl_primitive = prim;
4715 return (mp);
4716 }
4717
4718 /*
4719 * Allocate and initialize a DLPI notification. (May be called as writer.)
4720 */
4721 mblk_t *
4722 ip_dlnotify_alloc(uint_t notification, uint_t data)
4723 {
4724 dl_notify_ind_t *notifyp;
4725 mblk_t *mp;
4726
4727 if ((mp = ip_dlpi_alloc(DL_NOTIFY_IND_SIZE, DL_NOTIFY_IND)) == NULL)
4728 return (NULL);
4729
4730 notifyp = (dl_notify_ind_t *)mp->b_rptr;
4731 notifyp->dl_notification = notification;
4732 notifyp->dl_data = data;
4733 return (mp);
4734 }
4735
4736 mblk_t *
4737 ip_dlnotify_alloc2(uint_t notification, uint_t data1, uint_t data2)
4738 {
4739 dl_notify_ind_t *notifyp;
4740 mblk_t *mp;
4741
4742 if ((mp = ip_dlpi_alloc(DL_NOTIFY_IND_SIZE, DL_NOTIFY_IND)) == NULL)
4743 return (NULL);
4744
4745 notifyp = (dl_notify_ind_t *)mp->b_rptr;
4746 notifyp->dl_notification = notification;
4747 notifyp->dl_data1 = data1;
4748 notifyp->dl_data2 = data2;
4749 return (mp);
4750 }
4751
4752 /*
4753 * Debug formatting routine. Returns a character string representation of the
4754 * addr in buf, of the form xxx.xxx.xxx.xxx. This routine takes the address
4755 * in the form of a ipaddr_t and calls ip_dot_saddr with a pointer.
4756 *
4757 * Once the ndd table-printing interfaces are removed, this can be changed to
4758 * standard dotted-decimal form.
4759 */
4760 char *
4761 ip_dot_addr(ipaddr_t addr, char *buf)
4762 {
4763 uint8_t *ap = (uint8_t *)&addr;
4764
4765 (void) mi_sprintf(buf, "%03d.%03d.%03d.%03d",
4766 ap[0] & 0xFF, ap[1] & 0xFF, ap[2] & 0xFF, ap[3] & 0xFF);
4767 return (buf);
4768 }
4769
4770 /*
4771 * Write the given MAC address as a printable string in the usual colon-
4772 * separated format.
4773 */
4774 const char *
4775 mac_colon_addr(const uint8_t *addr, size_t alen, char *buf, size_t buflen)
4776 {
4777 char *bp;
4778
4779 if (alen == 0 || buflen < 4)
4780 return ("?");
4781 bp = buf;
4782 for (;;) {
4783 /*
4784 * If there are more MAC address bytes available, but we won't
4785 * have any room to print them, then add "..." to the string
4786 * instead. See below for the 'magic number' explanation.
4787 */
4788 if ((alen == 2 && buflen < 6) || (alen > 2 && buflen < 7)) {
4789 (void) strcpy(bp, "...");
4790 break;
4791 }
4792 (void) sprintf(bp, "%02x", *addr++);
4793 bp += 2;
4794 if (--alen == 0)
4795 break;
4796 *bp++ = ':';
4797 buflen -= 3;
4798 /*
4799 * At this point, based on the first 'if' statement above,
4800 * either alen == 1 and buflen >= 3, or alen > 1 and
4801 * buflen >= 4. The first case leaves room for the final "xx"
4802 * number and trailing NUL byte. The second leaves room for at
4803 * least "...". Thus the apparently 'magic' numbers chosen for
4804 * that statement.
4805 */
4806 }
4807 return (buf);
4808 }
4809
4810 /*
4811 * Called when it is conceptually a ULP that would sent the packet
4812 * e.g., port unreachable and protocol unreachable. Check that the packet
4813 * would have passed the IPsec global policy before sending the error.
4814 *
4815 * Send an ICMP error after patching up the packet appropriately.
4816 * Uses ip_drop_input and bumps the appropriate MIB.
4817 */
4818 void
4819 ip_fanout_send_icmp_v4(mblk_t *mp, uint_t icmp_type, uint_t icmp_code,
4820 ip_recv_attr_t *ira)
4821 {
4822 ipha_t *ipha;
4823 boolean_t secure;
4824 ill_t *ill = ira->ira_ill;
4825 ip_stack_t *ipst = ill->ill_ipst;
4826 netstack_t *ns = ipst->ips_netstack;
4827 ipsec_stack_t *ipss = ns->netstack_ipsec;
4828
4829 secure = ira->ira_flags & IRAF_IPSEC_SECURE;
4830
4831 /*
4832 * We are generating an icmp error for some inbound packet.
4833 * Called from all ip_fanout_(udp, tcp, proto) functions.
4834 * Before we generate an error, check with global policy
4835 * to see whether this is allowed to enter the system. As
4836 * there is no "conn", we are checking with global policy.
4837 */
4838 ipha = (ipha_t *)mp->b_rptr;
4839 if (secure || ipss->ipsec_inbound_v4_policy_present) {
4840 mp = ipsec_check_global_policy(mp, NULL, ipha, NULL, ira, ns);
4841 if (mp == NULL)
4842 return;
4843 }
4844
4845 /* We never send errors for protocols that we do implement */
4846 if (ira->ira_protocol == IPPROTO_ICMP ||
4847 ira->ira_protocol == IPPROTO_IGMP) {
4848 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
4849 ip_drop_input("ip_fanout_send_icmp_v4", mp, ill);
4850 freemsg(mp);
4851 return;
4852 }
4853 /*
4854 * Have to correct checksum since
4855 * the packet might have been
4856 * fragmented and the reassembly code in ip_rput
4857 * does not restore the IP checksum.
4858 */
4859 ipha->ipha_hdr_checksum = 0;
4860 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
4861
4862 switch (icmp_type) {
4863 case ICMP_DEST_UNREACHABLE:
4864 switch (icmp_code) {
4865 case ICMP_PROTOCOL_UNREACHABLE:
4866 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInUnknownProtos);
4867 ip_drop_input("ipIfStatsInUnknownProtos", mp, ill);
4868 break;
4869 case ICMP_PORT_UNREACHABLE:
4870 BUMP_MIB(ill->ill_ip_mib, udpIfStatsNoPorts);
4871 ip_drop_input("ipIfStatsNoPorts", mp, ill);
4872 break;
4873 }
4874
4875 icmp_unreachable(mp, icmp_code, ira);
4876 break;
4877 default:
4878 #ifdef DEBUG
4879 panic("ip_fanout_send_icmp_v4: wrong type");
4880 /*NOTREACHED*/
4881 #else
4882 freemsg(mp);
4883 break;
4884 #endif
4885 }
4886 }
4887
4888 /*
4889 * Used to send an ICMP error message when a packet is received for
4890 * a protocol that is not supported. The mblk passed as argument
4891 * is consumed by this function.
4892 */
4893 void
4894 ip_proto_not_sup(mblk_t *mp, ip_recv_attr_t *ira)
4895 {
4896 ipha_t *ipha;
4897
4898 ipha = (ipha_t *)mp->b_rptr;
4899 if (ira->ira_flags & IRAF_IS_IPV4) {
4900 ASSERT(IPH_HDR_VERSION(ipha) == IP_VERSION);
4901 ip_fanout_send_icmp_v4(mp, ICMP_DEST_UNREACHABLE,
4902 ICMP_PROTOCOL_UNREACHABLE, ira);
4903 } else {
4904 ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
4905 ip_fanout_send_icmp_v6(mp, ICMP6_PARAM_PROB,
4906 ICMP6_PARAMPROB_NEXTHEADER, ira);
4907 }
4908 }
4909
4910 /*
4911 * Deliver a rawip packet to the given conn, possibly applying ipsec policy.
4912 * Handles IPv4 and IPv6.
4913 * We are responsible for disposing of mp, such as by freemsg() or putnext()
4914 * Caller is responsible for dropping references to the conn.
4915 */
4916 void
4917 ip_fanout_proto_conn(conn_t *connp, mblk_t *mp, ipha_t *ipha, ip6_t *ip6h,
4918 ip_recv_attr_t *ira)
4919 {
4920 ill_t *ill = ira->ira_ill;
4921 ip_stack_t *ipst = ill->ill_ipst;
4922 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
4923 boolean_t secure;
4924 uint_t protocol = ira->ira_protocol;
4925 iaflags_t iraflags = ira->ira_flags;
4926 queue_t *rq;
4927
4928 secure = iraflags & IRAF_IPSEC_SECURE;
4929
4930 rq = connp->conn_rq;
4931 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld : !canputnext(rq)) {
4932 switch (protocol) {
4933 case IPPROTO_ICMPV6:
4934 BUMP_MIB(ill->ill_icmp6_mib, ipv6IfIcmpInOverflows);
4935 break;
4936 case IPPROTO_ICMP:
4937 BUMP_MIB(&ipst->ips_icmp_mib, icmpInOverflows);
4938 break;
4939 default:
4940 BUMP_MIB(ill->ill_ip_mib, rawipIfStatsInOverflows);
4941 break;
4942 }
4943 freemsg(mp);
4944 return;
4945 }
4946
4947 ASSERT(!(IPCL_IS_IPTUN(connp)));
4948
4949 if (((iraflags & IRAF_IS_IPV4) ?
4950 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
4951 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
4952 secure) {
4953 mp = ipsec_check_inbound_policy(mp, connp, ipha,
4954 ip6h, ira);
4955 if (mp == NULL) {
4956 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
4957 /* Note that mp is NULL */
4958 ip_drop_input("ipIfStatsInDiscards", mp, ill);
4959 return;
4960 }
4961 }
4962
4963 if (iraflags & IRAF_ICMP_ERROR) {
4964 (connp->conn_recvicmp)(connp, mp, NULL, ira);
4965 } else {
4966 ill_t *rill = ira->ira_rill;
4967
4968 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
4969 ira->ira_ill = ira->ira_rill = NULL;
4970 /* Send it upstream */
4971 (connp->conn_recv)(connp, mp, NULL, ira);
4972 ira->ira_ill = ill;
4973 ira->ira_rill = rill;
4974 }
4975 }
4976
4977 /*
4978 * Handle protocols with which IP is less intimate. There
4979 * can be more than one stream bound to a particular
4980 * protocol. When this is the case, normally each one gets a copy
4981 * of any incoming packets.
4982 *
4983 * IPsec NOTE :
4984 *
4985 * Don't allow a secure packet going up a non-secure connection.
4986 * We don't allow this because
4987 *
4988 * 1) Reply might go out in clear which will be dropped at
4989 * the sending side.
4990 * 2) If the reply goes out in clear it will give the
4991 * adversary enough information for getting the key in
4992 * most of the cases.
4993 *
4994 * Moreover getting a secure packet when we expect clear
4995 * implies that SA's were added without checking for
4996 * policy on both ends. This should not happen once ISAKMP
4997 * is used to negotiate SAs as SAs will be added only after
4998 * verifying the policy.
4999 *
5000 * Zones notes:
5001 * Earlier in ip_input on a system with multiple shared-IP zones we
5002 * duplicate the multicast and broadcast packets and send them up
5003 * with each explicit zoneid that exists on that ill.
5004 * This means that here we can match the zoneid with SO_ALLZONES being special.
5005 */
5006 void
5007 ip_fanout_proto_v4(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
5008 {
5009 mblk_t *mp1;
5010 ipaddr_t laddr;
5011 conn_t *connp, *first_connp, *next_connp;
5012 connf_t *connfp;
5013 ill_t *ill = ira->ira_ill;
5014 ip_stack_t *ipst = ill->ill_ipst;
5015
5016 laddr = ipha->ipha_dst;
5017
5018 connfp = &ipst->ips_ipcl_proto_fanout_v4[ira->ira_protocol];
5019 mutex_enter(&connfp->connf_lock);
5020 connp = connfp->connf_head;
5021 for (connp = connfp->connf_head; connp != NULL;
5022 connp = connp->conn_next) {
5023 /* Note: IPCL_PROTO_MATCH includes conn_wantpacket */
5024 if (IPCL_PROTO_MATCH(connp, ira, ipha) &&
5025 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5026 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp))) {
5027 break;
5028 }
5029 }
5030
5031 if (connp == NULL) {
5032 /*
5033 * No one bound to these addresses. Is
5034 * there a client that wants all
5035 * unclaimed datagrams?
5036 */
5037 mutex_exit(&connfp->connf_lock);
5038 ip_fanout_send_icmp_v4(mp, ICMP_DEST_UNREACHABLE,
5039 ICMP_PROTOCOL_UNREACHABLE, ira);
5040 return;
5041 }
5042
5043 ASSERT(IPCL_IS_NONSTR(connp) || connp->conn_rq != NULL);
5044
5045 CONN_INC_REF(connp);
5046 first_connp = connp;
5047 connp = connp->conn_next;
5048
5049 for (;;) {
5050 while (connp != NULL) {
5051 /* Note: IPCL_PROTO_MATCH includes conn_wantpacket */
5052 if (IPCL_PROTO_MATCH(connp, ira, ipha) &&
5053 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5054 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5055 ira, connp)))
5056 break;
5057 connp = connp->conn_next;
5058 }
5059
5060 if (connp == NULL) {
5061 /* No more interested clients */
5062 connp = first_connp;
5063 break;
5064 }
5065 if (((mp1 = dupmsg(mp)) == NULL) &&
5066 ((mp1 = copymsg(mp)) == NULL)) {
5067 /* Memory allocation failed */
5068 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5069 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5070 connp = first_connp;
5071 break;
5072 }
5073
5074 CONN_INC_REF(connp);
5075 mutex_exit(&connfp->connf_lock);
5076
5077 ip_fanout_proto_conn(connp, mp1, (ipha_t *)mp1->b_rptr, NULL,
5078 ira);
5079
5080 mutex_enter(&connfp->connf_lock);
5081 /* Follow the next pointer before releasing the conn. */
5082 next_connp = connp->conn_next;
5083 CONN_DEC_REF(connp);
5084 connp = next_connp;
5085 }
5086
5087 /* Last one. Send it upstream. */
5088 mutex_exit(&connfp->connf_lock);
5089
5090 ip_fanout_proto_conn(connp, mp, ipha, NULL, ira);
5091
5092 CONN_DEC_REF(connp);
5093 }
5094
5095 /*
5096 * If we have a IPsec NAT-Traversal packet, strip the zero-SPI or
5097 * pass it along to ESP if the SPI is non-zero. Returns the mblk if the mblk
5098 * is not consumed.
5099 *
5100 * One of three things can happen, all of which affect the passed-in mblk:
5101 *
5102 * 1.) The packet is stock UDP and gets its zero-SPI stripped. Return mblk..
5103 *
5104 * 2.) The packet is ESP-in-UDP, gets transformed into an equivalent
5105 * ESP packet, and is passed along to ESP for consumption. Return NULL.
5106 *
5107 * 3.) The packet is an ESP-in-UDP Keepalive. Drop it and return NULL.
5108 */
5109 mblk_t *
5110 zero_spi_check(mblk_t *mp, ip_recv_attr_t *ira)
5111 {
5112 int shift, plen, iph_len;
5113 ipha_t *ipha;
5114 udpha_t *udpha;
5115 uint32_t *spi;
5116 uint32_t esp_ports;
5117 uint8_t *orptr;
5118 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
5119 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
5120
5121 ipha = (ipha_t *)mp->b_rptr;
5122 iph_len = ira->ira_ip_hdr_length;
5123 plen = ira->ira_pktlen;
5124
5125 if (plen - iph_len - sizeof (udpha_t) < sizeof (uint32_t)) {
5126 /*
5127 * Most likely a keepalive for the benefit of an intervening
5128 * NAT. These aren't for us, per se, so drop it.
5129 *
5130 * RFC 3947/8 doesn't say for sure what to do for 2-3
5131 * byte packets (keepalives are 1-byte), but we'll drop them
5132 * also.
5133 */
5134 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
5135 DROPPER(ipss, ipds_esp_nat_t_ka), &ipss->ipsec_dropper);
5136 return (NULL);
5137 }
5138
5139 if (MBLKL(mp) < iph_len + sizeof (udpha_t) + sizeof (*spi)) {
5140 /* might as well pull it all up - it might be ESP. */
5141 if (!pullupmsg(mp, -1)) {
5142 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
5143 DROPPER(ipss, ipds_esp_nomem),
5144 &ipss->ipsec_dropper);
5145 return (NULL);
5146 }
5147
5148 ipha = (ipha_t *)mp->b_rptr;
5149 }
5150 spi = (uint32_t *)(mp->b_rptr + iph_len + sizeof (udpha_t));
5151 if (*spi == 0) {
5152 /* UDP packet - remove 0-spi. */
5153 shift = sizeof (uint32_t);
5154 } else {
5155 /* ESP-in-UDP packet - reduce to ESP. */
5156 ipha->ipha_protocol = IPPROTO_ESP;
5157 shift = sizeof (udpha_t);
5158 }
5159
5160 /* Fix IP header */
5161 ira->ira_pktlen = (plen - shift);
5162 ipha->ipha_length = htons(ira->ira_pktlen);
5163 ipha->ipha_hdr_checksum = 0;
5164
5165 orptr = mp->b_rptr;
5166 mp->b_rptr += shift;
5167
5168 udpha = (udpha_t *)(orptr + iph_len);
5169 if (*spi == 0) {
5170 ASSERT((uint8_t *)ipha == orptr);
5171 udpha->uha_length = htons(plen - shift - iph_len);
5172 iph_len += sizeof (udpha_t); /* For the call to ovbcopy(). */
5173 esp_ports = 0;
5174 } else {
5175 esp_ports = *((uint32_t *)udpha);
5176 ASSERT(esp_ports != 0);
5177 }
5178 ovbcopy(orptr, orptr + shift, iph_len);
5179 if (esp_ports != 0) /* Punt up for ESP processing. */ {
5180 ipha = (ipha_t *)(orptr + shift);
5181
5182 ira->ira_flags |= IRAF_ESP_UDP_PORTS;
5183 ira->ira_esp_udp_ports = esp_ports;
5184 ip_fanout_v4(mp, ipha, ira);
5185 return (NULL);
5186 }
5187 return (mp);
5188 }
5189
5190 /*
5191 * Deliver a udp packet to the given conn, possibly applying ipsec policy.
5192 * Handles IPv4 and IPv6.
5193 * We are responsible for disposing of mp, such as by freemsg() or putnext()
5194 * Caller is responsible for dropping references to the conn.
5195 */
5196 void
5197 ip_fanout_udp_conn(conn_t *connp, mblk_t *mp, ipha_t *ipha, ip6_t *ip6h,
5198 ip_recv_attr_t *ira)
5199 {
5200 ill_t *ill = ira->ira_ill;
5201 ip_stack_t *ipst = ill->ill_ipst;
5202 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
5203 boolean_t secure;
5204 iaflags_t iraflags = ira->ira_flags;
5205
5206 secure = iraflags & IRAF_IPSEC_SECURE;
5207
5208 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld :
5209 !canputnext(connp->conn_rq)) {
5210 BUMP_MIB(ill->ill_ip_mib, udpIfStatsInOverflows);
5211 freemsg(mp);
5212 return;
5213 }
5214
5215 if (((iraflags & IRAF_IS_IPV4) ?
5216 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
5217 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
5218 secure) {
5219 mp = ipsec_check_inbound_policy(mp, connp, ipha,
5220 ip6h, ira);
5221 if (mp == NULL) {
5222 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5223 /* Note that mp is NULL */
5224 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5225 return;
5226 }
5227 }
5228
5229 /*
5230 * Since this code is not used for UDP unicast we don't need a NAT_T
5231 * check. Only ip_fanout_v4 has that check.
5232 */
5233 if (ira->ira_flags & IRAF_ICMP_ERROR) {
5234 (connp->conn_recvicmp)(connp, mp, NULL, ira);
5235 } else {
5236 ill_t *rill = ira->ira_rill;
5237
5238 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
5239 ira->ira_ill = ira->ira_rill = NULL;
5240 /* Send it upstream */
5241 (connp->conn_recv)(connp, mp, NULL, ira);
5242 ira->ira_ill = ill;
5243 ira->ira_rill = rill;
5244 }
5245 }
5246
5247 /*
5248 * Fanout for UDP packets that are multicast or broadcast, and ICMP errors.
5249 * (Unicast fanout is handled in ip_input_v4.)
5250 *
5251 * If SO_REUSEADDR is set all multicast and broadcast packets
5252 * will be delivered to all conns bound to the same port.
5253 *
5254 * If there is at least one matching AF_INET receiver, then we will
5255 * ignore any AF_INET6 receivers.
5256 * In the special case where an AF_INET socket binds to 0.0.0.0/<port> and an
5257 * AF_INET6 socket binds to ::/<port>, only the AF_INET socket receives the IPv4
5258 * packets.
5259 *
5260 * Zones notes:
5261 * Earlier in ip_input on a system with multiple shared-IP zones we
5262 * duplicate the multicast and broadcast packets and send them up
5263 * with each explicit zoneid that exists on that ill.
5264 * This means that here we can match the zoneid with SO_ALLZONES being special.
5265 */
5266 void
5267 ip_fanout_udp_multi_v4(mblk_t *mp, ipha_t *ipha, uint16_t lport, uint16_t fport,
5268 ip_recv_attr_t *ira)
5269 {
5270 ipaddr_t laddr;
5271 in6_addr_t v6faddr;
5272 conn_t *connp;
5273 connf_t *connfp;
5274 ipaddr_t faddr;
5275 ill_t *ill = ira->ira_ill;
5276 ip_stack_t *ipst = ill->ill_ipst;
5277
5278 ASSERT(ira->ira_flags & (IRAF_MULTIBROADCAST|IRAF_ICMP_ERROR));
5279
5280 laddr = ipha->ipha_dst;
5281 faddr = ipha->ipha_src;
5282
5283 connfp = &ipst->ips_ipcl_udp_fanout[IPCL_UDP_HASH(lport, ipst)];
5284 mutex_enter(&connfp->connf_lock);
5285 connp = connfp->connf_head;
5286
5287 /*
5288 * If SO_REUSEADDR has been set on the first we send the
5289 * packet to all clients that have joined the group and
5290 * match the port.
5291 */
5292 while (connp != NULL) {
5293 if ((IPCL_UDP_MATCH(connp, lport, laddr, fport, faddr)) &&
5294 conn_wantpacket(connp, ira, ipha) &&
5295 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5296 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp)))
5297 break;
5298 connp = connp->conn_next;
5299 }
5300
5301 if (connp == NULL)
5302 goto notfound;
5303
5304 CONN_INC_REF(connp);
5305
5306 if (connp->conn_reuseaddr) {
5307 conn_t *first_connp = connp;
5308 conn_t *next_connp;
5309 mblk_t *mp1;
5310
5311 connp = connp->conn_next;
5312 for (;;) {
5313 while (connp != NULL) {
5314 if (IPCL_UDP_MATCH(connp, lport, laddr,
5315 fport, faddr) &&
5316 conn_wantpacket(connp, ira, ipha) &&
5317 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5318 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5319 ira, connp)))
5320 break;
5321 connp = connp->conn_next;
5322 }
5323 if (connp == NULL) {
5324 /* No more interested clients */
5325 connp = first_connp;
5326 break;
5327 }
5328 if (((mp1 = dupmsg(mp)) == NULL) &&
5329 ((mp1 = copymsg(mp)) == NULL)) {
5330 /* Memory allocation failed */
5331 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5332 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5333 connp = first_connp;
5334 break;
5335 }
5336 CONN_INC_REF(connp);
5337 mutex_exit(&connfp->connf_lock);
5338
5339 IP_STAT(ipst, ip_udp_fanmb);
5340 ip_fanout_udp_conn(connp, mp1, (ipha_t *)mp1->b_rptr,
5341 NULL, ira);
5342 mutex_enter(&connfp->connf_lock);
5343 /* Follow the next pointer before releasing the conn */
5344 next_connp = connp->conn_next;
5345 CONN_DEC_REF(connp);
5346 connp = next_connp;
5347 }
5348 }
5349
5350 /* Last one. Send it upstream. */
5351 mutex_exit(&connfp->connf_lock);
5352 IP_STAT(ipst, ip_udp_fanmb);
5353 ip_fanout_udp_conn(connp, mp, ipha, NULL, ira);
5354 CONN_DEC_REF(connp);
5355 return;
5356
5357 notfound:
5358 mutex_exit(&connfp->connf_lock);
5359 /*
5360 * IPv6 endpoints bound to multicast IPv4-mapped addresses
5361 * have already been matched above, since they live in the IPv4
5362 * fanout tables. This implies we only need to
5363 * check for IPv6 in6addr_any endpoints here.
5364 * Thus we compare using ipv6_all_zeros instead of the destination
5365 * address, except for the multicast group membership lookup which
5366 * uses the IPv4 destination.
5367 */
5368 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_src, &v6faddr);
5369 connfp = &ipst->ips_ipcl_udp_fanout[IPCL_UDP_HASH(lport, ipst)];
5370 mutex_enter(&connfp->connf_lock);
5371 connp = connfp->connf_head;
5372 /*
5373 * IPv4 multicast packet being delivered to an AF_INET6
5374 * in6addr_any endpoint.
5375 * Need to check conn_wantpacket(). Note that we use conn_wantpacket()
5376 * and not conn_wantpacket_v6() since any multicast membership is
5377 * for an IPv4-mapped multicast address.
5378 */
5379 while (connp != NULL) {
5380 if (IPCL_UDP_MATCH_V6(connp, lport, ipv6_all_zeros,
5381 fport, v6faddr) &&
5382 conn_wantpacket(connp, ira, ipha) &&
5383 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5384 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp)))
5385 break;
5386 connp = connp->conn_next;
5387 }
5388
5389 if (connp == NULL) {
5390 /*
5391 * No one bound to this port. Is
5392 * there a client that wants all
5393 * unclaimed datagrams?
5394 */
5395 mutex_exit(&connfp->connf_lock);
5396
5397 if (ipst->ips_ipcl_proto_fanout_v4[IPPROTO_UDP].connf_head !=
5398 NULL) {
5399 ASSERT(ira->ira_protocol == IPPROTO_UDP);
5400 ip_fanout_proto_v4(mp, ipha, ira);
5401 } else {
5402 /*
5403 * We used to attempt to send an icmp error here, but
5404 * since this is known to be a multicast packet
5405 * and we don't send icmp errors in response to
5406 * multicast, just drop the packet and give up sooner.
5407 */
5408 BUMP_MIB(ill->ill_ip_mib, udpIfStatsNoPorts);
5409 freemsg(mp);
5410 }
5411 return;
5412 }
5413 CONN_INC_REF(connp);
5414 ASSERT(IPCL_IS_NONSTR(connp) || connp->conn_rq != NULL);
5415
5416 /*
5417 * If SO_REUSEADDR has been set on the first we send the
5418 * packet to all clients that have joined the group and
5419 * match the port.
5420 */
5421 if (connp->conn_reuseaddr) {
5422 conn_t *first_connp = connp;
5423 conn_t *next_connp;
5424 mblk_t *mp1;
5425
5426 connp = connp->conn_next;
5427 for (;;) {
5428 while (connp != NULL) {
5429 if (IPCL_UDP_MATCH_V6(connp, lport,
5430 ipv6_all_zeros, fport, v6faddr) &&
5431 conn_wantpacket(connp, ira, ipha) &&
5432 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5433 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5434 ira, connp)))
5435 break;
5436 connp = connp->conn_next;
5437 }
5438 if (connp == NULL) {
5439 /* No more interested clients */
5440 connp = first_connp;
5441 break;
5442 }
5443 if (((mp1 = dupmsg(mp)) == NULL) &&
5444 ((mp1 = copymsg(mp)) == NULL)) {
5445 /* Memory allocation failed */
5446 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5447 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5448 connp = first_connp;
5449 break;
5450 }
5451 CONN_INC_REF(connp);
5452 mutex_exit(&connfp->connf_lock);
5453
5454 IP_STAT(ipst, ip_udp_fanmb);
5455 ip_fanout_udp_conn(connp, mp1, (ipha_t *)mp1->b_rptr,
5456 NULL, ira);
5457 mutex_enter(&connfp->connf_lock);
5458 /* Follow the next pointer before releasing the conn */
5459 next_connp = connp->conn_next;
5460 CONN_DEC_REF(connp);
5461 connp = next_connp;
5462 }
5463 }
5464
5465 /* Last one. Send it upstream. */
5466 mutex_exit(&connfp->connf_lock);
5467 IP_STAT(ipst, ip_udp_fanmb);
5468 ip_fanout_udp_conn(connp, mp, ipha, NULL, ira);
5469 CONN_DEC_REF(connp);
5470 }
5471
5472 /*
5473 * Split an incoming packet's IPv4 options into the label and the other options.
5474 * If 'allocate' is set it does memory allocation for the ip_pkt_t, including
5475 * clearing out any leftover label or options.
5476 * Otherwise it just makes ipp point into the packet.
5477 *
5478 * Returns zero if ok; ENOMEM if the buffer couldn't be allocated.
5479 */
5480 int
5481 ip_find_hdr_v4(ipha_t *ipha, ip_pkt_t *ipp, boolean_t allocate)
5482 {
5483 uchar_t *opt;
5484 uint32_t totallen;
5485 uint32_t optval;
5486 uint32_t optlen;
5487
5488 ipp->ipp_fields |= IPPF_HOPLIMIT | IPPF_TCLASS | IPPF_ADDR;
5489 ipp->ipp_hoplimit = ipha->ipha_ttl;
5490 ipp->ipp_type_of_service = ipha->ipha_type_of_service;
5491 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &ipp->ipp_addr);
5492
5493 /*
5494 * Get length (in 4 byte octets) of IP header options.
5495 */
5496 totallen = ipha->ipha_version_and_hdr_length -
5497 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
5498
5499 if (totallen == 0) {
5500 if (!allocate)
5501 return (0);
5502
5503 /* Clear out anything from a previous packet */
5504 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
5505 kmem_free(ipp->ipp_ipv4_options,
5506 ipp->ipp_ipv4_options_len);
5507 ipp->ipp_ipv4_options = NULL;
5508 ipp->ipp_ipv4_options_len = 0;
5509 ipp->ipp_fields &= ~IPPF_IPV4_OPTIONS;
5510 }
5511 if (ipp->ipp_fields & IPPF_LABEL_V4) {
5512 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
5513 ipp->ipp_label_v4 = NULL;
5514 ipp->ipp_label_len_v4 = 0;
5515 ipp->ipp_fields &= ~IPPF_LABEL_V4;
5516 }
5517 return (0);
5518 }
5519
5520 totallen <<= 2;
5521 opt = (uchar_t *)&ipha[1];
5522 if (!is_system_labeled()) {
5523
5524 copyall:
5525 if (!allocate) {
5526 if (totallen != 0) {
5527 ipp->ipp_ipv4_options = opt;
5528 ipp->ipp_ipv4_options_len = totallen;
5529 ipp->ipp_fields |= IPPF_IPV4_OPTIONS;
5530 }
5531 return (0);
5532 }
5533 /* Just copy all of options */
5534 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
5535 if (totallen == ipp->ipp_ipv4_options_len) {
5536 bcopy(opt, ipp->ipp_ipv4_options, totallen);
5537 return (0);
5538 }
5539 kmem_free(ipp->ipp_ipv4_options,
5540 ipp->ipp_ipv4_options_len);
5541 ipp->ipp_ipv4_options = NULL;
5542 ipp->ipp_ipv4_options_len = 0;
5543 ipp->ipp_fields &= ~IPPF_IPV4_OPTIONS;
5544 }
5545 if (totallen == 0)
5546 return (0);
5547
5548 ipp->ipp_ipv4_options = kmem_alloc(totallen, KM_NOSLEEP);
5549 if (ipp->ipp_ipv4_options == NULL)
5550 return (ENOMEM);
5551 ipp->ipp_ipv4_options_len = totallen;
5552 ipp->ipp_fields |= IPPF_IPV4_OPTIONS;
5553 bcopy(opt, ipp->ipp_ipv4_options, totallen);
5554 return (0);
5555 }
5556
5557 if (allocate && (ipp->ipp_fields & IPPF_LABEL_V4)) {
5558 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
5559 ipp->ipp_label_v4 = NULL;
5560 ipp->ipp_label_len_v4 = 0;
5561 ipp->ipp_fields &= ~IPPF_LABEL_V4;
5562 }
5563
5564 /*
5565 * Search for CIPSO option.
5566 * We assume CIPSO is first in options if it is present.
5567 * If it isn't, then ipp_opt_ipv4_options will not include the options
5568 * prior to the CIPSO option.
5569 */
5570 while (totallen != 0) {
5571 switch (optval = opt[IPOPT_OPTVAL]) {
5572 case IPOPT_EOL:
5573 return (0);
5574 case IPOPT_NOP:
5575 optlen = 1;
5576 break;
5577 default:
5578 if (totallen <= IPOPT_OLEN)
5579 return (EINVAL);
5580 optlen = opt[IPOPT_OLEN];
5581 if (optlen < 2)
5582 return (EINVAL);
5583 }
5584 if (optlen > totallen)
5585 return (EINVAL);
5586
5587 switch (optval) {
5588 case IPOPT_COMSEC:
5589 if (!allocate) {
5590 ipp->ipp_label_v4 = opt;
5591 ipp->ipp_label_len_v4 = optlen;
5592 ipp->ipp_fields |= IPPF_LABEL_V4;
5593 } else {
5594 ipp->ipp_label_v4 = kmem_alloc(optlen,
5595 KM_NOSLEEP);
5596 if (ipp->ipp_label_v4 == NULL)
5597 return (ENOMEM);
5598 ipp->ipp_label_len_v4 = optlen;
5599 ipp->ipp_fields |= IPPF_LABEL_V4;
5600 bcopy(opt, ipp->ipp_label_v4, optlen);
5601 }
5602 totallen -= optlen;
5603 opt += optlen;
5604
5605 /* Skip padding bytes until we get to a multiple of 4 */
5606 while ((totallen & 3) != 0 && opt[0] == IPOPT_NOP) {
5607 totallen--;
5608 opt++;
5609 }
5610 /* Remaining as ipp_ipv4_options */
5611 goto copyall;
5612 }
5613 totallen -= optlen;
5614 opt += optlen;
5615 }
5616 /* No CIPSO found; return everything as ipp_ipv4_options */
5617 totallen = ipha->ipha_version_and_hdr_length -
5618 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
5619 totallen <<= 2;
5620 opt = (uchar_t *)&ipha[1];
5621 goto copyall;
5622 }
5623
5624 /*
5625 * Efficient versions of lookup for an IRE when we only
5626 * match the address.
5627 * For RTF_REJECT or BLACKHOLE we return IRE_NOROUTE.
5628 * Does not handle multicast addresses.
5629 */
5630 uint_t
5631 ip_type_v4(ipaddr_t addr, ip_stack_t *ipst)
5632 {
5633 ire_t *ire;
5634 uint_t result;
5635
5636 ire = ire_ftable_lookup_simple_v4(addr, 0, ipst, NULL);
5637 ASSERT(ire != NULL);
5638 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))
5639 result = IRE_NOROUTE;
5640 else
5641 result = ire->ire_type;
5642 ire_refrele(ire);
5643 return (result);
5644 }
5645
5646 /*
5647 * Efficient versions of lookup for an IRE when we only
5648 * match the address.
5649 * For RTF_REJECT or BLACKHOLE we return IRE_NOROUTE.
5650 * Does not handle multicast addresses.
5651 */
5652 uint_t
5653 ip_type_v6(const in6_addr_t *addr, ip_stack_t *ipst)
5654 {
5655 ire_t *ire;
5656 uint_t result;
5657
5658 ire = ire_ftable_lookup_simple_v6(addr, 0, ipst, NULL);
5659 ASSERT(ire != NULL);
5660 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))
5661 result = IRE_NOROUTE;
5662 else
5663 result = ire->ire_type;
5664 ire_refrele(ire);
5665 return (result);
5666 }
5667
5668 /*
5669 * Nobody should be sending
5670 * packets up this stream
5671 */
5672 static int
5673 ip_lrput(queue_t *q, mblk_t *mp)
5674 {
5675 switch (mp->b_datap->db_type) {
5676 case M_FLUSH:
5677 /* Turn around */
5678 if (*mp->b_rptr & FLUSHW) {
5679 *mp->b_rptr &= ~FLUSHR;
5680 qreply(q, mp);
5681 return (0);
5682 }
5683 break;
5684 }
5685 freemsg(mp);
5686 return (0);
5687 }
5688
5689 /* Nobody should be sending packets down this stream */
5690 /* ARGSUSED */
5691 int
5692 ip_lwput(queue_t *q, mblk_t *mp)
5693 {
5694 freemsg(mp);
5695 return (0);
5696 }
5697
5698 /*
5699 * Move the first hop in any source route to ipha_dst and remove that part of
5700 * the source route. Called by other protocols. Errors in option formatting
5701 * are ignored - will be handled by ip_output_options. Return the final
5702 * destination (either ipha_dst or the last entry in a source route.)
5703 */
5704 ipaddr_t
5705 ip_massage_options(ipha_t *ipha, netstack_t *ns)
5706 {
5707 ipoptp_t opts;
5708 uchar_t *opt;
5709 uint8_t optval;
5710 uint8_t optlen;
5711 ipaddr_t dst;
5712 int i;
5713 ip_stack_t *ipst = ns->netstack_ip;
5714
5715 ip2dbg(("ip_massage_options\n"));
5716 dst = ipha->ipha_dst;
5717 for (optval = ipoptp_first(&opts, ipha);
5718 optval != IPOPT_EOL;
5719 optval = ipoptp_next(&opts)) {
5720 opt = opts.ipoptp_cur;
5721 switch (optval) {
5722 uint8_t off;
5723 case IPOPT_SSRR:
5724 case IPOPT_LSRR:
5725 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
5726 ip1dbg(("ip_massage_options: bad src route\n"));
5727 break;
5728 }
5729 optlen = opts.ipoptp_len;
5730 off = opt[IPOPT_OFFSET];
5731 off--;
5732 redo_srr:
5733 if (optlen < IP_ADDR_LEN ||
5734 off > optlen - IP_ADDR_LEN) {
5735 /* End of source route */
5736 ip1dbg(("ip_massage_options: end of SR\n"));
5737 break;
5738 }
5739 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
5740 ip1dbg(("ip_massage_options: next hop 0x%x\n",
5741 ntohl(dst)));
5742 /*
5743 * Check if our address is present more than
5744 * once as consecutive hops in source route.
5745 * XXX verify per-interface ip_forwarding
5746 * for source route?
5747 */
5748 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
5749 off += IP_ADDR_LEN;
5750 goto redo_srr;
5751 }
5752 if (dst == htonl(INADDR_LOOPBACK)) {
5753 ip1dbg(("ip_massage_options: loopback addr in "
5754 "source route!\n"));
5755 break;
5756 }
5757 /*
5758 * Update ipha_dst to be the first hop and remove the
5759 * first hop from the source route (by overwriting
5760 * part of the option with NOP options).
5761 */
5762 ipha->ipha_dst = dst;
5763 /* Put the last entry in dst */
5764 off = ((optlen - IP_ADDR_LEN - 3) & ~(IP_ADDR_LEN-1)) +
5765 3;
5766 bcopy(&opt[off], &dst, IP_ADDR_LEN);
5767
5768 ip1dbg(("ip_massage_options: last hop 0x%x\n",
5769 ntohl(dst)));
5770 /* Move down and overwrite */
5771 opt[IP_ADDR_LEN] = opt[0];
5772 opt[IP_ADDR_LEN+1] = opt[IPOPT_OLEN] - IP_ADDR_LEN;
5773 opt[IP_ADDR_LEN+2] = opt[IPOPT_OFFSET];
5774 for (i = 0; i < IP_ADDR_LEN; i++)
5775 opt[i] = IPOPT_NOP;
5776 break;
5777 }
5778 }
5779 return (dst);
5780 }
5781
5782 /*
5783 * Return the network mask
5784 * associated with the specified address.
5785 */
5786 ipaddr_t
5787 ip_net_mask(ipaddr_t addr)
5788 {
5789 uchar_t *up = (uchar_t *)&addr;
5790 ipaddr_t mask = 0;
5791 uchar_t *maskp = (uchar_t *)&mask;
5792
5793 #if defined(__i386) || defined(__amd64)
5794 #define TOTALLY_BRAIN_DAMAGED_C_COMPILER
5795 #endif
5796 #ifdef TOTALLY_BRAIN_DAMAGED_C_COMPILER
5797 maskp[0] = maskp[1] = maskp[2] = maskp[3] = 0;
5798 #endif
5799 if (CLASSD(addr)) {
5800 maskp[0] = 0xF0;
5801 return (mask);
5802 }
5803
5804 /* We assume Class E default netmask to be 32 */
5805 if (CLASSE(addr))
5806 return (0xffffffffU);
5807
5808 if (addr == 0)
5809 return (0);
5810 maskp[0] = 0xFF;
5811 if ((up[0] & 0x80) == 0)
5812 return (mask);
5813
5814 maskp[1] = 0xFF;
5815 if ((up[0] & 0xC0) == 0x80)
5816 return (mask);
5817
5818 maskp[2] = 0xFF;
5819 if ((up[0] & 0xE0) == 0xC0)
5820 return (mask);
5821
5822 /* Otherwise return no mask */
5823 return ((ipaddr_t)0);
5824 }
5825
5826 /* Name/Value Table Lookup Routine */
5827 char *
5828 ip_nv_lookup(nv_t *nv, int value)
5829 {
5830 if (!nv)
5831 return (NULL);
5832 for (; nv->nv_name; nv++) {
5833 if (nv->nv_value == value)
5834 return (nv->nv_name);
5835 }
5836 return ("unknown");
5837 }
5838
5839 static int
5840 ip_wait_for_info_ack(ill_t *ill)
5841 {
5842 int err;
5843
5844 mutex_enter(&ill->ill_lock);
5845 while (ill->ill_state_flags & ILL_LL_SUBNET_PENDING) {
5846 /*
5847 * Return value of 0 indicates a pending signal.
5848 */
5849 err = cv_wait_sig(&ill->ill_cv, &ill->ill_lock);
5850 if (err == 0) {
5851 mutex_exit(&ill->ill_lock);
5852 return (EINTR);
5853 }
5854 }
5855 mutex_exit(&ill->ill_lock);
5856 /*
5857 * ip_rput_other could have set an error in ill_error on
5858 * receipt of M_ERROR.
5859 */
5860 return (ill->ill_error);
5861 }
5862
5863 /*
5864 * This is a module open, i.e. this is a control stream for access
5865 * to a DLPI device. We allocate an ill_t as the instance data in
5866 * this case.
5867 */
5868 static int
5869 ip_modopen(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5870 {
5871 ill_t *ill;
5872 int err;
5873 zoneid_t zoneid;
5874 netstack_t *ns;
5875 ip_stack_t *ipst;
5876
5877 /*
5878 * Prevent unprivileged processes from pushing IP so that
5879 * they can't send raw IP.
5880 */
5881 if (secpolicy_net_rawaccess(credp) != 0)
5882 return (EPERM);
5883
5884 ns = netstack_find_by_cred(credp);
5885 ASSERT(ns != NULL);
5886 ipst = ns->netstack_ip;
5887 ASSERT(ipst != NULL);
5888
5889 /*
5890 * For exclusive stacks we set the zoneid to zero
5891 * to make IP operate as if in the global zone.
5892 */
5893 if (ipst->ips_netstack->netstack_stackid != GLOBAL_NETSTACKID)
5894 zoneid = GLOBAL_ZONEID;
5895 else
5896 zoneid = crgetzoneid(credp);
5897
5898 ill = (ill_t *)mi_open_alloc_sleep(sizeof (ill_t));
5899 q->q_ptr = WR(q)->q_ptr = ill;
5900 ill->ill_ipst = ipst;
5901 ill->ill_zoneid = zoneid;
5902
5903 /*
5904 * ill_init initializes the ill fields and then sends down
5905 * down a DL_INFO_REQ after calling qprocson.
5906 */
5907 err = ill_init(q, ill);
5908
5909 if (err != 0) {
5910 mi_free(ill);
5911 netstack_rele(ipst->ips_netstack);
5912 q->q_ptr = NULL;
5913 WR(q)->q_ptr = NULL;
5914 return (err);
5915 }
5916
5917 /*
5918 * Wait for the DL_INFO_ACK if a DL_INFO_REQ was sent.
5919 *
5920 * ill_init initializes the ipsq marking this thread as
5921 * writer
5922 */
5923 ipsq_exit(ill->ill_phyint->phyint_ipsq);
5924 err = ip_wait_for_info_ack(ill);
5925 if (err == 0)
5926 ill->ill_credp = credp;
5927 else
5928 goto fail;
5929
5930 crhold(credp);
5931
5932 mutex_enter(&ipst->ips_ip_mi_lock);
5933 err = mi_open_link(&ipst->ips_ip_g_head, (IDP)q->q_ptr, devp, flag,
5934 sflag, credp);
5935 mutex_exit(&ipst->ips_ip_mi_lock);
5936 fail:
5937 if (err) {
5938 (void) ip_close(q, 0, credp);
5939 return (err);
5940 }
5941 return (0);
5942 }
5943
5944 /* For /dev/ip aka AF_INET open */
5945 int
5946 ip_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5947 {
5948 return (ip_open(q, devp, flag, sflag, credp, B_FALSE));
5949 }
5950
5951 /* For /dev/ip6 aka AF_INET6 open */
5952 int
5953 ip_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5954 {
5955 return (ip_open(q, devp, flag, sflag, credp, B_TRUE));
5956 }
5957
5958 /* IP open routine. */
5959 int
5960 ip_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp,
5961 boolean_t isv6)
5962 {
5963 conn_t *connp;
5964 major_t maj;
5965 zoneid_t zoneid;
5966 netstack_t *ns;
5967 ip_stack_t *ipst;
5968
5969 /* Allow reopen. */
5970 if (q->q_ptr != NULL)
5971 return (0);
5972
5973 if (sflag & MODOPEN) {
5974 /* This is a module open */
5975 return (ip_modopen(q, devp, flag, sflag, credp));
5976 }
5977
5978 if ((flag & ~(FKLYR)) == IP_HELPER_STR) {
5979 /*
5980 * Non streams based socket looking for a stream
5981 * to access IP
5982 */
5983 return (ip_helper_stream_setup(q, devp, flag, sflag,
5984 credp, isv6));
5985 }
5986
5987 ns = netstack_find_by_cred(credp);
5988 ASSERT(ns != NULL);
5989 ipst = ns->netstack_ip;
5990 ASSERT(ipst != NULL);
5991
5992 /*
5993 * For exclusive stacks we set the zoneid to zero
5994 * to make IP operate as if in the global zone.
5995 */
5996 if (ipst->ips_netstack->netstack_stackid != GLOBAL_NETSTACKID)
5997 zoneid = GLOBAL_ZONEID;
5998 else
5999 zoneid = crgetzoneid(credp);
6000
6001 /*
6002 * We are opening as a device. This is an IP client stream, and we
6003 * allocate an conn_t as the instance data.
6004 */
6005 connp = ipcl_conn_create(IPCL_IPCCONN, KM_SLEEP, ipst->ips_netstack);
6006
6007 /*
6008 * ipcl_conn_create did a netstack_hold. Undo the hold that was
6009 * done by netstack_find_by_cred()
6010 */
6011 netstack_rele(ipst->ips_netstack);
6012
6013 connp->conn_ixa->ixa_flags |= IXAF_MULTICAST_LOOP | IXAF_SET_ULP_CKSUM;
6014 /* conn_allzones can not be set this early, hence no IPCL_ZONEID */
6015 connp->conn_ixa->ixa_zoneid = zoneid;
6016 connp->conn_zoneid = zoneid;
6017
6018 connp->conn_rq = q;
6019 q->q_ptr = WR(q)->q_ptr = connp;
6020
6021 /* Minor tells us which /dev entry was opened */
6022 if (isv6) {
6023 connp->conn_family = AF_INET6;
6024 connp->conn_ipversion = IPV6_VERSION;
6025 connp->conn_ixa->ixa_flags &= ~IXAF_IS_IPV4;
6026 connp->conn_ixa->ixa_src_preferences = IPV6_PREFER_SRC_DEFAULT;
6027 } else {
6028 connp->conn_family = AF_INET;
6029 connp->conn_ipversion = IPV4_VERSION;
6030 connp->conn_ixa->ixa_flags |= IXAF_IS_IPV4;
6031 }
6032
6033 if ((ip_minor_arena_la != NULL) && (flag & SO_SOCKSTR) &&
6034 ((connp->conn_dev = inet_minor_alloc(ip_minor_arena_la)) != 0)) {
6035 connp->conn_minor_arena = ip_minor_arena_la;
6036 } else {
6037 /*
6038 * Either minor numbers in the large arena were exhausted
6039 * or a non socket application is doing the open.
6040 * Try to allocate from the small arena.
6041 */
6042 if ((connp->conn_dev =
6043 inet_minor_alloc(ip_minor_arena_sa)) == 0) {
6044 /* CONN_DEC_REF takes care of netstack_rele() */
6045 q->q_ptr = WR(q)->q_ptr = NULL;
6046 CONN_DEC_REF(connp);
6047 return (EBUSY);
6048 }
6049 connp->conn_minor_arena = ip_minor_arena_sa;
6050 }
6051
6052 maj = getemajor(*devp);
6053 *devp = makedevice(maj, (minor_t)connp->conn_dev);
6054
6055 /*
6056 * connp->conn_cred is crfree()ed in ipcl_conn_destroy()
6057 */
6058 connp->conn_cred = credp;
6059 connp->conn_cpid = curproc->p_pid;
6060 /* Cache things in ixa without an extra refhold */
6061 ASSERT(!(connp->conn_ixa->ixa_free_flags & IXA_FREE_CRED));
6062 connp->conn_ixa->ixa_cred = connp->conn_cred;
6063 connp->conn_ixa->ixa_cpid = connp->conn_cpid;
6064 if (is_system_labeled())
6065 connp->conn_ixa->ixa_tsl = crgetlabel(connp->conn_cred);
6066
6067 /*
6068 * Handle IP_IOC_RTS_REQUEST and other ioctls which use conn_recv
6069 */
6070 connp->conn_recv = ip_conn_input;
6071 connp->conn_recvicmp = ip_conn_input_icmp;
6072
6073 crhold(connp->conn_cred);
6074
6075 /*
6076 * If the caller has the process-wide flag set, then default to MAC
6077 * exempt mode. This allows read-down to unlabeled hosts.
6078 */
6079 if (getpflags(NET_MAC_AWARE, credp) != 0)
6080 connp->conn_mac_mode = CONN_MAC_AWARE;
6081
6082 connp->conn_zone_is_global = (crgetzoneid(credp) == GLOBAL_ZONEID);
6083
6084 connp->conn_rq = q;
6085 connp->conn_wq = WR(q);
6086
6087 /* Non-zero default values */
6088 connp->conn_ixa->ixa_flags |= IXAF_MULTICAST_LOOP;
6089
6090 /*
6091 * Make the conn globally visible to walkers
6092 */
6093 ASSERT(connp->conn_ref == 1);
6094 mutex_enter(&connp->conn_lock);
6095 connp->conn_state_flags &= ~CONN_INCIPIENT;
6096 mutex_exit(&connp->conn_lock);
6097
6098 qprocson(q);
6099
6100 return (0);
6101 }
6102
6103 /*
6104 * Set IPsec policy from an ipsec_req_t. If the req is not "zero" and valid,
6105 * all of them are copied to the conn_t. If the req is "zero", the policy is
6106 * zeroed out. A "zero" policy has zero ipsr_{ah,req,self_encap}_req
6107 * fields.
6108 * We keep only the latest setting of the policy and thus policy setting
6109 * is not incremental/cumulative.
6110 *
6111 * Requests to set policies with multiple alternative actions will
6112 * go through a different API.
6113 */
6114 int
6115 ipsec_set_req(cred_t *cr, conn_t *connp, ipsec_req_t *req)
6116 {
6117 uint_t ah_req = 0;
6118 uint_t esp_req = 0;
6119 uint_t se_req = 0;
6120 ipsec_act_t *actp = NULL;
6121 uint_t nact;
6122 ipsec_policy_head_t *ph;
6123 boolean_t is_pol_reset, is_pol_inserted = B_FALSE;
6124 int error = 0;
6125 netstack_t *ns = connp->conn_netstack;
6126 ip_stack_t *ipst = ns->netstack_ip;
6127 ipsec_stack_t *ipss = ns->netstack_ipsec;
6128
6129 #define REQ_MASK (IPSEC_PREF_REQUIRED|IPSEC_PREF_NEVER)
6130
6131 /*
6132 * The IP_SEC_OPT option does not allow variable length parameters,
6133 * hence a request cannot be NULL.
6134 */
6135 if (req == NULL)
6136 return (EINVAL);
6137
6138 ah_req = req->ipsr_ah_req;
6139 esp_req = req->ipsr_esp_req;
6140 se_req = req->ipsr_self_encap_req;
6141
6142 /* Don't allow setting self-encap without one or more of AH/ESP. */
6143 if (se_req != 0 && esp_req == 0 && ah_req == 0)
6144 return (EINVAL);
6145
6146 /*
6147 * Are we dealing with a request to reset the policy (i.e.
6148 * zero requests).
6149 */
6150 is_pol_reset = ((ah_req & REQ_MASK) == 0 &&
6151 (esp_req & REQ_MASK) == 0 &&
6152 (se_req & REQ_MASK) == 0);
6153
6154 if (!is_pol_reset) {
6155 /*
6156 * If we couldn't load IPsec, fail with "protocol
6157 * not supported".
6158 * IPsec may not have been loaded for a request with zero
6159 * policies, so we don't fail in this case.
6160 */
6161 mutex_enter(&ipss->ipsec_loader_lock);
6162 if (ipss->ipsec_loader_state != IPSEC_LOADER_SUCCEEDED) {
6163 mutex_exit(&ipss->ipsec_loader_lock);
6164 return (EPROTONOSUPPORT);
6165 }
6166 mutex_exit(&ipss->ipsec_loader_lock);
6167
6168 /*
6169 * Test for valid requests. Invalid algorithms
6170 * need to be tested by IPsec code because new
6171 * algorithms can be added dynamically.
6172 */
6173 if ((ah_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0 ||
6174 (esp_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0 ||
6175 (se_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0) {
6176 return (EINVAL);
6177 }
6178
6179 /*
6180 * Only privileged users can issue these
6181 * requests.
6182 */
6183 if (((ah_req & IPSEC_PREF_NEVER) ||
6184 (esp_req & IPSEC_PREF_NEVER) ||
6185 (se_req & IPSEC_PREF_NEVER)) &&
6186 secpolicy_ip_config(cr, B_FALSE) != 0) {
6187 return (EPERM);
6188 }
6189
6190 /*
6191 * The IPSEC_PREF_REQUIRED and IPSEC_PREF_NEVER
6192 * are mutually exclusive.
6193 */
6194 if (((ah_req & REQ_MASK) == REQ_MASK) ||
6195 ((esp_req & REQ_MASK) == REQ_MASK) ||
6196 ((se_req & REQ_MASK) == REQ_MASK)) {
6197 /* Both of them are set */
6198 return (EINVAL);
6199 }
6200 }
6201
6202 ASSERT(MUTEX_HELD(&connp->conn_lock));
6203
6204 /*
6205 * If we have already cached policies in conn_connect(), don't
6206 * let them change now. We cache policies for connections
6207 * whose src,dst [addr, port] is known.
6208 */
6209 if (connp->conn_policy_cached) {
6210 return (EINVAL);
6211 }
6212
6213 /*
6214 * We have a zero policies, reset the connection policy if already
6215 * set. This will cause the connection to inherit the
6216 * global policy, if any.
6217 */
6218 if (is_pol_reset) {
6219 if (connp->conn_policy != NULL) {
6220 IPPH_REFRELE(connp->conn_policy, ipst->ips_netstack);
6221 connp->conn_policy = NULL;
6222 }
6223 connp->conn_in_enforce_policy = B_FALSE;
6224 connp->conn_out_enforce_policy = B_FALSE;
6225 return (0);
6226 }
6227
6228 ph = connp->conn_policy = ipsec_polhead_split(connp->conn_policy,
6229 ipst->ips_netstack);
6230 if (ph == NULL)
6231 goto enomem;
6232
6233 ipsec_actvec_from_req(req, &actp, &nact, ipst->ips_netstack);
6234 if (actp == NULL)
6235 goto enomem;
6236
6237 /*
6238 * Always insert IPv4 policy entries, since they can also apply to
6239 * ipv6 sockets being used in ipv4-compat mode.
6240 */
6241 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V4,
6242 IPSEC_TYPE_INBOUND, ns))
6243 goto enomem;
6244 is_pol_inserted = B_TRUE;
6245 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V4,
6246 IPSEC_TYPE_OUTBOUND, ns))
6247 goto enomem;
6248
6249 /*
6250 * We're looking at a v6 socket, also insert the v6-specific
6251 * entries.
6252 */
6253 if (connp->conn_family == AF_INET6) {
6254 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V6,
6255 IPSEC_TYPE_INBOUND, ns))
6256 goto enomem;
6257 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V6,
6258 IPSEC_TYPE_OUTBOUND, ns))
6259 goto enomem;
6260 }
6261
6262 ipsec_actvec_free(actp, nact);
6263
6264 /*
6265 * If the requests need security, set enforce_policy.
6266 * If the requests are IPSEC_PREF_NEVER, one should
6267 * still set conn_out_enforce_policy so that ip_set_destination
6268 * marks the ip_xmit_attr_t appropriatly. This is needed so that
6269 * for connections that we don't cache policy in at connect time,
6270 * if global policy matches in ip_output_attach_policy, we
6271 * don't wrongly inherit global policy. Similarly, we need
6272 * to set conn_in_enforce_policy also so that we don't verify
6273 * policy wrongly.
6274 */
6275 if ((ah_req & REQ_MASK) != 0 ||
6276 (esp_req & REQ_MASK) != 0 ||
6277 (se_req & REQ_MASK) != 0) {
6278 connp->conn_in_enforce_policy = B_TRUE;
6279 connp->conn_out_enforce_policy = B_TRUE;
6280 }
6281
6282 return (error);
6283 #undef REQ_MASK
6284
6285 /*
6286 * Common memory-allocation-failure exit path.
6287 */
6288 enomem:
6289 if (actp != NULL)
6290 ipsec_actvec_free(actp, nact);
6291 if (is_pol_inserted)
6292 ipsec_polhead_flush(ph, ns);
6293 return (ENOMEM);
6294 }
6295
6296 /*
6297 * Set socket options for joining and leaving multicast groups.
6298 * Common to IPv4 and IPv6; inet6 indicates the type of socket.
6299 * The caller has already check that the option name is consistent with
6300 * the address family of the socket.
6301 */
6302 int
6303 ip_opt_set_multicast_group(conn_t *connp, t_scalar_t name,
6304 uchar_t *invalp, boolean_t inet6, boolean_t checkonly)
6305 {
6306 int *i1 = (int *)invalp;
6307 int error = 0;
6308 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
6309 struct ip_mreq *v4_mreqp;
6310 struct ipv6_mreq *v6_mreqp;
6311 struct group_req *greqp;
6312 ire_t *ire;
6313 boolean_t done = B_FALSE;
6314 ipaddr_t ifaddr;
6315 in6_addr_t v6group;
6316 uint_t ifindex;
6317 boolean_t mcast_opt = B_TRUE;
6318 mcast_record_t fmode;
6319 int (*optfn)(conn_t *, boolean_t, const in6_addr_t *,
6320 ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *);
6321
6322 switch (name) {
6323 case IP_ADD_MEMBERSHIP:
6324 case IPV6_JOIN_GROUP:
6325 mcast_opt = B_FALSE;
6326 /* FALLTHROUGH */
6327 case MCAST_JOIN_GROUP:
6328 fmode = MODE_IS_EXCLUDE;
6329 optfn = ip_opt_add_group;
6330 break;
6331
6332 case IP_DROP_MEMBERSHIP:
6333 case IPV6_LEAVE_GROUP:
6334 mcast_opt = B_FALSE;
6335 /* FALLTHROUGH */
6336 case MCAST_LEAVE_GROUP:
6337 fmode = MODE_IS_INCLUDE;
6338 optfn = ip_opt_delete_group;
6339 break;
6340 default:
6341 ASSERT(0);
6342 }
6343
6344 if (mcast_opt) {
6345 struct sockaddr_in *sin;
6346 struct sockaddr_in6 *sin6;
6347
6348 greqp = (struct group_req *)i1;
6349 if (greqp->gr_group.ss_family == AF_INET) {
6350 sin = (struct sockaddr_in *)&(greqp->gr_group);
6351 IN6_INADDR_TO_V4MAPPED(&sin->sin_addr, &v6group);
6352 } else {
6353 if (!inet6)
6354 return (EINVAL); /* Not on INET socket */
6355
6356 sin6 = (struct sockaddr_in6 *)&(greqp->gr_group);
6357 v6group = sin6->sin6_addr;
6358 }
6359 ifaddr = INADDR_ANY;
6360 ifindex = greqp->gr_interface;
6361 } else if (inet6) {
6362 v6_mreqp = (struct ipv6_mreq *)i1;
6363 v6group = v6_mreqp->ipv6mr_multiaddr;
6364 ifaddr = INADDR_ANY;
6365 ifindex = v6_mreqp->ipv6mr_interface;
6366 } else {
6367 v4_mreqp = (struct ip_mreq *)i1;
6368 IN6_INADDR_TO_V4MAPPED(&v4_mreqp->imr_multiaddr, &v6group);
6369 ifaddr = (ipaddr_t)v4_mreqp->imr_interface.s_addr;
6370 ifindex = 0;
6371 }
6372
6373 /*
6374 * In the multirouting case, we need to replicate
6375 * the request on all interfaces that will take part
6376 * in replication. We do so because multirouting is
6377 * reflective, thus we will probably receive multi-
6378 * casts on those interfaces.
6379 * The ip_multirt_apply_membership() succeeds if
6380 * the operation succeeds on at least one interface.
6381 */
6382 if (IN6_IS_ADDR_V4MAPPED(&v6group)) {
6383 ipaddr_t group;
6384
6385 IN6_V4MAPPED_TO_IPADDR(&v6group, group);
6386
6387 ire = ire_ftable_lookup_v4(group, IP_HOST_MASK, 0,
6388 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6389 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6390 } else {
6391 ire = ire_ftable_lookup_v6(&v6group, &ipv6_all_ones, 0,
6392 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6393 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6394 }
6395 if (ire != NULL) {
6396 if (ire->ire_flags & RTF_MULTIRT) {
6397 error = ip_multirt_apply_membership(optfn, ire, connp,
6398 checkonly, &v6group, fmode, &ipv6_all_zeros);
6399 done = B_TRUE;
6400 }
6401 ire_refrele(ire);
6402 }
6403
6404 if (!done) {
6405 error = optfn(connp, checkonly, &v6group, ifaddr, ifindex,
6406 fmode, &ipv6_all_zeros);
6407 }
6408 return (error);
6409 }
6410
6411 /*
6412 * Set socket options for joining and leaving multicast groups
6413 * for specific sources.
6414 * Common to IPv4 and IPv6; inet6 indicates the type of socket.
6415 * The caller has already check that the option name is consistent with
6416 * the address family of the socket.
6417 */
6418 int
6419 ip_opt_set_multicast_sources(conn_t *connp, t_scalar_t name,
6420 uchar_t *invalp, boolean_t inet6, boolean_t checkonly)
6421 {
6422 int *i1 = (int *)invalp;
6423 int error = 0;
6424 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
6425 struct ip_mreq_source *imreqp;
6426 struct group_source_req *gsreqp;
6427 in6_addr_t v6group, v6src;
6428 uint32_t ifindex;
6429 ipaddr_t ifaddr;
6430 boolean_t mcast_opt = B_TRUE;
6431 mcast_record_t fmode;
6432 ire_t *ire;
6433 boolean_t done = B_FALSE;
6434 int (*optfn)(conn_t *, boolean_t, const in6_addr_t *,
6435 ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *);
6436
6437 switch (name) {
6438 case IP_BLOCK_SOURCE:
6439 mcast_opt = B_FALSE;
6440 /* FALLTHROUGH */
6441 case MCAST_BLOCK_SOURCE:
6442 fmode = MODE_IS_EXCLUDE;
6443 optfn = ip_opt_add_group;
6444 break;
6445
6446 case IP_UNBLOCK_SOURCE:
6447 mcast_opt = B_FALSE;
6448 /* FALLTHROUGH */
6449 case MCAST_UNBLOCK_SOURCE:
6450 fmode = MODE_IS_EXCLUDE;
6451 optfn = ip_opt_delete_group;
6452 break;
6453
6454 case IP_ADD_SOURCE_MEMBERSHIP:
6455 mcast_opt = B_FALSE;
6456 /* FALLTHROUGH */
6457 case MCAST_JOIN_SOURCE_GROUP:
6458 fmode = MODE_IS_INCLUDE;
6459 optfn = ip_opt_add_group;
6460 break;
6461
6462 case IP_DROP_SOURCE_MEMBERSHIP:
6463 mcast_opt = B_FALSE;
6464 /* FALLTHROUGH */
6465 case MCAST_LEAVE_SOURCE_GROUP:
6466 fmode = MODE_IS_INCLUDE;
6467 optfn = ip_opt_delete_group;
6468 break;
6469 default:
6470 ASSERT(0);
6471 }
6472
6473 if (mcast_opt) {
6474 gsreqp = (struct group_source_req *)i1;
6475 ifindex = gsreqp->gsr_interface;
6476 if (gsreqp->gsr_group.ss_family == AF_INET) {
6477 struct sockaddr_in *s;
6478 s = (struct sockaddr_in *)&gsreqp->gsr_group;
6479 IN6_INADDR_TO_V4MAPPED(&s->sin_addr, &v6group);
6480 s = (struct sockaddr_in *)&gsreqp->gsr_source;
6481 IN6_INADDR_TO_V4MAPPED(&s->sin_addr, &v6src);
6482 } else {
6483 struct sockaddr_in6 *s6;
6484
6485 if (!inet6)
6486 return (EINVAL); /* Not on INET socket */
6487
6488 s6 = (struct sockaddr_in6 *)&gsreqp->gsr_group;
6489 v6group = s6->sin6_addr;
6490 s6 = (struct sockaddr_in6 *)&gsreqp->gsr_source;
6491 v6src = s6->sin6_addr;
6492 }
6493 ifaddr = INADDR_ANY;
6494 } else {
6495 imreqp = (struct ip_mreq_source *)i1;
6496 IN6_INADDR_TO_V4MAPPED(&imreqp->imr_multiaddr, &v6group);
6497 IN6_INADDR_TO_V4MAPPED(&imreqp->imr_sourceaddr, &v6src);
6498 ifaddr = (ipaddr_t)imreqp->imr_interface.s_addr;
6499 ifindex = 0;
6500 }
6501
6502 /*
6503 * Handle src being mapped INADDR_ANY by changing it to unspecified.
6504 */
6505 if (IN6_IS_ADDR_V4MAPPED_ANY(&v6src))
6506 v6src = ipv6_all_zeros;
6507
6508 /*
6509 * In the multirouting case, we need to replicate
6510 * the request as noted in the mcast cases above.
6511 */
6512 if (IN6_IS_ADDR_V4MAPPED(&v6group)) {
6513 ipaddr_t group;
6514
6515 IN6_V4MAPPED_TO_IPADDR(&v6group, group);
6516
6517 ire = ire_ftable_lookup_v4(group, IP_HOST_MASK, 0,
6518 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6519 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6520 } else {
6521 ire = ire_ftable_lookup_v6(&v6group, &ipv6_all_ones, 0,
6522 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6523 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6524 }
6525 if (ire != NULL) {
6526 if (ire->ire_flags & RTF_MULTIRT) {
6527 error = ip_multirt_apply_membership(optfn, ire, connp,
6528 checkonly, &v6group, fmode, &v6src);
6529 done = B_TRUE;
6530 }
6531 ire_refrele(ire);
6532 }
6533 if (!done) {
6534 error = optfn(connp, checkonly, &v6group, ifaddr, ifindex,
6535 fmode, &v6src);
6536 }
6537 return (error);
6538 }
6539
6540 /*
6541 * Given a destination address and a pointer to where to put the information
6542 * this routine fills in the mtuinfo.
6543 * The socket must be connected.
6544 * For sctp conn_faddr is the primary address.
6545 */
6546 int
6547 ip_fill_mtuinfo(conn_t *connp, ip_xmit_attr_t *ixa, struct ip6_mtuinfo *mtuinfo)
6548 {
6549 uint32_t pmtu = IP_MAXPACKET;
6550 uint_t scopeid;
6551
6552 if (IN6_IS_ADDR_UNSPECIFIED(&connp->conn_faddr_v6))
6553 return (-1);
6554
6555 /* In case we never sent or called ip_set_destination_v4/v6 */
6556 if (ixa->ixa_ire != NULL)
6557 pmtu = ip_get_pmtu(ixa);
6558
6559 if (ixa->ixa_flags & IXAF_SCOPEID_SET)
6560 scopeid = ixa->ixa_scopeid;
6561 else
6562 scopeid = 0;
6563
6564 bzero(mtuinfo, sizeof (*mtuinfo));
6565 mtuinfo->ip6m_addr.sin6_family = AF_INET6;
6566 mtuinfo->ip6m_addr.sin6_port = connp->conn_fport;
6567 mtuinfo->ip6m_addr.sin6_addr = connp->conn_faddr_v6;
6568 mtuinfo->ip6m_addr.sin6_scope_id = scopeid;
6569 mtuinfo->ip6m_mtu = pmtu;
6570
6571 return (sizeof (struct ip6_mtuinfo));
6572 }
6573
6574 /*
6575 * When the src multihoming is changed from weak to [strong, preferred]
6576 * ip_ire_rebind_walker is called to walk the list of all ire_t entries
6577 * and identify routes that were created by user-applications in the
6578 * unbound state (i.e., without RTA_IFP), and for which an ire_ill is not
6579 * currently defined. These routes are then 'rebound', i.e., their ire_ill
6580 * is selected by finding an interface route for the gateway.
6581 */
6582 /* ARGSUSED */
6583 void
6584 ip_ire_rebind_walker(ire_t *ire, void *notused)
6585 {
6586 if (!ire->ire_unbound || ire->ire_ill != NULL)
6587 return;
6588 ire_rebind(ire);
6589 ire_delete(ire);
6590 }
6591
6592 /*
6593 * When the src multihoming is changed from [strong, preferred] to weak,
6594 * ip_ire_unbind_walker is called to walk the list of all ire_t entries, and
6595 * set any entries that were created by user-applications in the unbound state
6596 * (i.e., without RTA_IFP) back to having a NULL ire_ill.
6597 */
6598 /* ARGSUSED */
6599 void
6600 ip_ire_unbind_walker(ire_t *ire, void *notused)
6601 {
6602 ire_t *new_ire;
6603
6604 if (!ire->ire_unbound || ire->ire_ill == NULL)
6605 return;
6606 if (ire->ire_ipversion == IPV6_VERSION) {
6607 new_ire = ire_create_v6(&ire->ire_addr_v6, &ire->ire_mask_v6,
6608 &ire->ire_gateway_addr_v6, ire->ire_type, NULL,
6609 ire->ire_zoneid, ire->ire_flags, NULL, ire->ire_ipst);
6610 } else {
6611 new_ire = ire_create((uchar_t *)&ire->ire_addr,
6612 (uchar_t *)&ire->ire_mask,
6613 (uchar_t *)&ire->ire_gateway_addr, ire->ire_type, NULL,
6614 ire->ire_zoneid, ire->ire_flags, NULL, ire->ire_ipst);
6615 }
6616 if (new_ire == NULL)
6617 return;
6618 new_ire->ire_unbound = B_TRUE;
6619 /*
6620 * The bound ire must first be deleted so that we don't return
6621 * the existing one on the attempt to add the unbound new_ire.
6622 */
6623 ire_delete(ire);
6624 new_ire = ire_add(new_ire);
6625 if (new_ire != NULL)
6626 ire_refrele(new_ire);
6627 }
6628
6629 /*
6630 * When the settings of ip*_strict_src_multihoming tunables are changed,
6631 * all cached routes need to be recomputed. This recomputation needs to be
6632 * done when going from weaker to stronger modes so that the cached ire
6633 * for the connection does not violate the current ip*_strict_src_multihoming
6634 * setting. It also needs to be done when going from stronger to weaker modes,
6635 * so that we fall back to matching on the longest-matching-route (as opposed
6636 * to a shorter match that may have been selected in the strong mode
6637 * to satisfy src_multihoming settings).
6638 *
6639 * The cached ixa_ire entires for all conn_t entries are marked as
6640 * "verify" so that they will be recomputed for the next packet.
6641 */
6642 void
6643 conn_ire_revalidate(conn_t *connp, void *arg)
6644 {
6645 boolean_t isv6 = (boolean_t)arg;
6646
6647 if ((isv6 && connp->conn_ipversion != IPV6_VERSION) ||
6648 (!isv6 && connp->conn_ipversion != IPV4_VERSION))
6649 return;
6650 connp->conn_ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
6651 }
6652
6653 /*
6654 * Handles both IPv4 and IPv6 reassembly - doing the out-of-order cases,
6655 * When an ipf is passed here for the first time, if
6656 * we already have in-order fragments on the queue, we convert from the fast-
6657 * path reassembly scheme to the hard-case scheme. From then on, additional
6658 * fragments are reassembled here. We keep track of the start and end offsets
6659 * of each piece, and the number of holes in the chain. When the hole count
6660 * goes to zero, we are done!
6661 *
6662 * The ipf_count will be updated to account for any mblk(s) added (pointed to
6663 * by mp) or subtracted (freeb()ed dups), upon return the caller must update
6664 * ipfb_count and ill_frag_count by the difference of ipf_count before and
6665 * after the call to ip_reassemble().
6666 */
6667 int
6668 ip_reassemble(mblk_t *mp, ipf_t *ipf, uint_t start, boolean_t more, ill_t *ill,
6669 size_t msg_len)
6670 {
6671 uint_t end;
6672 mblk_t *next_mp;
6673 mblk_t *mp1;
6674 uint_t offset;
6675 boolean_t incr_dups = B_TRUE;
6676 boolean_t offset_zero_seen = B_FALSE;
6677 boolean_t pkt_boundary_checked = B_FALSE;
6678
6679 /* If start == 0 then ipf_nf_hdr_len has to be set. */
6680 ASSERT(start != 0 || ipf->ipf_nf_hdr_len != 0);
6681
6682 /* Add in byte count */
6683 ipf->ipf_count += msg_len;
6684 if (ipf->ipf_end) {
6685 /*
6686 * We were part way through in-order reassembly, but now there
6687 * is a hole. We walk through messages already queued, and
6688 * mark them for hard case reassembly. We know that up till
6689 * now they were in order starting from offset zero.
6690 */
6691 offset = 0;
6692 for (mp1 = ipf->ipf_mp->b_cont; mp1; mp1 = mp1->b_cont) {
6693 IP_REASS_SET_START(mp1, offset);
6694 if (offset == 0) {
6695 ASSERT(ipf->ipf_nf_hdr_len != 0);
6696 offset = -ipf->ipf_nf_hdr_len;
6697 }
6698 offset += mp1->b_wptr - mp1->b_rptr;
6699 IP_REASS_SET_END(mp1, offset);
6700 }
6701 /* One hole at the end. */
6702 ipf->ipf_hole_cnt = 1;
6703 /* Brand it as a hard case, forever. */
6704 ipf->ipf_end = 0;
6705 }
6706 /* Walk through all the new pieces. */
6707 do {
6708 end = start + (mp->b_wptr - mp->b_rptr);
6709 /*
6710 * If start is 0, decrease 'end' only for the first mblk of
6711 * the fragment. Otherwise 'end' can get wrong value in the
6712 * second pass of the loop if first mblk is exactly the
6713 * size of ipf_nf_hdr_len.
6714 */
6715 if (start == 0 && !offset_zero_seen) {
6716 /* First segment */
6717 ASSERT(ipf->ipf_nf_hdr_len != 0);
6718 end -= ipf->ipf_nf_hdr_len;
6719 offset_zero_seen = B_TRUE;
6720 }
6721 next_mp = mp->b_cont;
6722 /*
6723 * We are checking to see if there is any interesing data
6724 * to process. If there isn't and the mblk isn't the
6725 * one which carries the unfragmentable header then we
6726 * drop it. It's possible to have just the unfragmentable
6727 * header come through without any data. That needs to be
6728 * saved.
6729 *
6730 * If the assert at the top of this function holds then the
6731 * term "ipf->ipf_nf_hdr_len != 0" isn't needed. This code
6732 * is infrequently traveled enough that the test is left in
6733 * to protect against future code changes which break that
6734 * invariant.
6735 */
6736 if (start == end && start != 0 && ipf->ipf_nf_hdr_len != 0) {
6737 /* Empty. Blast it. */
6738 IP_REASS_SET_START(mp, 0);
6739 IP_REASS_SET_END(mp, 0);
6740 /*
6741 * If the ipf points to the mblk we are about to free,
6742 * update ipf to point to the next mblk (or NULL
6743 * if none).
6744 */
6745 if (ipf->ipf_mp->b_cont == mp)
6746 ipf->ipf_mp->b_cont = next_mp;
6747 freeb(mp);
6748 continue;
6749 }
6750 mp->b_cont = NULL;
6751 IP_REASS_SET_START(mp, start);
6752 IP_REASS_SET_END(mp, end);
6753 if (!ipf->ipf_tail_mp) {
6754 ipf->ipf_tail_mp = mp;
6755 ipf->ipf_mp->b_cont = mp;
6756 if (start == 0 || !more) {
6757 ipf->ipf_hole_cnt = 1;
6758 /*
6759 * if the first fragment comes in more than one
6760 * mblk, this loop will be executed for each
6761 * mblk. Need to adjust hole count so exiting
6762 * this routine will leave hole count at 1.
6763 */
6764 if (next_mp)
6765 ipf->ipf_hole_cnt++;
6766 } else
6767 ipf->ipf_hole_cnt = 2;
6768 continue;
6769 } else if (ipf->ipf_last_frag_seen && !more &&
6770 !pkt_boundary_checked) {
6771 /*
6772 * We check datagram boundary only if this fragment
6773 * claims to be the last fragment and we have seen a
6774 * last fragment in the past too. We do this only
6775 * once for a given fragment.
6776 *
6777 * start cannot be 0 here as fragments with start=0
6778 * and MF=0 gets handled as a complete packet. These
6779 * fragments should not reach here.
6780 */
6781
6782 if (start + msgdsize(mp) !=
6783 IP_REASS_END(ipf->ipf_tail_mp)) {
6784 /*
6785 * We have two fragments both of which claim
6786 * to be the last fragment but gives conflicting
6787 * information about the whole datagram size.
6788 * Something fishy is going on. Drop the
6789 * fragment and free up the reassembly list.
6790 */
6791 return (IP_REASS_FAILED);
6792 }
6793
6794 /*
6795 * We shouldn't come to this code block again for this
6796 * particular fragment.
6797 */
6798 pkt_boundary_checked = B_TRUE;
6799 }
6800
6801 /* New stuff at or beyond tail? */
6802 offset = IP_REASS_END(ipf->ipf_tail_mp);
6803 if (start >= offset) {
6804 if (ipf->ipf_last_frag_seen) {
6805 /* current fragment is beyond last fragment */
6806 return (IP_REASS_FAILED);
6807 }
6808 /* Link it on end. */
6809 ipf->ipf_tail_mp->b_cont = mp;
6810 ipf->ipf_tail_mp = mp;
6811 if (more) {
6812 if (start != offset)
6813 ipf->ipf_hole_cnt++;
6814 } else if (start == offset && next_mp == NULL)
6815 ipf->ipf_hole_cnt--;
6816 continue;
6817 }
6818 mp1 = ipf->ipf_mp->b_cont;
6819 offset = IP_REASS_START(mp1);
6820 /* New stuff at the front? */
6821 if (start < offset) {
6822 if (start == 0) {
6823 if (end >= offset) {
6824 /* Nailed the hole at the begining. */
6825 ipf->ipf_hole_cnt--;
6826 }
6827 } else if (end < offset) {
6828 /*
6829 * A hole, stuff, and a hole where there used
6830 * to be just a hole.
6831 */
6832 ipf->ipf_hole_cnt++;
6833 }
6834 mp->b_cont = mp1;
6835 /* Check for overlap. */
6836 while (end > offset) {
6837 if (end < IP_REASS_END(mp1)) {
6838 mp->b_wptr -= end - offset;
6839 IP_REASS_SET_END(mp, offset);
6840 BUMP_MIB(ill->ill_ip_mib,
6841 ipIfStatsReasmPartDups);
6842 break;
6843 }
6844 /* Did we cover another hole? */
6845 if ((mp1->b_cont &&
6846 IP_REASS_END(mp1) !=
6847 IP_REASS_START(mp1->b_cont) &&
6848 end >= IP_REASS_START(mp1->b_cont)) ||
6849 (!ipf->ipf_last_frag_seen && !more)) {
6850 ipf->ipf_hole_cnt--;
6851 }
6852 /* Clip out mp1. */
6853 if ((mp->b_cont = mp1->b_cont) == NULL) {
6854 /*
6855 * After clipping out mp1, this guy
6856 * is now hanging off the end.
6857 */
6858 ipf->ipf_tail_mp = mp;
6859 }
6860 IP_REASS_SET_START(mp1, 0);
6861 IP_REASS_SET_END(mp1, 0);
6862 /* Subtract byte count */
6863 ipf->ipf_count -= mp1->b_datap->db_lim -
6864 mp1->b_datap->db_base;
6865 freeb(mp1);
6866 BUMP_MIB(ill->ill_ip_mib,
6867 ipIfStatsReasmPartDups);
6868 mp1 = mp->b_cont;
6869 if (!mp1)
6870 break;
6871 offset = IP_REASS_START(mp1);
6872 }
6873 ipf->ipf_mp->b_cont = mp;
6874 continue;
6875 }
6876 /*
6877 * The new piece starts somewhere between the start of the head
6878 * and before the end of the tail.
6879 */
6880 for (; mp1; mp1 = mp1->b_cont) {
6881 offset = IP_REASS_END(mp1);
6882 if (start < offset) {
6883 if (end <= offset) {
6884 /* Nothing new. */
6885 IP_REASS_SET_START(mp, 0);
6886 IP_REASS_SET_END(mp, 0);
6887 /* Subtract byte count */
6888 ipf->ipf_count -= mp->b_datap->db_lim -
6889 mp->b_datap->db_base;
6890 if (incr_dups) {
6891 ipf->ipf_num_dups++;
6892 incr_dups = B_FALSE;
6893 }
6894 freeb(mp);
6895 BUMP_MIB(ill->ill_ip_mib,
6896 ipIfStatsReasmDuplicates);
6897 break;
6898 }
6899 /*
6900 * Trim redundant stuff off beginning of new
6901 * piece.
6902 */
6903 IP_REASS_SET_START(mp, offset);
6904 mp->b_rptr += offset - start;
6905 BUMP_MIB(ill->ill_ip_mib,
6906 ipIfStatsReasmPartDups);
6907 start = offset;
6908 if (!mp1->b_cont) {
6909 /*
6910 * After trimming, this guy is now
6911 * hanging off the end.
6912 */
6913 mp1->b_cont = mp;
6914 ipf->ipf_tail_mp = mp;
6915 if (!more) {
6916 ipf->ipf_hole_cnt--;
6917 }
6918 break;
6919 }
6920 }
6921 if (start >= IP_REASS_START(mp1->b_cont))
6922 continue;
6923 /* Fill a hole */
6924 if (start > offset)
6925 ipf->ipf_hole_cnt++;
6926 mp->b_cont = mp1->b_cont;
6927 mp1->b_cont = mp;
6928 mp1 = mp->b_cont;
6929 offset = IP_REASS_START(mp1);
6930 if (end >= offset) {
6931 ipf->ipf_hole_cnt--;
6932 /* Check for overlap. */
6933 while (end > offset) {
6934 if (end < IP_REASS_END(mp1)) {
6935 mp->b_wptr -= end - offset;
6936 IP_REASS_SET_END(mp, offset);
6937 /*
6938 * TODO we might bump
6939 * this up twice if there is
6940 * overlap at both ends.
6941 */
6942 BUMP_MIB(ill->ill_ip_mib,
6943 ipIfStatsReasmPartDups);
6944 break;
6945 }
6946 /* Did we cover another hole? */
6947 if ((mp1->b_cont &&
6948 IP_REASS_END(mp1)
6949 != IP_REASS_START(mp1->b_cont) &&
6950 end >=
6951 IP_REASS_START(mp1->b_cont)) ||
6952 (!ipf->ipf_last_frag_seen &&
6953 !more)) {
6954 ipf->ipf_hole_cnt--;
6955 }
6956 /* Clip out mp1. */
6957 if ((mp->b_cont = mp1->b_cont) ==
6958 NULL) {
6959 /*
6960 * After clipping out mp1,
6961 * this guy is now hanging
6962 * off the end.
6963 */
6964 ipf->ipf_tail_mp = mp;
6965 }
6966 IP_REASS_SET_START(mp1, 0);
6967 IP_REASS_SET_END(mp1, 0);
6968 /* Subtract byte count */
6969 ipf->ipf_count -=
6970 mp1->b_datap->db_lim -
6971 mp1->b_datap->db_base;
6972 freeb(mp1);
6973 BUMP_MIB(ill->ill_ip_mib,
6974 ipIfStatsReasmPartDups);
6975 mp1 = mp->b_cont;
6976 if (!mp1)
6977 break;
6978 offset = IP_REASS_START(mp1);
6979 }
6980 }
6981 break;
6982 }
6983 } while (start = end, mp = next_mp);
6984
6985 /* Fragment just processed could be the last one. Remember this fact */
6986 if (!more)
6987 ipf->ipf_last_frag_seen = B_TRUE;
6988
6989 /* Still got holes? */
6990 if (ipf->ipf_hole_cnt)
6991 return (IP_REASS_PARTIAL);
6992 /* Clean up overloaded fields to avoid upstream disasters. */
6993 for (mp1 = ipf->ipf_mp->b_cont; mp1; mp1 = mp1->b_cont) {
6994 IP_REASS_SET_START(mp1, 0);
6995 IP_REASS_SET_END(mp1, 0);
6996 }
6997 return (IP_REASS_COMPLETE);
6998 }
6999
7000 /*
7001 * Fragmentation reassembly. Each ILL has a hash table for
7002 * queuing packets undergoing reassembly for all IPIFs
7003 * associated with the ILL. The hash is based on the packet
7004 * IP ident field. The ILL frag hash table was allocated
7005 * as a timer block at the time the ILL was created. Whenever
7006 * there is anything on the reassembly queue, the timer will
7007 * be running. Returns the reassembled packet if reassembly completes.
7008 */
7009 mblk_t *
7010 ip_input_fragment(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
7011 {
7012 uint32_t frag_offset_flags;
7013 mblk_t *t_mp;
7014 ipaddr_t dst;
7015 uint8_t proto = ipha->ipha_protocol;
7016 uint32_t sum_val;
7017 uint16_t sum_flags;
7018 ipf_t *ipf;
7019 ipf_t **ipfp;
7020 ipfb_t *ipfb;
7021 uint16_t ident;
7022 uint32_t offset;
7023 ipaddr_t src;
7024 uint_t hdr_length;
7025 uint32_t end;
7026 mblk_t *mp1;
7027 mblk_t *tail_mp;
7028 size_t count;
7029 size_t msg_len;
7030 uint8_t ecn_info = 0;
7031 uint32_t packet_size;
7032 boolean_t pruned = B_FALSE;
7033 ill_t *ill = ira->ira_ill;
7034 ip_stack_t *ipst = ill->ill_ipst;
7035
7036 /*
7037 * Drop the fragmented as early as possible, if
7038 * we don't have resource(s) to re-assemble.
7039 */
7040 if (ipst->ips_ip_reass_queue_bytes == 0) {
7041 freemsg(mp);
7042 return (NULL);
7043 }
7044
7045 /* Check for fragmentation offset; return if there's none */
7046 if ((frag_offset_flags = ntohs(ipha->ipha_fragment_offset_and_flags) &
7047 (IPH_MF | IPH_OFFSET)) == 0)
7048 return (mp);
7049
7050 /*
7051 * We utilize hardware computed checksum info only for UDP since
7052 * IP fragmentation is a normal occurrence for the protocol. In
7053 * addition, checksum offload support for IP fragments carrying
7054 * UDP payload is commonly implemented across network adapters.
7055 */
7056 ASSERT(ira->ira_rill != NULL);
7057 if (proto == IPPROTO_UDP && dohwcksum &&
7058 ILL_HCKSUM_CAPABLE(ira->ira_rill) &&
7059 (DB_CKSUMFLAGS(mp) & (HCK_FULLCKSUM | HCK_PARTIALCKSUM))) {
7060 mblk_t *mp1 = mp->b_cont;
7061 int32_t len;
7062
7063 /* Record checksum information from the packet */
7064 sum_val = (uint32_t)DB_CKSUM16(mp);
7065 sum_flags = DB_CKSUMFLAGS(mp);
7066
7067 /* IP payload offset from beginning of mblk */
7068 offset = ((uchar_t *)ipha + IPH_HDR_LENGTH(ipha)) - mp->b_rptr;
7069
7070 if ((sum_flags & HCK_PARTIALCKSUM) &&
7071 (mp1 == NULL || mp1->b_cont == NULL) &&
7072 offset >= DB_CKSUMSTART(mp) &&
7073 ((len = offset - DB_CKSUMSTART(mp)) & 1) == 0) {
7074 uint32_t adj;
7075 /*
7076 * Partial checksum has been calculated by hardware
7077 * and attached to the packet; in addition, any
7078 * prepended extraneous data is even byte aligned.
7079 * If any such data exists, we adjust the checksum;
7080 * this would also handle any postpended data.
7081 */
7082 IP_ADJCKSUM_PARTIAL(mp->b_rptr + DB_CKSUMSTART(mp),
7083 mp, mp1, len, adj);
7084
7085 /* One's complement subtract extraneous checksum */
7086 if (adj >= sum_val)
7087 sum_val = ~(adj - sum_val) & 0xFFFF;
7088 else
7089 sum_val -= adj;
7090 }
7091 } else {
7092 sum_val = 0;
7093 sum_flags = 0;
7094 }
7095
7096 /* Clear hardware checksumming flag */
7097 DB_CKSUMFLAGS(mp) = 0;
7098
7099 ident = ipha->ipha_ident;
7100 offset = (frag_offset_flags << 3) & 0xFFFF;
7101 src = ipha->ipha_src;
7102 dst = ipha->ipha_dst;
7103 hdr_length = IPH_HDR_LENGTH(ipha);
7104 end = ntohs(ipha->ipha_length) - hdr_length;
7105
7106 /* If end == 0 then we have a packet with no data, so just free it */
7107 if (end == 0) {
7108 freemsg(mp);
7109 return (NULL);
7110 }
7111
7112 /* Record the ECN field info. */
7113 ecn_info = (ipha->ipha_type_of_service & 0x3);
7114 if (offset != 0) {
7115 /*
7116 * If this isn't the first piece, strip the header, and
7117 * add the offset to the end value.
7118 */
7119 mp->b_rptr += hdr_length;
7120 end += offset;
7121 }
7122
7123 /* Handle vnic loopback of fragments */
7124 if (mp->b_datap->db_ref > 2)
7125 msg_len = 0;
7126 else
7127 msg_len = MBLKSIZE(mp);
7128
7129 tail_mp = mp;
7130 while (tail_mp->b_cont != NULL) {
7131 tail_mp = tail_mp->b_cont;
7132 if (tail_mp->b_datap->db_ref <= 2)
7133 msg_len += MBLKSIZE(tail_mp);
7134 }
7135
7136 /* If the reassembly list for this ILL will get too big, prune it */
7137 if ((msg_len + sizeof (*ipf) + ill->ill_frag_count) >=
7138 ipst->ips_ip_reass_queue_bytes) {
7139 DTRACE_PROBE3(ip_reass_queue_bytes, uint_t, msg_len,
7140 uint_t, ill->ill_frag_count,
7141 uint_t, ipst->ips_ip_reass_queue_bytes);
7142 ill_frag_prune(ill,
7143 (ipst->ips_ip_reass_queue_bytes < msg_len) ? 0 :
7144 (ipst->ips_ip_reass_queue_bytes - msg_len));
7145 pruned = B_TRUE;
7146 }
7147
7148 ipfb = &ill->ill_frag_hash_tbl[ILL_FRAG_HASH(src, ident)];
7149 mutex_enter(&ipfb->ipfb_lock);
7150
7151 ipfp = &ipfb->ipfb_ipf;
7152 /* Try to find an existing fragment queue for this packet. */
7153 for (;;) {
7154 ipf = ipfp[0];
7155 if (ipf != NULL) {
7156 /*
7157 * It has to match on ident and src/dst address.
7158 */
7159 if (ipf->ipf_ident == ident &&
7160 ipf->ipf_src == src &&
7161 ipf->ipf_dst == dst &&
7162 ipf->ipf_protocol == proto) {
7163 /*
7164 * If we have received too many
7165 * duplicate fragments for this packet
7166 * free it.
7167 */
7168 if (ipf->ipf_num_dups > ip_max_frag_dups) {
7169 ill_frag_free_pkts(ill, ipfb, ipf, 1);
7170 freemsg(mp);
7171 mutex_exit(&ipfb->ipfb_lock);
7172 return (NULL);
7173 }
7174 /* Found it. */
7175 break;
7176 }
7177 ipfp = &ipf->ipf_hash_next;
7178 continue;
7179 }
7180
7181 /*
7182 * If we pruned the list, do we want to store this new
7183 * fragment?. We apply an optimization here based on the
7184 * fact that most fragments will be received in order.
7185 * So if the offset of this incoming fragment is zero,
7186 * it is the first fragment of a new packet. We will
7187 * keep it. Otherwise drop the fragment, as we have
7188 * probably pruned the packet already (since the
7189 * packet cannot be found).
7190 */
7191 if (pruned && offset != 0) {
7192 mutex_exit(&ipfb->ipfb_lock);
7193 freemsg(mp);
7194 return (NULL);
7195 }
7196
7197 if (ipfb->ipfb_frag_pkts >= MAX_FRAG_PKTS(ipst)) {
7198 /*
7199 * Too many fragmented packets in this hash
7200 * bucket. Free the oldest.
7201 */
7202 ill_frag_free_pkts(ill, ipfb, ipfb->ipfb_ipf, 1);
7203 }
7204
7205 /* New guy. Allocate a frag message. */
7206 mp1 = allocb(sizeof (*ipf), BPRI_MED);
7207 if (mp1 == NULL) {
7208 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7209 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7210 freemsg(mp);
7211 reass_done:
7212 mutex_exit(&ipfb->ipfb_lock);
7213 return (NULL);
7214 }
7215
7216 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmReqds);
7217 mp1->b_cont = mp;
7218
7219 /* Initialize the fragment header. */
7220 ipf = (ipf_t *)mp1->b_rptr;
7221 ipf->ipf_mp = mp1;
7222 ipf->ipf_ptphn = ipfp;
7223 ipfp[0] = ipf;
7224 ipf->ipf_hash_next = NULL;
7225 ipf->ipf_ident = ident;
7226 ipf->ipf_protocol = proto;
7227 ipf->ipf_src = src;
7228 ipf->ipf_dst = dst;
7229 ipf->ipf_nf_hdr_len = 0;
7230 /* Record reassembly start time. */
7231 ipf->ipf_timestamp = gethrestime_sec();
7232 /* Record ipf generation and account for frag header */
7233 ipf->ipf_gen = ill->ill_ipf_gen++;
7234 ipf->ipf_count = MBLKSIZE(mp1);
7235 ipf->ipf_last_frag_seen = B_FALSE;
7236 ipf->ipf_ecn = ecn_info;
7237 ipf->ipf_num_dups = 0;
7238 ipfb->ipfb_frag_pkts++;
7239 ipf->ipf_checksum = 0;
7240 ipf->ipf_checksum_flags = 0;
7241
7242 /* Store checksum value in fragment header */
7243 if (sum_flags != 0) {
7244 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7245 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7246 ipf->ipf_checksum = sum_val;
7247 ipf->ipf_checksum_flags = sum_flags;
7248 }
7249
7250 /*
7251 * We handle reassembly two ways. In the easy case,
7252 * where all the fragments show up in order, we do
7253 * minimal bookkeeping, and just clip new pieces on
7254 * the end. If we ever see a hole, then we go off
7255 * to ip_reassemble which has to mark the pieces and
7256 * keep track of the number of holes, etc. Obviously,
7257 * the point of having both mechanisms is so we can
7258 * handle the easy case as efficiently as possible.
7259 */
7260 if (offset == 0) {
7261 /* Easy case, in-order reassembly so far. */
7262 ipf->ipf_count += msg_len;
7263 ipf->ipf_tail_mp = tail_mp;
7264 /*
7265 * Keep track of next expected offset in
7266 * ipf_end.
7267 */
7268 ipf->ipf_end = end;
7269 ipf->ipf_nf_hdr_len = hdr_length;
7270 } else {
7271 /* Hard case, hole at the beginning. */
7272 ipf->ipf_tail_mp = NULL;
7273 /*
7274 * ipf_end == 0 means that we have given up
7275 * on easy reassembly.
7276 */
7277 ipf->ipf_end = 0;
7278
7279 /* Forget checksum offload from now on */
7280 ipf->ipf_checksum_flags = 0;
7281
7282 /*
7283 * ipf_hole_cnt is set by ip_reassemble.
7284 * ipf_count is updated by ip_reassemble.
7285 * No need to check for return value here
7286 * as we don't expect reassembly to complete
7287 * or fail for the first fragment itself.
7288 */
7289 (void) ip_reassemble(mp, ipf,
7290 (frag_offset_flags & IPH_OFFSET) << 3,
7291 (frag_offset_flags & IPH_MF), ill, msg_len);
7292 }
7293 /* Update per ipfb and ill byte counts */
7294 ipfb->ipfb_count += ipf->ipf_count;
7295 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7296 atomic_add_32(&ill->ill_frag_count, ipf->ipf_count);
7297 /* If the frag timer wasn't already going, start it. */
7298 mutex_enter(&ill->ill_lock);
7299 ill_frag_timer_start(ill);
7300 mutex_exit(&ill->ill_lock);
7301 goto reass_done;
7302 }
7303
7304 /*
7305 * If the packet's flag has changed (it could be coming up
7306 * from an interface different than the previous, therefore
7307 * possibly different checksum capability), then forget about
7308 * any stored checksum states. Otherwise add the value to
7309 * the existing one stored in the fragment header.
7310 */
7311 if (sum_flags != 0 && sum_flags == ipf->ipf_checksum_flags) {
7312 sum_val += ipf->ipf_checksum;
7313 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7314 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7315 ipf->ipf_checksum = sum_val;
7316 } else if (ipf->ipf_checksum_flags != 0) {
7317 /* Forget checksum offload from now on */
7318 ipf->ipf_checksum_flags = 0;
7319 }
7320
7321 /*
7322 * We have a new piece of a datagram which is already being
7323 * reassembled. Update the ECN info if all IP fragments
7324 * are ECN capable. If there is one which is not, clear
7325 * all the info. If there is at least one which has CE
7326 * code point, IP needs to report that up to transport.
7327 */
7328 if (ecn_info != IPH_ECN_NECT && ipf->ipf_ecn != IPH_ECN_NECT) {
7329 if (ecn_info == IPH_ECN_CE)
7330 ipf->ipf_ecn = IPH_ECN_CE;
7331 } else {
7332 ipf->ipf_ecn = IPH_ECN_NECT;
7333 }
7334 if (offset && ipf->ipf_end == offset) {
7335 /* The new fragment fits at the end */
7336 ipf->ipf_tail_mp->b_cont = mp;
7337 /* Update the byte count */
7338 ipf->ipf_count += msg_len;
7339 /* Update per ipfb and ill byte counts */
7340 ipfb->ipfb_count += msg_len;
7341 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7342 atomic_add_32(&ill->ill_frag_count, msg_len);
7343 if (frag_offset_flags & IPH_MF) {
7344 /* More to come. */
7345 ipf->ipf_end = end;
7346 ipf->ipf_tail_mp = tail_mp;
7347 goto reass_done;
7348 }
7349 } else {
7350 /* Go do the hard cases. */
7351 int ret;
7352
7353 if (offset == 0)
7354 ipf->ipf_nf_hdr_len = hdr_length;
7355
7356 /* Save current byte count */
7357 count = ipf->ipf_count;
7358 ret = ip_reassemble(mp, ipf,
7359 (frag_offset_flags & IPH_OFFSET) << 3,
7360 (frag_offset_flags & IPH_MF), ill, msg_len);
7361 /* Count of bytes added and subtracted (freeb()ed) */
7362 count = ipf->ipf_count - count;
7363 if (count) {
7364 /* Update per ipfb and ill byte counts */
7365 ipfb->ipfb_count += count;
7366 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7367 atomic_add_32(&ill->ill_frag_count, count);
7368 }
7369 if (ret == IP_REASS_PARTIAL) {
7370 goto reass_done;
7371 } else if (ret == IP_REASS_FAILED) {
7372 /* Reassembly failed. Free up all resources */
7373 ill_frag_free_pkts(ill, ipfb, ipf, 1);
7374 for (t_mp = mp; t_mp != NULL; t_mp = t_mp->b_cont) {
7375 IP_REASS_SET_START(t_mp, 0);
7376 IP_REASS_SET_END(t_mp, 0);
7377 }
7378 freemsg(mp);
7379 goto reass_done;
7380 }
7381 /* We will reach here iff 'ret' is IP_REASS_COMPLETE */
7382 }
7383 /*
7384 * We have completed reassembly. Unhook the frag header from
7385 * the reassembly list.
7386 *
7387 * Before we free the frag header, record the ECN info
7388 * to report back to the transport.
7389 */
7390 ecn_info = ipf->ipf_ecn;
7391 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmOKs);
7392 ipfp = ipf->ipf_ptphn;
7393
7394 /* We need to supply these to caller */
7395 if ((sum_flags = ipf->ipf_checksum_flags) != 0)
7396 sum_val = ipf->ipf_checksum;
7397 else
7398 sum_val = 0;
7399
7400 mp1 = ipf->ipf_mp;
7401 count = ipf->ipf_count;
7402 ipf = ipf->ipf_hash_next;
7403 if (ipf != NULL)
7404 ipf->ipf_ptphn = ipfp;
7405 ipfp[0] = ipf;
7406 atomic_add_32(&ill->ill_frag_count, -count);
7407 ASSERT(ipfb->ipfb_count >= count);
7408 ipfb->ipfb_count -= count;
7409 ipfb->ipfb_frag_pkts--;
7410 mutex_exit(&ipfb->ipfb_lock);
7411 /* Ditch the frag header. */
7412 mp = mp1->b_cont;
7413
7414 freeb(mp1);
7415
7416 /* Restore original IP length in header. */
7417 packet_size = (uint32_t)msgdsize(mp);
7418 if (packet_size > IP_MAXPACKET) {
7419 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7420 ip_drop_input("Reassembled packet too large", mp, ill);
7421 freemsg(mp);
7422 return (NULL);
7423 }
7424
7425 if (DB_REF(mp) > 1) {
7426 mblk_t *mp2 = copymsg(mp);
7427
7428 if (mp2 == NULL) {
7429 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7430 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7431 freemsg(mp);
7432 return (NULL);
7433 }
7434 freemsg(mp);
7435 mp = mp2;
7436 }
7437 ipha = (ipha_t *)mp->b_rptr;
7438
7439 ipha->ipha_length = htons((uint16_t)packet_size);
7440 /* We're now complete, zip the frag state */
7441 ipha->ipha_fragment_offset_and_flags = 0;
7442 /* Record the ECN info. */
7443 ipha->ipha_type_of_service &= 0xFC;
7444 ipha->ipha_type_of_service |= ecn_info;
7445
7446 /* Update the receive attributes */
7447 ira->ira_pktlen = packet_size;
7448 ira->ira_ip_hdr_length = IPH_HDR_LENGTH(ipha);
7449
7450 /* Reassembly is successful; set checksum information in packet */
7451 DB_CKSUM16(mp) = (uint16_t)sum_val;
7452 DB_CKSUMFLAGS(mp) = sum_flags;
7453 DB_CKSUMSTART(mp) = ira->ira_ip_hdr_length;
7454
7455 return (mp);
7456 }
7457
7458 /*
7459 * Pullup function that should be used for IP input in order to
7460 * ensure we do not loose the L2 source address; we need the l2 source
7461 * address for IP_RECVSLLA and for ndp_input.
7462 *
7463 * We return either NULL or b_rptr.
7464 */
7465 void *
7466 ip_pullup(mblk_t *mp, ssize_t len, ip_recv_attr_t *ira)
7467 {
7468 ill_t *ill = ira->ira_ill;
7469
7470 if (ip_rput_pullups++ == 0) {
7471 (void) mi_strlog(ill->ill_rq, 1, SL_ERROR|SL_TRACE,
7472 "ip_pullup: %s forced us to "
7473 " pullup pkt, hdr len %ld, hdr addr %p",
7474 ill->ill_name, len, (void *)mp->b_rptr);
7475 }
7476 if (!(ira->ira_flags & IRAF_L2SRC_SET))
7477 ip_setl2src(mp, ira, ira->ira_rill);
7478 ASSERT(ira->ira_flags & IRAF_L2SRC_SET);
7479 if (!pullupmsg(mp, len))
7480 return (NULL);
7481 else
7482 return (mp->b_rptr);
7483 }
7484
7485 /*
7486 * Make sure ira_l2src has an address. If we don't have one fill with zeros.
7487 * When called from the ULP ira_rill will be NULL hence the caller has to
7488 * pass in the ill.
7489 */
7490 /* ARGSUSED */
7491 void
7492 ip_setl2src(mblk_t *mp, ip_recv_attr_t *ira, ill_t *ill)
7493 {
7494 const uchar_t *addr;
7495 int alen;
7496
7497 if (ira->ira_flags & IRAF_L2SRC_SET)
7498 return;
7499
7500 ASSERT(ill != NULL);
7501 alen = ill->ill_phys_addr_length;
7502 ASSERT(alen <= sizeof (ira->ira_l2src));
7503 if (ira->ira_mhip != NULL &&
7504 (addr = ira->ira_mhip->mhi_saddr) != NULL) {
7505 bcopy(addr, ira->ira_l2src, alen);
7506 } else if ((ira->ira_flags & IRAF_L2SRC_LOOPBACK) &&
7507 (addr = ill->ill_phys_addr) != NULL) {
7508 bcopy(addr, ira->ira_l2src, alen);
7509 } else {
7510 bzero(ira->ira_l2src, alen);
7511 }
7512 ira->ira_flags |= IRAF_L2SRC_SET;
7513 }
7514
7515 /*
7516 * check ip header length and align it.
7517 */
7518 mblk_t *
7519 ip_check_and_align_header(mblk_t *mp, uint_t min_size, ip_recv_attr_t *ira)
7520 {
7521 ill_t *ill = ira->ira_ill;
7522 ssize_t len;
7523
7524 len = MBLKL(mp);
7525
7526 if (!OK_32PTR(mp->b_rptr))
7527 IP_STAT(ill->ill_ipst, ip_notaligned);
7528 else
7529 IP_STAT(ill->ill_ipst, ip_recv_pullup);
7530
7531 /* Guard against bogus device drivers */
7532 if (len < 0) {
7533 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7534 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7535 freemsg(mp);
7536 return (NULL);
7537 }
7538
7539 if (len == 0) {
7540 /* GLD sometimes sends up mblk with b_rptr == b_wptr! */
7541 mblk_t *mp1 = mp->b_cont;
7542
7543 if (!(ira->ira_flags & IRAF_L2SRC_SET))
7544 ip_setl2src(mp, ira, ira->ira_rill);
7545 ASSERT(ira->ira_flags & IRAF_L2SRC_SET);
7546
7547 freeb(mp);
7548 mp = mp1;
7549 if (mp == NULL)
7550 return (NULL);
7551
7552 if (OK_32PTR(mp->b_rptr) && MBLKL(mp) >= min_size)
7553 return (mp);
7554 }
7555 if (ip_pullup(mp, min_size, ira) == NULL) {
7556 if (msgdsize(mp) < min_size) {
7557 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7558 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7559 } else {
7560 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7561 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7562 }
7563 freemsg(mp);
7564 return (NULL);
7565 }
7566 return (mp);
7567 }
7568
7569 /*
7570 * Common code for IPv4 and IPv6 to check and pullup multi-mblks
7571 */
7572 mblk_t *
7573 ip_check_length(mblk_t *mp, uchar_t *rptr, ssize_t len, uint_t pkt_len,
7574 uint_t min_size, ip_recv_attr_t *ira)
7575 {
7576 ill_t *ill = ira->ira_ill;
7577
7578 /*
7579 * Make sure we have data length consistent
7580 * with the IP header.
7581 */
7582 if (mp->b_cont == NULL) {
7583 /* pkt_len is based on ipha_len, not the mblk length */
7584 if (pkt_len < min_size) {
7585 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7586 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7587 freemsg(mp);
7588 return (NULL);
7589 }
7590 if (len < 0) {
7591 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
7592 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
7593 freemsg(mp);
7594 return (NULL);
7595 }
7596 /* Drop any pad */
7597 mp->b_wptr = rptr + pkt_len;
7598 } else if ((len += msgdsize(mp->b_cont)) != 0) {
7599 ASSERT(pkt_len >= min_size);
7600 if (pkt_len < min_size) {
7601 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7602 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7603 freemsg(mp);
7604 return (NULL);
7605 }
7606 if (len < 0) {
7607 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
7608 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
7609 freemsg(mp);
7610 return (NULL);
7611 }
7612 /* Drop any pad */
7613 (void) adjmsg(mp, -len);
7614 /*
7615 * adjmsg may have freed an mblk from the chain, hence
7616 * invalidate any hw checksum here. This will force IP to
7617 * calculate the checksum in sw, but only for this packet.
7618 */
7619 DB_CKSUMFLAGS(mp) = 0;
7620 IP_STAT(ill->ill_ipst, ip_multimblk);
7621 }
7622 return (mp);
7623 }
7624
7625 /*
7626 * Check that the IPv4 opt_len is consistent with the packet and pullup
7627 * the options.
7628 */
7629 mblk_t *
7630 ip_check_optlen(mblk_t *mp, ipha_t *ipha, uint_t opt_len, uint_t pkt_len,
7631 ip_recv_attr_t *ira)
7632 {
7633 ill_t *ill = ira->ira_ill;
7634 ssize_t len;
7635
7636 /* Assume no IPv6 packets arrive over the IPv4 queue */
7637 if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
7638 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7639 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInWrongIPVersion);
7640 ip_drop_input("IPvN packet on IPv4 ill", mp, ill);
7641 freemsg(mp);
7642 return (NULL);
7643 }
7644
7645 if (opt_len > (15 - IP_SIMPLE_HDR_LENGTH_IN_WORDS)) {
7646 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7647 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7648 freemsg(mp);
7649 return (NULL);
7650 }
7651 /*
7652 * Recompute complete header length and make sure we
7653 * have access to all of it.
7654 */
7655 len = ((size_t)opt_len + IP_SIMPLE_HDR_LENGTH_IN_WORDS) << 2;
7656 if (len > (mp->b_wptr - mp->b_rptr)) {
7657 if (len > pkt_len) {
7658 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7659 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7660 freemsg(mp);
7661 return (NULL);
7662 }
7663 if (ip_pullup(mp, len, ira) == NULL) {
7664 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7665 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7666 freemsg(mp);
7667 return (NULL);
7668 }
7669 }
7670 return (mp);
7671 }
7672
7673 /*
7674 * Returns a new ire, or the same ire, or NULL.
7675 * If a different IRE is returned, then it is held; the caller
7676 * needs to release it.
7677 * In no case is there any hold/release on the ire argument.
7678 */
7679 ire_t *
7680 ip_check_multihome(void *addr, ire_t *ire, ill_t *ill)
7681 {
7682 ire_t *new_ire;
7683 ill_t *ire_ill;
7684 uint_t ifindex;
7685 ip_stack_t *ipst = ill->ill_ipst;
7686 boolean_t strict_check = B_FALSE;
7687
7688 /*
7689 * IPMP common case: if IRE and ILL are in the same group, there's no
7690 * issue (e.g. packet received on an underlying interface matched an
7691 * IRE_LOCAL on its associated group interface).
7692 */
7693 ASSERT(ire->ire_ill != NULL);
7694 if (IS_IN_SAME_ILLGRP(ill, ire->ire_ill))
7695 return (ire);
7696
7697 /*
7698 * Do another ire lookup here, using the ingress ill, to see if the
7699 * interface is in a usesrc group.
7700 * As long as the ills belong to the same group, we don't consider
7701 * them to be arriving on the wrong interface. Thus, if the switch
7702 * is doing inbound load spreading, we won't drop packets when the
7703 * ip*_strict_dst_multihoming switch is on.
7704 * We also need to check for IPIF_UNNUMBERED point2point interfaces
7705 * where the local address may not be unique. In this case we were
7706 * at the mercy of the initial ire lookup and the IRE_LOCAL it
7707 * actually returned. The new lookup, which is more specific, should
7708 * only find the IRE_LOCAL associated with the ingress ill if one
7709 * exists.
7710 */
7711 if (ire->ire_ipversion == IPV4_VERSION) {
7712 if (ipst->ips_ip_strict_dst_multihoming)
7713 strict_check = B_TRUE;
7714 new_ire = ire_ftable_lookup_v4(*((ipaddr_t *)addr), 0, 0,
7715 IRE_LOCAL, ill, ALL_ZONES, NULL,
7716 (MATCH_IRE_TYPE|MATCH_IRE_ILL), 0, ipst, NULL);
7717 } else {
7718 ASSERT(!IN6_IS_ADDR_MULTICAST((in6_addr_t *)addr));
7719 if (ipst->ips_ipv6_strict_dst_multihoming)
7720 strict_check = B_TRUE;
7721 new_ire = ire_ftable_lookup_v6((in6_addr_t *)addr, NULL, NULL,
7722 IRE_LOCAL, ill, ALL_ZONES, NULL,
7723 (MATCH_IRE_TYPE|MATCH_IRE_ILL), 0, ipst, NULL);
7724 }
7725 /*
7726 * If the same ire that was returned in ip_input() is found then this
7727 * is an indication that usesrc groups are in use. The packet
7728 * arrived on a different ill in the group than the one associated with
7729 * the destination address. If a different ire was found then the same
7730 * IP address must be hosted on multiple ills. This is possible with
7731 * unnumbered point2point interfaces. We switch to use this new ire in
7732 * order to have accurate interface statistics.
7733 */
7734 if (new_ire != NULL) {
7735 /* Note: held in one case but not the other? Caller handles */
7736 if (new_ire != ire)
7737 return (new_ire);
7738 /* Unchanged */
7739 ire_refrele(new_ire);
7740 return (ire);
7741 }
7742
7743 /*
7744 * Chase pointers once and store locally.
7745 */
7746 ASSERT(ire->ire_ill != NULL);
7747 ire_ill = ire->ire_ill;
7748 ifindex = ill->ill_usesrc_ifindex;
7749
7750 /*
7751 * Check if it's a legal address on the 'usesrc' interface.
7752 * For IPMP data addresses the IRE_LOCAL is the upper, hence we
7753 * can just check phyint_ifindex.
7754 */
7755 if (ifindex != 0 && ifindex == ire_ill->ill_phyint->phyint_ifindex) {
7756 return (ire);
7757 }
7758
7759 /*
7760 * If the ip*_strict_dst_multihoming switch is on then we can
7761 * only accept this packet if the interface is marked as routing.
7762 */
7763 if (!(strict_check))
7764 return (ire);
7765
7766 if ((ill->ill_flags & ire->ire_ill->ill_flags & ILLF_ROUTER) != 0) {
7767 return (ire);
7768 }
7769 return (NULL);
7770 }
7771
7772 /*
7773 * This function is used to construct a mac_header_info_s from a
7774 * DL_UNITDATA_IND message.
7775 * The address fields in the mhi structure points into the message,
7776 * thus the caller can't use those fields after freeing the message.
7777 *
7778 * We determine whether the packet received is a non-unicast packet
7779 * and in doing so, determine whether or not it is broadcast vs multicast.
7780 * For it to be a broadcast packet, we must have the appropriate mblk_t
7781 * hanging off the ill_t. If this is either not present or doesn't match
7782 * the destination mac address in the DL_UNITDATA_IND, the packet is deemed
7783 * to be multicast. Thus NICs that have no broadcast address (or no
7784 * capability for one, such as point to point links) cannot return as
7785 * the packet being broadcast.
7786 */
7787 void
7788 ip_dlur_to_mhi(ill_t *ill, mblk_t *mb, struct mac_header_info_s *mhip)
7789 {
7790 dl_unitdata_ind_t *ind = (dl_unitdata_ind_t *)mb->b_rptr;
7791 mblk_t *bmp;
7792 uint_t extra_offset;
7793
7794 bzero(mhip, sizeof (struct mac_header_info_s));
7795
7796 mhip->mhi_dsttype = MAC_ADDRTYPE_UNICAST;
7797
7798 if (ill->ill_sap_length < 0)
7799 extra_offset = 0;
7800 else
7801 extra_offset = ill->ill_sap_length;
7802
7803 mhip->mhi_daddr = (uchar_t *)ind + ind->dl_dest_addr_offset +
7804 extra_offset;
7805 mhip->mhi_saddr = (uchar_t *)ind + ind->dl_src_addr_offset +
7806 extra_offset;
7807
7808 if (!ind->dl_group_address)
7809 return;
7810
7811 /* Multicast or broadcast */
7812 mhip->mhi_dsttype = MAC_ADDRTYPE_MULTICAST;
7813
7814 if (ind->dl_dest_addr_offset > sizeof (*ind) &&
7815 ind->dl_dest_addr_offset + ind->dl_dest_addr_length < MBLKL(mb) &&
7816 (bmp = ill->ill_bcast_mp) != NULL) {
7817 dl_unitdata_req_t *dlur;
7818 uint8_t *bphys_addr;
7819
7820 dlur = (dl_unitdata_req_t *)bmp->b_rptr;
7821 bphys_addr = (uchar_t *)dlur + dlur->dl_dest_addr_offset +
7822 extra_offset;
7823
7824 if (bcmp(mhip->mhi_daddr, bphys_addr,
7825 ind->dl_dest_addr_length) == 0)
7826 mhip->mhi_dsttype = MAC_ADDRTYPE_BROADCAST;
7827 }
7828 }
7829
7830 /*
7831 * This function is used to construct a mac_header_info_s from a
7832 * M_DATA fastpath message from a DLPI driver.
7833 * The address fields in the mhi structure points into the message,
7834 * thus the caller can't use those fields after freeing the message.
7835 *
7836 * We determine whether the packet received is a non-unicast packet
7837 * and in doing so, determine whether or not it is broadcast vs multicast.
7838 * For it to be a broadcast packet, we must have the appropriate mblk_t
7839 * hanging off the ill_t. If this is either not present or doesn't match
7840 * the destination mac address in the DL_UNITDATA_IND, the packet is deemed
7841 * to be multicast. Thus NICs that have no broadcast address (or no
7842 * capability for one, such as point to point links) cannot return as
7843 * the packet being broadcast.
7844 */
7845 void
7846 ip_mdata_to_mhi(ill_t *ill, mblk_t *mp, struct mac_header_info_s *mhip)
7847 {
7848 mblk_t *bmp;
7849 struct ether_header *pether;
7850
7851 bzero(mhip, sizeof (struct mac_header_info_s));
7852
7853 mhip->mhi_dsttype = MAC_ADDRTYPE_UNICAST;
7854
7855 pether = (struct ether_header *)((char *)mp->b_rptr
7856 - sizeof (struct ether_header));
7857
7858 /*
7859 * Make sure the interface is an ethernet type, since we don't
7860 * know the header format for anything but Ethernet. Also make
7861 * sure we are pointing correctly above db_base.
7862 */
7863 if (ill->ill_type != IFT_ETHER)
7864 return;
7865
7866 retry:
7867 if ((uchar_t *)pether < mp->b_datap->db_base)
7868 return;
7869
7870 /* Is there a VLAN tag? */
7871 if (ill->ill_isv6) {
7872 if (pether->ether_type != htons(ETHERTYPE_IPV6)) {
7873 pether = (struct ether_header *)((char *)pether - 4);
7874 goto retry;
7875 }
7876 } else {
7877 if (pether->ether_type != htons(ETHERTYPE_IP)) {
7878 pether = (struct ether_header *)((char *)pether - 4);
7879 goto retry;
7880 }
7881 }
7882 mhip->mhi_daddr = (uchar_t *)&pether->ether_dhost;
7883 mhip->mhi_saddr = (uchar_t *)&pether->ether_shost;
7884
7885 if (!(mhip->mhi_daddr[0] & 0x01))
7886 return;
7887
7888 /* Multicast or broadcast */
7889 mhip->mhi_dsttype = MAC_ADDRTYPE_MULTICAST;
7890
7891 if ((bmp = ill->ill_bcast_mp) != NULL) {
7892 dl_unitdata_req_t *dlur;
7893 uint8_t *bphys_addr;
7894 uint_t addrlen;
7895
7896 dlur = (dl_unitdata_req_t *)bmp->b_rptr;
7897 addrlen = dlur->dl_dest_addr_length;
7898 if (ill->ill_sap_length < 0) {
7899 bphys_addr = (uchar_t *)dlur +
7900 dlur->dl_dest_addr_offset;
7901 addrlen += ill->ill_sap_length;
7902 } else {
7903 bphys_addr = (uchar_t *)dlur +
7904 dlur->dl_dest_addr_offset +
7905 ill->ill_sap_length;
7906 addrlen -= ill->ill_sap_length;
7907 }
7908 if (bcmp(mhip->mhi_daddr, bphys_addr, addrlen) == 0)
7909 mhip->mhi_dsttype = MAC_ADDRTYPE_BROADCAST;
7910 }
7911 }
7912
7913 /*
7914 * Handle anything but M_DATA messages
7915 * We see the DL_UNITDATA_IND which are part
7916 * of the data path, and also the other messages from the driver.
7917 */
7918 void
7919 ip_rput_notdata(ill_t *ill, mblk_t *mp)
7920 {
7921 mblk_t *first_mp;
7922 struct iocblk *iocp;
7923 struct mac_header_info_s mhi;
7924
7925 switch (DB_TYPE(mp)) {
7926 case M_PROTO:
7927 case M_PCPROTO: {
7928 if (((dl_unitdata_ind_t *)mp->b_rptr)->dl_primitive !=
7929 DL_UNITDATA_IND) {
7930 /* Go handle anything other than data elsewhere. */
7931 ip_rput_dlpi(ill, mp);
7932 return;
7933 }
7934
7935 first_mp = mp;
7936 mp = first_mp->b_cont;
7937 first_mp->b_cont = NULL;
7938
7939 if (mp == NULL) {
7940 freeb(first_mp);
7941 return;
7942 }
7943 ip_dlur_to_mhi(ill, first_mp, &mhi);
7944 if (ill->ill_isv6)
7945 ip_input_v6(ill, NULL, mp, &mhi);
7946 else
7947 ip_input(ill, NULL, mp, &mhi);
7948
7949 /* Ditch the DLPI header. */
7950 freeb(first_mp);
7951 return;
7952 }
7953 case M_IOCACK:
7954 iocp = (struct iocblk *)mp->b_rptr;
7955 switch (iocp->ioc_cmd) {
7956 case DL_IOC_HDR_INFO:
7957 ill_fastpath_ack(ill, mp);
7958 return;
7959 default:
7960 putnext(ill->ill_rq, mp);
7961 return;
7962 }
7963 /* FALLTHROUGH */
7964 case M_ERROR:
7965 case M_HANGUP:
7966 mutex_enter(&ill->ill_lock);
7967 if (ill->ill_state_flags & ILL_CONDEMNED) {
7968 mutex_exit(&ill->ill_lock);
7969 freemsg(mp);
7970 return;
7971 }
7972 ill_refhold_locked(ill);
7973 mutex_exit(&ill->ill_lock);
7974 qwriter_ip(ill, ill->ill_rq, mp, ip_rput_other, CUR_OP,
7975 B_FALSE);
7976 return;
7977 case M_CTL:
7978 putnext(ill->ill_rq, mp);
7979 return;
7980 case M_IOCNAK:
7981 ip1dbg(("got iocnak "));
7982 iocp = (struct iocblk *)mp->b_rptr;
7983 switch (iocp->ioc_cmd) {
7984 case DL_IOC_HDR_INFO:
7985 ip_rput_other(NULL, ill->ill_rq, mp, NULL);
7986 return;
7987 default:
7988 break;
7989 }
7990 /* FALLTHROUGH */
7991 default:
7992 putnext(ill->ill_rq, mp);
7993 return;
7994 }
7995 }
7996
7997 /* Read side put procedure. Packets coming from the wire arrive here. */
7998 int
7999 ip_rput(queue_t *q, mblk_t *mp)
8000 {
8001 ill_t *ill;
8002 union DL_primitives *dl;
8003
8004 ill = (ill_t *)q->q_ptr;
8005
8006 if (ill->ill_state_flags & (ILL_CONDEMNED | ILL_LL_SUBNET_PENDING)) {
8007 /*
8008 * If things are opening or closing, only accept high-priority
8009 * DLPI messages. (On open ill->ill_ipif has not yet been
8010 * created; on close, things hanging off the ill may have been
8011 * freed already.)
8012 */
8013 dl = (union DL_primitives *)mp->b_rptr;
8014 if (DB_TYPE(mp) != M_PCPROTO ||
8015 dl->dl_primitive == DL_UNITDATA_IND) {
8016 inet_freemsg(mp);
8017 return (0);
8018 }
8019 }
8020 if (DB_TYPE(mp) == M_DATA) {
8021 struct mac_header_info_s mhi;
8022
8023 ip_mdata_to_mhi(ill, mp, &mhi);
8024 ip_input(ill, NULL, mp, &mhi);
8025 } else {
8026 ip_rput_notdata(ill, mp);
8027 }
8028 return (0);
8029 }
8030
8031 /*
8032 * Move the information to a copy.
8033 */
8034 mblk_t *
8035 ip_fix_dbref(mblk_t *mp, ip_recv_attr_t *ira)
8036 {
8037 mblk_t *mp1;
8038 ill_t *ill = ira->ira_ill;
8039 ip_stack_t *ipst = ill->ill_ipst;
8040
8041 IP_STAT(ipst, ip_db_ref);
8042
8043 /* Make sure we have ira_l2src before we loose the original mblk */
8044 if (!(ira->ira_flags & IRAF_L2SRC_SET))
8045 ip_setl2src(mp, ira, ira->ira_rill);
8046
8047 mp1 = copymsg(mp);
8048 if (mp1 == NULL) {
8049 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
8050 ip_drop_input("ipIfStatsInDiscards", mp, ill);
8051 freemsg(mp);
8052 return (NULL);
8053 }
8054 /* preserve the hardware checksum flags and data, if present */
8055 if (DB_CKSUMFLAGS(mp) != 0) {
8056 DB_CKSUMFLAGS(mp1) = DB_CKSUMFLAGS(mp);
8057 DB_CKSUMSTART(mp1) = DB_CKSUMSTART(mp);
8058 DB_CKSUMSTUFF(mp1) = DB_CKSUMSTUFF(mp);
8059 DB_CKSUMEND(mp1) = DB_CKSUMEND(mp);
8060 DB_CKSUM16(mp1) = DB_CKSUM16(mp);
8061 }
8062 freemsg(mp);
8063 return (mp1);
8064 }
8065
8066 static void
8067 ip_dlpi_error(ill_t *ill, t_uscalar_t prim, t_uscalar_t dl_err,
8068 t_uscalar_t err)
8069 {
8070 if (dl_err == DL_SYSERR) {
8071 (void) mi_strlog(ill->ill_rq, 1, SL_CONSOLE|SL_ERROR|SL_TRACE,
8072 "%s: %s failed: DL_SYSERR (errno %u)\n",
8073 ill->ill_name, dl_primstr(prim), err);
8074 return;
8075 }
8076
8077 (void) mi_strlog(ill->ill_rq, 1, SL_CONSOLE|SL_ERROR|SL_TRACE,
8078 "%s: %s failed: %s\n", ill->ill_name, dl_primstr(prim),
8079 dl_errstr(dl_err));
8080 }
8081
8082 /*
8083 * ip_rput_dlpi is called by ip_rput to handle all DLPI messages other
8084 * than DL_UNITDATA_IND messages. If we need to process this message
8085 * exclusively, we call qwriter_ip, in which case we also need to call
8086 * ill_refhold before that, since qwriter_ip does an ill_refrele.
8087 */
8088 void
8089 ip_rput_dlpi(ill_t *ill, mblk_t *mp)
8090 {
8091 dl_ok_ack_t *dloa = (dl_ok_ack_t *)mp->b_rptr;
8092 dl_error_ack_t *dlea = (dl_error_ack_t *)dloa;
8093 queue_t *q = ill->ill_rq;
8094 t_uscalar_t prim = dloa->dl_primitive;
8095 t_uscalar_t reqprim = DL_PRIM_INVAL;
8096
8097 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi",
8098 char *, dl_primstr(prim), ill_t *, ill);
8099 ip1dbg(("ip_rput_dlpi"));
8100
8101 /*
8102 * If we received an ACK but didn't send a request for it, then it
8103 * can't be part of any pending operation; discard up-front.
8104 */
8105 switch (prim) {
8106 case DL_ERROR_ACK:
8107 reqprim = dlea->dl_error_primitive;
8108 ip2dbg(("ip_rput_dlpi(%s): DL_ERROR_ACK for %s (0x%x): %s "
8109 "(0x%x), unix %u\n", ill->ill_name, dl_primstr(reqprim),
8110 reqprim, dl_errstr(dlea->dl_errno), dlea->dl_errno,
8111 dlea->dl_unix_errno));
8112 break;
8113 case DL_OK_ACK:
8114 reqprim = dloa->dl_correct_primitive;
8115 break;
8116 case DL_INFO_ACK:
8117 reqprim = DL_INFO_REQ;
8118 break;
8119 case DL_BIND_ACK:
8120 reqprim = DL_BIND_REQ;
8121 break;
8122 case DL_PHYS_ADDR_ACK:
8123 reqprim = DL_PHYS_ADDR_REQ;
8124 break;
8125 case DL_NOTIFY_ACK:
8126 reqprim = DL_NOTIFY_REQ;
8127 break;
8128 case DL_CAPABILITY_ACK:
8129 reqprim = DL_CAPABILITY_REQ;
8130 break;
8131 }
8132
8133 if (prim != DL_NOTIFY_IND) {
8134 if (reqprim == DL_PRIM_INVAL ||
8135 !ill_dlpi_pending(ill, reqprim)) {
8136 /* Not a DLPI message we support or expected */
8137 freemsg(mp);
8138 return;
8139 }
8140 ip1dbg(("ip_rput: received %s for %s\n", dl_primstr(prim),
8141 dl_primstr(reqprim)));
8142 }
8143
8144 switch (reqprim) {
8145 case DL_UNBIND_REQ:
8146 /*
8147 * NOTE: we mark the unbind as complete even if we got a
8148 * DL_ERROR_ACK, since there's not much else we can do.
8149 */
8150 mutex_enter(&ill->ill_lock);
8151 ill->ill_state_flags &= ~ILL_DL_UNBIND_IN_PROGRESS;
8152 cv_signal(&ill->ill_cv);
8153 mutex_exit(&ill->ill_lock);
8154 break;
8155
8156 case DL_ENABMULTI_REQ:
8157 if (prim == DL_OK_ACK) {
8158 if (ill->ill_dlpi_multicast_state == IDS_INPROGRESS)
8159 ill->ill_dlpi_multicast_state = IDS_OK;
8160 }
8161 break;
8162 }
8163
8164 /*
8165 * The message is one we're waiting for (or DL_NOTIFY_IND), but we
8166 * need to become writer to continue to process it. Because an
8167 * exclusive operation doesn't complete until replies to all queued
8168 * DLPI messages have been received, we know we're in the middle of an
8169 * exclusive operation and pass CUR_OP (except for DL_NOTIFY_IND).
8170 *
8171 * As required by qwriter_ip(), we refhold the ill; it will refrele.
8172 * Since this is on the ill stream we unconditionally bump up the
8173 * refcount without doing ILL_CAN_LOOKUP().
8174 */
8175 ill_refhold(ill);
8176 if (prim == DL_NOTIFY_IND)
8177 qwriter_ip(ill, q, mp, ip_rput_dlpi_writer, NEW_OP, B_FALSE);
8178 else
8179 qwriter_ip(ill, q, mp, ip_rput_dlpi_writer, CUR_OP, B_FALSE);
8180 }
8181
8182 /*
8183 * Handling of DLPI messages that require exclusive access to the ipsq.
8184 *
8185 * Need to do ipsq_pending_mp_get on ioctl completion, which could
8186 * happen here. (along with mi_copy_done)
8187 */
8188 /* ARGSUSED */
8189 static void
8190 ip_rput_dlpi_writer(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
8191 {
8192 dl_ok_ack_t *dloa = (dl_ok_ack_t *)mp->b_rptr;
8193 dl_error_ack_t *dlea = (dl_error_ack_t *)dloa;
8194 int err = 0;
8195 ill_t *ill = (ill_t *)q->q_ptr;
8196 ipif_t *ipif = NULL;
8197 mblk_t *mp1 = NULL;
8198 conn_t *connp = NULL;
8199 t_uscalar_t paddrreq;
8200 mblk_t *mp_hw;
8201 boolean_t success;
8202 boolean_t ioctl_aborted = B_FALSE;
8203 boolean_t log = B_TRUE;
8204
8205 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer",
8206 char *, dl_primstr(dloa->dl_primitive), ill_t *, ill);
8207
8208 ip1dbg(("ip_rput_dlpi_writer .."));
8209 ASSERT(ipsq->ipsq_xop == ill->ill_phyint->phyint_ipsq->ipsq_xop);
8210 ASSERT(IAM_WRITER_ILL(ill));
8211
8212 ipif = ipsq->ipsq_xop->ipx_pending_ipif;
8213 /*
8214 * The current ioctl could have been aborted by the user and a new
8215 * ioctl to bring up another ill could have started. We could still
8216 * get a response from the driver later.
8217 */
8218 if (ipif != NULL && ipif->ipif_ill != ill)
8219 ioctl_aborted = B_TRUE;
8220
8221 switch (dloa->dl_primitive) {
8222 case DL_ERROR_ACK:
8223 ip1dbg(("ip_rput_dlpi_writer: got DL_ERROR_ACK for %s\n",
8224 dl_primstr(dlea->dl_error_primitive)));
8225
8226 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer error",
8227 char *, dl_primstr(dlea->dl_error_primitive),
8228 ill_t *, ill);
8229
8230 switch (dlea->dl_error_primitive) {
8231 case DL_DISABMULTI_REQ:
8232 ill_dlpi_done(ill, dlea->dl_error_primitive);
8233 break;
8234 case DL_PROMISCON_REQ:
8235 case DL_PROMISCOFF_REQ:
8236 case DL_UNBIND_REQ:
8237 case DL_ATTACH_REQ:
8238 case DL_INFO_REQ:
8239 ill_dlpi_done(ill, dlea->dl_error_primitive);
8240 break;
8241 case DL_NOTIFY_REQ:
8242 ill_dlpi_done(ill, DL_NOTIFY_REQ);
8243 log = B_FALSE;
8244 break;
8245 case DL_PHYS_ADDR_REQ:
8246 /*
8247 * For IPv6 only, there are two additional
8248 * phys_addr_req's sent to the driver to get the
8249 * IPv6 token and lla. This allows IP to acquire
8250 * the hardware address format for a given interface
8251 * without having built in knowledge of the hardware
8252 * address. ill_phys_addr_pend keeps track of the last
8253 * DL_PAR sent so we know which response we are
8254 * dealing with. ill_dlpi_done will update
8255 * ill_phys_addr_pend when it sends the next req.
8256 * We don't complete the IOCTL until all three DL_PARs
8257 * have been attempted, so set *_len to 0 and break.
8258 */
8259 paddrreq = ill->ill_phys_addr_pend;
8260 ill_dlpi_done(ill, DL_PHYS_ADDR_REQ);
8261 if (paddrreq == DL_IPV6_TOKEN) {
8262 ill->ill_token_length = 0;
8263 log = B_FALSE;
8264 break;
8265 } else if (paddrreq == DL_IPV6_LINK_LAYER_ADDR) {
8266 ill->ill_nd_lla_len = 0;
8267 log = B_FALSE;
8268 break;
8269 }
8270 /*
8271 * Something went wrong with the DL_PHYS_ADDR_REQ.
8272 * We presumably have an IOCTL hanging out waiting
8273 * for completion. Find it and complete the IOCTL
8274 * with the error noted.
8275 * However, ill_dl_phys was called on an ill queue
8276 * (from SIOCSLIFNAME), thus conn_pending_ill is not
8277 * set. But the ioctl is known to be pending on ill_wq.
8278 */
8279 if (!ill->ill_ifname_pending)
8280 break;
8281 ill->ill_ifname_pending = 0;
8282 if (!ioctl_aborted)
8283 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8284 if (mp1 != NULL) {
8285 /*
8286 * This operation (SIOCSLIFNAME) must have
8287 * happened on the ill. Assert there is no conn
8288 */
8289 ASSERT(connp == NULL);
8290 q = ill->ill_wq;
8291 }
8292 break;
8293 case DL_BIND_REQ:
8294 ill_dlpi_done(ill, DL_BIND_REQ);
8295 if (ill->ill_ifname_pending)
8296 break;
8297 mutex_enter(&ill->ill_lock);
8298 ill->ill_state_flags &= ~ILL_DOWN_IN_PROGRESS;
8299 mutex_exit(&ill->ill_lock);
8300 /*
8301 * Something went wrong with the bind. We presumably
8302 * have an IOCTL hanging out waiting for completion.
8303 * Find it, take down the interface that was coming
8304 * up, and complete the IOCTL with the error noted.
8305 */
8306 if (!ioctl_aborted)
8307 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8308 if (mp1 != NULL) {
8309 /*
8310 * This might be a result of a DL_NOTE_REPLUMB
8311 * notification. In that case, connp is NULL.
8312 */
8313 if (connp != NULL)
8314 q = CONNP_TO_WQ(connp);
8315
8316 (void) ipif_down(ipif, NULL, NULL);
8317 /* error is set below the switch */
8318 }
8319 break;
8320 case DL_ENABMULTI_REQ:
8321 ill_dlpi_done(ill, DL_ENABMULTI_REQ);
8322
8323 if (ill->ill_dlpi_multicast_state == IDS_INPROGRESS)
8324 ill->ill_dlpi_multicast_state = IDS_FAILED;
8325 if (ill->ill_dlpi_multicast_state == IDS_FAILED) {
8326
8327 printf("ip: joining multicasts failed (%d)"
8328 " on %s - will use link layer "
8329 "broadcasts for multicast\n",
8330 dlea->dl_errno, ill->ill_name);
8331
8332 /*
8333 * Set up for multi_bcast; We are the
8334 * writer, so ok to access ill->ill_ipif
8335 * without any lock.
8336 */
8337 mutex_enter(&ill->ill_phyint->phyint_lock);
8338 ill->ill_phyint->phyint_flags |=
8339 PHYI_MULTI_BCAST;
8340 mutex_exit(&ill->ill_phyint->phyint_lock);
8341
8342 }
8343 freemsg(mp); /* Don't want to pass this up */
8344 return;
8345 case DL_CAPABILITY_REQ:
8346 ip1dbg(("ip_rput_dlpi_writer: got DL_ERROR_ACK for "
8347 "DL_CAPABILITY REQ\n"));
8348 if (ill->ill_dlpi_capab_state == IDCS_PROBE_SENT)
8349 ill->ill_dlpi_capab_state = IDCS_FAILED;
8350 ill_capability_done(ill);
8351 freemsg(mp);
8352 return;
8353 }
8354 /*
8355 * Note the error for IOCTL completion (mp1 is set when
8356 * ready to complete ioctl). If ill_ifname_pending_err is
8357 * set, an error occured during plumbing (ill_ifname_pending),
8358 * so we want to report that error.
8359 *
8360 * NOTE: there are two addtional DL_PHYS_ADDR_REQ's
8361 * (DL_IPV6_TOKEN and DL_IPV6_LINK_LAYER_ADDR) that are
8362 * expected to get errack'd if the driver doesn't support
8363 * these flags (e.g. ethernet). log will be set to B_FALSE
8364 * if these error conditions are encountered.
8365 */
8366 if (mp1 != NULL) {
8367 if (ill->ill_ifname_pending_err != 0) {
8368 err = ill->ill_ifname_pending_err;
8369 ill->ill_ifname_pending_err = 0;
8370 } else {
8371 err = dlea->dl_unix_errno ?
8372 dlea->dl_unix_errno : ENXIO;
8373 }
8374 /*
8375 * If we're plumbing an interface and an error hasn't already
8376 * been saved, set ill_ifname_pending_err to the error passed
8377 * up. Ignore the error if log is B_FALSE (see comment above).
8378 */
8379 } else if (log && ill->ill_ifname_pending &&
8380 ill->ill_ifname_pending_err == 0) {
8381 ill->ill_ifname_pending_err = dlea->dl_unix_errno ?
8382 dlea->dl_unix_errno : ENXIO;
8383 }
8384
8385 if (log)
8386 ip_dlpi_error(ill, dlea->dl_error_primitive,
8387 dlea->dl_errno, dlea->dl_unix_errno);
8388 break;
8389 case DL_CAPABILITY_ACK:
8390 ill_capability_ack(ill, mp);
8391 /*
8392 * The message has been handed off to ill_capability_ack
8393 * and must not be freed below
8394 */
8395 mp = NULL;
8396 break;
8397
8398 case DL_INFO_ACK:
8399 /* Call a routine to handle this one. */
8400 ill_dlpi_done(ill, DL_INFO_REQ);
8401 ip_ll_subnet_defaults(ill, mp);
8402 ASSERT(!MUTEX_HELD(&ill->ill_phyint->phyint_ipsq->ipsq_lock));
8403 return;
8404 case DL_BIND_ACK:
8405 /*
8406 * We should have an IOCTL waiting on this unless
8407 * sent by ill_dl_phys, in which case just return
8408 */
8409 ill_dlpi_done(ill, DL_BIND_REQ);
8410
8411 if (ill->ill_ifname_pending) {
8412 DTRACE_PROBE2(ip__rput__dlpi__ifname__pending,
8413 ill_t *, ill, mblk_t *, mp);
8414 break;
8415 }
8416 mutex_enter(&ill->ill_lock);
8417 ill->ill_dl_up = 1;
8418 ill->ill_state_flags &= ~ILL_DOWN_IN_PROGRESS;
8419 mutex_exit(&ill->ill_lock);
8420
8421 if (!ioctl_aborted)
8422 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8423 if (mp1 == NULL) {
8424 DTRACE_PROBE1(ip__rput__dlpi__no__mblk, ill_t *, ill);
8425 break;
8426 }
8427 /*
8428 * mp1 was added by ill_dl_up(). if that is a result of
8429 * a DL_NOTE_REPLUMB notification, connp could be NULL.
8430 */
8431 if (connp != NULL)
8432 q = CONNP_TO_WQ(connp);
8433 /*
8434 * We are exclusive. So nothing can change even after
8435 * we get the pending mp.
8436 */
8437 ip1dbg(("ip_rput_dlpi: bind_ack %s\n", ill->ill_name));
8438 DTRACE_PROBE1(ip__rput__dlpi__bind__ack, ill_t *, ill);
8439 ill_nic_event_dispatch(ill, 0, NE_UP, NULL, 0);
8440
8441 /*
8442 * Now bring up the resolver; when that is complete, we'll
8443 * create IREs. Note that we intentionally mirror what
8444 * ipif_up() would have done, because we got here by way of
8445 * ill_dl_up(), which stopped ipif_up()'s processing.
8446 */
8447 if (ill->ill_isv6) {
8448 /*
8449 * v6 interfaces.
8450 * Unlike ARP which has to do another bind
8451 * and attach, once we get here we are
8452 * done with NDP
8453 */
8454 (void) ipif_resolver_up(ipif, Res_act_initial);
8455 if ((err = ipif_ndp_up(ipif, B_TRUE)) == 0)
8456 err = ipif_up_done_v6(ipif);
8457 } else if (ill->ill_net_type == IRE_IF_RESOLVER) {
8458 /*
8459 * ARP and other v4 external resolvers.
8460 * Leave the pending mblk intact so that
8461 * the ioctl completes in ip_rput().
8462 */
8463 if (connp != NULL)
8464 mutex_enter(&connp->conn_lock);
8465 mutex_enter(&ill->ill_lock);
8466 success = ipsq_pending_mp_add(connp, ipif, q, mp1, 0);
8467 mutex_exit(&ill->ill_lock);
8468 if (connp != NULL)
8469 mutex_exit(&connp->conn_lock);
8470 if (success) {
8471 err = ipif_resolver_up(ipif, Res_act_initial);
8472 if (err == EINPROGRESS) {
8473 freemsg(mp);
8474 return;
8475 }
8476 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8477 } else {
8478 /* The conn has started closing */
8479 err = EINTR;
8480 }
8481 } else {
8482 /*
8483 * This one is complete. Reply to pending ioctl.
8484 */
8485 (void) ipif_resolver_up(ipif, Res_act_initial);
8486 err = ipif_up_done(ipif);
8487 }
8488
8489 if ((err == 0) && (ill->ill_up_ipifs)) {
8490 err = ill_up_ipifs(ill, q, mp1);
8491 if (err == EINPROGRESS) {
8492 freemsg(mp);
8493 return;
8494 }
8495 }
8496
8497 /*
8498 * If we have a moved ipif to bring up, and everything has
8499 * succeeded to this point, bring it up on the IPMP ill.
8500 * Otherwise, leave it down -- the admin can try to bring it
8501 * up by hand if need be.
8502 */
8503 if (ill->ill_move_ipif != NULL) {
8504 if (err != 0) {
8505 ill->ill_move_ipif = NULL;
8506 } else {
8507 ipif = ill->ill_move_ipif;
8508 ill->ill_move_ipif = NULL;
8509 err = ipif_up(ipif, q, mp1);
8510 if (err == EINPROGRESS) {
8511 freemsg(mp);
8512 return;
8513 }
8514 }
8515 }
8516 break;
8517
8518 case DL_NOTIFY_IND: {
8519 dl_notify_ind_t *notify = (dl_notify_ind_t *)mp->b_rptr;
8520 uint_t orig_mtu, orig_mc_mtu;
8521
8522 switch (notify->dl_notification) {
8523 case DL_NOTE_PHYS_ADDR:
8524 err = ill_set_phys_addr(ill, mp);
8525 break;
8526
8527 case DL_NOTE_REPLUMB:
8528 /*
8529 * Directly return after calling ill_replumb().
8530 * Note that we should not free mp as it is reused
8531 * in the ill_replumb() function.
8532 */
8533 err = ill_replumb(ill, mp);
8534 return;
8535
8536 case DL_NOTE_FASTPATH_FLUSH:
8537 nce_flush(ill, B_FALSE);
8538 break;
8539
8540 case DL_NOTE_SDU_SIZE:
8541 case DL_NOTE_SDU_SIZE2:
8542 /*
8543 * The dce and fragmentation code can cope with
8544 * this changing while packets are being sent.
8545 * When packets are sent ip_output will discover
8546 * a change.
8547 *
8548 * Change the MTU size of the interface.
8549 */
8550 mutex_enter(&ill->ill_lock);
8551 orig_mtu = ill->ill_mtu;
8552 orig_mc_mtu = ill->ill_mc_mtu;
8553 switch (notify->dl_notification) {
8554 case DL_NOTE_SDU_SIZE:
8555 ill->ill_current_frag =
8556 (uint_t)notify->dl_data;
8557 ill->ill_mc_mtu = (uint_t)notify->dl_data;
8558 break;
8559 case DL_NOTE_SDU_SIZE2:
8560 ill->ill_current_frag =
8561 (uint_t)notify->dl_data1;
8562 ill->ill_mc_mtu = (uint_t)notify->dl_data2;
8563 break;
8564 }
8565 if (ill->ill_current_frag > ill->ill_max_frag)
8566 ill->ill_max_frag = ill->ill_current_frag;
8567
8568 if (!(ill->ill_flags & ILLF_FIXEDMTU)) {
8569 ill->ill_mtu = ill->ill_current_frag;
8570
8571 /*
8572 * If ill_user_mtu was set (via
8573 * SIOCSLIFLNKINFO), clamp ill_mtu at it.
8574 */
8575 if (ill->ill_user_mtu != 0 &&
8576 ill->ill_user_mtu < ill->ill_mtu)
8577 ill->ill_mtu = ill->ill_user_mtu;
8578
8579 if (ill->ill_user_mtu != 0 &&
8580 ill->ill_user_mtu < ill->ill_mc_mtu)
8581 ill->ill_mc_mtu = ill->ill_user_mtu;
8582
8583 if (ill->ill_isv6) {
8584 if (ill->ill_mtu < IPV6_MIN_MTU)
8585 ill->ill_mtu = IPV6_MIN_MTU;
8586 if (ill->ill_mc_mtu < IPV6_MIN_MTU)
8587 ill->ill_mc_mtu = IPV6_MIN_MTU;
8588 } else {
8589 if (ill->ill_mtu < IP_MIN_MTU)
8590 ill->ill_mtu = IP_MIN_MTU;
8591 if (ill->ill_mc_mtu < IP_MIN_MTU)
8592 ill->ill_mc_mtu = IP_MIN_MTU;
8593 }
8594 } else if (ill->ill_mc_mtu > ill->ill_mtu) {
8595 ill->ill_mc_mtu = ill->ill_mtu;
8596 }
8597
8598 mutex_exit(&ill->ill_lock);
8599 /*
8600 * Make sure all dce_generation checks find out
8601 * that ill_mtu/ill_mc_mtu has changed.
8602 */
8603 if (orig_mtu != ill->ill_mtu ||
8604 orig_mc_mtu != ill->ill_mc_mtu) {
8605 dce_increment_all_generations(ill->ill_isv6,
8606 ill->ill_ipst);
8607 }
8608
8609 /*
8610 * Refresh IPMP meta-interface MTU if necessary.
8611 */
8612 if (IS_UNDER_IPMP(ill))
8613 ipmp_illgrp_refresh_mtu(ill->ill_grp);
8614 break;
8615
8616 case DL_NOTE_LINK_UP:
8617 case DL_NOTE_LINK_DOWN: {
8618 /*
8619 * We are writer. ill / phyint / ipsq assocs stable.
8620 * The RUNNING flag reflects the state of the link.
8621 */
8622 phyint_t *phyint = ill->ill_phyint;
8623 uint64_t new_phyint_flags;
8624 boolean_t changed = B_FALSE;
8625 boolean_t went_up;
8626
8627 went_up = notify->dl_notification == DL_NOTE_LINK_UP;
8628 mutex_enter(&phyint->phyint_lock);
8629
8630 new_phyint_flags = went_up ?
8631 phyint->phyint_flags | PHYI_RUNNING :
8632 phyint->phyint_flags & ~PHYI_RUNNING;
8633
8634 if (IS_IPMP(ill)) {
8635 new_phyint_flags = went_up ?
8636 new_phyint_flags & ~PHYI_FAILED :
8637 new_phyint_flags | PHYI_FAILED;
8638 }
8639
8640 if (new_phyint_flags != phyint->phyint_flags) {
8641 phyint->phyint_flags = new_phyint_flags;
8642 changed = B_TRUE;
8643 }
8644 mutex_exit(&phyint->phyint_lock);
8645 /*
8646 * ill_restart_dad handles the DAD restart and routing
8647 * socket notification logic.
8648 */
8649 if (changed) {
8650 ill_restart_dad(phyint->phyint_illv4, went_up);
8651 ill_restart_dad(phyint->phyint_illv6, went_up);
8652 }
8653 break;
8654 }
8655 case DL_NOTE_PROMISC_ON_PHYS: {
8656 phyint_t *phyint = ill->ill_phyint;
8657
8658 mutex_enter(&phyint->phyint_lock);
8659 phyint->phyint_flags |= PHYI_PROMISC;
8660 mutex_exit(&phyint->phyint_lock);
8661 break;
8662 }
8663 case DL_NOTE_PROMISC_OFF_PHYS: {
8664 phyint_t *phyint = ill->ill_phyint;
8665
8666 mutex_enter(&phyint->phyint_lock);
8667 phyint->phyint_flags &= ~PHYI_PROMISC;
8668 mutex_exit(&phyint->phyint_lock);
8669 break;
8670 }
8671 case DL_NOTE_CAPAB_RENEG:
8672 /*
8673 * Something changed on the driver side.
8674 * It wants us to renegotiate the capabilities
8675 * on this ill. One possible cause is the aggregation
8676 * interface under us where a port got added or
8677 * went away.
8678 *
8679 * If the capability negotiation is already done
8680 * or is in progress, reset the capabilities and
8681 * mark the ill's ill_capab_reneg to be B_TRUE,
8682 * so that when the ack comes back, we can start
8683 * the renegotiation process.
8684 *
8685 * Note that if ill_capab_reneg is already B_TRUE
8686 * (ill_dlpi_capab_state is IDS_UNKNOWN in this case),
8687 * the capability resetting request has been sent
8688 * and the renegotiation has not been started yet;
8689 * nothing needs to be done in this case.
8690 */
8691 ipsq_current_start(ipsq, ill->ill_ipif, 0);
8692 ill_capability_reset(ill, B_TRUE);
8693 ipsq_current_finish(ipsq);
8694 break;
8695
8696 case DL_NOTE_ALLOWED_IPS:
8697 ill_set_allowed_ips(ill, mp);
8698 break;
8699 default:
8700 ip0dbg(("ip_rput_dlpi_writer: unknown notification "
8701 "type 0x%x for DL_NOTIFY_IND\n",
8702 notify->dl_notification));
8703 break;
8704 }
8705
8706 /*
8707 * As this is an asynchronous operation, we
8708 * should not call ill_dlpi_done
8709 */
8710 break;
8711 }
8712 case DL_NOTIFY_ACK: {
8713 dl_notify_ack_t *noteack = (dl_notify_ack_t *)mp->b_rptr;
8714
8715 if (noteack->dl_notifications & DL_NOTE_LINK_UP)
8716 ill->ill_note_link = 1;
8717 ill_dlpi_done(ill, DL_NOTIFY_REQ);
8718 break;
8719 }
8720 case DL_PHYS_ADDR_ACK: {
8721 /*
8722 * As part of plumbing the interface via SIOCSLIFNAME,
8723 * ill_dl_phys() will queue a series of DL_PHYS_ADDR_REQs,
8724 * whose answers we receive here. As each answer is received,
8725 * we call ill_dlpi_done() to dispatch the next request as
8726 * we're processing the current one. Once all answers have
8727 * been received, we use ipsq_pending_mp_get() to dequeue the
8728 * outstanding IOCTL and reply to it. (Because ill_dl_phys()
8729 * is invoked from an ill queue, conn_oper_pending_ill is not
8730 * available, but we know the ioctl is pending on ill_wq.)
8731 */
8732 uint_t paddrlen, paddroff;
8733 uint8_t *addr;
8734
8735 paddrreq = ill->ill_phys_addr_pend;
8736 paddrlen = ((dl_phys_addr_ack_t *)mp->b_rptr)->dl_addr_length;
8737 paddroff = ((dl_phys_addr_ack_t *)mp->b_rptr)->dl_addr_offset;
8738 addr = mp->b_rptr + paddroff;
8739
8740 ill_dlpi_done(ill, DL_PHYS_ADDR_REQ);
8741 if (paddrreq == DL_IPV6_TOKEN) {
8742 /*
8743 * bcopy to low-order bits of ill_token
8744 *
8745 * XXX Temporary hack - currently, all known tokens
8746 * are 64 bits, so I'll cheat for the moment.
8747 */
8748 bcopy(addr, &ill->ill_token.s6_addr32[2], paddrlen);
8749 ill->ill_token_length = paddrlen;
8750 break;
8751 } else if (paddrreq == DL_IPV6_LINK_LAYER_ADDR) {
8752 ASSERT(ill->ill_nd_lla_mp == NULL);
8753 ill_set_ndmp(ill, mp, paddroff, paddrlen);
8754 mp = NULL;
8755 break;
8756 } else if (paddrreq == DL_CURR_DEST_ADDR) {
8757 ASSERT(ill->ill_dest_addr_mp == NULL);
8758 ill->ill_dest_addr_mp = mp;
8759 ill->ill_dest_addr = addr;
8760 mp = NULL;
8761 if (ill->ill_isv6) {
8762 ill_setdesttoken(ill);
8763 ipif_setdestlinklocal(ill->ill_ipif);
8764 }
8765 break;
8766 }
8767
8768 ASSERT(paddrreq == DL_CURR_PHYS_ADDR);
8769 ASSERT(ill->ill_phys_addr_mp == NULL);
8770 if (!ill->ill_ifname_pending)
8771 break;
8772 ill->ill_ifname_pending = 0;
8773 if (!ioctl_aborted)
8774 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8775 if (mp1 != NULL) {
8776 ASSERT(connp == NULL);
8777 q = ill->ill_wq;
8778 }
8779 /*
8780 * If any error acks received during the plumbing sequence,
8781 * ill_ifname_pending_err will be set. Break out and send up
8782 * the error to the pending ioctl.
8783 */
8784 if (ill->ill_ifname_pending_err != 0) {
8785 err = ill->ill_ifname_pending_err;
8786 ill->ill_ifname_pending_err = 0;
8787 break;
8788 }
8789
8790 ill->ill_phys_addr_mp = mp;
8791 ill->ill_phys_addr = (paddrlen == 0 ? NULL : addr);
8792 mp = NULL;
8793
8794 /*
8795 * If paddrlen or ill_phys_addr_length is zero, the DLPI
8796 * provider doesn't support physical addresses. We check both
8797 * paddrlen and ill_phys_addr_length because sppp (PPP) does
8798 * not have physical addresses, but historically adversises a
8799 * physical address length of 0 in its DL_INFO_ACK, but 6 in
8800 * its DL_PHYS_ADDR_ACK.
8801 */
8802 if (paddrlen == 0 || ill->ill_phys_addr_length == 0) {
8803 ill->ill_phys_addr = NULL;
8804 } else if (paddrlen != ill->ill_phys_addr_length) {
8805 ip0dbg(("DL_PHYS_ADDR_ACK: got addrlen %d, expected %d",
8806 paddrlen, ill->ill_phys_addr_length));
8807 err = EINVAL;
8808 break;
8809 }
8810
8811 if (ill->ill_nd_lla_mp == NULL) {
8812 if ((mp_hw = copyb(ill->ill_phys_addr_mp)) == NULL) {
8813 err = ENOMEM;
8814 break;
8815 }
8816 ill_set_ndmp(ill, mp_hw, paddroff, paddrlen);
8817 }
8818
8819 if (ill->ill_isv6) {
8820 ill_setdefaulttoken(ill);
8821 ipif_setlinklocal(ill->ill_ipif);
8822 }
8823 break;
8824 }
8825 case DL_OK_ACK:
8826 ip2dbg(("DL_OK_ACK %s (0x%x)\n",
8827 dl_primstr((int)dloa->dl_correct_primitive),
8828 dloa->dl_correct_primitive));
8829 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer ok",
8830 char *, dl_primstr(dloa->dl_correct_primitive),
8831 ill_t *, ill);
8832
8833 switch (dloa->dl_correct_primitive) {
8834 case DL_ENABMULTI_REQ:
8835 case DL_DISABMULTI_REQ:
8836 ill_dlpi_done(ill, dloa->dl_correct_primitive);
8837 break;
8838 case DL_PROMISCON_REQ:
8839 case DL_PROMISCOFF_REQ:
8840 case DL_UNBIND_REQ:
8841 case DL_ATTACH_REQ:
8842 ill_dlpi_done(ill, dloa->dl_correct_primitive);
8843 break;
8844 }
8845 break;
8846 default:
8847 break;
8848 }
8849
8850 freemsg(mp);
8851 if (mp1 == NULL)
8852 return;
8853
8854 /*
8855 * The operation must complete without EINPROGRESS since
8856 * ipsq_pending_mp_get() has removed the mblk (mp1). Otherwise,
8857 * the operation will be stuck forever inside the IPSQ.
8858 */
8859 ASSERT(err != EINPROGRESS);
8860
8861 DTRACE_PROBE4(ipif__ioctl, char *, "ip_rput_dlpi_writer finish",
8862 int, ipsq->ipsq_xop->ipx_current_ioctl, ill_t *, ill,
8863 ipif_t *, NULL);
8864
8865 switch (ipsq->ipsq_xop->ipx_current_ioctl) {
8866 case 0:
8867 ipsq_current_finish(ipsq);
8868 break;
8869
8870 case SIOCSLIFNAME:
8871 case IF_UNITSEL: {
8872 ill_t *ill_other = ILL_OTHER(ill);
8873
8874 /*
8875 * If SIOCSLIFNAME or IF_UNITSEL is about to succeed, and the
8876 * ill has a peer which is in an IPMP group, then place ill
8877 * into the same group. One catch: although ifconfig plumbs
8878 * the appropriate IPMP meta-interface prior to plumbing this
8879 * ill, it is possible for multiple ifconfig applications to
8880 * race (or for another application to adjust plumbing), in
8881 * which case the IPMP meta-interface we need will be missing.
8882 * If so, kick the phyint out of the group.
8883 */
8884 if (err == 0 && ill_other != NULL && IS_UNDER_IPMP(ill_other)) {
8885 ipmp_grp_t *grp = ill->ill_phyint->phyint_grp;
8886 ipmp_illgrp_t *illg;
8887
8888 illg = ill->ill_isv6 ? grp->gr_v6 : grp->gr_v4;
8889 if (illg == NULL)
8890 ipmp_phyint_leave_grp(ill->ill_phyint);
8891 else
8892 ipmp_ill_join_illgrp(ill, illg);
8893 }
8894
8895 if (ipsq->ipsq_xop->ipx_current_ioctl == IF_UNITSEL)
8896 ip_ioctl_finish(q, mp1, err, NO_COPYOUT, ipsq);
8897 else
8898 ip_ioctl_finish(q, mp1, err, COPYOUT, ipsq);
8899 break;
8900 }
8901 case SIOCLIFADDIF:
8902 ip_ioctl_finish(q, mp1, err, COPYOUT, ipsq);
8903 break;
8904
8905 default:
8906 ip_ioctl_finish(q, mp1, err, NO_COPYOUT, ipsq);
8907 break;
8908 }
8909 }
8910
8911 /*
8912 * ip_rput_other is called by ip_rput to handle messages modifying the global
8913 * state in IP. If 'ipsq' is non-NULL, caller is writer on it.
8914 */
8915 /* ARGSUSED */
8916 void
8917 ip_rput_other(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
8918 {
8919 ill_t *ill = q->q_ptr;
8920 struct iocblk *iocp;
8921
8922 ip1dbg(("ip_rput_other "));
8923 if (ipsq != NULL) {
8924 ASSERT(IAM_WRITER_IPSQ(ipsq));
8925 ASSERT(ipsq->ipsq_xop ==
8926 ill->ill_phyint->phyint_ipsq->ipsq_xop);
8927 }
8928
8929 switch (mp->b_datap->db_type) {
8930 case M_ERROR:
8931 case M_HANGUP:
8932 /*
8933 * The device has a problem. We force the ILL down. It can
8934 * be brought up again manually using SIOCSIFFLAGS (via
8935 * ifconfig or equivalent).
8936 */
8937 ASSERT(ipsq != NULL);
8938 if (mp->b_rptr < mp->b_wptr)
8939 ill->ill_error = (int)(*mp->b_rptr & 0xFF);
8940 if (ill->ill_error == 0)
8941 ill->ill_error = ENXIO;
8942 if (!ill_down_start(q, mp))
8943 return;
8944 ipif_all_down_tail(ipsq, q, mp, NULL);
8945 break;
8946 case M_IOCNAK: {
8947 iocp = (struct iocblk *)mp->b_rptr;
8948
8949 ASSERT(iocp->ioc_cmd == DL_IOC_HDR_INFO);
8950 /*
8951 * If this was the first attempt, turn off the fastpath
8952 * probing.
8953 */
8954 mutex_enter(&ill->ill_lock);
8955 if (ill->ill_dlpi_fastpath_state == IDS_INPROGRESS) {
8956 ill->ill_dlpi_fastpath_state = IDS_FAILED;
8957 mutex_exit(&ill->ill_lock);
8958 /*
8959 * don't flush the nce_t entries: we use them
8960 * as an index to the ncec itself.
8961 */
8962 ip1dbg(("ip_rput: DLPI fastpath off on interface %s\n",
8963 ill->ill_name));
8964 } else {
8965 mutex_exit(&ill->ill_lock);
8966 }
8967 freemsg(mp);
8968 break;
8969 }
8970 default:
8971 ASSERT(0);
8972 break;
8973 }
8974 }
8975
8976 /*
8977 * Update any source route, record route or timestamp options
8978 * When it fails it has consumed the message and BUMPed the MIB.
8979 */
8980 boolean_t
8981 ip_forward_options(mblk_t *mp, ipha_t *ipha, ill_t *dst_ill,
8982 ip_recv_attr_t *ira)
8983 {
8984 ipoptp_t opts;
8985 uchar_t *opt;
8986 uint8_t optval;
8987 uint8_t optlen;
8988 ipaddr_t dst;
8989 ipaddr_t ifaddr;
8990 uint32_t ts;
8991 timestruc_t now;
8992 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
8993
8994 ip2dbg(("ip_forward_options\n"));
8995 dst = ipha->ipha_dst;
8996 for (optval = ipoptp_first(&opts, ipha);
8997 optval != IPOPT_EOL;
8998 optval = ipoptp_next(&opts)) {
8999 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
9000 opt = opts.ipoptp_cur;
9001 optlen = opts.ipoptp_len;
9002 ip2dbg(("ip_forward_options: opt %d, len %d\n",
9003 optval, opts.ipoptp_len));
9004 switch (optval) {
9005 uint32_t off;
9006 case IPOPT_SSRR:
9007 case IPOPT_LSRR:
9008 /* Check if adminstratively disabled */
9009 if (!ipst->ips_ip_forward_src_routed) {
9010 BUMP_MIB(dst_ill->ill_ip_mib,
9011 ipIfStatsForwProhibits);
9012 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED",
9013 mp, dst_ill);
9014 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED,
9015 ira);
9016 return (B_FALSE);
9017 }
9018 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9019 /*
9020 * Must be partial since ip_input_options
9021 * checked for strict.
9022 */
9023 break;
9024 }
9025 off = opt[IPOPT_OFFSET];
9026 off--;
9027 redo_srr:
9028 if (optlen < IP_ADDR_LEN ||
9029 off > optlen - IP_ADDR_LEN) {
9030 /* End of source route */
9031 ip1dbg((
9032 "ip_forward_options: end of SR\n"));
9033 break;
9034 }
9035 /* Pick a reasonable address on the outbound if */
9036 ASSERT(dst_ill != NULL);
9037 if (ip_select_source_v4(dst_ill, INADDR_ANY, dst,
9038 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9039 NULL) != 0) {
9040 /* No source! Shouldn't happen */
9041 ifaddr = INADDR_ANY;
9042 }
9043 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9044 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9045 ip1dbg(("ip_forward_options: next hop 0x%x\n",
9046 ntohl(dst)));
9047
9048 /*
9049 * Check if our address is present more than
9050 * once as consecutive hops in source route.
9051 */
9052 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
9053 off += IP_ADDR_LEN;
9054 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9055 goto redo_srr;
9056 }
9057 ipha->ipha_dst = dst;
9058 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9059 break;
9060 case IPOPT_RR:
9061 off = opt[IPOPT_OFFSET];
9062 off--;
9063 if (optlen < IP_ADDR_LEN ||
9064 off > optlen - IP_ADDR_LEN) {
9065 /* No more room - ignore */
9066 ip1dbg((
9067 "ip_forward_options: end of RR\n"));
9068 break;
9069 }
9070 /* Pick a reasonable address on the outbound if */
9071 ASSERT(dst_ill != NULL);
9072 if (ip_select_source_v4(dst_ill, INADDR_ANY, dst,
9073 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9074 NULL) != 0) {
9075 /* No source! Shouldn't happen */
9076 ifaddr = INADDR_ANY;
9077 }
9078 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9079 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9080 break;
9081 case IPOPT_TS:
9082 /* Insert timestamp if there is room */
9083 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9084 case IPOPT_TS_TSONLY:
9085 off = IPOPT_TS_TIMELEN;
9086 break;
9087 case IPOPT_TS_PRESPEC:
9088 case IPOPT_TS_PRESPEC_RFC791:
9089 /* Verify that the address matched */
9090 off = opt[IPOPT_OFFSET] - 1;
9091 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9092 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9093 /* Not for us */
9094 break;
9095 }
9096 /* FALLTHROUGH */
9097 case IPOPT_TS_TSANDADDR:
9098 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9099 break;
9100 default:
9101 /*
9102 * ip_*put_options should have already
9103 * dropped this packet.
9104 */
9105 cmn_err(CE_PANIC, "ip_forward_options: "
9106 "unknown IT - bug in ip_input_options?\n");
9107 return (B_TRUE); /* Keep "lint" happy */
9108 }
9109 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
9110 /* Increase overflow counter */
9111 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
9112 opt[IPOPT_POS_OV_FLG] =
9113 (uint8_t)((opt[IPOPT_POS_OV_FLG] & 0x0F) |
9114 (off << 4));
9115 break;
9116 }
9117 off = opt[IPOPT_OFFSET] - 1;
9118 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9119 case IPOPT_TS_PRESPEC:
9120 case IPOPT_TS_PRESPEC_RFC791:
9121 case IPOPT_TS_TSANDADDR:
9122 /* Pick a reasonable addr on the outbound if */
9123 ASSERT(dst_ill != NULL);
9124 if (ip_select_source_v4(dst_ill, INADDR_ANY,
9125 dst, INADDR_ANY, ALL_ZONES, ipst, &ifaddr,
9126 NULL, NULL) != 0) {
9127 /* No source! Shouldn't happen */
9128 ifaddr = INADDR_ANY;
9129 }
9130 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9131 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9132 /* FALLTHROUGH */
9133 case IPOPT_TS_TSONLY:
9134 off = opt[IPOPT_OFFSET] - 1;
9135 /* Compute # of milliseconds since midnight */
9136 gethrestime(&now);
9137 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
9138 NSEC2MSEC(now.tv_nsec);
9139 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
9140 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
9141 break;
9142 }
9143 break;
9144 }
9145 }
9146 return (B_TRUE);
9147 }
9148
9149 /*
9150 * Call ill_frag_timeout to do garbage collection. ill_frag_timeout
9151 * returns 'true' if there are still fragments left on the queue, in
9152 * which case we restart the timer.
9153 */
9154 void
9155 ill_frag_timer(void *arg)
9156 {
9157 ill_t *ill = (ill_t *)arg;
9158 boolean_t frag_pending;
9159 ip_stack_t *ipst = ill->ill_ipst;
9160 time_t timeout;
9161
9162 mutex_enter(&ill->ill_lock);
9163 ASSERT(!ill->ill_fragtimer_executing);
9164 if (ill->ill_state_flags & ILL_CONDEMNED) {
9165 ill->ill_frag_timer_id = 0;
9166 mutex_exit(&ill->ill_lock);
9167 return;
9168 }
9169 ill->ill_fragtimer_executing = 1;
9170 mutex_exit(&ill->ill_lock);
9171
9172 timeout = (ill->ill_isv6 ? ipst->ips_ipv6_reassembly_timeout :
9173 ipst->ips_ip_reassembly_timeout);
9174
9175 frag_pending = ill_frag_timeout(ill, timeout);
9176
9177 /*
9178 * Restart the timer, if we have fragments pending or if someone
9179 * wanted us to be scheduled again.
9180 */
9181 mutex_enter(&ill->ill_lock);
9182 ill->ill_fragtimer_executing = 0;
9183 ill->ill_frag_timer_id = 0;
9184 if (frag_pending || ill->ill_fragtimer_needrestart)
9185 ill_frag_timer_start(ill);
9186 mutex_exit(&ill->ill_lock);
9187 }
9188
9189 void
9190 ill_frag_timer_start(ill_t *ill)
9191 {
9192 ip_stack_t *ipst = ill->ill_ipst;
9193 clock_t timeo_ms;
9194
9195 ASSERT(MUTEX_HELD(&ill->ill_lock));
9196
9197 /* If the ill is closing or opening don't proceed */
9198 if (ill->ill_state_flags & ILL_CONDEMNED)
9199 return;
9200
9201 if (ill->ill_fragtimer_executing) {
9202 /*
9203 * ill_frag_timer is currently executing. Just record the
9204 * the fact that we want the timer to be restarted.
9205 * ill_frag_timer will post a timeout before it returns,
9206 * ensuring it will be called again.
9207 */
9208 ill->ill_fragtimer_needrestart = 1;
9209 return;
9210 }
9211
9212 if (ill->ill_frag_timer_id == 0) {
9213 timeo_ms = (ill->ill_isv6 ? ipst->ips_ipv6_reassembly_timeout :
9214 ipst->ips_ip_reassembly_timeout) * SECONDS;
9215
9216 /*
9217 * The timer is neither running nor is the timeout handler
9218 * executing. Post a timeout so that ill_frag_timer will be
9219 * called
9220 */
9221 ill->ill_frag_timer_id = timeout(ill_frag_timer, ill,
9222 MSEC_TO_TICK(timeo_ms >> 1));
9223 ill->ill_fragtimer_needrestart = 0;
9224 }
9225 }
9226
9227 /*
9228 * Update any source route, record route or timestamp options.
9229 * Check that we are at end of strict source route.
9230 * The options have already been checked for sanity in ip_input_options().
9231 */
9232 boolean_t
9233 ip_input_local_options(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
9234 {
9235 ipoptp_t opts;
9236 uchar_t *opt;
9237 uint8_t optval;
9238 uint8_t optlen;
9239 ipaddr_t dst;
9240 ipaddr_t ifaddr;
9241 uint32_t ts;
9242 timestruc_t now;
9243 ill_t *ill = ira->ira_ill;
9244 ip_stack_t *ipst = ill->ill_ipst;
9245
9246 ip2dbg(("ip_input_local_options\n"));
9247
9248 for (optval = ipoptp_first(&opts, ipha);
9249 optval != IPOPT_EOL;
9250 optval = ipoptp_next(&opts)) {
9251 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
9252 opt = opts.ipoptp_cur;
9253 optlen = opts.ipoptp_len;
9254 ip2dbg(("ip_input_local_options: opt %d, len %d\n",
9255 optval, optlen));
9256 switch (optval) {
9257 uint32_t off;
9258 case IPOPT_SSRR:
9259 case IPOPT_LSRR:
9260 off = opt[IPOPT_OFFSET];
9261 off--;
9262 if (optlen < IP_ADDR_LEN ||
9263 off > optlen - IP_ADDR_LEN) {
9264 /* End of source route */
9265 ip1dbg(("ip_input_local_options: end of SR\n"));
9266 break;
9267 }
9268 /*
9269 * This will only happen if two consecutive entries
9270 * in the source route contains our address or if
9271 * it is a packet with a loose source route which
9272 * reaches us before consuming the whole source route
9273 */
9274 ip1dbg(("ip_input_local_options: not end of SR\n"));
9275 if (optval == IPOPT_SSRR) {
9276 goto bad_src_route;
9277 }
9278 /*
9279 * Hack: instead of dropping the packet truncate the
9280 * source route to what has been used by filling the
9281 * rest with IPOPT_NOP.
9282 */
9283 opt[IPOPT_OLEN] = (uint8_t)off;
9284 while (off < optlen) {
9285 opt[off++] = IPOPT_NOP;
9286 }
9287 break;
9288 case IPOPT_RR:
9289 off = opt[IPOPT_OFFSET];
9290 off--;
9291 if (optlen < IP_ADDR_LEN ||
9292 off > optlen - IP_ADDR_LEN) {
9293 /* No more room - ignore */
9294 ip1dbg((
9295 "ip_input_local_options: end of RR\n"));
9296 break;
9297 }
9298 /* Pick a reasonable address on the outbound if */
9299 if (ip_select_source_v4(ill, INADDR_ANY, ipha->ipha_dst,
9300 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9301 NULL) != 0) {
9302 /* No source! Shouldn't happen */
9303 ifaddr = INADDR_ANY;
9304 }
9305 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9306 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9307 break;
9308 case IPOPT_TS:
9309 /* Insert timestamp if there is romm */
9310 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9311 case IPOPT_TS_TSONLY:
9312 off = IPOPT_TS_TIMELEN;
9313 break;
9314 case IPOPT_TS_PRESPEC:
9315 case IPOPT_TS_PRESPEC_RFC791:
9316 /* Verify that the address matched */
9317 off = opt[IPOPT_OFFSET] - 1;
9318 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9319 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9320 /* Not for us */
9321 break;
9322 }
9323 /* FALLTHROUGH */
9324 case IPOPT_TS_TSANDADDR:
9325 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9326 break;
9327 default:
9328 /*
9329 * ip_*put_options should have already
9330 * dropped this packet.
9331 */
9332 cmn_err(CE_PANIC, "ip_input_local_options: "
9333 "unknown IT - bug in ip_input_options?\n");
9334 return (B_TRUE); /* Keep "lint" happy */
9335 }
9336 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
9337 /* Increase overflow counter */
9338 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
9339 opt[IPOPT_POS_OV_FLG] =
9340 (uint8_t)((opt[IPOPT_POS_OV_FLG] & 0x0F) |
9341 (off << 4));
9342 break;
9343 }
9344 off = opt[IPOPT_OFFSET] - 1;
9345 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9346 case IPOPT_TS_PRESPEC:
9347 case IPOPT_TS_PRESPEC_RFC791:
9348 case IPOPT_TS_TSANDADDR:
9349 /* Pick a reasonable addr on the outbound if */
9350 if (ip_select_source_v4(ill, INADDR_ANY,
9351 ipha->ipha_dst, INADDR_ANY, ALL_ZONES, ipst,
9352 &ifaddr, NULL, NULL) != 0) {
9353 /* No source! Shouldn't happen */
9354 ifaddr = INADDR_ANY;
9355 }
9356 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9357 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9358 /* FALLTHROUGH */
9359 case IPOPT_TS_TSONLY:
9360 off = opt[IPOPT_OFFSET] - 1;
9361 /* Compute # of milliseconds since midnight */
9362 gethrestime(&now);
9363 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
9364 NSEC2MSEC(now.tv_nsec);
9365 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
9366 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
9367 break;
9368 }
9369 break;
9370 }
9371 }
9372 return (B_TRUE);
9373
9374 bad_src_route:
9375 /* make sure we clear any indication of a hardware checksum */
9376 DB_CKSUMFLAGS(mp) = 0;
9377 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ill);
9378 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
9379 return (B_FALSE);
9380
9381 }
9382
9383 /*
9384 * Process IP options in an inbound packet. Always returns the nexthop.
9385 * Normally this is the passed in nexthop, but if there is an option
9386 * that effects the nexthop (such as a source route) that will be returned.
9387 * Sets *errorp if there is an error, in which case an ICMP error has been sent
9388 * and mp freed.
9389 */
9390 ipaddr_t
9391 ip_input_options(ipha_t *ipha, ipaddr_t dst, mblk_t *mp,
9392 ip_recv_attr_t *ira, int *errorp)
9393 {
9394 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
9395 ipoptp_t opts;
9396 uchar_t *opt;
9397 uint8_t optval;
9398 uint8_t optlen;
9399 intptr_t code = 0;
9400 ire_t *ire;
9401
9402 ip2dbg(("ip_input_options\n"));
9403 *errorp = 0;
9404 for (optval = ipoptp_first(&opts, ipha);
9405 optval != IPOPT_EOL;
9406 optval = ipoptp_next(&opts)) {
9407 opt = opts.ipoptp_cur;
9408 optlen = opts.ipoptp_len;
9409 ip2dbg(("ip_input_options: opt %d, len %d\n",
9410 optval, optlen));
9411 /*
9412 * Note: we need to verify the checksum before we
9413 * modify anything thus this routine only extracts the next
9414 * hop dst from any source route.
9415 */
9416 switch (optval) {
9417 uint32_t off;
9418 case IPOPT_SSRR:
9419 case IPOPT_LSRR:
9420 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9421 if (optval == IPOPT_SSRR) {
9422 ip1dbg(("ip_input_options: not next"
9423 " strict source route 0x%x\n",
9424 ntohl(dst)));
9425 code = (char *)&ipha->ipha_dst -
9426 (char *)ipha;
9427 goto param_prob; /* RouterReq's */
9428 }
9429 ip2dbg(("ip_input_options: "
9430 "not next source route 0x%x\n",
9431 ntohl(dst)));
9432 break;
9433 }
9434
9435 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9436 ip1dbg((
9437 "ip_input_options: bad option offset\n"));
9438 code = (char *)&opt[IPOPT_OLEN] -
9439 (char *)ipha;
9440 goto param_prob;
9441 }
9442 off = opt[IPOPT_OFFSET];
9443 off--;
9444 redo_srr:
9445 if (optlen < IP_ADDR_LEN ||
9446 off > optlen - IP_ADDR_LEN) {
9447 /* End of source route */
9448 ip1dbg(("ip_input_options: end of SR\n"));
9449 break;
9450 }
9451 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9452 ip1dbg(("ip_input_options: next hop 0x%x\n",
9453 ntohl(dst)));
9454
9455 /*
9456 * Check if our address is present more than
9457 * once as consecutive hops in source route.
9458 * XXX verify per-interface ip_forwarding
9459 * for source route?
9460 */
9461 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
9462 off += IP_ADDR_LEN;
9463 goto redo_srr;
9464 }
9465
9466 if (dst == htonl(INADDR_LOOPBACK)) {
9467 ip1dbg(("ip_input_options: loopback addr in "
9468 "source route!\n"));
9469 goto bad_src_route;
9470 }
9471 /*
9472 * For strict: verify that dst is directly
9473 * reachable.
9474 */
9475 if (optval == IPOPT_SSRR) {
9476 ire = ire_ftable_lookup_v4(dst, 0, 0,
9477 IRE_INTERFACE, NULL, ALL_ZONES,
9478 ira->ira_tsl,
9479 MATCH_IRE_TYPE | MATCH_IRE_SECATTR, 0, ipst,
9480 NULL);
9481 if (ire == NULL) {
9482 ip1dbg(("ip_input_options: SSRR not "
9483 "directly reachable: 0x%x\n",
9484 ntohl(dst)));
9485 goto bad_src_route;
9486 }
9487 ire_refrele(ire);
9488 }
9489 /*
9490 * Defer update of the offset and the record route
9491 * until the packet is forwarded.
9492 */
9493 break;
9494 case IPOPT_RR:
9495 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9496 ip1dbg((
9497 "ip_input_options: bad option offset\n"));
9498 code = (char *)&opt[IPOPT_OLEN] -
9499 (char *)ipha;
9500 goto param_prob;
9501 }
9502 break;
9503 case IPOPT_TS:
9504 /*
9505 * Verify that length >= 5 and that there is either
9506 * room for another timestamp or that the overflow
9507 * counter is not maxed out.
9508 */
9509 code = (char *)&opt[IPOPT_OLEN] - (char *)ipha;
9510 if (optlen < IPOPT_MINLEN_IT) {
9511 goto param_prob;
9512 }
9513 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9514 ip1dbg((
9515 "ip_input_options: bad option offset\n"));
9516 code = (char *)&opt[IPOPT_OFFSET] -
9517 (char *)ipha;
9518 goto param_prob;
9519 }
9520 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9521 case IPOPT_TS_TSONLY:
9522 off = IPOPT_TS_TIMELEN;
9523 break;
9524 case IPOPT_TS_TSANDADDR:
9525 case IPOPT_TS_PRESPEC:
9526 case IPOPT_TS_PRESPEC_RFC791:
9527 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9528 break;
9529 default:
9530 code = (char *)&opt[IPOPT_POS_OV_FLG] -
9531 (char *)ipha;
9532 goto param_prob;
9533 }
9534 if (opt[IPOPT_OFFSET] - 1 + off > optlen &&
9535 (opt[IPOPT_POS_OV_FLG] & 0xF0) == 0xF0) {
9536 /*
9537 * No room and the overflow counter is 15
9538 * already.
9539 */
9540 goto param_prob;
9541 }
9542 break;
9543 }
9544 }
9545
9546 if ((opts.ipoptp_flags & IPOPTP_ERROR) == 0) {
9547 return (dst);
9548 }
9549
9550 ip1dbg(("ip_input_options: error processing IP options."));
9551 code = (char *)&opt[IPOPT_OFFSET] - (char *)ipha;
9552
9553 param_prob:
9554 /* make sure we clear any indication of a hardware checksum */
9555 DB_CKSUMFLAGS(mp) = 0;
9556 ip_drop_input("ICMP_PARAM_PROBLEM", mp, ira->ira_ill);
9557 icmp_param_problem(mp, (uint8_t)code, ira);
9558 *errorp = -1;
9559 return (dst);
9560
9561 bad_src_route:
9562 /* make sure we clear any indication of a hardware checksum */
9563 DB_CKSUMFLAGS(mp) = 0;
9564 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ira->ira_ill);
9565 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
9566 *errorp = -1;
9567 return (dst);
9568 }
9569
9570 /*
9571 * IP & ICMP info in >=14 msg's ...
9572 * - ip fixed part (mib2_ip_t)
9573 * - icmp fixed part (mib2_icmp_t)
9574 * - ipAddrEntryTable (ip 20) all IPv4 ipifs
9575 * - ipRouteEntryTable (ip 21) all IPv4 IREs
9576 * - ipNetToMediaEntryTable (ip 22) all IPv4 Neighbor Cache entries
9577 * - ipRouteAttributeTable (ip 102) labeled routes
9578 * - ip multicast membership (ip_member_t)
9579 * - ip multicast source filtering (ip_grpsrc_t)
9580 * - igmp fixed part (struct igmpstat)
9581 * - multicast routing stats (struct mrtstat)
9582 * - multicast routing vifs (array of struct vifctl)
9583 * - multicast routing routes (array of struct mfcctl)
9584 * - ip6 fixed part (mib2_ipv6IfStatsEntry_t)
9585 * One per ill plus one generic
9586 * - icmp6 fixed part (mib2_ipv6IfIcmpEntry_t)
9587 * One per ill plus one generic
9588 * - ipv6RouteEntry all IPv6 IREs
9589 * - ipv6RouteAttributeTable (ip6 102) labeled routes
9590 * - ipv6NetToMediaEntry all IPv6 Neighbor Cache entries
9591 * - ipv6AddrEntry all IPv6 ipifs
9592 * - ipv6 multicast membership (ipv6_member_t)
9593 * - ipv6 multicast source filtering (ipv6_grpsrc_t)
9594 *
9595 * NOTE: original mpctl is copied for msg's 2..N, since its ctl part is
9596 * already filled in by the caller.
9597 * If legacy_req is true then MIB structures needs to be truncated to their
9598 * legacy sizes before being returned.
9599 * Return value of 0 indicates that no messages were sent and caller
9600 * should free mpctl.
9601 */
9602 int
9603 ip_snmp_get(queue_t *q, mblk_t *mpctl, int level, boolean_t legacy_req)
9604 {
9605 ip_stack_t *ipst;
9606 sctp_stack_t *sctps;
9607
9608 if (q->q_next != NULL) {
9609 ipst = ILLQ_TO_IPST(q);
9610 } else {
9611 ipst = CONNQ_TO_IPST(q);
9612 }
9613 ASSERT(ipst != NULL);
9614 sctps = ipst->ips_netstack->netstack_sctp;
9615
9616 if (mpctl == NULL || mpctl->b_cont == NULL) {
9617 return (0);
9618 }
9619
9620 /*
9621 * For the purposes of the (broken) packet shell use
9622 * of the level we make sure MIB2_TCP/MIB2_UDP can be used
9623 * to make TCP and UDP appear first in the list of mib items.
9624 * TBD: We could expand this and use it in netstat so that
9625 * the kernel doesn't have to produce large tables (connections,
9626 * routes, etc) when netstat only wants the statistics or a particular
9627 * table.
9628 */
9629 if (!(level == MIB2_TCP || level == MIB2_UDP)) {
9630 if ((mpctl = icmp_snmp_get(q, mpctl)) == NULL) {
9631 return (1);
9632 }
9633 }
9634
9635 if (level != MIB2_TCP) {
9636 if ((mpctl = udp_snmp_get(q, mpctl, legacy_req)) == NULL) {
9637 return (1);
9638 }
9639 if (level == MIB2_UDP) {
9640 goto done;
9641 }
9642 }
9643
9644 if (level != MIB2_UDP) {
9645 if ((mpctl = tcp_snmp_get(q, mpctl, legacy_req)) == NULL) {
9646 return (1);
9647 }
9648 if (level == MIB2_TCP) {
9649 goto done;
9650 }
9651 }
9652
9653 if ((mpctl = ip_snmp_get_mib2_ip_traffic_stats(q, mpctl,
9654 ipst, legacy_req)) == NULL) {
9655 return (1);
9656 }
9657
9658 if ((mpctl = ip_snmp_get_mib2_ip6(q, mpctl, ipst,
9659 legacy_req)) == NULL) {
9660 return (1);
9661 }
9662
9663 if ((mpctl = ip_snmp_get_mib2_icmp(q, mpctl, ipst)) == NULL) {
9664 return (1);
9665 }
9666
9667 if ((mpctl = ip_snmp_get_mib2_icmp6(q, mpctl, ipst)) == NULL) {
9668 return (1);
9669 }
9670
9671 if ((mpctl = ip_snmp_get_mib2_igmp(q, mpctl, ipst)) == NULL) {
9672 return (1);
9673 }
9674
9675 if ((mpctl = ip_snmp_get_mib2_multi(q, mpctl, ipst)) == NULL) {
9676 return (1);
9677 }
9678
9679 if ((mpctl = ip_snmp_get_mib2_ip_addr(q, mpctl, ipst,
9680 legacy_req)) == NULL) {
9681 return (1);
9682 }
9683
9684 if ((mpctl = ip_snmp_get_mib2_ip6_addr(q, mpctl, ipst,
9685 legacy_req)) == NULL) {
9686 return (1);
9687 }
9688
9689 if ((mpctl = ip_snmp_get_mib2_ip_group_mem(q, mpctl, ipst)) == NULL) {
9690 return (1);
9691 }
9692
9693 if ((mpctl = ip_snmp_get_mib2_ip6_group_mem(q, mpctl, ipst)) == NULL) {
9694 return (1);
9695 }
9696
9697 if ((mpctl = ip_snmp_get_mib2_ip_group_src(q, mpctl, ipst)) == NULL) {
9698 return (1);
9699 }
9700
9701 if ((mpctl = ip_snmp_get_mib2_ip6_group_src(q, mpctl, ipst)) == NULL) {
9702 return (1);
9703 }
9704
9705 if ((mpctl = ip_snmp_get_mib2_virt_multi(q, mpctl, ipst)) == NULL) {
9706 return (1);
9707 }
9708
9709 if ((mpctl = ip_snmp_get_mib2_multi_rtable(q, mpctl, ipst)) == NULL) {
9710 return (1);
9711 }
9712
9713 mpctl = ip_snmp_get_mib2_ip_route_media(q, mpctl, level, ipst);
9714 if (mpctl == NULL)
9715 return (1);
9716
9717 mpctl = ip_snmp_get_mib2_ip6_route_media(q, mpctl, level, ipst);
9718 if (mpctl == NULL)
9719 return (1);
9720
9721 if ((mpctl = sctp_snmp_get_mib2(q, mpctl, sctps)) == NULL) {
9722 return (1);
9723 }
9724 if ((mpctl = ip_snmp_get_mib2_ip_dce(q, mpctl, ipst)) == NULL) {
9725 return (1);
9726 }
9727 done:
9728 freemsg(mpctl);
9729 return (1);
9730 }
9731
9732 /* Get global (legacy) IPv4 statistics */
9733 static mblk_t *
9734 ip_snmp_get_mib2_ip(queue_t *q, mblk_t *mpctl, mib2_ipIfStatsEntry_t *ipmib,
9735 ip_stack_t *ipst, boolean_t legacy_req)
9736 {
9737 mib2_ip_t old_ip_mib;
9738 struct opthdr *optp;
9739 mblk_t *mp2ctl;
9740 mib2_ipAddrEntry_t mae;
9741
9742 /*
9743 * make a copy of the original message
9744 */
9745 mp2ctl = copymsg(mpctl);
9746
9747 /* fixed length IP structure... */
9748 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9749 optp->level = MIB2_IP;
9750 optp->name = 0;
9751 SET_MIB(old_ip_mib.ipForwarding,
9752 (WE_ARE_FORWARDING(ipst) ? 1 : 2));
9753 SET_MIB(old_ip_mib.ipDefaultTTL,
9754 (uint32_t)ipst->ips_ip_def_ttl);
9755 SET_MIB(old_ip_mib.ipReasmTimeout,
9756 ipst->ips_ip_reassembly_timeout);
9757 SET_MIB(old_ip_mib.ipAddrEntrySize,
9758 (legacy_req) ? LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t) :
9759 sizeof (mib2_ipAddrEntry_t));
9760 SET_MIB(old_ip_mib.ipRouteEntrySize,
9761 sizeof (mib2_ipRouteEntry_t));
9762 SET_MIB(old_ip_mib.ipNetToMediaEntrySize,
9763 sizeof (mib2_ipNetToMediaEntry_t));
9764 SET_MIB(old_ip_mib.ipMemberEntrySize, sizeof (ip_member_t));
9765 SET_MIB(old_ip_mib.ipGroupSourceEntrySize, sizeof (ip_grpsrc_t));
9766 SET_MIB(old_ip_mib.ipRouteAttributeSize,
9767 sizeof (mib2_ipAttributeEntry_t));
9768 SET_MIB(old_ip_mib.transportMLPSize, sizeof (mib2_transportMLPEntry_t));
9769 SET_MIB(old_ip_mib.ipDestEntrySize, sizeof (dest_cache_entry_t));
9770
9771 /*
9772 * Grab the statistics from the new IP MIB
9773 */
9774 SET_MIB(old_ip_mib.ipInReceives,
9775 (uint32_t)ipmib->ipIfStatsHCInReceives);
9776 SET_MIB(old_ip_mib.ipInHdrErrors, ipmib->ipIfStatsInHdrErrors);
9777 SET_MIB(old_ip_mib.ipInAddrErrors, ipmib->ipIfStatsInAddrErrors);
9778 SET_MIB(old_ip_mib.ipForwDatagrams,
9779 (uint32_t)ipmib->ipIfStatsHCOutForwDatagrams);
9780 SET_MIB(old_ip_mib.ipInUnknownProtos,
9781 ipmib->ipIfStatsInUnknownProtos);
9782 SET_MIB(old_ip_mib.ipInDiscards, ipmib->ipIfStatsInDiscards);
9783 SET_MIB(old_ip_mib.ipInDelivers,
9784 (uint32_t)ipmib->ipIfStatsHCInDelivers);
9785 SET_MIB(old_ip_mib.ipOutRequests,
9786 (uint32_t)ipmib->ipIfStatsHCOutRequests);
9787 SET_MIB(old_ip_mib.ipOutDiscards, ipmib->ipIfStatsOutDiscards);
9788 SET_MIB(old_ip_mib.ipOutNoRoutes, ipmib->ipIfStatsOutNoRoutes);
9789 SET_MIB(old_ip_mib.ipReasmReqds, ipmib->ipIfStatsReasmReqds);
9790 SET_MIB(old_ip_mib.ipReasmOKs, ipmib->ipIfStatsReasmOKs);
9791 SET_MIB(old_ip_mib.ipReasmFails, ipmib->ipIfStatsReasmFails);
9792 SET_MIB(old_ip_mib.ipFragOKs, ipmib->ipIfStatsOutFragOKs);
9793 SET_MIB(old_ip_mib.ipFragFails, ipmib->ipIfStatsOutFragFails);
9794 SET_MIB(old_ip_mib.ipFragCreates, ipmib->ipIfStatsOutFragCreates);
9795
9796 /* ipRoutingDiscards is not being used */
9797 SET_MIB(old_ip_mib.ipRoutingDiscards, 0);
9798 SET_MIB(old_ip_mib.tcpInErrs, ipmib->tcpIfStatsInErrs);
9799 SET_MIB(old_ip_mib.udpNoPorts, ipmib->udpIfStatsNoPorts);
9800 SET_MIB(old_ip_mib.ipInCksumErrs, ipmib->ipIfStatsInCksumErrs);
9801 SET_MIB(old_ip_mib.ipReasmDuplicates,
9802 ipmib->ipIfStatsReasmDuplicates);
9803 SET_MIB(old_ip_mib.ipReasmPartDups, ipmib->ipIfStatsReasmPartDups);
9804 SET_MIB(old_ip_mib.ipForwProhibits, ipmib->ipIfStatsForwProhibits);
9805 SET_MIB(old_ip_mib.udpInCksumErrs, ipmib->udpIfStatsInCksumErrs);
9806 SET_MIB(old_ip_mib.udpInOverflows, ipmib->udpIfStatsInOverflows);
9807 SET_MIB(old_ip_mib.rawipInOverflows,
9808 ipmib->rawipIfStatsInOverflows);
9809
9810 SET_MIB(old_ip_mib.ipsecInSucceeded, ipmib->ipsecIfStatsInSucceeded);
9811 SET_MIB(old_ip_mib.ipsecInFailed, ipmib->ipsecIfStatsInFailed);
9812 SET_MIB(old_ip_mib.ipInIPv6, ipmib->ipIfStatsInWrongIPVersion);
9813 SET_MIB(old_ip_mib.ipOutIPv6, ipmib->ipIfStatsOutWrongIPVersion);
9814 SET_MIB(old_ip_mib.ipOutSwitchIPv6,
9815 ipmib->ipIfStatsOutSwitchIPVersion);
9816
9817 if (!snmp_append_data(mpctl->b_cont, (char *)&old_ip_mib,
9818 (int)sizeof (old_ip_mib))) {
9819 ip1dbg(("ip_snmp_get_mib2_ip: failed to allocate %u bytes\n",
9820 (uint_t)sizeof (old_ip_mib)));
9821 }
9822
9823 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9824 ip3dbg(("ip_snmp_get_mib2_ip: level %d, name %d, len %d\n",
9825 (int)optp->level, (int)optp->name, (int)optp->len));
9826 qreply(q, mpctl);
9827 return (mp2ctl);
9828 }
9829
9830 /* Per interface IPv4 statistics */
9831 static mblk_t *
9832 ip_snmp_get_mib2_ip_traffic_stats(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
9833 boolean_t legacy_req)
9834 {
9835 struct opthdr *optp;
9836 mblk_t *mp2ctl;
9837 ill_t *ill;
9838 ill_walk_context_t ctx;
9839 mblk_t *mp_tail = NULL;
9840 mib2_ipIfStatsEntry_t global_ip_mib;
9841 mib2_ipAddrEntry_t mae;
9842
9843 /*
9844 * Make a copy of the original message
9845 */
9846 mp2ctl = copymsg(mpctl);
9847
9848 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9849 optp->level = MIB2_IP;
9850 optp->name = MIB2_IP_TRAFFIC_STATS;
9851 /* Include "unknown interface" ip_mib */
9852 ipst->ips_ip_mib.ipIfStatsIPVersion = MIB2_INETADDRESSTYPE_ipv4;
9853 ipst->ips_ip_mib.ipIfStatsIfIndex =
9854 MIB2_UNKNOWN_INTERFACE; /* Flag to netstat */
9855 SET_MIB(ipst->ips_ip_mib.ipIfStatsForwarding,
9856 (ipst->ips_ip_forwarding ? 1 : 2));
9857 SET_MIB(ipst->ips_ip_mib.ipIfStatsDefaultTTL,
9858 (uint32_t)ipst->ips_ip_def_ttl);
9859 SET_MIB(ipst->ips_ip_mib.ipIfStatsEntrySize,
9860 sizeof (mib2_ipIfStatsEntry_t));
9861 SET_MIB(ipst->ips_ip_mib.ipIfStatsAddrEntrySize,
9862 sizeof (mib2_ipAddrEntry_t));
9863 SET_MIB(ipst->ips_ip_mib.ipIfStatsRouteEntrySize,
9864 sizeof (mib2_ipRouteEntry_t));
9865 SET_MIB(ipst->ips_ip_mib.ipIfStatsNetToMediaEntrySize,
9866 sizeof (mib2_ipNetToMediaEntry_t));
9867 SET_MIB(ipst->ips_ip_mib.ipIfStatsMemberEntrySize,
9868 sizeof (ip_member_t));
9869 SET_MIB(ipst->ips_ip_mib.ipIfStatsGroupSourceEntrySize,
9870 sizeof (ip_grpsrc_t));
9871
9872 bcopy(&ipst->ips_ip_mib, &global_ip_mib, sizeof (global_ip_mib));
9873
9874 if (legacy_req) {
9875 SET_MIB(global_ip_mib.ipIfStatsAddrEntrySize,
9876 LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t));
9877 }
9878
9879 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
9880 (char *)&global_ip_mib, (int)sizeof (global_ip_mib))) {
9881 ip1dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9882 "failed to allocate %u bytes\n",
9883 (uint_t)sizeof (global_ip_mib)));
9884 }
9885
9886 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
9887 ill = ILL_START_WALK_V4(&ctx, ipst);
9888 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
9889 ill->ill_ip_mib->ipIfStatsIfIndex =
9890 ill->ill_phyint->phyint_ifindex;
9891 SET_MIB(ill->ill_ip_mib->ipIfStatsForwarding,
9892 (ipst->ips_ip_forwarding ? 1 : 2));
9893 SET_MIB(ill->ill_ip_mib->ipIfStatsDefaultTTL,
9894 (uint32_t)ipst->ips_ip_def_ttl);
9895
9896 ip_mib2_add_ip_stats(&global_ip_mib, ill->ill_ip_mib);
9897 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
9898 (char *)ill->ill_ip_mib,
9899 (int)sizeof (*ill->ill_ip_mib))) {
9900 ip1dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9901 "failed to allocate %u bytes\n",
9902 (uint_t)sizeof (*ill->ill_ip_mib)));
9903 }
9904 }
9905 rw_exit(&ipst->ips_ill_g_lock);
9906
9907 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9908 ip3dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9909 "level %d, name %d, len %d\n",
9910 (int)optp->level, (int)optp->name, (int)optp->len));
9911 qreply(q, mpctl);
9912
9913 if (mp2ctl == NULL)
9914 return (NULL);
9915
9916 return (ip_snmp_get_mib2_ip(q, mp2ctl, &global_ip_mib, ipst,
9917 legacy_req));
9918 }
9919
9920 /* Global IPv4 ICMP statistics */
9921 static mblk_t *
9922 ip_snmp_get_mib2_icmp(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9923 {
9924 struct opthdr *optp;
9925 mblk_t *mp2ctl;
9926
9927 /*
9928 * Make a copy of the original message
9929 */
9930 mp2ctl = copymsg(mpctl);
9931
9932 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9933 optp->level = MIB2_ICMP;
9934 optp->name = 0;
9935 if (!snmp_append_data(mpctl->b_cont, (char *)&ipst->ips_icmp_mib,
9936 (int)sizeof (ipst->ips_icmp_mib))) {
9937 ip1dbg(("ip_snmp_get_mib2_icmp: failed to allocate %u bytes\n",
9938 (uint_t)sizeof (ipst->ips_icmp_mib)));
9939 }
9940 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9941 ip3dbg(("ip_snmp_get_mib2_icmp: level %d, name %d, len %d\n",
9942 (int)optp->level, (int)optp->name, (int)optp->len));
9943 qreply(q, mpctl);
9944 return (mp2ctl);
9945 }
9946
9947 /* Global IPv4 IGMP statistics */
9948 static mblk_t *
9949 ip_snmp_get_mib2_igmp(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9950 {
9951 struct opthdr *optp;
9952 mblk_t *mp2ctl;
9953
9954 /*
9955 * make a copy of the original message
9956 */
9957 mp2ctl = copymsg(mpctl);
9958
9959 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9960 optp->level = EXPER_IGMP;
9961 optp->name = 0;
9962 if (!snmp_append_data(mpctl->b_cont, (char *)&ipst->ips_igmpstat,
9963 (int)sizeof (ipst->ips_igmpstat))) {
9964 ip1dbg(("ip_snmp_get_mib2_igmp: failed to allocate %u bytes\n",
9965 (uint_t)sizeof (ipst->ips_igmpstat)));
9966 }
9967 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9968 ip3dbg(("ip_snmp_get_mib2_igmp: level %d, name %d, len %d\n",
9969 (int)optp->level, (int)optp->name, (int)optp->len));
9970 qreply(q, mpctl);
9971 return (mp2ctl);
9972 }
9973
9974 /* Global IPv4 Multicast Routing statistics */
9975 static mblk_t *
9976 ip_snmp_get_mib2_multi(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9977 {
9978 struct opthdr *optp;
9979 mblk_t *mp2ctl;
9980
9981 /*
9982 * make a copy of the original message
9983 */
9984 mp2ctl = copymsg(mpctl);
9985
9986 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9987 optp->level = EXPER_DVMRP;
9988 optp->name = 0;
9989 if (!ip_mroute_stats(mpctl->b_cont, ipst)) {
9990 ip0dbg(("ip_mroute_stats: failed\n"));
9991 }
9992 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9993 ip3dbg(("ip_snmp_get_mib2_multi: level %d, name %d, len %d\n",
9994 (int)optp->level, (int)optp->name, (int)optp->len));
9995 qreply(q, mpctl);
9996 return (mp2ctl);
9997 }
9998
9999 /* IPv4 address information */
10000 static mblk_t *
10001 ip_snmp_get_mib2_ip_addr(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
10002 boolean_t legacy_req)
10003 {
10004 struct opthdr *optp;
10005 mblk_t *mp2ctl;
10006 mblk_t *mp_tail = NULL;
10007 ill_t *ill;
10008 ipif_t *ipif;
10009 uint_t bitval;
10010 mib2_ipAddrEntry_t mae;
10011 size_t mae_size;
10012 zoneid_t zoneid;
10013 ill_walk_context_t ctx;
10014
10015 /*
10016 * make a copy of the original message
10017 */
10018 mp2ctl = copymsg(mpctl);
10019
10020 mae_size = (legacy_req) ? LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t) :
10021 sizeof (mib2_ipAddrEntry_t);
10022
10023 /* ipAddrEntryTable */
10024
10025 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10026 optp->level = MIB2_IP;
10027 optp->name = MIB2_IP_ADDR;
10028 zoneid = Q_TO_CONN(q)->conn_zoneid;
10029
10030 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10031 ill = ILL_START_WALK_V4(&ctx, ipst);
10032 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10033 for (ipif = ill->ill_ipif; ipif != NULL;
10034 ipif = ipif->ipif_next) {
10035 if (ipif->ipif_zoneid != zoneid &&
10036 ipif->ipif_zoneid != ALL_ZONES)
10037 continue;
10038 /* Sum of count from dead IRE_LO* and our current */
10039 mae.ipAdEntInfo.ae_ibcnt = ipif->ipif_ib_pkt_count;
10040 if (ipif->ipif_ire_local != NULL) {
10041 mae.ipAdEntInfo.ae_ibcnt +=
10042 ipif->ipif_ire_local->ire_ib_pkt_count;
10043 }
10044 mae.ipAdEntInfo.ae_obcnt = 0;
10045 mae.ipAdEntInfo.ae_focnt = 0;
10046
10047 ipif_get_name(ipif, mae.ipAdEntIfIndex.o_bytes,
10048 OCTET_LENGTH);
10049 mae.ipAdEntIfIndex.o_length =
10050 mi_strlen(mae.ipAdEntIfIndex.o_bytes);
10051 mae.ipAdEntAddr = ipif->ipif_lcl_addr;
10052 mae.ipAdEntNetMask = ipif->ipif_net_mask;
10053 mae.ipAdEntInfo.ae_subnet = ipif->ipif_subnet;
10054 mae.ipAdEntInfo.ae_subnet_len =
10055 ip_mask_to_plen(ipif->ipif_net_mask);
10056 mae.ipAdEntInfo.ae_src_addr = ipif->ipif_lcl_addr;
10057 for (bitval = 1;
10058 bitval &&
10059 !(bitval & ipif->ipif_brd_addr);
10060 bitval <<= 1)
10061 noop;
10062 mae.ipAdEntBcastAddr = bitval;
10063 mae.ipAdEntReasmMaxSize = IP_MAXPACKET;
10064 mae.ipAdEntInfo.ae_mtu = ipif->ipif_ill->ill_mtu;
10065 mae.ipAdEntInfo.ae_metric = ipif->ipif_ill->ill_metric;
10066 mae.ipAdEntInfo.ae_broadcast_addr =
10067 ipif->ipif_brd_addr;
10068 mae.ipAdEntInfo.ae_pp_dst_addr =
10069 ipif->ipif_pp_dst_addr;
10070 mae.ipAdEntInfo.ae_flags = ipif->ipif_flags |
10071 ill->ill_flags | ill->ill_phyint->phyint_flags;
10072 mae.ipAdEntRetransmitTime =
10073 ill->ill_reachable_retrans_time;
10074
10075 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10076 (char *)&mae, (int)mae_size)) {
10077 ip1dbg(("ip_snmp_get_mib2_ip_addr: failed to "
10078 "allocate %u bytes\n", (uint_t)mae_size));
10079 }
10080 }
10081 }
10082 rw_exit(&ipst->ips_ill_g_lock);
10083
10084 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10085 ip3dbg(("ip_snmp_get_mib2_ip_addr: level %d, name %d, len %d\n",
10086 (int)optp->level, (int)optp->name, (int)optp->len));
10087 qreply(q, mpctl);
10088 return (mp2ctl);
10089 }
10090
10091 /* IPv6 address information */
10092 static mblk_t *
10093 ip_snmp_get_mib2_ip6_addr(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
10094 boolean_t legacy_req)
10095 {
10096 struct opthdr *optp;
10097 mblk_t *mp2ctl;
10098 mblk_t *mp_tail = NULL;
10099 ill_t *ill;
10100 ipif_t *ipif;
10101 mib2_ipv6AddrEntry_t mae6;
10102 size_t mae6_size;
10103 zoneid_t zoneid;
10104 ill_walk_context_t ctx;
10105
10106 /*
10107 * make a copy of the original message
10108 */
10109 mp2ctl = copymsg(mpctl);
10110
10111 mae6_size = (legacy_req) ?
10112 LEGACY_MIB_SIZE(&mae6, mib2_ipv6AddrEntry_t) :
10113 sizeof (mib2_ipv6AddrEntry_t);
10114
10115 /* ipv6AddrEntryTable */
10116
10117 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10118 optp->level = MIB2_IP6;
10119 optp->name = MIB2_IP6_ADDR;
10120 zoneid = Q_TO_CONN(q)->conn_zoneid;
10121
10122 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10123 ill = ILL_START_WALK_V6(&ctx, ipst);
10124 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10125 for (ipif = ill->ill_ipif; ipif != NULL;
10126 ipif = ipif->ipif_next) {
10127 if (ipif->ipif_zoneid != zoneid &&
10128 ipif->ipif_zoneid != ALL_ZONES)
10129 continue;
10130 /* Sum of count from dead IRE_LO* and our current */
10131 mae6.ipv6AddrInfo.ae_ibcnt = ipif->ipif_ib_pkt_count;
10132 if (ipif->ipif_ire_local != NULL) {
10133 mae6.ipv6AddrInfo.ae_ibcnt +=
10134 ipif->ipif_ire_local->ire_ib_pkt_count;
10135 }
10136 mae6.ipv6AddrInfo.ae_obcnt = 0;
10137 mae6.ipv6AddrInfo.ae_focnt = 0;
10138
10139 ipif_get_name(ipif, mae6.ipv6AddrIfIndex.o_bytes,
10140 OCTET_LENGTH);
10141 mae6.ipv6AddrIfIndex.o_length =
10142 mi_strlen(mae6.ipv6AddrIfIndex.o_bytes);
10143 mae6.ipv6AddrAddress = ipif->ipif_v6lcl_addr;
10144 mae6.ipv6AddrPfxLength =
10145 ip_mask_to_plen_v6(&ipif->ipif_v6net_mask);
10146 mae6.ipv6AddrInfo.ae_subnet = ipif->ipif_v6subnet;
10147 mae6.ipv6AddrInfo.ae_subnet_len =
10148 mae6.ipv6AddrPfxLength;
10149 mae6.ipv6AddrInfo.ae_src_addr = ipif->ipif_v6lcl_addr;
10150
10151 /* Type: stateless(1), stateful(2), unknown(3) */
10152 if (ipif->ipif_flags & IPIF_ADDRCONF)
10153 mae6.ipv6AddrType = 1;
10154 else
10155 mae6.ipv6AddrType = 2;
10156 /* Anycast: true(1), false(2) */
10157 if (ipif->ipif_flags & IPIF_ANYCAST)
10158 mae6.ipv6AddrAnycastFlag = 1;
10159 else
10160 mae6.ipv6AddrAnycastFlag = 2;
10161
10162 /*
10163 * Address status: preferred(1), deprecated(2),
10164 * invalid(3), inaccessible(4), unknown(5)
10165 */
10166 if (ipif->ipif_flags & IPIF_NOLOCAL)
10167 mae6.ipv6AddrStatus = 3;
10168 else if (ipif->ipif_flags & IPIF_DEPRECATED)
10169 mae6.ipv6AddrStatus = 2;
10170 else
10171 mae6.ipv6AddrStatus = 1;
10172 mae6.ipv6AddrInfo.ae_mtu = ipif->ipif_ill->ill_mtu;
10173 mae6.ipv6AddrInfo.ae_metric =
10174 ipif->ipif_ill->ill_metric;
10175 mae6.ipv6AddrInfo.ae_pp_dst_addr =
10176 ipif->ipif_v6pp_dst_addr;
10177 mae6.ipv6AddrInfo.ae_flags = ipif->ipif_flags |
10178 ill->ill_flags | ill->ill_phyint->phyint_flags;
10179 mae6.ipv6AddrReasmMaxSize = IP_MAXPACKET;
10180 mae6.ipv6AddrIdentifier = ill->ill_token;
10181 mae6.ipv6AddrIdentifierLen = ill->ill_token_length;
10182 mae6.ipv6AddrReachableTime = ill->ill_reachable_time;
10183 mae6.ipv6AddrRetransmitTime =
10184 ill->ill_reachable_retrans_time;
10185 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10186 (char *)&mae6, (int)mae6_size)) {
10187 ip1dbg(("ip_snmp_get_mib2_ip6_addr: failed to "
10188 "allocate %u bytes\n",
10189 (uint_t)mae6_size));
10190 }
10191 }
10192 }
10193 rw_exit(&ipst->ips_ill_g_lock);
10194
10195 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10196 ip3dbg(("ip_snmp_get_mib2_ip6_addr: level %d, name %d, len %d\n",
10197 (int)optp->level, (int)optp->name, (int)optp->len));
10198 qreply(q, mpctl);
10199 return (mp2ctl);
10200 }
10201
10202 /* IPv4 multicast group membership. */
10203 static mblk_t *
10204 ip_snmp_get_mib2_ip_group_mem(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10205 {
10206 struct opthdr *optp;
10207 mblk_t *mp2ctl;
10208 ill_t *ill;
10209 ipif_t *ipif;
10210 ilm_t *ilm;
10211 ip_member_t ipm;
10212 mblk_t *mp_tail = NULL;
10213 ill_walk_context_t ctx;
10214 zoneid_t zoneid;
10215
10216 /*
10217 * make a copy of the original message
10218 */
10219 mp2ctl = copymsg(mpctl);
10220 zoneid = Q_TO_CONN(q)->conn_zoneid;
10221
10222 /* ipGroupMember table */
10223 optp = (struct opthdr *)&mpctl->b_rptr[
10224 sizeof (struct T_optmgmt_ack)];
10225 optp->level = MIB2_IP;
10226 optp->name = EXPER_IP_GROUP_MEMBERSHIP;
10227
10228 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10229 ill = ILL_START_WALK_V4(&ctx, ipst);
10230 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10231 /* Make sure the ill isn't going away. */
10232 if (!ill_check_and_refhold(ill))
10233 continue;
10234 rw_exit(&ipst->ips_ill_g_lock);
10235 rw_enter(&ill->ill_mcast_lock, RW_READER);
10236 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10237 if (ilm->ilm_zoneid != zoneid &&
10238 ilm->ilm_zoneid != ALL_ZONES)
10239 continue;
10240
10241 /* Is there an ipif for ilm_ifaddr? */
10242 for (ipif = ill->ill_ipif; ipif != NULL;
10243 ipif = ipif->ipif_next) {
10244 if (!IPIF_IS_CONDEMNED(ipif) &&
10245 ipif->ipif_lcl_addr == ilm->ilm_ifaddr &&
10246 ilm->ilm_ifaddr != INADDR_ANY)
10247 break;
10248 }
10249 if (ipif != NULL) {
10250 ipif_get_name(ipif,
10251 ipm.ipGroupMemberIfIndex.o_bytes,
10252 OCTET_LENGTH);
10253 } else {
10254 ill_get_name(ill,
10255 ipm.ipGroupMemberIfIndex.o_bytes,
10256 OCTET_LENGTH);
10257 }
10258 ipm.ipGroupMemberIfIndex.o_length =
10259 mi_strlen(ipm.ipGroupMemberIfIndex.o_bytes);
10260
10261 ipm.ipGroupMemberAddress = ilm->ilm_addr;
10262 ipm.ipGroupMemberRefCnt = ilm->ilm_refcnt;
10263 ipm.ipGroupMemberFilterMode = ilm->ilm_fmode;
10264 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10265 (char *)&ipm, (int)sizeof (ipm))) {
10266 ip1dbg(("ip_snmp_get_mib2_ip_group: "
10267 "failed to allocate %u bytes\n",
10268 (uint_t)sizeof (ipm)));
10269 }
10270 }
10271 rw_exit(&ill->ill_mcast_lock);
10272 ill_refrele(ill);
10273 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10274 }
10275 rw_exit(&ipst->ips_ill_g_lock);
10276 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10277 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10278 (int)optp->level, (int)optp->name, (int)optp->len));
10279 qreply(q, mpctl);
10280 return (mp2ctl);
10281 }
10282
10283 /* IPv6 multicast group membership. */
10284 static mblk_t *
10285 ip_snmp_get_mib2_ip6_group_mem(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10286 {
10287 struct opthdr *optp;
10288 mblk_t *mp2ctl;
10289 ill_t *ill;
10290 ilm_t *ilm;
10291 ipv6_member_t ipm6;
10292 mblk_t *mp_tail = NULL;
10293 ill_walk_context_t ctx;
10294 zoneid_t zoneid;
10295
10296 /*
10297 * make a copy of the original message
10298 */
10299 mp2ctl = copymsg(mpctl);
10300 zoneid = Q_TO_CONN(q)->conn_zoneid;
10301
10302 /* ip6GroupMember table */
10303 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10304 optp->level = MIB2_IP6;
10305 optp->name = EXPER_IP6_GROUP_MEMBERSHIP;
10306
10307 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10308 ill = ILL_START_WALK_V6(&ctx, ipst);
10309 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10310 /* Make sure the ill isn't going away. */
10311 if (!ill_check_and_refhold(ill))
10312 continue;
10313 rw_exit(&ipst->ips_ill_g_lock);
10314 /*
10315 * Normally we don't have any members on under IPMP interfaces.
10316 * We report them as a debugging aid.
10317 */
10318 rw_enter(&ill->ill_mcast_lock, RW_READER);
10319 ipm6.ipv6GroupMemberIfIndex = ill->ill_phyint->phyint_ifindex;
10320 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10321 if (ilm->ilm_zoneid != zoneid &&
10322 ilm->ilm_zoneid != ALL_ZONES)
10323 continue; /* not this zone */
10324 ipm6.ipv6GroupMemberAddress = ilm->ilm_v6addr;
10325 ipm6.ipv6GroupMemberRefCnt = ilm->ilm_refcnt;
10326 ipm6.ipv6GroupMemberFilterMode = ilm->ilm_fmode;
10327 if (!snmp_append_data2(mpctl->b_cont,
10328 &mp_tail,
10329 (char *)&ipm6, (int)sizeof (ipm6))) {
10330 ip1dbg(("ip_snmp_get_mib2_ip6_group: "
10331 "failed to allocate %u bytes\n",
10332 (uint_t)sizeof (ipm6)));
10333 }
10334 }
10335 rw_exit(&ill->ill_mcast_lock);
10336 ill_refrele(ill);
10337 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10338 }
10339 rw_exit(&ipst->ips_ill_g_lock);
10340
10341 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10342 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10343 (int)optp->level, (int)optp->name, (int)optp->len));
10344 qreply(q, mpctl);
10345 return (mp2ctl);
10346 }
10347
10348 /* IP multicast filtered sources */
10349 static mblk_t *
10350 ip_snmp_get_mib2_ip_group_src(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10351 {
10352 struct opthdr *optp;
10353 mblk_t *mp2ctl;
10354 ill_t *ill;
10355 ipif_t *ipif;
10356 ilm_t *ilm;
10357 ip_grpsrc_t ips;
10358 mblk_t *mp_tail = NULL;
10359 ill_walk_context_t ctx;
10360 zoneid_t zoneid;
10361 int i;
10362 slist_t *sl;
10363
10364 /*
10365 * make a copy of the original message
10366 */
10367 mp2ctl = copymsg(mpctl);
10368 zoneid = Q_TO_CONN(q)->conn_zoneid;
10369
10370 /* ipGroupSource table */
10371 optp = (struct opthdr *)&mpctl->b_rptr[
10372 sizeof (struct T_optmgmt_ack)];
10373 optp->level = MIB2_IP;
10374 optp->name = EXPER_IP_GROUP_SOURCES;
10375
10376 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10377 ill = ILL_START_WALK_V4(&ctx, ipst);
10378 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10379 /* Make sure the ill isn't going away. */
10380 if (!ill_check_and_refhold(ill))
10381 continue;
10382 rw_exit(&ipst->ips_ill_g_lock);
10383 rw_enter(&ill->ill_mcast_lock, RW_READER);
10384 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10385 sl = ilm->ilm_filter;
10386 if (ilm->ilm_zoneid != zoneid &&
10387 ilm->ilm_zoneid != ALL_ZONES)
10388 continue;
10389 if (SLIST_IS_EMPTY(sl))
10390 continue;
10391
10392 /* Is there an ipif for ilm_ifaddr? */
10393 for (ipif = ill->ill_ipif; ipif != NULL;
10394 ipif = ipif->ipif_next) {
10395 if (!IPIF_IS_CONDEMNED(ipif) &&
10396 ipif->ipif_lcl_addr == ilm->ilm_ifaddr &&
10397 ilm->ilm_ifaddr != INADDR_ANY)
10398 break;
10399 }
10400 if (ipif != NULL) {
10401 ipif_get_name(ipif,
10402 ips.ipGroupSourceIfIndex.o_bytes,
10403 OCTET_LENGTH);
10404 } else {
10405 ill_get_name(ill,
10406 ips.ipGroupSourceIfIndex.o_bytes,
10407 OCTET_LENGTH);
10408 }
10409 ips.ipGroupSourceIfIndex.o_length =
10410 mi_strlen(ips.ipGroupSourceIfIndex.o_bytes);
10411
10412 ips.ipGroupSourceGroup = ilm->ilm_addr;
10413 for (i = 0; i < sl->sl_numsrc; i++) {
10414 if (!IN6_IS_ADDR_V4MAPPED(&sl->sl_addr[i]))
10415 continue;
10416 IN6_V4MAPPED_TO_IPADDR(&sl->sl_addr[i],
10417 ips.ipGroupSourceAddress);
10418 if (snmp_append_data2(mpctl->b_cont, &mp_tail,
10419 (char *)&ips, (int)sizeof (ips)) == 0) {
10420 ip1dbg(("ip_snmp_get_mib2_ip_group_src:"
10421 " failed to allocate %u bytes\n",
10422 (uint_t)sizeof (ips)));
10423 }
10424 }
10425 }
10426 rw_exit(&ill->ill_mcast_lock);
10427 ill_refrele(ill);
10428 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10429 }
10430 rw_exit(&ipst->ips_ill_g_lock);
10431 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10432 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10433 (int)optp->level, (int)optp->name, (int)optp->len));
10434 qreply(q, mpctl);
10435 return (mp2ctl);
10436 }
10437
10438 /* IPv6 multicast filtered sources. */
10439 static mblk_t *
10440 ip_snmp_get_mib2_ip6_group_src(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10441 {
10442 struct opthdr *optp;
10443 mblk_t *mp2ctl;
10444 ill_t *ill;
10445 ilm_t *ilm;
10446 ipv6_grpsrc_t ips6;
10447 mblk_t *mp_tail = NULL;
10448 ill_walk_context_t ctx;
10449 zoneid_t zoneid;
10450 int i;
10451 slist_t *sl;
10452
10453 /*
10454 * make a copy of the original message
10455 */
10456 mp2ctl = copymsg(mpctl);
10457 zoneid = Q_TO_CONN(q)->conn_zoneid;
10458
10459 /* ip6GroupMember table */
10460 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10461 optp->level = MIB2_IP6;
10462 optp->name = EXPER_IP6_GROUP_SOURCES;
10463
10464 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10465 ill = ILL_START_WALK_V6(&ctx, ipst);
10466 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10467 /* Make sure the ill isn't going away. */
10468 if (!ill_check_and_refhold(ill))
10469 continue;
10470 rw_exit(&ipst->ips_ill_g_lock);
10471 /*
10472 * Normally we don't have any members on under IPMP interfaces.
10473 * We report them as a debugging aid.
10474 */
10475 rw_enter(&ill->ill_mcast_lock, RW_READER);
10476 ips6.ipv6GroupSourceIfIndex = ill->ill_phyint->phyint_ifindex;
10477 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10478 sl = ilm->ilm_filter;
10479 if (ilm->ilm_zoneid != zoneid &&
10480 ilm->ilm_zoneid != ALL_ZONES)
10481 continue;
10482 if (SLIST_IS_EMPTY(sl))
10483 continue;
10484 ips6.ipv6GroupSourceGroup = ilm->ilm_v6addr;
10485 for (i = 0; i < sl->sl_numsrc; i++) {
10486 ips6.ipv6GroupSourceAddress = sl->sl_addr[i];
10487 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10488 (char *)&ips6, (int)sizeof (ips6))) {
10489 ip1dbg(("ip_snmp_get_mib2_ip6_"
10490 "group_src: failed to allocate "
10491 "%u bytes\n",
10492 (uint_t)sizeof (ips6)));
10493 }
10494 }
10495 }
10496 rw_exit(&ill->ill_mcast_lock);
10497 ill_refrele(ill);
10498 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10499 }
10500 rw_exit(&ipst->ips_ill_g_lock);
10501
10502 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10503 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10504 (int)optp->level, (int)optp->name, (int)optp->len));
10505 qreply(q, mpctl);
10506 return (mp2ctl);
10507 }
10508
10509 /* Multicast routing virtual interface table. */
10510 static mblk_t *
10511 ip_snmp_get_mib2_virt_multi(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10512 {
10513 struct opthdr *optp;
10514 mblk_t *mp2ctl;
10515
10516 /*
10517 * make a copy of the original message
10518 */
10519 mp2ctl = copymsg(mpctl);
10520
10521 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10522 optp->level = EXPER_DVMRP;
10523 optp->name = EXPER_DVMRP_VIF;
10524 if (!ip_mroute_vif(mpctl->b_cont, ipst)) {
10525 ip0dbg(("ip_mroute_vif: failed\n"));
10526 }
10527 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10528 ip3dbg(("ip_snmp_get_mib2_virt_multi: level %d, name %d, len %d\n",
10529 (int)optp->level, (int)optp->name, (int)optp->len));
10530 qreply(q, mpctl);
10531 return (mp2ctl);
10532 }
10533
10534 /* Multicast routing table. */
10535 static mblk_t *
10536 ip_snmp_get_mib2_multi_rtable(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10537 {
10538 struct opthdr *optp;
10539 mblk_t *mp2ctl;
10540
10541 /*
10542 * make a copy of the original message
10543 */
10544 mp2ctl = copymsg(mpctl);
10545
10546 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10547 optp->level = EXPER_DVMRP;
10548 optp->name = EXPER_DVMRP_MRT;
10549 if (!ip_mroute_mrt(mpctl->b_cont, ipst)) {
10550 ip0dbg(("ip_mroute_mrt: failed\n"));
10551 }
10552 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10553 ip3dbg(("ip_snmp_get_mib2_multi_rtable: level %d, name %d, len %d\n",
10554 (int)optp->level, (int)optp->name, (int)optp->len));
10555 qreply(q, mpctl);
10556 return (mp2ctl);
10557 }
10558
10559 /*
10560 * Return ipRouteEntryTable, ipNetToMediaEntryTable, and ipRouteAttributeTable
10561 * in one IRE walk.
10562 */
10563 static mblk_t *
10564 ip_snmp_get_mib2_ip_route_media(queue_t *q, mblk_t *mpctl, int level,
10565 ip_stack_t *ipst)
10566 {
10567 struct opthdr *optp;
10568 mblk_t *mp2ctl; /* Returned */
10569 mblk_t *mp3ctl; /* nettomedia */
10570 mblk_t *mp4ctl; /* routeattrs */
10571 iproutedata_t ird;
10572 zoneid_t zoneid;
10573
10574 /*
10575 * make copies of the original message
10576 * - mp2ctl is returned unchanged to the caller for its use
10577 * - mpctl is sent upstream as ipRouteEntryTable
10578 * - mp3ctl is sent upstream as ipNetToMediaEntryTable
10579 * - mp4ctl is sent upstream as ipRouteAttributeTable
10580 */
10581 mp2ctl = copymsg(mpctl);
10582 mp3ctl = copymsg(mpctl);
10583 mp4ctl = copymsg(mpctl);
10584 if (mp3ctl == NULL || mp4ctl == NULL) {
10585 freemsg(mp4ctl);
10586 freemsg(mp3ctl);
10587 freemsg(mp2ctl);
10588 freemsg(mpctl);
10589 return (NULL);
10590 }
10591
10592 bzero(&ird, sizeof (ird));
10593
10594 ird.ird_route.lp_head = mpctl->b_cont;
10595 ird.ird_netmedia.lp_head = mp3ctl->b_cont;
10596 ird.ird_attrs.lp_head = mp4ctl->b_cont;
10597 /*
10598 * If the level has been set the special EXPER_IP_AND_ALL_IRES value,
10599 * then also include ire_testhidden IREs and IRE_IF_CLONE. This is
10600 * intended a temporary solution until a proper MIB API is provided
10601 * that provides complete filtering/caller-opt-in.
10602 */
10603 if (level == EXPER_IP_AND_ALL_IRES)
10604 ird.ird_flags |= IRD_REPORT_ALL;
10605
10606 zoneid = Q_TO_CONN(q)->conn_zoneid;
10607 ire_walk_v4(ip_snmp_get2_v4, &ird, zoneid, ipst);
10608
10609 /* ipRouteEntryTable in mpctl */
10610 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10611 optp->level = MIB2_IP;
10612 optp->name = MIB2_IP_ROUTE;
10613 optp->len = msgdsize(ird.ird_route.lp_head);
10614 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10615 (int)optp->level, (int)optp->name, (int)optp->len));
10616 qreply(q, mpctl);
10617
10618 /* ipNetToMediaEntryTable in mp3ctl */
10619 ncec_walk(NULL, ip_snmp_get2_v4_media, &ird, ipst);
10620
10621 optp = (struct opthdr *)&mp3ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10622 optp->level = MIB2_IP;
10623 optp->name = MIB2_IP_MEDIA;
10624 optp->len = msgdsize(ird.ird_netmedia.lp_head);
10625 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10626 (int)optp->level, (int)optp->name, (int)optp->len));
10627 qreply(q, mp3ctl);
10628
10629 /* ipRouteAttributeTable in mp4ctl */
10630 optp = (struct opthdr *)&mp4ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10631 optp->level = MIB2_IP;
10632 optp->name = EXPER_IP_RTATTR;
10633 optp->len = msgdsize(ird.ird_attrs.lp_head);
10634 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10635 (int)optp->level, (int)optp->name, (int)optp->len));
10636 if (optp->len == 0)
10637 freemsg(mp4ctl);
10638 else
10639 qreply(q, mp4ctl);
10640
10641 return (mp2ctl);
10642 }
10643
10644 /*
10645 * Return ipv6RouteEntryTable and ipv6RouteAttributeTable in one IRE walk, and
10646 * ipv6NetToMediaEntryTable in an NDP walk.
10647 */
10648 static mblk_t *
10649 ip_snmp_get_mib2_ip6_route_media(queue_t *q, mblk_t *mpctl, int level,
10650 ip_stack_t *ipst)
10651 {
10652 struct opthdr *optp;
10653 mblk_t *mp2ctl; /* Returned */
10654 mblk_t *mp3ctl; /* nettomedia */
10655 mblk_t *mp4ctl; /* routeattrs */
10656 iproutedata_t ird;
10657 zoneid_t zoneid;
10658
10659 /*
10660 * make copies of the original message
10661 * - mp2ctl is returned unchanged to the caller for its use
10662 * - mpctl is sent upstream as ipv6RouteEntryTable
10663 * - mp3ctl is sent upstream as ipv6NetToMediaEntryTable
10664 * - mp4ctl is sent upstream as ipv6RouteAttributeTable
10665 */
10666 mp2ctl = copymsg(mpctl);
10667 mp3ctl = copymsg(mpctl);
10668 mp4ctl = copymsg(mpctl);
10669 if (mp3ctl == NULL || mp4ctl == NULL) {
10670 freemsg(mp4ctl);
10671 freemsg(mp3ctl);
10672 freemsg(mp2ctl);
10673 freemsg(mpctl);
10674 return (NULL);
10675 }
10676
10677 bzero(&ird, sizeof (ird));
10678
10679 ird.ird_route.lp_head = mpctl->b_cont;
10680 ird.ird_netmedia.lp_head = mp3ctl->b_cont;
10681 ird.ird_attrs.lp_head = mp4ctl->b_cont;
10682 /*
10683 * If the level has been set the special EXPER_IP_AND_ALL_IRES value,
10684 * then also include ire_testhidden IREs and IRE_IF_CLONE. This is
10685 * intended a temporary solution until a proper MIB API is provided
10686 * that provides complete filtering/caller-opt-in.
10687 */
10688 if (level == EXPER_IP_AND_ALL_IRES)
10689 ird.ird_flags |= IRD_REPORT_ALL;
10690
10691 zoneid = Q_TO_CONN(q)->conn_zoneid;
10692 ire_walk_v6(ip_snmp_get2_v6_route, &ird, zoneid, ipst);
10693
10694 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10695 optp->level = MIB2_IP6;
10696 optp->name = MIB2_IP6_ROUTE;
10697 optp->len = msgdsize(ird.ird_route.lp_head);
10698 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10699 (int)optp->level, (int)optp->name, (int)optp->len));
10700 qreply(q, mpctl);
10701
10702 /* ipv6NetToMediaEntryTable in mp3ctl */
10703 ncec_walk(NULL, ip_snmp_get2_v6_media, &ird, ipst);
10704
10705 optp = (struct opthdr *)&mp3ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10706 optp->level = MIB2_IP6;
10707 optp->name = MIB2_IP6_MEDIA;
10708 optp->len = msgdsize(ird.ird_netmedia.lp_head);
10709 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10710 (int)optp->level, (int)optp->name, (int)optp->len));
10711 qreply(q, mp3ctl);
10712
10713 /* ipv6RouteAttributeTable in mp4ctl */
10714 optp = (struct opthdr *)&mp4ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10715 optp->level = MIB2_IP6;
10716 optp->name = EXPER_IP_RTATTR;
10717 optp->len = msgdsize(ird.ird_attrs.lp_head);
10718 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10719 (int)optp->level, (int)optp->name, (int)optp->len));
10720 if (optp->len == 0)
10721 freemsg(mp4ctl);
10722 else
10723 qreply(q, mp4ctl);
10724
10725 return (mp2ctl);
10726 }
10727
10728 /*
10729 * IPv6 mib: One per ill
10730 */
10731 static mblk_t *
10732 ip_snmp_get_mib2_ip6(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
10733 boolean_t legacy_req)
10734 {
10735 struct opthdr *optp;
10736 mblk_t *mp2ctl;
10737 ill_t *ill;
10738 ill_walk_context_t ctx;
10739 mblk_t *mp_tail = NULL;
10740 mib2_ipv6AddrEntry_t mae6;
10741 mib2_ipIfStatsEntry_t *ise;
10742 size_t ise_size, iae_size;
10743
10744 /*
10745 * Make a copy of the original message
10746 */
10747 mp2ctl = copymsg(mpctl);
10748
10749 /* fixed length IPv6 structure ... */
10750
10751 if (legacy_req) {
10752 ise_size = LEGACY_MIB_SIZE(&ipst->ips_ip6_mib,
10753 mib2_ipIfStatsEntry_t);
10754 iae_size = LEGACY_MIB_SIZE(&mae6, mib2_ipv6AddrEntry_t);
10755 } else {
10756 ise_size = sizeof (mib2_ipIfStatsEntry_t);
10757 iae_size = sizeof (mib2_ipv6AddrEntry_t);
10758 }
10759
10760 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10761 optp->level = MIB2_IP6;
10762 optp->name = 0;
10763 /* Include "unknown interface" ip6_mib */
10764 ipst->ips_ip6_mib.ipIfStatsIPVersion = MIB2_INETADDRESSTYPE_ipv6;
10765 ipst->ips_ip6_mib.ipIfStatsIfIndex =
10766 MIB2_UNKNOWN_INTERFACE; /* Flag to netstat */
10767 SET_MIB(ipst->ips_ip6_mib.ipIfStatsForwarding,
10768 ipst->ips_ipv6_forwarding ? 1 : 2);
10769 SET_MIB(ipst->ips_ip6_mib.ipIfStatsDefaultHopLimit,
10770 ipst->ips_ipv6_def_hops);
10771 SET_MIB(ipst->ips_ip6_mib.ipIfStatsEntrySize,
10772 sizeof (mib2_ipIfStatsEntry_t));
10773 SET_MIB(ipst->ips_ip6_mib.ipIfStatsAddrEntrySize,
10774 sizeof (mib2_ipv6AddrEntry_t));
10775 SET_MIB(ipst->ips_ip6_mib.ipIfStatsRouteEntrySize,
10776 sizeof (mib2_ipv6RouteEntry_t));
10777 SET_MIB(ipst->ips_ip6_mib.ipIfStatsNetToMediaEntrySize,
10778 sizeof (mib2_ipv6NetToMediaEntry_t));
10779 SET_MIB(ipst->ips_ip6_mib.ipIfStatsMemberEntrySize,
10780 sizeof (ipv6_member_t));
10781 SET_MIB(ipst->ips_ip6_mib.ipIfStatsGroupSourceEntrySize,
10782 sizeof (ipv6_grpsrc_t));
10783
10784 /*
10785 * Synchronize 64- and 32-bit counters
10786 */
10787 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInReceives,
10788 ipIfStatsHCInReceives);
10789 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInDelivers,
10790 ipIfStatsHCInDelivers);
10791 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutRequests,
10792 ipIfStatsHCOutRequests);
10793 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutForwDatagrams,
10794 ipIfStatsHCOutForwDatagrams);
10795 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutMcastPkts,
10796 ipIfStatsHCOutMcastPkts);
10797 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInMcastPkts,
10798 ipIfStatsHCInMcastPkts);
10799
10800 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10801 (char *)&ipst->ips_ip6_mib, (int)ise_size)) {
10802 ip1dbg(("ip_snmp_get_mib2_ip6: failed to allocate %u bytes\n",
10803 (uint_t)ise_size));
10804 } else if (legacy_req) {
10805 /* Adjust the EntrySize fields for legacy requests. */
10806 ise =
10807 (mib2_ipIfStatsEntry_t *)(mp_tail->b_wptr - (int)ise_size);
10808 SET_MIB(ise->ipIfStatsEntrySize, ise_size);
10809 SET_MIB(ise->ipIfStatsAddrEntrySize, iae_size);
10810 }
10811
10812 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10813 ill = ILL_START_WALK_V6(&ctx, ipst);
10814 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10815 ill->ill_ip_mib->ipIfStatsIfIndex =
10816 ill->ill_phyint->phyint_ifindex;
10817 SET_MIB(ill->ill_ip_mib->ipIfStatsForwarding,
10818 ipst->ips_ipv6_forwarding ? 1 : 2);
10819 SET_MIB(ill->ill_ip_mib->ipIfStatsDefaultHopLimit,
10820 ill->ill_max_hops);
10821
10822 /*
10823 * Synchronize 64- and 32-bit counters
10824 */
10825 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInReceives,
10826 ipIfStatsHCInReceives);
10827 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInDelivers,
10828 ipIfStatsHCInDelivers);
10829 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutRequests,
10830 ipIfStatsHCOutRequests);
10831 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutForwDatagrams,
10832 ipIfStatsHCOutForwDatagrams);
10833 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutMcastPkts,
10834 ipIfStatsHCOutMcastPkts);
10835 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInMcastPkts,
10836 ipIfStatsHCInMcastPkts);
10837
10838 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10839 (char *)ill->ill_ip_mib, (int)ise_size)) {
10840 ip1dbg(("ip_snmp_get_mib2_ip6: failed to allocate "
10841 "%u bytes\n", (uint_t)ise_size));
10842 } else if (legacy_req) {
10843 /* Adjust the EntrySize fields for legacy requests. */
10844 ise = (mib2_ipIfStatsEntry_t *)(mp_tail->b_wptr -
10845 (int)ise_size);
10846 SET_MIB(ise->ipIfStatsEntrySize, ise_size);
10847 SET_MIB(ise->ipIfStatsAddrEntrySize, iae_size);
10848 }
10849 }
10850 rw_exit(&ipst->ips_ill_g_lock);
10851
10852 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10853 ip3dbg(("ip_snmp_get_mib2_ip6: level %d, name %d, len %d\n",
10854 (int)optp->level, (int)optp->name, (int)optp->len));
10855 qreply(q, mpctl);
10856 return (mp2ctl);
10857 }
10858
10859 /*
10860 * ICMPv6 mib: One per ill
10861 */
10862 static mblk_t *
10863 ip_snmp_get_mib2_icmp6(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10864 {
10865 struct opthdr *optp;
10866 mblk_t *mp2ctl;
10867 ill_t *ill;
10868 ill_walk_context_t ctx;
10869 mblk_t *mp_tail = NULL;
10870 /*
10871 * Make a copy of the original message
10872 */
10873 mp2ctl = copymsg(mpctl);
10874
10875 /* fixed length ICMPv6 structure ... */
10876
10877 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10878 optp->level = MIB2_ICMP6;
10879 optp->name = 0;
10880 /* Include "unknown interface" icmp6_mib */
10881 ipst->ips_icmp6_mib.ipv6IfIcmpIfIndex =
10882 MIB2_UNKNOWN_INTERFACE; /* netstat flag */
10883 ipst->ips_icmp6_mib.ipv6IfIcmpEntrySize =
10884 sizeof (mib2_ipv6IfIcmpEntry_t);
10885 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10886 (char *)&ipst->ips_icmp6_mib,
10887 (int)sizeof (ipst->ips_icmp6_mib))) {
10888 ip1dbg(("ip_snmp_get_mib2_icmp6: failed to allocate %u bytes\n",
10889 (uint_t)sizeof (ipst->ips_icmp6_mib)));
10890 }
10891
10892 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10893 ill = ILL_START_WALK_V6(&ctx, ipst);
10894 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10895 ill->ill_icmp6_mib->ipv6IfIcmpIfIndex =
10896 ill->ill_phyint->phyint_ifindex;
10897 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10898 (char *)ill->ill_icmp6_mib,
10899 (int)sizeof (*ill->ill_icmp6_mib))) {
10900 ip1dbg(("ip_snmp_get_mib2_icmp6: failed to allocate "
10901 "%u bytes\n",
10902 (uint_t)sizeof (*ill->ill_icmp6_mib)));
10903 }
10904 }
10905 rw_exit(&ipst->ips_ill_g_lock);
10906
10907 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10908 ip3dbg(("ip_snmp_get_mib2_icmp6: level %d, name %d, len %d\n",
10909 (int)optp->level, (int)optp->name, (int)optp->len));
10910 qreply(q, mpctl);
10911 return (mp2ctl);
10912 }
10913
10914 /*
10915 * ire_walk routine to create both ipRouteEntryTable and
10916 * ipRouteAttributeTable in one IRE walk
10917 */
10918 static void
10919 ip_snmp_get2_v4(ire_t *ire, iproutedata_t *ird)
10920 {
10921 ill_t *ill;
10922 mib2_ipRouteEntry_t *re;
10923 mib2_ipAttributeEntry_t iaes;
10924 tsol_ire_gw_secattr_t *attrp;
10925 tsol_gc_t *gc = NULL;
10926 tsol_gcgrp_t *gcgrp = NULL;
10927 ip_stack_t *ipst = ire->ire_ipst;
10928
10929 ASSERT(ire->ire_ipversion == IPV4_VERSION);
10930
10931 if (!(ird->ird_flags & IRD_REPORT_ALL)) {
10932 if (ire->ire_testhidden)
10933 return;
10934 if (ire->ire_type & IRE_IF_CLONE)
10935 return;
10936 }
10937
10938 if ((re = kmem_zalloc(sizeof (*re), KM_NOSLEEP)) == NULL)
10939 return;
10940
10941 if ((attrp = ire->ire_gw_secattr) != NULL) {
10942 mutex_enter(&attrp->igsa_lock);
10943 if ((gc = attrp->igsa_gc) != NULL) {
10944 gcgrp = gc->gc_grp;
10945 ASSERT(gcgrp != NULL);
10946 rw_enter(&gcgrp->gcgrp_rwlock, RW_READER);
10947 }
10948 mutex_exit(&attrp->igsa_lock);
10949 }
10950 /*
10951 * Return all IRE types for route table... let caller pick and choose
10952 */
10953 re->ipRouteDest = ire->ire_addr;
10954 ill = ire->ire_ill;
10955 re->ipRouteIfIndex.o_length = 0;
10956 if (ill != NULL) {
10957 ill_get_name(ill, re->ipRouteIfIndex.o_bytes, OCTET_LENGTH);
10958 re->ipRouteIfIndex.o_length =
10959 mi_strlen(re->ipRouteIfIndex.o_bytes);
10960 }
10961 re->ipRouteMetric1 = -1;
10962 re->ipRouteMetric2 = -1;
10963 re->ipRouteMetric3 = -1;
10964 re->ipRouteMetric4 = -1;
10965
10966 re->ipRouteNextHop = ire->ire_gateway_addr;
10967 /* indirect(4), direct(3), or invalid(2) */
10968 if (ire->ire_flags & (RTF_REJECT | RTF_BLACKHOLE))
10969 re->ipRouteType = 2;
10970 else if (ire->ire_type & IRE_ONLINK)
10971 re->ipRouteType = 3;
10972 else
10973 re->ipRouteType = 4;
10974
10975 re->ipRouteProto = -1;
10976 re->ipRouteAge = gethrestime_sec() - ire->ire_create_time;
10977 re->ipRouteMask = ire->ire_mask;
10978 re->ipRouteMetric5 = -1;
10979 re->ipRouteInfo.re_max_frag = ire->ire_metrics.iulp_mtu;
10980 if (ire->ire_ill != NULL && re->ipRouteInfo.re_max_frag == 0)
10981 re->ipRouteInfo.re_max_frag = ire->ire_ill->ill_mtu;
10982
10983 re->ipRouteInfo.re_frag_flag = 0;
10984 re->ipRouteInfo.re_rtt = 0;
10985 re->ipRouteInfo.re_src_addr = 0;
10986 re->ipRouteInfo.re_ref = ire->ire_refcnt;
10987 re->ipRouteInfo.re_obpkt = ire->ire_ob_pkt_count;
10988 re->ipRouteInfo.re_ibpkt = ire->ire_ib_pkt_count;
10989 re->ipRouteInfo.re_flags = ire->ire_flags;
10990
10991 /* Add the IRE_IF_CLONE's counters to their parent IRE_INTERFACE */
10992 if (ire->ire_type & IRE_INTERFACE) {
10993 ire_t *child;
10994
10995 rw_enter(&ipst->ips_ire_dep_lock, RW_READER);
10996 child = ire->ire_dep_children;
10997 while (child != NULL) {
10998 re->ipRouteInfo.re_obpkt += child->ire_ob_pkt_count;
10999 re->ipRouteInfo.re_ibpkt += child->ire_ib_pkt_count;
11000 child = child->ire_dep_sib_next;
11001 }
11002 rw_exit(&ipst->ips_ire_dep_lock);
11003 }
11004
11005 if (ire->ire_flags & RTF_DYNAMIC) {
11006 re->ipRouteInfo.re_ire_type = IRE_HOST_REDIRECT;
11007 } else {
11008 re->ipRouteInfo.re_ire_type = ire->ire_type;
11009 }
11010
11011 if (!snmp_append_data2(ird->ird_route.lp_head, &ird->ird_route.lp_tail,
11012 (char *)re, (int)sizeof (*re))) {
11013 ip1dbg(("ip_snmp_get2_v4: failed to allocate %u bytes\n",
11014 (uint_t)sizeof (*re)));
11015 }
11016
11017 if (gc != NULL) {
11018 iaes.iae_routeidx = ird->ird_idx;
11019 iaes.iae_doi = gc->gc_db->gcdb_doi;
11020 iaes.iae_slrange = gc->gc_db->gcdb_slrange;
11021
11022 if (!snmp_append_data2(ird->ird_attrs.lp_head,
11023 &ird->ird_attrs.lp_tail, (char *)&iaes, sizeof (iaes))) {
11024 ip1dbg(("ip_snmp_get2_v4: failed to allocate %u "
11025 "bytes\n", (uint_t)sizeof (iaes)));
11026 }
11027 }
11028
11029 /* bump route index for next pass */
11030 ird->ird_idx++;
11031
11032 kmem_free(re, sizeof (*re));
11033 if (gcgrp != NULL)
11034 rw_exit(&gcgrp->gcgrp_rwlock);
11035 }
11036
11037 /*
11038 * ire_walk routine to create ipv6RouteEntryTable and ipRouteEntryTable.
11039 */
11040 static void
11041 ip_snmp_get2_v6_route(ire_t *ire, iproutedata_t *ird)
11042 {
11043 ill_t *ill;
11044 mib2_ipv6RouteEntry_t *re;
11045 mib2_ipAttributeEntry_t iaes;
11046 tsol_ire_gw_secattr_t *attrp;
11047 tsol_gc_t *gc = NULL;
11048 tsol_gcgrp_t *gcgrp = NULL;
11049 ip_stack_t *ipst = ire->ire_ipst;
11050
11051 ASSERT(ire->ire_ipversion == IPV6_VERSION);
11052
11053 if (!(ird->ird_flags & IRD_REPORT_ALL)) {
11054 if (ire->ire_testhidden)
11055 return;
11056 if (ire->ire_type & IRE_IF_CLONE)
11057 return;
11058 }
11059
11060 if ((re = kmem_zalloc(sizeof (*re), KM_NOSLEEP)) == NULL)
11061 return;
11062
11063 if ((attrp = ire->ire_gw_secattr) != NULL) {
11064 mutex_enter(&attrp->igsa_lock);
11065 if ((gc = attrp->igsa_gc) != NULL) {
11066 gcgrp = gc->gc_grp;
11067 ASSERT(gcgrp != NULL);
11068 rw_enter(&gcgrp->gcgrp_rwlock, RW_READER);
11069 }
11070 mutex_exit(&attrp->igsa_lock);
11071 }
11072 /*
11073 * Return all IRE types for route table... let caller pick and choose
11074 */
11075 re->ipv6RouteDest = ire->ire_addr_v6;
11076 re->ipv6RoutePfxLength = ip_mask_to_plen_v6(&ire->ire_mask_v6);
11077 re->ipv6RouteIndex = 0; /* Unique when multiple with same dest/plen */
11078 re->ipv6RouteIfIndex.o_length = 0;
11079 ill = ire->ire_ill;
11080 if (ill != NULL) {
11081 ill_get_name(ill, re->ipv6RouteIfIndex.o_bytes, OCTET_LENGTH);
11082 re->ipv6RouteIfIndex.o_length =
11083 mi_strlen(re->ipv6RouteIfIndex.o_bytes);
11084 }
11085
11086 ASSERT(!(ire->ire_type & IRE_BROADCAST));
11087
11088 mutex_enter(&ire->ire_lock);
11089 re->ipv6RouteNextHop = ire->ire_gateway_addr_v6;
11090 mutex_exit(&ire->ire_lock);
11091
11092 /* remote(4), local(3), or discard(2) */
11093 if (ire->ire_flags & (RTF_REJECT | RTF_BLACKHOLE))
11094 re->ipv6RouteType = 2;
11095 else if (ire->ire_type & IRE_ONLINK)
11096 re->ipv6RouteType = 3;
11097 else
11098 re->ipv6RouteType = 4;
11099
11100 re->ipv6RouteProtocol = -1;
11101 re->ipv6RoutePolicy = 0;
11102 re->ipv6RouteAge = gethrestime_sec() - ire->ire_create_time;
11103 re->ipv6RouteNextHopRDI = 0;
11104 re->ipv6RouteWeight = 0;
11105 re->ipv6RouteMetric = 0;
11106 re->ipv6RouteInfo.re_max_frag = ire->ire_metrics.iulp_mtu;
11107 if (ire->ire_ill != NULL && re->ipv6RouteInfo.re_max_frag == 0)
11108 re->ipv6RouteInfo.re_max_frag = ire->ire_ill->ill_mtu;
11109
11110 re->ipv6RouteInfo.re_frag_flag = 0;
11111 re->ipv6RouteInfo.re_rtt = 0;
11112 re->ipv6RouteInfo.re_src_addr = ipv6_all_zeros;
11113 re->ipv6RouteInfo.re_obpkt = ire->ire_ob_pkt_count;
11114 re->ipv6RouteInfo.re_ibpkt = ire->ire_ib_pkt_count;
11115 re->ipv6RouteInfo.re_ref = ire->ire_refcnt;
11116 re->ipv6RouteInfo.re_flags = ire->ire_flags;
11117
11118 /* Add the IRE_IF_CLONE's counters to their parent IRE_INTERFACE */
11119 if (ire->ire_type & IRE_INTERFACE) {
11120 ire_t *child;
11121
11122 rw_enter(&ipst->ips_ire_dep_lock, RW_READER);
11123 child = ire->ire_dep_children;
11124 while (child != NULL) {
11125 re->ipv6RouteInfo.re_obpkt += child->ire_ob_pkt_count;
11126 re->ipv6RouteInfo.re_ibpkt += child->ire_ib_pkt_count;
11127 child = child->ire_dep_sib_next;
11128 }
11129 rw_exit(&ipst->ips_ire_dep_lock);
11130 }
11131 if (ire->ire_flags & RTF_DYNAMIC) {
11132 re->ipv6RouteInfo.re_ire_type = IRE_HOST_REDIRECT;
11133 } else {
11134 re->ipv6RouteInfo.re_ire_type = ire->ire_type;
11135 }
11136
11137 if (!snmp_append_data2(ird->ird_route.lp_head, &ird->ird_route.lp_tail,
11138 (char *)re, (int)sizeof (*re))) {
11139 ip1dbg(("ip_snmp_get2_v6: failed to allocate %u bytes\n",
11140 (uint_t)sizeof (*re)));
11141 }
11142
11143 if (gc != NULL) {
11144 iaes.iae_routeidx = ird->ird_idx;
11145 iaes.iae_doi = gc->gc_db->gcdb_doi;
11146 iaes.iae_slrange = gc->gc_db->gcdb_slrange;
11147
11148 if (!snmp_append_data2(ird->ird_attrs.lp_head,
11149 &ird->ird_attrs.lp_tail, (char *)&iaes, sizeof (iaes))) {
11150 ip1dbg(("ip_snmp_get2_v6: failed to allocate %u "
11151 "bytes\n", (uint_t)sizeof (iaes)));
11152 }
11153 }
11154
11155 /* bump route index for next pass */
11156 ird->ird_idx++;
11157
11158 kmem_free(re, sizeof (*re));
11159 if (gcgrp != NULL)
11160 rw_exit(&gcgrp->gcgrp_rwlock);
11161 }
11162
11163 /*
11164 * ncec_walk routine to create ipv6NetToMediaEntryTable
11165 */
11166 static void
11167 ip_snmp_get2_v6_media(ncec_t *ncec, void *ptr)
11168 {
11169 iproutedata_t *ird = ptr;
11170 ill_t *ill;
11171 mib2_ipv6NetToMediaEntry_t ntme;
11172
11173 ill = ncec->ncec_ill;
11174 /* skip arpce entries, and loopback ncec entries */
11175 if (ill->ill_isv6 == B_FALSE || ill->ill_net_type == IRE_LOOPBACK)
11176 return;
11177 /*
11178 * Neighbor cache entry attached to IRE with on-link
11179 * destination.
11180 * We report all IPMP groups on ncec_ill which is normally the upper.
11181 */
11182 ntme.ipv6NetToMediaIfIndex = ill->ill_phyint->phyint_ifindex;
11183 ntme.ipv6NetToMediaNetAddress = ncec->ncec_addr;
11184 ntme.ipv6NetToMediaPhysAddress.o_length = ill->ill_phys_addr_length;
11185 if (ncec->ncec_lladdr != NULL) {
11186 bcopy(ncec->ncec_lladdr, ntme.ipv6NetToMediaPhysAddress.o_bytes,
11187 ntme.ipv6NetToMediaPhysAddress.o_length);
11188 }
11189 /*
11190 * Note: Returns ND_* states. Should be:
11191 * reachable(1), stale(2), delay(3), probe(4),
11192 * invalid(5), unknown(6)
11193 */
11194 ntme.ipv6NetToMediaState = ncec->ncec_state;
11195 ntme.ipv6NetToMediaLastUpdated = 0;
11196
11197 /* other(1), dynamic(2), static(3), local(4) */
11198 if (NCE_MYADDR(ncec)) {
11199 ntme.ipv6NetToMediaType = 4;
11200 } else if (ncec->ncec_flags & NCE_F_PUBLISH) {
11201 ntme.ipv6NetToMediaType = 1; /* proxy */
11202 } else if (ncec->ncec_flags & NCE_F_STATIC) {
11203 ntme.ipv6NetToMediaType = 3;
11204 } else if (ncec->ncec_flags & (NCE_F_MCAST|NCE_F_BCAST)) {
11205 ntme.ipv6NetToMediaType = 1;
11206 } else {
11207 ntme.ipv6NetToMediaType = 2;
11208 }
11209
11210 if (!snmp_append_data2(ird->ird_netmedia.lp_head,
11211 &ird->ird_netmedia.lp_tail, (char *)&ntme, sizeof (ntme))) {
11212 ip1dbg(("ip_snmp_get2_v6_media: failed to allocate %u bytes\n",
11213 (uint_t)sizeof (ntme)));
11214 }
11215 }
11216
11217 int
11218 nce2ace(ncec_t *ncec)
11219 {
11220 int flags = 0;
11221
11222 if (NCE_ISREACHABLE(ncec))
11223 flags |= ACE_F_RESOLVED;
11224 if (ncec->ncec_flags & NCE_F_AUTHORITY)
11225 flags |= ACE_F_AUTHORITY;
11226 if (ncec->ncec_flags & NCE_F_PUBLISH)
11227 flags |= ACE_F_PUBLISH;
11228 if ((ncec->ncec_flags & NCE_F_NONUD) != 0)
11229 flags |= ACE_F_PERMANENT;
11230 if (NCE_MYADDR(ncec))
11231 flags |= (ACE_F_MYADDR | ACE_F_AUTHORITY);
11232 if (ncec->ncec_flags & NCE_F_UNVERIFIED)
11233 flags |= ACE_F_UNVERIFIED;
11234 if (ncec->ncec_flags & NCE_F_AUTHORITY)
11235 flags |= ACE_F_AUTHORITY;
11236 if (ncec->ncec_flags & NCE_F_DELAYED)
11237 flags |= ACE_F_DELAYED;
11238 return (flags);
11239 }
11240
11241 /*
11242 * ncec_walk routine to create ipNetToMediaEntryTable
11243 */
11244 static void
11245 ip_snmp_get2_v4_media(ncec_t *ncec, void *ptr)
11246 {
11247 iproutedata_t *ird = ptr;
11248 ill_t *ill;
11249 mib2_ipNetToMediaEntry_t ntme;
11250 const char *name = "unknown";
11251 ipaddr_t ncec_addr;
11252
11253 ill = ncec->ncec_ill;
11254 if (ill->ill_isv6 || (ncec->ncec_flags & NCE_F_BCAST) ||
11255 ill->ill_net_type == IRE_LOOPBACK)
11256 return;
11257
11258 /* We report all IPMP groups on ncec_ill which is normally the upper. */
11259 name = ill->ill_name;
11260 /* Based on RFC 4293: other(1), inval(2), dyn(3), stat(4) */
11261 if (NCE_MYADDR(ncec)) {
11262 ntme.ipNetToMediaType = 4;
11263 } else if (ncec->ncec_flags & (NCE_F_MCAST|NCE_F_BCAST|NCE_F_PUBLISH)) {
11264 ntme.ipNetToMediaType = 1;
11265 } else {
11266 ntme.ipNetToMediaType = 3;
11267 }
11268 ntme.ipNetToMediaIfIndex.o_length = MIN(OCTET_LENGTH, strlen(name));
11269 bcopy(name, ntme.ipNetToMediaIfIndex.o_bytes,
11270 ntme.ipNetToMediaIfIndex.o_length);
11271
11272 IN6_V4MAPPED_TO_IPADDR(&ncec->ncec_addr, ncec_addr);
11273 bcopy(&ncec_addr, &ntme.ipNetToMediaNetAddress, sizeof (ncec_addr));
11274
11275 ntme.ipNetToMediaInfo.ntm_mask.o_length = sizeof (ipaddr_t);
11276 ncec_addr = INADDR_BROADCAST;
11277 bcopy(&ncec_addr, ntme.ipNetToMediaInfo.ntm_mask.o_bytes,
11278 sizeof (ncec_addr));
11279 /*
11280 * map all the flags to the ACE counterpart.
11281 */
11282 ntme.ipNetToMediaInfo.ntm_flags = nce2ace(ncec);
11283
11284 ntme.ipNetToMediaPhysAddress.o_length =
11285 MIN(OCTET_LENGTH, ill->ill_phys_addr_length);
11286
11287 if (!NCE_ISREACHABLE(ncec))
11288 ntme.ipNetToMediaPhysAddress.o_length = 0;
11289 else {
11290 if (ncec->ncec_lladdr != NULL) {
11291 bcopy(ncec->ncec_lladdr,
11292 ntme.ipNetToMediaPhysAddress.o_bytes,
11293 ntme.ipNetToMediaPhysAddress.o_length);
11294 }
11295 }
11296
11297 if (!snmp_append_data2(ird->ird_netmedia.lp_head,
11298 &ird->ird_netmedia.lp_tail, (char *)&ntme, sizeof (ntme))) {
11299 ip1dbg(("ip_snmp_get2_v4_media: failed to allocate %u bytes\n",
11300 (uint_t)sizeof (ntme)));
11301 }
11302 }
11303
11304 /*
11305 * return (0) if invalid set request, 1 otherwise, including non-tcp requests
11306 */
11307 /* ARGSUSED */
11308 int
11309 ip_snmp_set(queue_t *q, int level, int name, uchar_t *ptr, int len)
11310 {
11311 switch (level) {
11312 case MIB2_IP:
11313 case MIB2_ICMP:
11314 switch (name) {
11315 default:
11316 break;
11317 }
11318 return (1);
11319 default:
11320 return (1);
11321 }
11322 }
11323
11324 /*
11325 * When there exists both a 64- and 32-bit counter of a particular type
11326 * (i.e., InReceives), only the 64-bit counters are added.
11327 */
11328 void
11329 ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *o1, mib2_ipIfStatsEntry_t *o2)
11330 {
11331 UPDATE_MIB(o1, ipIfStatsInHdrErrors, o2->ipIfStatsInHdrErrors);
11332 UPDATE_MIB(o1, ipIfStatsInTooBigErrors, o2->ipIfStatsInTooBigErrors);
11333 UPDATE_MIB(o1, ipIfStatsInNoRoutes, o2->ipIfStatsInNoRoutes);
11334 UPDATE_MIB(o1, ipIfStatsInAddrErrors, o2->ipIfStatsInAddrErrors);
11335 UPDATE_MIB(o1, ipIfStatsInUnknownProtos, o2->ipIfStatsInUnknownProtos);
11336 UPDATE_MIB(o1, ipIfStatsInTruncatedPkts, o2->ipIfStatsInTruncatedPkts);
11337 UPDATE_MIB(o1, ipIfStatsInDiscards, o2->ipIfStatsInDiscards);
11338 UPDATE_MIB(o1, ipIfStatsOutDiscards, o2->ipIfStatsOutDiscards);
11339 UPDATE_MIB(o1, ipIfStatsOutFragOKs, o2->ipIfStatsOutFragOKs);
11340 UPDATE_MIB(o1, ipIfStatsOutFragFails, o2->ipIfStatsOutFragFails);
11341 UPDATE_MIB(o1, ipIfStatsOutFragCreates, o2->ipIfStatsOutFragCreates);
11342 UPDATE_MIB(o1, ipIfStatsReasmReqds, o2->ipIfStatsReasmReqds);
11343 UPDATE_MIB(o1, ipIfStatsReasmOKs, o2->ipIfStatsReasmOKs);
11344 UPDATE_MIB(o1, ipIfStatsReasmFails, o2->ipIfStatsReasmFails);
11345 UPDATE_MIB(o1, ipIfStatsOutNoRoutes, o2->ipIfStatsOutNoRoutes);
11346 UPDATE_MIB(o1, ipIfStatsReasmDuplicates, o2->ipIfStatsReasmDuplicates);
11347 UPDATE_MIB(o1, ipIfStatsReasmPartDups, o2->ipIfStatsReasmPartDups);
11348 UPDATE_MIB(o1, ipIfStatsForwProhibits, o2->ipIfStatsForwProhibits);
11349 UPDATE_MIB(o1, udpInCksumErrs, o2->udpInCksumErrs);
11350 UPDATE_MIB(o1, udpInOverflows, o2->udpInOverflows);
11351 UPDATE_MIB(o1, rawipInOverflows, o2->rawipInOverflows);
11352 UPDATE_MIB(o1, ipIfStatsInWrongIPVersion,
11353 o2->ipIfStatsInWrongIPVersion);
11354 UPDATE_MIB(o1, ipIfStatsOutWrongIPVersion,
11355 o2->ipIfStatsInWrongIPVersion);
11356 UPDATE_MIB(o1, ipIfStatsOutSwitchIPVersion,
11357 o2->ipIfStatsOutSwitchIPVersion);
11358 UPDATE_MIB(o1, ipIfStatsHCInReceives, o2->ipIfStatsHCInReceives);
11359 UPDATE_MIB(o1, ipIfStatsHCInOctets, o2->ipIfStatsHCInOctets);
11360 UPDATE_MIB(o1, ipIfStatsHCInForwDatagrams,
11361 o2->ipIfStatsHCInForwDatagrams);
11362 UPDATE_MIB(o1, ipIfStatsHCInDelivers, o2->ipIfStatsHCInDelivers);
11363 UPDATE_MIB(o1, ipIfStatsHCOutRequests, o2->ipIfStatsHCOutRequests);
11364 UPDATE_MIB(o1, ipIfStatsHCOutForwDatagrams,
11365 o2->ipIfStatsHCOutForwDatagrams);
11366 UPDATE_MIB(o1, ipIfStatsOutFragReqds, o2->ipIfStatsOutFragReqds);
11367 UPDATE_MIB(o1, ipIfStatsHCOutTransmits, o2->ipIfStatsHCOutTransmits);
11368 UPDATE_MIB(o1, ipIfStatsHCOutOctets, o2->ipIfStatsHCOutOctets);
11369 UPDATE_MIB(o1, ipIfStatsHCInMcastPkts, o2->ipIfStatsHCInMcastPkts);
11370 UPDATE_MIB(o1, ipIfStatsHCInMcastOctets, o2->ipIfStatsHCInMcastOctets);
11371 UPDATE_MIB(o1, ipIfStatsHCOutMcastPkts, o2->ipIfStatsHCOutMcastPkts);
11372 UPDATE_MIB(o1, ipIfStatsHCOutMcastOctets,
11373 o2->ipIfStatsHCOutMcastOctets);
11374 UPDATE_MIB(o1, ipIfStatsHCInBcastPkts, o2->ipIfStatsHCInBcastPkts);
11375 UPDATE_MIB(o1, ipIfStatsHCOutBcastPkts, o2->ipIfStatsHCOutBcastPkts);
11376 UPDATE_MIB(o1, ipsecInSucceeded, o2->ipsecInSucceeded);
11377 UPDATE_MIB(o1, ipsecInFailed, o2->ipsecInFailed);
11378 UPDATE_MIB(o1, ipInCksumErrs, o2->ipInCksumErrs);
11379 UPDATE_MIB(o1, tcpInErrs, o2->tcpInErrs);
11380 UPDATE_MIB(o1, udpNoPorts, o2->udpNoPorts);
11381 }
11382
11383 void
11384 ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *o1, mib2_ipv6IfIcmpEntry_t *o2)
11385 {
11386 UPDATE_MIB(o1, ipv6IfIcmpInMsgs, o2->ipv6IfIcmpInMsgs);
11387 UPDATE_MIB(o1, ipv6IfIcmpInErrors, o2->ipv6IfIcmpInErrors);
11388 UPDATE_MIB(o1, ipv6IfIcmpInDestUnreachs, o2->ipv6IfIcmpInDestUnreachs);
11389 UPDATE_MIB(o1, ipv6IfIcmpInAdminProhibs, o2->ipv6IfIcmpInAdminProhibs);
11390 UPDATE_MIB(o1, ipv6IfIcmpInTimeExcds, o2->ipv6IfIcmpInTimeExcds);
11391 UPDATE_MIB(o1, ipv6IfIcmpInParmProblems, o2->ipv6IfIcmpInParmProblems);
11392 UPDATE_MIB(o1, ipv6IfIcmpInPktTooBigs, o2->ipv6IfIcmpInPktTooBigs);
11393 UPDATE_MIB(o1, ipv6IfIcmpInEchos, o2->ipv6IfIcmpInEchos);
11394 UPDATE_MIB(o1, ipv6IfIcmpInEchoReplies, o2->ipv6IfIcmpInEchoReplies);
11395 UPDATE_MIB(o1, ipv6IfIcmpInRouterSolicits,
11396 o2->ipv6IfIcmpInRouterSolicits);
11397 UPDATE_MIB(o1, ipv6IfIcmpInRouterAdvertisements,
11398 o2->ipv6IfIcmpInRouterAdvertisements);
11399 UPDATE_MIB(o1, ipv6IfIcmpInNeighborSolicits,
11400 o2->ipv6IfIcmpInNeighborSolicits);
11401 UPDATE_MIB(o1, ipv6IfIcmpInNeighborAdvertisements,
11402 o2->ipv6IfIcmpInNeighborAdvertisements);
11403 UPDATE_MIB(o1, ipv6IfIcmpInRedirects, o2->ipv6IfIcmpInRedirects);
11404 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembQueries,
11405 o2->ipv6IfIcmpInGroupMembQueries);
11406 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembResponses,
11407 o2->ipv6IfIcmpInGroupMembResponses);
11408 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembReductions,
11409 o2->ipv6IfIcmpInGroupMembReductions);
11410 UPDATE_MIB(o1, ipv6IfIcmpOutMsgs, o2->ipv6IfIcmpOutMsgs);
11411 UPDATE_MIB(o1, ipv6IfIcmpOutErrors, o2->ipv6IfIcmpOutErrors);
11412 UPDATE_MIB(o1, ipv6IfIcmpOutDestUnreachs,
11413 o2->ipv6IfIcmpOutDestUnreachs);
11414 UPDATE_MIB(o1, ipv6IfIcmpOutAdminProhibs,
11415 o2->ipv6IfIcmpOutAdminProhibs);
11416 UPDATE_MIB(o1, ipv6IfIcmpOutTimeExcds, o2->ipv6IfIcmpOutTimeExcds);
11417 UPDATE_MIB(o1, ipv6IfIcmpOutParmProblems,
11418 o2->ipv6IfIcmpOutParmProblems);
11419 UPDATE_MIB(o1, ipv6IfIcmpOutPktTooBigs, o2->ipv6IfIcmpOutPktTooBigs);
11420 UPDATE_MIB(o1, ipv6IfIcmpOutEchos, o2->ipv6IfIcmpOutEchos);
11421 UPDATE_MIB(o1, ipv6IfIcmpOutEchoReplies, o2->ipv6IfIcmpOutEchoReplies);
11422 UPDATE_MIB(o1, ipv6IfIcmpOutRouterSolicits,
11423 o2->ipv6IfIcmpOutRouterSolicits);
11424 UPDATE_MIB(o1, ipv6IfIcmpOutRouterAdvertisements,
11425 o2->ipv6IfIcmpOutRouterAdvertisements);
11426 UPDATE_MIB(o1, ipv6IfIcmpOutNeighborSolicits,
11427 o2->ipv6IfIcmpOutNeighborSolicits);
11428 UPDATE_MIB(o1, ipv6IfIcmpOutNeighborAdvertisements,
11429 o2->ipv6IfIcmpOutNeighborAdvertisements);
11430 UPDATE_MIB(o1, ipv6IfIcmpOutRedirects, o2->ipv6IfIcmpOutRedirects);
11431 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembQueries,
11432 o2->ipv6IfIcmpOutGroupMembQueries);
11433 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembResponses,
11434 o2->ipv6IfIcmpOutGroupMembResponses);
11435 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembReductions,
11436 o2->ipv6IfIcmpOutGroupMembReductions);
11437 UPDATE_MIB(o1, ipv6IfIcmpInOverflows, o2->ipv6IfIcmpInOverflows);
11438 UPDATE_MIB(o1, ipv6IfIcmpBadHoplimit, o2->ipv6IfIcmpBadHoplimit);
11439 UPDATE_MIB(o1, ipv6IfIcmpInBadNeighborAdvertisements,
11440 o2->ipv6IfIcmpInBadNeighborAdvertisements);
11441 UPDATE_MIB(o1, ipv6IfIcmpInBadNeighborSolicitations,
11442 o2->ipv6IfIcmpInBadNeighborSolicitations);
11443 UPDATE_MIB(o1, ipv6IfIcmpInBadRedirects, o2->ipv6IfIcmpInBadRedirects);
11444 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembTotal,
11445 o2->ipv6IfIcmpInGroupMembTotal);
11446 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembBadQueries,
11447 o2->ipv6IfIcmpInGroupMembBadQueries);
11448 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembBadReports,
11449 o2->ipv6IfIcmpInGroupMembBadReports);
11450 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembOurReports,
11451 o2->ipv6IfIcmpInGroupMembOurReports);
11452 }
11453
11454 /*
11455 * Called before the options are updated to check if this packet will
11456 * be source routed from here.
11457 * This routine assumes that the options are well formed i.e. that they
11458 * have already been checked.
11459 */
11460 boolean_t
11461 ip_source_routed(ipha_t *ipha, ip_stack_t *ipst)
11462 {
11463 ipoptp_t opts;
11464 uchar_t *opt;
11465 uint8_t optval;
11466 uint8_t optlen;
11467 ipaddr_t dst;
11468
11469 if (IS_SIMPLE_IPH(ipha)) {
11470 ip2dbg(("not source routed\n"));
11471 return (B_FALSE);
11472 }
11473 dst = ipha->ipha_dst;
11474 for (optval = ipoptp_first(&opts, ipha);
11475 optval != IPOPT_EOL;
11476 optval = ipoptp_next(&opts)) {
11477 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11478 opt = opts.ipoptp_cur;
11479 optlen = opts.ipoptp_len;
11480 ip2dbg(("ip_source_routed: opt %d, len %d\n",
11481 optval, optlen));
11482 switch (optval) {
11483 uint32_t off;
11484 case IPOPT_SSRR:
11485 case IPOPT_LSRR:
11486 /*
11487 * If dst is one of our addresses and there are some
11488 * entries left in the source route return (true).
11489 */
11490 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
11491 ip2dbg(("ip_source_routed: not next"
11492 " source route 0x%x\n",
11493 ntohl(dst)));
11494 return (B_FALSE);
11495 }
11496 off = opt[IPOPT_OFFSET];
11497 off--;
11498 if (optlen < IP_ADDR_LEN ||
11499 off > optlen - IP_ADDR_LEN) {
11500 /* End of source route */
11501 ip1dbg(("ip_source_routed: end of SR\n"));
11502 return (B_FALSE);
11503 }
11504 return (B_TRUE);
11505 }
11506 }
11507 ip2dbg(("not source routed\n"));
11508 return (B_FALSE);
11509 }
11510
11511 /*
11512 * ip_unbind is called by the transports to remove a conn from
11513 * the fanout table.
11514 */
11515 void
11516 ip_unbind(conn_t *connp)
11517 {
11518
11519 ASSERT(!MUTEX_HELD(&connp->conn_lock));
11520
11521 if (is_system_labeled() && connp->conn_anon_port) {
11522 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
11523 connp->conn_mlp_type, connp->conn_proto,
11524 ntohs(connp->conn_lport), B_FALSE);
11525 connp->conn_anon_port = 0;
11526 }
11527 connp->conn_mlp_type = mlptSingle;
11528
11529 ipcl_hash_remove(connp);
11530 }
11531
11532 /*
11533 * Used for deciding the MSS size for the upper layer. Thus
11534 * we need to check the outbound policy values in the conn.
11535 */
11536 int
11537 conn_ipsec_length(conn_t *connp)
11538 {
11539 ipsec_latch_t *ipl;
11540
11541 ipl = connp->conn_latch;
11542 if (ipl == NULL)
11543 return (0);
11544
11545 if (connp->conn_ixa->ixa_ipsec_policy == NULL)
11546 return (0);
11547
11548 return (connp->conn_ixa->ixa_ipsec_policy->ipsp_act->ipa_ovhd);
11549 }
11550
11551 /*
11552 * Returns an estimate of the IPsec headers size. This is used if
11553 * we don't want to call into IPsec to get the exact size.
11554 */
11555 int
11556 ipsec_out_extra_length(ip_xmit_attr_t *ixa)
11557 {
11558 ipsec_action_t *a;
11559
11560 if (!(ixa->ixa_flags & IXAF_IPSEC_SECURE))
11561 return (0);
11562
11563 a = ixa->ixa_ipsec_action;
11564 if (a == NULL) {
11565 ASSERT(ixa->ixa_ipsec_policy != NULL);
11566 a = ixa->ixa_ipsec_policy->ipsp_act;
11567 }
11568 ASSERT(a != NULL);
11569
11570 return (a->ipa_ovhd);
11571 }
11572
11573 /*
11574 * If there are any source route options, return the true final
11575 * destination. Otherwise, return the destination.
11576 */
11577 ipaddr_t
11578 ip_get_dst(ipha_t *ipha)
11579 {
11580 ipoptp_t opts;
11581 uchar_t *opt;
11582 uint8_t optval;
11583 uint8_t optlen;
11584 ipaddr_t dst;
11585 uint32_t off;
11586
11587 dst = ipha->ipha_dst;
11588
11589 if (IS_SIMPLE_IPH(ipha))
11590 return (dst);
11591
11592 for (optval = ipoptp_first(&opts, ipha);
11593 optval != IPOPT_EOL;
11594 optval = ipoptp_next(&opts)) {
11595 opt = opts.ipoptp_cur;
11596 optlen = opts.ipoptp_len;
11597 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11598 switch (optval) {
11599 case IPOPT_SSRR:
11600 case IPOPT_LSRR:
11601 off = opt[IPOPT_OFFSET];
11602 /*
11603 * If one of the conditions is true, it means
11604 * end of options and dst already has the right
11605 * value.
11606 */
11607 if (!(optlen < IP_ADDR_LEN || off > optlen - 3)) {
11608 off = optlen - IP_ADDR_LEN;
11609 bcopy(&opt[off], &dst, IP_ADDR_LEN);
11610 }
11611 return (dst);
11612 default:
11613 break;
11614 }
11615 }
11616
11617 return (dst);
11618 }
11619
11620 /*
11621 * Outbound IP fragmentation routine.
11622 * Assumes the caller has checked whether or not fragmentation should
11623 * be allowed. Here we copy the DF bit from the header to all the generated
11624 * fragments.
11625 */
11626 int
11627 ip_fragment_v4(mblk_t *mp_orig, nce_t *nce, iaflags_t ixaflags,
11628 uint_t pkt_len, uint32_t max_frag, uint32_t xmit_hint, zoneid_t szone,
11629 zoneid_t nolzid, pfirepostfrag_t postfragfn, uintptr_t *ixa_cookie)
11630 {
11631 int i1;
11632 int hdr_len;
11633 mblk_t *hdr_mp;
11634 ipha_t *ipha;
11635 int ip_data_end;
11636 int len;
11637 mblk_t *mp = mp_orig;
11638 int offset;
11639 ill_t *ill = nce->nce_ill;
11640 ip_stack_t *ipst = ill->ill_ipst;
11641 mblk_t *carve_mp;
11642 uint32_t frag_flag;
11643 uint_t priority = mp->b_band;
11644 int error = 0;
11645
11646 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragReqds);
11647
11648 if (pkt_len != msgdsize(mp)) {
11649 ip0dbg(("Packet length mismatch: %d, %ld\n",
11650 pkt_len, msgdsize(mp)));
11651 freemsg(mp);
11652 return (EINVAL);
11653 }
11654
11655 if (max_frag == 0) {
11656 ip1dbg(("ip_fragment_v4: max_frag is zero. Dropping packet\n"));
11657 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11658 ip_drop_output("FragFails: zero max_frag", mp, ill);
11659 freemsg(mp);
11660 return (EINVAL);
11661 }
11662
11663 ASSERT(MBLKL(mp) >= sizeof (ipha_t));
11664 ipha = (ipha_t *)mp->b_rptr;
11665 ASSERT(ntohs(ipha->ipha_length) == pkt_len);
11666 frag_flag = ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_DF;
11667
11668 /*
11669 * Establish the starting offset. May not be zero if we are fragging
11670 * a fragment that is being forwarded.
11671 */
11672 offset = ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_OFFSET;
11673
11674 /* TODO why is this test needed? */
11675 if (((max_frag - ntohs(ipha->ipha_length)) & ~7) < 8) {
11676 /* TODO: notify ulp somehow */
11677 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11678 ip_drop_output("FragFails: bad starting offset", mp, ill);
11679 freemsg(mp);
11680 return (EINVAL);
11681 }
11682
11683 hdr_len = IPH_HDR_LENGTH(ipha);
11684 ipha->ipha_hdr_checksum = 0;
11685
11686 /*
11687 * Establish the number of bytes maximum per frag, after putting
11688 * in the header.
11689 */
11690 len = (max_frag - hdr_len) & ~7;
11691
11692 /* Get a copy of the header for the trailing frags */
11693 hdr_mp = ip_fragment_copyhdr((uchar_t *)ipha, hdr_len, offset, ipst,
11694 mp);
11695 if (hdr_mp == NULL) {
11696 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11697 ip_drop_output("FragFails: no hdr_mp", mp, ill);
11698 freemsg(mp);
11699 return (ENOBUFS);
11700 }
11701
11702 /* Store the starting offset, with the MoreFrags flag. */
11703 i1 = offset | IPH_MF | frag_flag;
11704 ipha->ipha_fragment_offset_and_flags = htons((uint16_t)i1);
11705
11706 /* Establish the ending byte offset, based on the starting offset. */
11707 offset <<= 3;
11708 ip_data_end = offset + ntohs(ipha->ipha_length) - hdr_len;
11709
11710 /* Store the length of the first fragment in the IP header. */
11711 i1 = len + hdr_len;
11712 ASSERT(i1 <= IP_MAXPACKET);
11713 ipha->ipha_length = htons((uint16_t)i1);
11714
11715 /*
11716 * Compute the IP header checksum for the first frag. We have to
11717 * watch out that we stop at the end of the header.
11718 */
11719 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
11720
11721 /*
11722 * Now carve off the first frag. Note that this will include the
11723 * original IP header.
11724 */
11725 if (!(mp = ip_carve_mp(&mp_orig, i1))) {
11726 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11727 ip_drop_output("FragFails: could not carve mp", mp_orig, ill);
11728 freeb(hdr_mp);
11729 freemsg(mp_orig);
11730 return (ENOBUFS);
11731 }
11732
11733 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragCreates);
11734
11735 error = postfragfn(mp, nce, ixaflags, i1, xmit_hint, szone, nolzid,
11736 ixa_cookie);
11737 if (error != 0 && error != EWOULDBLOCK) {
11738 /* No point in sending the other fragments */
11739 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11740 ip_drop_output("FragFails: postfragfn failed", mp_orig, ill);
11741 freeb(hdr_mp);
11742 freemsg(mp_orig);
11743 return (error);
11744 }
11745
11746 /* No need to redo state machine in loop */
11747 ixaflags &= ~IXAF_REACH_CONF;
11748
11749 /* Advance the offset to the second frag starting point. */
11750 offset += len;
11751 /*
11752 * Update hdr_len from the copied header - there might be less options
11753 * in the later fragments.
11754 */
11755 hdr_len = IPH_HDR_LENGTH(hdr_mp->b_rptr);
11756 /* Loop until done. */
11757 for (;;) {
11758 uint16_t offset_and_flags;
11759 uint16_t ip_len;
11760
11761 if (ip_data_end - offset > len) {
11762 /*
11763 * Carve off the appropriate amount from the original
11764 * datagram.
11765 */
11766 if (!(carve_mp = ip_carve_mp(&mp_orig, len))) {
11767 mp = NULL;
11768 break;
11769 }
11770 /*
11771 * More frags after this one. Get another copy
11772 * of the header.
11773 */
11774 if (carve_mp->b_datap->db_ref == 1 &&
11775 hdr_mp->b_wptr - hdr_mp->b_rptr <
11776 carve_mp->b_rptr - carve_mp->b_datap->db_base) {
11777 /* Inline IP header */
11778 carve_mp->b_rptr -= hdr_mp->b_wptr -
11779 hdr_mp->b_rptr;
11780 bcopy(hdr_mp->b_rptr, carve_mp->b_rptr,
11781 hdr_mp->b_wptr - hdr_mp->b_rptr);
11782 mp = carve_mp;
11783 } else {
11784 if (!(mp = copyb(hdr_mp))) {
11785 freemsg(carve_mp);
11786 break;
11787 }
11788 /* Get priority marking, if any. */
11789 mp->b_band = priority;
11790 mp->b_cont = carve_mp;
11791 }
11792 ipha = (ipha_t *)mp->b_rptr;
11793 offset_and_flags = IPH_MF;
11794 } else {
11795 /*
11796 * Last frag. Consume the header. Set len to
11797 * the length of this last piece.
11798 */
11799 len = ip_data_end - offset;
11800
11801 /*
11802 * Carve off the appropriate amount from the original
11803 * datagram.
11804 */
11805 if (!(carve_mp = ip_carve_mp(&mp_orig, len))) {
11806 mp = NULL;
11807 break;
11808 }
11809 if (carve_mp->b_datap->db_ref == 1 &&
11810 hdr_mp->b_wptr - hdr_mp->b_rptr <
11811 carve_mp->b_rptr - carve_mp->b_datap->db_base) {
11812 /* Inline IP header */
11813 carve_mp->b_rptr -= hdr_mp->b_wptr -
11814 hdr_mp->b_rptr;
11815 bcopy(hdr_mp->b_rptr, carve_mp->b_rptr,
11816 hdr_mp->b_wptr - hdr_mp->b_rptr);
11817 mp = carve_mp;
11818 freeb(hdr_mp);
11819 hdr_mp = mp;
11820 } else {
11821 mp = hdr_mp;
11822 /* Get priority marking, if any. */
11823 mp->b_band = priority;
11824 mp->b_cont = carve_mp;
11825 }
11826 ipha = (ipha_t *)mp->b_rptr;
11827 /* A frag of a frag might have IPH_MF non-zero */
11828 offset_and_flags =
11829 ntohs(ipha->ipha_fragment_offset_and_flags) &
11830 IPH_MF;
11831 }
11832 offset_and_flags |= (uint16_t)(offset >> 3);
11833 offset_and_flags |= (uint16_t)frag_flag;
11834 /* Store the offset and flags in the IP header. */
11835 ipha->ipha_fragment_offset_and_flags = htons(offset_and_flags);
11836
11837 /* Store the length in the IP header. */
11838 ip_len = (uint16_t)(len + hdr_len);
11839 ipha->ipha_length = htons(ip_len);
11840
11841 /*
11842 * Set the IP header checksum. Note that mp is just
11843 * the header, so this is easy to pass to ip_csum.
11844 */
11845 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
11846
11847 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragCreates);
11848
11849 error = postfragfn(mp, nce, ixaflags, ip_len, xmit_hint, szone,
11850 nolzid, ixa_cookie);
11851 /* All done if we just consumed the hdr_mp. */
11852 if (mp == hdr_mp) {
11853 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragOKs);
11854 return (error);
11855 }
11856 if (error != 0 && error != EWOULDBLOCK) {
11857 DTRACE_PROBE2(ip__xmit__frag__fail, ill_t *, ill,
11858 mblk_t *, hdr_mp);
11859 /* No point in sending the other fragments */
11860 break;
11861 }
11862
11863 /* Otherwise, advance and loop. */
11864 offset += len;
11865 }
11866 /* Clean up following allocation failure. */
11867 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11868 ip_drop_output("FragFails: loop ended", NULL, ill);
11869 if (mp != hdr_mp)
11870 freeb(hdr_mp);
11871 if (mp != mp_orig)
11872 freemsg(mp_orig);
11873 return (error);
11874 }
11875
11876 /*
11877 * Copy the header plus those options which have the copy bit set
11878 */
11879 static mblk_t *
11880 ip_fragment_copyhdr(uchar_t *rptr, int hdr_len, int offset, ip_stack_t *ipst,
11881 mblk_t *src)
11882 {
11883 mblk_t *mp;
11884 uchar_t *up;
11885
11886 /*
11887 * Quick check if we need to look for options without the copy bit
11888 * set
11889 */
11890 mp = allocb_tmpl(ipst->ips_ip_wroff_extra + hdr_len, src);
11891 if (!mp)
11892 return (mp);
11893 mp->b_rptr += ipst->ips_ip_wroff_extra;
11894 if (hdr_len == IP_SIMPLE_HDR_LENGTH || offset != 0) {
11895 bcopy(rptr, mp->b_rptr, hdr_len);
11896 mp->b_wptr += hdr_len + ipst->ips_ip_wroff_extra;
11897 return (mp);
11898 }
11899 up = mp->b_rptr;
11900 bcopy(rptr, up, IP_SIMPLE_HDR_LENGTH);
11901 up += IP_SIMPLE_HDR_LENGTH;
11902 rptr += IP_SIMPLE_HDR_LENGTH;
11903 hdr_len -= IP_SIMPLE_HDR_LENGTH;
11904 while (hdr_len > 0) {
11905 uint32_t optval;
11906 uint32_t optlen;
11907
11908 optval = *rptr;
11909 if (optval == IPOPT_EOL)
11910 break;
11911 if (optval == IPOPT_NOP)
11912 optlen = 1;
11913 else
11914 optlen = rptr[1];
11915 if (optval & IPOPT_COPY) {
11916 bcopy(rptr, up, optlen);
11917 up += optlen;
11918 }
11919 rptr += optlen;
11920 hdr_len -= optlen;
11921 }
11922 /*
11923 * Make sure that we drop an even number of words by filling
11924 * with EOL to the next word boundary.
11925 */
11926 for (hdr_len = up - (mp->b_rptr + IP_SIMPLE_HDR_LENGTH);
11927 hdr_len & 0x3; hdr_len++)
11928 *up++ = IPOPT_EOL;
11929 mp->b_wptr = up;
11930 /* Update header length */
11931 mp->b_rptr[0] = (uint8_t)((IP_VERSION << 4) | ((up - mp->b_rptr) >> 2));
11932 return (mp);
11933 }
11934
11935 /*
11936 * Update any source route, record route, or timestamp options when
11937 * sending a packet back to ourselves.
11938 * Check that we are at end of strict source route.
11939 * The options have been sanity checked by ip_output_options().
11940 */
11941 void
11942 ip_output_local_options(ipha_t *ipha, ip_stack_t *ipst)
11943 {
11944 ipoptp_t opts;
11945 uchar_t *opt;
11946 uint8_t optval;
11947 uint8_t optlen;
11948 ipaddr_t dst;
11949 uint32_t ts;
11950 timestruc_t now;
11951
11952 for (optval = ipoptp_first(&opts, ipha);
11953 optval != IPOPT_EOL;
11954 optval = ipoptp_next(&opts)) {
11955 opt = opts.ipoptp_cur;
11956 optlen = opts.ipoptp_len;
11957 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11958 switch (optval) {
11959 uint32_t off;
11960 case IPOPT_SSRR:
11961 case IPOPT_LSRR:
11962 off = opt[IPOPT_OFFSET];
11963 off--;
11964 if (optlen < IP_ADDR_LEN ||
11965 off > optlen - IP_ADDR_LEN) {
11966 /* End of source route */
11967 break;
11968 }
11969 /*
11970 * This will only happen if two consecutive entries
11971 * in the source route contains our address or if
11972 * it is a packet with a loose source route which
11973 * reaches us before consuming the whole source route
11974 */
11975
11976 if (optval == IPOPT_SSRR) {
11977 return;
11978 }
11979 /*
11980 * Hack: instead of dropping the packet truncate the
11981 * source route to what has been used by filling the
11982 * rest with IPOPT_NOP.
11983 */
11984 opt[IPOPT_OLEN] = (uint8_t)off;
11985 while (off < optlen) {
11986 opt[off++] = IPOPT_NOP;
11987 }
11988 break;
11989 case IPOPT_RR:
11990 off = opt[IPOPT_OFFSET];
11991 off--;
11992 if (optlen < IP_ADDR_LEN ||
11993 off > optlen - IP_ADDR_LEN) {
11994 /* No more room - ignore */
11995 ip1dbg((
11996 "ip_output_local_options: end of RR\n"));
11997 break;
11998 }
11999 dst = htonl(INADDR_LOOPBACK);
12000 bcopy(&dst, (char *)opt + off, IP_ADDR_LEN);
12001 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
12002 break;
12003 case IPOPT_TS:
12004 /* Insert timestamp if there is romm */
12005 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
12006 case IPOPT_TS_TSONLY:
12007 off = IPOPT_TS_TIMELEN;
12008 break;
12009 case IPOPT_TS_PRESPEC:
12010 case IPOPT_TS_PRESPEC_RFC791:
12011 /* Verify that the address matched */
12012 off = opt[IPOPT_OFFSET] - 1;
12013 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
12014 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
12015 /* Not for us */
12016 break;
12017 }
12018 /* FALLTHROUGH */
12019 case IPOPT_TS_TSANDADDR:
12020 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
12021 break;
12022 default:
12023 /*
12024 * ip_*put_options should have already
12025 * dropped this packet.
12026 */
12027 cmn_err(CE_PANIC, "ip_output_local_options: "
12028 "unknown IT - bug in ip_output_options?\n");
12029 return; /* Keep "lint" happy */
12030 }
12031 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
12032 /* Increase overflow counter */
12033 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
12034 opt[IPOPT_POS_OV_FLG] = (uint8_t)
12035 (opt[IPOPT_POS_OV_FLG] & 0x0F) |
12036 (off << 4);
12037 break;
12038 }
12039 off = opt[IPOPT_OFFSET] - 1;
12040 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
12041 case IPOPT_TS_PRESPEC:
12042 case IPOPT_TS_PRESPEC_RFC791:
12043 case IPOPT_TS_TSANDADDR:
12044 dst = htonl(INADDR_LOOPBACK);
12045 bcopy(&dst, (char *)opt + off, IP_ADDR_LEN);
12046 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
12047 /* FALLTHROUGH */
12048 case IPOPT_TS_TSONLY:
12049 off = opt[IPOPT_OFFSET] - 1;
12050 /* Compute # of milliseconds since midnight */
12051 gethrestime(&now);
12052 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
12053 NSEC2MSEC(now.tv_nsec);
12054 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
12055 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
12056 break;
12057 }
12058 break;
12059 }
12060 }
12061 }
12062
12063 /*
12064 * Prepend an M_DATA fastpath header, and if none present prepend a
12065 * DL_UNITDATA_REQ. Frees the mblk on failure.
12066 *
12067 * nce_dlur_mp and nce_fp_mp can not disappear once they have been set.
12068 * If there is a change to them, the nce will be deleted (condemned) and
12069 * a new nce_t will be created when packets are sent. Thus we need no locks
12070 * to access those fields.
12071 *
12072 * We preserve b_band to support IPQoS. If a DL_UNITDATA_REQ is prepended
12073 * we place b_band in dl_priority.dl_max.
12074 */
12075 static mblk_t *
12076 ip_xmit_attach_llhdr(mblk_t *mp, nce_t *nce)
12077 {
12078 uint_t hlen;
12079 mblk_t *mp1;
12080 uint_t priority;
12081 uchar_t *rptr;
12082
12083 rptr = mp->b_rptr;
12084
12085 ASSERT(DB_TYPE(mp) == M_DATA);
12086 priority = mp->b_band;
12087
12088 ASSERT(nce != NULL);
12089 if ((mp1 = nce->nce_fp_mp) != NULL) {
12090 hlen = MBLKL(mp1);
12091 /*
12092 * Check if we have enough room to prepend fastpath
12093 * header
12094 */
12095 if (hlen != 0 && (rptr - mp->b_datap->db_base) >= hlen) {
12096 rptr -= hlen;
12097 bcopy(mp1->b_rptr, rptr, hlen);
12098 /*
12099 * Set the b_rptr to the start of the link layer
12100 * header
12101 */
12102 mp->b_rptr = rptr;
12103 return (mp);
12104 }
12105 mp1 = copyb(mp1);
12106 if (mp1 == NULL) {
12107 ill_t *ill = nce->nce_ill;
12108
12109 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12110 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12111 freemsg(mp);
12112 return (NULL);
12113 }
12114 mp1->b_band = priority;
12115 mp1->b_cont = mp;
12116 DB_CKSUMSTART(mp1) = DB_CKSUMSTART(mp);
12117 DB_CKSUMSTUFF(mp1) = DB_CKSUMSTUFF(mp);
12118 DB_CKSUMEND(mp1) = DB_CKSUMEND(mp);
12119 DB_CKSUMFLAGS(mp1) = DB_CKSUMFLAGS(mp);
12120 DB_LSOMSS(mp1) = DB_LSOMSS(mp);
12121 DTRACE_PROBE1(ip__xmit__copyb, (mblk_t *), mp1);
12122 /*
12123 * XXX disable ICK_VALID and compute checksum
12124 * here; can happen if nce_fp_mp changes and
12125 * it can't be copied now due to insufficient
12126 * space. (unlikely, fp mp can change, but it
12127 * does not increase in length)
12128 */
12129 return (mp1);
12130 }
12131 mp1 = copyb(nce->nce_dlur_mp);
12132
12133 if (mp1 == NULL) {
12134 ill_t *ill = nce->nce_ill;
12135
12136 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12137 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12138 freemsg(mp);
12139 return (NULL);
12140 }
12141 mp1->b_cont = mp;
12142 if (priority != 0) {
12143 mp1->b_band = priority;
12144 ((dl_unitdata_req_t *)(mp1->b_rptr))->dl_priority.dl_max =
12145 priority;
12146 }
12147 return (mp1);
12148 }
12149
12150 /*
12151 * Finish the outbound IPsec processing. This function is called from
12152 * ipsec_out_process() if the IPsec packet was processed
12153 * synchronously, or from {ah,esp}_kcf_callback_outbound() if it was processed
12154 * asynchronously.
12155 *
12156 * This is common to IPv4 and IPv6.
12157 */
12158 int
12159 ip_output_post_ipsec(mblk_t *mp, ip_xmit_attr_t *ixa)
12160 {
12161 iaflags_t ixaflags = ixa->ixa_flags;
12162 uint_t pktlen;
12163
12164
12165 /* AH/ESP don't update ixa_pktlen when they modify the packet */
12166 if (ixaflags & IXAF_IS_IPV4) {
12167 ipha_t *ipha = (ipha_t *)mp->b_rptr;
12168
12169 ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
12170 pktlen = ntohs(ipha->ipha_length);
12171 } else {
12172 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
12173
12174 ASSERT(IPH_HDR_VERSION(mp->b_rptr) == IPV6_VERSION);
12175 pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
12176 }
12177
12178 /*
12179 * We release any hard reference on the SAs here to make
12180 * sure the SAs can be garbage collected. ipsr_sa has a soft reference
12181 * on the SAs.
12182 * If in the future we want the hard latching of the SAs in the
12183 * ip_xmit_attr_t then we should remove this.
12184 */
12185 if (ixa->ixa_ipsec_esp_sa != NULL) {
12186 IPSA_REFRELE(ixa->ixa_ipsec_esp_sa);
12187 ixa->ixa_ipsec_esp_sa = NULL;
12188 }
12189 if (ixa->ixa_ipsec_ah_sa != NULL) {
12190 IPSA_REFRELE(ixa->ixa_ipsec_ah_sa);
12191 ixa->ixa_ipsec_ah_sa = NULL;
12192 }
12193
12194 /* Do we need to fragment? */
12195 if ((ixa->ixa_flags & IXAF_IPV6_ADD_FRAGHDR) ||
12196 pktlen > ixa->ixa_fragsize) {
12197 if (ixaflags & IXAF_IS_IPV4) {
12198 ASSERT(!(ixa->ixa_flags & IXAF_IPV6_ADD_FRAGHDR));
12199 /*
12200 * We check for the DF case in ipsec_out_process
12201 * hence this only handles the non-DF case.
12202 */
12203 return (ip_fragment_v4(mp, ixa->ixa_nce, ixa->ixa_flags,
12204 pktlen, ixa->ixa_fragsize,
12205 ixa->ixa_xmit_hint, ixa->ixa_zoneid,
12206 ixa->ixa_no_loop_zoneid, ixa->ixa_postfragfn,
12207 &ixa->ixa_cookie));
12208 } else {
12209 mp = ip_fraghdr_add_v6(mp, ixa->ixa_ident, ixa);
12210 if (mp == NULL) {
12211 /* MIB and ip_drop_output already done */
12212 return (ENOMEM);
12213 }
12214 pktlen += sizeof (ip6_frag_t);
12215 if (pktlen > ixa->ixa_fragsize) {
12216 return (ip_fragment_v6(mp, ixa->ixa_nce,
12217 ixa->ixa_flags, pktlen,
12218 ixa->ixa_fragsize, ixa->ixa_xmit_hint,
12219 ixa->ixa_zoneid, ixa->ixa_no_loop_zoneid,
12220 ixa->ixa_postfragfn, &ixa->ixa_cookie));
12221 }
12222 }
12223 }
12224 return ((ixa->ixa_postfragfn)(mp, ixa->ixa_nce, ixa->ixa_flags,
12225 pktlen, ixa->ixa_xmit_hint, ixa->ixa_zoneid,
12226 ixa->ixa_no_loop_zoneid, NULL));
12227 }
12228
12229 /*
12230 * Finish the inbound IPsec processing. This function is called from
12231 * ipsec_out_process() if the IPsec packet was processed
12232 * synchronously, or from {ah,esp}_kcf_callback_outbound() if it was processed
12233 * asynchronously.
12234 *
12235 * This is common to IPv4 and IPv6.
12236 */
12237 void
12238 ip_input_post_ipsec(mblk_t *mp, ip_recv_attr_t *ira)
12239 {
12240 iaflags_t iraflags = ira->ira_flags;
12241
12242 /* Length might have changed */
12243 if (iraflags & IRAF_IS_IPV4) {
12244 ipha_t *ipha = (ipha_t *)mp->b_rptr;
12245
12246 ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
12247 ira->ira_pktlen = ntohs(ipha->ipha_length);
12248 ira->ira_ip_hdr_length = IPH_HDR_LENGTH(ipha);
12249 ira->ira_protocol = ipha->ipha_protocol;
12250
12251 ip_fanout_v4(mp, ipha, ira);
12252 } else {
12253 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
12254 uint8_t *nexthdrp;
12255
12256 ASSERT(IPH_HDR_VERSION(mp->b_rptr) == IPV6_VERSION);
12257 ira->ira_pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
12258 if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &ira->ira_ip_hdr_length,
12259 &nexthdrp)) {
12260 /* Malformed packet */
12261 BUMP_MIB(ira->ira_ill->ill_ip_mib, ipIfStatsInDiscards);
12262 ip_drop_input("ipIfStatsInDiscards", mp, ira->ira_ill);
12263 freemsg(mp);
12264 return;
12265 }
12266 ira->ira_protocol = *nexthdrp;
12267 ip_fanout_v6(mp, ip6h, ira);
12268 }
12269 }
12270
12271 /*
12272 * Select which AH & ESP SA's to use (if any) for the outbound packet.
12273 *
12274 * If this function returns B_TRUE, the requested SA's have been filled
12275 * into the ixa_ipsec_*_sa pointers.
12276 *
12277 * If the function returns B_FALSE, the packet has been "consumed", most
12278 * likely by an ACQUIRE sent up via PF_KEY to a key management daemon.
12279 *
12280 * The SA references created by the protocol-specific "select"
12281 * function will be released in ip_output_post_ipsec.
12282 */
12283 static boolean_t
12284 ipsec_out_select_sa(mblk_t *mp, ip_xmit_attr_t *ixa)
12285 {
12286 boolean_t need_ah_acquire = B_FALSE, need_esp_acquire = B_FALSE;
12287 ipsec_policy_t *pp;
12288 ipsec_action_t *ap;
12289
12290 ASSERT(ixa->ixa_flags & IXAF_IPSEC_SECURE);
12291 ASSERT((ixa->ixa_ipsec_policy != NULL) ||
12292 (ixa->ixa_ipsec_action != NULL));
12293
12294 ap = ixa->ixa_ipsec_action;
12295 if (ap == NULL) {
12296 pp = ixa->ixa_ipsec_policy;
12297 ASSERT(pp != NULL);
12298 ap = pp->ipsp_act;
12299 ASSERT(ap != NULL);
12300 }
12301
12302 /*
12303 * We have an action. now, let's select SA's.
12304 * A side effect of setting ixa_ipsec_*_sa is that it will
12305 * be cached in the conn_t.
12306 */
12307 if (ap->ipa_want_esp) {
12308 if (ixa->ixa_ipsec_esp_sa == NULL) {
12309 need_esp_acquire = !ipsec_outbound_sa(mp, ixa,
12310 IPPROTO_ESP);
12311 }
12312 ASSERT(need_esp_acquire || ixa->ixa_ipsec_esp_sa != NULL);
12313 }
12314
12315 if (ap->ipa_want_ah) {
12316 if (ixa->ixa_ipsec_ah_sa == NULL) {
12317 need_ah_acquire = !ipsec_outbound_sa(mp, ixa,
12318 IPPROTO_AH);
12319 }
12320 ASSERT(need_ah_acquire || ixa->ixa_ipsec_ah_sa != NULL);
12321 /*
12322 * The ESP and AH processing order needs to be preserved
12323 * when both protocols are required (ESP should be applied
12324 * before AH for an outbound packet). Force an ESP ACQUIRE
12325 * when both ESP and AH are required, and an AH ACQUIRE
12326 * is needed.
12327 */
12328 if (ap->ipa_want_esp && need_ah_acquire)
12329 need_esp_acquire = B_TRUE;
12330 }
12331
12332 /*
12333 * Send an ACQUIRE (extended, regular, or both) if we need one.
12334 * Release SAs that got referenced, but will not be used until we
12335 * acquire _all_ of the SAs we need.
12336 */
12337 if (need_ah_acquire || need_esp_acquire) {
12338 if (ixa->ixa_ipsec_ah_sa != NULL) {
12339 IPSA_REFRELE(ixa->ixa_ipsec_ah_sa);
12340 ixa->ixa_ipsec_ah_sa = NULL;
12341 }
12342 if (ixa->ixa_ipsec_esp_sa != NULL) {
12343 IPSA_REFRELE(ixa->ixa_ipsec_esp_sa);
12344 ixa->ixa_ipsec_esp_sa = NULL;
12345 }
12346
12347 sadb_acquire(mp, ixa, need_ah_acquire, need_esp_acquire);
12348 return (B_FALSE);
12349 }
12350
12351 return (B_TRUE);
12352 }
12353
12354 /*
12355 * Handle IPsec output processing.
12356 * This function is only entered once for a given packet.
12357 * We try to do things synchronously, but if we need to have user-level
12358 * set up SAs, or ESP or AH uses asynchronous kEF, then the operation
12359 * will be completed
12360 * - when the SAs are added in esp_add_sa_finish/ah_add_sa_finish
12361 * - when asynchronous ESP is done it will do AH
12362 *
12363 * In all cases we come back in ip_output_post_ipsec() to fragment and
12364 * send out the packet.
12365 */
12366 int
12367 ipsec_out_process(mblk_t *mp, ip_xmit_attr_t *ixa)
12368 {
12369 ill_t *ill = ixa->ixa_nce->nce_ill;
12370 ip_stack_t *ipst = ixa->ixa_ipst;
12371 ipsec_stack_t *ipss;
12372 ipsec_policy_t *pp;
12373 ipsec_action_t *ap;
12374
12375 ASSERT(ixa->ixa_flags & IXAF_IPSEC_SECURE);
12376
12377 ASSERT((ixa->ixa_ipsec_policy != NULL) ||
12378 (ixa->ixa_ipsec_action != NULL));
12379
12380 ipss = ipst->ips_netstack->netstack_ipsec;
12381 if (!ipsec_loaded(ipss)) {
12382 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12383 ip_drop_packet(mp, B_TRUE, ill,
12384 DROPPER(ipss, ipds_ip_ipsec_not_loaded),
12385 &ipss->ipsec_dropper);
12386 return (ENOTSUP);
12387 }
12388
12389 ap = ixa->ixa_ipsec_action;
12390 if (ap == NULL) {
12391 pp = ixa->ixa_ipsec_policy;
12392 ASSERT(pp != NULL);
12393 ap = pp->ipsp_act;
12394 ASSERT(ap != NULL);
12395 }
12396
12397 /* Handle explicit drop action and bypass. */
12398 switch (ap->ipa_act.ipa_type) {
12399 case IPSEC_ACT_DISCARD:
12400 case IPSEC_ACT_REJECT:
12401 ip_drop_packet(mp, B_FALSE, ill,
12402 DROPPER(ipss, ipds_spd_explicit), &ipss->ipsec_spd_dropper);
12403 return (EHOSTUNREACH); /* IPsec policy failure */
12404 case IPSEC_ACT_BYPASS:
12405 return (ip_output_post_ipsec(mp, ixa));
12406 }
12407
12408 /*
12409 * The order of processing is first insert a IP header if needed.
12410 * Then insert the ESP header and then the AH header.
12411 */
12412 if ((ixa->ixa_flags & IXAF_IS_IPV4) && ap->ipa_want_se) {
12413 /*
12414 * First get the outer IP header before sending
12415 * it to ESP.
12416 */
12417 ipha_t *oipha, *iipha;
12418 mblk_t *outer_mp, *inner_mp;
12419
12420 if ((outer_mp = allocb(sizeof (ipha_t), BPRI_HI)) == NULL) {
12421 (void) mi_strlog(ill->ill_rq, 0,
12422 SL_ERROR|SL_TRACE|SL_CONSOLE,
12423 "ipsec_out_process: "
12424 "Self-Encapsulation failed: Out of memory\n");
12425 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12426 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12427 freemsg(mp);
12428 return (ENOBUFS);
12429 }
12430 inner_mp = mp;
12431 ASSERT(inner_mp->b_datap->db_type == M_DATA);
12432 oipha = (ipha_t *)outer_mp->b_rptr;
12433 iipha = (ipha_t *)inner_mp->b_rptr;
12434 *oipha = *iipha;
12435 outer_mp->b_wptr += sizeof (ipha_t);
12436 oipha->ipha_length = htons(ntohs(iipha->ipha_length) +
12437 sizeof (ipha_t));
12438 oipha->ipha_protocol = IPPROTO_ENCAP;
12439 oipha->ipha_version_and_hdr_length =
12440 IP_SIMPLE_HDR_VERSION;
12441 oipha->ipha_hdr_checksum = 0;
12442 oipha->ipha_hdr_checksum = ip_csum_hdr(oipha);
12443 outer_mp->b_cont = inner_mp;
12444 mp = outer_mp;
12445
12446 ixa->ixa_flags |= IXAF_IPSEC_TUNNEL;
12447 }
12448
12449 /* If we need to wait for a SA then we can't return any errno */
12450 if (((ap->ipa_want_ah && (ixa->ixa_ipsec_ah_sa == NULL)) ||
12451 (ap->ipa_want_esp && (ixa->ixa_ipsec_esp_sa == NULL))) &&
12452 !ipsec_out_select_sa(mp, ixa))
12453 return (0);
12454
12455 /*
12456 * By now, we know what SA's to use. Toss over to ESP & AH
12457 * to do the heavy lifting.
12458 */
12459 if (ap->ipa_want_esp) {
12460 ASSERT(ixa->ixa_ipsec_esp_sa != NULL);
12461
12462 mp = ixa->ixa_ipsec_esp_sa->ipsa_output_func(mp, ixa);
12463 if (mp == NULL) {
12464 /*
12465 * Either it failed or is pending. In the former case
12466 * ipIfStatsInDiscards was increased.
12467 */
12468 return (0);
12469 }
12470 }
12471
12472 if (ap->ipa_want_ah) {
12473 ASSERT(ixa->ixa_ipsec_ah_sa != NULL);
12474
12475 mp = ixa->ixa_ipsec_ah_sa->ipsa_output_func(mp, ixa);
12476 if (mp == NULL) {
12477 /*
12478 * Either it failed or is pending. In the former case
12479 * ipIfStatsInDiscards was increased.
12480 */
12481 return (0);
12482 }
12483 }
12484 /*
12485 * We are done with IPsec processing. Send it over
12486 * the wire.
12487 */
12488 return (ip_output_post_ipsec(mp, ixa));
12489 }
12490
12491 /*
12492 * ioctls that go through a down/up sequence may need to wait for the down
12493 * to complete. This involves waiting for the ire and ipif refcnts to go down
12494 * to zero. Subsequently the ioctl is restarted from ipif_ill_refrele_tail.
12495 */
12496 /* ARGSUSED */
12497 void
12498 ip_reprocess_ioctl(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
12499 {
12500 struct iocblk *iocp;
12501 mblk_t *mp1;
12502 ip_ioctl_cmd_t *ipip;
12503 int err;
12504 sin_t *sin;
12505 struct lifreq *lifr;
12506 struct ifreq *ifr;
12507
12508 iocp = (struct iocblk *)mp->b_rptr;
12509 ASSERT(ipsq != NULL);
12510 /* Existence of mp1 verified in ip_wput_nondata */
12511 mp1 = mp->b_cont->b_cont;
12512 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12513 if (ipip->ipi_cmd == SIOCSLIFNAME || ipip->ipi_cmd == IF_UNITSEL) {
12514 /*
12515 * Special case where ipx_current_ipif is not set:
12516 * ill_phyint_reinit merged the v4 and v6 into a single ipsq.
12517 * We are here as were not able to complete the operation in
12518 * ipif_set_values because we could not become exclusive on
12519 * the new ipsq.
12520 */
12521 ill_t *ill = q->q_ptr;
12522 ipsq_current_start(ipsq, ill->ill_ipif, ipip->ipi_cmd);
12523 }
12524 ASSERT(ipsq->ipsq_xop->ipx_current_ipif != NULL);
12525
12526 if (ipip->ipi_cmd_type == IF_CMD) {
12527 /* This a old style SIOC[GS]IF* command */
12528 ifr = (struct ifreq *)mp1->b_rptr;
12529 sin = (sin_t *)&ifr->ifr_addr;
12530 } else if (ipip->ipi_cmd_type == LIF_CMD) {
12531 /* This a new style SIOC[GS]LIF* command */
12532 lifr = (struct lifreq *)mp1->b_rptr;
12533 sin = (sin_t *)&lifr->lifr_addr;
12534 } else {
12535 sin = NULL;
12536 }
12537
12538 err = (*ipip->ipi_func_restart)(ipsq->ipsq_xop->ipx_current_ipif, sin,
12539 q, mp, ipip, mp1->b_rptr);
12540
12541 DTRACE_PROBE4(ipif__ioctl, char *, "ip_reprocess_ioctl finish",
12542 int, ipip->ipi_cmd,
12543 ill_t *, ipsq->ipsq_xop->ipx_current_ipif->ipif_ill,
12544 ipif_t *, ipsq->ipsq_xop->ipx_current_ipif);
12545
12546 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), ipsq);
12547 }
12548
12549 /*
12550 * ioctl processing
12551 *
12552 * ioctl processing starts with ip_sioctl_copyin_setup(), which looks up
12553 * the ioctl command in the ioctl tables, determines the copyin data size
12554 * from the ipi_copyin_size field, and does an mi_copyin() of that size.
12555 *
12556 * ioctl processing then continues when the M_IOCDATA makes its way down to
12557 * ip_wput_nondata(). The ioctl is looked up again in the ioctl table, its
12558 * associated 'conn' is refheld till the end of the ioctl and the general
12559 * ioctl processing function ip_process_ioctl() is called to extract the
12560 * arguments and process the ioctl. To simplify extraction, ioctl commands
12561 * are "typed" based on the arguments they take (e.g., LIF_CMD which takes a
12562 * `struct lifreq'), and a common extract function (e.g., ip_extract_lifreq())
12563 * is used to extract the ioctl's arguments.
12564 *
12565 * ip_process_ioctl determines if the ioctl needs to be serialized, and if
12566 * so goes thru the serialization primitive ipsq_try_enter. Then the
12567 * appropriate function to handle the ioctl is called based on the entry in
12568 * the ioctl table. ioctl completion is encapsulated in ip_ioctl_finish
12569 * which also refreleases the 'conn' that was refheld at the start of the
12570 * ioctl. Finally ipsq_exit is called if needed to exit the ipsq.
12571 *
12572 * Many exclusive ioctls go thru an internal down up sequence as part of
12573 * the operation. For example an attempt to change the IP address of an
12574 * ipif entails ipif_down, set address, ipif_up. Bringing down the interface
12575 * does all the cleanup such as deleting all ires that use this address.
12576 * Then we need to wait till all references to the interface go away.
12577 */
12578 void
12579 ip_process_ioctl(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *arg)
12580 {
12581 struct iocblk *iocp = (struct iocblk *)mp->b_rptr;
12582 ip_ioctl_cmd_t *ipip = arg;
12583 ip_extract_func_t *extract_funcp;
12584 cmd_info_t ci;
12585 int err;
12586 boolean_t entered_ipsq = B_FALSE;
12587
12588 ip3dbg(("ip_process_ioctl: ioctl %X\n", iocp->ioc_cmd));
12589
12590 if (ipip == NULL)
12591 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12592
12593 /*
12594 * SIOCLIFADDIF needs to go thru a special path since the
12595 * ill may not exist yet. This happens in the case of lo0
12596 * which is created using this ioctl.
12597 */
12598 if (ipip->ipi_cmd == SIOCLIFADDIF) {
12599 err = ip_sioctl_addif(NULL, NULL, q, mp, NULL, NULL);
12600 DTRACE_PROBE4(ipif__ioctl, char *, "ip_process_ioctl finish",
12601 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12602 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12603 return;
12604 }
12605
12606 ci.ci_ipif = NULL;
12607 switch (ipip->ipi_cmd_type) {
12608 case MISC_CMD:
12609 case MSFILT_CMD:
12610 /*
12611 * All MISC_CMD ioctls come in here -- e.g. SIOCGLIFCONF.
12612 */
12613 if (ipip->ipi_cmd == IF_UNITSEL) {
12614 /* ioctl comes down the ill */
12615 ci.ci_ipif = ((ill_t *)q->q_ptr)->ill_ipif;
12616 ipif_refhold(ci.ci_ipif);
12617 }
12618 err = 0;
12619 ci.ci_sin = NULL;
12620 ci.ci_sin6 = NULL;
12621 ci.ci_lifr = NULL;
12622 extract_funcp = NULL;
12623 break;
12624
12625 case IF_CMD:
12626 case LIF_CMD:
12627 extract_funcp = ip_extract_lifreq;
12628 break;
12629
12630 case ARP_CMD:
12631 case XARP_CMD:
12632 extract_funcp = ip_extract_arpreq;
12633 break;
12634
12635 default:
12636 ASSERT(0);
12637 }
12638
12639 if (extract_funcp != NULL) {
12640 err = (*extract_funcp)(q, mp, ipip, &ci);
12641 if (err != 0) {
12642 DTRACE_PROBE4(ipif__ioctl,
12643 char *, "ip_process_ioctl finish err",
12644 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12645 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12646 return;
12647 }
12648
12649 /*
12650 * All of the extraction functions return a refheld ipif.
12651 */
12652 ASSERT(ci.ci_ipif != NULL);
12653 }
12654
12655 if (!(ipip->ipi_flags & IPI_WR)) {
12656 /*
12657 * A return value of EINPROGRESS means the ioctl is
12658 * either queued and waiting for some reason or has
12659 * already completed.
12660 */
12661 err = (*ipip->ipi_func)(ci.ci_ipif, ci.ci_sin, q, mp, ipip,
12662 ci.ci_lifr);
12663 if (ci.ci_ipif != NULL) {
12664 DTRACE_PROBE4(ipif__ioctl,
12665 char *, "ip_process_ioctl finish RD",
12666 int, ipip->ipi_cmd, ill_t *, ci.ci_ipif->ipif_ill,
12667 ipif_t *, ci.ci_ipif);
12668 ipif_refrele(ci.ci_ipif);
12669 } else {
12670 DTRACE_PROBE4(ipif__ioctl,
12671 char *, "ip_process_ioctl finish RD",
12672 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12673 }
12674 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12675 return;
12676 }
12677
12678 ASSERT(ci.ci_ipif != NULL);
12679
12680 /*
12681 * If ipsq is non-NULL, we are already being called exclusively
12682 */
12683 ASSERT(ipsq == NULL || IAM_WRITER_IPSQ(ipsq));
12684 if (ipsq == NULL) {
12685 ipsq = ipsq_try_enter(ci.ci_ipif, NULL, q, mp, ip_process_ioctl,
12686 NEW_OP, B_TRUE);
12687 if (ipsq == NULL) {
12688 ipif_refrele(ci.ci_ipif);
12689 return;
12690 }
12691 entered_ipsq = B_TRUE;
12692 }
12693 /*
12694 * Release the ipif so that ipif_down and friends that wait for
12695 * references to go away are not misled about the current ipif_refcnt
12696 * values. We are writer so we can access the ipif even after releasing
12697 * the ipif.
12698 */
12699 ipif_refrele(ci.ci_ipif);
12700
12701 ipsq_current_start(ipsq, ci.ci_ipif, ipip->ipi_cmd);
12702
12703 /*
12704 * A return value of EINPROGRESS means the ioctl is
12705 * either queued and waiting for some reason or has
12706 * already completed.
12707 */
12708 err = (*ipip->ipi_func)(ci.ci_ipif, ci.ci_sin, q, mp, ipip, ci.ci_lifr);
12709
12710 DTRACE_PROBE4(ipif__ioctl, char *, "ip_process_ioctl finish WR",
12711 int, ipip->ipi_cmd,
12712 ill_t *, ci.ci_ipif == NULL ? NULL : ci.ci_ipif->ipif_ill,
12713 ipif_t *, ci.ci_ipif);
12714 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), ipsq);
12715
12716 if (entered_ipsq)
12717 ipsq_exit(ipsq);
12718 }
12719
12720 /*
12721 * Complete the ioctl. Typically ioctls use the mi package and need to
12722 * do mi_copyout/mi_copy_done.
12723 */
12724 void
12725 ip_ioctl_finish(queue_t *q, mblk_t *mp, int err, int mode, ipsq_t *ipsq)
12726 {
12727 conn_t *connp = NULL;
12728
12729 if (err == EINPROGRESS)
12730 return;
12731
12732 if (CONN_Q(q)) {
12733 connp = Q_TO_CONN(q);
12734 ASSERT(connp->conn_ref >= 2);
12735 }
12736
12737 switch (mode) {
12738 case COPYOUT:
12739 if (err == 0)
12740 mi_copyout(q, mp);
12741 else
12742 mi_copy_done(q, mp, err);
12743 break;
12744
12745 case NO_COPYOUT:
12746 mi_copy_done(q, mp, err);
12747 break;
12748
12749 default:
12750 ASSERT(mode == CONN_CLOSE); /* aborted through CONN_CLOSE */
12751 break;
12752 }
12753
12754 /*
12755 * The conn refhold and ioctlref placed on the conn at the start of the
12756 * ioctl are released here.
12757 */
12758 if (connp != NULL) {
12759 CONN_DEC_IOCTLREF(connp);
12760 CONN_OPER_PENDING_DONE(connp);
12761 }
12762
12763 if (ipsq != NULL)
12764 ipsq_current_finish(ipsq);
12765 }
12766
12767 /* Handles all non data messages */
12768 int
12769 ip_wput_nondata(queue_t *q, mblk_t *mp)
12770 {
12771 mblk_t *mp1;
12772 struct iocblk *iocp;
12773 ip_ioctl_cmd_t *ipip;
12774 conn_t *connp;
12775 cred_t *cr;
12776 char *proto_str;
12777
12778 if (CONN_Q(q))
12779 connp = Q_TO_CONN(q);
12780 else
12781 connp = NULL;
12782
12783 switch (DB_TYPE(mp)) {
12784 case M_IOCTL:
12785 /*
12786 * IOCTL processing begins in ip_sioctl_copyin_setup which
12787 * will arrange to copy in associated control structures.
12788 */
12789 ip_sioctl_copyin_setup(q, mp);
12790 return (0);
12791 case M_IOCDATA:
12792 /*
12793 * Ensure that this is associated with one of our trans-
12794 * parent ioctls. If it's not ours, discard it if we're
12795 * running as a driver, or pass it on if we're a module.
12796 */
12797 iocp = (struct iocblk *)mp->b_rptr;
12798 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12799 if (ipip == NULL) {
12800 if (q->q_next == NULL) {
12801 goto nak;
12802 } else {
12803 putnext(q, mp);
12804 }
12805 return (0);
12806 }
12807 if ((q->q_next != NULL) && !(ipip->ipi_flags & IPI_MODOK)) {
12808 /*
12809 * The ioctl is one we recognise, but is not consumed
12810 * by IP as a module and we are a module, so we drop
12811 */
12812 goto nak;
12813 }
12814
12815 /* IOCTL continuation following copyin or copyout. */
12816 if (mi_copy_state(q, mp, NULL) == -1) {
12817 /*
12818 * The copy operation failed. mi_copy_state already
12819 * cleaned up, so we're out of here.
12820 */
12821 return (0);
12822 }
12823 /*
12824 * If we just completed a copy in, we become writer and
12825 * continue processing in ip_sioctl_copyin_done. If it
12826 * was a copy out, we call mi_copyout again. If there is
12827 * nothing more to copy out, it will complete the IOCTL.
12828 */
12829 if (MI_COPY_DIRECTION(mp) == MI_COPY_IN) {
12830 if (!(mp1 = mp->b_cont) || !(mp1 = mp1->b_cont)) {
12831 mi_copy_done(q, mp, EPROTO);
12832 return (0);
12833 }
12834 /*
12835 * Check for cases that need more copying. A return
12836 * value of 0 means a second copyin has been started,
12837 * so we return; a return value of 1 means no more
12838 * copying is needed, so we continue.
12839 */
12840 if (ipip->ipi_cmd_type == MSFILT_CMD &&
12841 MI_COPY_COUNT(mp) == 1) {
12842 if (ip_copyin_msfilter(q, mp) == 0)
12843 return (0);
12844 }
12845 /*
12846 * Refhold the conn, till the ioctl completes. This is
12847 * needed in case the ioctl ends up in the pending mp
12848 * list. Every mp in the ipx_pending_mp list must have
12849 * a refhold on the conn to resume processing. The
12850 * refhold is released when the ioctl completes
12851 * (whether normally or abnormally). An ioctlref is also
12852 * placed on the conn to prevent TCP from removing the
12853 * queue needed to send the ioctl reply back.
12854 * In all cases ip_ioctl_finish is called to finish
12855 * the ioctl and release the refholds.
12856 */
12857 if (connp != NULL) {
12858 /* This is not a reentry */
12859 CONN_INC_REF(connp);
12860 CONN_INC_IOCTLREF(connp);
12861 } else {
12862 if (!(ipip->ipi_flags & IPI_MODOK)) {
12863 mi_copy_done(q, mp, EINVAL);
12864 return (0);
12865 }
12866 }
12867
12868 ip_process_ioctl(NULL, q, mp, ipip);
12869
12870 } else {
12871 mi_copyout(q, mp);
12872 }
12873 return (0);
12874
12875 case M_IOCNAK:
12876 /*
12877 * The only way we could get here is if a resolver didn't like
12878 * an IOCTL we sent it. This shouldn't happen.
12879 */
12880 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
12881 "ip_wput_nondata: unexpected M_IOCNAK, ioc_cmd 0x%x",
12882 ((struct iocblk *)mp->b_rptr)->ioc_cmd);
12883 freemsg(mp);
12884 return (0);
12885 case M_IOCACK:
12886 /* /dev/ip shouldn't see this */
12887 goto nak;
12888 case M_FLUSH:
12889 if (*mp->b_rptr & FLUSHW)
12890 flushq(q, FLUSHALL);
12891 if (q->q_next) {
12892 putnext(q, mp);
12893 return (0);
12894 }
12895 if (*mp->b_rptr & FLUSHR) {
12896 *mp->b_rptr &= ~FLUSHW;
12897 qreply(q, mp);
12898 return (0);
12899 }
12900 freemsg(mp);
12901 return (0);
12902 case M_CTL:
12903 break;
12904 case M_PROTO:
12905 case M_PCPROTO:
12906 /*
12907 * The only PROTO messages we expect are SNMP-related.
12908 */
12909 switch (((union T_primitives *)mp->b_rptr)->type) {
12910 case T_SVR4_OPTMGMT_REQ:
12911 ip2dbg(("ip_wput_nondata: T_SVR4_OPTMGMT_REQ "
12912 "flags %x\n",
12913 ((struct T_optmgmt_req *)mp->b_rptr)->MGMT_flags));
12914
12915 if (connp == NULL) {
12916 proto_str = "T_SVR4_OPTMGMT_REQ";
12917 goto protonak;
12918 }
12919
12920 /*
12921 * All Solaris components should pass a db_credp
12922 * for this TPI message, hence we ASSERT.
12923 * But in case there is some other M_PROTO that looks
12924 * like a TPI message sent by some other kernel
12925 * component, we check and return an error.
12926 */
12927 cr = msg_getcred(mp, NULL);
12928 ASSERT(cr != NULL);
12929 if (cr == NULL) {
12930 mp = mi_tpi_err_ack_alloc(mp, TSYSERR, EINVAL);
12931 if (mp != NULL)
12932 qreply(q, mp);
12933 return (0);
12934 }
12935
12936 if (!snmpcom_req(q, mp, ip_snmp_set, ip_snmp_get, cr)) {
12937 proto_str = "Bad SNMPCOM request?";
12938 goto protonak;
12939 }
12940 return (0);
12941 default:
12942 ip1dbg(("ip_wput_nondata: dropping M_PROTO prim %u\n",
12943 (int)*(uint_t *)mp->b_rptr));
12944 freemsg(mp);
12945 return (0);
12946 }
12947 default:
12948 break;
12949 }
12950 if (q->q_next) {
12951 putnext(q, mp);
12952 } else
12953 freemsg(mp);
12954 return (0);
12955
12956 nak:
12957 iocp->ioc_error = EINVAL;
12958 mp->b_datap->db_type = M_IOCNAK;
12959 iocp->ioc_count = 0;
12960 qreply(q, mp);
12961 return (0);
12962
12963 protonak:
12964 cmn_err(CE_NOTE, "IP doesn't process %s as a module", proto_str);
12965 if ((mp = mi_tpi_err_ack_alloc(mp, TPROTO, EINVAL)) != NULL)
12966 qreply(q, mp);
12967 return (0);
12968 }
12969
12970 /*
12971 * Process IP options in an outbound packet. Verify that the nexthop in a
12972 * strict source route is onlink.
12973 * Returns non-zero if something fails in which case an ICMP error has been
12974 * sent and mp freed.
12975 *
12976 * Assumes the ULP has called ip_massage_options to move nexthop into ipha_dst.
12977 */
12978 int
12979 ip_output_options(mblk_t *mp, ipha_t *ipha, ip_xmit_attr_t *ixa, ill_t *ill)
12980 {
12981 ipoptp_t opts;
12982 uchar_t *opt;
12983 uint8_t optval;
12984 uint8_t optlen;
12985 ipaddr_t dst;
12986 intptr_t code = 0;
12987 ire_t *ire;
12988 ip_stack_t *ipst = ixa->ixa_ipst;
12989 ip_recv_attr_t iras;
12990
12991 ip2dbg(("ip_output_options\n"));
12992
12993 dst = ipha->ipha_dst;
12994 for (optval = ipoptp_first(&opts, ipha);
12995 optval != IPOPT_EOL;
12996 optval = ipoptp_next(&opts)) {
12997 opt = opts.ipoptp_cur;
12998 optlen = opts.ipoptp_len;
12999 ip2dbg(("ip_output_options: opt %d, len %d\n",
13000 optval, optlen));
13001 switch (optval) {
13002 uint32_t off;
13003 case IPOPT_SSRR:
13004 case IPOPT_LSRR:
13005 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
13006 ip1dbg((
13007 "ip_output_options: bad option offset\n"));
13008 code = (char *)&opt[IPOPT_OLEN] -
13009 (char *)ipha;
13010 goto param_prob;
13011 }
13012 off = opt[IPOPT_OFFSET];
13013 ip1dbg(("ip_output_options: next hop 0x%x\n",
13014 ntohl(dst)));
13015 /*
13016 * For strict: verify that dst is directly
13017 * reachable.
13018 */
13019 if (optval == IPOPT_SSRR) {
13020 ire = ire_ftable_lookup_v4(dst, 0, 0,
13021 IRE_INTERFACE, NULL, ALL_ZONES,
13022 ixa->ixa_tsl,
13023 MATCH_IRE_TYPE | MATCH_IRE_SECATTR, 0, ipst,
13024 NULL);
13025 if (ire == NULL) {
13026 ip1dbg(("ip_output_options: SSRR not"
13027 " directly reachable: 0x%x\n",
13028 ntohl(dst)));
13029 goto bad_src_route;
13030 }
13031 ire_refrele(ire);
13032 }
13033 break;
13034 case IPOPT_RR:
13035 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
13036 ip1dbg((
13037 "ip_output_options: bad option offset\n"));
13038 code = (char *)&opt[IPOPT_OLEN] -
13039 (char *)ipha;
13040 goto param_prob;
13041 }
13042 break;
13043 case IPOPT_TS:
13044 /*
13045 * Verify that length >=5 and that there is either
13046 * room for another timestamp or that the overflow
13047 * counter is not maxed out.
13048 */
13049 code = (char *)&opt[IPOPT_OLEN] - (char *)ipha;
13050 if (optlen < IPOPT_MINLEN_IT) {
13051 goto param_prob;
13052 }
13053 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
13054 ip1dbg((
13055 "ip_output_options: bad option offset\n"));
13056 code = (char *)&opt[IPOPT_OFFSET] -
13057 (char *)ipha;
13058 goto param_prob;
13059 }
13060 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
13061 case IPOPT_TS_TSONLY:
13062 off = IPOPT_TS_TIMELEN;
13063 break;
13064 case IPOPT_TS_TSANDADDR:
13065 case IPOPT_TS_PRESPEC:
13066 case IPOPT_TS_PRESPEC_RFC791:
13067 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
13068 break;
13069 default:
13070 code = (char *)&opt[IPOPT_POS_OV_FLG] -
13071 (char *)ipha;
13072 goto param_prob;
13073 }
13074 if (opt[IPOPT_OFFSET] - 1 + off > optlen &&
13075 (opt[IPOPT_POS_OV_FLG] & 0xF0) == 0xF0) {
13076 /*
13077 * No room and the overflow counter is 15
13078 * already.
13079 */
13080 goto param_prob;
13081 }
13082 break;
13083 }
13084 }
13085
13086 if ((opts.ipoptp_flags & IPOPTP_ERROR) == 0)
13087 return (0);
13088
13089 ip1dbg(("ip_output_options: error processing IP options."));
13090 code = (char *)&opt[IPOPT_OFFSET] - (char *)ipha;
13091
13092 param_prob:
13093 bzero(&iras, sizeof (iras));
13094 iras.ira_ill = iras.ira_rill = ill;
13095 iras.ira_ruifindex = ill->ill_phyint->phyint_ifindex;
13096 iras.ira_rifindex = iras.ira_ruifindex;
13097 iras.ira_flags = IRAF_IS_IPV4;
13098
13099 ip_drop_output("ip_output_options", mp, ill);
13100 icmp_param_problem(mp, (uint8_t)code, &iras);
13101 ASSERT(!(iras.ira_flags & IRAF_IPSEC_SECURE));
13102 return (-1);
13103
13104 bad_src_route:
13105 bzero(&iras, sizeof (iras));
13106 iras.ira_ill = iras.ira_rill = ill;
13107 iras.ira_ruifindex = ill->ill_phyint->phyint_ifindex;
13108 iras.ira_rifindex = iras.ira_ruifindex;
13109 iras.ira_flags = IRAF_IS_IPV4;
13110
13111 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ill);
13112 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, &iras);
13113 ASSERT(!(iras.ira_flags & IRAF_IPSEC_SECURE));
13114 return (-1);
13115 }
13116
13117 /*
13118 * The maximum value of conn_drain_list_cnt is CONN_MAXDRAINCNT.
13119 * conn_drain_list_cnt can be changed by setting conn_drain_nthreads
13120 * thru /etc/system.
13121 */
13122 #define CONN_MAXDRAINCNT 64
13123
13124 static void
13125 conn_drain_init(ip_stack_t *ipst)
13126 {
13127 int i, j;
13128 idl_tx_list_t *itl_tx;
13129
13130 ipst->ips_conn_drain_list_cnt = conn_drain_nthreads;
13131
13132 if ((ipst->ips_conn_drain_list_cnt == 0) ||
13133 (ipst->ips_conn_drain_list_cnt > CONN_MAXDRAINCNT)) {
13134 /*
13135 * Default value of the number of drainers is the
13136 * number of cpus, subject to maximum of 8 drainers.
13137 */
13138 if (boot_max_ncpus != -1)
13139 ipst->ips_conn_drain_list_cnt = MIN(boot_max_ncpus, 8);
13140 else
13141 ipst->ips_conn_drain_list_cnt = MIN(max_ncpus, 8);
13142 }
13143
13144 ipst->ips_idl_tx_list =
13145 kmem_zalloc(TX_FANOUT_SIZE * sizeof (idl_tx_list_t), KM_SLEEP);
13146 for (i = 0; i < TX_FANOUT_SIZE; i++) {
13147 itl_tx = &ipst->ips_idl_tx_list[i];
13148 itl_tx->txl_drain_list =
13149 kmem_zalloc(ipst->ips_conn_drain_list_cnt *
13150 sizeof (idl_t), KM_SLEEP);
13151 mutex_init(&itl_tx->txl_lock, NULL, MUTEX_DEFAULT, NULL);
13152 for (j = 0; j < ipst->ips_conn_drain_list_cnt; j++) {
13153 mutex_init(&itl_tx->txl_drain_list[j].idl_lock, NULL,
13154 MUTEX_DEFAULT, NULL);
13155 itl_tx->txl_drain_list[j].idl_itl = itl_tx;
13156 }
13157 }
13158 }
13159
13160 static void
13161 conn_drain_fini(ip_stack_t *ipst)
13162 {
13163 int i;
13164 idl_tx_list_t *itl_tx;
13165
13166 for (i = 0; i < TX_FANOUT_SIZE; i++) {
13167 itl_tx = &ipst->ips_idl_tx_list[i];
13168 kmem_free(itl_tx->txl_drain_list,
13169 ipst->ips_conn_drain_list_cnt * sizeof (idl_t));
13170 }
13171 kmem_free(ipst->ips_idl_tx_list,
13172 TX_FANOUT_SIZE * sizeof (idl_tx_list_t));
13173 ipst->ips_idl_tx_list = NULL;
13174 }
13175
13176 /*
13177 * Flow control has blocked us from proceeding. Insert the given conn in one
13178 * of the conn drain lists. When flow control is unblocked, either ip_wsrv()
13179 * (STREAMS) or ill_flow_enable() (direct) will be called back, which in turn
13180 * will call conn_walk_drain(). See the flow control notes at the top of this
13181 * file for more details.
13182 */
13183 void
13184 conn_drain_insert(conn_t *connp, idl_tx_list_t *tx_list)
13185 {
13186 idl_t *idl = tx_list->txl_drain_list;
13187 uint_t index;
13188 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
13189
13190 mutex_enter(&connp->conn_lock);
13191 if (connp->conn_state_flags & CONN_CLOSING) {
13192 /*
13193 * The conn is closing as a result of which CONN_CLOSING
13194 * is set. Return.
13195 */
13196 mutex_exit(&connp->conn_lock);
13197 return;
13198 } else if (connp->conn_idl == NULL) {
13199 /*
13200 * Assign the next drain list round robin. We dont' use
13201 * a lock, and thus it may not be strictly round robin.
13202 * Atomicity of load/stores is enough to make sure that
13203 * conn_drain_list_index is always within bounds.
13204 */
13205 index = tx_list->txl_drain_index;
13206 ASSERT(index < ipst->ips_conn_drain_list_cnt);
13207 connp->conn_idl = &tx_list->txl_drain_list[index];
13208 index++;
13209 if (index == ipst->ips_conn_drain_list_cnt)
13210 index = 0;
13211 tx_list->txl_drain_index = index;
13212 } else {
13213 ASSERT(connp->conn_idl->idl_itl == tx_list);
13214 }
13215 mutex_exit(&connp->conn_lock);
13216
13217 idl = connp->conn_idl;
13218 mutex_enter(&idl->idl_lock);
13219 if ((connp->conn_drain_prev != NULL) ||
13220 (connp->conn_state_flags & CONN_CLOSING)) {
13221 /*
13222 * The conn is either already in the drain list or closing.
13223 * (We needed to check for CONN_CLOSING again since close can
13224 * sneak in between dropping conn_lock and acquiring idl_lock.)
13225 */
13226 mutex_exit(&idl->idl_lock);
13227 return;
13228 }
13229
13230 /*
13231 * The conn is not in the drain list. Insert it at the
13232 * tail of the drain list. The drain list is circular
13233 * and doubly linked. idl_conn points to the 1st element
13234 * in the list.
13235 */
13236 if (idl->idl_conn == NULL) {
13237 idl->idl_conn = connp;
13238 connp->conn_drain_next = connp;
13239 connp->conn_drain_prev = connp;
13240 } else {
13241 conn_t *head = idl->idl_conn;
13242
13243 connp->conn_drain_next = head;
13244 connp->conn_drain_prev = head->conn_drain_prev;
13245 head->conn_drain_prev->conn_drain_next = connp;
13246 head->conn_drain_prev = connp;
13247 }
13248 /*
13249 * For non streams based sockets assert flow control.
13250 */
13251 conn_setqfull(connp, NULL);
13252 mutex_exit(&idl->idl_lock);
13253 }
13254
13255 static void
13256 conn_drain_remove(conn_t *connp)
13257 {
13258 idl_t *idl = connp->conn_idl;
13259
13260 if (idl != NULL) {
13261 /*
13262 * Remove ourself from the drain list.
13263 */
13264 if (connp->conn_drain_next == connp) {
13265 /* Singleton in the list */
13266 ASSERT(connp->conn_drain_prev == connp);
13267 idl->idl_conn = NULL;
13268 } else {
13269 connp->conn_drain_prev->conn_drain_next =
13270 connp->conn_drain_next;
13271 connp->conn_drain_next->conn_drain_prev =
13272 connp->conn_drain_prev;
13273 if (idl->idl_conn == connp)
13274 idl->idl_conn = connp->conn_drain_next;
13275 }
13276
13277 /*
13278 * NOTE: because conn_idl is associated with a specific drain
13279 * list which in turn is tied to the index the TX ring
13280 * (txl_cookie) hashes to, and because the TX ring can change
13281 * over the lifetime of the conn_t, we must clear conn_idl so
13282 * a subsequent conn_drain_insert() will set conn_idl again
13283 * based on the latest txl_cookie.
13284 */
13285 connp->conn_idl = NULL;
13286 }
13287 connp->conn_drain_next = NULL;
13288 connp->conn_drain_prev = NULL;
13289
13290 conn_clrqfull(connp, NULL);
13291 /*
13292 * For streams based sockets open up flow control.
13293 */
13294 if (!IPCL_IS_NONSTR(connp))
13295 enableok(connp->conn_wq);
13296 }
13297
13298 /*
13299 * This conn is closing, and we are called from ip_close. OR
13300 * this conn is draining because flow-control on the ill has been relieved.
13301 *
13302 * We must also need to remove conn's on this idl from the list, and also
13303 * inform the sockfs upcalls about the change in flow-control.
13304 */
13305 static void
13306 conn_drain(conn_t *connp, boolean_t closing)
13307 {
13308 idl_t *idl;
13309 conn_t *next_connp;
13310
13311 /*
13312 * connp->conn_idl is stable at this point, and no lock is needed
13313 * to check it. If we are called from ip_close, close has already
13314 * set CONN_CLOSING, thus freezing the value of conn_idl, and
13315 * called us only because conn_idl is non-null. If we are called thru
13316 * service, conn_idl could be null, but it cannot change because
13317 * service is single-threaded per queue, and there cannot be another
13318 * instance of service trying to call conn_drain_insert on this conn
13319 * now.
13320 */
13321 ASSERT(!closing || connp == NULL || connp->conn_idl != NULL);
13322
13323 /*
13324 * If the conn doesn't exist or is not on a drain list, bail.
13325 */
13326 if (connp == NULL || connp->conn_idl == NULL ||
13327 connp->conn_drain_prev == NULL) {
13328 return;
13329 }
13330
13331 idl = connp->conn_idl;
13332 ASSERT(MUTEX_HELD(&idl->idl_lock));
13333
13334 if (!closing) {
13335 next_connp = connp->conn_drain_next;
13336 while (next_connp != connp) {
13337 conn_t *delconnp = next_connp;
13338
13339 next_connp = next_connp->conn_drain_next;
13340 conn_drain_remove(delconnp);
13341 }
13342 ASSERT(connp->conn_drain_next == idl->idl_conn);
13343 }
13344 conn_drain_remove(connp);
13345 }
13346
13347 /*
13348 * Write service routine. Shared perimeter entry point.
13349 * The device queue's messages has fallen below the low water mark and STREAMS
13350 * has backenabled the ill_wq. Send sockfs notification about flow-control on
13351 * each waiting conn.
13352 */
13353 int
13354 ip_wsrv(queue_t *q)
13355 {
13356 ill_t *ill;
13357
13358 ill = (ill_t *)q->q_ptr;
13359 if (ill->ill_state_flags == 0) {
13360 ip_stack_t *ipst = ill->ill_ipst;
13361
13362 /*
13363 * The device flow control has opened up.
13364 * Walk through conn drain lists and qenable the
13365 * first conn in each list. This makes sense only
13366 * if the stream is fully plumbed and setup.
13367 * Hence the ill_state_flags check above.
13368 */
13369 ip1dbg(("ip_wsrv: walking\n"));
13370 conn_walk_drain(ipst, &ipst->ips_idl_tx_list[0]);
13371 enableok(ill->ill_wq);
13372 }
13373 return (0);
13374 }
13375
13376 /*
13377 * Callback to disable flow control in IP.
13378 *
13379 * This is a mac client callback added when the DLD_CAPAB_DIRECT capability
13380 * is enabled.
13381 *
13382 * When MAC_TX() is not able to send any more packets, dld sets its queue
13383 * to QFULL and enable the STREAMS flow control. Later, when the underlying
13384 * driver is able to continue to send packets, it calls mac_tx_(ring_)update()
13385 * function and wakes up corresponding mac worker threads, which in turn
13386 * calls this callback function, and disables flow control.
13387 */
13388 void
13389 ill_flow_enable(void *arg, ip_mac_tx_cookie_t cookie)
13390 {
13391 ill_t *ill = (ill_t *)arg;
13392 ip_stack_t *ipst = ill->ill_ipst;
13393 idl_tx_list_t *idl_txl;
13394
13395 idl_txl = &ipst->ips_idl_tx_list[IDLHASHINDEX(cookie)];
13396 mutex_enter(&idl_txl->txl_lock);
13397 /* add code to to set a flag to indicate idl_txl is enabled */
13398 conn_walk_drain(ipst, idl_txl);
13399 mutex_exit(&idl_txl->txl_lock);
13400 }
13401
13402 /*
13403 * Flow control has been relieved and STREAMS has backenabled us; drain
13404 * all the conn lists on `tx_list'.
13405 */
13406 static void
13407 conn_walk_drain(ip_stack_t *ipst, idl_tx_list_t *tx_list)
13408 {
13409 int i;
13410 idl_t *idl;
13411
13412 IP_STAT(ipst, ip_conn_walk_drain);
13413
13414 for (i = 0; i < ipst->ips_conn_drain_list_cnt; i++) {
13415 idl = &tx_list->txl_drain_list[i];
13416 mutex_enter(&idl->idl_lock);
13417 conn_drain(idl->idl_conn, B_FALSE);
13418 mutex_exit(&idl->idl_lock);
13419 }
13420 }
13421
13422 /*
13423 * Determine if the ill and multicast aspects of that packets
13424 * "matches" the conn.
13425 */
13426 boolean_t
13427 conn_wantpacket(conn_t *connp, ip_recv_attr_t *ira, ipha_t *ipha)
13428 {
13429 ill_t *ill = ira->ira_rill;
13430 zoneid_t zoneid = ira->ira_zoneid;
13431 uint_t in_ifindex;
13432 ipaddr_t dst, src;
13433
13434 dst = ipha->ipha_dst;
13435 src = ipha->ipha_src;
13436
13437 /*
13438 * conn_incoming_ifindex is set by IP_BOUND_IF which limits
13439 * unicast, broadcast and multicast reception to
13440 * conn_incoming_ifindex.
13441 * conn_wantpacket is called for unicast, broadcast and
13442 * multicast packets.
13443 */
13444 in_ifindex = connp->conn_incoming_ifindex;
13445
13446 /* mpathd can bind to the under IPMP interface, which we allow */
13447 if (in_ifindex != 0 && in_ifindex != ill->ill_phyint->phyint_ifindex) {
13448 if (!IS_UNDER_IPMP(ill))
13449 return (B_FALSE);
13450
13451 if (in_ifindex != ipmp_ill_get_ipmp_ifindex(ill))
13452 return (B_FALSE);
13453 }
13454
13455 if (!IPCL_ZONE_MATCH(connp, zoneid))
13456 return (B_FALSE);
13457
13458 if (!(ira->ira_flags & IRAF_MULTICAST))
13459 return (B_TRUE);
13460
13461 if (connp->conn_multi_router) {
13462 /* multicast packet and multicast router socket: send up */
13463 return (B_TRUE);
13464 }
13465
13466 if (ipha->ipha_protocol == IPPROTO_PIM ||
13467 ipha->ipha_protocol == IPPROTO_RSVP)
13468 return (B_TRUE);
13469
13470 return (conn_hasmembers_ill_withsrc_v4(connp, dst, src, ira->ira_ill));
13471 }
13472
13473 void
13474 conn_setqfull(conn_t *connp, boolean_t *flow_stopped)
13475 {
13476 if (IPCL_IS_NONSTR(connp)) {
13477 (*connp->conn_upcalls->su_txq_full)
13478 (connp->conn_upper_handle, B_TRUE);
13479 if (flow_stopped != NULL)
13480 *flow_stopped = B_TRUE;
13481 } else {
13482 queue_t *q = connp->conn_wq;
13483
13484 ASSERT(q != NULL);
13485 if (!(q->q_flag & QFULL)) {
13486 mutex_enter(QLOCK(q));
13487 if (!(q->q_flag & QFULL)) {
13488 /* still need to set QFULL */
13489 q->q_flag |= QFULL;
13490 /* set flow_stopped to true under QLOCK */
13491 if (flow_stopped != NULL)
13492 *flow_stopped = B_TRUE;
13493 mutex_exit(QLOCK(q));
13494 } else {
13495 /* flow_stopped is left unchanged */
13496 mutex_exit(QLOCK(q));
13497 }
13498 }
13499 }
13500 }
13501
13502 void
13503 conn_clrqfull(conn_t *connp, boolean_t *flow_stopped)
13504 {
13505 if (IPCL_IS_NONSTR(connp)) {
13506 (*connp->conn_upcalls->su_txq_full)
13507 (connp->conn_upper_handle, B_FALSE);
13508 if (flow_stopped != NULL)
13509 *flow_stopped = B_FALSE;
13510 } else {
13511 queue_t *q = connp->conn_wq;
13512
13513 ASSERT(q != NULL);
13514 if (q->q_flag & QFULL) {
13515 mutex_enter(QLOCK(q));
13516 if (q->q_flag & QFULL) {
13517 q->q_flag &= ~QFULL;
13518 /* set flow_stopped to false under QLOCK */
13519 if (flow_stopped != NULL)
13520 *flow_stopped = B_FALSE;
13521 mutex_exit(QLOCK(q));
13522 if (q->q_flag & QWANTW)
13523 qbackenable(q, 0);
13524 } else {
13525 /* flow_stopped is left unchanged */
13526 mutex_exit(QLOCK(q));
13527 }
13528 }
13529 }
13530
13531 mutex_enter(&connp->conn_lock);
13532 connp->conn_blocked = B_FALSE;
13533 mutex_exit(&connp->conn_lock);
13534 }
13535
13536 /*
13537 * Return the length in bytes of the IPv4 headers (base header, label, and
13538 * other IP options) that will be needed based on the
13539 * ip_pkt_t structure passed by the caller.
13540 *
13541 * The returned length does not include the length of the upper level
13542 * protocol (ULP) header.
13543 * The caller needs to check that the length doesn't exceed the max for IPv4.
13544 */
13545 int
13546 ip_total_hdrs_len_v4(const ip_pkt_t *ipp)
13547 {
13548 int len;
13549
13550 len = IP_SIMPLE_HDR_LENGTH;
13551 if (ipp->ipp_fields & IPPF_LABEL_V4) {
13552 ASSERT(ipp->ipp_label_len_v4 != 0);
13553 /* We need to round up here */
13554 len += (ipp->ipp_label_len_v4 + 3) & ~3;
13555 }
13556
13557 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
13558 ASSERT(ipp->ipp_ipv4_options_len != 0);
13559 ASSERT((ipp->ipp_ipv4_options_len & 3) == 0);
13560 len += ipp->ipp_ipv4_options_len;
13561 }
13562 return (len);
13563 }
13564
13565 /*
13566 * All-purpose routine to build an IPv4 header with options based
13567 * on the abstract ip_pkt_t.
13568 *
13569 * The caller has to set the source and destination address as well as
13570 * ipha_length. The caller has to massage any source route and compensate
13571 * for the ULP pseudo-header checksum due to the source route.
13572 */
13573 void
13574 ip_build_hdrs_v4(uchar_t *buf, uint_t buf_len, const ip_pkt_t *ipp,
13575 uint8_t protocol)
13576 {
13577 ipha_t *ipha = (ipha_t *)buf;
13578 uint8_t *cp;
13579
13580 /* Initialize IPv4 header */
13581 ipha->ipha_type_of_service = ipp->ipp_type_of_service;
13582 ipha->ipha_length = 0; /* Caller will set later */
13583 ipha->ipha_ident = 0;
13584 ipha->ipha_fragment_offset_and_flags = 0;
13585 ipha->ipha_ttl = ipp->ipp_unicast_hops;
13586 ipha->ipha_protocol = protocol;
13587 ipha->ipha_hdr_checksum = 0;
13588
13589 if ((ipp->ipp_fields & IPPF_ADDR) &&
13590 IN6_IS_ADDR_V4MAPPED(&ipp->ipp_addr))
13591 ipha->ipha_src = ipp->ipp_addr_v4;
13592
13593 cp = (uint8_t *)&ipha[1];
13594 if (ipp->ipp_fields & IPPF_LABEL_V4) {
13595 ASSERT(ipp->ipp_label_len_v4 != 0);
13596 bcopy(ipp->ipp_label_v4, cp, ipp->ipp_label_len_v4);
13597 cp += ipp->ipp_label_len_v4;
13598 /* We need to round up here */
13599 while ((uintptr_t)cp & 0x3) {
13600 *cp++ = IPOPT_NOP;
13601 }
13602 }
13603
13604 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
13605 ASSERT(ipp->ipp_ipv4_options_len != 0);
13606 ASSERT((ipp->ipp_ipv4_options_len & 3) == 0);
13607 bcopy(ipp->ipp_ipv4_options, cp, ipp->ipp_ipv4_options_len);
13608 cp += ipp->ipp_ipv4_options_len;
13609 }
13610 ipha->ipha_version_and_hdr_length =
13611 (uint8_t)((IP_VERSION << 4) + buf_len / 4);
13612
13613 ASSERT((int)(cp - buf) == buf_len);
13614 }
13615
13616 /* Allocate the private structure */
13617 static int
13618 ip_priv_alloc(void **bufp)
13619 {
13620 void *buf;
13621
13622 if ((buf = kmem_alloc(sizeof (ip_priv_t), KM_NOSLEEP)) == NULL)
13623 return (ENOMEM);
13624
13625 *bufp = buf;
13626 return (0);
13627 }
13628
13629 /* Function to delete the private structure */
13630 void
13631 ip_priv_free(void *buf)
13632 {
13633 ASSERT(buf != NULL);
13634 kmem_free(buf, sizeof (ip_priv_t));
13635 }
13636
13637 /*
13638 * The entry point for IPPF processing.
13639 * If the classifier (IPGPC_CLASSIFY) is not loaded and configured, the
13640 * routine just returns.
13641 *
13642 * When called, ip_process generates an ipp_packet_t structure
13643 * which holds the state information for this packet and invokes the
13644 * the classifier (via ipp_packet_process). The classification, depending on
13645 * configured filters, results in a list of actions for this packet. Invoking
13646 * an action may cause the packet to be dropped, in which case we return NULL.
13647 * proc indicates the callout position for
13648 * this packet and ill is the interface this packet arrived on or will leave
13649 * on (inbound and outbound resp.).
13650 *
13651 * We do the processing on the rill (mapped to the upper if ipmp), but MIB
13652 * on the ill corrsponding to the destination IP address.
13653 */
13654 mblk_t *
13655 ip_process(ip_proc_t proc, mblk_t *mp, ill_t *rill, ill_t *ill)
13656 {
13657 ip_priv_t *priv;
13658 ipp_action_id_t aid;
13659 int rc = 0;
13660 ipp_packet_t *pp;
13661
13662 /* If the classifier is not loaded, return */
13663 if ((aid = ipp_action_lookup(IPGPC_CLASSIFY)) == IPP_ACTION_INVAL) {
13664 return (mp);
13665 }
13666
13667 ASSERT(mp != NULL);
13668
13669 /* Allocate the packet structure */
13670 rc = ipp_packet_alloc(&pp, "ip", aid);
13671 if (rc != 0)
13672 goto drop;
13673
13674 /* Allocate the private structure */
13675 rc = ip_priv_alloc((void **)&priv);
13676 if (rc != 0) {
13677 ipp_packet_free(pp);
13678 goto drop;
13679 }
13680 priv->proc = proc;
13681 priv->ill_index = ill_get_upper_ifindex(rill);
13682
13683 ipp_packet_set_private(pp, priv, ip_priv_free);
13684 ipp_packet_set_data(pp, mp);
13685
13686 /* Invoke the classifier */
13687 rc = ipp_packet_process(&pp);
13688 if (pp != NULL) {
13689 mp = ipp_packet_get_data(pp);
13690 ipp_packet_free(pp);
13691 if (rc != 0)
13692 goto drop;
13693 return (mp);
13694 } else {
13695 /* No mp to trace in ip_drop_input/ip_drop_output */
13696 mp = NULL;
13697 }
13698 drop:
13699 if (proc == IPP_LOCAL_IN || proc == IPP_FWD_IN) {
13700 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
13701 ip_drop_input("ip_process", mp, ill);
13702 } else {
13703 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
13704 ip_drop_output("ip_process", mp, ill);
13705 }
13706 freemsg(mp);
13707 return (NULL);
13708 }
13709
13710 /*
13711 * Propagate a multicast group membership operation (add/drop) on
13712 * all the interfaces crossed by the related multirt routes.
13713 * The call is considered successful if the operation succeeds
13714 * on at least one interface.
13715 *
13716 * This assumes that a set of IRE_HOST/RTF_MULTIRT has been created for the
13717 * multicast addresses with the ire argument being the first one.
13718 * We walk the bucket to find all the of those.
13719 *
13720 * Common to IPv4 and IPv6.
13721 */
13722 static int
13723 ip_multirt_apply_membership(int (*fn)(conn_t *, boolean_t,
13724 const in6_addr_t *, ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *),
13725 ire_t *ire, conn_t *connp, boolean_t checkonly, const in6_addr_t *v6group,
13726 mcast_record_t fmode, const in6_addr_t *v6src)
13727 {
13728 ire_t *ire_gw;
13729 irb_t *irb;
13730 int ifindex;
13731 int error = 0;
13732 int result;
13733 ip_stack_t *ipst = ire->ire_ipst;
13734 ipaddr_t group;
13735 boolean_t isv6;
13736 int match_flags;
13737
13738 if (IN6_IS_ADDR_V4MAPPED(v6group)) {
13739 IN6_V4MAPPED_TO_IPADDR(v6group, group);
13740 isv6 = B_FALSE;
13741 } else {
13742 isv6 = B_TRUE;
13743 }
13744
13745 irb = ire->ire_bucket;
13746 ASSERT(irb != NULL);
13747
13748 result = 0;
13749 irb_refhold(irb);
13750 for (; ire != NULL; ire = ire->ire_next) {
13751 if ((ire->ire_flags & RTF_MULTIRT) == 0)
13752 continue;
13753
13754 /* We handle -ifp routes by matching on the ill if set */
13755 match_flags = MATCH_IRE_TYPE;
13756 if (ire->ire_ill != NULL)
13757 match_flags |= MATCH_IRE_ILL;
13758
13759 if (isv6) {
13760 if (!IN6_ARE_ADDR_EQUAL(&ire->ire_addr_v6, v6group))
13761 continue;
13762
13763 ire_gw = ire_ftable_lookup_v6(&ire->ire_gateway_addr_v6,
13764 0, 0, IRE_INTERFACE, ire->ire_ill, ALL_ZONES, NULL,
13765 match_flags, 0, ipst, NULL);
13766 } else {
13767 if (ire->ire_addr != group)
13768 continue;
13769
13770 ire_gw = ire_ftable_lookup_v4(ire->ire_gateway_addr,
13771 0, 0, IRE_INTERFACE, ire->ire_ill, ALL_ZONES, NULL,
13772 match_flags, 0, ipst, NULL);
13773 }
13774 /* No interface route exists for the gateway; skip this ire. */
13775 if (ire_gw == NULL)
13776 continue;
13777 if (ire_gw->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
13778 ire_refrele(ire_gw);
13779 continue;
13780 }
13781 ASSERT(ire_gw->ire_ill != NULL); /* IRE_INTERFACE */
13782 ifindex = ire_gw->ire_ill->ill_phyint->phyint_ifindex;
13783
13784 /*
13785 * The operation is considered a success if
13786 * it succeeds at least once on any one interface.
13787 */
13788 error = fn(connp, checkonly, v6group, INADDR_ANY, ifindex,
13789 fmode, v6src);
13790 if (error == 0)
13791 result = CGTP_MCAST_SUCCESS;
13792
13793 ire_refrele(ire_gw);
13794 }
13795 irb_refrele(irb);
13796 /*
13797 * Consider the call as successful if we succeeded on at least
13798 * one interface. Otherwise, return the last encountered error.
13799 */
13800 return (result == CGTP_MCAST_SUCCESS ? 0 : error);
13801 }
13802
13803 /*
13804 * Return the expected CGTP hooks version number.
13805 */
13806 int
13807 ip_cgtp_filter_supported(void)
13808 {
13809 return (ip_cgtp_filter_rev);
13810 }
13811
13812 /*
13813 * CGTP hooks can be registered by invoking this function.
13814 * Checks that the version number matches.
13815 */
13816 int
13817 ip_cgtp_filter_register(netstackid_t stackid, cgtp_filter_ops_t *ops)
13818 {
13819 netstack_t *ns;
13820 ip_stack_t *ipst;
13821
13822 if (ops->cfo_filter_rev != CGTP_FILTER_REV)
13823 return (ENOTSUP);
13824
13825 ns = netstack_find_by_stackid(stackid);
13826 if (ns == NULL)
13827 return (EINVAL);
13828 ipst = ns->netstack_ip;
13829 ASSERT(ipst != NULL);
13830
13831 if (ipst->ips_ip_cgtp_filter_ops != NULL) {
13832 netstack_rele(ns);
13833 return (EALREADY);
13834 }
13835
13836 ipst->ips_ip_cgtp_filter_ops = ops;
13837
13838 ill_set_inputfn_all(ipst);
13839
13840 netstack_rele(ns);
13841 return (0);
13842 }
13843
13844 /*
13845 * CGTP hooks can be unregistered by invoking this function.
13846 * Returns ENXIO if there was no registration.
13847 * Returns EBUSY if the ndd variable has not been turned off.
13848 */
13849 int
13850 ip_cgtp_filter_unregister(netstackid_t stackid)
13851 {
13852 netstack_t *ns;
13853 ip_stack_t *ipst;
13854
13855 ns = netstack_find_by_stackid(stackid);
13856 if (ns == NULL)
13857 return (EINVAL);
13858 ipst = ns->netstack_ip;
13859 ASSERT(ipst != NULL);
13860
13861 if (ipst->ips_ip_cgtp_filter) {
13862 netstack_rele(ns);
13863 return (EBUSY);
13864 }
13865
13866 if (ipst->ips_ip_cgtp_filter_ops == NULL) {
13867 netstack_rele(ns);
13868 return (ENXIO);
13869 }
13870 ipst->ips_ip_cgtp_filter_ops = NULL;
13871
13872 ill_set_inputfn_all(ipst);
13873
13874 netstack_rele(ns);
13875 return (0);
13876 }
13877
13878 /*
13879 * Check whether there is a CGTP filter registration.
13880 * Returns non-zero if there is a registration, otherwise returns zero.
13881 * Note: returns zero if bad stackid.
13882 */
13883 int
13884 ip_cgtp_filter_is_registered(netstackid_t stackid)
13885 {
13886 netstack_t *ns;
13887 ip_stack_t *ipst;
13888 int ret;
13889
13890 ns = netstack_find_by_stackid(stackid);
13891 if (ns == NULL)
13892 return (0);
13893 ipst = ns->netstack_ip;
13894 ASSERT(ipst != NULL);
13895
13896 if (ipst->ips_ip_cgtp_filter_ops != NULL)
13897 ret = 1;
13898 else
13899 ret = 0;
13900
13901 netstack_rele(ns);
13902 return (ret);
13903 }
13904
13905 static int
13906 ip_squeue_switch(int val)
13907 {
13908 int rval;
13909
13910 switch (val) {
13911 case IP_SQUEUE_ENTER_NODRAIN:
13912 rval = SQ_NODRAIN;
13913 break;
13914 case IP_SQUEUE_ENTER:
13915 rval = SQ_PROCESS;
13916 break;
13917 case IP_SQUEUE_FILL:
13918 default:
13919 rval = SQ_FILL;
13920 break;
13921 }
13922 return (rval);
13923 }
13924
13925 static void *
13926 ip_kstat2_init(netstackid_t stackid, ip_stat_t *ip_statisticsp)
13927 {
13928 kstat_t *ksp;
13929
13930 ip_stat_t template = {
13931 { "ip_udp_fannorm", KSTAT_DATA_UINT64 },
13932 { "ip_udp_fanmb", KSTAT_DATA_UINT64 },
13933 { "ip_recv_pullup", KSTAT_DATA_UINT64 },
13934 { "ip_db_ref", KSTAT_DATA_UINT64 },
13935 { "ip_notaligned", KSTAT_DATA_UINT64 },
13936 { "ip_multimblk", KSTAT_DATA_UINT64 },
13937 { "ip_opt", KSTAT_DATA_UINT64 },
13938 { "ipsec_proto_ahesp", KSTAT_DATA_UINT64 },
13939 { "ip_conn_flputbq", KSTAT_DATA_UINT64 },
13940 { "ip_conn_walk_drain", KSTAT_DATA_UINT64 },
13941 { "ip_out_sw_cksum", KSTAT_DATA_UINT64 },
13942 { "ip_out_sw_cksum_bytes", KSTAT_DATA_UINT64 },
13943 { "ip_in_sw_cksum", KSTAT_DATA_UINT64 },
13944 { "ip_ire_reclaim_calls", KSTAT_DATA_UINT64 },
13945 { "ip_ire_reclaim_deleted", KSTAT_DATA_UINT64 },
13946 { "ip_nce_reclaim_calls", KSTAT_DATA_UINT64 },
13947 { "ip_nce_reclaim_deleted", KSTAT_DATA_UINT64 },
13948 { "ip_nce_mcast_reclaim_calls", KSTAT_DATA_UINT64 },
13949 { "ip_nce_mcast_reclaim_deleted", KSTAT_DATA_UINT64 },
13950 { "ip_nce_mcast_reclaim_tqfail", KSTAT_DATA_UINT64 },
13951 { "ip_dce_reclaim_calls", KSTAT_DATA_UINT64 },
13952 { "ip_dce_reclaim_deleted", KSTAT_DATA_UINT64 },
13953 { "ip_tcp_in_full_hw_cksum_err", KSTAT_DATA_UINT64 },
13954 { "ip_tcp_in_part_hw_cksum_err", KSTAT_DATA_UINT64 },
13955 { "ip_tcp_in_sw_cksum_err", KSTAT_DATA_UINT64 },
13956 { "ip_udp_in_full_hw_cksum_err", KSTAT_DATA_UINT64 },
13957 { "ip_udp_in_part_hw_cksum_err", KSTAT_DATA_UINT64 },
13958 { "ip_udp_in_sw_cksum_err", KSTAT_DATA_UINT64 },
13959 { "conn_in_recvdstaddr", KSTAT_DATA_UINT64 },
13960 { "conn_in_recvopts", KSTAT_DATA_UINT64 },
13961 { "conn_in_recvif", KSTAT_DATA_UINT64 },
13962 { "conn_in_recvslla", KSTAT_DATA_UINT64 },
13963 { "conn_in_recvucred", KSTAT_DATA_UINT64 },
13964 { "conn_in_recvttl", KSTAT_DATA_UINT64 },
13965 { "conn_in_recvhopopts", KSTAT_DATA_UINT64 },
13966 { "conn_in_recvhoplimit", KSTAT_DATA_UINT64 },
13967 { "conn_in_recvdstopts", KSTAT_DATA_UINT64 },
13968 { "conn_in_recvrthdrdstopts", KSTAT_DATA_UINT64 },
13969 { "conn_in_recvrthdr", KSTAT_DATA_UINT64 },
13970 { "conn_in_recvpktinfo", KSTAT_DATA_UINT64 },
13971 { "conn_in_recvtclass", KSTAT_DATA_UINT64 },
13972 { "conn_in_timestamp", KSTAT_DATA_UINT64 },
13973 };
13974
13975 ksp = kstat_create_netstack("ip", 0, "ipstat", "net",
13976 KSTAT_TYPE_NAMED, sizeof (template) / sizeof (kstat_named_t),
13977 KSTAT_FLAG_VIRTUAL, stackid);
13978
13979 if (ksp == NULL)
13980 return (NULL);
13981
13982 bcopy(&template, ip_statisticsp, sizeof (template));
13983 ksp->ks_data = (void *)ip_statisticsp;
13984 ksp->ks_private = (void *)(uintptr_t)stackid;
13985
13986 kstat_install(ksp);
13987 return (ksp);
13988 }
13989
13990 static void
13991 ip_kstat2_fini(netstackid_t stackid, kstat_t *ksp)
13992 {
13993 if (ksp != NULL) {
13994 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
13995 kstat_delete_netstack(ksp, stackid);
13996 }
13997 }
13998
13999 static void *
14000 ip_kstat_init(netstackid_t stackid, ip_stack_t *ipst)
14001 {
14002 kstat_t *ksp;
14003
14004 ip_named_kstat_t template = {
14005 { "forwarding", KSTAT_DATA_UINT32, 0 },
14006 { "defaultTTL", KSTAT_DATA_UINT32, 0 },
14007 { "inReceives", KSTAT_DATA_UINT64, 0 },
14008 { "inHdrErrors", KSTAT_DATA_UINT32, 0 },
14009 { "inAddrErrors", KSTAT_DATA_UINT32, 0 },
14010 { "forwDatagrams", KSTAT_DATA_UINT64, 0 },
14011 { "inUnknownProtos", KSTAT_DATA_UINT32, 0 },
14012 { "inDiscards", KSTAT_DATA_UINT32, 0 },
14013 { "inDelivers", KSTAT_DATA_UINT64, 0 },
14014 { "outRequests", KSTAT_DATA_UINT64, 0 },
14015 { "outDiscards", KSTAT_DATA_UINT32, 0 },
14016 { "outNoRoutes", KSTAT_DATA_UINT32, 0 },
14017 { "reasmTimeout", KSTAT_DATA_UINT32, 0 },
14018 { "reasmReqds", KSTAT_DATA_UINT32, 0 },
14019 { "reasmOKs", KSTAT_DATA_UINT32, 0 },
14020 { "reasmFails", KSTAT_DATA_UINT32, 0 },
14021 { "fragOKs", KSTAT_DATA_UINT32, 0 },
14022 { "fragFails", KSTAT_DATA_UINT32, 0 },
14023 { "fragCreates", KSTAT_DATA_UINT32, 0 },
14024 { "addrEntrySize", KSTAT_DATA_INT32, 0 },
14025 { "routeEntrySize", KSTAT_DATA_INT32, 0 },
14026 { "netToMediaEntrySize", KSTAT_DATA_INT32, 0 },
14027 { "routingDiscards", KSTAT_DATA_UINT32, 0 },
14028 { "inErrs", KSTAT_DATA_UINT32, 0 },
14029 { "noPorts", KSTAT_DATA_UINT32, 0 },
14030 { "inCksumErrs", KSTAT_DATA_UINT32, 0 },
14031 { "reasmDuplicates", KSTAT_DATA_UINT32, 0 },
14032 { "reasmPartDups", KSTAT_DATA_UINT32, 0 },
14033 { "forwProhibits", KSTAT_DATA_UINT32, 0 },
14034 { "udpInCksumErrs", KSTAT_DATA_UINT32, 0 },
14035 { "udpInOverflows", KSTAT_DATA_UINT32, 0 },
14036 { "rawipInOverflows", KSTAT_DATA_UINT32, 0 },
14037 { "ipsecInSucceeded", KSTAT_DATA_UINT32, 0 },
14038 { "ipsecInFailed", KSTAT_DATA_INT32, 0 },
14039 { "memberEntrySize", KSTAT_DATA_INT32, 0 },
14040 { "inIPv6", KSTAT_DATA_UINT32, 0 },
14041 { "outIPv6", KSTAT_DATA_UINT32, 0 },
14042 { "outSwitchIPv6", KSTAT_DATA_UINT32, 0 },
14043 };
14044
14045 ksp = kstat_create_netstack("ip", 0, "ip", "mib2", KSTAT_TYPE_NAMED,
14046 NUM_OF_FIELDS(ip_named_kstat_t), 0, stackid);
14047 if (ksp == NULL || ksp->ks_data == NULL)
14048 return (NULL);
14049
14050 template.forwarding.value.ui32 = WE_ARE_FORWARDING(ipst) ? 1:2;
14051 template.defaultTTL.value.ui32 = (uint32_t)ipst->ips_ip_def_ttl;
14052 template.reasmTimeout.value.ui32 = ipst->ips_ip_reassembly_timeout;
14053 template.addrEntrySize.value.i32 = sizeof (mib2_ipAddrEntry_t);
14054 template.routeEntrySize.value.i32 = sizeof (mib2_ipRouteEntry_t);
14055
14056 template.netToMediaEntrySize.value.i32 =
14057 sizeof (mib2_ipNetToMediaEntry_t);
14058
14059 template.memberEntrySize.value.i32 = sizeof (ipv6_member_t);
14060
14061 bcopy(&template, ksp->ks_data, sizeof (template));
14062 ksp->ks_update = ip_kstat_update;
14063 ksp->ks_private = (void *)(uintptr_t)stackid;
14064
14065 kstat_install(ksp);
14066 return (ksp);
14067 }
14068
14069 static void
14070 ip_kstat_fini(netstackid_t stackid, kstat_t *ksp)
14071 {
14072 if (ksp != NULL) {
14073 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
14074 kstat_delete_netstack(ksp, stackid);
14075 }
14076 }
14077
14078 static int
14079 ip_kstat_update(kstat_t *kp, int rw)
14080 {
14081 ip_named_kstat_t *ipkp;
14082 mib2_ipIfStatsEntry_t ipmib;
14083 ill_walk_context_t ctx;
14084 ill_t *ill;
14085 netstackid_t stackid = (zoneid_t)(uintptr_t)kp->ks_private;
14086 netstack_t *ns;
14087 ip_stack_t *ipst;
14088
14089 if (kp->ks_data == NULL)
14090 return (EIO);
14091
14092 if (rw == KSTAT_WRITE)
14093 return (EACCES);
14094
14095 ns = netstack_find_by_stackid(stackid);
14096 if (ns == NULL)
14097 return (-1);
14098 ipst = ns->netstack_ip;
14099 if (ipst == NULL) {
14100 netstack_rele(ns);
14101 return (-1);
14102 }
14103 ipkp = (ip_named_kstat_t *)kp->ks_data;
14104
14105 bcopy(&ipst->ips_ip_mib, &ipmib, sizeof (ipmib));
14106 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
14107 ill = ILL_START_WALK_V4(&ctx, ipst);
14108 for (; ill != NULL; ill = ill_next(&ctx, ill))
14109 ip_mib2_add_ip_stats(&ipmib, ill->ill_ip_mib);
14110 rw_exit(&ipst->ips_ill_g_lock);
14111
14112 ipkp->forwarding.value.ui32 = ipmib.ipIfStatsForwarding;
14113 ipkp->defaultTTL.value.ui32 = ipmib.ipIfStatsDefaultTTL;
14114 ipkp->inReceives.value.ui64 = ipmib.ipIfStatsHCInReceives;
14115 ipkp->inHdrErrors.value.ui32 = ipmib.ipIfStatsInHdrErrors;
14116 ipkp->inAddrErrors.value.ui32 = ipmib.ipIfStatsInAddrErrors;
14117 ipkp->forwDatagrams.value.ui64 = ipmib.ipIfStatsHCOutForwDatagrams;
14118 ipkp->inUnknownProtos.value.ui32 = ipmib.ipIfStatsInUnknownProtos;
14119 ipkp->inDiscards.value.ui32 = ipmib.ipIfStatsInDiscards;
14120 ipkp->inDelivers.value.ui64 = ipmib.ipIfStatsHCInDelivers;
14121 ipkp->outRequests.value.ui64 = ipmib.ipIfStatsHCOutRequests;
14122 ipkp->outDiscards.value.ui32 = ipmib.ipIfStatsOutDiscards;
14123 ipkp->outNoRoutes.value.ui32 = ipmib.ipIfStatsOutNoRoutes;
14124 ipkp->reasmTimeout.value.ui32 = ipst->ips_ip_reassembly_timeout;
14125 ipkp->reasmReqds.value.ui32 = ipmib.ipIfStatsReasmReqds;
14126 ipkp->reasmOKs.value.ui32 = ipmib.ipIfStatsReasmOKs;
14127 ipkp->reasmFails.value.ui32 = ipmib.ipIfStatsReasmFails;
14128 ipkp->fragOKs.value.ui32 = ipmib.ipIfStatsOutFragOKs;
14129 ipkp->fragFails.value.ui32 = ipmib.ipIfStatsOutFragFails;
14130 ipkp->fragCreates.value.ui32 = ipmib.ipIfStatsOutFragCreates;
14131
14132 ipkp->routingDiscards.value.ui32 = 0;
14133 ipkp->inErrs.value.ui32 = ipmib.tcpIfStatsInErrs;
14134 ipkp->noPorts.value.ui32 = ipmib.udpIfStatsNoPorts;
14135 ipkp->inCksumErrs.value.ui32 = ipmib.ipIfStatsInCksumErrs;
14136 ipkp->reasmDuplicates.value.ui32 = ipmib.ipIfStatsReasmDuplicates;
14137 ipkp->reasmPartDups.value.ui32 = ipmib.ipIfStatsReasmPartDups;
14138 ipkp->forwProhibits.value.ui32 = ipmib.ipIfStatsForwProhibits;
14139 ipkp->udpInCksumErrs.value.ui32 = ipmib.udpIfStatsInCksumErrs;
14140 ipkp->udpInOverflows.value.ui32 = ipmib.udpIfStatsInOverflows;
14141 ipkp->rawipInOverflows.value.ui32 = ipmib.rawipIfStatsInOverflows;
14142 ipkp->ipsecInSucceeded.value.ui32 = ipmib.ipsecIfStatsInSucceeded;
14143 ipkp->ipsecInFailed.value.i32 = ipmib.ipsecIfStatsInFailed;
14144
14145 ipkp->inIPv6.value.ui32 = ipmib.ipIfStatsInWrongIPVersion;
14146 ipkp->outIPv6.value.ui32 = ipmib.ipIfStatsOutWrongIPVersion;
14147 ipkp->outSwitchIPv6.value.ui32 = ipmib.ipIfStatsOutSwitchIPVersion;
14148
14149 netstack_rele(ns);
14150
14151 return (0);
14152 }
14153
14154 static void *
14155 icmp_kstat_init(netstackid_t stackid)
14156 {
14157 kstat_t *ksp;
14158
14159 icmp_named_kstat_t template = {
14160 { "inMsgs", KSTAT_DATA_UINT32 },
14161 { "inErrors", KSTAT_DATA_UINT32 },
14162 { "inDestUnreachs", KSTAT_DATA_UINT32 },
14163 { "inTimeExcds", KSTAT_DATA_UINT32 },
14164 { "inParmProbs", KSTAT_DATA_UINT32 },
14165 { "inSrcQuenchs", KSTAT_DATA_UINT32 },
14166 { "inRedirects", KSTAT_DATA_UINT32 },
14167 { "inEchos", KSTAT_DATA_UINT32 },
14168 { "inEchoReps", KSTAT_DATA_UINT32 },
14169 { "inTimestamps", KSTAT_DATA_UINT32 },
14170 { "inTimestampReps", KSTAT_DATA_UINT32 },
14171 { "inAddrMasks", KSTAT_DATA_UINT32 },
14172 { "inAddrMaskReps", KSTAT_DATA_UINT32 },
14173 { "outMsgs", KSTAT_DATA_UINT32 },
14174 { "outErrors", KSTAT_DATA_UINT32 },
14175 { "outDestUnreachs", KSTAT_DATA_UINT32 },
14176 { "outTimeExcds", KSTAT_DATA_UINT32 },
14177 { "outParmProbs", KSTAT_DATA_UINT32 },
14178 { "outSrcQuenchs", KSTAT_DATA_UINT32 },
14179 { "outRedirects", KSTAT_DATA_UINT32 },
14180 { "outEchos", KSTAT_DATA_UINT32 },
14181 { "outEchoReps", KSTAT_DATA_UINT32 },
14182 { "outTimestamps", KSTAT_DATA_UINT32 },
14183 { "outTimestampReps", KSTAT_DATA_UINT32 },
14184 { "outAddrMasks", KSTAT_DATA_UINT32 },
14185 { "outAddrMaskReps", KSTAT_DATA_UINT32 },
14186 { "inChksumErrs", KSTAT_DATA_UINT32 },
14187 { "inUnknowns", KSTAT_DATA_UINT32 },
14188 { "inFragNeeded", KSTAT_DATA_UINT32 },
14189 { "outFragNeeded", KSTAT_DATA_UINT32 },
14190 { "outDrops", KSTAT_DATA_UINT32 },
14191 { "inOverFlows", KSTAT_DATA_UINT32 },
14192 { "inBadRedirects", KSTAT_DATA_UINT32 },
14193 };
14194
14195 ksp = kstat_create_netstack("ip", 0, "icmp", "mib2", KSTAT_TYPE_NAMED,
14196 NUM_OF_FIELDS(icmp_named_kstat_t), 0, stackid);
14197 if (ksp == NULL || ksp->ks_data == NULL)
14198 return (NULL);
14199
14200 bcopy(&template, ksp->ks_data, sizeof (template));
14201
14202 ksp->ks_update = icmp_kstat_update;
14203 ksp->ks_private = (void *)(uintptr_t)stackid;
14204
14205 kstat_install(ksp);
14206 return (ksp);
14207 }
14208
14209 static void
14210 icmp_kstat_fini(netstackid_t stackid, kstat_t *ksp)
14211 {
14212 if (ksp != NULL) {
14213 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
14214 kstat_delete_netstack(ksp, stackid);
14215 }
14216 }
14217
14218 static int
14219 icmp_kstat_update(kstat_t *kp, int rw)
14220 {
14221 icmp_named_kstat_t *icmpkp;
14222 netstackid_t stackid = (zoneid_t)(uintptr_t)kp->ks_private;
14223 netstack_t *ns;
14224 ip_stack_t *ipst;
14225
14226 if (kp->ks_data == NULL)
14227 return (EIO);
14228
14229 if (rw == KSTAT_WRITE)
14230 return (EACCES);
14231
14232 ns = netstack_find_by_stackid(stackid);
14233 if (ns == NULL)
14234 return (-1);
14235 ipst = ns->netstack_ip;
14236 if (ipst == NULL) {
14237 netstack_rele(ns);
14238 return (-1);
14239 }
14240 icmpkp = (icmp_named_kstat_t *)kp->ks_data;
14241
14242 icmpkp->inMsgs.value.ui32 = ipst->ips_icmp_mib.icmpInMsgs;
14243 icmpkp->inErrors.value.ui32 = ipst->ips_icmp_mib.icmpInErrors;
14244 icmpkp->inDestUnreachs.value.ui32 =
14245 ipst->ips_icmp_mib.icmpInDestUnreachs;
14246 icmpkp->inTimeExcds.value.ui32 = ipst->ips_icmp_mib.icmpInTimeExcds;
14247 icmpkp->inParmProbs.value.ui32 = ipst->ips_icmp_mib.icmpInParmProbs;
14248 icmpkp->inSrcQuenchs.value.ui32 = ipst->ips_icmp_mib.icmpInSrcQuenchs;
14249 icmpkp->inRedirects.value.ui32 = ipst->ips_icmp_mib.icmpInRedirects;
14250 icmpkp->inEchos.value.ui32 = ipst->ips_icmp_mib.icmpInEchos;
14251 icmpkp->inEchoReps.value.ui32 = ipst->ips_icmp_mib.icmpInEchoReps;
14252 icmpkp->inTimestamps.value.ui32 = ipst->ips_icmp_mib.icmpInTimestamps;
14253 icmpkp->inTimestampReps.value.ui32 =
14254 ipst->ips_icmp_mib.icmpInTimestampReps;
14255 icmpkp->inAddrMasks.value.ui32 = ipst->ips_icmp_mib.icmpInAddrMasks;
14256 icmpkp->inAddrMaskReps.value.ui32 =
14257 ipst->ips_icmp_mib.icmpInAddrMaskReps;
14258 icmpkp->outMsgs.value.ui32 = ipst->ips_icmp_mib.icmpOutMsgs;
14259 icmpkp->outErrors.value.ui32 = ipst->ips_icmp_mib.icmpOutErrors;
14260 icmpkp->outDestUnreachs.value.ui32 =
14261 ipst->ips_icmp_mib.icmpOutDestUnreachs;
14262 icmpkp->outTimeExcds.value.ui32 = ipst->ips_icmp_mib.icmpOutTimeExcds;
14263 icmpkp->outParmProbs.value.ui32 = ipst->ips_icmp_mib.icmpOutParmProbs;
14264 icmpkp->outSrcQuenchs.value.ui32 =
14265 ipst->ips_icmp_mib.icmpOutSrcQuenchs;
14266 icmpkp->outRedirects.value.ui32 = ipst->ips_icmp_mib.icmpOutRedirects;
14267 icmpkp->outEchos.value.ui32 = ipst->ips_icmp_mib.icmpOutEchos;
14268 icmpkp->outEchoReps.value.ui32 = ipst->ips_icmp_mib.icmpOutEchoReps;
14269 icmpkp->outTimestamps.value.ui32 =
14270 ipst->ips_icmp_mib.icmpOutTimestamps;
14271 icmpkp->outTimestampReps.value.ui32 =
14272 ipst->ips_icmp_mib.icmpOutTimestampReps;
14273 icmpkp->outAddrMasks.value.ui32 =
14274 ipst->ips_icmp_mib.icmpOutAddrMasks;
14275 icmpkp->outAddrMaskReps.value.ui32 =
14276 ipst->ips_icmp_mib.icmpOutAddrMaskReps;
14277 icmpkp->inCksumErrs.value.ui32 = ipst->ips_icmp_mib.icmpInCksumErrs;
14278 icmpkp->inUnknowns.value.ui32 = ipst->ips_icmp_mib.icmpInUnknowns;
14279 icmpkp->inFragNeeded.value.ui32 = ipst->ips_icmp_mib.icmpInFragNeeded;
14280 icmpkp->outFragNeeded.value.ui32 =
14281 ipst->ips_icmp_mib.icmpOutFragNeeded;
14282 icmpkp->outDrops.value.ui32 = ipst->ips_icmp_mib.icmpOutDrops;
14283 icmpkp->inOverflows.value.ui32 = ipst->ips_icmp_mib.icmpInOverflows;
14284 icmpkp->inBadRedirects.value.ui32 =
14285 ipst->ips_icmp_mib.icmpInBadRedirects;
14286
14287 netstack_rele(ns);
14288 return (0);
14289 }
14290
14291 /*
14292 * This is the fanout function for raw socket opened for SCTP. Note
14293 * that it is called after SCTP checks that there is no socket which
14294 * wants a packet. Then before SCTP handles this out of the blue packet,
14295 * this function is called to see if there is any raw socket for SCTP.
14296 * If there is and it is bound to the correct address, the packet will
14297 * be sent to that socket. Note that only one raw socket can be bound to
14298 * a port. This is assured in ipcl_sctp_hash_insert();
14299 */
14300 void
14301 ip_fanout_sctp_raw(mblk_t *mp, ipha_t *ipha, ip6_t *ip6h, uint32_t ports,
14302 ip_recv_attr_t *ira)
14303 {
14304 conn_t *connp;
14305 queue_t *rq;
14306 boolean_t secure;
14307 ill_t *ill = ira->ira_ill;
14308 ip_stack_t *ipst = ill->ill_ipst;
14309 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
14310 sctp_stack_t *sctps = ipst->ips_netstack->netstack_sctp;
14311 iaflags_t iraflags = ira->ira_flags;
14312 ill_t *rill = ira->ira_rill;
14313
14314 secure = iraflags & IRAF_IPSEC_SECURE;
14315
14316 connp = ipcl_classify_raw(mp, IPPROTO_SCTP, ports, ipha, ip6h,
14317 ira, ipst);
14318 if (connp == NULL) {
14319 /*
14320 * Although raw sctp is not summed, OOB chunks must be.
14321 * Drop the packet here if the sctp checksum failed.
14322 */
14323 if (iraflags & IRAF_SCTP_CSUM_ERR) {
14324 SCTPS_BUMP_MIB(sctps, sctpChecksumError);
14325 freemsg(mp);
14326 return;
14327 }
14328 ira->ira_ill = ira->ira_rill = NULL;
14329 sctp_ootb_input(mp, ira, ipst);
14330 ira->ira_ill = ill;
14331 ira->ira_rill = rill;
14332 return;
14333 }
14334 rq = connp->conn_rq;
14335 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld : !canputnext(rq)) {
14336 CONN_DEC_REF(connp);
14337 BUMP_MIB(ill->ill_ip_mib, rawipIfStatsInOverflows);
14338 freemsg(mp);
14339 return;
14340 }
14341 if (((iraflags & IRAF_IS_IPV4) ?
14342 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
14343 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
14344 secure) {
14345 mp = ipsec_check_inbound_policy(mp, connp, ipha,
14346 ip6h, ira);
14347 if (mp == NULL) {
14348 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
14349 /* Note that mp is NULL */
14350 ip_drop_input("ipIfStatsInDiscards", mp, ill);
14351 CONN_DEC_REF(connp);
14352 return;
14353 }
14354 }
14355
14356 if (iraflags & IRAF_ICMP_ERROR) {
14357 (connp->conn_recvicmp)(connp, mp, NULL, ira);
14358 } else {
14359 ill_t *rill = ira->ira_rill;
14360
14361 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
14362 /* This is the SOCK_RAW, IPPROTO_SCTP case. */
14363 ira->ira_ill = ira->ira_rill = NULL;
14364 (connp->conn_recv)(connp, mp, NULL, ira);
14365 ira->ira_ill = ill;
14366 ira->ira_rill = rill;
14367 }
14368 CONN_DEC_REF(connp);
14369 }
14370
14371 /*
14372 * Free a packet that has the link-layer dl_unitdata_req_t or fast-path
14373 * header before the ip payload.
14374 */
14375 static void
14376 ip_xmit_flowctl_drop(ill_t *ill, mblk_t *mp, boolean_t is_fp_mp, int fp_mp_len)
14377 {
14378 int len = (mp->b_wptr - mp->b_rptr);
14379 mblk_t *ip_mp;
14380
14381 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14382 if (is_fp_mp || len != fp_mp_len) {
14383 if (len > fp_mp_len) {
14384 /*
14385 * fastpath header and ip header in the first mblk
14386 */
14387 mp->b_rptr += fp_mp_len;
14388 } else {
14389 /*
14390 * ip_xmit_attach_llhdr had to prepend an mblk to
14391 * attach the fastpath header before ip header.
14392 */
14393 ip_mp = mp->b_cont;
14394 freeb(mp);
14395 mp = ip_mp;
14396 mp->b_rptr += (fp_mp_len - len);
14397 }
14398 } else {
14399 ip_mp = mp->b_cont;
14400 freeb(mp);
14401 mp = ip_mp;
14402 }
14403 ip_drop_output("ipIfStatsOutDiscards - flow ctl", mp, ill);
14404 freemsg(mp);
14405 }
14406
14407 /*
14408 * Normal post fragmentation function.
14409 *
14410 * Send a packet using the passed in nce. This handles both IPv4 and IPv6
14411 * using the same state machine.
14412 *
14413 * We return an error on failure. In particular we return EWOULDBLOCK
14414 * when the driver flow controls. In that case this ensures that ip_wsrv runs
14415 * (currently by canputnext failure resulting in backenabling from GLD.)
14416 * This allows the callers of conn_ip_output() to use EWOULDBLOCK as an
14417 * indication that they can flow control until ip_wsrv() tells then to restart.
14418 *
14419 * If the nce passed by caller is incomplete, this function
14420 * queues the packet and if necessary, sends ARP request and bails.
14421 * If the Neighbor Cache passed is fully resolved, we simply prepend
14422 * the link-layer header to the packet, do ipsec hw acceleration
14423 * work if necessary, and send the packet out on the wire.
14424 */
14425 /* ARGSUSED6 */
14426 int
14427 ip_xmit(mblk_t *mp, nce_t *nce, iaflags_t ixaflags, uint_t pkt_len,
14428 uint32_t xmit_hint, zoneid_t szone, zoneid_t nolzid, uintptr_t *ixacookie)
14429 {
14430 queue_t *wq;
14431 ill_t *ill = nce->nce_ill;
14432 ip_stack_t *ipst = ill->ill_ipst;
14433 uint64_t delta;
14434 boolean_t isv6 = ill->ill_isv6;
14435 boolean_t fp_mp;
14436 ncec_t *ncec = nce->nce_common;
14437 int64_t now = LBOLT_FASTPATH64;
14438 boolean_t is_probe;
14439
14440 DTRACE_PROBE1(ip__xmit, nce_t *, nce);
14441
14442 ASSERT(mp != NULL);
14443 ASSERT(mp->b_datap->db_type == M_DATA);
14444 ASSERT(pkt_len == msgdsize(mp));
14445
14446 /*
14447 * If we have already been here and are coming back after ARP/ND.
14448 * the IXAF_NO_TRACE flag is set. We skip FW_HOOKS, DTRACE and ipobs
14449 * in that case since they have seen the packet when it came here
14450 * the first time.
14451 */
14452 if (ixaflags & IXAF_NO_TRACE)
14453 goto sendit;
14454
14455 if (ixaflags & IXAF_IS_IPV4) {
14456 ipha_t *ipha = (ipha_t *)mp->b_rptr;
14457
14458 ASSERT(!isv6);
14459 ASSERT(pkt_len == ntohs(((ipha_t *)mp->b_rptr)->ipha_length));
14460 if (HOOKS4_INTERESTED_PHYSICAL_OUT(ipst) &&
14461 !(ixaflags & IXAF_NO_PFHOOK)) {
14462 int error;
14463
14464 FW_HOOKS(ipst->ips_ip4_physical_out_event,
14465 ipst->ips_ipv4firewall_physical_out,
14466 NULL, ill, ipha, mp, mp, 0, ipst, error);
14467 DTRACE_PROBE1(ip4__physical__out__end,
14468 mblk_t *, mp);
14469 if (mp == NULL)
14470 return (error);
14471
14472 /* The length could have changed */
14473 pkt_len = msgdsize(mp);
14474 }
14475 if (ipst->ips_ip4_observe.he_interested) {
14476 /*
14477 * Note that for TX the zoneid is the sending
14478 * zone, whether or not MLP is in play.
14479 * Since the szone argument is the IP zoneid (i.e.,
14480 * zero for exclusive-IP zones) and ipobs wants
14481 * the system zoneid, we map it here.
14482 */
14483 szone = IP_REAL_ZONEID(szone, ipst);
14484
14485 /*
14486 * On the outbound path the destination zone will be
14487 * unknown as we're sending this packet out on the
14488 * wire.
14489 */
14490 ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone, ALL_ZONES,
14491 ill, ipst);
14492 }
14493 DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
14494 void_ip_t *, ipha, __dtrace_ipsr_ill_t *, ill,
14495 ipha_t *, ipha, ip6_t *, NULL, int, 0);
14496 } else {
14497 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
14498
14499 ASSERT(isv6);
14500 ASSERT(pkt_len ==
14501 ntohs(((ip6_t *)mp->b_rptr)->ip6_plen) + IPV6_HDR_LEN);
14502 if (HOOKS6_INTERESTED_PHYSICAL_OUT(ipst) &&
14503 !(ixaflags & IXAF_NO_PFHOOK)) {
14504 int error;
14505
14506 FW_HOOKS6(ipst->ips_ip6_physical_out_event,
14507 ipst->ips_ipv6firewall_physical_out,
14508 NULL, ill, ip6h, mp, mp, 0, ipst, error);
14509 DTRACE_PROBE1(ip6__physical__out__end,
14510 mblk_t *, mp);
14511 if (mp == NULL)
14512 return (error);
14513
14514 /* The length could have changed */
14515 pkt_len = msgdsize(mp);
14516 }
14517 if (ipst->ips_ip6_observe.he_interested) {
14518 /* See above */
14519 szone = IP_REAL_ZONEID(szone, ipst);
14520
14521 ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone, ALL_ZONES,
14522 ill, ipst);
14523 }
14524 DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
14525 void_ip_t *, ip6h, __dtrace_ipsr_ill_t *, ill,
14526 ipha_t *, NULL, ip6_t *, ip6h, int, 0);
14527 }
14528
14529 sendit:
14530 /*
14531 * We check the state without a lock because the state can never
14532 * move "backwards" to initial or incomplete.
14533 */
14534 switch (ncec->ncec_state) {
14535 case ND_REACHABLE:
14536 case ND_STALE:
14537 case ND_DELAY:
14538 case ND_PROBE:
14539 mp = ip_xmit_attach_llhdr(mp, nce);
14540 if (mp == NULL) {
14541 /*
14542 * ip_xmit_attach_llhdr has increased
14543 * ipIfStatsOutDiscards and called ip_drop_output()
14544 */
14545 return (ENOBUFS);
14546 }
14547 /*
14548 * check if nce_fastpath completed and we tagged on a
14549 * copy of nce_fp_mp in ip_xmit_attach_llhdr().
14550 */
14551 fp_mp = (mp->b_datap->db_type == M_DATA);
14552
14553 if (fp_mp &&
14554 (ill->ill_capabilities & ILL_CAPAB_DLD_DIRECT)) {
14555 ill_dld_direct_t *idd;
14556
14557 idd = &ill->ill_dld_capab->idc_direct;
14558 /*
14559 * Send the packet directly to DLD, where it
14560 * may be queued depending on the availability
14561 * of transmit resources at the media layer.
14562 * Return value should be taken into
14563 * account and flow control the TCP.
14564 */
14565 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
14566 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
14567 pkt_len);
14568
14569 if (ixaflags & IXAF_NO_DEV_FLOW_CTL) {
14570 (void) idd->idd_tx_df(idd->idd_tx_dh, mp,
14571 (uintptr_t)xmit_hint, IP_DROP_ON_NO_DESC);
14572 } else {
14573 uintptr_t cookie;
14574
14575 if ((cookie = idd->idd_tx_df(idd->idd_tx_dh,
14576 mp, (uintptr_t)xmit_hint, 0)) != 0) {
14577 if (ixacookie != NULL)
14578 *ixacookie = cookie;
14579 return (EWOULDBLOCK);
14580 }
14581 }
14582 } else {
14583 wq = ill->ill_wq;
14584
14585 if (!(ixaflags & IXAF_NO_DEV_FLOW_CTL) &&
14586 !canputnext(wq)) {
14587 if (ixacookie != NULL)
14588 *ixacookie = 0;
14589 ip_xmit_flowctl_drop(ill, mp, fp_mp,
14590 nce->nce_fp_mp != NULL ?
14591 MBLKL(nce->nce_fp_mp) : 0);
14592 return (EWOULDBLOCK);
14593 }
14594 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
14595 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
14596 pkt_len);
14597 putnext(wq, mp);
14598 }
14599
14600 /*
14601 * The rest of this function implements Neighbor Unreachability
14602 * detection. Determine if the ncec is eligible for NUD.
14603 */
14604 if (ncec->ncec_flags & NCE_F_NONUD)
14605 return (0);
14606
14607 ASSERT(ncec->ncec_state != ND_INCOMPLETE);
14608
14609 /*
14610 * Check for upper layer advice
14611 */
14612 if (ixaflags & IXAF_REACH_CONF) {
14613 timeout_id_t tid;
14614
14615 /*
14616 * It should be o.k. to check the state without
14617 * a lock here, at most we lose an advice.
14618 */
14619 ncec->ncec_last = TICK_TO_MSEC(now);
14620 if (ncec->ncec_state != ND_REACHABLE) {
14621 mutex_enter(&ncec->ncec_lock);
14622 ncec->ncec_state = ND_REACHABLE;
14623 tid = ncec->ncec_timeout_id;
14624 ncec->ncec_timeout_id = 0;
14625 mutex_exit(&ncec->ncec_lock);
14626 (void) untimeout(tid);
14627 if (ip_debug > 2) {
14628 /* ip1dbg */
14629 pr_addr_dbg("ip_xmit: state"
14630 " for %s changed to"
14631 " REACHABLE\n", AF_INET6,
14632 &ncec->ncec_addr);
14633 }
14634 }
14635 return (0);
14636 }
14637
14638 delta = TICK_TO_MSEC(now) - ncec->ncec_last;
14639 ip1dbg(("ip_xmit: delta = %" PRId64
14640 " ill_reachable_time = %d \n", delta,
14641 ill->ill_reachable_time));
14642 if (delta > (uint64_t)ill->ill_reachable_time) {
14643 mutex_enter(&ncec->ncec_lock);
14644 switch (ncec->ncec_state) {
14645 case ND_REACHABLE:
14646 ASSERT((ncec->ncec_flags & NCE_F_NONUD) == 0);
14647 /* FALLTHROUGH */
14648 case ND_STALE:
14649 /*
14650 * ND_REACHABLE is identical to
14651 * ND_STALE in this specific case. If
14652 * reachable time has expired for this
14653 * neighbor (delta is greater than
14654 * reachable time), conceptually, the
14655 * neighbor cache is no longer in
14656 * REACHABLE state, but already in
14657 * STALE state. So the correct
14658 * transition here is to ND_DELAY.
14659 */
14660 ncec->ncec_state = ND_DELAY;
14661 mutex_exit(&ncec->ncec_lock);
14662 nce_restart_timer(ncec,
14663 ipst->ips_delay_first_probe_time);
14664 if (ip_debug > 3) {
14665 /* ip2dbg */
14666 pr_addr_dbg("ip_xmit: state"
14667 " for %s changed to"
14668 " DELAY\n", AF_INET6,
14669 &ncec->ncec_addr);
14670 }
14671 break;
14672 case ND_DELAY:
14673 case ND_PROBE:
14674 mutex_exit(&ncec->ncec_lock);
14675 /* Timers have already started */
14676 break;
14677 case ND_UNREACHABLE:
14678 /*
14679 * nce_timer has detected that this ncec
14680 * is unreachable and initiated deleting
14681 * this ncec.
14682 * This is a harmless race where we found the
14683 * ncec before it was deleted and have
14684 * just sent out a packet using this
14685 * unreachable ncec.
14686 */
14687 mutex_exit(&ncec->ncec_lock);
14688 break;
14689 default:
14690 ASSERT(0);
14691 mutex_exit(&ncec->ncec_lock);
14692 }
14693 }
14694 return (0);
14695
14696 case ND_INCOMPLETE:
14697 /*
14698 * the state could have changed since we didn't hold the lock.
14699 * Re-verify state under lock.
14700 */
14701 is_probe = ipmp_packet_is_probe(mp, nce->nce_ill);
14702 mutex_enter(&ncec->ncec_lock);
14703 if (NCE_ISREACHABLE(ncec)) {
14704 mutex_exit(&ncec->ncec_lock);
14705 goto sendit;
14706 }
14707 /* queue the packet */
14708 nce_queue_mp(ncec, mp, is_probe);
14709 mutex_exit(&ncec->ncec_lock);
14710 DTRACE_PROBE2(ip__xmit__incomplete,
14711 (ncec_t *), ncec, (mblk_t *), mp);
14712 return (0);
14713
14714 case ND_INITIAL:
14715 /*
14716 * State could have changed since we didn't hold the lock, so
14717 * re-verify state.
14718 */
14719 is_probe = ipmp_packet_is_probe(mp, nce->nce_ill);
14720 mutex_enter(&ncec->ncec_lock);
14721 if (NCE_ISREACHABLE(ncec)) {
14722 mutex_exit(&ncec->ncec_lock);
14723 goto sendit;
14724 }
14725 nce_queue_mp(ncec, mp, is_probe);
14726 if (ncec->ncec_state == ND_INITIAL) {
14727 ncec->ncec_state = ND_INCOMPLETE;
14728 mutex_exit(&ncec->ncec_lock);
14729 /*
14730 * figure out the source we want to use
14731 * and resolve it.
14732 */
14733 ip_ndp_resolve(ncec);
14734 } else {
14735 mutex_exit(&ncec->ncec_lock);
14736 }
14737 return (0);
14738
14739 case ND_UNREACHABLE:
14740 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14741 ip_drop_output("ipIfStatsOutDiscards - ND_UNREACHABLE",
14742 mp, ill);
14743 freemsg(mp);
14744 return (0);
14745
14746 default:
14747 ASSERT(0);
14748 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14749 ip_drop_output("ipIfStatsOutDiscards - ND_other",
14750 mp, ill);
14751 freemsg(mp);
14752 return (ENETUNREACH);
14753 }
14754 }
14755
14756 /*
14757 * Return B_TRUE if the buffers differ in length or content.
14758 * This is used for comparing extension header buffers.
14759 * Note that an extension header would be declared different
14760 * even if all that changed was the next header value in that header i.e.
14761 * what really changed is the next extension header.
14762 */
14763 boolean_t
14764 ip_cmpbuf(const void *abuf, uint_t alen, boolean_t b_valid, const void *bbuf,
14765 uint_t blen)
14766 {
14767 if (!b_valid)
14768 blen = 0;
14769
14770 if (alen != blen)
14771 return (B_TRUE);
14772 if (alen == 0)
14773 return (B_FALSE); /* Both zero length */
14774 return (bcmp(abuf, bbuf, alen));
14775 }
14776
14777 /*
14778 * Preallocate memory for ip_savebuf(). Returns B_TRUE if ok.
14779 * Return B_FALSE if memory allocation fails - don't change any state!
14780 */
14781 boolean_t
14782 ip_allocbuf(void **dstp, uint_t *dstlenp, boolean_t src_valid,
14783 const void *src, uint_t srclen)
14784 {
14785 void *dst;
14786
14787 if (!src_valid)
14788 srclen = 0;
14789
14790 ASSERT(*dstlenp == 0);
14791 if (src != NULL && srclen != 0) {
14792 dst = mi_alloc(srclen, BPRI_MED);
14793 if (dst == NULL)
14794 return (B_FALSE);
14795 } else {
14796 dst = NULL;
14797 }
14798 if (*dstp != NULL)
14799 mi_free(*dstp);
14800 *dstp = dst;
14801 *dstlenp = dst == NULL ? 0 : srclen;
14802 return (B_TRUE);
14803 }
14804
14805 /*
14806 * Replace what is in *dst, *dstlen with the source.
14807 * Assumes ip_allocbuf has already been called.
14808 */
14809 void
14810 ip_savebuf(void **dstp, uint_t *dstlenp, boolean_t src_valid,
14811 const void *src, uint_t srclen)
14812 {
14813 if (!src_valid)
14814 srclen = 0;
14815
14816 ASSERT(*dstlenp == srclen);
14817 if (src != NULL && srclen != 0)
14818 bcopy(src, *dstp, srclen);
14819 }
14820
14821 /*
14822 * Free the storage pointed to by the members of an ip_pkt_t.
14823 */
14824 void
14825 ip_pkt_free(ip_pkt_t *ipp)
14826 {
14827 uint_t fields = ipp->ipp_fields;
14828
14829 if (fields & IPPF_HOPOPTS) {
14830 kmem_free(ipp->ipp_hopopts, ipp->ipp_hopoptslen);
14831 ipp->ipp_hopopts = NULL;
14832 ipp->ipp_hopoptslen = 0;
14833 }
14834 if (fields & IPPF_RTHDRDSTOPTS) {
14835 kmem_free(ipp->ipp_rthdrdstopts, ipp->ipp_rthdrdstoptslen);
14836 ipp->ipp_rthdrdstopts = NULL;
14837 ipp->ipp_rthdrdstoptslen = 0;
14838 }
14839 if (fields & IPPF_DSTOPTS) {
14840 kmem_free(ipp->ipp_dstopts, ipp->ipp_dstoptslen);
14841 ipp->ipp_dstopts = NULL;
14842 ipp->ipp_dstoptslen = 0;
14843 }
14844 if (fields & IPPF_RTHDR) {
14845 kmem_free(ipp->ipp_rthdr, ipp->ipp_rthdrlen);
14846 ipp->ipp_rthdr = NULL;
14847 ipp->ipp_rthdrlen = 0;
14848 }
14849 if (fields & IPPF_IPV4_OPTIONS) {
14850 kmem_free(ipp->ipp_ipv4_options, ipp->ipp_ipv4_options_len);
14851 ipp->ipp_ipv4_options = NULL;
14852 ipp->ipp_ipv4_options_len = 0;
14853 }
14854 if (fields & IPPF_LABEL_V4) {
14855 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
14856 ipp->ipp_label_v4 = NULL;
14857 ipp->ipp_label_len_v4 = 0;
14858 }
14859 if (fields & IPPF_LABEL_V6) {
14860 kmem_free(ipp->ipp_label_v6, ipp->ipp_label_len_v6);
14861 ipp->ipp_label_v6 = NULL;
14862 ipp->ipp_label_len_v6 = 0;
14863 }
14864 ipp->ipp_fields &= ~(IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14865 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6);
14866 }
14867
14868 /*
14869 * Copy from src to dst and allocate as needed.
14870 * Returns zero or ENOMEM.
14871 *
14872 * The caller must initialize dst to zero.
14873 */
14874 int
14875 ip_pkt_copy(ip_pkt_t *src, ip_pkt_t *dst, int kmflag)
14876 {
14877 uint_t fields = src->ipp_fields;
14878
14879 /* Start with fields that don't require memory allocation */
14880 dst->ipp_fields = fields &
14881 ~(IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14882 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6);
14883
14884 dst->ipp_addr = src->ipp_addr;
14885 dst->ipp_unicast_hops = src->ipp_unicast_hops;
14886 dst->ipp_hoplimit = src->ipp_hoplimit;
14887 dst->ipp_tclass = src->ipp_tclass;
14888 dst->ipp_type_of_service = src->ipp_type_of_service;
14889
14890 if (!(fields & (IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14891 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6)))
14892 return (0);
14893
14894 if (fields & IPPF_HOPOPTS) {
14895 dst->ipp_hopopts = kmem_alloc(src->ipp_hopoptslen, kmflag);
14896 if (dst->ipp_hopopts == NULL) {
14897 ip_pkt_free(dst);
14898 return (ENOMEM);
14899 }
14900 dst->ipp_fields |= IPPF_HOPOPTS;
14901 bcopy(src->ipp_hopopts, dst->ipp_hopopts,
14902 src->ipp_hopoptslen);
14903 dst->ipp_hopoptslen = src->ipp_hopoptslen;
14904 }
14905 if (fields & IPPF_RTHDRDSTOPTS) {
14906 dst->ipp_rthdrdstopts = kmem_alloc(src->ipp_rthdrdstoptslen,
14907 kmflag);
14908 if (dst->ipp_rthdrdstopts == NULL) {
14909 ip_pkt_free(dst);
14910 return (ENOMEM);
14911 }
14912 dst->ipp_fields |= IPPF_RTHDRDSTOPTS;
14913 bcopy(src->ipp_rthdrdstopts, dst->ipp_rthdrdstopts,
14914 src->ipp_rthdrdstoptslen);
14915 dst->ipp_rthdrdstoptslen = src->ipp_rthdrdstoptslen;
14916 }
14917 if (fields & IPPF_DSTOPTS) {
14918 dst->ipp_dstopts = kmem_alloc(src->ipp_dstoptslen, kmflag);
14919 if (dst->ipp_dstopts == NULL) {
14920 ip_pkt_free(dst);
14921 return (ENOMEM);
14922 }
14923 dst->ipp_fields |= IPPF_DSTOPTS;
14924 bcopy(src->ipp_dstopts, dst->ipp_dstopts,
14925 src->ipp_dstoptslen);
14926 dst->ipp_dstoptslen = src->ipp_dstoptslen;
14927 }
14928 if (fields & IPPF_RTHDR) {
14929 dst->ipp_rthdr = kmem_alloc(src->ipp_rthdrlen, kmflag);
14930 if (dst->ipp_rthdr == NULL) {
14931 ip_pkt_free(dst);
14932 return (ENOMEM);
14933 }
14934 dst->ipp_fields |= IPPF_RTHDR;
14935 bcopy(src->ipp_rthdr, dst->ipp_rthdr,
14936 src->ipp_rthdrlen);
14937 dst->ipp_rthdrlen = src->ipp_rthdrlen;
14938 }
14939 if (fields & IPPF_IPV4_OPTIONS) {
14940 dst->ipp_ipv4_options = kmem_alloc(src->ipp_ipv4_options_len,
14941 kmflag);
14942 if (dst->ipp_ipv4_options == NULL) {
14943 ip_pkt_free(dst);
14944 return (ENOMEM);
14945 }
14946 dst->ipp_fields |= IPPF_IPV4_OPTIONS;
14947 bcopy(src->ipp_ipv4_options, dst->ipp_ipv4_options,
14948 src->ipp_ipv4_options_len);
14949 dst->ipp_ipv4_options_len = src->ipp_ipv4_options_len;
14950 }
14951 if (fields & IPPF_LABEL_V4) {
14952 dst->ipp_label_v4 = kmem_alloc(src->ipp_label_len_v4, kmflag);
14953 if (dst->ipp_label_v4 == NULL) {
14954 ip_pkt_free(dst);
14955 return (ENOMEM);
14956 }
14957 dst->ipp_fields |= IPPF_LABEL_V4;
14958 bcopy(src->ipp_label_v4, dst->ipp_label_v4,
14959 src->ipp_label_len_v4);
14960 dst->ipp_label_len_v4 = src->ipp_label_len_v4;
14961 }
14962 if (fields & IPPF_LABEL_V6) {
14963 dst->ipp_label_v6 = kmem_alloc(src->ipp_label_len_v6, kmflag);
14964 if (dst->ipp_label_v6 == NULL) {
14965 ip_pkt_free(dst);
14966 return (ENOMEM);
14967 }
14968 dst->ipp_fields |= IPPF_LABEL_V6;
14969 bcopy(src->ipp_label_v6, dst->ipp_label_v6,
14970 src->ipp_label_len_v6);
14971 dst->ipp_label_len_v6 = src->ipp_label_len_v6;
14972 }
14973 if (fields & IPPF_FRAGHDR) {
14974 dst->ipp_fraghdr = kmem_alloc(src->ipp_fraghdrlen, kmflag);
14975 if (dst->ipp_fraghdr == NULL) {
14976 ip_pkt_free(dst);
14977 return (ENOMEM);
14978 }
14979 dst->ipp_fields |= IPPF_FRAGHDR;
14980 bcopy(src->ipp_fraghdr, dst->ipp_fraghdr,
14981 src->ipp_fraghdrlen);
14982 dst->ipp_fraghdrlen = src->ipp_fraghdrlen;
14983 }
14984 return (0);
14985 }
14986
14987 /*
14988 * Returns INADDR_ANY if no source route
14989 */
14990 ipaddr_t
14991 ip_pkt_source_route_v4(const ip_pkt_t *ipp)
14992 {
14993 ipaddr_t nexthop = INADDR_ANY;
14994 ipoptp_t opts;
14995 uchar_t *opt;
14996 uint8_t optval;
14997 uint8_t optlen;
14998 uint32_t totallen;
14999
15000 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
15001 return (INADDR_ANY);
15002
15003 totallen = ipp->ipp_ipv4_options_len;
15004 if (totallen & 0x3)
15005 return (INADDR_ANY);
15006
15007 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
15008 optval != IPOPT_EOL;
15009 optval = ipoptp_next(&opts)) {
15010 opt = opts.ipoptp_cur;
15011 switch (optval) {
15012 uint8_t off;
15013 case IPOPT_SSRR:
15014 case IPOPT_LSRR:
15015 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
15016 break;
15017 }
15018 optlen = opts.ipoptp_len;
15019 off = opt[IPOPT_OFFSET];
15020 off--;
15021 if (optlen < IP_ADDR_LEN ||
15022 off > optlen - IP_ADDR_LEN) {
15023 /* End of source route */
15024 break;
15025 }
15026 bcopy((char *)opt + off, &nexthop, IP_ADDR_LEN);
15027 if (nexthop == htonl(INADDR_LOOPBACK)) {
15028 /* Ignore */
15029 nexthop = INADDR_ANY;
15030 break;
15031 }
15032 break;
15033 }
15034 }
15035 return (nexthop);
15036 }
15037
15038 /*
15039 * Reverse a source route.
15040 */
15041 void
15042 ip_pkt_source_route_reverse_v4(ip_pkt_t *ipp)
15043 {
15044 ipaddr_t tmp;
15045 ipoptp_t opts;
15046 uchar_t *opt;
15047 uint8_t optval;
15048 uint32_t totallen;
15049
15050 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
15051 return;
15052
15053 totallen = ipp->ipp_ipv4_options_len;
15054 if (totallen & 0x3)
15055 return;
15056
15057 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
15058 optval != IPOPT_EOL;
15059 optval = ipoptp_next(&opts)) {
15060 uint8_t off1, off2;
15061
15062 opt = opts.ipoptp_cur;
15063 switch (optval) {
15064 case IPOPT_SSRR:
15065 case IPOPT_LSRR:
15066 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
15067 break;
15068 }
15069 off1 = IPOPT_MINOFF_SR - 1;
15070 off2 = opt[IPOPT_OFFSET] - IP_ADDR_LEN - 1;
15071 while (off2 > off1) {
15072 bcopy(opt + off2, &tmp, IP_ADDR_LEN);
15073 bcopy(opt + off1, opt + off2, IP_ADDR_LEN);
15074 bcopy(&tmp, opt + off2, IP_ADDR_LEN);
15075 off2 -= IP_ADDR_LEN;
15076 off1 += IP_ADDR_LEN;
15077 }
15078 opt[IPOPT_OFFSET] = IPOPT_MINOFF_SR;
15079 break;
15080 }
15081 }
15082 }
15083
15084 /*
15085 * Returns NULL if no routing header
15086 */
15087 in6_addr_t *
15088 ip_pkt_source_route_v6(const ip_pkt_t *ipp)
15089 {
15090 in6_addr_t *nexthop = NULL;
15091 ip6_rthdr0_t *rthdr;
15092
15093 if (!(ipp->ipp_fields & IPPF_RTHDR))
15094 return (NULL);
15095
15096 rthdr = (ip6_rthdr0_t *)ipp->ipp_rthdr;
15097 if (rthdr->ip6r0_segleft == 0)
15098 return (NULL);
15099
15100 nexthop = (in6_addr_t *)((char *)rthdr + sizeof (*rthdr));
15101 return (nexthop);
15102 }
15103
15104 zoneid_t
15105 ip_get_zoneid_v4(ipaddr_t addr, mblk_t *mp, ip_recv_attr_t *ira,
15106 zoneid_t lookup_zoneid)
15107 {
15108 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
15109 ire_t *ire;
15110 int ire_flags = MATCH_IRE_TYPE;
15111 zoneid_t zoneid = ALL_ZONES;
15112
15113 if (is_system_labeled() && !tsol_can_accept_raw(mp, ira, B_FALSE))
15114 return (ALL_ZONES);
15115
15116 if (lookup_zoneid != ALL_ZONES)
15117 ire_flags |= MATCH_IRE_ZONEONLY;
15118 ire = ire_ftable_lookup_v4(addr, 0, 0, IRE_LOCAL | IRE_LOOPBACK,
15119 NULL, lookup_zoneid, NULL, ire_flags, 0, ipst, NULL);
15120 if (ire != NULL) {
15121 zoneid = IP_REAL_ZONEID(ire->ire_zoneid, ipst);
15122 ire_refrele(ire);
15123 }
15124 return (zoneid);
15125 }
15126
15127 zoneid_t
15128 ip_get_zoneid_v6(in6_addr_t *addr, mblk_t *mp, const ill_t *ill,
15129 ip_recv_attr_t *ira, zoneid_t lookup_zoneid)
15130 {
15131 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
15132 ire_t *ire;
15133 int ire_flags = MATCH_IRE_TYPE;
15134 zoneid_t zoneid = ALL_ZONES;
15135
15136 if (is_system_labeled() && !tsol_can_accept_raw(mp, ira, B_FALSE))
15137 return (ALL_ZONES);
15138
15139 if (IN6_IS_ADDR_LINKLOCAL(addr))
15140 ire_flags |= MATCH_IRE_ILL;
15141
15142 if (lookup_zoneid != ALL_ZONES)
15143 ire_flags |= MATCH_IRE_ZONEONLY;
15144 ire = ire_ftable_lookup_v6(addr, NULL, NULL, IRE_LOCAL | IRE_LOOPBACK,
15145 ill, lookup_zoneid, NULL, ire_flags, 0, ipst, NULL);
15146 if (ire != NULL) {
15147 zoneid = IP_REAL_ZONEID(ire->ire_zoneid, ipst);
15148 ire_refrele(ire);
15149 }
15150 return (zoneid);
15151 }
15152
15153 /*
15154 * IP obserability hook support functions.
15155 */
15156 static void
15157 ipobs_init(ip_stack_t *ipst)
15158 {
15159 netid_t id;
15160
15161 id = net_getnetidbynetstackid(ipst->ips_netstack->netstack_stackid);
15162
15163 ipst->ips_ip4_observe_pr = net_protocol_lookup(id, NHF_INET);
15164 VERIFY(ipst->ips_ip4_observe_pr != NULL);
15165
15166 ipst->ips_ip6_observe_pr = net_protocol_lookup(id, NHF_INET6);
15167 VERIFY(ipst->ips_ip6_observe_pr != NULL);
15168 }
15169
15170 static void
15171 ipobs_fini(ip_stack_t *ipst)
15172 {
15173
15174 VERIFY(net_protocol_release(ipst->ips_ip4_observe_pr) == 0);
15175 VERIFY(net_protocol_release(ipst->ips_ip6_observe_pr) == 0);
15176 }
15177
15178 /*
15179 * hook_pkt_observe_t is composed in network byte order so that the
15180 * entire mblk_t chain handed into hook_run can be used as-is.
15181 * The caveat is that use of the fields, such as the zone fields,
15182 * requires conversion into host byte order first.
15183 */
15184 void
15185 ipobs_hook(mblk_t *mp, int htype, zoneid_t zsrc, zoneid_t zdst,
15186 const ill_t *ill, ip_stack_t *ipst)
15187 {
15188 hook_pkt_observe_t *hdr;
15189 uint64_t grifindex;
15190 mblk_t *imp;
15191
15192 imp = allocb(sizeof (*hdr), BPRI_HI);
15193 if (imp == NULL)
15194 return;
15195
15196 hdr = (hook_pkt_observe_t *)imp->b_rptr;
15197 /*
15198 * b_wptr is set to make the apparent size of the data in the mblk_t
15199 * to exclude the pointers at the end of hook_pkt_observer_t.
15200 */
15201 imp->b_wptr = imp->b_rptr + sizeof (dl_ipnetinfo_t);
15202 imp->b_cont = mp;
15203
15204 ASSERT(DB_TYPE(mp) == M_DATA);
15205
15206 if (IS_UNDER_IPMP(ill))
15207 grifindex = ipmp_ill_get_ipmp_ifindex(ill);
15208 else
15209 grifindex = 0;
15210
15211 hdr->hpo_version = 1;
15212 hdr->hpo_htype = htons(htype);
15213 hdr->hpo_pktlen = htonl((ulong_t)msgdsize(mp));
15214 hdr->hpo_ifindex = htonl(ill->ill_phyint->phyint_ifindex);
15215 hdr->hpo_grifindex = htonl(grifindex);
15216 hdr->hpo_zsrc = htonl(zsrc);
15217 hdr->hpo_zdst = htonl(zdst);
15218 hdr->hpo_pkt = imp;
15219 hdr->hpo_ctx = ipst->ips_netstack;
15220
15221 if (ill->ill_isv6) {
15222 hdr->hpo_family = AF_INET6;
15223 (void) hook_run(ipst->ips_ipv6_net_data->netd_hooks,
15224 ipst->ips_ipv6observing, (hook_data_t)hdr);
15225 } else {
15226 hdr->hpo_family = AF_INET;
15227 (void) hook_run(ipst->ips_ipv4_net_data->netd_hooks,
15228 ipst->ips_ipv4observing, (hook_data_t)hdr);
15229 }
15230
15231 imp->b_cont = NULL;
15232 freemsg(imp);
15233 }
15234
15235 /*
15236 * Utility routine that checks if `v4srcp' is a valid address on underlying
15237 * interface `ill'. If `ipifp' is non-NULL, it's set to a held ipif
15238 * associated with `v4srcp' on success. NOTE: if this is not called from
15239 * inside the IPSQ (ill_g_lock is not held), `ill' may be removed from the
15240 * group during or after this lookup.
15241 */
15242 boolean_t
15243 ipif_lookup_testaddr_v4(ill_t *ill, const in_addr_t *v4srcp, ipif_t **ipifp)
15244 {
15245 ipif_t *ipif;
15246
15247 ipif = ipif_lookup_addr_exact(*v4srcp, ill, ill->ill_ipst);
15248 if (ipif != NULL) {
15249 if (ipifp != NULL)
15250 *ipifp = ipif;
15251 else
15252 ipif_refrele(ipif);
15253 return (B_TRUE);
15254 }
15255
15256 ip1dbg(("ipif_lookup_testaddr_v4: cannot find ipif for src %x\n",
15257 *v4srcp));
15258 return (B_FALSE);
15259 }
15260
15261 /*
15262 * Transport protocol call back function for CPU state change.
15263 */
15264 /* ARGSUSED */
15265 static int
15266 ip_tp_cpu_update(cpu_setup_t what, int id, void *arg)
15267 {
15268 processorid_t cpu_seqid;
15269 netstack_handle_t nh;
15270 netstack_t *ns;
15271
15272 ASSERT(MUTEX_HELD(&cpu_lock));
15273
15274 switch (what) {
15275 case CPU_CONFIG:
15276 case CPU_ON:
15277 case CPU_INIT:
15278 case CPU_CPUPART_IN:
15279 cpu_seqid = cpu[id]->cpu_seqid;
15280 netstack_next_init(&nh);
15281 while ((ns = netstack_next(&nh)) != NULL) {
15282 tcp_stack_cpu_add(ns->netstack_tcp, cpu_seqid);
15283 sctp_stack_cpu_add(ns->netstack_sctp, cpu_seqid);
15284 udp_stack_cpu_add(ns->netstack_udp, cpu_seqid);
15285 netstack_rele(ns);
15286 }
15287 netstack_next_fini(&nh);
15288 break;
15289 case CPU_UNCONFIG:
15290 case CPU_OFF:
15291 case CPU_CPUPART_OUT:
15292 /*
15293 * Nothing to do. We don't remove the per CPU stats from
15294 * the IP stack even when the CPU goes offline.
15295 */
15296 break;
15297 default:
15298 break;
15299 }
15300 return (0);
15301 }