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) 2018 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
99 #include <net/pfkeyv2.h>
100 #include <inet/sadb.h>
101 #include <inet/ipsec_impl.h>
102 #include <inet/iptun/iptun_impl.h>
103 #include <inet/ipdrop.h>
104 #include <inet/ip_netinfo.h>
105 #include <inet/ilb_ip.h>
106
107 #include <sys/ethernet.h>
108 #include <net/if_types.h>
109 #include <sys/cpuvar.h>
110
111 #include <ipp/ipp.h>
112 #include <ipp/ipp_impl.h>
113 #include <ipp/ipgpc/ipgpc.h>
114
115 #include <sys/pattr.h>
116 #include <inet/ipclassifier.h>
117 #include <inet/sctp_ip.h>
118 #include <inet/sctp/sctp_impl.h>
119 #include <inet/udp_impl.h>
120 #include <inet/rawip_impl.h>
121 #include <inet/rts_impl.h>
122
123 #include <sys/tsol/label.h>
124 #include <sys/tsol/tnet.h>
125
126 #include <sys/squeue_impl.h>
127 #include <inet/ip_arp.h>
128
129 #include <sys/clock_impl.h> /* For LBOLT_FASTPATH{,64} */
130
131 /*
132 * Values for squeue switch:
133 * IP_SQUEUE_ENTER_NODRAIN: SQ_NODRAIN
134 * IP_SQUEUE_ENTER: SQ_PROCESS
135 * IP_SQUEUE_FILL: SQ_FILL
136 */
137 int ip_squeue_enter = IP_SQUEUE_ENTER; /* Setable in /etc/system */
138
139 int ip_squeue_flag;
140
141 /*
142 * Setable in /etc/system
143 */
144 int ip_poll_normal_ms = 100;
145 int ip_poll_normal_ticks = 0;
146 int ip_modclose_ackwait_ms = 3000;
147
148 /*
149 * It would be nice to have these present only in DEBUG systems, but the
150 * current design of the global symbol checking logic requires them to be
151 * unconditionally present.
152 */
153 uint_t ip_thread_data; /* TSD key for debug support */
154 krwlock_t ip_thread_rwlock;
155 list_t ip_thread_list;
156
157 /*
158 * Structure to represent a linked list of msgblks. Used by ip_snmp_ functions.
159 */
160
161 struct listptr_s {
162 mblk_t *lp_head; /* pointer to the head of the list */
163 mblk_t *lp_tail; /* pointer to the tail of the list */
164 };
165
166 typedef struct listptr_s listptr_t;
167
168 /*
169 * This is used by ip_snmp_get_mib2_ip_route_media and
170 * ip_snmp_get_mib2_ip6_route_media to carry the lists of return data.
171 */
172 typedef struct iproutedata_s {
173 uint_t ird_idx;
174 uint_t ird_flags; /* see below */
175 listptr_t ird_route; /* ipRouteEntryTable */
176 listptr_t ird_netmedia; /* ipNetToMediaEntryTable */
177 listptr_t ird_attrs; /* ipRouteAttributeTable */
178 } iproutedata_t;
179
180 /* Include ire_testhidden and IRE_IF_CLONE routes */
181 #define IRD_REPORT_ALL 0x01
182
183 /*
184 * Cluster specific hooks. These should be NULL when booted as a non-cluster
185 */
186
187 /*
188 * Hook functions to enable cluster networking
189 * On non-clustered systems these vectors must always be NULL.
190 *
191 * Hook function to Check ip specified ip address is a shared ip address
192 * in the cluster
193 *
194 */
195 int (*cl_inet_isclusterwide)(netstackid_t stack_id, uint8_t protocol,
196 sa_family_t addr_family, uint8_t *laddrp, void *args) = NULL;
197
198 /*
199 * Hook function to generate cluster wide ip fragment identifier
200 */
201 uint32_t (*cl_inet_ipident)(netstackid_t stack_id, uint8_t protocol,
202 sa_family_t addr_family, uint8_t *laddrp, uint8_t *faddrp,
203 void *args) = NULL;
204
205 /*
206 * Hook function to generate cluster wide SPI.
207 */
208 void (*cl_inet_getspi)(netstackid_t, uint8_t, uint8_t *, size_t,
209 void *) = NULL;
210
211 /*
212 * Hook function to verify if the SPI is already utlized.
213 */
214
215 int (*cl_inet_checkspi)(netstackid_t, uint8_t, uint32_t, void *) = NULL;
216
217 /*
218 * Hook function to delete the SPI from the cluster wide repository.
219 */
220
221 void (*cl_inet_deletespi)(netstackid_t, uint8_t, uint32_t, void *) = NULL;
222
223 /*
224 * Hook function to inform the cluster when packet received on an IDLE SA
225 */
226
227 void (*cl_inet_idlesa)(netstackid_t, uint8_t, uint32_t, sa_family_t,
228 in6_addr_t, in6_addr_t, void *) = NULL;
229
230 /*
231 * Synchronization notes:
232 *
233 * IP is a fully D_MP STREAMS module/driver. Thus it does not depend on any
234 * MT level protection given by STREAMS. IP uses a combination of its own
235 * internal serialization mechanism and standard Solaris locking techniques.
236 * The internal serialization is per phyint. This is used to serialize
237 * plumbing operations, IPMP operations, most set ioctls, etc.
238 *
239 * Plumbing is a long sequence of operations involving message
240 * exchanges between IP, ARP and device drivers. Many set ioctls are typically
241 * involved in plumbing operations. A natural model is to serialize these
242 * ioctls one per ill. For example plumbing of hme0 and qfe0 can go on in
243 * parallel without any interference. But various set ioctls on hme0 are best
244 * serialized, along with IPMP operations and processing of DLPI control
245 * messages received from drivers on a per phyint basis. This serialization is
246 * provided by the ipsq_t and primitives operating on this. Details can
247 * be found in ip_if.c above the core primitives operating on ipsq_t.
248 *
249 * Lookups of an ipif or ill by a thread return a refheld ipif / ill.
250 * Simiarly lookup of an ire by a thread also returns a refheld ire.
251 * In addition ipif's and ill's referenced by the ire are also indirectly
252 * refheld. Thus no ipif or ill can vanish as long as an ipif is refheld
253 * directly or indirectly. For example an SIOCSLIFADDR ioctl that changes the
254 * address of an ipif has to go through the ipsq_t. This ensures that only
255 * one such exclusive operation proceeds at any time on the ipif. It then
256 * waits for all refcnts
257 * associated with this ipif to come down to zero. The address is changed
258 * only after the ipif has been quiesced. Then the ipif is brought up again.
259 * More details are described above the comment in ip_sioctl_flags.
260 *
261 * Packet processing is based mostly on IREs and are fully multi-threaded
262 * using standard Solaris MT techniques.
263 *
264 * There are explicit locks in IP to handle:
265 * - The ip_g_head list maintained by mi_open_link() and friends.
266 *
267 * - The reassembly data structures (one lock per hash bucket)
268 *
269 * - conn_lock is meant to protect conn_t fields. The fields actually
270 * protected by conn_lock are documented in the conn_t definition.
271 *
272 * - ire_lock to protect some of the fields of the ire, IRE tables
273 * (one lock per hash bucket). Refer to ip_ire.c for details.
274 *
275 * - ndp_g_lock and ncec_lock for protecting NCEs.
276 *
277 * - ill_lock protects fields of the ill and ipif. Details in ip.h
278 *
279 * - ill_g_lock: This is a global reader/writer lock. Protects the following
280 * * The AVL tree based global multi list of all ills.
281 * * The linked list of all ipifs of an ill
282 * * The <ipsq-xop> mapping
283 * * <ill-phyint> association
284 * Insertion/deletion of an ill in the system, insertion/deletion of an ipif
285 * into an ill, changing the <ipsq-xop> mapping of an ill, changing the
286 * <ill-phyint> assoc of an ill will all have to hold the ill_g_lock as
287 * writer for the actual duration of the insertion/deletion/change.
288 *
289 * - ill_lock: This is a per ill mutex.
290 * It protects some members of the ill_t struct; see ip.h for details.
291 * It also protects the <ill-phyint> assoc.
292 * It also protects the list of ipifs hanging off the ill.
293 *
294 * - ipsq_lock: This is a per ipsq_t mutex lock.
295 * This protects some members of the ipsq_t struct; see ip.h for details.
296 * It also protects the <ipsq-ipxop> mapping
297 *
298 * - ipx_lock: This is a per ipxop_t mutex lock.
299 * This protects some members of the ipxop_t struct; see ip.h for details.
300 *
301 * - phyint_lock: This is a per phyint mutex lock. Protects just the
302 * phyint_flags
303 *
304 * - ip_addr_avail_lock: This is used to ensure the uniqueness of IP addresses.
305 * This lock is held in ipif_up_done and the ipif is marked IPIF_UP and the
306 * uniqueness check also done atomically.
307 *
308 * - ill_g_usesrc_lock: This readers/writer lock protects the usesrc
309 * group list linked by ill_usesrc_grp_next. It also protects the
310 * ill_usesrc_ifindex field. It is taken as a writer when a member of the
311 * group is being added or deleted. This lock is taken as a reader when
312 * walking the list/group(eg: to get the number of members in a usesrc group).
313 * Note, it is only necessary to take this lock if the ill_usesrc_grp_next
314 * field is changing state i.e from NULL to non-NULL or vice-versa. For
315 * example, it is not necessary to take this lock in the initial portion
316 * of ip_sioctl_slifusesrc or at all in ip_sioctl_flags since these
317 * operations are executed exclusively and that ensures that the "usesrc
318 * group state" cannot change. The "usesrc group state" change can happen
319 * only in the latter part of ip_sioctl_slifusesrc and in ill_delete.
320 *
321 * Changing <ill-phyint>, <ipsq-xop> assocications:
322 *
323 * To change the <ill-phyint> association, the ill_g_lock must be held
324 * as writer, and the ill_locks of both the v4 and v6 instance of the ill
325 * must be held.
326 *
327 * To change the <ipsq-xop> association, the ill_g_lock must be held as
328 * writer, the ipsq_lock must be held, and one must be writer on the ipsq.
329 * This is only done when ills are added or removed from IPMP groups.
330 *
331 * To add or delete an ipif from the list of ipifs hanging off the ill,
332 * ill_g_lock (writer) and ill_lock must be held and the thread must be
333 * a writer on the associated ipsq.
334 *
335 * To add or delete an ill to the system, the ill_g_lock must be held as
336 * writer and the thread must be a writer on the associated ipsq.
337 *
338 * To add or delete an ilm to an ill, the ill_lock must be held and the thread
339 * must be a writer on the associated ipsq.
340 *
341 * Lock hierarchy
342 *
343 * Some lock hierarchy scenarios are listed below.
344 *
345 * ill_g_lock -> conn_lock -> ill_lock -> ipsq_lock -> ipx_lock
346 * ill_g_lock -> ill_lock(s) -> phyint_lock
347 * ill_g_lock -> ndp_g_lock -> ill_lock -> ncec_lock
348 * ill_g_lock -> ip_addr_avail_lock
349 * conn_lock -> irb_lock -> ill_lock -> ire_lock
350 * ill_g_lock -> ip_g_nd_lock
351 * ill_g_lock -> ips_ipmp_lock -> ill_lock -> nce_lock
352 * ill_g_lock -> ndp_g_lock -> ill_lock -> ncec_lock -> nce_lock
353 * arl_lock -> ill_lock
354 * ips_ire_dep_lock -> irb_lock
355 *
356 * When more than 1 ill lock is needed to be held, all ill lock addresses
357 * are sorted on address and locked starting from highest addressed lock
358 * downward.
359 *
360 * Multicast scenarios
361 * ips_ill_g_lock -> ill_mcast_lock
362 * conn_ilg_lock -> ips_ill_g_lock -> ill_lock
363 * ill_mcast_serializer -> ill_mcast_lock -> ips_ipmp_lock -> ill_lock
364 * ill_mcast_serializer -> ill_mcast_lock -> connf_lock -> conn_lock
365 * ill_mcast_serializer -> ill_mcast_lock -> conn_ilg_lock
366 * ill_mcast_serializer -> ill_mcast_lock -> ips_igmp_timer_lock
367 *
368 * IPsec scenarios
369 *
370 * ipsa_lock -> ill_g_lock -> ill_lock
371 * ill_g_usesrc_lock -> ill_g_lock -> ill_lock
372 *
373 * Trusted Solaris scenarios
374 *
375 * igsa_lock -> gcgrp_rwlock -> gcgrp_lock
376 * igsa_lock -> gcdb_lock
377 * gcgrp_rwlock -> ire_lock
378 * gcgrp_rwlock -> gcdb_lock
379 *
380 * squeue(sq_lock), flow related (ft_lock, fe_lock) locking
381 *
382 * cpu_lock --> ill_lock --> sqset_lock --> sq_lock
383 * sq_lock -> conn_lock -> QLOCK(q)
384 * ill_lock -> ft_lock -> fe_lock
385 *
386 * Routing/forwarding table locking notes:
387 *
388 * Lock acquisition order: Radix tree lock, irb_lock.
389 * Requirements:
390 * i. Walker must not hold any locks during the walker callback.
391 * ii Walker must not see a truncated tree during the walk because of any node
392 * deletion.
393 * iii Existing code assumes ire_bucket is valid if it is non-null and is used
394 * in many places in the code to walk the irb list. Thus even if all the
395 * ires in a bucket have been deleted, we still can't free the radix node
396 * until the ires have actually been inactive'd (freed).
397 *
398 * Tree traversal - Need to hold the global tree lock in read mode.
399 * Before dropping the global tree lock, need to either increment the ire_refcnt
400 * to ensure that the radix node can't be deleted.
401 *
402 * Tree add - Need to hold the global tree lock in write mode to add a
403 * radix node. To prevent the node from being deleted, increment the
404 * irb_refcnt, after the node is added to the tree. The ire itself is
405 * added later while holding the irb_lock, but not the tree lock.
406 *
407 * Tree delete - Need to hold the global tree lock and irb_lock in write mode.
408 * All associated ires must be inactive (i.e. freed), and irb_refcnt
409 * must be zero.
410 *
411 * Walker - Increment irb_refcnt before calling the walker callback. Hold the
412 * global tree lock (read mode) for traversal.
413 *
414 * IRE dependencies - In some cases we hold ips_ire_dep_lock across ire_refrele
415 * hence we will acquire irb_lock while holding ips_ire_dep_lock.
416 *
417 * IPsec notes :
418 *
419 * IP interacts with the IPsec code (AH/ESP) by storing IPsec attributes
420 * in the ip_xmit_attr_t ip_recv_attr_t. For outbound datagrams, the
421 * ip_xmit_attr_t has the
422 * information used by the IPsec code for applying the right level of
423 * protection. The information initialized by IP in the ip_xmit_attr_t
424 * is determined by the per-socket policy or global policy in the system.
425 * For inbound datagrams, the ip_recv_attr_t
426 * starts out with nothing in it. It gets filled
427 * with the right information if it goes through the AH/ESP code, which
428 * happens if the incoming packet is secure. The information initialized
429 * by AH/ESP, is later used by IP (during fanouts to ULP) to see whether
430 * the policy requirements needed by per-socket policy or global policy
431 * is met or not.
432 *
433 * For fully connected sockets i.e dst, src [addr, port] is known,
434 * conn_policy_cached is set indicating that policy has been cached.
435 * conn_in_enforce_policy may or may not be set depending on whether
436 * there is a global policy match or per-socket policy match.
437 * Policy inheriting happpens in ip_policy_set once the destination is known.
438 * Once the right policy is set on the conn_t, policy cannot change for
439 * this socket. This makes life simpler for TCP (UDP ?) where
440 * re-transmissions go out with the same policy. For symmetry, policy
441 * is cached for fully connected UDP sockets also. Thus if policy is cached,
442 * it also implies that policy is latched i.e policy cannot change
443 * on these sockets. As we have the right policy on the conn, we don't
444 * have to lookup global policy for every outbound and inbound datagram
445 * and thus serving as an optimization. Note that a global policy change
446 * does not affect fully connected sockets if they have policy. If fully
447 * connected sockets did not have any policy associated with it, global
448 * policy change may affect them.
449 *
450 * IP Flow control notes:
451 * ---------------------
452 * Non-TCP streams are flow controlled by IP. The way this is accomplished
453 * differs when ILL_CAPAB_DLD_DIRECT is enabled for that IP instance. When
454 * ILL_DIRECT_CAPABLE(ill) is TRUE, IP can do direct function calls into
455 * GLDv3. Otherwise packets are sent down to lower layers using STREAMS
456 * functions.
457 *
458 * Per Tx ring udp flow control:
459 * This is applicable only when ILL_CAPAB_DLD_DIRECT capability is set in
460 * the ill (i.e. ILL_DIRECT_CAPABLE(ill) is true).
461 *
462 * The underlying link can expose multiple Tx rings to the GLDv3 mac layer.
463 * To achieve best performance, outgoing traffic need to be fanned out among
464 * these Tx ring. mac_tx() is called (via str_mdata_fastpath_put()) to send
465 * traffic out of the NIC and it takes a fanout hint. UDP connections pass
466 * the address of connp as fanout hint to mac_tx(). Under flow controlled
467 * condition, mac_tx() returns a non-NULL cookie (ip_mac_tx_cookie_t). This
468 * cookie points to a specific Tx ring that is blocked. The cookie is used to
469 * hash into an idl_tx_list[] entry in idl_tx_list[] array. Each idl_tx_list_t
470 * point to drain_lists (idl_t's). These drain list will store the blocked UDP
471 * connp's. The drain list is not a single list but a configurable number of
472 * lists.
473 *
474 * The diagram below shows idl_tx_list_t's and their drain_lists. ip_stack_t
475 * has an array of idl_tx_list_t. The size of the array is TX_FANOUT_SIZE
476 * which is equal to 128. This array in turn contains a pointer to idl_t[],
477 * the ip drain list. The idl_t[] array size is MIN(max_ncpus, 8). The drain
478 * list will point to the list of connp's that are flow controlled.
479 *
480 * --------------- ------- ------- -------
481 * |->|drain_list[0]|-->|connp|-->|connp|-->|connp|-->
482 * | --------------- ------- ------- -------
483 * | --------------- ------- ------- -------
484 * |->|drain_list[1]|-->|connp|-->|connp|-->|connp|-->
485 * ---------------- | --------------- ------- ------- -------
486 * |idl_tx_list[0]|->| --------------- ------- ------- -------
487 * ---------------- |->|drain_list[2]|-->|connp|-->|connp|-->|connp|-->
488 * | --------------- ------- ------- -------
489 * . . . . .
490 * | --------------- ------- ------- -------
491 * |->|drain_list[n]|-->|connp|-->|connp|-->|connp|-->
492 * --------------- ------- ------- -------
493 * --------------- ------- ------- -------
494 * |->|drain_list[0]|-->|connp|-->|connp|-->|connp|-->
495 * | --------------- ------- ------- -------
496 * | --------------- ------- ------- -------
497 * ---------------- |->|drain_list[1]|-->|connp|-->|connp|-->|connp|-->
498 * |idl_tx_list[1]|->| --------------- ------- ------- -------
499 * ---------------- | . . . .
500 * | --------------- ------- ------- -------
501 * |->|drain_list[n]|-->|connp|-->|connp|-->|connp|-->
502 * --------------- ------- ------- -------
503 * .....
504 * ----------------
505 * |idl_tx_list[n]|-> ...
506 * ----------------
507 *
508 * When mac_tx() returns a cookie, the cookie is hashed into an index into
509 * ips_idl_tx_list[], and conn_drain_insert() is called with the idl_tx_list
510 * to insert the conn onto. conn_drain_insert() asserts flow control for the
511 * sockets via su_txq_full() (non-STREAMS) or QFULL on conn_wq (STREAMS).
512 * Further, conn_blocked is set to indicate that the conn is blocked.
513 *
514 * GLDv3 calls ill_flow_enable() when flow control is relieved. The cookie
515 * passed in the call to ill_flow_enable() identifies the blocked Tx ring and
516 * is again hashed to locate the appropriate idl_tx_list, which is then
517 * drained via conn_walk_drain(). conn_walk_drain() goes through each conn in
518 * the drain list and calls conn_drain_remove() to clear flow control (via
519 * calling su_txq_full() or clearing QFULL), and remove the conn from the
520 * drain list.
521 *
522 * Note that the drain list is not a single list but a (configurable) array of
523 * lists (8 elements by default). Synchronization between drain insertion and
524 * flow control wakeup is handled by using idl_txl->txl_lock, and only
525 * conn_drain_insert() and conn_drain_remove() manipulate the drain list.
526 *
527 * Flow control via STREAMS is used when ILL_DIRECT_CAPABLE() returns FALSE.
528 * On the send side, if the packet cannot be sent down to the driver by IP
529 * (canput() fails), ip_xmit() drops the packet and returns EWOULDBLOCK to the
530 * caller, who may then invoke ixa_check_drain_insert() to insert the conn on
531 * the 0'th drain list. When ip_wsrv() runs on the ill_wq because flow
532 * control has been relieved, the blocked conns in the 0'th drain list are
533 * drained as in the non-STREAMS case.
534 *
535 * In both the STREAMS and non-STREAMS cases, the sockfs upcall to set QFULL
536 * is done when the conn is inserted into the drain list (conn_drain_insert())
537 * and cleared when the conn is removed from the it (conn_drain_remove()).
538 *
539 * IPQOS notes:
540 *
541 * IPQoS Policies are applied to packets using IPPF (IP Policy framework)
542 * and IPQoS modules. IPPF includes hooks in IP at different control points
543 * (callout positions) which direct packets to IPQoS modules for policy
544 * processing. Policies, if present, are global.
545 *
546 * The callout positions are located in the following paths:
547 * o local_in (packets destined for this host)
548 * o local_out (packets orginating from this host )
549 * o fwd_in (packets forwarded by this m/c - inbound)
550 * o fwd_out (packets forwarded by this m/c - outbound)
551 * Hooks at these callout points can be enabled/disabled using the ndd variable
552 * ip_policy_mask (a bit mask with the 4 LSB indicating the callout positions).
553 * By default all the callout positions are enabled.
554 *
555 * Outbound (local_out)
556 * Hooks are placed in ire_send_wire_v4 and ire_send_wire_v6.
557 *
558 * Inbound (local_in)
559 * Hooks are placed in ip_fanout_v4 and ip_fanout_v6.
560 *
561 * Forwarding (in and out)
562 * Hooks are placed in ire_recv_forward_v4/v6.
563 *
564 * IP Policy Framework processing (IPPF processing)
565 * Policy processing for a packet is initiated by ip_process, which ascertains
566 * that the classifier (ipgpc) is loaded and configured, failing which the
567 * packet resumes normal processing in IP. If the clasifier is present, the
568 * packet is acted upon by one or more IPQoS modules (action instances), per
569 * filters configured in ipgpc and resumes normal IP processing thereafter.
570 * An action instance can drop a packet in course of its processing.
571 *
572 * Zones notes:
573 *
574 * The partitioning rules for networking are as follows:
575 * 1) Packets coming from a zone must have a source address belonging to that
576 * zone.
577 * 2) Packets coming from a zone can only be sent on a physical interface on
578 * which the zone has an IP address.
579 * 3) Between two zones on the same machine, packet delivery is only allowed if
580 * there's a matching route for the destination and zone in the forwarding
581 * table.
582 * 4) The TCP and UDP port spaces are per-zone; that is, two processes in
583 * different zones can bind to the same port with the wildcard address
584 * (INADDR_ANY).
585 *
586 * The granularity of interface partitioning is at the logical interface level.
587 * Therefore, every zone has its own IP addresses, and incoming packets can be
588 * attributed to a zone unambiguously. A logical interface is placed into a zone
589 * using the SIOCSLIFZONE ioctl; this sets the ipif_zoneid field in the ipif_t
590 * structure. Rule (1) is implemented by modifying the source address selection
591 * algorithm so that the list of eligible addresses is filtered based on the
592 * sending process zone.
593 *
594 * The Internet Routing Entries (IREs) are either exclusive to a zone or shared
595 * across all zones, depending on their type. Here is the break-up:
596 *
597 * IRE type Shared/exclusive
598 * -------- ----------------
599 * IRE_BROADCAST Exclusive
600 * IRE_DEFAULT (default routes) Shared (*)
601 * IRE_LOCAL Exclusive (x)
602 * IRE_LOOPBACK Exclusive
603 * IRE_PREFIX (net routes) Shared (*)
604 * IRE_IF_NORESOLVER (interface routes) Exclusive
605 * IRE_IF_RESOLVER (interface routes) Exclusive
606 * IRE_IF_CLONE (interface routes) Exclusive
607 * IRE_HOST (host routes) Shared (*)
608 *
609 * (*) A zone can only use a default or off-subnet route if the gateway is
610 * directly reachable from the zone, that is, if the gateway's address matches
611 * one of the zone's logical interfaces.
612 *
613 * (x) IRE_LOCAL are handled a bit differently.
614 * When ip_restrict_interzone_loopback is set (the default),
615 * ire_route_recursive restricts loopback using an IRE_LOCAL
616 * between zone to the case when L2 would have conceptually looped the packet
617 * back, i.e. the loopback which is required since neither Ethernet drivers
618 * nor Ethernet hardware loops them back. This is the case when the normal
619 * routes (ignoring IREs with different zoneids) would send out the packet on
620 * the same ill as the ill with which is IRE_LOCAL is associated.
621 *
622 * Multiple zones can share a common broadcast address; typically all zones
623 * share the 255.255.255.255 address. Incoming as well as locally originated
624 * broadcast packets must be dispatched to all the zones on the broadcast
625 * network. For directed broadcasts (e.g. 10.16.72.255) this is not trivial
626 * since some zones may not be on the 10.16.72/24 network. To handle this, each
627 * zone has its own set of IRE_BROADCAST entries; then, broadcast packets are
628 * sent to every zone that has an IRE_BROADCAST entry for the destination
629 * address on the input ill, see ip_input_broadcast().
630 *
631 * Applications in different zones can join the same multicast group address.
632 * The same logic applies for multicast as for broadcast. ip_input_multicast
633 * dispatches packets to all zones that have members on the physical interface.
634 */
635
636 /*
637 * Squeue Fanout flags:
638 * 0: No fanout.
639 * 1: Fanout across all squeues
640 */
641 boolean_t ip_squeue_fanout = 0;
642
643 /*
644 * Maximum dups allowed per packet.
645 */
646 uint_t ip_max_frag_dups = 10;
647
648 static int ip_open(queue_t *q, dev_t *devp, int flag, int sflag,
649 cred_t *credp, boolean_t isv6);
650 static mblk_t *ip_xmit_attach_llhdr(mblk_t *, nce_t *);
651
652 static boolean_t icmp_inbound_verify_v4(mblk_t *, icmph_t *, ip_recv_attr_t *);
653 static void icmp_inbound_too_big_v4(icmph_t *, ip_recv_attr_t *);
654 static void icmp_inbound_error_fanout_v4(mblk_t *, icmph_t *,
655 ip_recv_attr_t *);
656 static void icmp_options_update(ipha_t *);
657 static void icmp_param_problem(mblk_t *, uint8_t, ip_recv_attr_t *);
658 static void icmp_pkt(mblk_t *, void *, size_t, ip_recv_attr_t *);
659 static mblk_t *icmp_pkt_err_ok(mblk_t *, ip_recv_attr_t *);
660 static void icmp_redirect_v4(mblk_t *mp, ipha_t *, icmph_t *,
661 ip_recv_attr_t *);
662 static void icmp_send_redirect(mblk_t *, ipaddr_t, ip_recv_attr_t *);
663 static void icmp_send_reply_v4(mblk_t *, ipha_t *, icmph_t *,
664 ip_recv_attr_t *);
665
666 mblk_t *ip_dlpi_alloc(size_t, t_uscalar_t);
667 char *ip_dot_addr(ipaddr_t, char *);
668 mblk_t *ip_carve_mp(mblk_t **, ssize_t);
669 int ip_close(queue_t *, int);
670 static char *ip_dot_saddr(uchar_t *, char *);
671 static void ip_lrput(queue_t *, mblk_t *);
672 ipaddr_t ip_net_mask(ipaddr_t);
673 char *ip_nv_lookup(nv_t *, int);
674 void 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 int ip_snmp_get2_v4_media(ncec_t *, iproutedata_t *);
711 static int ip_snmp_get2_v6_media(ncec_t *, iproutedata_t *);
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 (pfi_t)ip_rput, NULL, ip_openv4, ip_close, NULL,
1189 &ip_mod_info
1190 };
1191
1192 struct qinit iprinitv6 = {
1193 (pfi_t)ip_rput_v6, NULL, ip_openv6, ip_close, NULL,
1194 &ip_mod_info
1195 };
1196
1197 static struct qinit ipwinit = {
1198 (pfi_t)ip_wput_nondata, (pfi_t)ip_wsrv, NULL, NULL, NULL,
1199 &ip_mod_info
1200 };
1201
1202 static struct qinit iplrinit = {
1203 (pfi_t)ip_lrput, NULL, ip_openv4, ip_close, NULL,
1204 &ip_mod_info
1205 };
1206
1207 static struct qinit iplwinit = {
1208 (pfi_t)ip_lwput, NULL, NULL, NULL, NULL,
1209 &ip_mod_info
1210 };
1211
1212 /* For AF_INET aka /dev/ip */
1213 struct streamtab ipinfov4 = {
1214 &iprinitv4, &ipwinit, &iplrinit, &iplwinit
1215 };
1216
1217 /* For AF_INET6 aka /dev/ip6 */
1218 struct streamtab ipinfov6 = {
1219 &iprinitv6, &ipwinit, &iplrinit, &iplwinit
1220 };
1221
1222 #ifdef DEBUG
1223 boolean_t skip_sctp_cksum = B_FALSE;
1224 #endif
1225
1226 /*
1227 * Generate an ICMP fragmentation needed message.
1228 * When called from ip_output side a minimal ip_recv_attr_t needs to be
1229 * constructed by the caller.
1230 */
1231 void
1232 icmp_frag_needed(mblk_t *mp, int mtu, ip_recv_attr_t *ira)
1233 {
1234 icmph_t icmph;
1235 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
1236
1237 mp = icmp_pkt_err_ok(mp, ira);
1238 if (mp == NULL)
1239 return;
1240
1241 bzero(&icmph, sizeof (icmph_t));
1242 icmph.icmph_type = ICMP_DEST_UNREACHABLE;
1243 icmph.icmph_code = ICMP_FRAGMENTATION_NEEDED;
1244 icmph.icmph_du_mtu = htons((uint16_t)mtu);
1245 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutFragNeeded);
1246 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDestUnreachs);
1247
1248 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
1249 }
1250
1251 /*
1252 * icmp_inbound_v4 deals with ICMP messages that are handled by IP.
1253 * If the ICMP message is consumed by IP, i.e., it should not be delivered
1254 * to any IPPROTO_ICMP raw sockets, then it returns NULL.
1255 * Likewise, if the ICMP error is misformed (too short, etc), then it
1256 * returns NULL. The caller uses this to determine whether or not to send
1257 * to raw sockets.
1258 *
1259 * All error messages are passed to the matching transport stream.
1260 *
1261 * The following cases are handled by icmp_inbound:
1262 * 1) It needs to send a reply back and possibly delivering it
1263 * to the "interested" upper clients.
1264 * 2) Return the mblk so that the caller can pass it to the RAW socket clients.
1265 * 3) It needs to change some values in IP only.
1266 * 4) It needs to change some values in IP and upper layers e.g TCP
1267 * by delivering an error to the upper layers.
1268 *
1269 * We handle the above three cases in the context of IPsec in the
1270 * following way :
1271 *
1272 * 1) Send the reply back in the same way as the request came in.
1273 * If it came in encrypted, it goes out encrypted. If it came in
1274 * clear, it goes out in clear. Thus, this will prevent chosen
1275 * plain text attack.
1276 * 2) The client may or may not expect things to come in secure.
1277 * If it comes in secure, the policy constraints are checked
1278 * before delivering it to the upper layers. If it comes in
1279 * clear, ipsec_inbound_accept_clear will decide whether to
1280 * accept this in clear or not. In both the cases, if the returned
1281 * message (IP header + 8 bytes) that caused the icmp message has
1282 * AH/ESP headers, it is sent up to AH/ESP for validation before
1283 * sending up. If there are only 8 bytes of returned message, then
1284 * upper client will not be notified.
1285 * 3) Check with global policy to see whether it matches the constaints.
1286 * But this will be done only if icmp_accept_messages_in_clear is
1287 * zero.
1288 * 4) If we need to change both in IP and ULP, then the decision taken
1289 * while affecting the values in IP and while delivering up to TCP
1290 * should be the same.
1291 *
1292 * There are two cases.
1293 *
1294 * a) If we reject data at the IP layer (ipsec_check_global_policy()
1295 * failed), we will not deliver it to the ULP, even though they
1296 * are *willing* to accept in *clear*. This is fine as our global
1297 * disposition to icmp messages asks us reject the datagram.
1298 *
1299 * b) If we accept data at the IP layer (ipsec_check_global_policy()
1300 * succeeded or icmp_accept_messages_in_clear is 1), and not able
1301 * to deliver it to ULP (policy failed), it can lead to
1302 * consistency problems. The cases known at this time are
1303 * ICMP_DESTINATION_UNREACHABLE messages with following code
1304 * values :
1305 *
1306 * - ICMP_FRAGMENTATION_NEEDED : IP adapts to the new value
1307 * and Upper layer rejects. Then the communication will
1308 * come to a stop. This is solved by making similar decisions
1309 * at both levels. Currently, when we are unable to deliver
1310 * to the Upper Layer (due to policy failures) while IP has
1311 * adjusted dce_pmtu, the next outbound datagram would
1312 * generate a local ICMP_FRAGMENTATION_NEEDED message - which
1313 * will be with the right level of protection. Thus the right
1314 * value will be communicated even if we are not able to
1315 * communicate when we get from the wire initially. But this
1316 * assumes there would be at least one outbound datagram after
1317 * IP has adjusted its dce_pmtu value. To make things
1318 * simpler, we accept in clear after the validation of
1319 * AH/ESP headers.
1320 *
1321 * - Other ICMP ERRORS : We may not be able to deliver it to the
1322 * upper layer depending on the level of protection the upper
1323 * layer expects and the disposition in ipsec_inbound_accept_clear().
1324 * ipsec_inbound_accept_clear() decides whether a given ICMP error
1325 * should be accepted in clear when the Upper layer expects secure.
1326 * Thus the communication may get aborted by some bad ICMP
1327 * packets.
1328 */
1329 mblk_t *
1330 icmp_inbound_v4(mblk_t *mp, ip_recv_attr_t *ira)
1331 {
1332 icmph_t *icmph;
1333 ipha_t *ipha; /* Outer header */
1334 int ip_hdr_length; /* Outer header length */
1335 boolean_t interested;
1336 ipif_t *ipif;
1337 uint32_t ts;
1338 uint32_t *tsp;
1339 timestruc_t now;
1340 ill_t *ill = ira->ira_ill;
1341 ip_stack_t *ipst = ill->ill_ipst;
1342 zoneid_t zoneid = ira->ira_zoneid;
1343 int len_needed;
1344 mblk_t *mp_ret = NULL;
1345
1346 ipha = (ipha_t *)mp->b_rptr;
1347
1348 BUMP_MIB(&ipst->ips_icmp_mib, icmpInMsgs);
1349
1350 ip_hdr_length = ira->ira_ip_hdr_length;
1351 if ((mp->b_wptr - mp->b_rptr) < (ip_hdr_length + ICMPH_SIZE)) {
1352 if (ira->ira_pktlen < (ip_hdr_length + ICMPH_SIZE)) {
1353 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
1354 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
1355 freemsg(mp);
1356 return (NULL);
1357 }
1358 /* Last chance to get real. */
1359 ipha = ip_pullup(mp, ip_hdr_length + ICMPH_SIZE, ira);
1360 if (ipha == NULL) {
1361 BUMP_MIB(&ipst->ips_icmp_mib, icmpInErrors);
1362 freemsg(mp);
1363 return (NULL);
1364 }
1365 }
1366
1367 /* The IP header will always be a multiple of four bytes */
1368 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1369 ip2dbg(("icmp_inbound_v4: type %d code %d\n", icmph->icmph_type,
1370 icmph->icmph_code));
1371
1372 /*
1373 * We will set "interested" to "true" if we should pass a copy to
1374 * the transport or if we handle the packet locally.
1375 */
1376 interested = B_FALSE;
1377 switch (icmph->icmph_type) {
1378 case ICMP_ECHO_REPLY:
1379 BUMP_MIB(&ipst->ips_icmp_mib, icmpInEchoReps);
1380 break;
1381 case ICMP_DEST_UNREACHABLE:
1382 if (icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED)
1383 BUMP_MIB(&ipst->ips_icmp_mib, icmpInFragNeeded);
1384 interested = B_TRUE; /* Pass up to transport */
1385 BUMP_MIB(&ipst->ips_icmp_mib, icmpInDestUnreachs);
1386 break;
1387 case ICMP_SOURCE_QUENCH:
1388 interested = B_TRUE; /* Pass up to transport */
1389 BUMP_MIB(&ipst->ips_icmp_mib, icmpInSrcQuenchs);
1390 break;
1391 case ICMP_REDIRECT:
1392 if (!ipst->ips_ip_ignore_redirect)
1393 interested = B_TRUE;
1394 BUMP_MIB(&ipst->ips_icmp_mib, icmpInRedirects);
1395 break;
1396 case ICMP_ECHO_REQUEST:
1397 /*
1398 * Whether to respond to echo requests that come in as IP
1399 * broadcasts or as IP multicast is subject to debate
1400 * (what isn't?). We aim to please, you pick it.
1401 * Default is do it.
1402 */
1403 if (ira->ira_flags & IRAF_MULTICAST) {
1404 /* multicast: respond based on tunable */
1405 interested = ipst->ips_ip_g_resp_to_echo_mcast;
1406 } else if (ira->ira_flags & IRAF_BROADCAST) {
1407 /* broadcast: respond based on tunable */
1408 interested = ipst->ips_ip_g_resp_to_echo_bcast;
1409 } else {
1410 /* unicast: always respond */
1411 interested = B_TRUE;
1412 }
1413 BUMP_MIB(&ipst->ips_icmp_mib, icmpInEchos);
1414 if (!interested) {
1415 /* We never pass these to RAW sockets */
1416 freemsg(mp);
1417 return (NULL);
1418 }
1419
1420 /* Check db_ref to make sure we can modify the packet. */
1421 if (mp->b_datap->db_ref > 1) {
1422 mblk_t *mp1;
1423
1424 mp1 = copymsg(mp);
1425 freemsg(mp);
1426 if (!mp1) {
1427 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1428 return (NULL);
1429 }
1430 mp = mp1;
1431 ipha = (ipha_t *)mp->b_rptr;
1432 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1433 }
1434 icmph->icmph_type = ICMP_ECHO_REPLY;
1435 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutEchoReps);
1436 icmp_send_reply_v4(mp, ipha, icmph, ira);
1437 return (NULL);
1438
1439 case ICMP_ROUTER_ADVERTISEMENT:
1440 case ICMP_ROUTER_SOLICITATION:
1441 break;
1442 case ICMP_TIME_EXCEEDED:
1443 interested = B_TRUE; /* Pass up to transport */
1444 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimeExcds);
1445 break;
1446 case ICMP_PARAM_PROBLEM:
1447 interested = B_TRUE; /* Pass up to transport */
1448 BUMP_MIB(&ipst->ips_icmp_mib, icmpInParmProbs);
1449 break;
1450 case ICMP_TIME_STAMP_REQUEST:
1451 /* Response to Time Stamp Requests is local policy. */
1452 if (ipst->ips_ip_g_resp_to_timestamp) {
1453 if (ira->ira_flags & IRAF_MULTIBROADCAST)
1454 interested =
1455 ipst->ips_ip_g_resp_to_timestamp_bcast;
1456 else
1457 interested = B_TRUE;
1458 }
1459 if (!interested) {
1460 /* We never pass these to RAW sockets */
1461 freemsg(mp);
1462 return (NULL);
1463 }
1464
1465 /* Make sure we have enough of the packet */
1466 len_needed = ip_hdr_length + ICMPH_SIZE +
1467 3 * sizeof (uint32_t);
1468
1469 if (mp->b_wptr - mp->b_rptr < len_needed) {
1470 ipha = ip_pullup(mp, len_needed, ira);
1471 if (ipha == NULL) {
1472 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1473 ip_drop_input("ipIfStatsInDiscards - ip_pullup",
1474 mp, ill);
1475 freemsg(mp);
1476 return (NULL);
1477 }
1478 /* Refresh following the pullup. */
1479 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1480 }
1481 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimestamps);
1482 /* Check db_ref to make sure we can modify the packet. */
1483 if (mp->b_datap->db_ref > 1) {
1484 mblk_t *mp1;
1485
1486 mp1 = copymsg(mp);
1487 freemsg(mp);
1488 if (!mp1) {
1489 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1490 return (NULL);
1491 }
1492 mp = mp1;
1493 ipha = (ipha_t *)mp->b_rptr;
1494 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1495 }
1496 icmph->icmph_type = ICMP_TIME_STAMP_REPLY;
1497 tsp = (uint32_t *)&icmph[1];
1498 tsp++; /* Skip past 'originate time' */
1499 /* Compute # of milliseconds since midnight */
1500 gethrestime(&now);
1501 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
1502 NSEC2MSEC(now.tv_nsec);
1503 *tsp++ = htonl(ts); /* Lay in 'receive time' */
1504 *tsp++ = htonl(ts); /* Lay in 'send time' */
1505 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutTimestampReps);
1506 icmp_send_reply_v4(mp, ipha, icmph, ira);
1507 return (NULL);
1508
1509 case ICMP_TIME_STAMP_REPLY:
1510 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimestampReps);
1511 break;
1512 case ICMP_INFO_REQUEST:
1513 /* Per RFC 1122 3.2.2.7, ignore this. */
1514 case ICMP_INFO_REPLY:
1515 break;
1516 case ICMP_ADDRESS_MASK_REQUEST:
1517 if (ira->ira_flags & IRAF_MULTIBROADCAST) {
1518 interested =
1519 ipst->ips_ip_respond_to_address_mask_broadcast;
1520 } else {
1521 interested = B_TRUE;
1522 }
1523 if (!interested) {
1524 /* We never pass these to RAW sockets */
1525 freemsg(mp);
1526 return (NULL);
1527 }
1528 len_needed = ip_hdr_length + ICMPH_SIZE + IP_ADDR_LEN;
1529 if (mp->b_wptr - mp->b_rptr < len_needed) {
1530 ipha = ip_pullup(mp, len_needed, ira);
1531 if (ipha == NULL) {
1532 BUMP_MIB(ill->ill_ip_mib,
1533 ipIfStatsInTruncatedPkts);
1534 ip_drop_input("ipIfStatsInTruncatedPkts", mp,
1535 ill);
1536 freemsg(mp);
1537 return (NULL);
1538 }
1539 /* Refresh following the pullup. */
1540 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1541 }
1542 BUMP_MIB(&ipst->ips_icmp_mib, icmpInAddrMasks);
1543 /* Check db_ref to make sure we can modify the packet. */
1544 if (mp->b_datap->db_ref > 1) {
1545 mblk_t *mp1;
1546
1547 mp1 = copymsg(mp);
1548 freemsg(mp);
1549 if (!mp1) {
1550 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1551 return (NULL);
1552 }
1553 mp = mp1;
1554 ipha = (ipha_t *)mp->b_rptr;
1555 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1556 }
1557 /*
1558 * Need the ipif with the mask be the same as the source
1559 * address of the mask reply. For unicast we have a specific
1560 * ipif. For multicast/broadcast we only handle onlink
1561 * senders, and use the source address to pick an ipif.
1562 */
1563 ipif = ipif_lookup_addr(ipha->ipha_dst, ill, zoneid, ipst);
1564 if (ipif == NULL) {
1565 /* Broadcast or multicast */
1566 ipif = ipif_lookup_remote(ill, ipha->ipha_src, zoneid);
1567 if (ipif == NULL) {
1568 freemsg(mp);
1569 return (NULL);
1570 }
1571 }
1572 icmph->icmph_type = ICMP_ADDRESS_MASK_REPLY;
1573 bcopy(&ipif->ipif_net_mask, &icmph[1], IP_ADDR_LEN);
1574 ipif_refrele(ipif);
1575 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutAddrMaskReps);
1576 icmp_send_reply_v4(mp, ipha, icmph, ira);
1577 return (NULL);
1578
1579 case ICMP_ADDRESS_MASK_REPLY:
1580 BUMP_MIB(&ipst->ips_icmp_mib, icmpInAddrMaskReps);
1581 break;
1582 default:
1583 interested = B_TRUE; /* Pass up to transport */
1584 BUMP_MIB(&ipst->ips_icmp_mib, icmpInUnknowns);
1585 break;
1586 }
1587 /*
1588 * See if there is an ICMP client to avoid an extra copymsg/freemsg
1589 * if there isn't one.
1590 */
1591 if (ipst->ips_ipcl_proto_fanout_v4[IPPROTO_ICMP].connf_head != NULL) {
1592 /* If there is an ICMP client and we want one too, copy it. */
1593
1594 if (!interested) {
1595 /* Caller will deliver to RAW sockets */
1596 return (mp);
1597 }
1598 mp_ret = copymsg(mp);
1599 if (mp_ret == NULL) {
1600 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1601 ip_drop_input("ipIfStatsInDiscards - copymsg", mp, ill);
1602 }
1603 } else if (!interested) {
1604 /* Neither we nor raw sockets are interested. Drop packet now */
1605 freemsg(mp);
1606 return (NULL);
1607 }
1608
1609 /*
1610 * ICMP error or redirect packet. Make sure we have enough of
1611 * the header and that db_ref == 1 since we might end up modifying
1612 * the packet.
1613 */
1614 if (mp->b_cont != NULL) {
1615 if (ip_pullup(mp, -1, ira) == NULL) {
1616 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1617 ip_drop_input("ipIfStatsInDiscards - ip_pullup",
1618 mp, ill);
1619 freemsg(mp);
1620 return (mp_ret);
1621 }
1622 }
1623
1624 if (mp->b_datap->db_ref > 1) {
1625 mblk_t *mp1;
1626
1627 mp1 = copymsg(mp);
1628 if (mp1 == NULL) {
1629 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1630 ip_drop_input("ipIfStatsInDiscards - copymsg", mp, ill);
1631 freemsg(mp);
1632 return (mp_ret);
1633 }
1634 freemsg(mp);
1635 mp = mp1;
1636 }
1637
1638 /*
1639 * In case mp has changed, verify the message before any further
1640 * processes.
1641 */
1642 ipha = (ipha_t *)mp->b_rptr;
1643 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1644 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
1645 freemsg(mp);
1646 return (mp_ret);
1647 }
1648
1649 switch (icmph->icmph_type) {
1650 case ICMP_REDIRECT:
1651 icmp_redirect_v4(mp, ipha, icmph, ira);
1652 break;
1653 case ICMP_DEST_UNREACHABLE:
1654 if (icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED) {
1655 /* Update DCE and adjust MTU is icmp header if needed */
1656 icmp_inbound_too_big_v4(icmph, ira);
1657 }
1658 /* FALLTHRU */
1659 default:
1660 icmp_inbound_error_fanout_v4(mp, icmph, ira);
1661 break;
1662 }
1663 return (mp_ret);
1664 }
1665
1666 /*
1667 * Send an ICMP echo, timestamp or address mask reply.
1668 * The caller has already updated the payload part of the packet.
1669 * We handle the ICMP checksum, IP source address selection and feed
1670 * the packet into ip_output_simple.
1671 */
1672 static void
1673 icmp_send_reply_v4(mblk_t *mp, ipha_t *ipha, icmph_t *icmph,
1674 ip_recv_attr_t *ira)
1675 {
1676 uint_t ip_hdr_length = ira->ira_ip_hdr_length;
1677 ill_t *ill = ira->ira_ill;
1678 ip_stack_t *ipst = ill->ill_ipst;
1679 ip_xmit_attr_t ixas;
1680
1681 /* Send out an ICMP packet */
1682 icmph->icmph_checksum = 0;
1683 icmph->icmph_checksum = IP_CSUM(mp, ip_hdr_length, 0);
1684 /* Reset time to live. */
1685 ipha->ipha_ttl = ipst->ips_ip_def_ttl;
1686 {
1687 /* Swap source and destination addresses */
1688 ipaddr_t tmp;
1689
1690 tmp = ipha->ipha_src;
1691 ipha->ipha_src = ipha->ipha_dst;
1692 ipha->ipha_dst = tmp;
1693 }
1694 ipha->ipha_ident = 0;
1695 if (!IS_SIMPLE_IPH(ipha))
1696 icmp_options_update(ipha);
1697
1698 bzero(&ixas, sizeof (ixas));
1699 ixas.ixa_flags = IXAF_BASIC_SIMPLE_V4;
1700 ixas.ixa_zoneid = ira->ira_zoneid;
1701 ixas.ixa_cred = kcred;
1702 ixas.ixa_cpid = NOPID;
1703 ixas.ixa_tsl = ira->ira_tsl; /* Behave as a multi-level responder */
1704 ixas.ixa_ifindex = 0;
1705 ixas.ixa_ipst = ipst;
1706 ixas.ixa_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1707
1708 if (!(ira->ira_flags & IRAF_IPSEC_SECURE)) {
1709 /*
1710 * This packet should go out the same way as it
1711 * came in i.e in clear, independent of the IPsec policy
1712 * for transmitting packets.
1713 */
1714 ixas.ixa_flags |= IXAF_NO_IPSEC;
1715 } else {
1716 if (!ipsec_in_to_out(ira, &ixas, mp, ipha, NULL)) {
1717 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1718 /* Note: mp already consumed and ip_drop_packet done */
1719 return;
1720 }
1721 }
1722 if (ira->ira_flags & IRAF_MULTIBROADCAST) {
1723 /*
1724 * Not one or our addresses (IRE_LOCALs), thus we let
1725 * ip_output_simple pick the source.
1726 */
1727 ipha->ipha_src = INADDR_ANY;
1728 ixas.ixa_flags |= IXAF_SET_SOURCE;
1729 }
1730 /* Should we send with DF and use dce_pmtu? */
1731 if (ipst->ips_ipv4_icmp_return_pmtu) {
1732 ixas.ixa_flags |= IXAF_PMTU_DISCOVERY;
1733 ipha->ipha_fragment_offset_and_flags |= IPH_DF_HTONS;
1734 }
1735
1736 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutMsgs);
1737
1738 (void) ip_output_simple(mp, &ixas);
1739 ixa_cleanup(&ixas);
1740 }
1741
1742 /*
1743 * Verify the ICMP messages for either for ICMP error or redirect packet.
1744 * The caller should have fully pulled up the message. If it's a redirect
1745 * packet, only basic checks on IP header will be done; otherwise, verify
1746 * the packet by looking at the included ULP header.
1747 *
1748 * Called before icmp_inbound_error_fanout_v4 is called.
1749 */
1750 static boolean_t
1751 icmp_inbound_verify_v4(mblk_t *mp, icmph_t *icmph, ip_recv_attr_t *ira)
1752 {
1753 ill_t *ill = ira->ira_ill;
1754 int hdr_length;
1755 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
1756 conn_t *connp;
1757 ipha_t *ipha; /* Inner IP header */
1758
1759 ipha = (ipha_t *)&icmph[1];
1760 if ((uchar_t *)ipha + IP_SIMPLE_HDR_LENGTH > mp->b_wptr)
1761 goto truncated;
1762
1763 hdr_length = IPH_HDR_LENGTH(ipha);
1764
1765 if ((IPH_HDR_VERSION(ipha) != IPV4_VERSION))
1766 goto discard_pkt;
1767
1768 if (hdr_length < sizeof (ipha_t))
1769 goto truncated;
1770
1771 if ((uchar_t *)ipha + hdr_length > mp->b_wptr)
1772 goto truncated;
1773
1774 /*
1775 * Stop here for ICMP_REDIRECT.
1776 */
1777 if (icmph->icmph_type == ICMP_REDIRECT)
1778 return (B_TRUE);
1779
1780 /*
1781 * ICMP errors only.
1782 */
1783 switch (ipha->ipha_protocol) {
1784 case IPPROTO_UDP:
1785 /*
1786 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1787 * transport header.
1788 */
1789 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1790 mp->b_wptr)
1791 goto truncated;
1792 break;
1793 case IPPROTO_TCP: {
1794 tcpha_t *tcpha;
1795
1796 /*
1797 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1798 * transport header.
1799 */
1800 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1801 mp->b_wptr)
1802 goto truncated;
1803
1804 tcpha = (tcpha_t *)((uchar_t *)ipha + hdr_length);
1805 connp = ipcl_tcp_lookup_reversed_ipv4(ipha, tcpha, TCPS_LISTEN,
1806 ipst);
1807 if (connp == NULL)
1808 goto discard_pkt;
1809
1810 if ((connp->conn_verifyicmp != NULL) &&
1811 !connp->conn_verifyicmp(connp, tcpha, icmph, NULL, ira)) {
1812 CONN_DEC_REF(connp);
1813 goto discard_pkt;
1814 }
1815 CONN_DEC_REF(connp);
1816 break;
1817 }
1818 case IPPROTO_SCTP:
1819 /*
1820 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1821 * transport header.
1822 */
1823 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1824 mp->b_wptr)
1825 goto truncated;
1826 break;
1827 case IPPROTO_ESP:
1828 case IPPROTO_AH:
1829 break;
1830 case IPPROTO_ENCAP:
1831 if ((uchar_t *)ipha + hdr_length + sizeof (ipha_t) >
1832 mp->b_wptr)
1833 goto truncated;
1834 break;
1835 default:
1836 break;
1837 }
1838
1839 return (B_TRUE);
1840
1841 discard_pkt:
1842 /* Bogus ICMP error. */
1843 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1844 return (B_FALSE);
1845
1846 truncated:
1847 /* We pulled up everthing already. Must be truncated */
1848 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
1849 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
1850 return (B_FALSE);
1851 }
1852
1853 /* Table from RFC 1191 */
1854 static int icmp_frag_size_table[] =
1855 { 32000, 17914, 8166, 4352, 2002, 1496, 1006, 508, 296, 68 };
1856
1857 /*
1858 * Process received ICMP Packet too big.
1859 * Just handles the DCE create/update, including using the above table of
1860 * PMTU guesses. The caller is responsible for validating the packet before
1861 * passing it in and also to fanout the ICMP error to any matching transport
1862 * conns. Assumes the message has been fully pulled up and verified.
1863 *
1864 * Before getting here, the caller has called icmp_inbound_verify_v4()
1865 * that should have verified with ULP to prevent undoing the changes we're
1866 * going to make to DCE. For example, TCP might have verified that the packet
1867 * which generated error is in the send window.
1868 *
1869 * In some cases modified this MTU in the ICMP header packet; the caller
1870 * should pass to the matching ULP after this returns.
1871 */
1872 static void
1873 icmp_inbound_too_big_v4(icmph_t *icmph, ip_recv_attr_t *ira)
1874 {
1875 dce_t *dce;
1876 int old_mtu;
1877 int mtu, orig_mtu;
1878 ipaddr_t dst;
1879 boolean_t disable_pmtud;
1880 ill_t *ill = ira->ira_ill;
1881 ip_stack_t *ipst = ill->ill_ipst;
1882 uint_t hdr_length;
1883 ipha_t *ipha;
1884
1885 /* Caller already pulled up everything. */
1886 ipha = (ipha_t *)&icmph[1];
1887 ASSERT(icmph->icmph_type == ICMP_DEST_UNREACHABLE &&
1888 icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED);
1889 ASSERT(ill != NULL);
1890
1891 hdr_length = IPH_HDR_LENGTH(ipha);
1892
1893 /*
1894 * We handle path MTU for source routed packets since the DCE
1895 * is looked up using the final destination.
1896 */
1897 dst = ip_get_dst(ipha);
1898
1899 dce = dce_lookup_and_add_v4(dst, ipst);
1900 if (dce == NULL) {
1901 /* Couldn't add a unique one - ENOMEM */
1902 ip1dbg(("icmp_inbound_too_big_v4: no dce for 0x%x\n",
1903 ntohl(dst)));
1904 return;
1905 }
1906
1907 /* Check for MTU discovery advice as described in RFC 1191 */
1908 mtu = ntohs(icmph->icmph_du_mtu);
1909 orig_mtu = mtu;
1910 disable_pmtud = B_FALSE;
1911
1912 mutex_enter(&dce->dce_lock);
1913 if (dce->dce_flags & DCEF_PMTU)
1914 old_mtu = dce->dce_pmtu;
1915 else
1916 old_mtu = ill->ill_mtu;
1917
1918 if (icmph->icmph_du_zero != 0 || mtu < ipst->ips_ip_pmtu_min) {
1919 uint32_t length;
1920 int i;
1921
1922 /*
1923 * Use the table from RFC 1191 to figure out
1924 * the next "plateau" based on the length in
1925 * the original IP packet.
1926 */
1927 length = ntohs(ipha->ipha_length);
1928 DTRACE_PROBE2(ip4__pmtu__guess, dce_t *, dce,
1929 uint32_t, length);
1930 if (old_mtu <= length &&
1931 old_mtu >= length - hdr_length) {
1932 /*
1933 * Handle broken BSD 4.2 systems that
1934 * return the wrong ipha_length in ICMP
1935 * errors.
1936 */
1937 ip1dbg(("Wrong mtu: sent %d, dce %d\n",
1938 length, old_mtu));
1939 length -= hdr_length;
1940 }
1941 for (i = 0; i < A_CNT(icmp_frag_size_table); i++) {
1942 if (length > icmp_frag_size_table[i])
1943 break;
1944 }
1945 if (i == A_CNT(icmp_frag_size_table)) {
1946 /* Smaller than IP_MIN_MTU! */
1947 ip1dbg(("Too big for packet size %d\n",
1948 length));
1949 disable_pmtud = B_TRUE;
1950 mtu = ipst->ips_ip_pmtu_min;
1951 } else {
1952 mtu = icmp_frag_size_table[i];
1953 ip1dbg(("Calculated mtu %d, packet size %d, "
1954 "before %d\n", mtu, length, old_mtu));
1955 if (mtu < ipst->ips_ip_pmtu_min) {
1956 mtu = ipst->ips_ip_pmtu_min;
1957 disable_pmtud = B_TRUE;
1958 }
1959 }
1960 }
1961 if (disable_pmtud)
1962 dce->dce_flags |= DCEF_TOO_SMALL_PMTU;
1963 else
1964 dce->dce_flags &= ~DCEF_TOO_SMALL_PMTU;
1965
1966 dce->dce_pmtu = MIN(old_mtu, mtu);
1967 /* Prepare to send the new max frag size for the ULP. */
1968 icmph->icmph_du_zero = 0;
1969 icmph->icmph_du_mtu = htons((uint16_t)dce->dce_pmtu);
1970 DTRACE_PROBE4(ip4__pmtu__change, icmph_t *, icmph, dce_t *,
1971 dce, int, orig_mtu, int, mtu);
1972
1973 /* We now have a PMTU for sure */
1974 dce->dce_flags |= DCEF_PMTU;
1975 dce->dce_last_change_time = TICK_TO_SEC(ddi_get_lbolt64());
1976 mutex_exit(&dce->dce_lock);
1977 /*
1978 * After dropping the lock the new value is visible to everyone.
1979 * Then we bump the generation number so any cached values reinspect
1980 * the dce_t.
1981 */
1982 dce_increment_generation(dce);
1983 dce_refrele(dce);
1984 }
1985
1986 /*
1987 * If the packet in error is Self-Encapsulated, icmp_inbound_error_fanout_v4
1988 * calls this function.
1989 */
1990 static mblk_t *
1991 icmp_inbound_self_encap_error_v4(mblk_t *mp, ipha_t *ipha, ipha_t *in_ipha)
1992 {
1993 int length;
1994
1995 ASSERT(mp->b_datap->db_type == M_DATA);
1996
1997 /* icmp_inbound_v4 has already pulled up the whole error packet */
1998 ASSERT(mp->b_cont == NULL);
1999
2000 /*
2001 * The length that we want to overlay is the inner header
2002 * and what follows it.
2003 */
2004 length = msgdsize(mp) - ((uchar_t *)in_ipha - mp->b_rptr);
2005
2006 /*
2007 * Overlay the inner header and whatever follows it over the
2008 * outer header.
2009 */
2010 bcopy((uchar_t *)in_ipha, (uchar_t *)ipha, length);
2011
2012 /* Adjust for what we removed */
2013 mp->b_wptr -= (uchar_t *)in_ipha - (uchar_t *)ipha;
2014 return (mp);
2015 }
2016
2017 /*
2018 * Try to pass the ICMP message upstream in case the ULP cares.
2019 *
2020 * If the packet that caused the ICMP error is secure, we send
2021 * it to AH/ESP to make sure that the attached packet has a
2022 * valid association. ipha in the code below points to the
2023 * IP header of the packet that caused the error.
2024 *
2025 * For IPsec cases, we let the next-layer-up (which has access to
2026 * cached policy on the conn_t, or can query the SPD directly)
2027 * subtract out any IPsec overhead if they must. We therefore make no
2028 * adjustments here for IPsec overhead.
2029 *
2030 * IFN could have been generated locally or by some router.
2031 *
2032 * LOCAL : ire_send_wire (before calling ipsec_out_process) can call
2033 * icmp_frag_needed/icmp_pkt2big_v6 to generated a local IFN.
2034 * This happens because IP adjusted its value of MTU on an
2035 * earlier IFN message and could not tell the upper layer,
2036 * the new adjusted value of MTU e.g. Packet was encrypted
2037 * or there was not enough information to fanout to upper
2038 * layers. Thus on the next outbound datagram, ire_send_wire
2039 * generates the IFN, where IPsec processing has *not* been
2040 * done.
2041 *
2042 * Note that we retain ixa_fragsize across IPsec thus once
2043 * we have picking ixa_fragsize and entered ipsec_out_process we do
2044 * no change the fragsize even if the path MTU changes before
2045 * we reach ip_output_post_ipsec.
2046 *
2047 * In the local case, IRAF_LOOPBACK will be set indicating
2048 * that IFN was generated locally.
2049 *
2050 * ROUTER : IFN could be secure or non-secure.
2051 *
2052 * * SECURE : We use the IPSEC_IN to fanout to AH/ESP if the
2053 * packet in error has AH/ESP headers to validate the AH/ESP
2054 * headers. AH/ESP will verify whether there is a valid SA or
2055 * not and send it back. We will fanout again if we have more
2056 * data in the packet.
2057 *
2058 * If the packet in error does not have AH/ESP, we handle it
2059 * like any other case.
2060 *
2061 * * NON_SECURE : If the packet in error has AH/ESP headers, we send it
2062 * up to AH/ESP for validation. AH/ESP will verify whether there is a
2063 * valid SA or not and send it back. We will fanout again if
2064 * we have more data in the packet.
2065 *
2066 * If the packet in error does not have AH/ESP, we handle it
2067 * like any other case.
2068 *
2069 * The caller must have called icmp_inbound_verify_v4.
2070 */
2071 static void
2072 icmp_inbound_error_fanout_v4(mblk_t *mp, icmph_t *icmph, ip_recv_attr_t *ira)
2073 {
2074 uint16_t *up; /* Pointer to ports in ULP header */
2075 uint32_t ports; /* reversed ports for fanout */
2076 ipha_t ripha; /* With reversed addresses */
2077 ipha_t *ipha; /* Inner IP header */
2078 uint_t hdr_length; /* Inner IP header length */
2079 tcpha_t *tcpha;
2080 conn_t *connp;
2081 ill_t *ill = ira->ira_ill;
2082 ip_stack_t *ipst = ill->ill_ipst;
2083 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
2084 ill_t *rill = ira->ira_rill;
2085
2086 /* Caller already pulled up everything. */
2087 ipha = (ipha_t *)&icmph[1];
2088 ASSERT((uchar_t *)&ipha[1] <= mp->b_wptr);
2089 ASSERT(mp->b_cont == NULL);
2090
2091 hdr_length = IPH_HDR_LENGTH(ipha);
2092 ira->ira_protocol = ipha->ipha_protocol;
2093
2094 /*
2095 * We need a separate IP header with the source and destination
2096 * addresses reversed to do fanout/classification because the ipha in
2097 * the ICMP error is in the form we sent it out.
2098 */
2099 ripha.ipha_src = ipha->ipha_dst;
2100 ripha.ipha_dst = ipha->ipha_src;
2101 ripha.ipha_protocol = ipha->ipha_protocol;
2102 ripha.ipha_version_and_hdr_length = ipha->ipha_version_and_hdr_length;
2103
2104 ip2dbg(("icmp_inbound_error_v4: proto %d %x to %x: %d/%d\n",
2105 ripha.ipha_protocol, ntohl(ipha->ipha_src),
2106 ntohl(ipha->ipha_dst),
2107 icmph->icmph_type, icmph->icmph_code));
2108
2109 switch (ipha->ipha_protocol) {
2110 case IPPROTO_UDP:
2111 up = (uint16_t *)((uchar_t *)ipha + hdr_length);
2112
2113 /* Attempt to find a client stream based on port. */
2114 ip2dbg(("icmp_inbound_error_v4: UDP ports %d to %d\n",
2115 ntohs(up[0]), ntohs(up[1])));
2116
2117 /* Note that we send error to all matches. */
2118 ira->ira_flags |= IRAF_ICMP_ERROR;
2119 ip_fanout_udp_multi_v4(mp, &ripha, up[0], up[1], ira);
2120 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2121 return;
2122
2123 case IPPROTO_TCP:
2124 /*
2125 * Find a TCP client stream for this packet.
2126 * Note that we do a reverse lookup since the header is
2127 * in the form we sent it out.
2128 */
2129 tcpha = (tcpha_t *)((uchar_t *)ipha + hdr_length);
2130 connp = ipcl_tcp_lookup_reversed_ipv4(ipha, tcpha, TCPS_LISTEN,
2131 ipst);
2132 if (connp == NULL)
2133 goto discard_pkt;
2134
2135 if (CONN_INBOUND_POLICY_PRESENT(connp, ipss) ||
2136 (ira->ira_flags & IRAF_IPSEC_SECURE)) {
2137 mp = ipsec_check_inbound_policy(mp, connp,
2138 ipha, NULL, ira);
2139 if (mp == NULL) {
2140 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2141 /* Note that mp is NULL */
2142 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2143 CONN_DEC_REF(connp);
2144 return;
2145 }
2146 }
2147
2148 ira->ira_flags |= IRAF_ICMP_ERROR;
2149 ira->ira_ill = ira->ira_rill = NULL;
2150 if (IPCL_IS_TCP(connp)) {
2151 SQUEUE_ENTER_ONE(connp->conn_sqp, mp,
2152 connp->conn_recvicmp, connp, ira, SQ_FILL,
2153 SQTAG_TCP_INPUT_ICMP_ERR);
2154 } else {
2155 /* Not TCP; must be SOCK_RAW, IPPROTO_TCP */
2156 (connp->conn_recv)(connp, mp, NULL, ira);
2157 CONN_DEC_REF(connp);
2158 }
2159 ira->ira_ill = ill;
2160 ira->ira_rill = rill;
2161 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2162 return;
2163
2164 case IPPROTO_SCTP:
2165 up = (uint16_t *)((uchar_t *)ipha + hdr_length);
2166 /* Find a SCTP client stream for this packet. */
2167 ((uint16_t *)&ports)[0] = up[1];
2168 ((uint16_t *)&ports)[1] = up[0];
2169
2170 ira->ira_flags |= IRAF_ICMP_ERROR;
2171 ip_fanout_sctp(mp, &ripha, NULL, ports, ira);
2172 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2173 return;
2174
2175 case IPPROTO_ESP:
2176 case IPPROTO_AH:
2177 if (!ipsec_loaded(ipss)) {
2178 ip_proto_not_sup(mp, ira);
2179 return;
2180 }
2181
2182 if (ipha->ipha_protocol == IPPROTO_ESP)
2183 mp = ipsecesp_icmp_error(mp, ira);
2184 else
2185 mp = ipsecah_icmp_error(mp, ira);
2186 if (mp == NULL)
2187 return;
2188
2189 /* Just in case ipsec didn't preserve the NULL b_cont */
2190 if (mp->b_cont != NULL) {
2191 if (!pullupmsg(mp, -1))
2192 goto discard_pkt;
2193 }
2194
2195 /*
2196 * Note that ira_pktlen and ira_ip_hdr_length are no longer
2197 * correct, but we don't use them any more here.
2198 *
2199 * If succesful, the mp has been modified to not include
2200 * the ESP/AH header so we can fanout to the ULP's icmp
2201 * error handler.
2202 */
2203 if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH)
2204 goto truncated;
2205
2206 /* Verify the modified message before any further processes. */
2207 ipha = (ipha_t *)mp->b_rptr;
2208 hdr_length = IPH_HDR_LENGTH(ipha);
2209 icmph = (icmph_t *)&mp->b_rptr[hdr_length];
2210 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
2211 freemsg(mp);
2212 return;
2213 }
2214
2215 icmp_inbound_error_fanout_v4(mp, icmph, ira);
2216 return;
2217
2218 case IPPROTO_ENCAP: {
2219 /* Look for self-encapsulated packets that caused an error */
2220 ipha_t *in_ipha;
2221
2222 /*
2223 * Caller has verified that length has to be
2224 * at least the size of IP header.
2225 */
2226 ASSERT(hdr_length >= sizeof (ipha_t));
2227 /*
2228 * Check the sanity of the inner IP header like
2229 * we did for the outer header.
2230 */
2231 in_ipha = (ipha_t *)((uchar_t *)ipha + hdr_length);
2232 if ((IPH_HDR_VERSION(in_ipha) != IPV4_VERSION)) {
2233 goto discard_pkt;
2234 }
2235 if (IPH_HDR_LENGTH(in_ipha) < sizeof (ipha_t)) {
2236 goto discard_pkt;
2237 }
2238 /* Check for Self-encapsulated tunnels */
2239 if (in_ipha->ipha_src == ipha->ipha_src &&
2240 in_ipha->ipha_dst == ipha->ipha_dst) {
2241
2242 mp = icmp_inbound_self_encap_error_v4(mp, ipha,
2243 in_ipha);
2244 if (mp == NULL)
2245 goto discard_pkt;
2246
2247 /*
2248 * Just in case self_encap didn't preserve the NULL
2249 * b_cont
2250 */
2251 if (mp->b_cont != NULL) {
2252 if (!pullupmsg(mp, -1))
2253 goto discard_pkt;
2254 }
2255 /*
2256 * Note that ira_pktlen and ira_ip_hdr_length are no
2257 * longer correct, but we don't use them any more here.
2258 */
2259 if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH)
2260 goto truncated;
2261
2262 /*
2263 * Verify the modified message before any further
2264 * processes.
2265 */
2266 ipha = (ipha_t *)mp->b_rptr;
2267 hdr_length = IPH_HDR_LENGTH(ipha);
2268 icmph = (icmph_t *)&mp->b_rptr[hdr_length];
2269 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
2270 freemsg(mp);
2271 return;
2272 }
2273
2274 /*
2275 * The packet in error is self-encapsualted.
2276 * And we are finding it further encapsulated
2277 * which we could not have possibly generated.
2278 */
2279 if (ipha->ipha_protocol == IPPROTO_ENCAP) {
2280 goto discard_pkt;
2281 }
2282 icmp_inbound_error_fanout_v4(mp, icmph, ira);
2283 return;
2284 }
2285 /* No self-encapsulated */
2286 /* FALLTHRU */
2287 }
2288 case IPPROTO_IPV6:
2289 if ((connp = ipcl_iptun_classify_v4(&ripha.ipha_src,
2290 &ripha.ipha_dst, ipst)) != NULL) {
2291 ira->ira_flags |= IRAF_ICMP_ERROR;
2292 connp->conn_recvicmp(connp, mp, NULL, ira);
2293 CONN_DEC_REF(connp);
2294 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2295 return;
2296 }
2297 /*
2298 * No IP tunnel is interested, fallthrough and see
2299 * if a raw socket will want it.
2300 */
2301 /* FALLTHRU */
2302 default:
2303 ira->ira_flags |= IRAF_ICMP_ERROR;
2304 ip_fanout_proto_v4(mp, &ripha, ira);
2305 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2306 return;
2307 }
2308 /* NOTREACHED */
2309 discard_pkt:
2310 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2311 ip1dbg(("icmp_inbound_error_fanout_v4: drop pkt\n"));
2312 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2313 freemsg(mp);
2314 return;
2315
2316 truncated:
2317 /* We pulled up everthing already. Must be truncated */
2318 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
2319 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
2320 freemsg(mp);
2321 }
2322
2323 /*
2324 * Common IP options parser.
2325 *
2326 * Setup routine: fill in *optp with options-parsing state, then
2327 * tail-call ipoptp_next to return the first option.
2328 */
2329 uint8_t
2330 ipoptp_first(ipoptp_t *optp, ipha_t *ipha)
2331 {
2332 uint32_t totallen; /* total length of all options */
2333
2334 totallen = ipha->ipha_version_and_hdr_length -
2335 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
2336 totallen <<= 2;
2337 optp->ipoptp_next = (uint8_t *)(&ipha[1]);
2338 optp->ipoptp_end = optp->ipoptp_next + totallen;
2339 optp->ipoptp_flags = 0;
2340 return (ipoptp_next(optp));
2341 }
2342
2343 /* Like above but without an ipha_t */
2344 uint8_t
2345 ipoptp_first2(ipoptp_t *optp, uint32_t totallen, uint8_t *opt)
2346 {
2347 optp->ipoptp_next = opt;
2348 optp->ipoptp_end = optp->ipoptp_next + totallen;
2349 optp->ipoptp_flags = 0;
2350 return (ipoptp_next(optp));
2351 }
2352
2353 /*
2354 * Common IP options parser: extract next option.
2355 */
2356 uint8_t
2357 ipoptp_next(ipoptp_t *optp)
2358 {
2359 uint8_t *end = optp->ipoptp_end;
2360 uint8_t *cur = optp->ipoptp_next;
2361 uint8_t opt, len, pointer;
2362
2363 /*
2364 * If cur > end already, then the ipoptp_end or ipoptp_next pointer
2365 * has been corrupted.
2366 */
2367 ASSERT(cur <= end);
2368
2369 if (cur == end)
2370 return (IPOPT_EOL);
2371
2372 opt = cur[IPOPT_OPTVAL];
2373
2374 /*
2375 * Skip any NOP options.
2376 */
2377 while (opt == IPOPT_NOP) {
2378 cur++;
2379 if (cur == end)
2380 return (IPOPT_EOL);
2381 opt = cur[IPOPT_OPTVAL];
2382 }
2383
2384 if (opt == IPOPT_EOL)
2385 return (IPOPT_EOL);
2386
2387 /*
2388 * Option requiring a length.
2389 */
2390 if ((cur + 1) >= end) {
2391 optp->ipoptp_flags |= IPOPTP_ERROR;
2392 return (IPOPT_EOL);
2393 }
2394 len = cur[IPOPT_OLEN];
2395 if (len < 2) {
2396 optp->ipoptp_flags |= IPOPTP_ERROR;
2397 return (IPOPT_EOL);
2398 }
2399 optp->ipoptp_cur = cur;
2400 optp->ipoptp_len = len;
2401 optp->ipoptp_next = cur + len;
2402 if (cur + len > end) {
2403 optp->ipoptp_flags |= IPOPTP_ERROR;
2404 return (IPOPT_EOL);
2405 }
2406
2407 /*
2408 * For the options which require a pointer field, make sure
2409 * its there, and make sure it points to either something
2410 * inside this option, or the end of the option.
2411 */
2412 switch (opt) {
2413 case IPOPT_RR:
2414 case IPOPT_TS:
2415 case IPOPT_LSRR:
2416 case IPOPT_SSRR:
2417 if (len <= IPOPT_OFFSET) {
2418 optp->ipoptp_flags |= IPOPTP_ERROR;
2419 return (opt);
2420 }
2421 pointer = cur[IPOPT_OFFSET];
2422 if (pointer - 1 > len) {
2423 optp->ipoptp_flags |= IPOPTP_ERROR;
2424 return (opt);
2425 }
2426 break;
2427 }
2428
2429 /*
2430 * Sanity check the pointer field based on the type of the
2431 * option.
2432 */
2433 switch (opt) {
2434 case IPOPT_RR:
2435 case IPOPT_SSRR:
2436 case IPOPT_LSRR:
2437 if (pointer < IPOPT_MINOFF_SR)
2438 optp->ipoptp_flags |= IPOPTP_ERROR;
2439 break;
2440 case IPOPT_TS:
2441 if (pointer < IPOPT_MINOFF_IT)
2442 optp->ipoptp_flags |= IPOPTP_ERROR;
2443 /*
2444 * Note that the Internet Timestamp option also
2445 * contains two four bit fields (the Overflow field,
2446 * and the Flag field), which follow the pointer
2447 * field. We don't need to check that these fields
2448 * fall within the length of the option because this
2449 * was implicitely done above. We've checked that the
2450 * pointer value is at least IPOPT_MINOFF_IT, and that
2451 * it falls within the option. Since IPOPT_MINOFF_IT >
2452 * IPOPT_POS_OV_FLG, we don't need the explicit check.
2453 */
2454 ASSERT(len > IPOPT_POS_OV_FLG);
2455 break;
2456 }
2457
2458 return (opt);
2459 }
2460
2461 /*
2462 * Use the outgoing IP header to create an IP_OPTIONS option the way
2463 * it was passed down from the application.
2464 *
2465 * This is compatible with BSD in that it returns
2466 * the reverse source route with the final destination
2467 * as the last entry. The first 4 bytes of the option
2468 * will contain the final destination.
2469 */
2470 int
2471 ip_opt_get_user(conn_t *connp, uchar_t *buf)
2472 {
2473 ipoptp_t opts;
2474 uchar_t *opt;
2475 uint8_t optval;
2476 uint8_t optlen;
2477 uint32_t len = 0;
2478 uchar_t *buf1 = buf;
2479 uint32_t totallen;
2480 ipaddr_t dst;
2481 ip_pkt_t *ipp = &connp->conn_xmit_ipp;
2482
2483 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
2484 return (0);
2485
2486 totallen = ipp->ipp_ipv4_options_len;
2487 if (totallen & 0x3)
2488 return (0);
2489
2490 buf += IP_ADDR_LEN; /* Leave room for final destination */
2491 len += IP_ADDR_LEN;
2492 bzero(buf1, IP_ADDR_LEN);
2493
2494 dst = connp->conn_faddr_v4;
2495
2496 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
2497 optval != IPOPT_EOL;
2498 optval = ipoptp_next(&opts)) {
2499 int off;
2500
2501 opt = opts.ipoptp_cur;
2502 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
2503 break;
2504 }
2505 optlen = opts.ipoptp_len;
2506
2507 switch (optval) {
2508 case IPOPT_SSRR:
2509 case IPOPT_LSRR:
2510
2511 /*
2512 * Insert destination as the first entry in the source
2513 * route and move down the entries on step.
2514 * The last entry gets placed at buf1.
2515 */
2516 buf[IPOPT_OPTVAL] = optval;
2517 buf[IPOPT_OLEN] = optlen;
2518 buf[IPOPT_OFFSET] = optlen;
2519
2520 off = optlen - IP_ADDR_LEN;
2521 if (off < 0) {
2522 /* No entries in source route */
2523 break;
2524 }
2525 /* Last entry in source route if not already set */
2526 if (dst == INADDR_ANY)
2527 bcopy(opt + off, buf1, IP_ADDR_LEN);
2528 off -= IP_ADDR_LEN;
2529
2530 while (off > 0) {
2531 bcopy(opt + off,
2532 buf + off + IP_ADDR_LEN,
2533 IP_ADDR_LEN);
2534 off -= IP_ADDR_LEN;
2535 }
2536 /* ipha_dst into first slot */
2537 bcopy(&dst, buf + off + IP_ADDR_LEN,
2538 IP_ADDR_LEN);
2539 buf += optlen;
2540 len += optlen;
2541 break;
2542
2543 default:
2544 bcopy(opt, buf, optlen);
2545 buf += optlen;
2546 len += optlen;
2547 break;
2548 }
2549 }
2550 done:
2551 /* Pad the resulting options */
2552 while (len & 0x3) {
2553 *buf++ = IPOPT_EOL;
2554 len++;
2555 }
2556 return (len);
2557 }
2558
2559 /*
2560 * Update any record route or timestamp options to include this host.
2561 * Reverse any source route option.
2562 * This routine assumes that the options are well formed i.e. that they
2563 * have already been checked.
2564 */
2565 static void
2566 icmp_options_update(ipha_t *ipha)
2567 {
2568 ipoptp_t opts;
2569 uchar_t *opt;
2570 uint8_t optval;
2571 ipaddr_t src; /* Our local address */
2572 ipaddr_t dst;
2573
2574 ip2dbg(("icmp_options_update\n"));
2575 src = ipha->ipha_src;
2576 dst = ipha->ipha_dst;
2577
2578 for (optval = ipoptp_first(&opts, ipha);
2579 optval != IPOPT_EOL;
2580 optval = ipoptp_next(&opts)) {
2581 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
2582 opt = opts.ipoptp_cur;
2583 ip2dbg(("icmp_options_update: opt %d, len %d\n",
2584 optval, opts.ipoptp_len));
2585 switch (optval) {
2586 int off1, off2;
2587 case IPOPT_SSRR:
2588 case IPOPT_LSRR:
2589 /*
2590 * Reverse the source route. The first entry
2591 * should be the next to last one in the current
2592 * source route (the last entry is our address).
2593 * The last entry should be the final destination.
2594 */
2595 off1 = IPOPT_MINOFF_SR - 1;
2596 off2 = opt[IPOPT_OFFSET] - IP_ADDR_LEN - 1;
2597 if (off2 < 0) {
2598 /* No entries in source route */
2599 ip1dbg((
2600 "icmp_options_update: bad src route\n"));
2601 break;
2602 }
2603 bcopy((char *)opt + off2, &dst, IP_ADDR_LEN);
2604 bcopy(&ipha->ipha_dst, (char *)opt + off2, IP_ADDR_LEN);
2605 bcopy(&dst, &ipha->ipha_dst, IP_ADDR_LEN);
2606 off2 -= IP_ADDR_LEN;
2607
2608 while (off1 < off2) {
2609 bcopy((char *)opt + off1, &src, IP_ADDR_LEN);
2610 bcopy((char *)opt + off2, (char *)opt + off1,
2611 IP_ADDR_LEN);
2612 bcopy(&src, (char *)opt + off2, IP_ADDR_LEN);
2613 off1 += IP_ADDR_LEN;
2614 off2 -= IP_ADDR_LEN;
2615 }
2616 opt[IPOPT_OFFSET] = IPOPT_MINOFF_SR;
2617 break;
2618 }
2619 }
2620 }
2621
2622 /*
2623 * Process received ICMP Redirect messages.
2624 * Assumes the caller has verified that the headers are in the pulled up mblk.
2625 * Consumes mp.
2626 */
2627 static void
2628 icmp_redirect_v4(mblk_t *mp, ipha_t *ipha, icmph_t *icmph, ip_recv_attr_t *ira)
2629 {
2630 ire_t *ire, *nire;
2631 ire_t *prev_ire;
2632 ipaddr_t src, dst, gateway;
2633 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2634 ipha_t *inner_ipha; /* Inner IP header */
2635
2636 /* Caller already pulled up everything. */
2637 inner_ipha = (ipha_t *)&icmph[1];
2638 src = ipha->ipha_src;
2639 dst = inner_ipha->ipha_dst;
2640 gateway = icmph->icmph_rd_gateway;
2641 /* Make sure the new gateway is reachable somehow. */
2642 ire = ire_ftable_lookup_v4(gateway, 0, 0, IRE_ONLINK, NULL,
2643 ALL_ZONES, NULL, MATCH_IRE_TYPE, 0, ipst, NULL);
2644 /*
2645 * Make sure we had a route for the dest in question and that
2646 * that route was pointing to the old gateway (the source of the
2647 * redirect packet.)
2648 * We do longest match and then compare ire_gateway_addr below.
2649 */
2650 prev_ire = ire_ftable_lookup_v4(dst, 0, 0, 0, NULL, ALL_ZONES,
2651 NULL, MATCH_IRE_DSTONLY, 0, ipst, NULL);
2652 /*
2653 * Check that
2654 * the redirect was not from ourselves
2655 * the new gateway and the old gateway are directly reachable
2656 */
2657 if (prev_ire == NULL || ire == NULL ||
2658 (prev_ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK)) ||
2659 (prev_ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) ||
2660 !(ire->ire_type & IRE_IF_ALL) ||
2661 prev_ire->ire_gateway_addr != src) {
2662 BUMP_MIB(&ipst->ips_icmp_mib, icmpInBadRedirects);
2663 ip_drop_input("icmpInBadRedirects - ire", mp, ira->ira_ill);
2664 freemsg(mp);
2665 if (ire != NULL)
2666 ire_refrele(ire);
2667 if (prev_ire != NULL)
2668 ire_refrele(prev_ire);
2669 return;
2670 }
2671
2672 ire_refrele(prev_ire);
2673 ire_refrele(ire);
2674
2675 /*
2676 * TODO: more precise handling for cases 0, 2, 3, the latter two
2677 * require TOS routing
2678 */
2679 switch (icmph->icmph_code) {
2680 case 0:
2681 case 1:
2682 /* TODO: TOS specificity for cases 2 and 3 */
2683 case 2:
2684 case 3:
2685 break;
2686 default:
2687 BUMP_MIB(&ipst->ips_icmp_mib, icmpInBadRedirects);
2688 ip_drop_input("icmpInBadRedirects - code", mp, ira->ira_ill);
2689 freemsg(mp);
2690 return;
2691 }
2692 /*
2693 * Create a Route Association. This will allow us to remember that
2694 * someone we believe told us to use the particular gateway.
2695 */
2696 ire = ire_create(
2697 (uchar_t *)&dst, /* dest addr */
2698 (uchar_t *)&ip_g_all_ones, /* mask */
2699 (uchar_t *)&gateway, /* gateway addr */
2700 IRE_HOST,
2701 NULL, /* ill */
2702 ALL_ZONES,
2703 (RTF_DYNAMIC | RTF_GATEWAY | RTF_HOST),
2704 NULL, /* tsol_gc_t */
2705 ipst);
2706
2707 if (ire == NULL) {
2708 freemsg(mp);
2709 return;
2710 }
2711 nire = ire_add(ire);
2712 /* Check if it was a duplicate entry */
2713 if (nire != NULL && nire != ire) {
2714 ASSERT(nire->ire_identical_ref > 1);
2715 ire_delete(nire);
2716 ire_refrele(nire);
2717 nire = NULL;
2718 }
2719 ire = nire;
2720 if (ire != NULL) {
2721 ire_refrele(ire); /* Held in ire_add */
2722
2723 /* tell routing sockets that we received a redirect */
2724 ip_rts_change(RTM_REDIRECT, dst, gateway, IP_HOST_MASK, 0, src,
2725 (RTF_DYNAMIC | RTF_GATEWAY | RTF_HOST), 0,
2726 (RTA_DST | RTA_GATEWAY | RTA_NETMASK | RTA_AUTHOR), ipst);
2727 }
2728
2729 /*
2730 * Delete any existing IRE_HOST type redirect ires for this destination.
2731 * This together with the added IRE has the effect of
2732 * modifying an existing redirect.
2733 */
2734 prev_ire = ire_ftable_lookup_v4(dst, 0, src, IRE_HOST, NULL,
2735 ALL_ZONES, NULL, (MATCH_IRE_GW | MATCH_IRE_TYPE), 0, ipst, NULL);
2736 if (prev_ire != NULL) {
2737 if (prev_ire ->ire_flags & RTF_DYNAMIC)
2738 ire_delete(prev_ire);
2739 ire_refrele(prev_ire);
2740 }
2741
2742 freemsg(mp);
2743 }
2744
2745 /*
2746 * Generate an ICMP parameter problem message.
2747 * When called from ip_output side a minimal ip_recv_attr_t needs to be
2748 * constructed by the caller.
2749 */
2750 static void
2751 icmp_param_problem(mblk_t *mp, uint8_t ptr, ip_recv_attr_t *ira)
2752 {
2753 icmph_t icmph;
2754 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2755
2756 mp = icmp_pkt_err_ok(mp, ira);
2757 if (mp == NULL)
2758 return;
2759
2760 bzero(&icmph, sizeof (icmph_t));
2761 icmph.icmph_type = ICMP_PARAM_PROBLEM;
2762 icmph.icmph_pp_ptr = ptr;
2763 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutParmProbs);
2764 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
2765 }
2766
2767 /*
2768 * Build and ship an IPv4 ICMP message using the packet data in mp, and
2769 * the ICMP header pointed to by "stuff". (May be called as writer.)
2770 * Note: assumes that icmp_pkt_err_ok has been called to verify that
2771 * an icmp error packet can be sent.
2772 * Assigns an appropriate source address to the packet. If ipha_dst is
2773 * one of our addresses use it for source. Otherwise let ip_output_simple
2774 * pick the source address.
2775 */
2776 static void
2777 icmp_pkt(mblk_t *mp, void *stuff, size_t len, ip_recv_attr_t *ira)
2778 {
2779 ipaddr_t dst;
2780 icmph_t *icmph;
2781 ipha_t *ipha;
2782 uint_t len_needed;
2783 size_t msg_len;
2784 mblk_t *mp1;
2785 ipaddr_t src;
2786 ire_t *ire;
2787 ip_xmit_attr_t ixas;
2788 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2789
2790 ipha = (ipha_t *)mp->b_rptr;
2791
2792 bzero(&ixas, sizeof (ixas));
2793 ixas.ixa_flags = IXAF_BASIC_SIMPLE_V4;
2794 ixas.ixa_zoneid = ira->ira_zoneid;
2795 ixas.ixa_ifindex = 0;
2796 ixas.ixa_ipst = ipst;
2797 ixas.ixa_cred = kcred;
2798 ixas.ixa_cpid = NOPID;
2799 ixas.ixa_tsl = ira->ira_tsl; /* Behave as a multi-level responder */
2800 ixas.ixa_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
2801
2802 if (ira->ira_flags & IRAF_IPSEC_SECURE) {
2803 /*
2804 * Apply IPsec based on how IPsec was applied to
2805 * the packet that had the error.
2806 *
2807 * If it was an outbound packet that caused the ICMP
2808 * error, then the caller will have setup the IRA
2809 * appropriately.
2810 */
2811 if (!ipsec_in_to_out(ira, &ixas, mp, ipha, NULL)) {
2812 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
2813 /* Note: mp already consumed and ip_drop_packet done */
2814 return;
2815 }
2816 } else {
2817 /*
2818 * This is in clear. The icmp message we are building
2819 * here should go out in clear, independent of our policy.
2820 */
2821 ixas.ixa_flags |= IXAF_NO_IPSEC;
2822 }
2823
2824 /* Remember our eventual destination */
2825 dst = ipha->ipha_src;
2826
2827 /*
2828 * If the packet was for one of our unicast addresses, make
2829 * sure we respond with that as the source. Otherwise
2830 * have ip_output_simple pick the source address.
2831 */
2832 ire = ire_ftable_lookup_v4(ipha->ipha_dst, 0, 0,
2833 (IRE_LOCAL|IRE_LOOPBACK), NULL, ira->ira_zoneid, NULL,
2834 MATCH_IRE_TYPE|MATCH_IRE_ZONEONLY, 0, ipst, NULL);
2835 if (ire != NULL) {
2836 ire_refrele(ire);
2837 src = ipha->ipha_dst;
2838 } else {
2839 src = INADDR_ANY;
2840 ixas.ixa_flags |= IXAF_SET_SOURCE;
2841 }
2842
2843 /*
2844 * Check if we can send back more then 8 bytes in addition to
2845 * the IP header. We try to send 64 bytes of data and the internal
2846 * header in the special cases of ipv4 encapsulated ipv4 or ipv6.
2847 */
2848 len_needed = IPH_HDR_LENGTH(ipha);
2849 if (ipha->ipha_protocol == IPPROTO_ENCAP ||
2850 ipha->ipha_protocol == IPPROTO_IPV6) {
2851 if (!pullupmsg(mp, -1)) {
2852 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
2853 ip_drop_output("ipIfStatsOutDiscards", mp, NULL);
2854 freemsg(mp);
2855 return;
2856 }
2857 ipha = (ipha_t *)mp->b_rptr;
2858
2859 if (ipha->ipha_protocol == IPPROTO_ENCAP) {
2860 len_needed += IPH_HDR_LENGTH(((uchar_t *)ipha +
2861 len_needed));
2862 } else {
2863 ip6_t *ip6h = (ip6_t *)((uchar_t *)ipha + len_needed);
2864
2865 ASSERT(ipha->ipha_protocol == IPPROTO_IPV6);
2866 len_needed += ip_hdr_length_v6(mp, ip6h);
2867 }
2868 }
2869 len_needed += ipst->ips_ip_icmp_return;
2870 msg_len = msgdsize(mp);
2871 if (msg_len > len_needed) {
2872 (void) adjmsg(mp, len_needed - msg_len);
2873 msg_len = len_needed;
2874 }
2875 mp1 = allocb(sizeof (icmp_ipha) + len, BPRI_MED);
2876 if (mp1 == NULL) {
2877 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutErrors);
2878 freemsg(mp);
2879 return;
2880 }
2881 mp1->b_cont = mp;
2882 mp = mp1;
2883
2884 /*
2885 * Set IXAF_TRUSTED_ICMP so we can let the ICMP messages this
2886 * node generates be accepted in peace by all on-host destinations.
2887 * If we do NOT assume that all on-host destinations trust
2888 * self-generated ICMP messages, then rework here, ip6.c, and spd.c.
2889 * (Look for IXAF_TRUSTED_ICMP).
2890 */
2891 ixas.ixa_flags |= IXAF_TRUSTED_ICMP;
2892
2893 ipha = (ipha_t *)mp->b_rptr;
2894 mp1->b_wptr = (uchar_t *)ipha + (sizeof (icmp_ipha) + len);
2895 *ipha = icmp_ipha;
2896 ipha->ipha_src = src;
2897 ipha->ipha_dst = dst;
2898 ipha->ipha_ttl = ipst->ips_ip_def_ttl;
2899 msg_len += sizeof (icmp_ipha) + len;
2900 if (msg_len > IP_MAXPACKET) {
2901 (void) adjmsg(mp, IP_MAXPACKET - msg_len);
2902 msg_len = IP_MAXPACKET;
2903 }
2904 ipha->ipha_length = htons((uint16_t)msg_len);
2905 icmph = (icmph_t *)&ipha[1];
2906 bcopy(stuff, icmph, len);
2907 icmph->icmph_checksum = 0;
2908 icmph->icmph_checksum = IP_CSUM(mp, (int32_t)sizeof (ipha_t), 0);
2909 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutMsgs);
2910
2911 (void) ip_output_simple(mp, &ixas);
2912 ixa_cleanup(&ixas);
2913 }
2914
2915 /*
2916 * Determine if an ICMP error packet can be sent given the rate limit.
2917 * The limit consists of an average frequency (icmp_pkt_err_interval measured
2918 * in milliseconds) and a burst size. Burst size number of packets can
2919 * be sent arbitrarely closely spaced.
2920 * The state is tracked using two variables to implement an approximate
2921 * token bucket filter:
2922 * icmp_pkt_err_last - lbolt value when the last burst started
2923 * icmp_pkt_err_sent - number of packets sent in current burst
2924 */
2925 boolean_t
2926 icmp_err_rate_limit(ip_stack_t *ipst)
2927 {
2928 clock_t now = TICK_TO_MSEC(ddi_get_lbolt());
2929 uint_t refilled; /* Number of packets refilled in tbf since last */
2930 /* Guard against changes by loading into local variable */
2931 uint_t err_interval = ipst->ips_ip_icmp_err_interval;
2932
2933 if (err_interval == 0)
2934 return (B_FALSE);
2935
2936 if (ipst->ips_icmp_pkt_err_last > now) {
2937 /* 100HZ lbolt in ms for 32bit arch wraps every 49.7 days */
2938 ipst->ips_icmp_pkt_err_last = 0;
2939 ipst->ips_icmp_pkt_err_sent = 0;
2940 }
2941 /*
2942 * If we are in a burst update the token bucket filter.
2943 * Update the "last" time to be close to "now" but make sure
2944 * we don't loose precision.
2945 */
2946 if (ipst->ips_icmp_pkt_err_sent != 0) {
2947 refilled = (now - ipst->ips_icmp_pkt_err_last)/err_interval;
2948 if (refilled > ipst->ips_icmp_pkt_err_sent) {
2949 ipst->ips_icmp_pkt_err_sent = 0;
2950 } else {
2951 ipst->ips_icmp_pkt_err_sent -= refilled;
2952 ipst->ips_icmp_pkt_err_last += refilled * err_interval;
2953 }
2954 }
2955 if (ipst->ips_icmp_pkt_err_sent == 0) {
2956 /* Start of new burst */
2957 ipst->ips_icmp_pkt_err_last = now;
2958 }
2959 if (ipst->ips_icmp_pkt_err_sent < ipst->ips_ip_icmp_err_burst) {
2960 ipst->ips_icmp_pkt_err_sent++;
2961 ip1dbg(("icmp_err_rate_limit: %d sent in burst\n",
2962 ipst->ips_icmp_pkt_err_sent));
2963 return (B_FALSE);
2964 }
2965 ip1dbg(("icmp_err_rate_limit: dropped\n"));
2966 return (B_TRUE);
2967 }
2968
2969 /*
2970 * Check if it is ok to send an IPv4 ICMP error packet in
2971 * response to the IPv4 packet in mp.
2972 * Free the message and return null if no
2973 * ICMP error packet should be sent.
2974 */
2975 static mblk_t *
2976 icmp_pkt_err_ok(mblk_t *mp, ip_recv_attr_t *ira)
2977 {
2978 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2979 icmph_t *icmph;
2980 ipha_t *ipha;
2981 uint_t len_needed;
2982
2983 if (!mp)
2984 return (NULL);
2985 ipha = (ipha_t *)mp->b_rptr;
2986 if (ip_csum_hdr(ipha)) {
2987 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsInCksumErrs);
2988 ip_drop_input("ipIfStatsInCksumErrs", mp, NULL);
2989 freemsg(mp);
2990 return (NULL);
2991 }
2992 if (ip_type_v4(ipha->ipha_dst, ipst) == IRE_BROADCAST ||
2993 ip_type_v4(ipha->ipha_src, ipst) == IRE_BROADCAST ||
2994 CLASSD(ipha->ipha_dst) ||
2995 CLASSD(ipha->ipha_src) ||
2996 (ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_OFFSET)) {
2997 /* Note: only errors to the fragment with offset 0 */
2998 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
2999 freemsg(mp);
3000 return (NULL);
3001 }
3002 if (ipha->ipha_protocol == IPPROTO_ICMP) {
3003 /*
3004 * Check the ICMP type. RFC 1122 sez: don't send ICMP
3005 * errors in response to any ICMP errors.
3006 */
3007 len_needed = IPH_HDR_LENGTH(ipha) + ICMPH_SIZE;
3008 if (mp->b_wptr - mp->b_rptr < len_needed) {
3009 if (!pullupmsg(mp, len_needed)) {
3010 BUMP_MIB(&ipst->ips_icmp_mib, icmpInErrors);
3011 freemsg(mp);
3012 return (NULL);
3013 }
3014 ipha = (ipha_t *)mp->b_rptr;
3015 }
3016 icmph = (icmph_t *)
3017 (&((char *)ipha)[IPH_HDR_LENGTH(ipha)]);
3018 switch (icmph->icmph_type) {
3019 case ICMP_DEST_UNREACHABLE:
3020 case ICMP_SOURCE_QUENCH:
3021 case ICMP_TIME_EXCEEDED:
3022 case ICMP_PARAM_PROBLEM:
3023 case ICMP_REDIRECT:
3024 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
3025 freemsg(mp);
3026 return (NULL);
3027 default:
3028 break;
3029 }
3030 }
3031 /*
3032 * If this is a labeled system, then check to see if we're allowed to
3033 * send a response to this particular sender. If not, then just drop.
3034 */
3035 if (is_system_labeled() && !tsol_can_reply_error(mp, ira)) {
3036 ip2dbg(("icmp_pkt_err_ok: can't respond to packet\n"));
3037 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
3038 freemsg(mp);
3039 return (NULL);
3040 }
3041 if (icmp_err_rate_limit(ipst)) {
3042 /*
3043 * Only send ICMP error packets every so often.
3044 * This should be done on a per port/source basis,
3045 * but for now this will suffice.
3046 */
3047 freemsg(mp);
3048 return (NULL);
3049 }
3050 return (mp);
3051 }
3052
3053 /*
3054 * Called when a packet was sent out the same link that it arrived on.
3055 * Check if it is ok to send a redirect and then send it.
3056 */
3057 void
3058 ip_send_potential_redirect_v4(mblk_t *mp, ipha_t *ipha, ire_t *ire,
3059 ip_recv_attr_t *ira)
3060 {
3061 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3062 ipaddr_t src, nhop;
3063 mblk_t *mp1;
3064 ire_t *nhop_ire;
3065
3066 /*
3067 * Check the source address to see if it originated
3068 * on the same logical subnet it is going back out on.
3069 * If so, we should be able to send it a redirect.
3070 * Avoid sending a redirect if the destination
3071 * is directly connected (i.e., we matched an IRE_ONLINK),
3072 * or if the packet was source routed out this interface.
3073 *
3074 * We avoid sending a redirect if the
3075 * destination is directly connected
3076 * because it is possible that multiple
3077 * IP subnets may have been configured on
3078 * the link, and the source may not
3079 * be on the same subnet as ip destination,
3080 * even though they are on the same
3081 * physical link.
3082 */
3083 if ((ire->ire_type & IRE_ONLINK) ||
3084 ip_source_routed(ipha, ipst))
3085 return;
3086
3087 nhop_ire = ire_nexthop(ire);
3088 if (nhop_ire == NULL)
3089 return;
3090
3091 nhop = nhop_ire->ire_addr;
3092
3093 if (nhop_ire->ire_type & IRE_IF_CLONE) {
3094 ire_t *ire2;
3095
3096 /* Follow ire_dep_parent to find non-clone IRE_INTERFACE */
3097 mutex_enter(&nhop_ire->ire_lock);
3098 ire2 = nhop_ire->ire_dep_parent;
3099 if (ire2 != NULL)
3100 ire_refhold(ire2);
3101 mutex_exit(&nhop_ire->ire_lock);
3102 ire_refrele(nhop_ire);
3103 nhop_ire = ire2;
3104 }
3105 if (nhop_ire == NULL)
3106 return;
3107
3108 ASSERT(!(nhop_ire->ire_type & IRE_IF_CLONE));
3109
3110 src = ipha->ipha_src;
3111
3112 /*
3113 * We look at the interface ire for the nexthop,
3114 * to see if ipha_src is in the same subnet
3115 * as the nexthop.
3116 */
3117 if ((src & nhop_ire->ire_mask) == (nhop & nhop_ire->ire_mask)) {
3118 /*
3119 * The source is directly connected.
3120 */
3121 mp1 = copymsg(mp);
3122 if (mp1 != NULL) {
3123 icmp_send_redirect(mp1, nhop, ira);
3124 }
3125 }
3126 ire_refrele(nhop_ire);
3127 }
3128
3129 /*
3130 * Generate an ICMP redirect message.
3131 */
3132 static void
3133 icmp_send_redirect(mblk_t *mp, ipaddr_t gateway, ip_recv_attr_t *ira)
3134 {
3135 icmph_t icmph;
3136 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3137
3138 mp = icmp_pkt_err_ok(mp, ira);
3139 if (mp == NULL)
3140 return;
3141
3142 bzero(&icmph, sizeof (icmph_t));
3143 icmph.icmph_type = ICMP_REDIRECT;
3144 icmph.icmph_code = 1;
3145 icmph.icmph_rd_gateway = gateway;
3146 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutRedirects);
3147 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3148 }
3149
3150 /*
3151 * Generate an ICMP time exceeded message.
3152 */
3153 void
3154 icmp_time_exceeded(mblk_t *mp, uint8_t code, ip_recv_attr_t *ira)
3155 {
3156 icmph_t icmph;
3157 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3158
3159 mp = icmp_pkt_err_ok(mp, ira);
3160 if (mp == NULL)
3161 return;
3162
3163 bzero(&icmph, sizeof (icmph_t));
3164 icmph.icmph_type = ICMP_TIME_EXCEEDED;
3165 icmph.icmph_code = code;
3166 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutTimeExcds);
3167 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3168 }
3169
3170 /*
3171 * Generate an ICMP unreachable message.
3172 * When called from ip_output side a minimal ip_recv_attr_t needs to be
3173 * constructed by the caller.
3174 */
3175 void
3176 icmp_unreachable(mblk_t *mp, uint8_t code, ip_recv_attr_t *ira)
3177 {
3178 icmph_t icmph;
3179 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3180
3181 mp = icmp_pkt_err_ok(mp, ira);
3182 if (mp == NULL)
3183 return;
3184
3185 bzero(&icmph, sizeof (icmph_t));
3186 icmph.icmph_type = ICMP_DEST_UNREACHABLE;
3187 icmph.icmph_code = code;
3188 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDestUnreachs);
3189 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3190 }
3191
3192 /*
3193 * Latch in the IPsec state for a stream based the policy in the listener
3194 * and the actions in the ip_recv_attr_t.
3195 * Called directly from TCP and SCTP.
3196 */
3197 boolean_t
3198 ip_ipsec_policy_inherit(conn_t *connp, conn_t *lconnp, ip_recv_attr_t *ira)
3199 {
3200 ASSERT(lconnp->conn_policy != NULL);
3201 ASSERT(connp->conn_policy == NULL);
3202
3203 IPPH_REFHOLD(lconnp->conn_policy);
3204 connp->conn_policy = lconnp->conn_policy;
3205
3206 if (ira->ira_ipsec_action != NULL) {
3207 if (connp->conn_latch == NULL) {
3208 connp->conn_latch = iplatch_create();
3209 if (connp->conn_latch == NULL)
3210 return (B_FALSE);
3211 }
3212 ipsec_latch_inbound(connp, ira);
3213 }
3214 return (B_TRUE);
3215 }
3216
3217 /*
3218 * Verify whether or not the IP address is a valid local address.
3219 * Could be a unicast, including one for a down interface.
3220 * If allow_mcbc then a multicast or broadcast address is also
3221 * acceptable.
3222 *
3223 * In the case of a broadcast/multicast address, however, the
3224 * upper protocol is expected to reset the src address
3225 * to zero when we return IPVL_MCAST/IPVL_BCAST so that
3226 * no packets are emitted with broadcast/multicast address as
3227 * source address (that violates hosts requirements RFC 1122)
3228 * The addresses valid for bind are:
3229 * (1) - INADDR_ANY (0)
3230 * (2) - IP address of an UP interface
3231 * (3) - IP address of a DOWN interface
3232 * (4) - valid local IP broadcast addresses. In this case
3233 * the conn will only receive packets destined to
3234 * the specified broadcast address.
3235 * (5) - a multicast address. In this case
3236 * the conn will only receive packets destined to
3237 * the specified multicast address. Note: the
3238 * application still has to issue an
3239 * IP_ADD_MEMBERSHIP socket option.
3240 *
3241 * In all the above cases, the bound address must be valid in the current zone.
3242 * When the address is loopback, multicast or broadcast, there might be many
3243 * matching IREs so bind has to look up based on the zone.
3244 */
3245 ip_laddr_t
3246 ip_laddr_verify_v4(ipaddr_t src_addr, zoneid_t zoneid,
3247 ip_stack_t *ipst, boolean_t allow_mcbc)
3248 {
3249 ire_t *src_ire;
3250
3251 ASSERT(src_addr != INADDR_ANY);
3252
3253 src_ire = ire_ftable_lookup_v4(src_addr, 0, 0, 0,
3254 NULL, zoneid, NULL, MATCH_IRE_ZONEONLY, 0, ipst, NULL);
3255
3256 /*
3257 * If an address other than in6addr_any is requested,
3258 * we verify that it is a valid address for bind
3259 * Note: Following code is in if-else-if form for
3260 * readability compared to a condition check.
3261 */
3262 if (src_ire != NULL && (src_ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK))) {
3263 /*
3264 * (2) Bind to address of local UP interface
3265 */
3266 ire_refrele(src_ire);
3267 return (IPVL_UNICAST_UP);
3268 } else if (src_ire != NULL && src_ire->ire_type & IRE_BROADCAST) {
3269 /*
3270 * (4) Bind to broadcast address
3271 */
3272 ire_refrele(src_ire);
3273 if (allow_mcbc)
3274 return (IPVL_BCAST);
3275 else
3276 return (IPVL_BAD);
3277 } else if (CLASSD(src_addr)) {
3278 /* (5) bind to multicast address. */
3279 if (src_ire != NULL)
3280 ire_refrele(src_ire);
3281
3282 if (allow_mcbc)
3283 return (IPVL_MCAST);
3284 else
3285 return (IPVL_BAD);
3286 } else {
3287 ipif_t *ipif;
3288
3289 /*
3290 * (3) Bind to address of local DOWN interface?
3291 * (ipif_lookup_addr() looks up all interfaces
3292 * but we do not get here for UP interfaces
3293 * - case (2) above)
3294 */
3295 if (src_ire != NULL)
3296 ire_refrele(src_ire);
3297
3298 ipif = ipif_lookup_addr(src_addr, NULL, zoneid, ipst);
3299 if (ipif == NULL)
3300 return (IPVL_BAD);
3301
3302 /* Not a useful source? */
3303 if (ipif->ipif_flags & (IPIF_NOLOCAL | IPIF_ANYCAST)) {
3304 ipif_refrele(ipif);
3305 return (IPVL_BAD);
3306 }
3307 ipif_refrele(ipif);
3308 return (IPVL_UNICAST_DOWN);
3309 }
3310 }
3311
3312 /*
3313 * Insert in the bind fanout for IPv4 and IPv6.
3314 * The caller should already have used ip_laddr_verify_v*() before calling
3315 * this.
3316 */
3317 int
3318 ip_laddr_fanout_insert(conn_t *connp)
3319 {
3320 int error;
3321
3322 /*
3323 * Allow setting new policies. For example, disconnects result
3324 * in us being called. As we would have set conn_policy_cached
3325 * to B_TRUE before, we should set it to B_FALSE, so that policy
3326 * can change after the disconnect.
3327 */
3328 connp->conn_policy_cached = B_FALSE;
3329
3330 error = ipcl_bind_insert(connp);
3331 if (error != 0) {
3332 if (connp->conn_anon_port) {
3333 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
3334 connp->conn_mlp_type, connp->conn_proto,
3335 ntohs(connp->conn_lport), B_FALSE);
3336 }
3337 connp->conn_mlp_type = mlptSingle;
3338 }
3339 return (error);
3340 }
3341
3342 /*
3343 * Verify that both the source and destination addresses are valid. If
3344 * IPDF_VERIFY_DST is not set, then the destination address may be unreachable,
3345 * i.e. have no route to it. Protocols like TCP want to verify destination
3346 * reachability, while tunnels do not.
3347 *
3348 * Determine the route, the interface, and (optionally) the source address
3349 * to use to reach a given destination.
3350 * Note that we allow connect to broadcast and multicast addresses when
3351 * IPDF_ALLOW_MCBC is set.
3352 * first_hop and dst_addr are normally the same, but if source routing
3353 * they will differ; in that case the first_hop is what we'll use for the
3354 * routing lookup but the dce and label checks will be done on dst_addr,
3355 *
3356 * If uinfo is set, then we fill in the best available information
3357 * we have for the destination. This is based on (in priority order) any
3358 * metrics and path MTU stored in a dce_t, route metrics, and finally the
3359 * ill_mtu/ill_mc_mtu.
3360 *
3361 * Tsol note: If we have a source route then dst_addr != firsthop. But we
3362 * always do the label check on dst_addr.
3363 */
3364 int
3365 ip_set_destination_v4(ipaddr_t *src_addrp, ipaddr_t dst_addr, ipaddr_t firsthop,
3366 ip_xmit_attr_t *ixa, iulp_t *uinfo, uint32_t flags, uint_t mac_mode)
3367 {
3368 ire_t *ire = NULL;
3369 int error = 0;
3370 ipaddr_t setsrc; /* RTF_SETSRC */
3371 zoneid_t zoneid = ixa->ixa_zoneid; /* Honors SO_ALLZONES */
3372 ip_stack_t *ipst = ixa->ixa_ipst;
3373 dce_t *dce;
3374 uint_t pmtu;
3375 uint_t generation;
3376 nce_t *nce;
3377 ill_t *ill = NULL;
3378 boolean_t multirt = B_FALSE;
3379
3380 ASSERT(ixa->ixa_flags & IXAF_IS_IPV4);
3381
3382 /*
3383 * We never send to zero; the ULPs map it to the loopback address.
3384 * We can't allow it since we use zero to mean unitialized in some
3385 * places.
3386 */
3387 ASSERT(dst_addr != INADDR_ANY);
3388
3389 if (is_system_labeled()) {
3390 ts_label_t *tsl = NULL;
3391
3392 error = tsol_check_dest(ixa->ixa_tsl, &dst_addr, IPV4_VERSION,
3393 mac_mode, (flags & IPDF_ZONE_IS_GLOBAL) != 0, &tsl);
3394 if (error != 0)
3395 return (error);
3396 if (tsl != NULL) {
3397 /* Update the label */
3398 ip_xmit_attr_replace_tsl(ixa, tsl);
3399 }
3400 }
3401
3402 setsrc = INADDR_ANY;
3403 /*
3404 * Select a route; For IPMP interfaces, we would only select
3405 * a "hidden" route (i.e., going through a specific under_ill)
3406 * if ixa_ifindex has been specified.
3407 */
3408 ire = ip_select_route_v4(firsthop, *src_addrp, ixa,
3409 &generation, &setsrc, &error, &multirt);
3410 ASSERT(ire != NULL); /* IRE_NOROUTE if none found */
3411 if (error != 0)
3412 goto bad_addr;
3413
3414 /*
3415 * ire can't be a broadcast or multicast unless IPDF_ALLOW_MCBC is set.
3416 * If IPDF_VERIFY_DST is set, the destination must be reachable;
3417 * Otherwise the destination needn't be reachable.
3418 *
3419 * If we match on a reject or black hole, then we've got a
3420 * local failure. May as well fail out the connect() attempt,
3421 * since it's never going to succeed.
3422 */
3423 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
3424 /*
3425 * If we're verifying destination reachability, we always want
3426 * to complain here.
3427 *
3428 * If we're not verifying destination reachability but the
3429 * destination has a route, we still want to fail on the
3430 * temporary address and broadcast address tests.
3431 *
3432 * In both cases do we let the code continue so some reasonable
3433 * information is returned to the caller. That enables the
3434 * caller to use (and even cache) the IRE. conn_ip_ouput will
3435 * use the generation mismatch path to check for the unreachable
3436 * case thereby avoiding any specific check in the main path.
3437 */
3438 ASSERT(generation == IRE_GENERATION_VERIFY);
3439 if (flags & IPDF_VERIFY_DST) {
3440 /*
3441 * Set errno but continue to set up ixa_ire to be
3442 * the RTF_REJECT|RTF_BLACKHOLE IRE.
3443 * That allows callers to use ip_output to get an
3444 * ICMP error back.
3445 */
3446 if (!(ire->ire_type & IRE_HOST))
3447 error = ENETUNREACH;
3448 else
3449 error = EHOSTUNREACH;
3450 }
3451 }
3452
3453 if ((ire->ire_type & (IRE_BROADCAST|IRE_MULTICAST)) &&
3454 !(flags & IPDF_ALLOW_MCBC)) {
3455 ire_refrele(ire);
3456 ire = ire_reject(ipst, B_FALSE);
3457 generation = IRE_GENERATION_VERIFY;
3458 error = ENETUNREACH;
3459 }
3460
3461 /* Cache things */
3462 if (ixa->ixa_ire != NULL)
3463 ire_refrele_notr(ixa->ixa_ire);
3464 #ifdef DEBUG
3465 ire_refhold_notr(ire);
3466 ire_refrele(ire);
3467 #endif
3468 ixa->ixa_ire = ire;
3469 ixa->ixa_ire_generation = generation;
3470
3471 /*
3472 * Ensure that ixa_dce is always set any time that ixa_ire is set,
3473 * since some callers will send a packet to conn_ip_output() even if
3474 * there's an error.
3475 */
3476 if (flags & IPDF_UNIQUE_DCE) {
3477 /* Fallback to the default dce if allocation fails */
3478 dce = dce_lookup_and_add_v4(dst_addr, ipst);
3479 if (dce != NULL)
3480 generation = dce->dce_generation;
3481 else
3482 dce = dce_lookup_v4(dst_addr, ipst, &generation);
3483 } else {
3484 dce = dce_lookup_v4(dst_addr, ipst, &generation);
3485 }
3486 ASSERT(dce != NULL);
3487 if (ixa->ixa_dce != NULL)
3488 dce_refrele_notr(ixa->ixa_dce);
3489 #ifdef DEBUG
3490 dce_refhold_notr(dce);
3491 dce_refrele(dce);
3492 #endif
3493 ixa->ixa_dce = dce;
3494 ixa->ixa_dce_generation = generation;
3495
3496 /*
3497 * For multicast with multirt we have a flag passed back from
3498 * ire_lookup_multi_ill_v4 since we don't have an IRE for each
3499 * possible multicast address.
3500 * We also need a flag for multicast since we can't check
3501 * whether RTF_MULTIRT is set in ixa_ire for multicast.
3502 */
3503 if (multirt) {
3504 ixa->ixa_postfragfn = ip_postfrag_multirt_v4;
3505 ixa->ixa_flags |= IXAF_MULTIRT_MULTICAST;
3506 } else {
3507 ixa->ixa_postfragfn = ire->ire_postfragfn;
3508 ixa->ixa_flags &= ~IXAF_MULTIRT_MULTICAST;
3509 }
3510 if (!(ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))) {
3511 /* Get an nce to cache. */
3512 nce = ire_to_nce(ire, firsthop, NULL);
3513 if (nce == NULL) {
3514 /* Allocation failure? */
3515 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3516 } else {
3517 if (ixa->ixa_nce != NULL)
3518 nce_refrele(ixa->ixa_nce);
3519 ixa->ixa_nce = nce;
3520 }
3521 }
3522
3523 /*
3524 * If the source address is a loopback address, the
3525 * destination had best be local or multicast.
3526 * If we are sending to an IRE_LOCAL using a loopback source then
3527 * it had better be the same zoneid.
3528 */
3529 if (*src_addrp == htonl(INADDR_LOOPBACK)) {
3530 if ((ire->ire_type & IRE_LOCAL) && ire->ire_zoneid != zoneid) {
3531 ire = NULL; /* Stored in ixa_ire */
3532 error = EADDRNOTAVAIL;
3533 goto bad_addr;
3534 }
3535 if (!(ire->ire_type & (IRE_LOOPBACK|IRE_LOCAL|IRE_MULTICAST))) {
3536 ire = NULL; /* Stored in ixa_ire */
3537 error = EADDRNOTAVAIL;
3538 goto bad_addr;
3539 }
3540 }
3541 if (ire->ire_type & IRE_BROADCAST) {
3542 /*
3543 * If the ULP didn't have a specified source, then we
3544 * make sure we reselect the source when sending
3545 * broadcasts out different interfaces.
3546 */
3547 if (flags & IPDF_SELECT_SRC)
3548 ixa->ixa_flags |= IXAF_SET_SOURCE;
3549 else
3550 ixa->ixa_flags &= ~IXAF_SET_SOURCE;
3551 }
3552
3553 /*
3554 * Does the caller want us to pick a source address?
3555 */
3556 if (flags & IPDF_SELECT_SRC) {
3557 ipaddr_t src_addr;
3558
3559 /*
3560 * We use use ire_nexthop_ill to avoid the under ipmp
3561 * interface for source address selection. Note that for ipmp
3562 * probe packets, ixa_ifindex would have been specified, and
3563 * the ip_select_route() invocation would have picked an ire
3564 * will ire_ill pointing at an under interface.
3565 */
3566 ill = ire_nexthop_ill(ire);
3567
3568 /* If unreachable we have no ill but need some source */
3569 if (ill == NULL) {
3570 src_addr = htonl(INADDR_LOOPBACK);
3571 /* Make sure we look for a better source address */
3572 generation = SRC_GENERATION_VERIFY;
3573 } else {
3574 error = ip_select_source_v4(ill, setsrc, dst_addr,
3575 ixa->ixa_multicast_ifaddr, zoneid,
3576 ipst, &src_addr, &generation, NULL);
3577 if (error != 0) {
3578 ire = NULL; /* Stored in ixa_ire */
3579 goto bad_addr;
3580 }
3581 }
3582
3583 /*
3584 * We allow the source address to to down.
3585 * However, we check that we don't use the loopback address
3586 * as a source when sending out on the wire.
3587 */
3588 if ((src_addr == htonl(INADDR_LOOPBACK)) &&
3589 !(ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK|IRE_MULTICAST)) &&
3590 !(ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))) {
3591 ire = NULL; /* Stored in ixa_ire */
3592 error = EADDRNOTAVAIL;
3593 goto bad_addr;
3594 }
3595
3596 *src_addrp = src_addr;
3597 ixa->ixa_src_generation = generation;
3598 }
3599
3600 /*
3601 * Make sure we don't leave an unreachable ixa_nce in place
3602 * since ip_select_route is used when we unplumb i.e., remove
3603 * references on ixa_ire, ixa_nce, and ixa_dce.
3604 */
3605 nce = ixa->ixa_nce;
3606 if (nce != NULL && nce->nce_is_condemned) {
3607 nce_refrele(nce);
3608 ixa->ixa_nce = NULL;
3609 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3610 }
3611
3612 /*
3613 * The caller has set IXAF_PMTU_DISCOVERY if path MTU is desired.
3614 * However, we can't do it for IPv4 multicast or broadcast.
3615 */
3616 if (ire->ire_type & (IRE_BROADCAST|IRE_MULTICAST))
3617 ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
3618
3619 /*
3620 * Set initial value for fragmentation limit. Either conn_ip_output
3621 * or ULP might updates it when there are routing changes.
3622 * Handles a NULL ixa_ire->ire_ill or a NULL ixa_nce for RTF_REJECT.
3623 */
3624 pmtu = ip_get_pmtu(ixa);
3625 ixa->ixa_fragsize = pmtu;
3626 /* Make sure ixa_fragsize and ixa_pmtu remain identical */
3627 if (ixa->ixa_flags & IXAF_VERIFY_PMTU)
3628 ixa->ixa_pmtu = pmtu;
3629
3630 /*
3631 * Extract information useful for some transports.
3632 * First we look for DCE metrics. Then we take what we have in
3633 * the metrics in the route, where the offlink is used if we have
3634 * one.
3635 */
3636 if (uinfo != NULL) {
3637 bzero(uinfo, sizeof (*uinfo));
3638
3639 if (dce->dce_flags & DCEF_UINFO)
3640 *uinfo = dce->dce_uinfo;
3641
3642 rts_merge_metrics(uinfo, &ire->ire_metrics);
3643
3644 /* Allow ire_metrics to decrease the path MTU from above */
3645 if (uinfo->iulp_mtu == 0 || uinfo->iulp_mtu > pmtu)
3646 uinfo->iulp_mtu = pmtu;
3647
3648 uinfo->iulp_localnet = (ire->ire_type & IRE_ONLINK) != 0;
3649 uinfo->iulp_loopback = (ire->ire_type & IRE_LOOPBACK) != 0;
3650 uinfo->iulp_local = (ire->ire_type & IRE_LOCAL) != 0;
3651 }
3652
3653 if (ill != NULL)
3654 ill_refrele(ill);
3655
3656 return (error);
3657
3658 bad_addr:
3659 if (ire != NULL)
3660 ire_refrele(ire);
3661
3662 if (ill != NULL)
3663 ill_refrele(ill);
3664
3665 /*
3666 * Make sure we don't leave an unreachable ixa_nce in place
3667 * since ip_select_route is used when we unplumb i.e., remove
3668 * references on ixa_ire, ixa_nce, and ixa_dce.
3669 */
3670 nce = ixa->ixa_nce;
3671 if (nce != NULL && nce->nce_is_condemned) {
3672 nce_refrele(nce);
3673 ixa->ixa_nce = NULL;
3674 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3675 }
3676
3677 return (error);
3678 }
3679
3680
3681 /*
3682 * Get the base MTU for the case when path MTU discovery is not used.
3683 * Takes the MTU of the IRE into account.
3684 */
3685 uint_t
3686 ip_get_base_mtu(ill_t *ill, ire_t *ire)
3687 {
3688 uint_t mtu;
3689 uint_t iremtu = ire->ire_metrics.iulp_mtu;
3690
3691 if (ire->ire_type & (IRE_MULTICAST|IRE_BROADCAST))
3692 mtu = ill->ill_mc_mtu;
3693 else
3694 mtu = ill->ill_mtu;
3695
3696 if (iremtu != 0 && iremtu < mtu)
3697 mtu = iremtu;
3698
3699 return (mtu);
3700 }
3701
3702 /*
3703 * Get the PMTU for the attributes. Handles both IPv4 and IPv6.
3704 * Assumes that ixa_ire, dce, and nce have already been set up.
3705 *
3706 * The caller has set IXAF_PMTU_DISCOVERY if path MTU discovery is desired.
3707 * We avoid path MTU discovery if it is disabled with ndd.
3708 * Furtermore, if the path MTU is too small, then we don't set DF for IPv4.
3709 *
3710 * NOTE: We also used to turn it off for source routed packets. That
3711 * is no longer required since the dce is per final destination.
3712 */
3713 uint_t
3714 ip_get_pmtu(ip_xmit_attr_t *ixa)
3715 {
3716 ip_stack_t *ipst = ixa->ixa_ipst;
3717 dce_t *dce;
3718 nce_t *nce;
3719 ire_t *ire;
3720 uint_t pmtu;
3721
3722 ire = ixa->ixa_ire;
3723 dce = ixa->ixa_dce;
3724 nce = ixa->ixa_nce;
3725
3726 /*
3727 * If path MTU discovery has been turned off by ndd, then we ignore
3728 * any dce_pmtu and for IPv4 we will not set DF.
3729 */
3730 if (!ipst->ips_ip_path_mtu_discovery)
3731 ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
3732
3733 pmtu = IP_MAXPACKET;
3734 /*
3735 * Decide whether whether IPv4 sets DF
3736 * For IPv6 "no DF" means to use the 1280 mtu
3737 */
3738 if (ixa->ixa_flags & IXAF_PMTU_DISCOVERY) {
3739 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3740 } else {
3741 ixa->ixa_flags &= ~IXAF_PMTU_IPV4_DF;
3742 if (!(ixa->ixa_flags & IXAF_IS_IPV4))
3743 pmtu = IPV6_MIN_MTU;
3744 }
3745
3746 /* Check if the PMTU is to old before we use it */
3747 if ((dce->dce_flags & DCEF_PMTU) &&
3748 TICK_TO_SEC(ddi_get_lbolt64()) - dce->dce_last_change_time >
3749 ipst->ips_ip_pathmtu_interval) {
3750 /*
3751 * Older than 20 minutes. Drop the path MTU information.
3752 */
3753 mutex_enter(&dce->dce_lock);
3754 dce->dce_flags &= ~(DCEF_PMTU|DCEF_TOO_SMALL_PMTU);
3755 dce->dce_last_change_time = TICK_TO_SEC(ddi_get_lbolt64());
3756 mutex_exit(&dce->dce_lock);
3757 dce_increment_generation(dce);
3758 }
3759
3760 /* The metrics on the route can lower the path MTU */
3761 if (ire->ire_metrics.iulp_mtu != 0 &&
3762 ire->ire_metrics.iulp_mtu < pmtu)
3763 pmtu = ire->ire_metrics.iulp_mtu;
3764
3765 /*
3766 * If the path MTU is smaller than some minimum, we still use dce_pmtu
3767 * above (would be 576 for IPv4 and 1280 for IPv6), but we clear
3768 * IXAF_PMTU_IPV4_DF so that we avoid setting DF for IPv4.
3769 */
3770 if (ixa->ixa_flags & IXAF_PMTU_DISCOVERY) {
3771 if (dce->dce_flags & DCEF_PMTU) {
3772 if (dce->dce_pmtu < pmtu)
3773 pmtu = dce->dce_pmtu;
3774
3775 if (dce->dce_flags & DCEF_TOO_SMALL_PMTU) {
3776 ixa->ixa_flags |= IXAF_PMTU_TOO_SMALL;
3777 ixa->ixa_flags &= ~IXAF_PMTU_IPV4_DF;
3778 } else {
3779 ixa->ixa_flags &= ~IXAF_PMTU_TOO_SMALL;
3780 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3781 }
3782 } else {
3783 ixa->ixa_flags &= ~IXAF_PMTU_TOO_SMALL;
3784 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3785 }
3786 }
3787
3788 /*
3789 * If we have an IRE_LOCAL we use the loopback mtu instead of
3790 * the ill for going out the wire i.e., IRE_LOCAL gets the same
3791 * mtu as IRE_LOOPBACK.
3792 */
3793 if (ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK)) {
3794 uint_t loopback_mtu;
3795
3796 loopback_mtu = (ire->ire_ipversion == IPV6_VERSION) ?
3797 ip_loopback_mtu_v6plus : ip_loopback_mtuplus;
3798
3799 if (loopback_mtu < pmtu)
3800 pmtu = loopback_mtu;
3801 } else if (nce != NULL) {
3802 /*
3803 * Make sure we don't exceed the interface MTU.
3804 * In the case of RTF_REJECT or RTF_BLACKHOLE we might not have
3805 * an ill. We'd use the above IP_MAXPACKET in that case just
3806 * to tell the transport something larger than zero.
3807 */
3808 if (ire->ire_type & (IRE_MULTICAST|IRE_BROADCAST)) {
3809 if (nce->nce_common->ncec_ill->ill_mc_mtu < pmtu)
3810 pmtu = nce->nce_common->ncec_ill->ill_mc_mtu;
3811 if (nce->nce_common->ncec_ill != nce->nce_ill &&
3812 nce->nce_ill->ill_mc_mtu < pmtu) {
3813 /*
3814 * for interfaces in an IPMP group, the mtu of
3815 * the nce_ill (under_ill) could be different
3816 * from the mtu of the ncec_ill, so we take the
3817 * min of the two.
3818 */
3819 pmtu = nce->nce_ill->ill_mc_mtu;
3820 }
3821 } else {
3822 if (nce->nce_common->ncec_ill->ill_mtu < pmtu)
3823 pmtu = nce->nce_common->ncec_ill->ill_mtu;
3824 if (nce->nce_common->ncec_ill != nce->nce_ill &&
3825 nce->nce_ill->ill_mtu < pmtu) {
3826 /*
3827 * for interfaces in an IPMP group, the mtu of
3828 * the nce_ill (under_ill) could be different
3829 * from the mtu of the ncec_ill, so we take the
3830 * min of the two.
3831 */
3832 pmtu = nce->nce_ill->ill_mtu;
3833 }
3834 }
3835 }
3836
3837 /*
3838 * Handle the IPV6_USE_MIN_MTU socket option or ancillary data.
3839 * Only applies to IPv6.
3840 */
3841 if (!(ixa->ixa_flags & IXAF_IS_IPV4)) {
3842 if (ixa->ixa_flags & IXAF_USE_MIN_MTU) {
3843 switch (ixa->ixa_use_min_mtu) {
3844 case IPV6_USE_MIN_MTU_MULTICAST:
3845 if (ire->ire_type & IRE_MULTICAST)
3846 pmtu = IPV6_MIN_MTU;
3847 break;
3848 case IPV6_USE_MIN_MTU_ALWAYS:
3849 pmtu = IPV6_MIN_MTU;
3850 break;
3851 case IPV6_USE_MIN_MTU_NEVER:
3852 break;
3853 }
3854 } else {
3855 /* Default is IPV6_USE_MIN_MTU_MULTICAST */
3856 if (ire->ire_type & IRE_MULTICAST)
3857 pmtu = IPV6_MIN_MTU;
3858 }
3859 }
3860
3861 /*
3862 * For multirouted IPv6 packets, the IP layer will insert a 8-byte
3863 * fragment header in every packet. We compensate for those cases by
3864 * returning a smaller path MTU to the ULP.
3865 *
3866 * In the case of CGTP then ip_output will add a fragment header.
3867 * Make sure there is room for it by telling a smaller number
3868 * to the transport.
3869 *
3870 * When IXAF_IPV6_ADDR_FRAGHDR we subtract the frag hdr here
3871 * so the ULPs consistently see a iulp_pmtu and ip_get_pmtu()
3872 * which is the size of the packets it can send.
3873 */
3874 if (!(ixa->ixa_flags & IXAF_IS_IPV4)) {
3875 if ((ire->ire_flags & RTF_MULTIRT) ||
3876 (ixa->ixa_flags & IXAF_MULTIRT_MULTICAST)) {
3877 pmtu -= sizeof (ip6_frag_t);
3878 ixa->ixa_flags |= IXAF_IPV6_ADD_FRAGHDR;
3879 }
3880 }
3881
3882 return (pmtu);
3883 }
3884
3885 /*
3886 * Carve "len" bytes out of an mblk chain, consuming any we empty, and duping
3887 * the final piece where we don't. Return a pointer to the first mblk in the
3888 * result, and update the pointer to the next mblk to chew on. If anything
3889 * goes wrong (i.e., dupb fails), we waste everything in sight and return a
3890 * NULL pointer.
3891 */
3892 mblk_t *
3893 ip_carve_mp(mblk_t **mpp, ssize_t len)
3894 {
3895 mblk_t *mp0;
3896 mblk_t *mp1;
3897 mblk_t *mp2;
3898
3899 if (!len || !mpp || !(mp0 = *mpp))
3900 return (NULL);
3901 /* If we aren't going to consume the first mblk, we need a dup. */
3902 if (mp0->b_wptr - mp0->b_rptr > len) {
3903 mp1 = dupb(mp0);
3904 if (mp1) {
3905 /* Partition the data between the two mblks. */
3906 mp1->b_wptr = mp1->b_rptr + len;
3907 mp0->b_rptr = mp1->b_wptr;
3908 /*
3909 * after adjustments if mblk not consumed is now
3910 * unaligned, try to align it. If this fails free
3911 * all messages and let upper layer recover.
3912 */
3913 if (!OK_32PTR(mp0->b_rptr)) {
3914 if (!pullupmsg(mp0, -1)) {
3915 freemsg(mp0);
3916 freemsg(mp1);
3917 *mpp = NULL;
3918 return (NULL);
3919 }
3920 }
3921 }
3922 return (mp1);
3923 }
3924 /* Eat through as many mblks as we need to get len bytes. */
3925 len -= mp0->b_wptr - mp0->b_rptr;
3926 for (mp2 = mp1 = mp0; (mp2 = mp2->b_cont) != 0 && len; mp1 = mp2) {
3927 if (mp2->b_wptr - mp2->b_rptr > len) {
3928 /*
3929 * We won't consume the entire last mblk. Like
3930 * above, dup and partition it.
3931 */
3932 mp1->b_cont = dupb(mp2);
3933 mp1 = mp1->b_cont;
3934 if (!mp1) {
3935 /*
3936 * Trouble. Rather than go to a lot of
3937 * trouble to clean up, we free the messages.
3938 * This won't be any worse than losing it on
3939 * the wire.
3940 */
3941 freemsg(mp0);
3942 freemsg(mp2);
3943 *mpp = NULL;
3944 return (NULL);
3945 }
3946 mp1->b_wptr = mp1->b_rptr + len;
3947 mp2->b_rptr = mp1->b_wptr;
3948 /*
3949 * after adjustments if mblk not consumed is now
3950 * unaligned, try to align it. If this fails free
3951 * all messages and let upper layer recover.
3952 */
3953 if (!OK_32PTR(mp2->b_rptr)) {
3954 if (!pullupmsg(mp2, -1)) {
3955 freemsg(mp0);
3956 freemsg(mp2);
3957 *mpp = NULL;
3958 return (NULL);
3959 }
3960 }
3961 *mpp = mp2;
3962 return (mp0);
3963 }
3964 /* Decrement len by the amount we just got. */
3965 len -= mp2->b_wptr - mp2->b_rptr;
3966 }
3967 /*
3968 * len should be reduced to zero now. If not our caller has
3969 * screwed up.
3970 */
3971 if (len) {
3972 /* Shouldn't happen! */
3973 freemsg(mp0);
3974 *mpp = NULL;
3975 return (NULL);
3976 }
3977 /*
3978 * We consumed up to exactly the end of an mblk. Detach the part
3979 * we are returning from the rest of the chain.
3980 */
3981 mp1->b_cont = NULL;
3982 *mpp = mp2;
3983 return (mp0);
3984 }
3985
3986 /* The ill stream is being unplumbed. Called from ip_close */
3987 int
3988 ip_modclose(ill_t *ill)
3989 {
3990 boolean_t success;
3991 ipsq_t *ipsq;
3992 ipif_t *ipif;
3993 queue_t *q = ill->ill_rq;
3994 ip_stack_t *ipst = ill->ill_ipst;
3995 int i;
3996 arl_ill_common_t *ai = ill->ill_common;
3997
3998 /*
3999 * The punlink prior to this may have initiated a capability
4000 * negotiation. But ipsq_enter will block until that finishes or
4001 * times out.
4002 */
4003 success = ipsq_enter(ill, B_FALSE, NEW_OP);
4004
4005 /*
4006 * Open/close/push/pop is guaranteed to be single threaded
4007 * per stream by STREAMS. FS guarantees that all references
4008 * from top are gone before close is called. So there can't
4009 * be another close thread that has set CONDEMNED on this ill.
4010 * and cause ipsq_enter to return failure.
4011 */
4012 ASSERT(success);
4013 ipsq = ill->ill_phyint->phyint_ipsq;
4014
4015 /*
4016 * Mark it condemned. No new reference will be made to this ill.
4017 * Lookup functions will return an error. Threads that try to
4018 * increment the refcnt must check for ILL_CAN_LOOKUP. This ensures
4019 * that the refcnt will drop down to zero.
4020 */
4021 mutex_enter(&ill->ill_lock);
4022 ill->ill_state_flags |= ILL_CONDEMNED;
4023 for (ipif = ill->ill_ipif; ipif != NULL;
4024 ipif = ipif->ipif_next) {
4025 ipif->ipif_state_flags |= IPIF_CONDEMNED;
4026 }
4027 /*
4028 * Wake up anybody waiting to enter the ipsq. ipsq_enter
4029 * returns error if ILL_CONDEMNED is set
4030 */
4031 cv_broadcast(&ill->ill_cv);
4032 mutex_exit(&ill->ill_lock);
4033
4034 /*
4035 * Send all the deferred DLPI messages downstream which came in
4036 * during the small window right before ipsq_enter(). We do this
4037 * without waiting for the ACKs because all the ACKs for M_PROTO
4038 * messages are ignored in ip_rput() when ILL_CONDEMNED is set.
4039 */
4040 ill_dlpi_send_deferred(ill);
4041
4042 /*
4043 * Shut down fragmentation reassembly.
4044 * ill_frag_timer won't start a timer again.
4045 * Now cancel any existing timer
4046 */
4047 (void) untimeout(ill->ill_frag_timer_id);
4048 (void) ill_frag_timeout(ill, 0);
4049
4050 /*
4051 * Call ill_delete to bring down the ipifs, ilms and ill on
4052 * this ill. Then wait for the refcnts to drop to zero.
4053 * ill_is_freeable checks whether the ill is really quiescent.
4054 * Then make sure that threads that are waiting to enter the
4055 * ipsq have seen the error returned by ipsq_enter and have
4056 * gone away. Then we call ill_delete_tail which does the
4057 * DL_UNBIND_REQ with the driver and then qprocsoff.
4058 */
4059 ill_delete(ill);
4060 mutex_enter(&ill->ill_lock);
4061 while (!ill_is_freeable(ill))
4062 cv_wait(&ill->ill_cv, &ill->ill_lock);
4063
4064 while (ill->ill_waiters)
4065 cv_wait(&ill->ill_cv, &ill->ill_lock);
4066
4067 mutex_exit(&ill->ill_lock);
4068
4069 /*
4070 * ill_delete_tail drops reference on ill_ipst, but we need to keep
4071 * it held until the end of the function since the cleanup
4072 * below needs to be able to use the ip_stack_t.
4073 */
4074 netstack_hold(ipst->ips_netstack);
4075
4076 /* qprocsoff is done via ill_delete_tail */
4077 ill_delete_tail(ill);
4078 /*
4079 * synchronously wait for arp stream to unbind. After this, we
4080 * cannot get any data packets up from the driver.
4081 */
4082 arp_unbind_complete(ill);
4083 ASSERT(ill->ill_ipst == NULL);
4084
4085 /*
4086 * Walk through all conns and qenable those that have queued data.
4087 * Close synchronization needs this to
4088 * be done to ensure that all upper layers blocked
4089 * due to flow control to the closing device
4090 * get unblocked.
4091 */
4092 ip1dbg(("ip_wsrv: walking\n"));
4093 for (i = 0; i < TX_FANOUT_SIZE; i++) {
4094 conn_walk_drain(ipst, &ipst->ips_idl_tx_list[i]);
4095 }
4096
4097 /*
4098 * ai can be null if this is an IPv6 ill, or if the IPv4
4099 * stream is being torn down before ARP was plumbed (e.g.,
4100 * /sbin/ifconfig plumbing a stream twice, and encountering
4101 * an error
4102 */
4103 if (ai != NULL) {
4104 ASSERT(!ill->ill_isv6);
4105 mutex_enter(&ai->ai_lock);
4106 ai->ai_ill = NULL;
4107 if (ai->ai_arl == NULL) {
4108 mutex_destroy(&ai->ai_lock);
4109 kmem_free(ai, sizeof (*ai));
4110 } else {
4111 cv_signal(&ai->ai_ill_unplumb_done);
4112 mutex_exit(&ai->ai_lock);
4113 }
4114 }
4115
4116 mutex_enter(&ipst->ips_ip_mi_lock);
4117 mi_close_unlink(&ipst->ips_ip_g_head, (IDP)ill);
4118 mutex_exit(&ipst->ips_ip_mi_lock);
4119
4120 /*
4121 * credp could be null if the open didn't succeed and ip_modopen
4122 * itself calls ip_close.
4123 */
4124 if (ill->ill_credp != NULL)
4125 crfree(ill->ill_credp);
4126
4127 mutex_destroy(&ill->ill_saved_ire_lock);
4128 mutex_destroy(&ill->ill_lock);
4129 rw_destroy(&ill->ill_mcast_lock);
4130 mutex_destroy(&ill->ill_mcast_serializer);
4131 list_destroy(&ill->ill_nce);
4132
4133 /*
4134 * Now we are done with the module close pieces that
4135 * need the netstack_t.
4136 */
4137 netstack_rele(ipst->ips_netstack);
4138
4139 mi_close_free((IDP)ill);
4140 q->q_ptr = WR(q)->q_ptr = NULL;
4141
4142 ipsq_exit(ipsq);
4143
4144 return (0);
4145 }
4146
4147 /*
4148 * This is called as part of close() for IP, UDP, ICMP, and RTS
4149 * in order to quiesce the conn.
4150 */
4151 void
4152 ip_quiesce_conn(conn_t *connp)
4153 {
4154 boolean_t drain_cleanup_reqd = B_FALSE;
4155 boolean_t conn_ioctl_cleanup_reqd = B_FALSE;
4156 boolean_t ilg_cleanup_reqd = B_FALSE;
4157 ip_stack_t *ipst;
4158
4159 ASSERT(!IPCL_IS_TCP(connp));
4160 ipst = connp->conn_netstack->netstack_ip;
4161
4162 /*
4163 * Mark the conn as closing, and this conn must not be
4164 * inserted in future into any list. Eg. conn_drain_insert(),
4165 * won't insert this conn into the conn_drain_list.
4166 *
4167 * conn_idl, and conn_ilg cannot get set henceforth.
4168 */
4169 mutex_enter(&connp->conn_lock);
4170 ASSERT(!(connp->conn_state_flags & CONN_QUIESCED));
4171 connp->conn_state_flags |= CONN_CLOSING;
4172 if (connp->conn_idl != NULL)
4173 drain_cleanup_reqd = B_TRUE;
4174 if (connp->conn_oper_pending_ill != NULL)
4175 conn_ioctl_cleanup_reqd = B_TRUE;
4176 if (connp->conn_dhcpinit_ill != NULL) {
4177 ASSERT(connp->conn_dhcpinit_ill->ill_dhcpinit != 0);
4178 atomic_dec_32(&connp->conn_dhcpinit_ill->ill_dhcpinit);
4179 ill_set_inputfn(connp->conn_dhcpinit_ill);
4180 connp->conn_dhcpinit_ill = NULL;
4181 }
4182 if (connp->conn_ilg != NULL)
4183 ilg_cleanup_reqd = B_TRUE;
4184 mutex_exit(&connp->conn_lock);
4185
4186 if (conn_ioctl_cleanup_reqd)
4187 conn_ioctl_cleanup(connp);
4188
4189 if (is_system_labeled() && connp->conn_anon_port) {
4190 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
4191 connp->conn_mlp_type, connp->conn_proto,
4192 ntohs(connp->conn_lport), B_FALSE);
4193 connp->conn_anon_port = 0;
4194 }
4195 connp->conn_mlp_type = mlptSingle;
4196
4197 /*
4198 * Remove this conn from any fanout list it is on.
4199 * and then wait for any threads currently operating
4200 * on this endpoint to finish
4201 */
4202 ipcl_hash_remove(connp);
4203
4204 /*
4205 * Remove this conn from the drain list, and do any other cleanup that
4206 * may be required. (TCP conns are never flow controlled, and
4207 * conn_idl will be NULL.)
4208 */
4209 if (drain_cleanup_reqd && connp->conn_idl != NULL) {
4210 idl_t *idl = connp->conn_idl;
4211
4212 mutex_enter(&idl->idl_lock);
4213 conn_drain(connp, B_TRUE);
4214 mutex_exit(&idl->idl_lock);
4215 }
4216
4217 if (connp == ipst->ips_ip_g_mrouter)
4218 (void) ip_mrouter_done(ipst);
4219
4220 if (ilg_cleanup_reqd)
4221 ilg_delete_all(connp);
4222
4223 /*
4224 * Now conn refcnt can increase only thru CONN_INC_REF_LOCKED.
4225 * callers from write side can't be there now because close
4226 * is in progress. The only other caller is ipcl_walk
4227 * which checks for the condemned flag.
4228 */
4229 mutex_enter(&connp->conn_lock);
4230 connp->conn_state_flags |= CONN_CONDEMNED;
4231 while (connp->conn_ref != 1)
4232 cv_wait(&connp->conn_cv, &connp->conn_lock);
4233 connp->conn_state_flags |= CONN_QUIESCED;
4234 mutex_exit(&connp->conn_lock);
4235 }
4236
4237 /* ARGSUSED */
4238 int
4239 ip_close(queue_t *q, int flags)
4240 {
4241 conn_t *connp;
4242
4243 /*
4244 * Call the appropriate delete routine depending on whether this is
4245 * a module or device.
4246 */
4247 if (WR(q)->q_next != NULL) {
4248 /* This is a module close */
4249 return (ip_modclose((ill_t *)q->q_ptr));
4250 }
4251
4252 connp = q->q_ptr;
4253 ip_quiesce_conn(connp);
4254
4255 qprocsoff(q);
4256
4257 /*
4258 * Now we are truly single threaded on this stream, and can
4259 * delete the things hanging off the connp, and finally the connp.
4260 * We removed this connp from the fanout list, it cannot be
4261 * accessed thru the fanouts, and we already waited for the
4262 * conn_ref to drop to 0. We are already in close, so
4263 * there cannot be any other thread from the top. qprocsoff
4264 * has completed, and service has completed or won't run in
4265 * future.
4266 */
4267 ASSERT(connp->conn_ref == 1);
4268
4269 inet_minor_free(connp->conn_minor_arena, connp->conn_dev);
4270
4271 connp->conn_ref--;
4272 ipcl_conn_destroy(connp);
4273
4274 q->q_ptr = WR(q)->q_ptr = NULL;
4275 return (0);
4276 }
4277
4278 /*
4279 * Wapper around putnext() so that ip_rts_request can merely use
4280 * conn_recv.
4281 */
4282 /*ARGSUSED2*/
4283 static void
4284 ip_conn_input(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira)
4285 {
4286 conn_t *connp = (conn_t *)arg1;
4287
4288 putnext(connp->conn_rq, mp);
4289 }
4290
4291 /* Dummy in case ICMP error delivery is attempted to a /dev/ip instance */
4292 /* ARGSUSED */
4293 static void
4294 ip_conn_input_icmp(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira)
4295 {
4296 freemsg(mp);
4297 }
4298
4299 /*
4300 * Called when the module is about to be unloaded
4301 */
4302 void
4303 ip_ddi_destroy(void)
4304 {
4305 /* This needs to be called before destroying any transports. */
4306 mutex_enter(&cpu_lock);
4307 unregister_cpu_setup_func(ip_tp_cpu_update, NULL);
4308 mutex_exit(&cpu_lock);
4309
4310 tnet_fini();
4311
4312 icmp_ddi_g_destroy();
4313 rts_ddi_g_destroy();
4314 udp_ddi_g_destroy();
4315 sctp_ddi_g_destroy();
4316 tcp_ddi_g_destroy();
4317 ilb_ddi_g_destroy();
4318 dce_g_destroy();
4319 ipsec_policy_g_destroy();
4320 ipcl_g_destroy();
4321 ip_net_g_destroy();
4322 ip_ire_g_fini();
4323 inet_minor_destroy(ip_minor_arena_sa);
4324 #if defined(_LP64)
4325 inet_minor_destroy(ip_minor_arena_la);
4326 #endif
4327
4328 #ifdef DEBUG
4329 list_destroy(&ip_thread_list);
4330 rw_destroy(&ip_thread_rwlock);
4331 tsd_destroy(&ip_thread_data);
4332 #endif
4333
4334 netstack_unregister(NS_IP);
4335 }
4336
4337 /*
4338 * First step in cleanup.
4339 */
4340 /* ARGSUSED */
4341 static void
4342 ip_stack_shutdown(netstackid_t stackid, void *arg)
4343 {
4344 ip_stack_t *ipst = (ip_stack_t *)arg;
4345 kt_did_t ktid;
4346
4347 #ifdef NS_DEBUG
4348 printf("ip_stack_shutdown(%p, stack %d)\n", (void *)ipst, stackid);
4349 #endif
4350
4351 /*
4352 * Perform cleanup for special interfaces (loopback and IPMP).
4353 */
4354 ip_interface_cleanup(ipst);
4355
4356 /*
4357 * The *_hook_shutdown()s start the process of notifying any
4358 * consumers that things are going away.... nothing is destroyed.
4359 */
4360 ipv4_hook_shutdown(ipst);
4361 ipv6_hook_shutdown(ipst);
4362 arp_hook_shutdown(ipst);
4363
4364 mutex_enter(&ipst->ips_capab_taskq_lock);
4365 ktid = ipst->ips_capab_taskq_thread->t_did;
4366 ipst->ips_capab_taskq_quit = B_TRUE;
4367 cv_signal(&ipst->ips_capab_taskq_cv);
4368 mutex_exit(&ipst->ips_capab_taskq_lock);
4369
4370 /*
4371 * In rare occurrences, particularly on virtual hardware where CPUs can
4372 * be de-scheduled, the thread that we just signaled will not run until
4373 * after we have gotten through parts of ip_stack_fini. If that happens
4374 * then we'll try to grab the ips_capab_taskq_lock as part of returning
4375 * from cv_wait which no longer exists.
4376 */
4377 thread_join(ktid);
4378 }
4379
4380 /*
4381 * Free the IP stack instance.
4382 */
4383 static void
4384 ip_stack_fini(netstackid_t stackid, void *arg)
4385 {
4386 ip_stack_t *ipst = (ip_stack_t *)arg;
4387 int ret;
4388
4389 #ifdef NS_DEBUG
4390 printf("ip_stack_fini(%p, stack %d)\n", (void *)ipst, stackid);
4391 #endif
4392 /*
4393 * At this point, all of the notifications that the events and
4394 * protocols are going away have been run, meaning that we can
4395 * now set about starting to clean things up.
4396 */
4397 ipobs_fini(ipst);
4398 ipv4_hook_destroy(ipst);
4399 ipv6_hook_destroy(ipst);
4400 arp_hook_destroy(ipst);
4401 ip_net_destroy(ipst);
4402
4403 ipmp_destroy(ipst);
4404
4405 ip_kstat_fini(stackid, ipst->ips_ip_mibkp);
4406 ipst->ips_ip_mibkp = NULL;
4407 icmp_kstat_fini(stackid, ipst->ips_icmp_mibkp);
4408 ipst->ips_icmp_mibkp = NULL;
4409 ip_kstat2_fini(stackid, ipst->ips_ip_kstat);
4410 ipst->ips_ip_kstat = NULL;
4411 bzero(&ipst->ips_ip_statistics, sizeof (ipst->ips_ip_statistics));
4412 ip6_kstat_fini(stackid, ipst->ips_ip6_kstat);
4413 ipst->ips_ip6_kstat = NULL;
4414 bzero(&ipst->ips_ip6_statistics, sizeof (ipst->ips_ip6_statistics));
4415
4416 kmem_free(ipst->ips_propinfo_tbl,
4417 ip_propinfo_count * sizeof (mod_prop_info_t));
4418 ipst->ips_propinfo_tbl = NULL;
4419
4420 dce_stack_destroy(ipst);
4421 ip_mrouter_stack_destroy(ipst);
4422
4423 /*
4424 * Quiesce all of our timers. Note we set the quiesce flags before we
4425 * call untimeout. The slowtimers may actually kick off another instance
4426 * of the non-slow timers.
4427 */
4428 mutex_enter(&ipst->ips_igmp_timer_lock);
4429 ipst->ips_igmp_timer_quiesce = B_TRUE;
4430 mutex_exit(&ipst->ips_igmp_timer_lock);
4431
4432 mutex_enter(&ipst->ips_mld_timer_lock);
4433 ipst->ips_mld_timer_quiesce = B_TRUE;
4434 mutex_exit(&ipst->ips_mld_timer_lock);
4435
4436 mutex_enter(&ipst->ips_igmp_slowtimeout_lock);
4437 ipst->ips_igmp_slowtimeout_quiesce = B_TRUE;
4438 mutex_exit(&ipst->ips_igmp_slowtimeout_lock);
4439
4440 mutex_enter(&ipst->ips_mld_slowtimeout_lock);
4441 ipst->ips_mld_slowtimeout_quiesce = B_TRUE;
4442 mutex_exit(&ipst->ips_mld_slowtimeout_lock);
4443
4444 ret = untimeout(ipst->ips_igmp_timeout_id);
4445 if (ret == -1) {
4446 ASSERT(ipst->ips_igmp_timeout_id == 0);
4447 } else {
4448 ASSERT(ipst->ips_igmp_timeout_id != 0);
4449 ipst->ips_igmp_timeout_id = 0;
4450 }
4451 ret = untimeout(ipst->ips_igmp_slowtimeout_id);
4452 if (ret == -1) {
4453 ASSERT(ipst->ips_igmp_slowtimeout_id == 0);
4454 } else {
4455 ASSERT(ipst->ips_igmp_slowtimeout_id != 0);
4456 ipst->ips_igmp_slowtimeout_id = 0;
4457 }
4458 ret = untimeout(ipst->ips_mld_timeout_id);
4459 if (ret == -1) {
4460 ASSERT(ipst->ips_mld_timeout_id == 0);
4461 } else {
4462 ASSERT(ipst->ips_mld_timeout_id != 0);
4463 ipst->ips_mld_timeout_id = 0;
4464 }
4465 ret = untimeout(ipst->ips_mld_slowtimeout_id);
4466 if (ret == -1) {
4467 ASSERT(ipst->ips_mld_slowtimeout_id == 0);
4468 } else {
4469 ASSERT(ipst->ips_mld_slowtimeout_id != 0);
4470 ipst->ips_mld_slowtimeout_id = 0;
4471 }
4472
4473 ip_ire_fini(ipst);
4474 ip6_asp_free(ipst);
4475 conn_drain_fini(ipst);
4476 ipcl_destroy(ipst);
4477
4478 mutex_destroy(&ipst->ips_ndp4->ndp_g_lock);
4479 mutex_destroy(&ipst->ips_ndp6->ndp_g_lock);
4480 kmem_free(ipst->ips_ndp4, sizeof (ndp_g_t));
4481 ipst->ips_ndp4 = NULL;
4482 kmem_free(ipst->ips_ndp6, sizeof (ndp_g_t));
4483 ipst->ips_ndp6 = NULL;
4484
4485 if (ipst->ips_loopback_ksp != NULL) {
4486 kstat_delete_netstack(ipst->ips_loopback_ksp, stackid);
4487 ipst->ips_loopback_ksp = NULL;
4488 }
4489
4490 mutex_destroy(&ipst->ips_capab_taskq_lock);
4491 cv_destroy(&ipst->ips_capab_taskq_cv);
4492
4493 rw_destroy(&ipst->ips_srcid_lock);
4494
4495 mutex_destroy(&ipst->ips_ip_mi_lock);
4496 rw_destroy(&ipst->ips_ill_g_usesrc_lock);
4497
4498 mutex_destroy(&ipst->ips_igmp_timer_lock);
4499 mutex_destroy(&ipst->ips_mld_timer_lock);
4500 mutex_destroy(&ipst->ips_igmp_slowtimeout_lock);
4501 mutex_destroy(&ipst->ips_mld_slowtimeout_lock);
4502 mutex_destroy(&ipst->ips_ip_addr_avail_lock);
4503 rw_destroy(&ipst->ips_ill_g_lock);
4504
4505 kmem_free(ipst->ips_phyint_g_list, sizeof (phyint_list_t));
4506 ipst->ips_phyint_g_list = NULL;
4507 kmem_free(ipst->ips_ill_g_heads, sizeof (ill_g_head_t) * MAX_G_HEADS);
4508 ipst->ips_ill_g_heads = NULL;
4509
4510 ldi_ident_release(ipst->ips_ldi_ident);
4511 kmem_free(ipst, sizeof (*ipst));
4512 }
4513
4514 /*
4515 * This function is called from the TSD destructor, and is used to debug
4516 * reference count issues in IP. See block comment in <inet/ip_if.h> for
4517 * details.
4518 */
4519 static void
4520 ip_thread_exit(void *phash)
4521 {
4522 th_hash_t *thh = phash;
4523
4524 rw_enter(&ip_thread_rwlock, RW_WRITER);
4525 list_remove(&ip_thread_list, thh);
4526 rw_exit(&ip_thread_rwlock);
4527 mod_hash_destroy_hash(thh->thh_hash);
4528 kmem_free(thh, sizeof (*thh));
4529 }
4530
4531 /*
4532 * Called when the IP kernel module is loaded into the kernel
4533 */
4534 void
4535 ip_ddi_init(void)
4536 {
4537 ip_squeue_flag = ip_squeue_switch(ip_squeue_enter);
4538
4539 /*
4540 * For IP and TCP the minor numbers should start from 2 since we have 4
4541 * initial devices: ip, ip6, tcp, tcp6.
4542 */
4543 /*
4544 * If this is a 64-bit kernel, then create two separate arenas -
4545 * one for TLIs in the range of INET_MIN_DEV+2 through 2^^18-1, and the
4546 * other for socket apps in the range 2^^18 through 2^^32-1.
4547 */
4548 ip_minor_arena_la = NULL;
4549 ip_minor_arena_sa = NULL;
4550 #if defined(_LP64)
4551 if ((ip_minor_arena_sa = inet_minor_create("ip_minor_arena_sa",
4552 INET_MIN_DEV + 2, MAXMIN32, KM_SLEEP)) == NULL) {
4553 cmn_err(CE_PANIC,
4554 "ip_ddi_init: ip_minor_arena_sa creation failed\n");
4555 }
4556 if ((ip_minor_arena_la = inet_minor_create("ip_minor_arena_la",
4557 MAXMIN32 + 1, MAXMIN64, KM_SLEEP)) == NULL) {
4558 cmn_err(CE_PANIC,
4559 "ip_ddi_init: ip_minor_arena_la creation failed\n");
4560 }
4561 #else
4562 if ((ip_minor_arena_sa = inet_minor_create("ip_minor_arena_sa",
4563 INET_MIN_DEV + 2, MAXMIN, KM_SLEEP)) == NULL) {
4564 cmn_err(CE_PANIC,
4565 "ip_ddi_init: ip_minor_arena_sa creation failed\n");
4566 }
4567 #endif
4568 ip_poll_normal_ticks = MSEC_TO_TICK_ROUNDUP(ip_poll_normal_ms);
4569
4570 ipcl_g_init();
4571 ip_ire_g_init();
4572 ip_net_g_init();
4573
4574 #ifdef DEBUG
4575 tsd_create(&ip_thread_data, ip_thread_exit);
4576 rw_init(&ip_thread_rwlock, NULL, RW_DEFAULT, NULL);
4577 list_create(&ip_thread_list, sizeof (th_hash_t),
4578 offsetof(th_hash_t, thh_link));
4579 #endif
4580 ipsec_policy_g_init();
4581 tcp_ddi_g_init();
4582 sctp_ddi_g_init();
4583 dce_g_init();
4584
4585 /*
4586 * We want to be informed each time a stack is created or
4587 * destroyed in the kernel, so we can maintain the
4588 * set of udp_stack_t's.
4589 */
4590 netstack_register(NS_IP, ip_stack_init, ip_stack_shutdown,
4591 ip_stack_fini);
4592
4593 tnet_init();
4594
4595 udp_ddi_g_init();
4596 rts_ddi_g_init();
4597 icmp_ddi_g_init();
4598 ilb_ddi_g_init();
4599
4600 /* This needs to be called after all transports are initialized. */
4601 mutex_enter(&cpu_lock);
4602 register_cpu_setup_func(ip_tp_cpu_update, NULL);
4603 mutex_exit(&cpu_lock);
4604 }
4605
4606 /*
4607 * Initialize the IP stack instance.
4608 */
4609 static void *
4610 ip_stack_init(netstackid_t stackid, netstack_t *ns)
4611 {
4612 ip_stack_t *ipst;
4613 size_t arrsz;
4614 major_t major;
4615
4616 #ifdef NS_DEBUG
4617 printf("ip_stack_init(stack %d)\n", stackid);
4618 #endif
4619
4620 ipst = (ip_stack_t *)kmem_zalloc(sizeof (*ipst), KM_SLEEP);
4621 ipst->ips_netstack = ns;
4622
4623 ipst->ips_ill_g_heads = kmem_zalloc(sizeof (ill_g_head_t) * MAX_G_HEADS,
4624 KM_SLEEP);
4625 ipst->ips_phyint_g_list = kmem_zalloc(sizeof (phyint_list_t),
4626 KM_SLEEP);
4627 ipst->ips_ndp4 = kmem_zalloc(sizeof (ndp_g_t), KM_SLEEP);
4628 ipst->ips_ndp6 = kmem_zalloc(sizeof (ndp_g_t), KM_SLEEP);
4629 mutex_init(&ipst->ips_ndp4->ndp_g_lock, NULL, MUTEX_DEFAULT, NULL);
4630 mutex_init(&ipst->ips_ndp6->ndp_g_lock, NULL, MUTEX_DEFAULT, NULL);
4631
4632 mutex_init(&ipst->ips_igmp_timer_lock, NULL, MUTEX_DEFAULT, NULL);
4633 ipst->ips_igmp_deferred_next = INFINITY;
4634 mutex_init(&ipst->ips_mld_timer_lock, NULL, MUTEX_DEFAULT, NULL);
4635 ipst->ips_mld_deferred_next = INFINITY;
4636 mutex_init(&ipst->ips_igmp_slowtimeout_lock, NULL, MUTEX_DEFAULT, NULL);
4637 mutex_init(&ipst->ips_mld_slowtimeout_lock, NULL, MUTEX_DEFAULT, NULL);
4638 mutex_init(&ipst->ips_ip_mi_lock, NULL, MUTEX_DEFAULT, NULL);
4639 mutex_init(&ipst->ips_ip_addr_avail_lock, NULL, MUTEX_DEFAULT, NULL);
4640 rw_init(&ipst->ips_ill_g_lock, NULL, RW_DEFAULT, NULL);
4641 rw_init(&ipst->ips_ill_g_usesrc_lock, NULL, RW_DEFAULT, NULL);
4642
4643 ipcl_init(ipst);
4644 ip_ire_init(ipst);
4645 ip6_asp_init(ipst);
4646 ipif_init(ipst);
4647 conn_drain_init(ipst);
4648 ip_mrouter_stack_init(ipst);
4649 dce_stack_init(ipst);
4650
4651 ipst->ips_ip_multirt_log_interval = 1000;
4652
4653 ipst->ips_ill_index = 1;
4654
4655 ipst->ips_saved_ip_forwarding = -1;
4656 ipst->ips_reg_vif_num = ALL_VIFS; /* Index to Register vif */
4657
4658 arrsz = ip_propinfo_count * sizeof (mod_prop_info_t);
4659 ipst->ips_propinfo_tbl = (mod_prop_info_t *)kmem_alloc(arrsz, KM_SLEEP);
4660 bcopy(ip_propinfo_tbl, ipst->ips_propinfo_tbl, arrsz);
4661
4662 ipst->ips_ip_mibkp = ip_kstat_init(stackid, ipst);
4663 ipst->ips_icmp_mibkp = icmp_kstat_init(stackid);
4664 ipst->ips_ip_kstat = ip_kstat2_init(stackid, &ipst->ips_ip_statistics);
4665 ipst->ips_ip6_kstat =
4666 ip6_kstat_init(stackid, &ipst->ips_ip6_statistics);
4667
4668 ipst->ips_ip_src_id = 1;
4669 rw_init(&ipst->ips_srcid_lock, NULL, RW_DEFAULT, NULL);
4670
4671 ipst->ips_src_generation = SRC_GENERATION_INITIAL;
4672
4673 ip_net_init(ipst, ns);
4674 ipv4_hook_init(ipst);
4675 ipv6_hook_init(ipst);
4676 arp_hook_init(ipst);
4677 ipmp_init(ipst);
4678 ipobs_init(ipst);
4679
4680 /*
4681 * Create the taskq dispatcher thread and initialize related stuff.
4682 */
4683 mutex_init(&ipst->ips_capab_taskq_lock, NULL, MUTEX_DEFAULT, NULL);
4684 cv_init(&ipst->ips_capab_taskq_cv, NULL, CV_DEFAULT, NULL);
4685 ipst->ips_capab_taskq_thread = thread_create(NULL, 0,
4686 ill_taskq_dispatch, ipst, 0, &p0, TS_RUN, minclsyspri);
4687
4688 major = mod_name_to_major(INET_NAME);
4689 (void) ldi_ident_from_major(major, &ipst->ips_ldi_ident);
4690 return (ipst);
4691 }
4692
4693 /*
4694 * Allocate and initialize a DLPI template of the specified length. (May be
4695 * called as writer.)
4696 */
4697 mblk_t *
4698 ip_dlpi_alloc(size_t len, t_uscalar_t prim)
4699 {
4700 mblk_t *mp;
4701
4702 mp = allocb(len, BPRI_MED);
4703 if (!mp)
4704 return (NULL);
4705
4706 /*
4707 * DLPIv2 says that DL_INFO_REQ and DL_TOKEN_REQ (the latter
4708 * of which we don't seem to use) are sent with M_PCPROTO, and
4709 * that other DLPI are M_PROTO.
4710 */
4711 if (prim == DL_INFO_REQ) {
4712 mp->b_datap->db_type = M_PCPROTO;
4713 } else {
4714 mp->b_datap->db_type = M_PROTO;
4715 }
4716
4717 mp->b_wptr = mp->b_rptr + len;
4718 bzero(mp->b_rptr, len);
4719 ((dl_unitdata_req_t *)mp->b_rptr)->dl_primitive = prim;
4720 return (mp);
4721 }
4722
4723 /*
4724 * Allocate and initialize a DLPI notification. (May be called as writer.)
4725 */
4726 mblk_t *
4727 ip_dlnotify_alloc(uint_t notification, uint_t data)
4728 {
4729 dl_notify_ind_t *notifyp;
4730 mblk_t *mp;
4731
4732 if ((mp = ip_dlpi_alloc(DL_NOTIFY_IND_SIZE, DL_NOTIFY_IND)) == NULL)
4733 return (NULL);
4734
4735 notifyp = (dl_notify_ind_t *)mp->b_rptr;
4736 notifyp->dl_notification = notification;
4737 notifyp->dl_data = data;
4738 return (mp);
4739 }
4740
4741 mblk_t *
4742 ip_dlnotify_alloc2(uint_t notification, uint_t data1, uint_t data2)
4743 {
4744 dl_notify_ind_t *notifyp;
4745 mblk_t *mp;
4746
4747 if ((mp = ip_dlpi_alloc(DL_NOTIFY_IND_SIZE, DL_NOTIFY_IND)) == NULL)
4748 return (NULL);
4749
4750 notifyp = (dl_notify_ind_t *)mp->b_rptr;
4751 notifyp->dl_notification = notification;
4752 notifyp->dl_data1 = data1;
4753 notifyp->dl_data2 = data2;
4754 return (mp);
4755 }
4756
4757 /*
4758 * Debug formatting routine. Returns a character string representation of the
4759 * addr in buf, of the form xxx.xxx.xxx.xxx. This routine takes the address
4760 * in the form of a ipaddr_t and calls ip_dot_saddr with a pointer.
4761 *
4762 * Once the ndd table-printing interfaces are removed, this can be changed to
4763 * standard dotted-decimal form.
4764 */
4765 char *
4766 ip_dot_addr(ipaddr_t addr, char *buf)
4767 {
4768 uint8_t *ap = (uint8_t *)&addr;
4769
4770 (void) mi_sprintf(buf, "%03d.%03d.%03d.%03d",
4771 ap[0] & 0xFF, ap[1] & 0xFF, ap[2] & 0xFF, ap[3] & 0xFF);
4772 return (buf);
4773 }
4774
4775 /*
4776 * Write the given MAC address as a printable string in the usual colon-
4777 * separated format.
4778 */
4779 const char *
4780 mac_colon_addr(const uint8_t *addr, size_t alen, char *buf, size_t buflen)
4781 {
4782 char *bp;
4783
4784 if (alen == 0 || buflen < 4)
4785 return ("?");
4786 bp = buf;
4787 for (;;) {
4788 /*
4789 * If there are more MAC address bytes available, but we won't
4790 * have any room to print them, then add "..." to the string
4791 * instead. See below for the 'magic number' explanation.
4792 */
4793 if ((alen == 2 && buflen < 6) || (alen > 2 && buflen < 7)) {
4794 (void) strcpy(bp, "...");
4795 break;
4796 }
4797 (void) sprintf(bp, "%02x", *addr++);
4798 bp += 2;
4799 if (--alen == 0)
4800 break;
4801 *bp++ = ':';
4802 buflen -= 3;
4803 /*
4804 * At this point, based on the first 'if' statement above,
4805 * either alen == 1 and buflen >= 3, or alen > 1 and
4806 * buflen >= 4. The first case leaves room for the final "xx"
4807 * number and trailing NUL byte. The second leaves room for at
4808 * least "...". Thus the apparently 'magic' numbers chosen for
4809 * that statement.
4810 */
4811 }
4812 return (buf);
4813 }
4814
4815 /*
4816 * Called when it is conceptually a ULP that would sent the packet
4817 * e.g., port unreachable and protocol unreachable. Check that the packet
4818 * would have passed the IPsec global policy before sending the error.
4819 *
4820 * Send an ICMP error after patching up the packet appropriately.
4821 * Uses ip_drop_input and bumps the appropriate MIB.
4822 */
4823 void
4824 ip_fanout_send_icmp_v4(mblk_t *mp, uint_t icmp_type, uint_t icmp_code,
4825 ip_recv_attr_t *ira)
4826 {
4827 ipha_t *ipha;
4828 boolean_t secure;
4829 ill_t *ill = ira->ira_ill;
4830 ip_stack_t *ipst = ill->ill_ipst;
4831 netstack_t *ns = ipst->ips_netstack;
4832 ipsec_stack_t *ipss = ns->netstack_ipsec;
4833
4834 secure = ira->ira_flags & IRAF_IPSEC_SECURE;
4835
4836 /*
4837 * We are generating an icmp error for some inbound packet.
4838 * Called from all ip_fanout_(udp, tcp, proto) functions.
4839 * Before we generate an error, check with global policy
4840 * to see whether this is allowed to enter the system. As
4841 * there is no "conn", we are checking with global policy.
4842 */
4843 ipha = (ipha_t *)mp->b_rptr;
4844 if (secure || ipss->ipsec_inbound_v4_policy_present) {
4845 mp = ipsec_check_global_policy(mp, NULL, ipha, NULL, ira, ns);
4846 if (mp == NULL)
4847 return;
4848 }
4849
4850 /* We never send errors for protocols that we do implement */
4851 if (ira->ira_protocol == IPPROTO_ICMP ||
4852 ira->ira_protocol == IPPROTO_IGMP) {
4853 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
4854 ip_drop_input("ip_fanout_send_icmp_v4", mp, ill);
4855 freemsg(mp);
4856 return;
4857 }
4858 /*
4859 * Have to correct checksum since
4860 * the packet might have been
4861 * fragmented and the reassembly code in ip_rput
4862 * does not restore the IP checksum.
4863 */
4864 ipha->ipha_hdr_checksum = 0;
4865 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
4866
4867 switch (icmp_type) {
4868 case ICMP_DEST_UNREACHABLE:
4869 switch (icmp_code) {
4870 case ICMP_PROTOCOL_UNREACHABLE:
4871 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInUnknownProtos);
4872 ip_drop_input("ipIfStatsInUnknownProtos", mp, ill);
4873 break;
4874 case ICMP_PORT_UNREACHABLE:
4875 BUMP_MIB(ill->ill_ip_mib, udpIfStatsNoPorts);
4876 ip_drop_input("ipIfStatsNoPorts", mp, ill);
4877 break;
4878 }
4879
4880 icmp_unreachable(mp, icmp_code, ira);
4881 break;
4882 default:
4883 #ifdef DEBUG
4884 panic("ip_fanout_send_icmp_v4: wrong type");
4885 /*NOTREACHED*/
4886 #else
4887 freemsg(mp);
4888 break;
4889 #endif
4890 }
4891 }
4892
4893 /*
4894 * Used to send an ICMP error message when a packet is received for
4895 * a protocol that is not supported. The mblk passed as argument
4896 * is consumed by this function.
4897 */
4898 void
4899 ip_proto_not_sup(mblk_t *mp, ip_recv_attr_t *ira)
4900 {
4901 ipha_t *ipha;
4902
4903 ipha = (ipha_t *)mp->b_rptr;
4904 if (ira->ira_flags & IRAF_IS_IPV4) {
4905 ASSERT(IPH_HDR_VERSION(ipha) == IP_VERSION);
4906 ip_fanout_send_icmp_v4(mp, ICMP_DEST_UNREACHABLE,
4907 ICMP_PROTOCOL_UNREACHABLE, ira);
4908 } else {
4909 ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
4910 ip_fanout_send_icmp_v6(mp, ICMP6_PARAM_PROB,
4911 ICMP6_PARAMPROB_NEXTHEADER, ira);
4912 }
4913 }
4914
4915 /*
4916 * Deliver a rawip packet to the given conn, possibly applying ipsec policy.
4917 * Handles IPv4 and IPv6.
4918 * We are responsible for disposing of mp, such as by freemsg() or putnext()
4919 * Caller is responsible for dropping references to the conn.
4920 */
4921 void
4922 ip_fanout_proto_conn(conn_t *connp, mblk_t *mp, ipha_t *ipha, ip6_t *ip6h,
4923 ip_recv_attr_t *ira)
4924 {
4925 ill_t *ill = ira->ira_ill;
4926 ip_stack_t *ipst = ill->ill_ipst;
4927 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
4928 boolean_t secure;
4929 uint_t protocol = ira->ira_protocol;
4930 iaflags_t iraflags = ira->ira_flags;
4931 queue_t *rq;
4932
4933 secure = iraflags & IRAF_IPSEC_SECURE;
4934
4935 rq = connp->conn_rq;
4936 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld : !canputnext(rq)) {
4937 switch (protocol) {
4938 case IPPROTO_ICMPV6:
4939 BUMP_MIB(ill->ill_icmp6_mib, ipv6IfIcmpInOverflows);
4940 break;
4941 case IPPROTO_ICMP:
4942 BUMP_MIB(&ipst->ips_icmp_mib, icmpInOverflows);
4943 break;
4944 default:
4945 BUMP_MIB(ill->ill_ip_mib, rawipIfStatsInOverflows);
4946 break;
4947 }
4948 freemsg(mp);
4949 return;
4950 }
4951
4952 ASSERT(!(IPCL_IS_IPTUN(connp)));
4953
4954 if (((iraflags & IRAF_IS_IPV4) ?
4955 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
4956 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
4957 secure) {
4958 mp = ipsec_check_inbound_policy(mp, connp, ipha,
4959 ip6h, ira);
4960 if (mp == NULL) {
4961 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
4962 /* Note that mp is NULL */
4963 ip_drop_input("ipIfStatsInDiscards", mp, ill);
4964 return;
4965 }
4966 }
4967
4968 if (iraflags & IRAF_ICMP_ERROR) {
4969 (connp->conn_recvicmp)(connp, mp, NULL, ira);
4970 } else {
4971 ill_t *rill = ira->ira_rill;
4972
4973 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
4974 ira->ira_ill = ira->ira_rill = NULL;
4975 /* Send it upstream */
4976 (connp->conn_recv)(connp, mp, NULL, ira);
4977 ira->ira_ill = ill;
4978 ira->ira_rill = rill;
4979 }
4980 }
4981
4982 /*
4983 * Handle protocols with which IP is less intimate. There
4984 * can be more than one stream bound to a particular
4985 * protocol. When this is the case, normally each one gets a copy
4986 * of any incoming packets.
4987 *
4988 * IPsec NOTE :
4989 *
4990 * Don't allow a secure packet going up a non-secure connection.
4991 * We don't allow this because
4992 *
4993 * 1) Reply might go out in clear which will be dropped at
4994 * the sending side.
4995 * 2) If the reply goes out in clear it will give the
4996 * adversary enough information for getting the key in
4997 * most of the cases.
4998 *
4999 * Moreover getting a secure packet when we expect clear
5000 * implies that SA's were added without checking for
5001 * policy on both ends. This should not happen once ISAKMP
5002 * is used to negotiate SAs as SAs will be added only after
5003 * verifying the policy.
5004 *
5005 * Zones notes:
5006 * Earlier in ip_input on a system with multiple shared-IP zones we
5007 * duplicate the multicast and broadcast packets and send them up
5008 * with each explicit zoneid that exists on that ill.
5009 * This means that here we can match the zoneid with SO_ALLZONES being special.
5010 */
5011 void
5012 ip_fanout_proto_v4(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
5013 {
5014 mblk_t *mp1;
5015 ipaddr_t laddr;
5016 conn_t *connp, *first_connp, *next_connp;
5017 connf_t *connfp;
5018 ill_t *ill = ira->ira_ill;
5019 ip_stack_t *ipst = ill->ill_ipst;
5020
5021 laddr = ipha->ipha_dst;
5022
5023 connfp = &ipst->ips_ipcl_proto_fanout_v4[ira->ira_protocol];
5024 mutex_enter(&connfp->connf_lock);
5025 connp = connfp->connf_head;
5026 for (connp = connfp->connf_head; connp != NULL;
5027 connp = connp->conn_next) {
5028 /* Note: IPCL_PROTO_MATCH includes conn_wantpacket */
5029 if (IPCL_PROTO_MATCH(connp, ira, ipha) &&
5030 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5031 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp))) {
5032 break;
5033 }
5034 }
5035
5036 if (connp == NULL) {
5037 /*
5038 * No one bound to these addresses. Is
5039 * there a client that wants all
5040 * unclaimed datagrams?
5041 */
5042 mutex_exit(&connfp->connf_lock);
5043 ip_fanout_send_icmp_v4(mp, ICMP_DEST_UNREACHABLE,
5044 ICMP_PROTOCOL_UNREACHABLE, ira);
5045 return;
5046 }
5047
5048 ASSERT(IPCL_IS_NONSTR(connp) || connp->conn_rq != NULL);
5049
5050 CONN_INC_REF(connp);
5051 first_connp = connp;
5052 connp = connp->conn_next;
5053
5054 for (;;) {
5055 while (connp != NULL) {
5056 /* Note: IPCL_PROTO_MATCH includes conn_wantpacket */
5057 if (IPCL_PROTO_MATCH(connp, ira, ipha) &&
5058 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5059 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5060 ira, connp)))
5061 break;
5062 connp = connp->conn_next;
5063 }
5064
5065 if (connp == NULL) {
5066 /* No more interested clients */
5067 connp = first_connp;
5068 break;
5069 }
5070 if (((mp1 = dupmsg(mp)) == NULL) &&
5071 ((mp1 = copymsg(mp)) == NULL)) {
5072 /* Memory allocation failed */
5073 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5074 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5075 connp = first_connp;
5076 break;
5077 }
5078
5079 CONN_INC_REF(connp);
5080 mutex_exit(&connfp->connf_lock);
5081
5082 ip_fanout_proto_conn(connp, mp1, (ipha_t *)mp1->b_rptr, NULL,
5083 ira);
5084
5085 mutex_enter(&connfp->connf_lock);
5086 /* Follow the next pointer before releasing the conn. */
5087 next_connp = connp->conn_next;
5088 CONN_DEC_REF(connp);
5089 connp = next_connp;
5090 }
5091
5092 /* Last one. Send it upstream. */
5093 mutex_exit(&connfp->connf_lock);
5094
5095 ip_fanout_proto_conn(connp, mp, ipha, NULL, ira);
5096
5097 CONN_DEC_REF(connp);
5098 }
5099
5100 /*
5101 * If we have a IPsec NAT-Traversal packet, strip the zero-SPI or
5102 * pass it along to ESP if the SPI is non-zero. Returns the mblk if the mblk
5103 * is not consumed.
5104 *
5105 * One of three things can happen, all of which affect the passed-in mblk:
5106 *
5107 * 1.) The packet is stock UDP and gets its zero-SPI stripped. Return mblk..
5108 *
5109 * 2.) The packet is ESP-in-UDP, gets transformed into an equivalent
5110 * ESP packet, and is passed along to ESP for consumption. Return NULL.
5111 *
5112 * 3.) The packet is an ESP-in-UDP Keepalive. Drop it and return NULL.
5113 */
5114 mblk_t *
5115 zero_spi_check(mblk_t *mp, ip_recv_attr_t *ira)
5116 {
5117 int shift, plen, iph_len;
5118 ipha_t *ipha;
5119 udpha_t *udpha;
5120 uint32_t *spi;
5121 uint32_t esp_ports;
5122 uint8_t *orptr;
5123 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
5124 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
5125
5126 ipha = (ipha_t *)mp->b_rptr;
5127 iph_len = ira->ira_ip_hdr_length;
5128 plen = ira->ira_pktlen;
5129
5130 if (plen - iph_len - sizeof (udpha_t) < sizeof (uint32_t)) {
5131 /*
5132 * Most likely a keepalive for the benefit of an intervening
5133 * NAT. These aren't for us, per se, so drop it.
5134 *
5135 * RFC 3947/8 doesn't say for sure what to do for 2-3
5136 * byte packets (keepalives are 1-byte), but we'll drop them
5137 * also.
5138 */
5139 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
5140 DROPPER(ipss, ipds_esp_nat_t_ka), &ipss->ipsec_dropper);
5141 return (NULL);
5142 }
5143
5144 if (MBLKL(mp) < iph_len + sizeof (udpha_t) + sizeof (*spi)) {
5145 /* might as well pull it all up - it might be ESP. */
5146 if (!pullupmsg(mp, -1)) {
5147 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
5148 DROPPER(ipss, ipds_esp_nomem),
5149 &ipss->ipsec_dropper);
5150 return (NULL);
5151 }
5152
5153 ipha = (ipha_t *)mp->b_rptr;
5154 }
5155 spi = (uint32_t *)(mp->b_rptr + iph_len + sizeof (udpha_t));
5156 if (*spi == 0) {
5157 /* UDP packet - remove 0-spi. */
5158 shift = sizeof (uint32_t);
5159 } else {
5160 /* ESP-in-UDP packet - reduce to ESP. */
5161 ipha->ipha_protocol = IPPROTO_ESP;
5162 shift = sizeof (udpha_t);
5163 }
5164
5165 /* Fix IP header */
5166 ira->ira_pktlen = (plen - shift);
5167 ipha->ipha_length = htons(ira->ira_pktlen);
5168 ipha->ipha_hdr_checksum = 0;
5169
5170 orptr = mp->b_rptr;
5171 mp->b_rptr += shift;
5172
5173 udpha = (udpha_t *)(orptr + iph_len);
5174 if (*spi == 0) {
5175 ASSERT((uint8_t *)ipha == orptr);
5176 udpha->uha_length = htons(plen - shift - iph_len);
5177 iph_len += sizeof (udpha_t); /* For the call to ovbcopy(). */
5178 esp_ports = 0;
5179 } else {
5180 esp_ports = *((uint32_t *)udpha);
5181 ASSERT(esp_ports != 0);
5182 }
5183 ovbcopy(orptr, orptr + shift, iph_len);
5184 if (esp_ports != 0) /* Punt up for ESP processing. */ {
5185 ipha = (ipha_t *)(orptr + shift);
5186
5187 ira->ira_flags |= IRAF_ESP_UDP_PORTS;
5188 ira->ira_esp_udp_ports = esp_ports;
5189 ip_fanout_v4(mp, ipha, ira);
5190 return (NULL);
5191 }
5192 return (mp);
5193 }
5194
5195 /*
5196 * Deliver a udp packet to the given conn, possibly applying ipsec policy.
5197 * Handles IPv4 and IPv6.
5198 * We are responsible for disposing of mp, such as by freemsg() or putnext()
5199 * Caller is responsible for dropping references to the conn.
5200 */
5201 void
5202 ip_fanout_udp_conn(conn_t *connp, mblk_t *mp, ipha_t *ipha, ip6_t *ip6h,
5203 ip_recv_attr_t *ira)
5204 {
5205 ill_t *ill = ira->ira_ill;
5206 ip_stack_t *ipst = ill->ill_ipst;
5207 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
5208 boolean_t secure;
5209 iaflags_t iraflags = ira->ira_flags;
5210
5211 secure = iraflags & IRAF_IPSEC_SECURE;
5212
5213 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld :
5214 !canputnext(connp->conn_rq)) {
5215 BUMP_MIB(ill->ill_ip_mib, udpIfStatsInOverflows);
5216 freemsg(mp);
5217 return;
5218 }
5219
5220 if (((iraflags & IRAF_IS_IPV4) ?
5221 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
5222 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
5223 secure) {
5224 mp = ipsec_check_inbound_policy(mp, connp, ipha,
5225 ip6h, ira);
5226 if (mp == NULL) {
5227 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5228 /* Note that mp is NULL */
5229 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5230 return;
5231 }
5232 }
5233
5234 /*
5235 * Since this code is not used for UDP unicast we don't need a NAT_T
5236 * check. Only ip_fanout_v4 has that check.
5237 */
5238 if (ira->ira_flags & IRAF_ICMP_ERROR) {
5239 (connp->conn_recvicmp)(connp, mp, NULL, ira);
5240 } else {
5241 ill_t *rill = ira->ira_rill;
5242
5243 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
5244 ira->ira_ill = ira->ira_rill = NULL;
5245 /* Send it upstream */
5246 (connp->conn_recv)(connp, mp, NULL, ira);
5247 ira->ira_ill = ill;
5248 ira->ira_rill = rill;
5249 }
5250 }
5251
5252 /*
5253 * Fanout for UDP packets that are multicast or broadcast, and ICMP errors.
5254 * (Unicast fanout is handled in ip_input_v4.)
5255 *
5256 * If SO_REUSEADDR is set all multicast and broadcast packets
5257 * will be delivered to all conns bound to the same port.
5258 *
5259 * If there is at least one matching AF_INET receiver, then we will
5260 * ignore any AF_INET6 receivers.
5261 * In the special case where an AF_INET socket binds to 0.0.0.0/<port> and an
5262 * AF_INET6 socket binds to ::/<port>, only the AF_INET socket receives the IPv4
5263 * packets.
5264 *
5265 * Zones notes:
5266 * Earlier in ip_input on a system with multiple shared-IP zones we
5267 * duplicate the multicast and broadcast packets and send them up
5268 * with each explicit zoneid that exists on that ill.
5269 * This means that here we can match the zoneid with SO_ALLZONES being special.
5270 */
5271 void
5272 ip_fanout_udp_multi_v4(mblk_t *mp, ipha_t *ipha, uint16_t lport, uint16_t fport,
5273 ip_recv_attr_t *ira)
5274 {
5275 ipaddr_t laddr;
5276 in6_addr_t v6faddr;
5277 conn_t *connp;
5278 connf_t *connfp;
5279 ipaddr_t faddr;
5280 ill_t *ill = ira->ira_ill;
5281 ip_stack_t *ipst = ill->ill_ipst;
5282
5283 ASSERT(ira->ira_flags & (IRAF_MULTIBROADCAST|IRAF_ICMP_ERROR));
5284
5285 laddr = ipha->ipha_dst;
5286 faddr = ipha->ipha_src;
5287
5288 connfp = &ipst->ips_ipcl_udp_fanout[IPCL_UDP_HASH(lport, ipst)];
5289 mutex_enter(&connfp->connf_lock);
5290 connp = connfp->connf_head;
5291
5292 /*
5293 * If SO_REUSEADDR has been set on the first we send the
5294 * packet to all clients that have joined the group and
5295 * match the port.
5296 */
5297 while (connp != NULL) {
5298 if ((IPCL_UDP_MATCH(connp, lport, laddr, fport, faddr)) &&
5299 conn_wantpacket(connp, ira, ipha) &&
5300 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5301 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp)))
5302 break;
5303 connp = connp->conn_next;
5304 }
5305
5306 if (connp == NULL)
5307 goto notfound;
5308
5309 CONN_INC_REF(connp);
5310
5311 if (connp->conn_reuseaddr) {
5312 conn_t *first_connp = connp;
5313 conn_t *next_connp;
5314 mblk_t *mp1;
5315
5316 connp = connp->conn_next;
5317 for (;;) {
5318 while (connp != NULL) {
5319 if (IPCL_UDP_MATCH(connp, lport, laddr,
5320 fport, faddr) &&
5321 conn_wantpacket(connp, ira, ipha) &&
5322 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5323 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5324 ira, connp)))
5325 break;
5326 connp = connp->conn_next;
5327 }
5328 if (connp == NULL) {
5329 /* No more interested clients */
5330 connp = first_connp;
5331 break;
5332 }
5333 if (((mp1 = dupmsg(mp)) == NULL) &&
5334 ((mp1 = copymsg(mp)) == NULL)) {
5335 /* Memory allocation failed */
5336 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5337 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5338 connp = first_connp;
5339 break;
5340 }
5341 CONN_INC_REF(connp);
5342 mutex_exit(&connfp->connf_lock);
5343
5344 IP_STAT(ipst, ip_udp_fanmb);
5345 ip_fanout_udp_conn(connp, mp1, (ipha_t *)mp1->b_rptr,
5346 NULL, ira);
5347 mutex_enter(&connfp->connf_lock);
5348 /* Follow the next pointer before releasing the conn */
5349 next_connp = connp->conn_next;
5350 CONN_DEC_REF(connp);
5351 connp = next_connp;
5352 }
5353 }
5354
5355 /* Last one. Send it upstream. */
5356 mutex_exit(&connfp->connf_lock);
5357 IP_STAT(ipst, ip_udp_fanmb);
5358 ip_fanout_udp_conn(connp, mp, ipha, NULL, ira);
5359 CONN_DEC_REF(connp);
5360 return;
5361
5362 notfound:
5363 mutex_exit(&connfp->connf_lock);
5364 /*
5365 * IPv6 endpoints bound to multicast IPv4-mapped addresses
5366 * have already been matched above, since they live in the IPv4
5367 * fanout tables. This implies we only need to
5368 * check for IPv6 in6addr_any endpoints here.
5369 * Thus we compare using ipv6_all_zeros instead of the destination
5370 * address, except for the multicast group membership lookup which
5371 * uses the IPv4 destination.
5372 */
5373 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_src, &v6faddr);
5374 connfp = &ipst->ips_ipcl_udp_fanout[IPCL_UDP_HASH(lport, ipst)];
5375 mutex_enter(&connfp->connf_lock);
5376 connp = connfp->connf_head;
5377 /*
5378 * IPv4 multicast packet being delivered to an AF_INET6
5379 * in6addr_any endpoint.
5380 * Need to check conn_wantpacket(). Note that we use conn_wantpacket()
5381 * and not conn_wantpacket_v6() since any multicast membership is
5382 * for an IPv4-mapped multicast address.
5383 */
5384 while (connp != NULL) {
5385 if (IPCL_UDP_MATCH_V6(connp, lport, ipv6_all_zeros,
5386 fport, v6faddr) &&
5387 conn_wantpacket(connp, ira, ipha) &&
5388 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5389 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp)))
5390 break;
5391 connp = connp->conn_next;
5392 }
5393
5394 if (connp == NULL) {
5395 /*
5396 * No one bound to this port. Is
5397 * there a client that wants all
5398 * unclaimed datagrams?
5399 */
5400 mutex_exit(&connfp->connf_lock);
5401
5402 if (ipst->ips_ipcl_proto_fanout_v4[IPPROTO_UDP].connf_head !=
5403 NULL) {
5404 ASSERT(ira->ira_protocol == IPPROTO_UDP);
5405 ip_fanout_proto_v4(mp, ipha, ira);
5406 } else {
5407 /*
5408 * We used to attempt to send an icmp error here, but
5409 * since this is known to be a multicast packet
5410 * and we don't send icmp errors in response to
5411 * multicast, just drop the packet and give up sooner.
5412 */
5413 BUMP_MIB(ill->ill_ip_mib, udpIfStatsNoPorts);
5414 freemsg(mp);
5415 }
5416 return;
5417 }
5418 CONN_INC_REF(connp);
5419 ASSERT(IPCL_IS_NONSTR(connp) || connp->conn_rq != NULL);
5420
5421 /*
5422 * If SO_REUSEADDR has been set on the first we send the
5423 * packet to all clients that have joined the group and
5424 * match the port.
5425 */
5426 if (connp->conn_reuseaddr) {
5427 conn_t *first_connp = connp;
5428 conn_t *next_connp;
5429 mblk_t *mp1;
5430
5431 connp = connp->conn_next;
5432 for (;;) {
5433 while (connp != NULL) {
5434 if (IPCL_UDP_MATCH_V6(connp, lport,
5435 ipv6_all_zeros, fport, v6faddr) &&
5436 conn_wantpacket(connp, ira, ipha) &&
5437 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5438 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5439 ira, connp)))
5440 break;
5441 connp = connp->conn_next;
5442 }
5443 if (connp == NULL) {
5444 /* No more interested clients */
5445 connp = first_connp;
5446 break;
5447 }
5448 if (((mp1 = dupmsg(mp)) == NULL) &&
5449 ((mp1 = copymsg(mp)) == NULL)) {
5450 /* Memory allocation failed */
5451 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5452 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5453 connp = first_connp;
5454 break;
5455 }
5456 CONN_INC_REF(connp);
5457 mutex_exit(&connfp->connf_lock);
5458
5459 IP_STAT(ipst, ip_udp_fanmb);
5460 ip_fanout_udp_conn(connp, mp1, (ipha_t *)mp1->b_rptr,
5461 NULL, ira);
5462 mutex_enter(&connfp->connf_lock);
5463 /* Follow the next pointer before releasing the conn */
5464 next_connp = connp->conn_next;
5465 CONN_DEC_REF(connp);
5466 connp = next_connp;
5467 }
5468 }
5469
5470 /* Last one. Send it upstream. */
5471 mutex_exit(&connfp->connf_lock);
5472 IP_STAT(ipst, ip_udp_fanmb);
5473 ip_fanout_udp_conn(connp, mp, ipha, NULL, ira);
5474 CONN_DEC_REF(connp);
5475 }
5476
5477 /*
5478 * Split an incoming packet's IPv4 options into the label and the other options.
5479 * If 'allocate' is set it does memory allocation for the ip_pkt_t, including
5480 * clearing out any leftover label or options.
5481 * Otherwise it just makes ipp point into the packet.
5482 *
5483 * Returns zero if ok; ENOMEM if the buffer couldn't be allocated.
5484 */
5485 int
5486 ip_find_hdr_v4(ipha_t *ipha, ip_pkt_t *ipp, boolean_t allocate)
5487 {
5488 uchar_t *opt;
5489 uint32_t totallen;
5490 uint32_t optval;
5491 uint32_t optlen;
5492
5493 ipp->ipp_fields |= IPPF_HOPLIMIT | IPPF_TCLASS | IPPF_ADDR;
5494 ipp->ipp_hoplimit = ipha->ipha_ttl;
5495 ipp->ipp_type_of_service = ipha->ipha_type_of_service;
5496 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &ipp->ipp_addr);
5497
5498 /*
5499 * Get length (in 4 byte octets) of IP header options.
5500 */
5501 totallen = ipha->ipha_version_and_hdr_length -
5502 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
5503
5504 if (totallen == 0) {
5505 if (!allocate)
5506 return (0);
5507
5508 /* Clear out anything from a previous packet */
5509 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
5510 kmem_free(ipp->ipp_ipv4_options,
5511 ipp->ipp_ipv4_options_len);
5512 ipp->ipp_ipv4_options = NULL;
5513 ipp->ipp_ipv4_options_len = 0;
5514 ipp->ipp_fields &= ~IPPF_IPV4_OPTIONS;
5515 }
5516 if (ipp->ipp_fields & IPPF_LABEL_V4) {
5517 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
5518 ipp->ipp_label_v4 = NULL;
5519 ipp->ipp_label_len_v4 = 0;
5520 ipp->ipp_fields &= ~IPPF_LABEL_V4;
5521 }
5522 return (0);
5523 }
5524
5525 totallen <<= 2;
5526 opt = (uchar_t *)&ipha[1];
5527 if (!is_system_labeled()) {
5528
5529 copyall:
5530 if (!allocate) {
5531 if (totallen != 0) {
5532 ipp->ipp_ipv4_options = opt;
5533 ipp->ipp_ipv4_options_len = totallen;
5534 ipp->ipp_fields |= IPPF_IPV4_OPTIONS;
5535 }
5536 return (0);
5537 }
5538 /* Just copy all of options */
5539 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
5540 if (totallen == ipp->ipp_ipv4_options_len) {
5541 bcopy(opt, ipp->ipp_ipv4_options, totallen);
5542 return (0);
5543 }
5544 kmem_free(ipp->ipp_ipv4_options,
5545 ipp->ipp_ipv4_options_len);
5546 ipp->ipp_ipv4_options = NULL;
5547 ipp->ipp_ipv4_options_len = 0;
5548 ipp->ipp_fields &= ~IPPF_IPV4_OPTIONS;
5549 }
5550 if (totallen == 0)
5551 return (0);
5552
5553 ipp->ipp_ipv4_options = kmem_alloc(totallen, KM_NOSLEEP);
5554 if (ipp->ipp_ipv4_options == NULL)
5555 return (ENOMEM);
5556 ipp->ipp_ipv4_options_len = totallen;
5557 ipp->ipp_fields |= IPPF_IPV4_OPTIONS;
5558 bcopy(opt, ipp->ipp_ipv4_options, totallen);
5559 return (0);
5560 }
5561
5562 if (allocate && (ipp->ipp_fields & IPPF_LABEL_V4)) {
5563 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
5564 ipp->ipp_label_v4 = NULL;
5565 ipp->ipp_label_len_v4 = 0;
5566 ipp->ipp_fields &= ~IPPF_LABEL_V4;
5567 }
5568
5569 /*
5570 * Search for CIPSO option.
5571 * We assume CIPSO is first in options if it is present.
5572 * If it isn't, then ipp_opt_ipv4_options will not include the options
5573 * prior to the CIPSO option.
5574 */
5575 while (totallen != 0) {
5576 switch (optval = opt[IPOPT_OPTVAL]) {
5577 case IPOPT_EOL:
5578 return (0);
5579 case IPOPT_NOP:
5580 optlen = 1;
5581 break;
5582 default:
5583 if (totallen <= IPOPT_OLEN)
5584 return (EINVAL);
5585 optlen = opt[IPOPT_OLEN];
5586 if (optlen < 2)
5587 return (EINVAL);
5588 }
5589 if (optlen > totallen)
5590 return (EINVAL);
5591
5592 switch (optval) {
5593 case IPOPT_COMSEC:
5594 if (!allocate) {
5595 ipp->ipp_label_v4 = opt;
5596 ipp->ipp_label_len_v4 = optlen;
5597 ipp->ipp_fields |= IPPF_LABEL_V4;
5598 } else {
5599 ipp->ipp_label_v4 = kmem_alloc(optlen,
5600 KM_NOSLEEP);
5601 if (ipp->ipp_label_v4 == NULL)
5602 return (ENOMEM);
5603 ipp->ipp_label_len_v4 = optlen;
5604 ipp->ipp_fields |= IPPF_LABEL_V4;
5605 bcopy(opt, ipp->ipp_label_v4, optlen);
5606 }
5607 totallen -= optlen;
5608 opt += optlen;
5609
5610 /* Skip padding bytes until we get to a multiple of 4 */
5611 while ((totallen & 3) != 0 && opt[0] == IPOPT_NOP) {
5612 totallen--;
5613 opt++;
5614 }
5615 /* Remaining as ipp_ipv4_options */
5616 goto copyall;
5617 }
5618 totallen -= optlen;
5619 opt += optlen;
5620 }
5621 /* No CIPSO found; return everything as ipp_ipv4_options */
5622 totallen = ipha->ipha_version_and_hdr_length -
5623 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
5624 totallen <<= 2;
5625 opt = (uchar_t *)&ipha[1];
5626 goto copyall;
5627 }
5628
5629 /*
5630 * Efficient versions of lookup for an IRE when we only
5631 * match the address.
5632 * For RTF_REJECT or BLACKHOLE we return IRE_NOROUTE.
5633 * Does not handle multicast addresses.
5634 */
5635 uint_t
5636 ip_type_v4(ipaddr_t addr, ip_stack_t *ipst)
5637 {
5638 ire_t *ire;
5639 uint_t result;
5640
5641 ire = ire_ftable_lookup_simple_v4(addr, 0, ipst, NULL);
5642 ASSERT(ire != NULL);
5643 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))
5644 result = IRE_NOROUTE;
5645 else
5646 result = ire->ire_type;
5647 ire_refrele(ire);
5648 return (result);
5649 }
5650
5651 /*
5652 * Efficient versions of lookup for an IRE when we only
5653 * match the address.
5654 * For RTF_REJECT or BLACKHOLE we return IRE_NOROUTE.
5655 * Does not handle multicast addresses.
5656 */
5657 uint_t
5658 ip_type_v6(const in6_addr_t *addr, ip_stack_t *ipst)
5659 {
5660 ire_t *ire;
5661 uint_t result;
5662
5663 ire = ire_ftable_lookup_simple_v6(addr, 0, ipst, NULL);
5664 ASSERT(ire != NULL);
5665 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))
5666 result = IRE_NOROUTE;
5667 else
5668 result = ire->ire_type;
5669 ire_refrele(ire);
5670 return (result);
5671 }
5672
5673 /*
5674 * Nobody should be sending
5675 * packets up this stream
5676 */
5677 static void
5678 ip_lrput(queue_t *q, mblk_t *mp)
5679 {
5680 switch (mp->b_datap->db_type) {
5681 case M_FLUSH:
5682 /* Turn around */
5683 if (*mp->b_rptr & FLUSHW) {
5684 *mp->b_rptr &= ~FLUSHR;
5685 qreply(q, mp);
5686 return;
5687 }
5688 break;
5689 }
5690 freemsg(mp);
5691 }
5692
5693 /* Nobody should be sending packets down this stream */
5694 /* ARGSUSED */
5695 void
5696 ip_lwput(queue_t *q, mblk_t *mp)
5697 {
5698 freemsg(mp);
5699 }
5700
5701 /*
5702 * Move the first hop in any source route to ipha_dst and remove that part of
5703 * the source route. Called by other protocols. Errors in option formatting
5704 * are ignored - will be handled by ip_output_options. Return the final
5705 * destination (either ipha_dst or the last entry in a source route.)
5706 */
5707 ipaddr_t
5708 ip_massage_options(ipha_t *ipha, netstack_t *ns)
5709 {
5710 ipoptp_t opts;
5711 uchar_t *opt;
5712 uint8_t optval;
5713 uint8_t optlen;
5714 ipaddr_t dst;
5715 int i;
5716 ip_stack_t *ipst = ns->netstack_ip;
5717
5718 ip2dbg(("ip_massage_options\n"));
5719 dst = ipha->ipha_dst;
5720 for (optval = ipoptp_first(&opts, ipha);
5721 optval != IPOPT_EOL;
5722 optval = ipoptp_next(&opts)) {
5723 opt = opts.ipoptp_cur;
5724 switch (optval) {
5725 uint8_t off;
5726 case IPOPT_SSRR:
5727 case IPOPT_LSRR:
5728 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
5729 ip1dbg(("ip_massage_options: bad src route\n"));
5730 break;
5731 }
5732 optlen = opts.ipoptp_len;
5733 off = opt[IPOPT_OFFSET];
5734 off--;
5735 redo_srr:
5736 if (optlen < IP_ADDR_LEN ||
5737 off > optlen - IP_ADDR_LEN) {
5738 /* End of source route */
5739 ip1dbg(("ip_massage_options: end of SR\n"));
5740 break;
5741 }
5742 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
5743 ip1dbg(("ip_massage_options: next hop 0x%x\n",
5744 ntohl(dst)));
5745 /*
5746 * Check if our address is present more than
5747 * once as consecutive hops in source route.
5748 * XXX verify per-interface ip_forwarding
5749 * for source route?
5750 */
5751 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
5752 off += IP_ADDR_LEN;
5753 goto redo_srr;
5754 }
5755 if (dst == htonl(INADDR_LOOPBACK)) {
5756 ip1dbg(("ip_massage_options: loopback addr in "
5757 "source route!\n"));
5758 break;
5759 }
5760 /*
5761 * Update ipha_dst to be the first hop and remove the
5762 * first hop from the source route (by overwriting
5763 * part of the option with NOP options).
5764 */
5765 ipha->ipha_dst = dst;
5766 /* Put the last entry in dst */
5767 off = ((optlen - IP_ADDR_LEN - 3) & ~(IP_ADDR_LEN-1)) +
5768 3;
5769 bcopy(&opt[off], &dst, IP_ADDR_LEN);
5770
5771 ip1dbg(("ip_massage_options: last hop 0x%x\n",
5772 ntohl(dst)));
5773 /* Move down and overwrite */
5774 opt[IP_ADDR_LEN] = opt[0];
5775 opt[IP_ADDR_LEN+1] = opt[IPOPT_OLEN] - IP_ADDR_LEN;
5776 opt[IP_ADDR_LEN+2] = opt[IPOPT_OFFSET];
5777 for (i = 0; i < IP_ADDR_LEN; i++)
5778 opt[i] = IPOPT_NOP;
5779 break;
5780 }
5781 }
5782 return (dst);
5783 }
5784
5785 /*
5786 * Return the network mask
5787 * associated with the specified address.
5788 */
5789 ipaddr_t
5790 ip_net_mask(ipaddr_t addr)
5791 {
5792 uchar_t *up = (uchar_t *)&addr;
5793 ipaddr_t mask = 0;
5794 uchar_t *maskp = (uchar_t *)&mask;
5795
5796 #if defined(__i386) || defined(__amd64)
5797 #define TOTALLY_BRAIN_DAMAGED_C_COMPILER
5798 #endif
5799 #ifdef TOTALLY_BRAIN_DAMAGED_C_COMPILER
5800 maskp[0] = maskp[1] = maskp[2] = maskp[3] = 0;
5801 #endif
5802 if (CLASSD(addr)) {
5803 maskp[0] = 0xF0;
5804 return (mask);
5805 }
5806
5807 /* We assume Class E default netmask to be 32 */
5808 if (CLASSE(addr))
5809 return (0xffffffffU);
5810
5811 if (addr == 0)
5812 return (0);
5813 maskp[0] = 0xFF;
5814 if ((up[0] & 0x80) == 0)
5815 return (mask);
5816
5817 maskp[1] = 0xFF;
5818 if ((up[0] & 0xC0) == 0x80)
5819 return (mask);
5820
5821 maskp[2] = 0xFF;
5822 if ((up[0] & 0xE0) == 0xC0)
5823 return (mask);
5824
5825 /* Otherwise return no mask */
5826 return ((ipaddr_t)0);
5827 }
5828
5829 /* Name/Value Table Lookup Routine */
5830 char *
5831 ip_nv_lookup(nv_t *nv, int value)
5832 {
5833 if (!nv)
5834 return (NULL);
5835 for (; nv->nv_name; nv++) {
5836 if (nv->nv_value == value)
5837 return (nv->nv_name);
5838 }
5839 return ("unknown");
5840 }
5841
5842 static int
5843 ip_wait_for_info_ack(ill_t *ill)
5844 {
5845 int err;
5846
5847 mutex_enter(&ill->ill_lock);
5848 while (ill->ill_state_flags & ILL_LL_SUBNET_PENDING) {
5849 /*
5850 * Return value of 0 indicates a pending signal.
5851 */
5852 err = cv_wait_sig(&ill->ill_cv, &ill->ill_lock);
5853 if (err == 0) {
5854 mutex_exit(&ill->ill_lock);
5855 return (EINTR);
5856 }
5857 }
5858 mutex_exit(&ill->ill_lock);
5859 /*
5860 * ip_rput_other could have set an error in ill_error on
5861 * receipt of M_ERROR.
5862 */
5863 return (ill->ill_error);
5864 }
5865
5866 /*
5867 * This is a module open, i.e. this is a control stream for access
5868 * to a DLPI device. We allocate an ill_t as the instance data in
5869 * this case.
5870 */
5871 static int
5872 ip_modopen(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5873 {
5874 ill_t *ill;
5875 int err;
5876 zoneid_t zoneid;
5877 netstack_t *ns;
5878 ip_stack_t *ipst;
5879
5880 /*
5881 * Prevent unprivileged processes from pushing IP so that
5882 * they can't send raw IP.
5883 */
5884 if (secpolicy_net_rawaccess(credp) != 0)
5885 return (EPERM);
5886
5887 ns = netstack_find_by_cred(credp);
5888 ASSERT(ns != NULL);
5889 ipst = ns->netstack_ip;
5890 ASSERT(ipst != NULL);
5891
5892 /*
5893 * For exclusive stacks we set the zoneid to zero
5894 * to make IP operate as if in the global zone.
5895 */
5896 if (ipst->ips_netstack->netstack_stackid != GLOBAL_NETSTACKID)
5897 zoneid = GLOBAL_ZONEID;
5898 else
5899 zoneid = crgetzoneid(credp);
5900
5901 ill = (ill_t *)mi_open_alloc_sleep(sizeof (ill_t));
5902 q->q_ptr = WR(q)->q_ptr = ill;
5903 ill->ill_ipst = ipst;
5904 ill->ill_zoneid = zoneid;
5905
5906 /*
5907 * ill_init initializes the ill fields and then sends down
5908 * down a DL_INFO_REQ after calling qprocson.
5909 */
5910 err = ill_init(q, ill);
5911
5912 if (err != 0) {
5913 mi_free(ill);
5914 netstack_rele(ipst->ips_netstack);
5915 q->q_ptr = NULL;
5916 WR(q)->q_ptr = NULL;
5917 return (err);
5918 }
5919
5920 /*
5921 * Wait for the DL_INFO_ACK if a DL_INFO_REQ was sent.
5922 *
5923 * ill_init initializes the ipsq marking this thread as
5924 * writer
5925 */
5926 ipsq_exit(ill->ill_phyint->phyint_ipsq);
5927 err = ip_wait_for_info_ack(ill);
5928 if (err == 0)
5929 ill->ill_credp = credp;
5930 else
5931 goto fail;
5932
5933 crhold(credp);
5934
5935 mutex_enter(&ipst->ips_ip_mi_lock);
5936 err = mi_open_link(&ipst->ips_ip_g_head, (IDP)q->q_ptr, devp, flag,
5937 sflag, credp);
5938 mutex_exit(&ipst->ips_ip_mi_lock);
5939 fail:
5940 if (err) {
5941 (void) ip_close(q, 0);
5942 return (err);
5943 }
5944 return (0);
5945 }
5946
5947 /* For /dev/ip aka AF_INET open */
5948 int
5949 ip_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5950 {
5951 return (ip_open(q, devp, flag, sflag, credp, B_FALSE));
5952 }
5953
5954 /* For /dev/ip6 aka AF_INET6 open */
5955 int
5956 ip_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5957 {
5958 return (ip_open(q, devp, flag, sflag, credp, B_TRUE));
5959 }
5960
5961 /* IP open routine. */
5962 int
5963 ip_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp,
5964 boolean_t isv6)
5965 {
5966 conn_t *connp;
5967 major_t maj;
5968 zoneid_t zoneid;
5969 netstack_t *ns;
5970 ip_stack_t *ipst;
5971
5972 /* Allow reopen. */
5973 if (q->q_ptr != NULL)
5974 return (0);
5975
5976 if (sflag & MODOPEN) {
5977 /* This is a module open */
5978 return (ip_modopen(q, devp, flag, sflag, credp));
5979 }
5980
5981 if ((flag & ~(FKLYR)) == IP_HELPER_STR) {
5982 /*
5983 * Non streams based socket looking for a stream
5984 * to access IP
5985 */
5986 return (ip_helper_stream_setup(q, devp, flag, sflag,
5987 credp, isv6));
5988 }
5989
5990 ns = netstack_find_by_cred(credp);
5991 ASSERT(ns != NULL);
5992 ipst = ns->netstack_ip;
5993 ASSERT(ipst != NULL);
5994
5995 /*
5996 * For exclusive stacks we set the zoneid to zero
5997 * to make IP operate as if in the global zone.
5998 */
5999 if (ipst->ips_netstack->netstack_stackid != GLOBAL_NETSTACKID)
6000 zoneid = GLOBAL_ZONEID;
6001 else
6002 zoneid = crgetzoneid(credp);
6003
6004 /*
6005 * We are opening as a device. This is an IP client stream, and we
6006 * allocate an conn_t as the instance data.
6007 */
6008 connp = ipcl_conn_create(IPCL_IPCCONN, KM_SLEEP, ipst->ips_netstack);
6009
6010 /*
6011 * ipcl_conn_create did a netstack_hold. Undo the hold that was
6012 * done by netstack_find_by_cred()
6013 */
6014 netstack_rele(ipst->ips_netstack);
6015
6016 connp->conn_ixa->ixa_flags |= IXAF_MULTICAST_LOOP | IXAF_SET_ULP_CKSUM;
6017 /* conn_allzones can not be set this early, hence no IPCL_ZONEID */
6018 connp->conn_ixa->ixa_zoneid = zoneid;
6019 connp->conn_zoneid = zoneid;
6020
6021 connp->conn_rq = q;
6022 q->q_ptr = WR(q)->q_ptr = connp;
6023
6024 /* Minor tells us which /dev entry was opened */
6025 if (isv6) {
6026 connp->conn_family = AF_INET6;
6027 connp->conn_ipversion = IPV6_VERSION;
6028 connp->conn_ixa->ixa_flags &= ~IXAF_IS_IPV4;
6029 connp->conn_ixa->ixa_src_preferences = IPV6_PREFER_SRC_DEFAULT;
6030 } else {
6031 connp->conn_family = AF_INET;
6032 connp->conn_ipversion = IPV4_VERSION;
6033 connp->conn_ixa->ixa_flags |= IXAF_IS_IPV4;
6034 }
6035
6036 if ((ip_minor_arena_la != NULL) && (flag & SO_SOCKSTR) &&
6037 ((connp->conn_dev = inet_minor_alloc(ip_minor_arena_la)) != 0)) {
6038 connp->conn_minor_arena = ip_minor_arena_la;
6039 } else {
6040 /*
6041 * Either minor numbers in the large arena were exhausted
6042 * or a non socket application is doing the open.
6043 * Try to allocate from the small arena.
6044 */
6045 if ((connp->conn_dev =
6046 inet_minor_alloc(ip_minor_arena_sa)) == 0) {
6047 /* CONN_DEC_REF takes care of netstack_rele() */
6048 q->q_ptr = WR(q)->q_ptr = NULL;
6049 CONN_DEC_REF(connp);
6050 return (EBUSY);
6051 }
6052 connp->conn_minor_arena = ip_minor_arena_sa;
6053 }
6054
6055 maj = getemajor(*devp);
6056 *devp = makedevice(maj, (minor_t)connp->conn_dev);
6057
6058 /*
6059 * connp->conn_cred is crfree()ed in ipcl_conn_destroy()
6060 */
6061 connp->conn_cred = credp;
6062 connp->conn_cpid = curproc->p_pid;
6063 /* Cache things in ixa without an extra refhold */
6064 ASSERT(!(connp->conn_ixa->ixa_free_flags & IXA_FREE_CRED));
6065 connp->conn_ixa->ixa_cred = connp->conn_cred;
6066 connp->conn_ixa->ixa_cpid = connp->conn_cpid;
6067 if (is_system_labeled())
6068 connp->conn_ixa->ixa_tsl = crgetlabel(connp->conn_cred);
6069
6070 /*
6071 * Handle IP_IOC_RTS_REQUEST and other ioctls which use conn_recv
6072 */
6073 connp->conn_recv = ip_conn_input;
6074 connp->conn_recvicmp = ip_conn_input_icmp;
6075
6076 crhold(connp->conn_cred);
6077
6078 /*
6079 * If the caller has the process-wide flag set, then default to MAC
6080 * exempt mode. This allows read-down to unlabeled hosts.
6081 */
6082 if (getpflags(NET_MAC_AWARE, credp) != 0)
6083 connp->conn_mac_mode = CONN_MAC_AWARE;
6084
6085 connp->conn_zone_is_global = (crgetzoneid(credp) == GLOBAL_ZONEID);
6086
6087 connp->conn_rq = q;
6088 connp->conn_wq = WR(q);
6089
6090 /* Non-zero default values */
6091 connp->conn_ixa->ixa_flags |= IXAF_MULTICAST_LOOP;
6092
6093 /*
6094 * Make the conn globally visible to walkers
6095 */
6096 ASSERT(connp->conn_ref == 1);
6097 mutex_enter(&connp->conn_lock);
6098 connp->conn_state_flags &= ~CONN_INCIPIENT;
6099 mutex_exit(&connp->conn_lock);
6100
6101 qprocson(q);
6102
6103 return (0);
6104 }
6105
6106 /*
6107 * Set IPsec policy from an ipsec_req_t. If the req is not "zero" and valid,
6108 * all of them are copied to the conn_t. If the req is "zero", the policy is
6109 * zeroed out. A "zero" policy has zero ipsr_{ah,req,self_encap}_req
6110 * fields.
6111 * We keep only the latest setting of the policy and thus policy setting
6112 * is not incremental/cumulative.
6113 *
6114 * Requests to set policies with multiple alternative actions will
6115 * go through a different API.
6116 */
6117 int
6118 ipsec_set_req(cred_t *cr, conn_t *connp, ipsec_req_t *req)
6119 {
6120 uint_t ah_req = 0;
6121 uint_t esp_req = 0;
6122 uint_t se_req = 0;
6123 ipsec_act_t *actp = NULL;
6124 uint_t nact;
6125 ipsec_policy_head_t *ph;
6126 boolean_t is_pol_reset, is_pol_inserted = B_FALSE;
6127 int error = 0;
6128 netstack_t *ns = connp->conn_netstack;
6129 ip_stack_t *ipst = ns->netstack_ip;
6130 ipsec_stack_t *ipss = ns->netstack_ipsec;
6131
6132 #define REQ_MASK (IPSEC_PREF_REQUIRED|IPSEC_PREF_NEVER)
6133
6134 /*
6135 * The IP_SEC_OPT option does not allow variable length parameters,
6136 * hence a request cannot be NULL.
6137 */
6138 if (req == NULL)
6139 return (EINVAL);
6140
6141 ah_req = req->ipsr_ah_req;
6142 esp_req = req->ipsr_esp_req;
6143 se_req = req->ipsr_self_encap_req;
6144
6145 /* Don't allow setting self-encap without one or more of AH/ESP. */
6146 if (se_req != 0 && esp_req == 0 && ah_req == 0)
6147 return (EINVAL);
6148
6149 /*
6150 * Are we dealing with a request to reset the policy (i.e.
6151 * zero requests).
6152 */
6153 is_pol_reset = ((ah_req & REQ_MASK) == 0 &&
6154 (esp_req & REQ_MASK) == 0 &&
6155 (se_req & REQ_MASK) == 0);
6156
6157 if (!is_pol_reset) {
6158 /*
6159 * If we couldn't load IPsec, fail with "protocol
6160 * not supported".
6161 * IPsec may not have been loaded for a request with zero
6162 * policies, so we don't fail in this case.
6163 */
6164 mutex_enter(&ipss->ipsec_loader_lock);
6165 if (ipss->ipsec_loader_state != IPSEC_LOADER_SUCCEEDED) {
6166 mutex_exit(&ipss->ipsec_loader_lock);
6167 return (EPROTONOSUPPORT);
6168 }
6169 mutex_exit(&ipss->ipsec_loader_lock);
6170
6171 /*
6172 * Test for valid requests. Invalid algorithms
6173 * need to be tested by IPsec code because new
6174 * algorithms can be added dynamically.
6175 */
6176 if ((ah_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0 ||
6177 (esp_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0 ||
6178 (se_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0) {
6179 return (EINVAL);
6180 }
6181
6182 /*
6183 * Only privileged users can issue these
6184 * requests.
6185 */
6186 if (((ah_req & IPSEC_PREF_NEVER) ||
6187 (esp_req & IPSEC_PREF_NEVER) ||
6188 (se_req & IPSEC_PREF_NEVER)) &&
6189 secpolicy_ip_config(cr, B_FALSE) != 0) {
6190 return (EPERM);
6191 }
6192
6193 /*
6194 * The IPSEC_PREF_REQUIRED and IPSEC_PREF_NEVER
6195 * are mutually exclusive.
6196 */
6197 if (((ah_req & REQ_MASK) == REQ_MASK) ||
6198 ((esp_req & REQ_MASK) == REQ_MASK) ||
6199 ((se_req & REQ_MASK) == REQ_MASK)) {
6200 /* Both of them are set */
6201 return (EINVAL);
6202 }
6203 }
6204
6205 ASSERT(MUTEX_HELD(&connp->conn_lock));
6206
6207 /*
6208 * If we have already cached policies in conn_connect(), don't
6209 * let them change now. We cache policies for connections
6210 * whose src,dst [addr, port] is known.
6211 */
6212 if (connp->conn_policy_cached) {
6213 return (EINVAL);
6214 }
6215
6216 /*
6217 * We have a zero policies, reset the connection policy if already
6218 * set. This will cause the connection to inherit the
6219 * global policy, if any.
6220 */
6221 if (is_pol_reset) {
6222 if (connp->conn_policy != NULL) {
6223 IPPH_REFRELE(connp->conn_policy, ipst->ips_netstack);
6224 connp->conn_policy = NULL;
6225 }
6226 connp->conn_in_enforce_policy = B_FALSE;
6227 connp->conn_out_enforce_policy = B_FALSE;
6228 return (0);
6229 }
6230
6231 ph = connp->conn_policy = ipsec_polhead_split(connp->conn_policy,
6232 ipst->ips_netstack);
6233 if (ph == NULL)
6234 goto enomem;
6235
6236 ipsec_actvec_from_req(req, &actp, &nact, ipst->ips_netstack);
6237 if (actp == NULL)
6238 goto enomem;
6239
6240 /*
6241 * Always insert IPv4 policy entries, since they can also apply to
6242 * ipv6 sockets being used in ipv4-compat mode.
6243 */
6244 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V4,
6245 IPSEC_TYPE_INBOUND, ns))
6246 goto enomem;
6247 is_pol_inserted = B_TRUE;
6248 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V4,
6249 IPSEC_TYPE_OUTBOUND, ns))
6250 goto enomem;
6251
6252 /*
6253 * We're looking at a v6 socket, also insert the v6-specific
6254 * entries.
6255 */
6256 if (connp->conn_family == AF_INET6) {
6257 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V6,
6258 IPSEC_TYPE_INBOUND, ns))
6259 goto enomem;
6260 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V6,
6261 IPSEC_TYPE_OUTBOUND, ns))
6262 goto enomem;
6263 }
6264
6265 ipsec_actvec_free(actp, nact);
6266
6267 /*
6268 * If the requests need security, set enforce_policy.
6269 * If the requests are IPSEC_PREF_NEVER, one should
6270 * still set conn_out_enforce_policy so that ip_set_destination
6271 * marks the ip_xmit_attr_t appropriatly. This is needed so that
6272 * for connections that we don't cache policy in at connect time,
6273 * if global policy matches in ip_output_attach_policy, we
6274 * don't wrongly inherit global policy. Similarly, we need
6275 * to set conn_in_enforce_policy also so that we don't verify
6276 * policy wrongly.
6277 */
6278 if ((ah_req & REQ_MASK) != 0 ||
6279 (esp_req & REQ_MASK) != 0 ||
6280 (se_req & REQ_MASK) != 0) {
6281 connp->conn_in_enforce_policy = B_TRUE;
6282 connp->conn_out_enforce_policy = B_TRUE;
6283 }
6284
6285 return (error);
6286 #undef REQ_MASK
6287
6288 /*
6289 * Common memory-allocation-failure exit path.
6290 */
6291 enomem:
6292 if (actp != NULL)
6293 ipsec_actvec_free(actp, nact);
6294 if (is_pol_inserted)
6295 ipsec_polhead_flush(ph, ns);
6296 return (ENOMEM);
6297 }
6298
6299 /*
6300 * Set socket options for joining and leaving multicast groups.
6301 * Common to IPv4 and IPv6; inet6 indicates the type of socket.
6302 * The caller has already check that the option name is consistent with
6303 * the address family of the socket.
6304 */
6305 int
6306 ip_opt_set_multicast_group(conn_t *connp, t_scalar_t name,
6307 uchar_t *invalp, boolean_t inet6, boolean_t checkonly)
6308 {
6309 int *i1 = (int *)invalp;
6310 int error = 0;
6311 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
6312 struct ip_mreq *v4_mreqp;
6313 struct ipv6_mreq *v6_mreqp;
6314 struct group_req *greqp;
6315 ire_t *ire;
6316 boolean_t done = B_FALSE;
6317 ipaddr_t ifaddr;
6318 in6_addr_t v6group;
6319 uint_t ifindex;
6320 boolean_t mcast_opt = B_TRUE;
6321 mcast_record_t fmode;
6322 int (*optfn)(conn_t *, boolean_t, const in6_addr_t *,
6323 ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *);
6324
6325 switch (name) {
6326 case IP_ADD_MEMBERSHIP:
6327 case IPV6_JOIN_GROUP:
6328 mcast_opt = B_FALSE;
6329 /* FALLTHRU */
6330 case MCAST_JOIN_GROUP:
6331 fmode = MODE_IS_EXCLUDE;
6332 optfn = ip_opt_add_group;
6333 break;
6334
6335 case IP_DROP_MEMBERSHIP:
6336 case IPV6_LEAVE_GROUP:
6337 mcast_opt = B_FALSE;
6338 /* FALLTHRU */
6339 case MCAST_LEAVE_GROUP:
6340 fmode = MODE_IS_INCLUDE;
6341 optfn = ip_opt_delete_group;
6342 break;
6343 default:
6344 ASSERT(0);
6345 }
6346
6347 if (mcast_opt) {
6348 struct sockaddr_in *sin;
6349 struct sockaddr_in6 *sin6;
6350
6351 greqp = (struct group_req *)i1;
6352 if (greqp->gr_group.ss_family == AF_INET) {
6353 sin = (struct sockaddr_in *)&(greqp->gr_group);
6354 IN6_INADDR_TO_V4MAPPED(&sin->sin_addr, &v6group);
6355 } else {
6356 if (!inet6)
6357 return (EINVAL); /* Not on INET socket */
6358
6359 sin6 = (struct sockaddr_in6 *)&(greqp->gr_group);
6360 v6group = sin6->sin6_addr;
6361 }
6362 ifaddr = INADDR_ANY;
6363 ifindex = greqp->gr_interface;
6364 } else if (inet6) {
6365 v6_mreqp = (struct ipv6_mreq *)i1;
6366 v6group = v6_mreqp->ipv6mr_multiaddr;
6367 ifaddr = INADDR_ANY;
6368 ifindex = v6_mreqp->ipv6mr_interface;
6369 } else {
6370 v4_mreqp = (struct ip_mreq *)i1;
6371 IN6_INADDR_TO_V4MAPPED(&v4_mreqp->imr_multiaddr, &v6group);
6372 ifaddr = (ipaddr_t)v4_mreqp->imr_interface.s_addr;
6373 ifindex = 0;
6374 }
6375
6376 /*
6377 * In the multirouting case, we need to replicate
6378 * the request on all interfaces that will take part
6379 * in replication. We do so because multirouting is
6380 * reflective, thus we will probably receive multi-
6381 * casts on those interfaces.
6382 * The ip_multirt_apply_membership() succeeds if
6383 * the operation succeeds on at least one interface.
6384 */
6385 if (IN6_IS_ADDR_V4MAPPED(&v6group)) {
6386 ipaddr_t group;
6387
6388 IN6_V4MAPPED_TO_IPADDR(&v6group, group);
6389
6390 ire = ire_ftable_lookup_v4(group, IP_HOST_MASK, 0,
6391 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6392 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6393 } else {
6394 ire = ire_ftable_lookup_v6(&v6group, &ipv6_all_ones, 0,
6395 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6396 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6397 }
6398 if (ire != NULL) {
6399 if (ire->ire_flags & RTF_MULTIRT) {
6400 error = ip_multirt_apply_membership(optfn, ire, connp,
6401 checkonly, &v6group, fmode, &ipv6_all_zeros);
6402 done = B_TRUE;
6403 }
6404 ire_refrele(ire);
6405 }
6406
6407 if (!done) {
6408 error = optfn(connp, checkonly, &v6group, ifaddr, ifindex,
6409 fmode, &ipv6_all_zeros);
6410 }
6411 return (error);
6412 }
6413
6414 /*
6415 * Set socket options for joining and leaving multicast groups
6416 * for specific sources.
6417 * Common to IPv4 and IPv6; inet6 indicates the type of socket.
6418 * The caller has already check that the option name is consistent with
6419 * the address family of the socket.
6420 */
6421 int
6422 ip_opt_set_multicast_sources(conn_t *connp, t_scalar_t name,
6423 uchar_t *invalp, boolean_t inet6, boolean_t checkonly)
6424 {
6425 int *i1 = (int *)invalp;
6426 int error = 0;
6427 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
6428 struct ip_mreq_source *imreqp;
6429 struct group_source_req *gsreqp;
6430 in6_addr_t v6group, v6src;
6431 uint32_t ifindex;
6432 ipaddr_t ifaddr;
6433 boolean_t mcast_opt = B_TRUE;
6434 mcast_record_t fmode;
6435 ire_t *ire;
6436 boolean_t done = B_FALSE;
6437 int (*optfn)(conn_t *, boolean_t, const in6_addr_t *,
6438 ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *);
6439
6440 switch (name) {
6441 case IP_BLOCK_SOURCE:
6442 mcast_opt = B_FALSE;
6443 /* FALLTHRU */
6444 case MCAST_BLOCK_SOURCE:
6445 fmode = MODE_IS_EXCLUDE;
6446 optfn = ip_opt_add_group;
6447 break;
6448
6449 case IP_UNBLOCK_SOURCE:
6450 mcast_opt = B_FALSE;
6451 /* FALLTHRU */
6452 case MCAST_UNBLOCK_SOURCE:
6453 fmode = MODE_IS_EXCLUDE;
6454 optfn = ip_opt_delete_group;
6455 break;
6456
6457 case IP_ADD_SOURCE_MEMBERSHIP:
6458 mcast_opt = B_FALSE;
6459 /* FALLTHRU */
6460 case MCAST_JOIN_SOURCE_GROUP:
6461 fmode = MODE_IS_INCLUDE;
6462 optfn = ip_opt_add_group;
6463 break;
6464
6465 case IP_DROP_SOURCE_MEMBERSHIP:
6466 mcast_opt = B_FALSE;
6467 /* FALLTHRU */
6468 case MCAST_LEAVE_SOURCE_GROUP:
6469 fmode = MODE_IS_INCLUDE;
6470 optfn = ip_opt_delete_group;
6471 break;
6472 default:
6473 ASSERT(0);
6474 }
6475
6476 if (mcast_opt) {
6477 gsreqp = (struct group_source_req *)i1;
6478 ifindex = gsreqp->gsr_interface;
6479 if (gsreqp->gsr_group.ss_family == AF_INET) {
6480 struct sockaddr_in *s;
6481 s = (struct sockaddr_in *)&gsreqp->gsr_group;
6482 IN6_INADDR_TO_V4MAPPED(&s->sin_addr, &v6group);
6483 s = (struct sockaddr_in *)&gsreqp->gsr_source;
6484 IN6_INADDR_TO_V4MAPPED(&s->sin_addr, &v6src);
6485 } else {
6486 struct sockaddr_in6 *s6;
6487
6488 if (!inet6)
6489 return (EINVAL); /* Not on INET socket */
6490
6491 s6 = (struct sockaddr_in6 *)&gsreqp->gsr_group;
6492 v6group = s6->sin6_addr;
6493 s6 = (struct sockaddr_in6 *)&gsreqp->gsr_source;
6494 v6src = s6->sin6_addr;
6495 }
6496 ifaddr = INADDR_ANY;
6497 } else {
6498 imreqp = (struct ip_mreq_source *)i1;
6499 IN6_INADDR_TO_V4MAPPED(&imreqp->imr_multiaddr, &v6group);
6500 IN6_INADDR_TO_V4MAPPED(&imreqp->imr_sourceaddr, &v6src);
6501 ifaddr = (ipaddr_t)imreqp->imr_interface.s_addr;
6502 ifindex = 0;
6503 }
6504
6505 /*
6506 * Handle src being mapped INADDR_ANY by changing it to unspecified.
6507 */
6508 if (IN6_IS_ADDR_V4MAPPED_ANY(&v6src))
6509 v6src = ipv6_all_zeros;
6510
6511 /*
6512 * In the multirouting case, we need to replicate
6513 * the request as noted in the mcast cases above.
6514 */
6515 if (IN6_IS_ADDR_V4MAPPED(&v6group)) {
6516 ipaddr_t group;
6517
6518 IN6_V4MAPPED_TO_IPADDR(&v6group, group);
6519
6520 ire = ire_ftable_lookup_v4(group, IP_HOST_MASK, 0,
6521 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6522 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6523 } else {
6524 ire = ire_ftable_lookup_v6(&v6group, &ipv6_all_ones, 0,
6525 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6526 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6527 }
6528 if (ire != NULL) {
6529 if (ire->ire_flags & RTF_MULTIRT) {
6530 error = ip_multirt_apply_membership(optfn, ire, connp,
6531 checkonly, &v6group, fmode, &v6src);
6532 done = B_TRUE;
6533 }
6534 ire_refrele(ire);
6535 }
6536 if (!done) {
6537 error = optfn(connp, checkonly, &v6group, ifaddr, ifindex,
6538 fmode, &v6src);
6539 }
6540 return (error);
6541 }
6542
6543 /*
6544 * Given a destination address and a pointer to where to put the information
6545 * this routine fills in the mtuinfo.
6546 * The socket must be connected.
6547 * For sctp conn_faddr is the primary address.
6548 */
6549 int
6550 ip_fill_mtuinfo(conn_t *connp, ip_xmit_attr_t *ixa, struct ip6_mtuinfo *mtuinfo)
6551 {
6552 uint32_t pmtu = IP_MAXPACKET;
6553 uint_t scopeid;
6554
6555 if (IN6_IS_ADDR_UNSPECIFIED(&connp->conn_faddr_v6))
6556 return (-1);
6557
6558 /* In case we never sent or called ip_set_destination_v4/v6 */
6559 if (ixa->ixa_ire != NULL)
6560 pmtu = ip_get_pmtu(ixa);
6561
6562 if (ixa->ixa_flags & IXAF_SCOPEID_SET)
6563 scopeid = ixa->ixa_scopeid;
6564 else
6565 scopeid = 0;
6566
6567 bzero(mtuinfo, sizeof (*mtuinfo));
6568 mtuinfo->ip6m_addr.sin6_family = AF_INET6;
6569 mtuinfo->ip6m_addr.sin6_port = connp->conn_fport;
6570 mtuinfo->ip6m_addr.sin6_addr = connp->conn_faddr_v6;
6571 mtuinfo->ip6m_addr.sin6_scope_id = scopeid;
6572 mtuinfo->ip6m_mtu = pmtu;
6573
6574 return (sizeof (struct ip6_mtuinfo));
6575 }
6576
6577 /*
6578 * When the src multihoming is changed from weak to [strong, preferred]
6579 * ip_ire_rebind_walker is called to walk the list of all ire_t entries
6580 * and identify routes that were created by user-applications in the
6581 * unbound state (i.e., without RTA_IFP), and for which an ire_ill is not
6582 * currently defined. These routes are then 'rebound', i.e., their ire_ill
6583 * is selected by finding an interface route for the gateway.
6584 */
6585 /* ARGSUSED */
6586 void
6587 ip_ire_rebind_walker(ire_t *ire, void *notused)
6588 {
6589 if (!ire->ire_unbound || ire->ire_ill != NULL)
6590 return;
6591 ire_rebind(ire);
6592 ire_delete(ire);
6593 }
6594
6595 /*
6596 * When the src multihoming is changed from [strong, preferred] to weak,
6597 * ip_ire_unbind_walker is called to walk the list of all ire_t entries, and
6598 * set any entries that were created by user-applications in the unbound state
6599 * (i.e., without RTA_IFP) back to having a NULL ire_ill.
6600 */
6601 /* ARGSUSED */
6602 void
6603 ip_ire_unbind_walker(ire_t *ire, void *notused)
6604 {
6605 ire_t *new_ire;
6606
6607 if (!ire->ire_unbound || ire->ire_ill == NULL)
6608 return;
6609 if (ire->ire_ipversion == IPV6_VERSION) {
6610 new_ire = ire_create_v6(&ire->ire_addr_v6, &ire->ire_mask_v6,
6611 &ire->ire_gateway_addr_v6, ire->ire_type, NULL,
6612 ire->ire_zoneid, ire->ire_flags, NULL, ire->ire_ipst);
6613 } else {
6614 new_ire = ire_create((uchar_t *)&ire->ire_addr,
6615 (uchar_t *)&ire->ire_mask,
6616 (uchar_t *)&ire->ire_gateway_addr, ire->ire_type, NULL,
6617 ire->ire_zoneid, ire->ire_flags, NULL, ire->ire_ipst);
6618 }
6619 if (new_ire == NULL)
6620 return;
6621 new_ire->ire_unbound = B_TRUE;
6622 /*
6623 * The bound ire must first be deleted so that we don't return
6624 * the existing one on the attempt to add the unbound new_ire.
6625 */
6626 ire_delete(ire);
6627 new_ire = ire_add(new_ire);
6628 if (new_ire != NULL)
6629 ire_refrele(new_ire);
6630 }
6631
6632 /*
6633 * When the settings of ip*_strict_src_multihoming tunables are changed,
6634 * all cached routes need to be recomputed. This recomputation needs to be
6635 * done when going from weaker to stronger modes so that the cached ire
6636 * for the connection does not violate the current ip*_strict_src_multihoming
6637 * setting. It also needs to be done when going from stronger to weaker modes,
6638 * so that we fall back to matching on the longest-matching-route (as opposed
6639 * to a shorter match that may have been selected in the strong mode
6640 * to satisfy src_multihoming settings).
6641 *
6642 * The cached ixa_ire entires for all conn_t entries are marked as
6643 * "verify" so that they will be recomputed for the next packet.
6644 */
6645 void
6646 conn_ire_revalidate(conn_t *connp, void *arg)
6647 {
6648 boolean_t isv6 = (boolean_t)arg;
6649
6650 if ((isv6 && connp->conn_ipversion != IPV6_VERSION) ||
6651 (!isv6 && connp->conn_ipversion != IPV4_VERSION))
6652 return;
6653 connp->conn_ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
6654 }
6655
6656 /*
6657 * Handles both IPv4 and IPv6 reassembly - doing the out-of-order cases,
6658 * When an ipf is passed here for the first time, if
6659 * we already have in-order fragments on the queue, we convert from the fast-
6660 * path reassembly scheme to the hard-case scheme. From then on, additional
6661 * fragments are reassembled here. We keep track of the start and end offsets
6662 * of each piece, and the number of holes in the chain. When the hole count
6663 * goes to zero, we are done!
6664 *
6665 * The ipf_count will be updated to account for any mblk(s) added (pointed to
6666 * by mp) or subtracted (freeb()ed dups), upon return the caller must update
6667 * ipfb_count and ill_frag_count by the difference of ipf_count before and
6668 * after the call to ip_reassemble().
6669 */
6670 int
6671 ip_reassemble(mblk_t *mp, ipf_t *ipf, uint_t start, boolean_t more, ill_t *ill,
6672 size_t msg_len)
6673 {
6674 uint_t end;
6675 mblk_t *next_mp;
6676 mblk_t *mp1;
6677 uint_t offset;
6678 boolean_t incr_dups = B_TRUE;
6679 boolean_t offset_zero_seen = B_FALSE;
6680 boolean_t pkt_boundary_checked = B_FALSE;
6681
6682 /* If start == 0 then ipf_nf_hdr_len has to be set. */
6683 ASSERT(start != 0 || ipf->ipf_nf_hdr_len != 0);
6684
6685 /* Add in byte count */
6686 ipf->ipf_count += msg_len;
6687 if (ipf->ipf_end) {
6688 /*
6689 * We were part way through in-order reassembly, but now there
6690 * is a hole. We walk through messages already queued, and
6691 * mark them for hard case reassembly. We know that up till
6692 * now they were in order starting from offset zero.
6693 */
6694 offset = 0;
6695 for (mp1 = ipf->ipf_mp->b_cont; mp1; mp1 = mp1->b_cont) {
6696 IP_REASS_SET_START(mp1, offset);
6697 if (offset == 0) {
6698 ASSERT(ipf->ipf_nf_hdr_len != 0);
6699 offset = -ipf->ipf_nf_hdr_len;
6700 }
6701 offset += mp1->b_wptr - mp1->b_rptr;
6702 IP_REASS_SET_END(mp1, offset);
6703 }
6704 /* One hole at the end. */
6705 ipf->ipf_hole_cnt = 1;
6706 /* Brand it as a hard case, forever. */
6707 ipf->ipf_end = 0;
6708 }
6709 /* Walk through all the new pieces. */
6710 do {
6711 end = start + (mp->b_wptr - mp->b_rptr);
6712 /*
6713 * If start is 0, decrease 'end' only for the first mblk of
6714 * the fragment. Otherwise 'end' can get wrong value in the
6715 * second pass of the loop if first mblk is exactly the
6716 * size of ipf_nf_hdr_len.
6717 */
6718 if (start == 0 && !offset_zero_seen) {
6719 /* First segment */
6720 ASSERT(ipf->ipf_nf_hdr_len != 0);
6721 end -= ipf->ipf_nf_hdr_len;
6722 offset_zero_seen = B_TRUE;
6723 }
6724 next_mp = mp->b_cont;
6725 /*
6726 * We are checking to see if there is any interesing data
6727 * to process. If there isn't and the mblk isn't the
6728 * one which carries the unfragmentable header then we
6729 * drop it. It's possible to have just the unfragmentable
6730 * header come through without any data. That needs to be
6731 * saved.
6732 *
6733 * If the assert at the top of this function holds then the
6734 * term "ipf->ipf_nf_hdr_len != 0" isn't needed. This code
6735 * is infrequently traveled enough that the test is left in
6736 * to protect against future code changes which break that
6737 * invariant.
6738 */
6739 if (start == end && start != 0 && ipf->ipf_nf_hdr_len != 0) {
6740 /* Empty. Blast it. */
6741 IP_REASS_SET_START(mp, 0);
6742 IP_REASS_SET_END(mp, 0);
6743 /*
6744 * If the ipf points to the mblk we are about to free,
6745 * update ipf to point to the next mblk (or NULL
6746 * if none).
6747 */
6748 if (ipf->ipf_mp->b_cont == mp)
6749 ipf->ipf_mp->b_cont = next_mp;
6750 freeb(mp);
6751 continue;
6752 }
6753 mp->b_cont = NULL;
6754 IP_REASS_SET_START(mp, start);
6755 IP_REASS_SET_END(mp, end);
6756 if (!ipf->ipf_tail_mp) {
6757 ipf->ipf_tail_mp = mp;
6758 ipf->ipf_mp->b_cont = mp;
6759 if (start == 0 || !more) {
6760 ipf->ipf_hole_cnt = 1;
6761 /*
6762 * if the first fragment comes in more than one
6763 * mblk, this loop will be executed for each
6764 * mblk. Need to adjust hole count so exiting
6765 * this routine will leave hole count at 1.
6766 */
6767 if (next_mp)
6768 ipf->ipf_hole_cnt++;
6769 } else
6770 ipf->ipf_hole_cnt = 2;
6771 continue;
6772 } else if (ipf->ipf_last_frag_seen && !more &&
6773 !pkt_boundary_checked) {
6774 /*
6775 * We check datagram boundary only if this fragment
6776 * claims to be the last fragment and we have seen a
6777 * last fragment in the past too. We do this only
6778 * once for a given fragment.
6779 *
6780 * start cannot be 0 here as fragments with start=0
6781 * and MF=0 gets handled as a complete packet. These
6782 * fragments should not reach here.
6783 */
6784
6785 if (start + msgdsize(mp) !=
6786 IP_REASS_END(ipf->ipf_tail_mp)) {
6787 /*
6788 * We have two fragments both of which claim
6789 * to be the last fragment but gives conflicting
6790 * information about the whole datagram size.
6791 * Something fishy is going on. Drop the
6792 * fragment and free up the reassembly list.
6793 */
6794 return (IP_REASS_FAILED);
6795 }
6796
6797 /*
6798 * We shouldn't come to this code block again for this
6799 * particular fragment.
6800 */
6801 pkt_boundary_checked = B_TRUE;
6802 }
6803
6804 /* New stuff at or beyond tail? */
6805 offset = IP_REASS_END(ipf->ipf_tail_mp);
6806 if (start >= offset) {
6807 if (ipf->ipf_last_frag_seen) {
6808 /* current fragment is beyond last fragment */
6809 return (IP_REASS_FAILED);
6810 }
6811 /* Link it on end. */
6812 ipf->ipf_tail_mp->b_cont = mp;
6813 ipf->ipf_tail_mp = mp;
6814 if (more) {
6815 if (start != offset)
6816 ipf->ipf_hole_cnt++;
6817 } else if (start == offset && next_mp == NULL)
6818 ipf->ipf_hole_cnt--;
6819 continue;
6820 }
6821 mp1 = ipf->ipf_mp->b_cont;
6822 offset = IP_REASS_START(mp1);
6823 /* New stuff at the front? */
6824 if (start < offset) {
6825 if (start == 0) {
6826 if (end >= offset) {
6827 /* Nailed the hole at the begining. */
6828 ipf->ipf_hole_cnt--;
6829 }
6830 } else if (end < offset) {
6831 /*
6832 * A hole, stuff, and a hole where there used
6833 * to be just a hole.
6834 */
6835 ipf->ipf_hole_cnt++;
6836 }
6837 mp->b_cont = mp1;
6838 /* Check for overlap. */
6839 while (end > offset) {
6840 if (end < IP_REASS_END(mp1)) {
6841 mp->b_wptr -= end - offset;
6842 IP_REASS_SET_END(mp, offset);
6843 BUMP_MIB(ill->ill_ip_mib,
6844 ipIfStatsReasmPartDups);
6845 break;
6846 }
6847 /* Did we cover another hole? */
6848 if ((mp1->b_cont &&
6849 IP_REASS_END(mp1) !=
6850 IP_REASS_START(mp1->b_cont) &&
6851 end >= IP_REASS_START(mp1->b_cont)) ||
6852 (!ipf->ipf_last_frag_seen && !more)) {
6853 ipf->ipf_hole_cnt--;
6854 }
6855 /* Clip out mp1. */
6856 if ((mp->b_cont = mp1->b_cont) == NULL) {
6857 /*
6858 * After clipping out mp1, this guy
6859 * is now hanging off the end.
6860 */
6861 ipf->ipf_tail_mp = mp;
6862 }
6863 IP_REASS_SET_START(mp1, 0);
6864 IP_REASS_SET_END(mp1, 0);
6865 /* Subtract byte count */
6866 ipf->ipf_count -= mp1->b_datap->db_lim -
6867 mp1->b_datap->db_base;
6868 freeb(mp1);
6869 BUMP_MIB(ill->ill_ip_mib,
6870 ipIfStatsReasmPartDups);
6871 mp1 = mp->b_cont;
6872 if (!mp1)
6873 break;
6874 offset = IP_REASS_START(mp1);
6875 }
6876 ipf->ipf_mp->b_cont = mp;
6877 continue;
6878 }
6879 /*
6880 * The new piece starts somewhere between the start of the head
6881 * and before the end of the tail.
6882 */
6883 for (; mp1; mp1 = mp1->b_cont) {
6884 offset = IP_REASS_END(mp1);
6885 if (start < offset) {
6886 if (end <= offset) {
6887 /* Nothing new. */
6888 IP_REASS_SET_START(mp, 0);
6889 IP_REASS_SET_END(mp, 0);
6890 /* Subtract byte count */
6891 ipf->ipf_count -= mp->b_datap->db_lim -
6892 mp->b_datap->db_base;
6893 if (incr_dups) {
6894 ipf->ipf_num_dups++;
6895 incr_dups = B_FALSE;
6896 }
6897 freeb(mp);
6898 BUMP_MIB(ill->ill_ip_mib,
6899 ipIfStatsReasmDuplicates);
6900 break;
6901 }
6902 /*
6903 * Trim redundant stuff off beginning of new
6904 * piece.
6905 */
6906 IP_REASS_SET_START(mp, offset);
6907 mp->b_rptr += offset - start;
6908 BUMP_MIB(ill->ill_ip_mib,
6909 ipIfStatsReasmPartDups);
6910 start = offset;
6911 if (!mp1->b_cont) {
6912 /*
6913 * After trimming, this guy is now
6914 * hanging off the end.
6915 */
6916 mp1->b_cont = mp;
6917 ipf->ipf_tail_mp = mp;
6918 if (!more) {
6919 ipf->ipf_hole_cnt--;
6920 }
6921 break;
6922 }
6923 }
6924 if (start >= IP_REASS_START(mp1->b_cont))
6925 continue;
6926 /* Fill a hole */
6927 if (start > offset)
6928 ipf->ipf_hole_cnt++;
6929 mp->b_cont = mp1->b_cont;
6930 mp1->b_cont = mp;
6931 mp1 = mp->b_cont;
6932 offset = IP_REASS_START(mp1);
6933 if (end >= offset) {
6934 ipf->ipf_hole_cnt--;
6935 /* Check for overlap. */
6936 while (end > offset) {
6937 if (end < IP_REASS_END(mp1)) {
6938 mp->b_wptr -= end - offset;
6939 IP_REASS_SET_END(mp, offset);
6940 /*
6941 * TODO we might bump
6942 * this up twice if there is
6943 * overlap at both ends.
6944 */
6945 BUMP_MIB(ill->ill_ip_mib,
6946 ipIfStatsReasmPartDups);
6947 break;
6948 }
6949 /* Did we cover another hole? */
6950 if ((mp1->b_cont &&
6951 IP_REASS_END(mp1)
6952 != IP_REASS_START(mp1->b_cont) &&
6953 end >=
6954 IP_REASS_START(mp1->b_cont)) ||
6955 (!ipf->ipf_last_frag_seen &&
6956 !more)) {
6957 ipf->ipf_hole_cnt--;
6958 }
6959 /* Clip out mp1. */
6960 if ((mp->b_cont = mp1->b_cont) ==
6961 NULL) {
6962 /*
6963 * After clipping out mp1,
6964 * this guy is now hanging
6965 * off the end.
6966 */
6967 ipf->ipf_tail_mp = mp;
6968 }
6969 IP_REASS_SET_START(mp1, 0);
6970 IP_REASS_SET_END(mp1, 0);
6971 /* Subtract byte count */
6972 ipf->ipf_count -=
6973 mp1->b_datap->db_lim -
6974 mp1->b_datap->db_base;
6975 freeb(mp1);
6976 BUMP_MIB(ill->ill_ip_mib,
6977 ipIfStatsReasmPartDups);
6978 mp1 = mp->b_cont;
6979 if (!mp1)
6980 break;
6981 offset = IP_REASS_START(mp1);
6982 }
6983 }
6984 break;
6985 }
6986 } while (start = end, mp = next_mp);
6987
6988 /* Fragment just processed could be the last one. Remember this fact */
6989 if (!more)
6990 ipf->ipf_last_frag_seen = B_TRUE;
6991
6992 /* Still got holes? */
6993 if (ipf->ipf_hole_cnt)
6994 return (IP_REASS_PARTIAL);
6995 /* Clean up overloaded fields to avoid upstream disasters. */
6996 for (mp1 = ipf->ipf_mp->b_cont; mp1; mp1 = mp1->b_cont) {
6997 IP_REASS_SET_START(mp1, 0);
6998 IP_REASS_SET_END(mp1, 0);
6999 }
7000 return (IP_REASS_COMPLETE);
7001 }
7002
7003 /*
7004 * Fragmentation reassembly. Each ILL has a hash table for
7005 * queuing packets undergoing reassembly for all IPIFs
7006 * associated with the ILL. The hash is based on the packet
7007 * IP ident field. The ILL frag hash table was allocated
7008 * as a timer block at the time the ILL was created. Whenever
7009 * there is anything on the reassembly queue, the timer will
7010 * be running. Returns the reassembled packet if reassembly completes.
7011 */
7012 mblk_t *
7013 ip_input_fragment(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
7014 {
7015 uint32_t frag_offset_flags;
7016 mblk_t *t_mp;
7017 ipaddr_t dst;
7018 uint8_t proto = ipha->ipha_protocol;
7019 uint32_t sum_val;
7020 uint16_t sum_flags;
7021 ipf_t *ipf;
7022 ipf_t **ipfp;
7023 ipfb_t *ipfb;
7024 uint16_t ident;
7025 uint32_t offset;
7026 ipaddr_t src;
7027 uint_t hdr_length;
7028 uint32_t end;
7029 mblk_t *mp1;
7030 mblk_t *tail_mp;
7031 size_t count;
7032 size_t msg_len;
7033 uint8_t ecn_info = 0;
7034 uint32_t packet_size;
7035 boolean_t pruned = B_FALSE;
7036 ill_t *ill = ira->ira_ill;
7037 ip_stack_t *ipst = ill->ill_ipst;
7038
7039 /*
7040 * Drop the fragmented as early as possible, if
7041 * we don't have resource(s) to re-assemble.
7042 */
7043 if (ipst->ips_ip_reass_queue_bytes == 0) {
7044 freemsg(mp);
7045 return (NULL);
7046 }
7047
7048 /* Check for fragmentation offset; return if there's none */
7049 if ((frag_offset_flags = ntohs(ipha->ipha_fragment_offset_and_flags) &
7050 (IPH_MF | IPH_OFFSET)) == 0)
7051 return (mp);
7052
7053 /*
7054 * We utilize hardware computed checksum info only for UDP since
7055 * IP fragmentation is a normal occurrence for the protocol. In
7056 * addition, checksum offload support for IP fragments carrying
7057 * UDP payload is commonly implemented across network adapters.
7058 */
7059 ASSERT(ira->ira_rill != NULL);
7060 if (proto == IPPROTO_UDP && dohwcksum &&
7061 ILL_HCKSUM_CAPABLE(ira->ira_rill) &&
7062 (DB_CKSUMFLAGS(mp) & (HCK_FULLCKSUM | HCK_PARTIALCKSUM))) {
7063 mblk_t *mp1 = mp->b_cont;
7064 int32_t len;
7065
7066 /* Record checksum information from the packet */
7067 sum_val = (uint32_t)DB_CKSUM16(mp);
7068 sum_flags = DB_CKSUMFLAGS(mp);
7069
7070 /* IP payload offset from beginning of mblk */
7071 offset = ((uchar_t *)ipha + IPH_HDR_LENGTH(ipha)) - mp->b_rptr;
7072
7073 if ((sum_flags & HCK_PARTIALCKSUM) &&
7074 (mp1 == NULL || mp1->b_cont == NULL) &&
7075 offset >= DB_CKSUMSTART(mp) &&
7076 ((len = offset - DB_CKSUMSTART(mp)) & 1) == 0) {
7077 uint32_t adj;
7078 /*
7079 * Partial checksum has been calculated by hardware
7080 * and attached to the packet; in addition, any
7081 * prepended extraneous data is even byte aligned.
7082 * If any such data exists, we adjust the checksum;
7083 * this would also handle any postpended data.
7084 */
7085 IP_ADJCKSUM_PARTIAL(mp->b_rptr + DB_CKSUMSTART(mp),
7086 mp, mp1, len, adj);
7087
7088 /* One's complement subtract extraneous checksum */
7089 if (adj >= sum_val)
7090 sum_val = ~(adj - sum_val) & 0xFFFF;
7091 else
7092 sum_val -= adj;
7093 }
7094 } else {
7095 sum_val = 0;
7096 sum_flags = 0;
7097 }
7098
7099 /* Clear hardware checksumming flag */
7100 DB_CKSUMFLAGS(mp) = 0;
7101
7102 ident = ipha->ipha_ident;
7103 offset = (frag_offset_flags << 3) & 0xFFFF;
7104 src = ipha->ipha_src;
7105 dst = ipha->ipha_dst;
7106 hdr_length = IPH_HDR_LENGTH(ipha);
7107 end = ntohs(ipha->ipha_length) - hdr_length;
7108
7109 /* If end == 0 then we have a packet with no data, so just free it */
7110 if (end == 0) {
7111 freemsg(mp);
7112 return (NULL);
7113 }
7114
7115 /* Record the ECN field info. */
7116 ecn_info = (ipha->ipha_type_of_service & 0x3);
7117 if (offset != 0) {
7118 /*
7119 * If this isn't the first piece, strip the header, and
7120 * add the offset to the end value.
7121 */
7122 mp->b_rptr += hdr_length;
7123 end += offset;
7124 }
7125
7126 /* Handle vnic loopback of fragments */
7127 if (mp->b_datap->db_ref > 2)
7128 msg_len = 0;
7129 else
7130 msg_len = MBLKSIZE(mp);
7131
7132 tail_mp = mp;
7133 while (tail_mp->b_cont != NULL) {
7134 tail_mp = tail_mp->b_cont;
7135 if (tail_mp->b_datap->db_ref <= 2)
7136 msg_len += MBLKSIZE(tail_mp);
7137 }
7138
7139 /* If the reassembly list for this ILL will get too big, prune it */
7140 if ((msg_len + sizeof (*ipf) + ill->ill_frag_count) >=
7141 ipst->ips_ip_reass_queue_bytes) {
7142 DTRACE_PROBE3(ip_reass_queue_bytes, uint_t, msg_len,
7143 uint_t, ill->ill_frag_count,
7144 uint_t, ipst->ips_ip_reass_queue_bytes);
7145 ill_frag_prune(ill,
7146 (ipst->ips_ip_reass_queue_bytes < msg_len) ? 0 :
7147 (ipst->ips_ip_reass_queue_bytes - msg_len));
7148 pruned = B_TRUE;
7149 }
7150
7151 ipfb = &ill->ill_frag_hash_tbl[ILL_FRAG_HASH(src, ident)];
7152 mutex_enter(&ipfb->ipfb_lock);
7153
7154 ipfp = &ipfb->ipfb_ipf;
7155 /* Try to find an existing fragment queue for this packet. */
7156 for (;;) {
7157 ipf = ipfp[0];
7158 if (ipf != NULL) {
7159 /*
7160 * It has to match on ident and src/dst address.
7161 */
7162 if (ipf->ipf_ident == ident &&
7163 ipf->ipf_src == src &&
7164 ipf->ipf_dst == dst &&
7165 ipf->ipf_protocol == proto) {
7166 /*
7167 * If we have received too many
7168 * duplicate fragments for this packet
7169 * free it.
7170 */
7171 if (ipf->ipf_num_dups > ip_max_frag_dups) {
7172 ill_frag_free_pkts(ill, ipfb, ipf, 1);
7173 freemsg(mp);
7174 mutex_exit(&ipfb->ipfb_lock);
7175 return (NULL);
7176 }
7177 /* Found it. */
7178 break;
7179 }
7180 ipfp = &ipf->ipf_hash_next;
7181 continue;
7182 }
7183
7184 /*
7185 * If we pruned the list, do we want to store this new
7186 * fragment?. We apply an optimization here based on the
7187 * fact that most fragments will be received in order.
7188 * So if the offset of this incoming fragment is zero,
7189 * it is the first fragment of a new packet. We will
7190 * keep it. Otherwise drop the fragment, as we have
7191 * probably pruned the packet already (since the
7192 * packet cannot be found).
7193 */
7194 if (pruned && offset != 0) {
7195 mutex_exit(&ipfb->ipfb_lock);
7196 freemsg(mp);
7197 return (NULL);
7198 }
7199
7200 if (ipfb->ipfb_frag_pkts >= MAX_FRAG_PKTS(ipst)) {
7201 /*
7202 * Too many fragmented packets in this hash
7203 * bucket. Free the oldest.
7204 */
7205 ill_frag_free_pkts(ill, ipfb, ipfb->ipfb_ipf, 1);
7206 }
7207
7208 /* New guy. Allocate a frag message. */
7209 mp1 = allocb(sizeof (*ipf), BPRI_MED);
7210 if (mp1 == NULL) {
7211 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7212 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7213 freemsg(mp);
7214 reass_done:
7215 mutex_exit(&ipfb->ipfb_lock);
7216 return (NULL);
7217 }
7218
7219 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmReqds);
7220 mp1->b_cont = mp;
7221
7222 /* Initialize the fragment header. */
7223 ipf = (ipf_t *)mp1->b_rptr;
7224 ipf->ipf_mp = mp1;
7225 ipf->ipf_ptphn = ipfp;
7226 ipfp[0] = ipf;
7227 ipf->ipf_hash_next = NULL;
7228 ipf->ipf_ident = ident;
7229 ipf->ipf_protocol = proto;
7230 ipf->ipf_src = src;
7231 ipf->ipf_dst = dst;
7232 ipf->ipf_nf_hdr_len = 0;
7233 /* Record reassembly start time. */
7234 ipf->ipf_timestamp = gethrestime_sec();
7235 /* Record ipf generation and account for frag header */
7236 ipf->ipf_gen = ill->ill_ipf_gen++;
7237 ipf->ipf_count = MBLKSIZE(mp1);
7238 ipf->ipf_last_frag_seen = B_FALSE;
7239 ipf->ipf_ecn = ecn_info;
7240 ipf->ipf_num_dups = 0;
7241 ipfb->ipfb_frag_pkts++;
7242 ipf->ipf_checksum = 0;
7243 ipf->ipf_checksum_flags = 0;
7244
7245 /* Store checksum value in fragment header */
7246 if (sum_flags != 0) {
7247 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7248 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7249 ipf->ipf_checksum = sum_val;
7250 ipf->ipf_checksum_flags = sum_flags;
7251 }
7252
7253 /*
7254 * We handle reassembly two ways. In the easy case,
7255 * where all the fragments show up in order, we do
7256 * minimal bookkeeping, and just clip new pieces on
7257 * the end. If we ever see a hole, then we go off
7258 * to ip_reassemble which has to mark the pieces and
7259 * keep track of the number of holes, etc. Obviously,
7260 * the point of having both mechanisms is so we can
7261 * handle the easy case as efficiently as possible.
7262 */
7263 if (offset == 0) {
7264 /* Easy case, in-order reassembly so far. */
7265 ipf->ipf_count += msg_len;
7266 ipf->ipf_tail_mp = tail_mp;
7267 /*
7268 * Keep track of next expected offset in
7269 * ipf_end.
7270 */
7271 ipf->ipf_end = end;
7272 ipf->ipf_nf_hdr_len = hdr_length;
7273 } else {
7274 /* Hard case, hole at the beginning. */
7275 ipf->ipf_tail_mp = NULL;
7276 /*
7277 * ipf_end == 0 means that we have given up
7278 * on easy reassembly.
7279 */
7280 ipf->ipf_end = 0;
7281
7282 /* Forget checksum offload from now on */
7283 ipf->ipf_checksum_flags = 0;
7284
7285 /*
7286 * ipf_hole_cnt is set by ip_reassemble.
7287 * ipf_count is updated by ip_reassemble.
7288 * No need to check for return value here
7289 * as we don't expect reassembly to complete
7290 * or fail for the first fragment itself.
7291 */
7292 (void) ip_reassemble(mp, ipf,
7293 (frag_offset_flags & IPH_OFFSET) << 3,
7294 (frag_offset_flags & IPH_MF), ill, msg_len);
7295 }
7296 /* Update per ipfb and ill byte counts */
7297 ipfb->ipfb_count += ipf->ipf_count;
7298 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7299 atomic_add_32(&ill->ill_frag_count, ipf->ipf_count);
7300 /* If the frag timer wasn't already going, start it. */
7301 mutex_enter(&ill->ill_lock);
7302 ill_frag_timer_start(ill);
7303 mutex_exit(&ill->ill_lock);
7304 goto reass_done;
7305 }
7306
7307 /*
7308 * If the packet's flag has changed (it could be coming up
7309 * from an interface different than the previous, therefore
7310 * possibly different checksum capability), then forget about
7311 * any stored checksum states. Otherwise add the value to
7312 * the existing one stored in the fragment header.
7313 */
7314 if (sum_flags != 0 && sum_flags == ipf->ipf_checksum_flags) {
7315 sum_val += ipf->ipf_checksum;
7316 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7317 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7318 ipf->ipf_checksum = sum_val;
7319 } else if (ipf->ipf_checksum_flags != 0) {
7320 /* Forget checksum offload from now on */
7321 ipf->ipf_checksum_flags = 0;
7322 }
7323
7324 /*
7325 * We have a new piece of a datagram which is already being
7326 * reassembled. Update the ECN info if all IP fragments
7327 * are ECN capable. If there is one which is not, clear
7328 * all the info. If there is at least one which has CE
7329 * code point, IP needs to report that up to transport.
7330 */
7331 if (ecn_info != IPH_ECN_NECT && ipf->ipf_ecn != IPH_ECN_NECT) {
7332 if (ecn_info == IPH_ECN_CE)
7333 ipf->ipf_ecn = IPH_ECN_CE;
7334 } else {
7335 ipf->ipf_ecn = IPH_ECN_NECT;
7336 }
7337 if (offset && ipf->ipf_end == offset) {
7338 /* The new fragment fits at the end */
7339 ipf->ipf_tail_mp->b_cont = mp;
7340 /* Update the byte count */
7341 ipf->ipf_count += msg_len;
7342 /* Update per ipfb and ill byte counts */
7343 ipfb->ipfb_count += msg_len;
7344 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7345 atomic_add_32(&ill->ill_frag_count, msg_len);
7346 if (frag_offset_flags & IPH_MF) {
7347 /* More to come. */
7348 ipf->ipf_end = end;
7349 ipf->ipf_tail_mp = tail_mp;
7350 goto reass_done;
7351 }
7352 } else {
7353 /* Go do the hard cases. */
7354 int ret;
7355
7356 if (offset == 0)
7357 ipf->ipf_nf_hdr_len = hdr_length;
7358
7359 /* Save current byte count */
7360 count = ipf->ipf_count;
7361 ret = ip_reassemble(mp, ipf,
7362 (frag_offset_flags & IPH_OFFSET) << 3,
7363 (frag_offset_flags & IPH_MF), ill, msg_len);
7364 /* Count of bytes added and subtracted (freeb()ed) */
7365 count = ipf->ipf_count - count;
7366 if (count) {
7367 /* Update per ipfb and ill byte counts */
7368 ipfb->ipfb_count += count;
7369 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7370 atomic_add_32(&ill->ill_frag_count, count);
7371 }
7372 if (ret == IP_REASS_PARTIAL) {
7373 goto reass_done;
7374 } else if (ret == IP_REASS_FAILED) {
7375 /* Reassembly failed. Free up all resources */
7376 ill_frag_free_pkts(ill, ipfb, ipf, 1);
7377 for (t_mp = mp; t_mp != NULL; t_mp = t_mp->b_cont) {
7378 IP_REASS_SET_START(t_mp, 0);
7379 IP_REASS_SET_END(t_mp, 0);
7380 }
7381 freemsg(mp);
7382 goto reass_done;
7383 }
7384 /* We will reach here iff 'ret' is IP_REASS_COMPLETE */
7385 }
7386 /*
7387 * We have completed reassembly. Unhook the frag header from
7388 * the reassembly list.
7389 *
7390 * Before we free the frag header, record the ECN info
7391 * to report back to the transport.
7392 */
7393 ecn_info = ipf->ipf_ecn;
7394 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmOKs);
7395 ipfp = ipf->ipf_ptphn;
7396
7397 /* We need to supply these to caller */
7398 if ((sum_flags = ipf->ipf_checksum_flags) != 0)
7399 sum_val = ipf->ipf_checksum;
7400 else
7401 sum_val = 0;
7402
7403 mp1 = ipf->ipf_mp;
7404 count = ipf->ipf_count;
7405 ipf = ipf->ipf_hash_next;
7406 if (ipf != NULL)
7407 ipf->ipf_ptphn = ipfp;
7408 ipfp[0] = ipf;
7409 atomic_add_32(&ill->ill_frag_count, -count);
7410 ASSERT(ipfb->ipfb_count >= count);
7411 ipfb->ipfb_count -= count;
7412 ipfb->ipfb_frag_pkts--;
7413 mutex_exit(&ipfb->ipfb_lock);
7414 /* Ditch the frag header. */
7415 mp = mp1->b_cont;
7416
7417 freeb(mp1);
7418
7419 /* Restore original IP length in header. */
7420 packet_size = (uint32_t)msgdsize(mp);
7421 if (packet_size > IP_MAXPACKET) {
7422 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7423 ip_drop_input("Reassembled packet too large", mp, ill);
7424 freemsg(mp);
7425 return (NULL);
7426 }
7427
7428 if (DB_REF(mp) > 1) {
7429 mblk_t *mp2 = copymsg(mp);
7430
7431 if (mp2 == NULL) {
7432 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7433 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7434 freemsg(mp);
7435 return (NULL);
7436 }
7437 freemsg(mp);
7438 mp = mp2;
7439 }
7440 ipha = (ipha_t *)mp->b_rptr;
7441
7442 ipha->ipha_length = htons((uint16_t)packet_size);
7443 /* We're now complete, zip the frag state */
7444 ipha->ipha_fragment_offset_and_flags = 0;
7445 /* Record the ECN info. */
7446 ipha->ipha_type_of_service &= 0xFC;
7447 ipha->ipha_type_of_service |= ecn_info;
7448
7449 /* Update the receive attributes */
7450 ira->ira_pktlen = packet_size;
7451 ira->ira_ip_hdr_length = IPH_HDR_LENGTH(ipha);
7452
7453 /* Reassembly is successful; set checksum information in packet */
7454 DB_CKSUM16(mp) = (uint16_t)sum_val;
7455 DB_CKSUMFLAGS(mp) = sum_flags;
7456 DB_CKSUMSTART(mp) = ira->ira_ip_hdr_length;
7457
7458 return (mp);
7459 }
7460
7461 /*
7462 * Pullup function that should be used for IP input in order to
7463 * ensure we do not loose the L2 source address; we need the l2 source
7464 * address for IP_RECVSLLA and for ndp_input.
7465 *
7466 * We return either NULL or b_rptr.
7467 */
7468 void *
7469 ip_pullup(mblk_t *mp, ssize_t len, ip_recv_attr_t *ira)
7470 {
7471 ill_t *ill = ira->ira_ill;
7472
7473 if (ip_rput_pullups++ == 0) {
7474 (void) mi_strlog(ill->ill_rq, 1, SL_ERROR|SL_TRACE,
7475 "ip_pullup: %s forced us to "
7476 " pullup pkt, hdr len %ld, hdr addr %p",
7477 ill->ill_name, len, (void *)mp->b_rptr);
7478 }
7479 if (!(ira->ira_flags & IRAF_L2SRC_SET))
7480 ip_setl2src(mp, ira, ira->ira_rill);
7481 ASSERT(ira->ira_flags & IRAF_L2SRC_SET);
7482 if (!pullupmsg(mp, len))
7483 return (NULL);
7484 else
7485 return (mp->b_rptr);
7486 }
7487
7488 /*
7489 * Make sure ira_l2src has an address. If we don't have one fill with zeros.
7490 * When called from the ULP ira_rill will be NULL hence the caller has to
7491 * pass in the ill.
7492 */
7493 /* ARGSUSED */
7494 void
7495 ip_setl2src(mblk_t *mp, ip_recv_attr_t *ira, ill_t *ill)
7496 {
7497 const uchar_t *addr;
7498 int alen;
7499
7500 if (ira->ira_flags & IRAF_L2SRC_SET)
7501 return;
7502
7503 ASSERT(ill != NULL);
7504 alen = ill->ill_phys_addr_length;
7505 ASSERT(alen <= sizeof (ira->ira_l2src));
7506 if (ira->ira_mhip != NULL &&
7507 (addr = ira->ira_mhip->mhi_saddr) != NULL) {
7508 bcopy(addr, ira->ira_l2src, alen);
7509 } else if ((ira->ira_flags & IRAF_L2SRC_LOOPBACK) &&
7510 (addr = ill->ill_phys_addr) != NULL) {
7511 bcopy(addr, ira->ira_l2src, alen);
7512 } else {
7513 bzero(ira->ira_l2src, alen);
7514 }
7515 ira->ira_flags |= IRAF_L2SRC_SET;
7516 }
7517
7518 /*
7519 * check ip header length and align it.
7520 */
7521 mblk_t *
7522 ip_check_and_align_header(mblk_t *mp, uint_t min_size, ip_recv_attr_t *ira)
7523 {
7524 ill_t *ill = ira->ira_ill;
7525 ssize_t len;
7526
7527 len = MBLKL(mp);
7528
7529 if (!OK_32PTR(mp->b_rptr))
7530 IP_STAT(ill->ill_ipst, ip_notaligned);
7531 else
7532 IP_STAT(ill->ill_ipst, ip_recv_pullup);
7533
7534 /* Guard against bogus device drivers */
7535 if (len < 0) {
7536 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7537 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7538 freemsg(mp);
7539 return (NULL);
7540 }
7541
7542 if (len == 0) {
7543 /* GLD sometimes sends up mblk with b_rptr == b_wptr! */
7544 mblk_t *mp1 = mp->b_cont;
7545
7546 if (!(ira->ira_flags & IRAF_L2SRC_SET))
7547 ip_setl2src(mp, ira, ira->ira_rill);
7548 ASSERT(ira->ira_flags & IRAF_L2SRC_SET);
7549
7550 freeb(mp);
7551 mp = mp1;
7552 if (mp == NULL)
7553 return (NULL);
7554
7555 if (OK_32PTR(mp->b_rptr) && MBLKL(mp) >= min_size)
7556 return (mp);
7557 }
7558 if (ip_pullup(mp, min_size, ira) == NULL) {
7559 if (msgdsize(mp) < min_size) {
7560 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7561 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7562 } else {
7563 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7564 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7565 }
7566 freemsg(mp);
7567 return (NULL);
7568 }
7569 return (mp);
7570 }
7571
7572 /*
7573 * Common code for IPv4 and IPv6 to check and pullup multi-mblks
7574 */
7575 mblk_t *
7576 ip_check_length(mblk_t *mp, uchar_t *rptr, ssize_t len, uint_t pkt_len,
7577 uint_t min_size, ip_recv_attr_t *ira)
7578 {
7579 ill_t *ill = ira->ira_ill;
7580
7581 /*
7582 * Make sure we have data length consistent
7583 * with the IP header.
7584 */
7585 if (mp->b_cont == NULL) {
7586 /* pkt_len is based on ipha_len, not the mblk length */
7587 if (pkt_len < min_size) {
7588 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7589 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7590 freemsg(mp);
7591 return (NULL);
7592 }
7593 if (len < 0) {
7594 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
7595 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
7596 freemsg(mp);
7597 return (NULL);
7598 }
7599 /* Drop any pad */
7600 mp->b_wptr = rptr + pkt_len;
7601 } else if ((len += msgdsize(mp->b_cont)) != 0) {
7602 ASSERT(pkt_len >= min_size);
7603 if (pkt_len < min_size) {
7604 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7605 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7606 freemsg(mp);
7607 return (NULL);
7608 }
7609 if (len < 0) {
7610 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
7611 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
7612 freemsg(mp);
7613 return (NULL);
7614 }
7615 /* Drop any pad */
7616 (void) adjmsg(mp, -len);
7617 /*
7618 * adjmsg may have freed an mblk from the chain, hence
7619 * invalidate any hw checksum here. This will force IP to
7620 * calculate the checksum in sw, but only for this packet.
7621 */
7622 DB_CKSUMFLAGS(mp) = 0;
7623 IP_STAT(ill->ill_ipst, ip_multimblk);
7624 }
7625 return (mp);
7626 }
7627
7628 /*
7629 * Check that the IPv4 opt_len is consistent with the packet and pullup
7630 * the options.
7631 */
7632 mblk_t *
7633 ip_check_optlen(mblk_t *mp, ipha_t *ipha, uint_t opt_len, uint_t pkt_len,
7634 ip_recv_attr_t *ira)
7635 {
7636 ill_t *ill = ira->ira_ill;
7637 ssize_t len;
7638
7639 /* Assume no IPv6 packets arrive over the IPv4 queue */
7640 if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
7641 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7642 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInWrongIPVersion);
7643 ip_drop_input("IPvN packet on IPv4 ill", mp, ill);
7644 freemsg(mp);
7645 return (NULL);
7646 }
7647
7648 if (opt_len > (15 - IP_SIMPLE_HDR_LENGTH_IN_WORDS)) {
7649 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7650 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7651 freemsg(mp);
7652 return (NULL);
7653 }
7654 /*
7655 * Recompute complete header length and make sure we
7656 * have access to all of it.
7657 */
7658 len = ((size_t)opt_len + IP_SIMPLE_HDR_LENGTH_IN_WORDS) << 2;
7659 if (len > (mp->b_wptr - mp->b_rptr)) {
7660 if (len > pkt_len) {
7661 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7662 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7663 freemsg(mp);
7664 return (NULL);
7665 }
7666 if (ip_pullup(mp, len, ira) == NULL) {
7667 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7668 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7669 freemsg(mp);
7670 return (NULL);
7671 }
7672 }
7673 return (mp);
7674 }
7675
7676 /*
7677 * Returns a new ire, or the same ire, or NULL.
7678 * If a different IRE is returned, then it is held; the caller
7679 * needs to release it.
7680 * In no case is there any hold/release on the ire argument.
7681 */
7682 ire_t *
7683 ip_check_multihome(void *addr, ire_t *ire, ill_t *ill)
7684 {
7685 ire_t *new_ire;
7686 ill_t *ire_ill;
7687 uint_t ifindex;
7688 ip_stack_t *ipst = ill->ill_ipst;
7689 boolean_t strict_check = B_FALSE;
7690
7691 /*
7692 * IPMP common case: if IRE and ILL are in the same group, there's no
7693 * issue (e.g. packet received on an underlying interface matched an
7694 * IRE_LOCAL on its associated group interface).
7695 */
7696 ASSERT(ire->ire_ill != NULL);
7697 if (IS_IN_SAME_ILLGRP(ill, ire->ire_ill))
7698 return (ire);
7699
7700 /*
7701 * Do another ire lookup here, using the ingress ill, to see if the
7702 * interface is in a usesrc group.
7703 * As long as the ills belong to the same group, we don't consider
7704 * them to be arriving on the wrong interface. Thus, if the switch
7705 * is doing inbound load spreading, we won't drop packets when the
7706 * ip*_strict_dst_multihoming switch is on.
7707 * We also need to check for IPIF_UNNUMBERED point2point interfaces
7708 * where the local address may not be unique. In this case we were
7709 * at the mercy of the initial ire lookup and the IRE_LOCAL it
7710 * actually returned. The new lookup, which is more specific, should
7711 * only find the IRE_LOCAL associated with the ingress ill if one
7712 * exists.
7713 */
7714 if (ire->ire_ipversion == IPV4_VERSION) {
7715 if (ipst->ips_ip_strict_dst_multihoming)
7716 strict_check = B_TRUE;
7717 new_ire = ire_ftable_lookup_v4(*((ipaddr_t *)addr), 0, 0,
7718 IRE_LOCAL, ill, ALL_ZONES, NULL,
7719 (MATCH_IRE_TYPE|MATCH_IRE_ILL), 0, ipst, NULL);
7720 } else {
7721 ASSERT(!IN6_IS_ADDR_MULTICAST((in6_addr_t *)addr));
7722 if (ipst->ips_ipv6_strict_dst_multihoming)
7723 strict_check = B_TRUE;
7724 new_ire = ire_ftable_lookup_v6((in6_addr_t *)addr, NULL, NULL,
7725 IRE_LOCAL, ill, ALL_ZONES, NULL,
7726 (MATCH_IRE_TYPE|MATCH_IRE_ILL), 0, ipst, NULL);
7727 }
7728 /*
7729 * If the same ire that was returned in ip_input() is found then this
7730 * is an indication that usesrc groups are in use. The packet
7731 * arrived on a different ill in the group than the one associated with
7732 * the destination address. If a different ire was found then the same
7733 * IP address must be hosted on multiple ills. This is possible with
7734 * unnumbered point2point interfaces. We switch to use this new ire in
7735 * order to have accurate interface statistics.
7736 */
7737 if (new_ire != NULL) {
7738 /* Note: held in one case but not the other? Caller handles */
7739 if (new_ire != ire)
7740 return (new_ire);
7741 /* Unchanged */
7742 ire_refrele(new_ire);
7743 return (ire);
7744 }
7745
7746 /*
7747 * Chase pointers once and store locally.
7748 */
7749 ASSERT(ire->ire_ill != NULL);
7750 ire_ill = ire->ire_ill;
7751 ifindex = ill->ill_usesrc_ifindex;
7752
7753 /*
7754 * Check if it's a legal address on the 'usesrc' interface.
7755 * For IPMP data addresses the IRE_LOCAL is the upper, hence we
7756 * can just check phyint_ifindex.
7757 */
7758 if (ifindex != 0 && ifindex == ire_ill->ill_phyint->phyint_ifindex) {
7759 return (ire);
7760 }
7761
7762 /*
7763 * If the ip*_strict_dst_multihoming switch is on then we can
7764 * only accept this packet if the interface is marked as routing.
7765 */
7766 if (!(strict_check))
7767 return (ire);
7768
7769 if ((ill->ill_flags & ire->ire_ill->ill_flags & ILLF_ROUTER) != 0) {
7770 return (ire);
7771 }
7772 return (NULL);
7773 }
7774
7775 /*
7776 * This function is used to construct a mac_header_info_s from a
7777 * DL_UNITDATA_IND message.
7778 * The address fields in the mhi structure points into the message,
7779 * thus the caller can't use those fields after freeing the message.
7780 *
7781 * We determine whether the packet received is a non-unicast packet
7782 * and in doing so, determine whether or not it is broadcast vs multicast.
7783 * For it to be a broadcast packet, we must have the appropriate mblk_t
7784 * hanging off the ill_t. If this is either not present or doesn't match
7785 * the destination mac address in the DL_UNITDATA_IND, the packet is deemed
7786 * to be multicast. Thus NICs that have no broadcast address (or no
7787 * capability for one, such as point to point links) cannot return as
7788 * the packet being broadcast.
7789 */
7790 void
7791 ip_dlur_to_mhi(ill_t *ill, mblk_t *mb, struct mac_header_info_s *mhip)
7792 {
7793 dl_unitdata_ind_t *ind = (dl_unitdata_ind_t *)mb->b_rptr;
7794 mblk_t *bmp;
7795 uint_t extra_offset;
7796
7797 bzero(mhip, sizeof (struct mac_header_info_s));
7798
7799 mhip->mhi_dsttype = MAC_ADDRTYPE_UNICAST;
7800
7801 if (ill->ill_sap_length < 0)
7802 extra_offset = 0;
7803 else
7804 extra_offset = ill->ill_sap_length;
7805
7806 mhip->mhi_daddr = (uchar_t *)ind + ind->dl_dest_addr_offset +
7807 extra_offset;
7808 mhip->mhi_saddr = (uchar_t *)ind + ind->dl_src_addr_offset +
7809 extra_offset;
7810
7811 if (!ind->dl_group_address)
7812 return;
7813
7814 /* Multicast or broadcast */
7815 mhip->mhi_dsttype = MAC_ADDRTYPE_MULTICAST;
7816
7817 if (ind->dl_dest_addr_offset > sizeof (*ind) &&
7818 ind->dl_dest_addr_offset + ind->dl_dest_addr_length < MBLKL(mb) &&
7819 (bmp = ill->ill_bcast_mp) != NULL) {
7820 dl_unitdata_req_t *dlur;
7821 uint8_t *bphys_addr;
7822
7823 dlur = (dl_unitdata_req_t *)bmp->b_rptr;
7824 bphys_addr = (uchar_t *)dlur + dlur->dl_dest_addr_offset +
7825 extra_offset;
7826
7827 if (bcmp(mhip->mhi_daddr, bphys_addr,
7828 ind->dl_dest_addr_length) == 0)
7829 mhip->mhi_dsttype = MAC_ADDRTYPE_BROADCAST;
7830 }
7831 }
7832
7833 /*
7834 * This function is used to construct a mac_header_info_s from a
7835 * M_DATA fastpath message from a DLPI driver.
7836 * The address fields in the mhi structure points into the message,
7837 * thus the caller can't use those fields after freeing the message.
7838 *
7839 * We determine whether the packet received is a non-unicast packet
7840 * and in doing so, determine whether or not it is broadcast vs multicast.
7841 * For it to be a broadcast packet, we must have the appropriate mblk_t
7842 * hanging off the ill_t. If this is either not present or doesn't match
7843 * the destination mac address in the DL_UNITDATA_IND, the packet is deemed
7844 * to be multicast. Thus NICs that have no broadcast address (or no
7845 * capability for one, such as point to point links) cannot return as
7846 * the packet being broadcast.
7847 */
7848 void
7849 ip_mdata_to_mhi(ill_t *ill, mblk_t *mp, struct mac_header_info_s *mhip)
7850 {
7851 mblk_t *bmp;
7852 struct ether_header *pether;
7853
7854 bzero(mhip, sizeof (struct mac_header_info_s));
7855
7856 mhip->mhi_dsttype = MAC_ADDRTYPE_UNICAST;
7857
7858 pether = (struct ether_header *)((char *)mp->b_rptr
7859 - sizeof (struct ether_header));
7860
7861 /*
7862 * Make sure the interface is an ethernet type, since we don't
7863 * know the header format for anything but Ethernet. Also make
7864 * sure we are pointing correctly above db_base.
7865 */
7866 if (ill->ill_type != IFT_ETHER)
7867 return;
7868
7869 retry:
7870 if ((uchar_t *)pether < mp->b_datap->db_base)
7871 return;
7872
7873 /* Is there a VLAN tag? */
7874 if (ill->ill_isv6) {
7875 if (pether->ether_type != htons(ETHERTYPE_IPV6)) {
7876 pether = (struct ether_header *)((char *)pether - 4);
7877 goto retry;
7878 }
7879 } else {
7880 if (pether->ether_type != htons(ETHERTYPE_IP)) {
7881 pether = (struct ether_header *)((char *)pether - 4);
7882 goto retry;
7883 }
7884 }
7885 mhip->mhi_daddr = (uchar_t *)&pether->ether_dhost;
7886 mhip->mhi_saddr = (uchar_t *)&pether->ether_shost;
7887
7888 if (!(mhip->mhi_daddr[0] & 0x01))
7889 return;
7890
7891 /* Multicast or broadcast */
7892 mhip->mhi_dsttype = MAC_ADDRTYPE_MULTICAST;
7893
7894 if ((bmp = ill->ill_bcast_mp) != NULL) {
7895 dl_unitdata_req_t *dlur;
7896 uint8_t *bphys_addr;
7897 uint_t addrlen;
7898
7899 dlur = (dl_unitdata_req_t *)bmp->b_rptr;
7900 addrlen = dlur->dl_dest_addr_length;
7901 if (ill->ill_sap_length < 0) {
7902 bphys_addr = (uchar_t *)dlur +
7903 dlur->dl_dest_addr_offset;
7904 addrlen += ill->ill_sap_length;
7905 } else {
7906 bphys_addr = (uchar_t *)dlur +
7907 dlur->dl_dest_addr_offset +
7908 ill->ill_sap_length;
7909 addrlen -= ill->ill_sap_length;
7910 }
7911 if (bcmp(mhip->mhi_daddr, bphys_addr, addrlen) == 0)
7912 mhip->mhi_dsttype = MAC_ADDRTYPE_BROADCAST;
7913 }
7914 }
7915
7916 /*
7917 * Handle anything but M_DATA messages
7918 * We see the DL_UNITDATA_IND which are part
7919 * of the data path, and also the other messages from the driver.
7920 */
7921 void
7922 ip_rput_notdata(ill_t *ill, mblk_t *mp)
7923 {
7924 mblk_t *first_mp;
7925 struct iocblk *iocp;
7926 struct mac_header_info_s mhi;
7927
7928 switch (DB_TYPE(mp)) {
7929 case M_PROTO:
7930 case M_PCPROTO: {
7931 if (((dl_unitdata_ind_t *)mp->b_rptr)->dl_primitive !=
7932 DL_UNITDATA_IND) {
7933 /* Go handle anything other than data elsewhere. */
7934 ip_rput_dlpi(ill, mp);
7935 return;
7936 }
7937
7938 first_mp = mp;
7939 mp = first_mp->b_cont;
7940 first_mp->b_cont = NULL;
7941
7942 if (mp == NULL) {
7943 freeb(first_mp);
7944 return;
7945 }
7946 ip_dlur_to_mhi(ill, first_mp, &mhi);
7947 if (ill->ill_isv6)
7948 ip_input_v6(ill, NULL, mp, &mhi);
7949 else
7950 ip_input(ill, NULL, mp, &mhi);
7951
7952 /* Ditch the DLPI header. */
7953 freeb(first_mp);
7954 return;
7955 }
7956 case M_IOCACK:
7957 iocp = (struct iocblk *)mp->b_rptr;
7958 switch (iocp->ioc_cmd) {
7959 case DL_IOC_HDR_INFO:
7960 ill_fastpath_ack(ill, mp);
7961 return;
7962 default:
7963 putnext(ill->ill_rq, mp);
7964 return;
7965 }
7966 /* FALLTHRU */
7967 case M_ERROR:
7968 case M_HANGUP:
7969 mutex_enter(&ill->ill_lock);
7970 if (ill->ill_state_flags & ILL_CONDEMNED) {
7971 mutex_exit(&ill->ill_lock);
7972 freemsg(mp);
7973 return;
7974 }
7975 ill_refhold_locked(ill);
7976 mutex_exit(&ill->ill_lock);
7977 qwriter_ip(ill, ill->ill_rq, mp, ip_rput_other, CUR_OP,
7978 B_FALSE);
7979 return;
7980 case M_CTL:
7981 putnext(ill->ill_rq, mp);
7982 return;
7983 case M_IOCNAK:
7984 ip1dbg(("got iocnak "));
7985 iocp = (struct iocblk *)mp->b_rptr;
7986 switch (iocp->ioc_cmd) {
7987 case DL_IOC_HDR_INFO:
7988 ip_rput_other(NULL, ill->ill_rq, mp, NULL);
7989 return;
7990 default:
7991 break;
7992 }
7993 /* FALLTHRU */
7994 default:
7995 putnext(ill->ill_rq, mp);
7996 return;
7997 }
7998 }
7999
8000 /* Read side put procedure. Packets coming from the wire arrive here. */
8001 void
8002 ip_rput(queue_t *q, mblk_t *mp)
8003 {
8004 ill_t *ill;
8005 union DL_primitives *dl;
8006
8007 ill = (ill_t *)q->q_ptr;
8008
8009 if (ill->ill_state_flags & (ILL_CONDEMNED | ILL_LL_SUBNET_PENDING)) {
8010 /*
8011 * If things are opening or closing, only accept high-priority
8012 * DLPI messages. (On open ill->ill_ipif has not yet been
8013 * created; on close, things hanging off the ill may have been
8014 * freed already.)
8015 */
8016 dl = (union DL_primitives *)mp->b_rptr;
8017 if (DB_TYPE(mp) != M_PCPROTO ||
8018 dl->dl_primitive == DL_UNITDATA_IND) {
8019 inet_freemsg(mp);
8020 return;
8021 }
8022 }
8023 if (DB_TYPE(mp) == M_DATA) {
8024 struct mac_header_info_s mhi;
8025
8026 ip_mdata_to_mhi(ill, mp, &mhi);
8027 ip_input(ill, NULL, mp, &mhi);
8028 } else {
8029 ip_rput_notdata(ill, mp);
8030 }
8031 }
8032
8033 /*
8034 * Move the information to a copy.
8035 */
8036 mblk_t *
8037 ip_fix_dbref(mblk_t *mp, ip_recv_attr_t *ira)
8038 {
8039 mblk_t *mp1;
8040 ill_t *ill = ira->ira_ill;
8041 ip_stack_t *ipst = ill->ill_ipst;
8042
8043 IP_STAT(ipst, ip_db_ref);
8044
8045 /* Make sure we have ira_l2src before we loose the original mblk */
8046 if (!(ira->ira_flags & IRAF_L2SRC_SET))
8047 ip_setl2src(mp, ira, ira->ira_rill);
8048
8049 mp1 = copymsg(mp);
8050 if (mp1 == NULL) {
8051 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
8052 ip_drop_input("ipIfStatsInDiscards", mp, ill);
8053 freemsg(mp);
8054 return (NULL);
8055 }
8056 /* preserve the hardware checksum flags and data, if present */
8057 if (DB_CKSUMFLAGS(mp) != 0) {
8058 DB_CKSUMFLAGS(mp1) = DB_CKSUMFLAGS(mp);
8059 DB_CKSUMSTART(mp1) = DB_CKSUMSTART(mp);
8060 DB_CKSUMSTUFF(mp1) = DB_CKSUMSTUFF(mp);
8061 DB_CKSUMEND(mp1) = DB_CKSUMEND(mp);
8062 DB_CKSUM16(mp1) = DB_CKSUM16(mp);
8063 }
8064 freemsg(mp);
8065 return (mp1);
8066 }
8067
8068 static void
8069 ip_dlpi_error(ill_t *ill, t_uscalar_t prim, t_uscalar_t dl_err,
8070 t_uscalar_t err)
8071 {
8072 if (dl_err == DL_SYSERR) {
8073 (void) mi_strlog(ill->ill_rq, 1, SL_CONSOLE|SL_ERROR|SL_TRACE,
8074 "%s: %s failed: DL_SYSERR (errno %u)\n",
8075 ill->ill_name, dl_primstr(prim), err);
8076 return;
8077 }
8078
8079 (void) mi_strlog(ill->ill_rq, 1, SL_CONSOLE|SL_ERROR|SL_TRACE,
8080 "%s: %s failed: %s\n", ill->ill_name, dl_primstr(prim),
8081 dl_errstr(dl_err));
8082 }
8083
8084 /*
8085 * ip_rput_dlpi is called by ip_rput to handle all DLPI messages other
8086 * than DL_UNITDATA_IND messages. If we need to process this message
8087 * exclusively, we call qwriter_ip, in which case we also need to call
8088 * ill_refhold before that, since qwriter_ip does an ill_refrele.
8089 */
8090 void
8091 ip_rput_dlpi(ill_t *ill, mblk_t *mp)
8092 {
8093 dl_ok_ack_t *dloa = (dl_ok_ack_t *)mp->b_rptr;
8094 dl_error_ack_t *dlea = (dl_error_ack_t *)dloa;
8095 queue_t *q = ill->ill_rq;
8096 t_uscalar_t prim = dloa->dl_primitive;
8097 t_uscalar_t reqprim = DL_PRIM_INVAL;
8098
8099 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi",
8100 char *, dl_primstr(prim), ill_t *, ill);
8101 ip1dbg(("ip_rput_dlpi"));
8102
8103 /*
8104 * If we received an ACK but didn't send a request for it, then it
8105 * can't be part of any pending operation; discard up-front.
8106 */
8107 switch (prim) {
8108 case DL_ERROR_ACK:
8109 reqprim = dlea->dl_error_primitive;
8110 ip2dbg(("ip_rput_dlpi(%s): DL_ERROR_ACK for %s (0x%x): %s "
8111 "(0x%x), unix %u\n", ill->ill_name, dl_primstr(reqprim),
8112 reqprim, dl_errstr(dlea->dl_errno), dlea->dl_errno,
8113 dlea->dl_unix_errno));
8114 break;
8115 case DL_OK_ACK:
8116 reqprim = dloa->dl_correct_primitive;
8117 break;
8118 case DL_INFO_ACK:
8119 reqprim = DL_INFO_REQ;
8120 break;
8121 case DL_BIND_ACK:
8122 reqprim = DL_BIND_REQ;
8123 break;
8124 case DL_PHYS_ADDR_ACK:
8125 reqprim = DL_PHYS_ADDR_REQ;
8126 break;
8127 case DL_NOTIFY_ACK:
8128 reqprim = DL_NOTIFY_REQ;
8129 break;
8130 case DL_CAPABILITY_ACK:
8131 reqprim = DL_CAPABILITY_REQ;
8132 break;
8133 }
8134
8135 if (prim != DL_NOTIFY_IND) {
8136 if (reqprim == DL_PRIM_INVAL ||
8137 !ill_dlpi_pending(ill, reqprim)) {
8138 /* Not a DLPI message we support or expected */
8139 freemsg(mp);
8140 return;
8141 }
8142 ip1dbg(("ip_rput: received %s for %s\n", dl_primstr(prim),
8143 dl_primstr(reqprim)));
8144 }
8145
8146 switch (reqprim) {
8147 case DL_UNBIND_REQ:
8148 /*
8149 * NOTE: we mark the unbind as complete even if we got a
8150 * DL_ERROR_ACK, since there's not much else we can do.
8151 */
8152 mutex_enter(&ill->ill_lock);
8153 ill->ill_state_flags &= ~ILL_DL_UNBIND_IN_PROGRESS;
8154 cv_signal(&ill->ill_cv);
8155 mutex_exit(&ill->ill_lock);
8156 break;
8157
8158 case DL_ENABMULTI_REQ:
8159 if (prim == DL_OK_ACK) {
8160 if (ill->ill_dlpi_multicast_state == IDS_INPROGRESS)
8161 ill->ill_dlpi_multicast_state = IDS_OK;
8162 }
8163 break;
8164 }
8165
8166 /*
8167 * The message is one we're waiting for (or DL_NOTIFY_IND), but we
8168 * need to become writer to continue to process it. Because an
8169 * exclusive operation doesn't complete until replies to all queued
8170 * DLPI messages have been received, we know we're in the middle of an
8171 * exclusive operation and pass CUR_OP (except for DL_NOTIFY_IND).
8172 *
8173 * As required by qwriter_ip(), we refhold the ill; it will refrele.
8174 * Since this is on the ill stream we unconditionally bump up the
8175 * refcount without doing ILL_CAN_LOOKUP().
8176 */
8177 ill_refhold(ill);
8178 if (prim == DL_NOTIFY_IND)
8179 qwriter_ip(ill, q, mp, ip_rput_dlpi_writer, NEW_OP, B_FALSE);
8180 else
8181 qwriter_ip(ill, q, mp, ip_rput_dlpi_writer, CUR_OP, B_FALSE);
8182 }
8183
8184 /*
8185 * Handling of DLPI messages that require exclusive access to the ipsq.
8186 *
8187 * Need to do ipsq_pending_mp_get on ioctl completion, which could
8188 * happen here. (along with mi_copy_done)
8189 */
8190 /* ARGSUSED */
8191 static void
8192 ip_rput_dlpi_writer(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
8193 {
8194 dl_ok_ack_t *dloa = (dl_ok_ack_t *)mp->b_rptr;
8195 dl_error_ack_t *dlea = (dl_error_ack_t *)dloa;
8196 int err = 0;
8197 ill_t *ill = (ill_t *)q->q_ptr;
8198 ipif_t *ipif = NULL;
8199 mblk_t *mp1 = NULL;
8200 conn_t *connp = NULL;
8201 t_uscalar_t paddrreq;
8202 mblk_t *mp_hw;
8203 boolean_t success;
8204 boolean_t ioctl_aborted = B_FALSE;
8205 boolean_t log = B_TRUE;
8206
8207 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer",
8208 char *, dl_primstr(dloa->dl_primitive), ill_t *, ill);
8209
8210 ip1dbg(("ip_rput_dlpi_writer .."));
8211 ASSERT(ipsq->ipsq_xop == ill->ill_phyint->phyint_ipsq->ipsq_xop);
8212 ASSERT(IAM_WRITER_ILL(ill));
8213
8214 ipif = ipsq->ipsq_xop->ipx_pending_ipif;
8215 /*
8216 * The current ioctl could have been aborted by the user and a new
8217 * ioctl to bring up another ill could have started. We could still
8218 * get a response from the driver later.
8219 */
8220 if (ipif != NULL && ipif->ipif_ill != ill)
8221 ioctl_aborted = B_TRUE;
8222
8223 switch (dloa->dl_primitive) {
8224 case DL_ERROR_ACK:
8225 ip1dbg(("ip_rput_dlpi_writer: got DL_ERROR_ACK for %s\n",
8226 dl_primstr(dlea->dl_error_primitive)));
8227
8228 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer error",
8229 char *, dl_primstr(dlea->dl_error_primitive),
8230 ill_t *, ill);
8231
8232 switch (dlea->dl_error_primitive) {
8233 case DL_DISABMULTI_REQ:
8234 ill_dlpi_done(ill, dlea->dl_error_primitive);
8235 break;
8236 case DL_PROMISCON_REQ:
8237 case DL_PROMISCOFF_REQ:
8238 case DL_UNBIND_REQ:
8239 case DL_ATTACH_REQ:
8240 case DL_INFO_REQ:
8241 ill_dlpi_done(ill, dlea->dl_error_primitive);
8242 break;
8243 case DL_NOTIFY_REQ:
8244 ill_dlpi_done(ill, DL_NOTIFY_REQ);
8245 log = B_FALSE;
8246 break;
8247 case DL_PHYS_ADDR_REQ:
8248 /*
8249 * For IPv6 only, there are two additional
8250 * phys_addr_req's sent to the driver to get the
8251 * IPv6 token and lla. This allows IP to acquire
8252 * the hardware address format for a given interface
8253 * without having built in knowledge of the hardware
8254 * address. ill_phys_addr_pend keeps track of the last
8255 * DL_PAR sent so we know which response we are
8256 * dealing with. ill_dlpi_done will update
8257 * ill_phys_addr_pend when it sends the next req.
8258 * We don't complete the IOCTL until all three DL_PARs
8259 * have been attempted, so set *_len to 0 and break.
8260 */
8261 paddrreq = ill->ill_phys_addr_pend;
8262 ill_dlpi_done(ill, DL_PHYS_ADDR_REQ);
8263 if (paddrreq == DL_IPV6_TOKEN) {
8264 ill->ill_token_length = 0;
8265 log = B_FALSE;
8266 break;
8267 } else if (paddrreq == DL_IPV6_LINK_LAYER_ADDR) {
8268 ill->ill_nd_lla_len = 0;
8269 log = B_FALSE;
8270 break;
8271 }
8272 /*
8273 * Something went wrong with the DL_PHYS_ADDR_REQ.
8274 * We presumably have an IOCTL hanging out waiting
8275 * for completion. Find it and complete the IOCTL
8276 * with the error noted.
8277 * However, ill_dl_phys was called on an ill queue
8278 * (from SIOCSLIFNAME), thus conn_pending_ill is not
8279 * set. But the ioctl is known to be pending on ill_wq.
8280 */
8281 if (!ill->ill_ifname_pending)
8282 break;
8283 ill->ill_ifname_pending = 0;
8284 if (!ioctl_aborted)
8285 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8286 if (mp1 != NULL) {
8287 /*
8288 * This operation (SIOCSLIFNAME) must have
8289 * happened on the ill. Assert there is no conn
8290 */
8291 ASSERT(connp == NULL);
8292 q = ill->ill_wq;
8293 }
8294 break;
8295 case DL_BIND_REQ:
8296 ill_dlpi_done(ill, DL_BIND_REQ);
8297 if (ill->ill_ifname_pending)
8298 break;
8299 mutex_enter(&ill->ill_lock);
8300 ill->ill_state_flags &= ~ILL_DOWN_IN_PROGRESS;
8301 mutex_exit(&ill->ill_lock);
8302 /*
8303 * Something went wrong with the bind. We presumably
8304 * have an IOCTL hanging out waiting for completion.
8305 * Find it, take down the interface that was coming
8306 * up, and complete the IOCTL with the error noted.
8307 */
8308 if (!ioctl_aborted)
8309 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8310 if (mp1 != NULL) {
8311 /*
8312 * This might be a result of a DL_NOTE_REPLUMB
8313 * notification. In that case, connp is NULL.
8314 */
8315 if (connp != NULL)
8316 q = CONNP_TO_WQ(connp);
8317
8318 (void) ipif_down(ipif, NULL, NULL);
8319 /* error is set below the switch */
8320 }
8321 break;
8322 case DL_ENABMULTI_REQ:
8323 ill_dlpi_done(ill, DL_ENABMULTI_REQ);
8324
8325 if (ill->ill_dlpi_multicast_state == IDS_INPROGRESS)
8326 ill->ill_dlpi_multicast_state = IDS_FAILED;
8327 if (ill->ill_dlpi_multicast_state == IDS_FAILED) {
8328
8329 printf("ip: joining multicasts failed (%d)"
8330 " on %s - will use link layer "
8331 "broadcasts for multicast\n",
8332 dlea->dl_errno, ill->ill_name);
8333
8334 /*
8335 * Set up for multi_bcast; We are the
8336 * writer, so ok to access ill->ill_ipif
8337 * without any lock.
8338 */
8339 mutex_enter(&ill->ill_phyint->phyint_lock);
8340 ill->ill_phyint->phyint_flags |=
8341 PHYI_MULTI_BCAST;
8342 mutex_exit(&ill->ill_phyint->phyint_lock);
8343
8344 }
8345 freemsg(mp); /* Don't want to pass this up */
8346 return;
8347 case DL_CAPABILITY_REQ:
8348 ip1dbg(("ip_rput_dlpi_writer: got DL_ERROR_ACK for "
8349 "DL_CAPABILITY REQ\n"));
8350 if (ill->ill_dlpi_capab_state == IDCS_PROBE_SENT)
8351 ill->ill_dlpi_capab_state = IDCS_FAILED;
8352 ill_capability_done(ill);
8353 freemsg(mp);
8354 return;
8355 }
8356 /*
8357 * Note the error for IOCTL completion (mp1 is set when
8358 * ready to complete ioctl). If ill_ifname_pending_err is
8359 * set, an error occured during plumbing (ill_ifname_pending),
8360 * so we want to report that error.
8361 *
8362 * NOTE: there are two addtional DL_PHYS_ADDR_REQ's
8363 * (DL_IPV6_TOKEN and DL_IPV6_LINK_LAYER_ADDR) that are
8364 * expected to get errack'd if the driver doesn't support
8365 * these flags (e.g. ethernet). log will be set to B_FALSE
8366 * if these error conditions are encountered.
8367 */
8368 if (mp1 != NULL) {
8369 if (ill->ill_ifname_pending_err != 0) {
8370 err = ill->ill_ifname_pending_err;
8371 ill->ill_ifname_pending_err = 0;
8372 } else {
8373 err = dlea->dl_unix_errno ?
8374 dlea->dl_unix_errno : ENXIO;
8375 }
8376 /*
8377 * If we're plumbing an interface and an error hasn't already
8378 * been saved, set ill_ifname_pending_err to the error passed
8379 * up. Ignore the error if log is B_FALSE (see comment above).
8380 */
8381 } else if (log && ill->ill_ifname_pending &&
8382 ill->ill_ifname_pending_err == 0) {
8383 ill->ill_ifname_pending_err = dlea->dl_unix_errno ?
8384 dlea->dl_unix_errno : ENXIO;
8385 }
8386
8387 if (log)
8388 ip_dlpi_error(ill, dlea->dl_error_primitive,
8389 dlea->dl_errno, dlea->dl_unix_errno);
8390 break;
8391 case DL_CAPABILITY_ACK:
8392 ill_capability_ack(ill, mp);
8393 /*
8394 * The message has been handed off to ill_capability_ack
8395 * and must not be freed below
8396 */
8397 mp = NULL;
8398 break;
8399
8400 case DL_INFO_ACK:
8401 /* Call a routine to handle this one. */
8402 ill_dlpi_done(ill, DL_INFO_REQ);
8403 ip_ll_subnet_defaults(ill, mp);
8404 ASSERT(!MUTEX_HELD(&ill->ill_phyint->phyint_ipsq->ipsq_lock));
8405 return;
8406 case DL_BIND_ACK:
8407 /*
8408 * We should have an IOCTL waiting on this unless
8409 * sent by ill_dl_phys, in which case just return
8410 */
8411 ill_dlpi_done(ill, DL_BIND_REQ);
8412
8413 if (ill->ill_ifname_pending) {
8414 DTRACE_PROBE2(ip__rput__dlpi__ifname__pending,
8415 ill_t *, ill, mblk_t *, mp);
8416 break;
8417 }
8418 mutex_enter(&ill->ill_lock);
8419 ill->ill_dl_up = 1;
8420 ill->ill_state_flags &= ~ILL_DOWN_IN_PROGRESS;
8421 mutex_exit(&ill->ill_lock);
8422
8423 if (!ioctl_aborted)
8424 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8425 if (mp1 == NULL) {
8426 DTRACE_PROBE1(ip__rput__dlpi__no__mblk, ill_t *, ill);
8427 break;
8428 }
8429 /*
8430 * mp1 was added by ill_dl_up(). if that is a result of
8431 * a DL_NOTE_REPLUMB notification, connp could be NULL.
8432 */
8433 if (connp != NULL)
8434 q = CONNP_TO_WQ(connp);
8435 /*
8436 * We are exclusive. So nothing can change even after
8437 * we get the pending mp.
8438 */
8439 ip1dbg(("ip_rput_dlpi: bind_ack %s\n", ill->ill_name));
8440 DTRACE_PROBE1(ip__rput__dlpi__bind__ack, ill_t *, ill);
8441 ill_nic_event_dispatch(ill, 0, NE_UP, NULL, 0);
8442
8443 /*
8444 * Now bring up the resolver; when that is complete, we'll
8445 * create IREs. Note that we intentionally mirror what
8446 * ipif_up() would have done, because we got here by way of
8447 * ill_dl_up(), which stopped ipif_up()'s processing.
8448 */
8449 if (ill->ill_isv6) {
8450 /*
8451 * v6 interfaces.
8452 * Unlike ARP which has to do another bind
8453 * and attach, once we get here we are
8454 * done with NDP
8455 */
8456 (void) ipif_resolver_up(ipif, Res_act_initial);
8457 if ((err = ipif_ndp_up(ipif, B_TRUE)) == 0)
8458 err = ipif_up_done_v6(ipif);
8459 } else if (ill->ill_net_type == IRE_IF_RESOLVER) {
8460 /*
8461 * ARP and other v4 external resolvers.
8462 * Leave the pending mblk intact so that
8463 * the ioctl completes in ip_rput().
8464 */
8465 if (connp != NULL)
8466 mutex_enter(&connp->conn_lock);
8467 mutex_enter(&ill->ill_lock);
8468 success = ipsq_pending_mp_add(connp, ipif, q, mp1, 0);
8469 mutex_exit(&ill->ill_lock);
8470 if (connp != NULL)
8471 mutex_exit(&connp->conn_lock);
8472 if (success) {
8473 err = ipif_resolver_up(ipif, Res_act_initial);
8474 if (err == EINPROGRESS) {
8475 freemsg(mp);
8476 return;
8477 }
8478 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8479 } else {
8480 /* The conn has started closing */
8481 err = EINTR;
8482 }
8483 } else {
8484 /*
8485 * This one is complete. Reply to pending ioctl.
8486 */
8487 (void) ipif_resolver_up(ipif, Res_act_initial);
8488 err = ipif_up_done(ipif);
8489 }
8490
8491 if ((err == 0) && (ill->ill_up_ipifs)) {
8492 err = ill_up_ipifs(ill, q, mp1);
8493 if (err == EINPROGRESS) {
8494 freemsg(mp);
8495 return;
8496 }
8497 }
8498
8499 /*
8500 * If we have a moved ipif to bring up, and everything has
8501 * succeeded to this point, bring it up on the IPMP ill.
8502 * Otherwise, leave it down -- the admin can try to bring it
8503 * up by hand if need be.
8504 */
8505 if (ill->ill_move_ipif != NULL) {
8506 if (err != 0) {
8507 ill->ill_move_ipif = NULL;
8508 } else {
8509 ipif = ill->ill_move_ipif;
8510 ill->ill_move_ipif = NULL;
8511 err = ipif_up(ipif, q, mp1);
8512 if (err == EINPROGRESS) {
8513 freemsg(mp);
8514 return;
8515 }
8516 }
8517 }
8518 break;
8519
8520 case DL_NOTIFY_IND: {
8521 dl_notify_ind_t *notify = (dl_notify_ind_t *)mp->b_rptr;
8522 uint_t orig_mtu, orig_mc_mtu;
8523
8524 switch (notify->dl_notification) {
8525 case DL_NOTE_PHYS_ADDR:
8526 err = ill_set_phys_addr(ill, mp);
8527 break;
8528
8529 case DL_NOTE_REPLUMB:
8530 /*
8531 * Directly return after calling ill_replumb().
8532 * Note that we should not free mp as it is reused
8533 * in the ill_replumb() function.
8534 */
8535 err = ill_replumb(ill, mp);
8536 return;
8537
8538 case DL_NOTE_FASTPATH_FLUSH:
8539 nce_flush(ill, B_FALSE);
8540 break;
8541
8542 case DL_NOTE_SDU_SIZE:
8543 case DL_NOTE_SDU_SIZE2:
8544 /*
8545 * The dce and fragmentation code can cope with
8546 * this changing while packets are being sent.
8547 * When packets are sent ip_output will discover
8548 * a change.
8549 *
8550 * Change the MTU size of the interface.
8551 */
8552 mutex_enter(&ill->ill_lock);
8553 orig_mtu = ill->ill_mtu;
8554 orig_mc_mtu = ill->ill_mc_mtu;
8555 switch (notify->dl_notification) {
8556 case DL_NOTE_SDU_SIZE:
8557 ill->ill_current_frag =
8558 (uint_t)notify->dl_data;
8559 ill->ill_mc_mtu = (uint_t)notify->dl_data;
8560 break;
8561 case DL_NOTE_SDU_SIZE2:
8562 ill->ill_current_frag =
8563 (uint_t)notify->dl_data1;
8564 ill->ill_mc_mtu = (uint_t)notify->dl_data2;
8565 break;
8566 }
8567 if (ill->ill_current_frag > ill->ill_max_frag)
8568 ill->ill_max_frag = ill->ill_current_frag;
8569
8570 if (!(ill->ill_flags & ILLF_FIXEDMTU)) {
8571 ill->ill_mtu = ill->ill_current_frag;
8572
8573 /*
8574 * If ill_user_mtu was set (via
8575 * SIOCSLIFLNKINFO), clamp ill_mtu at it.
8576 */
8577 if (ill->ill_user_mtu != 0 &&
8578 ill->ill_user_mtu < ill->ill_mtu)
8579 ill->ill_mtu = ill->ill_user_mtu;
8580
8581 if (ill->ill_user_mtu != 0 &&
8582 ill->ill_user_mtu < ill->ill_mc_mtu)
8583 ill->ill_mc_mtu = ill->ill_user_mtu;
8584
8585 if (ill->ill_isv6) {
8586 if (ill->ill_mtu < IPV6_MIN_MTU)
8587 ill->ill_mtu = IPV6_MIN_MTU;
8588 if (ill->ill_mc_mtu < IPV6_MIN_MTU)
8589 ill->ill_mc_mtu = IPV6_MIN_MTU;
8590 } else {
8591 if (ill->ill_mtu < IP_MIN_MTU)
8592 ill->ill_mtu = IP_MIN_MTU;
8593 if (ill->ill_mc_mtu < IP_MIN_MTU)
8594 ill->ill_mc_mtu = IP_MIN_MTU;
8595 }
8596 } else if (ill->ill_mc_mtu > ill->ill_mtu) {
8597 ill->ill_mc_mtu = ill->ill_mtu;
8598 }
8599
8600 mutex_exit(&ill->ill_lock);
8601 /*
8602 * Make sure all dce_generation checks find out
8603 * that ill_mtu/ill_mc_mtu has changed.
8604 */
8605 if (orig_mtu != ill->ill_mtu ||
8606 orig_mc_mtu != ill->ill_mc_mtu) {
8607 dce_increment_all_generations(ill->ill_isv6,
8608 ill->ill_ipst);
8609 }
8610
8611 /*
8612 * Refresh IPMP meta-interface MTU if necessary.
8613 */
8614 if (IS_UNDER_IPMP(ill))
8615 ipmp_illgrp_refresh_mtu(ill->ill_grp);
8616 break;
8617
8618 case DL_NOTE_LINK_UP:
8619 case DL_NOTE_LINK_DOWN: {
8620 /*
8621 * We are writer. ill / phyint / ipsq assocs stable.
8622 * The RUNNING flag reflects the state of the link.
8623 */
8624 phyint_t *phyint = ill->ill_phyint;
8625 uint64_t new_phyint_flags;
8626 boolean_t changed = B_FALSE;
8627 boolean_t went_up;
8628
8629 went_up = notify->dl_notification == DL_NOTE_LINK_UP;
8630 mutex_enter(&phyint->phyint_lock);
8631
8632 new_phyint_flags = went_up ?
8633 phyint->phyint_flags | PHYI_RUNNING :
8634 phyint->phyint_flags & ~PHYI_RUNNING;
8635
8636 if (IS_IPMP(ill)) {
8637 new_phyint_flags = went_up ?
8638 new_phyint_flags & ~PHYI_FAILED :
8639 new_phyint_flags | PHYI_FAILED;
8640 }
8641
8642 if (new_phyint_flags != phyint->phyint_flags) {
8643 phyint->phyint_flags = new_phyint_flags;
8644 changed = B_TRUE;
8645 }
8646 mutex_exit(&phyint->phyint_lock);
8647 /*
8648 * ill_restart_dad handles the DAD restart and routing
8649 * socket notification logic.
8650 */
8651 if (changed) {
8652 ill_restart_dad(phyint->phyint_illv4, went_up);
8653 ill_restart_dad(phyint->phyint_illv6, went_up);
8654 }
8655 break;
8656 }
8657 case DL_NOTE_PROMISC_ON_PHYS: {
8658 phyint_t *phyint = ill->ill_phyint;
8659
8660 mutex_enter(&phyint->phyint_lock);
8661 phyint->phyint_flags |= PHYI_PROMISC;
8662 mutex_exit(&phyint->phyint_lock);
8663 break;
8664 }
8665 case DL_NOTE_PROMISC_OFF_PHYS: {
8666 phyint_t *phyint = ill->ill_phyint;
8667
8668 mutex_enter(&phyint->phyint_lock);
8669 phyint->phyint_flags &= ~PHYI_PROMISC;
8670 mutex_exit(&phyint->phyint_lock);
8671 break;
8672 }
8673 case DL_NOTE_CAPAB_RENEG:
8674 /*
8675 * Something changed on the driver side.
8676 * It wants us to renegotiate the capabilities
8677 * on this ill. One possible cause is the aggregation
8678 * interface under us where a port got added or
8679 * went away.
8680 *
8681 * If the capability negotiation is already done
8682 * or is in progress, reset the capabilities and
8683 * mark the ill's ill_capab_reneg to be B_TRUE,
8684 * so that when the ack comes back, we can start
8685 * the renegotiation process.
8686 *
8687 * Note that if ill_capab_reneg is already B_TRUE
8688 * (ill_dlpi_capab_state is IDS_UNKNOWN in this case),
8689 * the capability resetting request has been sent
8690 * and the renegotiation has not been started yet;
8691 * nothing needs to be done in this case.
8692 */
8693 ipsq_current_start(ipsq, ill->ill_ipif, 0);
8694 ill_capability_reset(ill, B_TRUE);
8695 ipsq_current_finish(ipsq);
8696 break;
8697
8698 case DL_NOTE_ALLOWED_IPS:
8699 ill_set_allowed_ips(ill, mp);
8700 break;
8701 default:
8702 ip0dbg(("ip_rput_dlpi_writer: unknown notification "
8703 "type 0x%x for DL_NOTIFY_IND\n",
8704 notify->dl_notification));
8705 break;
8706 }
8707
8708 /*
8709 * As this is an asynchronous operation, we
8710 * should not call ill_dlpi_done
8711 */
8712 break;
8713 }
8714 case DL_NOTIFY_ACK: {
8715 dl_notify_ack_t *noteack = (dl_notify_ack_t *)mp->b_rptr;
8716
8717 if (noteack->dl_notifications & DL_NOTE_LINK_UP)
8718 ill->ill_note_link = 1;
8719 ill_dlpi_done(ill, DL_NOTIFY_REQ);
8720 break;
8721 }
8722 case DL_PHYS_ADDR_ACK: {
8723 /*
8724 * As part of plumbing the interface via SIOCSLIFNAME,
8725 * ill_dl_phys() will queue a series of DL_PHYS_ADDR_REQs,
8726 * whose answers we receive here. As each answer is received,
8727 * we call ill_dlpi_done() to dispatch the next request as
8728 * we're processing the current one. Once all answers have
8729 * been received, we use ipsq_pending_mp_get() to dequeue the
8730 * outstanding IOCTL and reply to it. (Because ill_dl_phys()
8731 * is invoked from an ill queue, conn_oper_pending_ill is not
8732 * available, but we know the ioctl is pending on ill_wq.)
8733 */
8734 uint_t paddrlen, paddroff;
8735 uint8_t *addr;
8736
8737 paddrreq = ill->ill_phys_addr_pend;
8738 paddrlen = ((dl_phys_addr_ack_t *)mp->b_rptr)->dl_addr_length;
8739 paddroff = ((dl_phys_addr_ack_t *)mp->b_rptr)->dl_addr_offset;
8740 addr = mp->b_rptr + paddroff;
8741
8742 ill_dlpi_done(ill, DL_PHYS_ADDR_REQ);
8743 if (paddrreq == DL_IPV6_TOKEN) {
8744 /*
8745 * bcopy to low-order bits of ill_token
8746 *
8747 * XXX Temporary hack - currently, all known tokens
8748 * are 64 bits, so I'll cheat for the moment.
8749 */
8750 bcopy(addr, &ill->ill_token.s6_addr32[2], paddrlen);
8751 ill->ill_token_length = paddrlen;
8752 break;
8753 } else if (paddrreq == DL_IPV6_LINK_LAYER_ADDR) {
8754 ASSERT(ill->ill_nd_lla_mp == NULL);
8755 ill_set_ndmp(ill, mp, paddroff, paddrlen);
8756 mp = NULL;
8757 break;
8758 } else if (paddrreq == DL_CURR_DEST_ADDR) {
8759 ASSERT(ill->ill_dest_addr_mp == NULL);
8760 ill->ill_dest_addr_mp = mp;
8761 ill->ill_dest_addr = addr;
8762 mp = NULL;
8763 if (ill->ill_isv6) {
8764 ill_setdesttoken(ill);
8765 ipif_setdestlinklocal(ill->ill_ipif);
8766 }
8767 break;
8768 }
8769
8770 ASSERT(paddrreq == DL_CURR_PHYS_ADDR);
8771 ASSERT(ill->ill_phys_addr_mp == NULL);
8772 if (!ill->ill_ifname_pending)
8773 break;
8774 ill->ill_ifname_pending = 0;
8775 if (!ioctl_aborted)
8776 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8777 if (mp1 != NULL) {
8778 ASSERT(connp == NULL);
8779 q = ill->ill_wq;
8780 }
8781 /*
8782 * If any error acks received during the plumbing sequence,
8783 * ill_ifname_pending_err will be set. Break out and send up
8784 * the error to the pending ioctl.
8785 */
8786 if (ill->ill_ifname_pending_err != 0) {
8787 err = ill->ill_ifname_pending_err;
8788 ill->ill_ifname_pending_err = 0;
8789 break;
8790 }
8791
8792 ill->ill_phys_addr_mp = mp;
8793 ill->ill_phys_addr = (paddrlen == 0 ? NULL : addr);
8794 mp = NULL;
8795
8796 /*
8797 * If paddrlen or ill_phys_addr_length is zero, the DLPI
8798 * provider doesn't support physical addresses. We check both
8799 * paddrlen and ill_phys_addr_length because sppp (PPP) does
8800 * not have physical addresses, but historically adversises a
8801 * physical address length of 0 in its DL_INFO_ACK, but 6 in
8802 * its DL_PHYS_ADDR_ACK.
8803 */
8804 if (paddrlen == 0 || ill->ill_phys_addr_length == 0) {
8805 ill->ill_phys_addr = NULL;
8806 } else if (paddrlen != ill->ill_phys_addr_length) {
8807 ip0dbg(("DL_PHYS_ADDR_ACK: got addrlen %d, expected %d",
8808 paddrlen, ill->ill_phys_addr_length));
8809 err = EINVAL;
8810 break;
8811 }
8812
8813 if (ill->ill_nd_lla_mp == NULL) {
8814 if ((mp_hw = copyb(ill->ill_phys_addr_mp)) == NULL) {
8815 err = ENOMEM;
8816 break;
8817 }
8818 ill_set_ndmp(ill, mp_hw, paddroff, paddrlen);
8819 }
8820
8821 if (ill->ill_isv6) {
8822 ill_setdefaulttoken(ill);
8823 ipif_setlinklocal(ill->ill_ipif);
8824 }
8825 break;
8826 }
8827 case DL_OK_ACK:
8828 ip2dbg(("DL_OK_ACK %s (0x%x)\n",
8829 dl_primstr((int)dloa->dl_correct_primitive),
8830 dloa->dl_correct_primitive));
8831 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer ok",
8832 char *, dl_primstr(dloa->dl_correct_primitive),
8833 ill_t *, ill);
8834
8835 switch (dloa->dl_correct_primitive) {
8836 case DL_ENABMULTI_REQ:
8837 case DL_DISABMULTI_REQ:
8838 ill_dlpi_done(ill, dloa->dl_correct_primitive);
8839 break;
8840 case DL_PROMISCON_REQ:
8841 case DL_PROMISCOFF_REQ:
8842 case DL_UNBIND_REQ:
8843 case DL_ATTACH_REQ:
8844 ill_dlpi_done(ill, dloa->dl_correct_primitive);
8845 break;
8846 }
8847 break;
8848 default:
8849 break;
8850 }
8851
8852 freemsg(mp);
8853 if (mp1 == NULL)
8854 return;
8855
8856 /*
8857 * The operation must complete without EINPROGRESS since
8858 * ipsq_pending_mp_get() has removed the mblk (mp1). Otherwise,
8859 * the operation will be stuck forever inside the IPSQ.
8860 */
8861 ASSERT(err != EINPROGRESS);
8862
8863 DTRACE_PROBE4(ipif__ioctl, char *, "ip_rput_dlpi_writer finish",
8864 int, ipsq->ipsq_xop->ipx_current_ioctl, ill_t *, ill,
8865 ipif_t *, NULL);
8866
8867 switch (ipsq->ipsq_xop->ipx_current_ioctl) {
8868 case 0:
8869 ipsq_current_finish(ipsq);
8870 break;
8871
8872 case SIOCSLIFNAME:
8873 case IF_UNITSEL: {
8874 ill_t *ill_other = ILL_OTHER(ill);
8875
8876 /*
8877 * If SIOCSLIFNAME or IF_UNITSEL is about to succeed, and the
8878 * ill has a peer which is in an IPMP group, then place ill
8879 * into the same group. One catch: although ifconfig plumbs
8880 * the appropriate IPMP meta-interface prior to plumbing this
8881 * ill, it is possible for multiple ifconfig applications to
8882 * race (or for another application to adjust plumbing), in
8883 * which case the IPMP meta-interface we need will be missing.
8884 * If so, kick the phyint out of the group.
8885 */
8886 if (err == 0 && ill_other != NULL && IS_UNDER_IPMP(ill_other)) {
8887 ipmp_grp_t *grp = ill->ill_phyint->phyint_grp;
8888 ipmp_illgrp_t *illg;
8889
8890 illg = ill->ill_isv6 ? grp->gr_v6 : grp->gr_v4;
8891 if (illg == NULL)
8892 ipmp_phyint_leave_grp(ill->ill_phyint);
8893 else
8894 ipmp_ill_join_illgrp(ill, illg);
8895 }
8896
8897 if (ipsq->ipsq_xop->ipx_current_ioctl == IF_UNITSEL)
8898 ip_ioctl_finish(q, mp1, err, NO_COPYOUT, ipsq);
8899 else
8900 ip_ioctl_finish(q, mp1, err, COPYOUT, ipsq);
8901 break;
8902 }
8903 case SIOCLIFADDIF:
8904 ip_ioctl_finish(q, mp1, err, COPYOUT, ipsq);
8905 break;
8906
8907 default:
8908 ip_ioctl_finish(q, mp1, err, NO_COPYOUT, ipsq);
8909 break;
8910 }
8911 }
8912
8913 /*
8914 * ip_rput_other is called by ip_rput to handle messages modifying the global
8915 * state in IP. If 'ipsq' is non-NULL, caller is writer on it.
8916 */
8917 /* ARGSUSED */
8918 void
8919 ip_rput_other(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
8920 {
8921 ill_t *ill = q->q_ptr;
8922 struct iocblk *iocp;
8923
8924 ip1dbg(("ip_rput_other "));
8925 if (ipsq != NULL) {
8926 ASSERT(IAM_WRITER_IPSQ(ipsq));
8927 ASSERT(ipsq->ipsq_xop ==
8928 ill->ill_phyint->phyint_ipsq->ipsq_xop);
8929 }
8930
8931 switch (mp->b_datap->db_type) {
8932 case M_ERROR:
8933 case M_HANGUP:
8934 /*
8935 * The device has a problem. We force the ILL down. It can
8936 * be brought up again manually using SIOCSIFFLAGS (via
8937 * ifconfig or equivalent).
8938 */
8939 ASSERT(ipsq != NULL);
8940 if (mp->b_rptr < mp->b_wptr)
8941 ill->ill_error = (int)(*mp->b_rptr & 0xFF);
8942 if (ill->ill_error == 0)
8943 ill->ill_error = ENXIO;
8944 if (!ill_down_start(q, mp))
8945 return;
8946 ipif_all_down_tail(ipsq, q, mp, NULL);
8947 break;
8948 case M_IOCNAK: {
8949 iocp = (struct iocblk *)mp->b_rptr;
8950
8951 ASSERT(iocp->ioc_cmd == DL_IOC_HDR_INFO);
8952 /*
8953 * If this was the first attempt, turn off the fastpath
8954 * probing.
8955 */
8956 mutex_enter(&ill->ill_lock);
8957 if (ill->ill_dlpi_fastpath_state == IDS_INPROGRESS) {
8958 ill->ill_dlpi_fastpath_state = IDS_FAILED;
8959 mutex_exit(&ill->ill_lock);
8960 /*
8961 * don't flush the nce_t entries: we use them
8962 * as an index to the ncec itself.
8963 */
8964 ip1dbg(("ip_rput: DLPI fastpath off on interface %s\n",
8965 ill->ill_name));
8966 } else {
8967 mutex_exit(&ill->ill_lock);
8968 }
8969 freemsg(mp);
8970 break;
8971 }
8972 default:
8973 ASSERT(0);
8974 break;
8975 }
8976 }
8977
8978 /*
8979 * Update any source route, record route or timestamp options
8980 * When it fails it has consumed the message and BUMPed the MIB.
8981 */
8982 boolean_t
8983 ip_forward_options(mblk_t *mp, ipha_t *ipha, ill_t *dst_ill,
8984 ip_recv_attr_t *ira)
8985 {
8986 ipoptp_t opts;
8987 uchar_t *opt;
8988 uint8_t optval;
8989 uint8_t optlen;
8990 ipaddr_t dst;
8991 ipaddr_t ifaddr;
8992 uint32_t ts;
8993 timestruc_t now;
8994 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
8995
8996 ip2dbg(("ip_forward_options\n"));
8997 dst = ipha->ipha_dst;
8998 for (optval = ipoptp_first(&opts, ipha);
8999 optval != IPOPT_EOL;
9000 optval = ipoptp_next(&opts)) {
9001 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
9002 opt = opts.ipoptp_cur;
9003 optlen = opts.ipoptp_len;
9004 ip2dbg(("ip_forward_options: opt %d, len %d\n",
9005 optval, opts.ipoptp_len));
9006 switch (optval) {
9007 uint32_t off;
9008 case IPOPT_SSRR:
9009 case IPOPT_LSRR:
9010 /* Check if adminstratively disabled */
9011 if (!ipst->ips_ip_forward_src_routed) {
9012 BUMP_MIB(dst_ill->ill_ip_mib,
9013 ipIfStatsForwProhibits);
9014 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED",
9015 mp, dst_ill);
9016 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED,
9017 ira);
9018 return (B_FALSE);
9019 }
9020 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9021 /*
9022 * Must be partial since ip_input_options
9023 * checked for strict.
9024 */
9025 break;
9026 }
9027 off = opt[IPOPT_OFFSET];
9028 off--;
9029 redo_srr:
9030 if (optlen < IP_ADDR_LEN ||
9031 off > optlen - IP_ADDR_LEN) {
9032 /* End of source route */
9033 ip1dbg((
9034 "ip_forward_options: end of SR\n"));
9035 break;
9036 }
9037 /* Pick a reasonable address on the outbound if */
9038 ASSERT(dst_ill != NULL);
9039 if (ip_select_source_v4(dst_ill, INADDR_ANY, dst,
9040 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9041 NULL) != 0) {
9042 /* No source! Shouldn't happen */
9043 ifaddr = INADDR_ANY;
9044 }
9045 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9046 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9047 ip1dbg(("ip_forward_options: next hop 0x%x\n",
9048 ntohl(dst)));
9049
9050 /*
9051 * Check if our address is present more than
9052 * once as consecutive hops in source route.
9053 */
9054 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
9055 off += IP_ADDR_LEN;
9056 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9057 goto redo_srr;
9058 }
9059 ipha->ipha_dst = dst;
9060 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9061 break;
9062 case IPOPT_RR:
9063 off = opt[IPOPT_OFFSET];
9064 off--;
9065 if (optlen < IP_ADDR_LEN ||
9066 off > optlen - IP_ADDR_LEN) {
9067 /* No more room - ignore */
9068 ip1dbg((
9069 "ip_forward_options: end of RR\n"));
9070 break;
9071 }
9072 /* Pick a reasonable address on the outbound if */
9073 ASSERT(dst_ill != NULL);
9074 if (ip_select_source_v4(dst_ill, INADDR_ANY, dst,
9075 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9076 NULL) != 0) {
9077 /* No source! Shouldn't happen */
9078 ifaddr = INADDR_ANY;
9079 }
9080 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9081 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9082 break;
9083 case IPOPT_TS:
9084 /* Insert timestamp if there is room */
9085 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9086 case IPOPT_TS_TSONLY:
9087 off = IPOPT_TS_TIMELEN;
9088 break;
9089 case IPOPT_TS_PRESPEC:
9090 case IPOPT_TS_PRESPEC_RFC791:
9091 /* Verify that the address matched */
9092 off = opt[IPOPT_OFFSET] - 1;
9093 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9094 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9095 /* Not for us */
9096 break;
9097 }
9098 /* FALLTHRU */
9099 case IPOPT_TS_TSANDADDR:
9100 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9101 break;
9102 default:
9103 /*
9104 * ip_*put_options should have already
9105 * dropped this packet.
9106 */
9107 cmn_err(CE_PANIC, "ip_forward_options: "
9108 "unknown IT - bug in ip_input_options?\n");
9109 return (B_TRUE); /* Keep "lint" happy */
9110 }
9111 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
9112 /* Increase overflow counter */
9113 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
9114 opt[IPOPT_POS_OV_FLG] =
9115 (uint8_t)((opt[IPOPT_POS_OV_FLG] & 0x0F) |
9116 (off << 4));
9117 break;
9118 }
9119 off = opt[IPOPT_OFFSET] - 1;
9120 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9121 case IPOPT_TS_PRESPEC:
9122 case IPOPT_TS_PRESPEC_RFC791:
9123 case IPOPT_TS_TSANDADDR:
9124 /* Pick a reasonable addr on the outbound if */
9125 ASSERT(dst_ill != NULL);
9126 if (ip_select_source_v4(dst_ill, INADDR_ANY,
9127 dst, INADDR_ANY, ALL_ZONES, ipst, &ifaddr,
9128 NULL, NULL) != 0) {
9129 /* No source! Shouldn't happen */
9130 ifaddr = INADDR_ANY;
9131 }
9132 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9133 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9134 /* FALLTHRU */
9135 case IPOPT_TS_TSONLY:
9136 off = opt[IPOPT_OFFSET] - 1;
9137 /* Compute # of milliseconds since midnight */
9138 gethrestime(&now);
9139 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
9140 NSEC2MSEC(now.tv_nsec);
9141 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
9142 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
9143 break;
9144 }
9145 break;
9146 }
9147 }
9148 return (B_TRUE);
9149 }
9150
9151 /*
9152 * Call ill_frag_timeout to do garbage collection. ill_frag_timeout
9153 * returns 'true' if there are still fragments left on the queue, in
9154 * which case we restart the timer.
9155 */
9156 void
9157 ill_frag_timer(void *arg)
9158 {
9159 ill_t *ill = (ill_t *)arg;
9160 boolean_t frag_pending;
9161 ip_stack_t *ipst = ill->ill_ipst;
9162 time_t timeout;
9163
9164 mutex_enter(&ill->ill_lock);
9165 ASSERT(!ill->ill_fragtimer_executing);
9166 if (ill->ill_state_flags & ILL_CONDEMNED) {
9167 ill->ill_frag_timer_id = 0;
9168 mutex_exit(&ill->ill_lock);
9169 return;
9170 }
9171 ill->ill_fragtimer_executing = 1;
9172 mutex_exit(&ill->ill_lock);
9173
9174 timeout = (ill->ill_isv6 ? ipst->ips_ipv6_reassembly_timeout :
9175 ipst->ips_ip_reassembly_timeout);
9176
9177 frag_pending = ill_frag_timeout(ill, timeout);
9178
9179 /*
9180 * Restart the timer, if we have fragments pending or if someone
9181 * wanted us to be scheduled again.
9182 */
9183 mutex_enter(&ill->ill_lock);
9184 ill->ill_fragtimer_executing = 0;
9185 ill->ill_frag_timer_id = 0;
9186 if (frag_pending || ill->ill_fragtimer_needrestart)
9187 ill_frag_timer_start(ill);
9188 mutex_exit(&ill->ill_lock);
9189 }
9190
9191 void
9192 ill_frag_timer_start(ill_t *ill)
9193 {
9194 ip_stack_t *ipst = ill->ill_ipst;
9195 clock_t timeo_ms;
9196
9197 ASSERT(MUTEX_HELD(&ill->ill_lock));
9198
9199 /* If the ill is closing or opening don't proceed */
9200 if (ill->ill_state_flags & ILL_CONDEMNED)
9201 return;
9202
9203 if (ill->ill_fragtimer_executing) {
9204 /*
9205 * ill_frag_timer is currently executing. Just record the
9206 * the fact that we want the timer to be restarted.
9207 * ill_frag_timer will post a timeout before it returns,
9208 * ensuring it will be called again.
9209 */
9210 ill->ill_fragtimer_needrestart = 1;
9211 return;
9212 }
9213
9214 if (ill->ill_frag_timer_id == 0) {
9215 timeo_ms = (ill->ill_isv6 ? ipst->ips_ipv6_reassembly_timeout :
9216 ipst->ips_ip_reassembly_timeout) * SECONDS;
9217
9218 /*
9219 * The timer is neither running nor is the timeout handler
9220 * executing. Post a timeout so that ill_frag_timer will be
9221 * called
9222 */
9223 ill->ill_frag_timer_id = timeout(ill_frag_timer, ill,
9224 MSEC_TO_TICK(timeo_ms >> 1));
9225 ill->ill_fragtimer_needrestart = 0;
9226 }
9227 }
9228
9229 /*
9230 * Update any source route, record route or timestamp options.
9231 * Check that we are at end of strict source route.
9232 * The options have already been checked for sanity in ip_input_options().
9233 */
9234 boolean_t
9235 ip_input_local_options(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
9236 {
9237 ipoptp_t opts;
9238 uchar_t *opt;
9239 uint8_t optval;
9240 uint8_t optlen;
9241 ipaddr_t dst;
9242 ipaddr_t ifaddr;
9243 uint32_t ts;
9244 timestruc_t now;
9245 ill_t *ill = ira->ira_ill;
9246 ip_stack_t *ipst = ill->ill_ipst;
9247
9248 ip2dbg(("ip_input_local_options\n"));
9249
9250 for (optval = ipoptp_first(&opts, ipha);
9251 optval != IPOPT_EOL;
9252 optval = ipoptp_next(&opts)) {
9253 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
9254 opt = opts.ipoptp_cur;
9255 optlen = opts.ipoptp_len;
9256 ip2dbg(("ip_input_local_options: opt %d, len %d\n",
9257 optval, optlen));
9258 switch (optval) {
9259 uint32_t off;
9260 case IPOPT_SSRR:
9261 case IPOPT_LSRR:
9262 off = opt[IPOPT_OFFSET];
9263 off--;
9264 if (optlen < IP_ADDR_LEN ||
9265 off > optlen - IP_ADDR_LEN) {
9266 /* End of source route */
9267 ip1dbg(("ip_input_local_options: end of SR\n"));
9268 break;
9269 }
9270 /*
9271 * This will only happen if two consecutive entries
9272 * in the source route contains our address or if
9273 * it is a packet with a loose source route which
9274 * reaches us before consuming the whole source route
9275 */
9276 ip1dbg(("ip_input_local_options: not end of SR\n"));
9277 if (optval == IPOPT_SSRR) {
9278 goto bad_src_route;
9279 }
9280 /*
9281 * Hack: instead of dropping the packet truncate the
9282 * source route to what has been used by filling the
9283 * rest with IPOPT_NOP.
9284 */
9285 opt[IPOPT_OLEN] = (uint8_t)off;
9286 while (off < optlen) {
9287 opt[off++] = IPOPT_NOP;
9288 }
9289 break;
9290 case IPOPT_RR:
9291 off = opt[IPOPT_OFFSET];
9292 off--;
9293 if (optlen < IP_ADDR_LEN ||
9294 off > optlen - IP_ADDR_LEN) {
9295 /* No more room - ignore */
9296 ip1dbg((
9297 "ip_input_local_options: end of RR\n"));
9298 break;
9299 }
9300 /* Pick a reasonable address on the outbound if */
9301 if (ip_select_source_v4(ill, INADDR_ANY, ipha->ipha_dst,
9302 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9303 NULL) != 0) {
9304 /* No source! Shouldn't happen */
9305 ifaddr = INADDR_ANY;
9306 }
9307 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9308 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9309 break;
9310 case IPOPT_TS:
9311 /* Insert timestamp if there is romm */
9312 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9313 case IPOPT_TS_TSONLY:
9314 off = IPOPT_TS_TIMELEN;
9315 break;
9316 case IPOPT_TS_PRESPEC:
9317 case IPOPT_TS_PRESPEC_RFC791:
9318 /* Verify that the address matched */
9319 off = opt[IPOPT_OFFSET] - 1;
9320 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9321 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9322 /* Not for us */
9323 break;
9324 }
9325 /* FALLTHRU */
9326 case IPOPT_TS_TSANDADDR:
9327 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9328 break;
9329 default:
9330 /*
9331 * ip_*put_options should have already
9332 * dropped this packet.
9333 */
9334 cmn_err(CE_PANIC, "ip_input_local_options: "
9335 "unknown IT - bug in ip_input_options?\n");
9336 return (B_TRUE); /* Keep "lint" happy */
9337 }
9338 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
9339 /* Increase overflow counter */
9340 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
9341 opt[IPOPT_POS_OV_FLG] =
9342 (uint8_t)((opt[IPOPT_POS_OV_FLG] & 0x0F) |
9343 (off << 4));
9344 break;
9345 }
9346 off = opt[IPOPT_OFFSET] - 1;
9347 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9348 case IPOPT_TS_PRESPEC:
9349 case IPOPT_TS_PRESPEC_RFC791:
9350 case IPOPT_TS_TSANDADDR:
9351 /* Pick a reasonable addr on the outbound if */
9352 if (ip_select_source_v4(ill, INADDR_ANY,
9353 ipha->ipha_dst, INADDR_ANY, ALL_ZONES, ipst,
9354 &ifaddr, NULL, NULL) != 0) {
9355 /* No source! Shouldn't happen */
9356 ifaddr = INADDR_ANY;
9357 }
9358 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9359 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9360 /* FALLTHRU */
9361 case IPOPT_TS_TSONLY:
9362 off = opt[IPOPT_OFFSET] - 1;
9363 /* Compute # of milliseconds since midnight */
9364 gethrestime(&now);
9365 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
9366 NSEC2MSEC(now.tv_nsec);
9367 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
9368 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
9369 break;
9370 }
9371 break;
9372 }
9373 }
9374 return (B_TRUE);
9375
9376 bad_src_route:
9377 /* make sure we clear any indication of a hardware checksum */
9378 DB_CKSUMFLAGS(mp) = 0;
9379 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ill);
9380 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
9381 return (B_FALSE);
9382
9383 }
9384
9385 /*
9386 * Process IP options in an inbound packet. Always returns the nexthop.
9387 * Normally this is the passed in nexthop, but if there is an option
9388 * that effects the nexthop (such as a source route) that will be returned.
9389 * Sets *errorp if there is an error, in which case an ICMP error has been sent
9390 * and mp freed.
9391 */
9392 ipaddr_t
9393 ip_input_options(ipha_t *ipha, ipaddr_t dst, mblk_t *mp,
9394 ip_recv_attr_t *ira, int *errorp)
9395 {
9396 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
9397 ipoptp_t opts;
9398 uchar_t *opt;
9399 uint8_t optval;
9400 uint8_t optlen;
9401 intptr_t code = 0;
9402 ire_t *ire;
9403
9404 ip2dbg(("ip_input_options\n"));
9405 *errorp = 0;
9406 for (optval = ipoptp_first(&opts, ipha);
9407 optval != IPOPT_EOL;
9408 optval = ipoptp_next(&opts)) {
9409 opt = opts.ipoptp_cur;
9410 optlen = opts.ipoptp_len;
9411 ip2dbg(("ip_input_options: opt %d, len %d\n",
9412 optval, optlen));
9413 /*
9414 * Note: we need to verify the checksum before we
9415 * modify anything thus this routine only extracts the next
9416 * hop dst from any source route.
9417 */
9418 switch (optval) {
9419 uint32_t off;
9420 case IPOPT_SSRR:
9421 case IPOPT_LSRR:
9422 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9423 if (optval == IPOPT_SSRR) {
9424 ip1dbg(("ip_input_options: not next"
9425 " strict source route 0x%x\n",
9426 ntohl(dst)));
9427 code = (char *)&ipha->ipha_dst -
9428 (char *)ipha;
9429 goto param_prob; /* RouterReq's */
9430 }
9431 ip2dbg(("ip_input_options: "
9432 "not next source route 0x%x\n",
9433 ntohl(dst)));
9434 break;
9435 }
9436
9437 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9438 ip1dbg((
9439 "ip_input_options: bad option offset\n"));
9440 code = (char *)&opt[IPOPT_OLEN] -
9441 (char *)ipha;
9442 goto param_prob;
9443 }
9444 off = opt[IPOPT_OFFSET];
9445 off--;
9446 redo_srr:
9447 if (optlen < IP_ADDR_LEN ||
9448 off > optlen - IP_ADDR_LEN) {
9449 /* End of source route */
9450 ip1dbg(("ip_input_options: end of SR\n"));
9451 break;
9452 }
9453 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9454 ip1dbg(("ip_input_options: next hop 0x%x\n",
9455 ntohl(dst)));
9456
9457 /*
9458 * Check if our address is present more than
9459 * once as consecutive hops in source route.
9460 * XXX verify per-interface ip_forwarding
9461 * for source route?
9462 */
9463 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
9464 off += IP_ADDR_LEN;
9465 goto redo_srr;
9466 }
9467
9468 if (dst == htonl(INADDR_LOOPBACK)) {
9469 ip1dbg(("ip_input_options: loopback addr in "
9470 "source route!\n"));
9471 goto bad_src_route;
9472 }
9473 /*
9474 * For strict: verify that dst is directly
9475 * reachable.
9476 */
9477 if (optval == IPOPT_SSRR) {
9478 ire = ire_ftable_lookup_v4(dst, 0, 0,
9479 IRE_INTERFACE, NULL, ALL_ZONES,
9480 ira->ira_tsl,
9481 MATCH_IRE_TYPE | MATCH_IRE_SECATTR, 0, ipst,
9482 NULL);
9483 if (ire == NULL) {
9484 ip1dbg(("ip_input_options: SSRR not "
9485 "directly reachable: 0x%x\n",
9486 ntohl(dst)));
9487 goto bad_src_route;
9488 }
9489 ire_refrele(ire);
9490 }
9491 /*
9492 * Defer update of the offset and the record route
9493 * until the packet is forwarded.
9494 */
9495 break;
9496 case IPOPT_RR:
9497 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9498 ip1dbg((
9499 "ip_input_options: bad option offset\n"));
9500 code = (char *)&opt[IPOPT_OLEN] -
9501 (char *)ipha;
9502 goto param_prob;
9503 }
9504 break;
9505 case IPOPT_TS:
9506 /*
9507 * Verify that length >= 5 and that there is either
9508 * room for another timestamp or that the overflow
9509 * counter is not maxed out.
9510 */
9511 code = (char *)&opt[IPOPT_OLEN] - (char *)ipha;
9512 if (optlen < IPOPT_MINLEN_IT) {
9513 goto param_prob;
9514 }
9515 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9516 ip1dbg((
9517 "ip_input_options: bad option offset\n"));
9518 code = (char *)&opt[IPOPT_OFFSET] -
9519 (char *)ipha;
9520 goto param_prob;
9521 }
9522 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9523 case IPOPT_TS_TSONLY:
9524 off = IPOPT_TS_TIMELEN;
9525 break;
9526 case IPOPT_TS_TSANDADDR:
9527 case IPOPT_TS_PRESPEC:
9528 case IPOPT_TS_PRESPEC_RFC791:
9529 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9530 break;
9531 default:
9532 code = (char *)&opt[IPOPT_POS_OV_FLG] -
9533 (char *)ipha;
9534 goto param_prob;
9535 }
9536 if (opt[IPOPT_OFFSET] - 1 + off > optlen &&
9537 (opt[IPOPT_POS_OV_FLG] & 0xF0) == 0xF0) {
9538 /*
9539 * No room and the overflow counter is 15
9540 * already.
9541 */
9542 goto param_prob;
9543 }
9544 break;
9545 }
9546 }
9547
9548 if ((opts.ipoptp_flags & IPOPTP_ERROR) == 0) {
9549 return (dst);
9550 }
9551
9552 ip1dbg(("ip_input_options: error processing IP options."));
9553 code = (char *)&opt[IPOPT_OFFSET] - (char *)ipha;
9554
9555 param_prob:
9556 /* make sure we clear any indication of a hardware checksum */
9557 DB_CKSUMFLAGS(mp) = 0;
9558 ip_drop_input("ICMP_PARAM_PROBLEM", mp, ira->ira_ill);
9559 icmp_param_problem(mp, (uint8_t)code, ira);
9560 *errorp = -1;
9561 return (dst);
9562
9563 bad_src_route:
9564 /* make sure we clear any indication of a hardware checksum */
9565 DB_CKSUMFLAGS(mp) = 0;
9566 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ira->ira_ill);
9567 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
9568 *errorp = -1;
9569 return (dst);
9570 }
9571
9572 /*
9573 * IP & ICMP info in >=14 msg's ...
9574 * - ip fixed part (mib2_ip_t)
9575 * - icmp fixed part (mib2_icmp_t)
9576 * - ipAddrEntryTable (ip 20) all IPv4 ipifs
9577 * - ipRouteEntryTable (ip 21) all IPv4 IREs
9578 * - ipNetToMediaEntryTable (ip 22) all IPv4 Neighbor Cache entries
9579 * - ipRouteAttributeTable (ip 102) labeled routes
9580 * - ip multicast membership (ip_member_t)
9581 * - ip multicast source filtering (ip_grpsrc_t)
9582 * - igmp fixed part (struct igmpstat)
9583 * - multicast routing stats (struct mrtstat)
9584 * - multicast routing vifs (array of struct vifctl)
9585 * - multicast routing routes (array of struct mfcctl)
9586 * - ip6 fixed part (mib2_ipv6IfStatsEntry_t)
9587 * One per ill plus one generic
9588 * - icmp6 fixed part (mib2_ipv6IfIcmpEntry_t)
9589 * One per ill plus one generic
9590 * - ipv6RouteEntry all IPv6 IREs
9591 * - ipv6RouteAttributeTable (ip6 102) labeled routes
9592 * - ipv6NetToMediaEntry all IPv6 Neighbor Cache entries
9593 * - ipv6AddrEntry all IPv6 ipifs
9594 * - ipv6 multicast membership (ipv6_member_t)
9595 * - ipv6 multicast source filtering (ipv6_grpsrc_t)
9596 *
9597 * NOTE: original mpctl is copied for msg's 2..N, since its ctl part is
9598 * already filled in by the caller.
9599 * If legacy_req is true then MIB structures needs to be truncated to their
9600 * legacy sizes before being returned.
9601 * Return value of 0 indicates that no messages were sent and caller
9602 * should free mpctl.
9603 */
9604 int
9605 ip_snmp_get(queue_t *q, mblk_t *mpctl, int level, boolean_t legacy_req)
9606 {
9607 ip_stack_t *ipst;
9608 sctp_stack_t *sctps;
9609
9610 if (q->q_next != NULL) {
9611 ipst = ILLQ_TO_IPST(q);
9612 } else {
9613 ipst = CONNQ_TO_IPST(q);
9614 }
9615 ASSERT(ipst != NULL);
9616 sctps = ipst->ips_netstack->netstack_sctp;
9617
9618 if (mpctl == NULL || mpctl->b_cont == NULL) {
9619 return (0);
9620 }
9621
9622 /*
9623 * For the purposes of the (broken) packet shell use
9624 * of the level we make sure MIB2_TCP/MIB2_UDP can be used
9625 * to make TCP and UDP appear first in the list of mib items.
9626 * TBD: We could expand this and use it in netstat so that
9627 * the kernel doesn't have to produce large tables (connections,
9628 * routes, etc) when netstat only wants the statistics or a particular
9629 * table.
9630 */
9631 if (!(level == MIB2_TCP || level == MIB2_UDP)) {
9632 if ((mpctl = icmp_snmp_get(q, mpctl)) == NULL) {
9633 return (1);
9634 }
9635 }
9636
9637 if (level != MIB2_TCP) {
9638 if ((mpctl = udp_snmp_get(q, mpctl, legacy_req)) == NULL) {
9639 return (1);
9640 }
9641 }
9642
9643 if (level != MIB2_UDP) {
9644 if ((mpctl = tcp_snmp_get(q, mpctl, legacy_req)) == NULL) {
9645 return (1);
9646 }
9647 }
9648
9649 if ((mpctl = ip_snmp_get_mib2_ip_traffic_stats(q, mpctl,
9650 ipst, legacy_req)) == NULL) {
9651 return (1);
9652 }
9653
9654 if ((mpctl = ip_snmp_get_mib2_ip6(q, mpctl, ipst,
9655 legacy_req)) == NULL) {
9656 return (1);
9657 }
9658
9659 if ((mpctl = ip_snmp_get_mib2_icmp(q, mpctl, ipst)) == NULL) {
9660 return (1);
9661 }
9662
9663 if ((mpctl = ip_snmp_get_mib2_icmp6(q, mpctl, ipst)) == NULL) {
9664 return (1);
9665 }
9666
9667 if ((mpctl = ip_snmp_get_mib2_igmp(q, mpctl, ipst)) == NULL) {
9668 return (1);
9669 }
9670
9671 if ((mpctl = ip_snmp_get_mib2_multi(q, mpctl, ipst)) == NULL) {
9672 return (1);
9673 }
9674
9675 if ((mpctl = ip_snmp_get_mib2_ip_addr(q, mpctl, ipst,
9676 legacy_req)) == NULL) {
9677 return (1);
9678 }
9679
9680 if ((mpctl = ip_snmp_get_mib2_ip6_addr(q, mpctl, ipst,
9681 legacy_req)) == NULL) {
9682 return (1);
9683 }
9684
9685 if ((mpctl = ip_snmp_get_mib2_ip_group_mem(q, mpctl, ipst)) == NULL) {
9686 return (1);
9687 }
9688
9689 if ((mpctl = ip_snmp_get_mib2_ip6_group_mem(q, mpctl, ipst)) == NULL) {
9690 return (1);
9691 }
9692
9693 if ((mpctl = ip_snmp_get_mib2_ip_group_src(q, mpctl, ipst)) == NULL) {
9694 return (1);
9695 }
9696
9697 if ((mpctl = ip_snmp_get_mib2_ip6_group_src(q, mpctl, ipst)) == NULL) {
9698 return (1);
9699 }
9700
9701 if ((mpctl = ip_snmp_get_mib2_virt_multi(q, mpctl, ipst)) == NULL) {
9702 return (1);
9703 }
9704
9705 if ((mpctl = ip_snmp_get_mib2_multi_rtable(q, mpctl, ipst)) == NULL) {
9706 return (1);
9707 }
9708
9709 mpctl = ip_snmp_get_mib2_ip_route_media(q, mpctl, level, ipst);
9710 if (mpctl == NULL)
9711 return (1);
9712
9713 mpctl = ip_snmp_get_mib2_ip6_route_media(q, mpctl, level, ipst);
9714 if (mpctl == NULL)
9715 return (1);
9716
9717 if ((mpctl = sctp_snmp_get_mib2(q, mpctl, sctps)) == NULL) {
9718 return (1);
9719 }
9720 if ((mpctl = ip_snmp_get_mib2_ip_dce(q, mpctl, ipst)) == NULL) {
9721 return (1);
9722 }
9723 freemsg(mpctl);
9724 return (1);
9725 }
9726
9727 /* Get global (legacy) IPv4 statistics */
9728 static mblk_t *
9729 ip_snmp_get_mib2_ip(queue_t *q, mblk_t *mpctl, mib2_ipIfStatsEntry_t *ipmib,
9730 ip_stack_t *ipst, boolean_t legacy_req)
9731 {
9732 mib2_ip_t old_ip_mib;
9733 struct opthdr *optp;
9734 mblk_t *mp2ctl;
9735 mib2_ipAddrEntry_t mae;
9736
9737 /*
9738 * make a copy of the original message
9739 */
9740 mp2ctl = copymsg(mpctl);
9741
9742 /* fixed length IP structure... */
9743 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9744 optp->level = MIB2_IP;
9745 optp->name = 0;
9746 SET_MIB(old_ip_mib.ipForwarding,
9747 (WE_ARE_FORWARDING(ipst) ? 1 : 2));
9748 SET_MIB(old_ip_mib.ipDefaultTTL,
9749 (uint32_t)ipst->ips_ip_def_ttl);
9750 SET_MIB(old_ip_mib.ipReasmTimeout,
9751 ipst->ips_ip_reassembly_timeout);
9752 SET_MIB(old_ip_mib.ipAddrEntrySize,
9753 (legacy_req) ? LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t) :
9754 sizeof (mib2_ipAddrEntry_t));
9755 SET_MIB(old_ip_mib.ipRouteEntrySize,
9756 sizeof (mib2_ipRouteEntry_t));
9757 SET_MIB(old_ip_mib.ipNetToMediaEntrySize,
9758 sizeof (mib2_ipNetToMediaEntry_t));
9759 SET_MIB(old_ip_mib.ipMemberEntrySize, sizeof (ip_member_t));
9760 SET_MIB(old_ip_mib.ipGroupSourceEntrySize, sizeof (ip_grpsrc_t));
9761 SET_MIB(old_ip_mib.ipRouteAttributeSize,
9762 sizeof (mib2_ipAttributeEntry_t));
9763 SET_MIB(old_ip_mib.transportMLPSize, sizeof (mib2_transportMLPEntry_t));
9764 SET_MIB(old_ip_mib.ipDestEntrySize, sizeof (dest_cache_entry_t));
9765
9766 /*
9767 * Grab the statistics from the new IP MIB
9768 */
9769 SET_MIB(old_ip_mib.ipInReceives,
9770 (uint32_t)ipmib->ipIfStatsHCInReceives);
9771 SET_MIB(old_ip_mib.ipInHdrErrors, ipmib->ipIfStatsInHdrErrors);
9772 SET_MIB(old_ip_mib.ipInAddrErrors, ipmib->ipIfStatsInAddrErrors);
9773 SET_MIB(old_ip_mib.ipForwDatagrams,
9774 (uint32_t)ipmib->ipIfStatsHCOutForwDatagrams);
9775 SET_MIB(old_ip_mib.ipInUnknownProtos,
9776 ipmib->ipIfStatsInUnknownProtos);
9777 SET_MIB(old_ip_mib.ipInDiscards, ipmib->ipIfStatsInDiscards);
9778 SET_MIB(old_ip_mib.ipInDelivers,
9779 (uint32_t)ipmib->ipIfStatsHCInDelivers);
9780 SET_MIB(old_ip_mib.ipOutRequests,
9781 (uint32_t)ipmib->ipIfStatsHCOutRequests);
9782 SET_MIB(old_ip_mib.ipOutDiscards, ipmib->ipIfStatsOutDiscards);
9783 SET_MIB(old_ip_mib.ipOutNoRoutes, ipmib->ipIfStatsOutNoRoutes);
9784 SET_MIB(old_ip_mib.ipReasmReqds, ipmib->ipIfStatsReasmReqds);
9785 SET_MIB(old_ip_mib.ipReasmOKs, ipmib->ipIfStatsReasmOKs);
9786 SET_MIB(old_ip_mib.ipReasmFails, ipmib->ipIfStatsReasmFails);
9787 SET_MIB(old_ip_mib.ipFragOKs, ipmib->ipIfStatsOutFragOKs);
9788 SET_MIB(old_ip_mib.ipFragFails, ipmib->ipIfStatsOutFragFails);
9789 SET_MIB(old_ip_mib.ipFragCreates, ipmib->ipIfStatsOutFragCreates);
9790
9791 /* ipRoutingDiscards is not being used */
9792 SET_MIB(old_ip_mib.ipRoutingDiscards, 0);
9793 SET_MIB(old_ip_mib.tcpInErrs, ipmib->tcpIfStatsInErrs);
9794 SET_MIB(old_ip_mib.udpNoPorts, ipmib->udpIfStatsNoPorts);
9795 SET_MIB(old_ip_mib.ipInCksumErrs, ipmib->ipIfStatsInCksumErrs);
9796 SET_MIB(old_ip_mib.ipReasmDuplicates,
9797 ipmib->ipIfStatsReasmDuplicates);
9798 SET_MIB(old_ip_mib.ipReasmPartDups, ipmib->ipIfStatsReasmPartDups);
9799 SET_MIB(old_ip_mib.ipForwProhibits, ipmib->ipIfStatsForwProhibits);
9800 SET_MIB(old_ip_mib.udpInCksumErrs, ipmib->udpIfStatsInCksumErrs);
9801 SET_MIB(old_ip_mib.udpInOverflows, ipmib->udpIfStatsInOverflows);
9802 SET_MIB(old_ip_mib.rawipInOverflows,
9803 ipmib->rawipIfStatsInOverflows);
9804
9805 SET_MIB(old_ip_mib.ipsecInSucceeded, ipmib->ipsecIfStatsInSucceeded);
9806 SET_MIB(old_ip_mib.ipsecInFailed, ipmib->ipsecIfStatsInFailed);
9807 SET_MIB(old_ip_mib.ipInIPv6, ipmib->ipIfStatsInWrongIPVersion);
9808 SET_MIB(old_ip_mib.ipOutIPv6, ipmib->ipIfStatsOutWrongIPVersion);
9809 SET_MIB(old_ip_mib.ipOutSwitchIPv6,
9810 ipmib->ipIfStatsOutSwitchIPVersion);
9811
9812 if (!snmp_append_data(mpctl->b_cont, (char *)&old_ip_mib,
9813 (int)sizeof (old_ip_mib))) {
9814 ip1dbg(("ip_snmp_get_mib2_ip: failed to allocate %u bytes\n",
9815 (uint_t)sizeof (old_ip_mib)));
9816 }
9817
9818 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9819 ip3dbg(("ip_snmp_get_mib2_ip: level %d, name %d, len %d\n",
9820 (int)optp->level, (int)optp->name, (int)optp->len));
9821 qreply(q, mpctl);
9822 return (mp2ctl);
9823 }
9824
9825 /* Per interface IPv4 statistics */
9826 static mblk_t *
9827 ip_snmp_get_mib2_ip_traffic_stats(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
9828 boolean_t legacy_req)
9829 {
9830 struct opthdr *optp;
9831 mblk_t *mp2ctl;
9832 ill_t *ill;
9833 ill_walk_context_t ctx;
9834 mblk_t *mp_tail = NULL;
9835 mib2_ipIfStatsEntry_t global_ip_mib;
9836 mib2_ipAddrEntry_t mae;
9837
9838 /*
9839 * Make a copy of the original message
9840 */
9841 mp2ctl = copymsg(mpctl);
9842
9843 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9844 optp->level = MIB2_IP;
9845 optp->name = MIB2_IP_TRAFFIC_STATS;
9846 /* Include "unknown interface" ip_mib */
9847 ipst->ips_ip_mib.ipIfStatsIPVersion = MIB2_INETADDRESSTYPE_ipv4;
9848 ipst->ips_ip_mib.ipIfStatsIfIndex =
9849 MIB2_UNKNOWN_INTERFACE; /* Flag to netstat */
9850 SET_MIB(ipst->ips_ip_mib.ipIfStatsForwarding,
9851 (ipst->ips_ip_forwarding ? 1 : 2));
9852 SET_MIB(ipst->ips_ip_mib.ipIfStatsDefaultTTL,
9853 (uint32_t)ipst->ips_ip_def_ttl);
9854 SET_MIB(ipst->ips_ip_mib.ipIfStatsEntrySize,
9855 sizeof (mib2_ipIfStatsEntry_t));
9856 SET_MIB(ipst->ips_ip_mib.ipIfStatsAddrEntrySize,
9857 sizeof (mib2_ipAddrEntry_t));
9858 SET_MIB(ipst->ips_ip_mib.ipIfStatsRouteEntrySize,
9859 sizeof (mib2_ipRouteEntry_t));
9860 SET_MIB(ipst->ips_ip_mib.ipIfStatsNetToMediaEntrySize,
9861 sizeof (mib2_ipNetToMediaEntry_t));
9862 SET_MIB(ipst->ips_ip_mib.ipIfStatsMemberEntrySize,
9863 sizeof (ip_member_t));
9864 SET_MIB(ipst->ips_ip_mib.ipIfStatsGroupSourceEntrySize,
9865 sizeof (ip_grpsrc_t));
9866
9867 bcopy(&ipst->ips_ip_mib, &global_ip_mib, sizeof (global_ip_mib));
9868
9869 if (legacy_req) {
9870 SET_MIB(global_ip_mib.ipIfStatsAddrEntrySize,
9871 LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t));
9872 }
9873
9874 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
9875 (char *)&global_ip_mib, (int)sizeof (global_ip_mib))) {
9876 ip1dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9877 "failed to allocate %u bytes\n",
9878 (uint_t)sizeof (global_ip_mib)));
9879 }
9880
9881 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
9882 ill = ILL_START_WALK_V4(&ctx, ipst);
9883 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
9884 ill->ill_ip_mib->ipIfStatsIfIndex =
9885 ill->ill_phyint->phyint_ifindex;
9886 SET_MIB(ill->ill_ip_mib->ipIfStatsForwarding,
9887 (ipst->ips_ip_forwarding ? 1 : 2));
9888 SET_MIB(ill->ill_ip_mib->ipIfStatsDefaultTTL,
9889 (uint32_t)ipst->ips_ip_def_ttl);
9890
9891 ip_mib2_add_ip_stats(&global_ip_mib, ill->ill_ip_mib);
9892 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
9893 (char *)ill->ill_ip_mib,
9894 (int)sizeof (*ill->ill_ip_mib))) {
9895 ip1dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9896 "failed to allocate %u bytes\n",
9897 (uint_t)sizeof (*ill->ill_ip_mib)));
9898 }
9899 }
9900 rw_exit(&ipst->ips_ill_g_lock);
9901
9902 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9903 ip3dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9904 "level %d, name %d, len %d\n",
9905 (int)optp->level, (int)optp->name, (int)optp->len));
9906 qreply(q, mpctl);
9907
9908 if (mp2ctl == NULL)
9909 return (NULL);
9910
9911 return (ip_snmp_get_mib2_ip(q, mp2ctl, &global_ip_mib, ipst,
9912 legacy_req));
9913 }
9914
9915 /* Global IPv4 ICMP statistics */
9916 static mblk_t *
9917 ip_snmp_get_mib2_icmp(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9918 {
9919 struct opthdr *optp;
9920 mblk_t *mp2ctl;
9921
9922 /*
9923 * Make a copy of the original message
9924 */
9925 mp2ctl = copymsg(mpctl);
9926
9927 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9928 optp->level = MIB2_ICMP;
9929 optp->name = 0;
9930 if (!snmp_append_data(mpctl->b_cont, (char *)&ipst->ips_icmp_mib,
9931 (int)sizeof (ipst->ips_icmp_mib))) {
9932 ip1dbg(("ip_snmp_get_mib2_icmp: failed to allocate %u bytes\n",
9933 (uint_t)sizeof (ipst->ips_icmp_mib)));
9934 }
9935 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9936 ip3dbg(("ip_snmp_get_mib2_icmp: level %d, name %d, len %d\n",
9937 (int)optp->level, (int)optp->name, (int)optp->len));
9938 qreply(q, mpctl);
9939 return (mp2ctl);
9940 }
9941
9942 /* Global IPv4 IGMP statistics */
9943 static mblk_t *
9944 ip_snmp_get_mib2_igmp(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9945 {
9946 struct opthdr *optp;
9947 mblk_t *mp2ctl;
9948
9949 /*
9950 * make a copy of the original message
9951 */
9952 mp2ctl = copymsg(mpctl);
9953
9954 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9955 optp->level = EXPER_IGMP;
9956 optp->name = 0;
9957 if (!snmp_append_data(mpctl->b_cont, (char *)&ipst->ips_igmpstat,
9958 (int)sizeof (ipst->ips_igmpstat))) {
9959 ip1dbg(("ip_snmp_get_mib2_igmp: failed to allocate %u bytes\n",
9960 (uint_t)sizeof (ipst->ips_igmpstat)));
9961 }
9962 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9963 ip3dbg(("ip_snmp_get_mib2_igmp: level %d, name %d, len %d\n",
9964 (int)optp->level, (int)optp->name, (int)optp->len));
9965 qreply(q, mpctl);
9966 return (mp2ctl);
9967 }
9968
9969 /* Global IPv4 Multicast Routing statistics */
9970 static mblk_t *
9971 ip_snmp_get_mib2_multi(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9972 {
9973 struct opthdr *optp;
9974 mblk_t *mp2ctl;
9975
9976 /*
9977 * make a copy of the original message
9978 */
9979 mp2ctl = copymsg(mpctl);
9980
9981 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9982 optp->level = EXPER_DVMRP;
9983 optp->name = 0;
9984 if (!ip_mroute_stats(mpctl->b_cont, ipst)) {
9985 ip0dbg(("ip_mroute_stats: failed\n"));
9986 }
9987 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9988 ip3dbg(("ip_snmp_get_mib2_multi: level %d, name %d, len %d\n",
9989 (int)optp->level, (int)optp->name, (int)optp->len));
9990 qreply(q, mpctl);
9991 return (mp2ctl);
9992 }
9993
9994 /* IPv4 address information */
9995 static mblk_t *
9996 ip_snmp_get_mib2_ip_addr(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
9997 boolean_t legacy_req)
9998 {
9999 struct opthdr *optp;
10000 mblk_t *mp2ctl;
10001 mblk_t *mp_tail = NULL;
10002 ill_t *ill;
10003 ipif_t *ipif;
10004 uint_t bitval;
10005 mib2_ipAddrEntry_t mae;
10006 size_t mae_size;
10007 zoneid_t zoneid;
10008 ill_walk_context_t ctx;
10009
10010 /*
10011 * make a copy of the original message
10012 */
10013 mp2ctl = copymsg(mpctl);
10014
10015 mae_size = (legacy_req) ? LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t) :
10016 sizeof (mib2_ipAddrEntry_t);
10017
10018 /* ipAddrEntryTable */
10019
10020 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10021 optp->level = MIB2_IP;
10022 optp->name = MIB2_IP_ADDR;
10023 zoneid = Q_TO_CONN(q)->conn_zoneid;
10024
10025 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10026 ill = ILL_START_WALK_V4(&ctx, ipst);
10027 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10028 for (ipif = ill->ill_ipif; ipif != NULL;
10029 ipif = ipif->ipif_next) {
10030 if (ipif->ipif_zoneid != zoneid &&
10031 ipif->ipif_zoneid != ALL_ZONES)
10032 continue;
10033 /* Sum of count from dead IRE_LO* and our current */
10034 mae.ipAdEntInfo.ae_ibcnt = ipif->ipif_ib_pkt_count;
10035 if (ipif->ipif_ire_local != NULL) {
10036 mae.ipAdEntInfo.ae_ibcnt +=
10037 ipif->ipif_ire_local->ire_ib_pkt_count;
10038 }
10039 mae.ipAdEntInfo.ae_obcnt = 0;
10040 mae.ipAdEntInfo.ae_focnt = 0;
10041
10042 ipif_get_name(ipif, mae.ipAdEntIfIndex.o_bytes,
10043 OCTET_LENGTH);
10044 mae.ipAdEntIfIndex.o_length =
10045 mi_strlen(mae.ipAdEntIfIndex.o_bytes);
10046 mae.ipAdEntAddr = ipif->ipif_lcl_addr;
10047 mae.ipAdEntNetMask = ipif->ipif_net_mask;
10048 mae.ipAdEntInfo.ae_subnet = ipif->ipif_subnet;
10049 mae.ipAdEntInfo.ae_subnet_len =
10050 ip_mask_to_plen(ipif->ipif_net_mask);
10051 mae.ipAdEntInfo.ae_src_addr = ipif->ipif_lcl_addr;
10052 for (bitval = 1;
10053 bitval &&
10054 !(bitval & ipif->ipif_brd_addr);
10055 bitval <<= 1)
10056 noop;
10057 mae.ipAdEntBcastAddr = bitval;
10058 mae.ipAdEntReasmMaxSize = IP_MAXPACKET;
10059 mae.ipAdEntInfo.ae_mtu = ipif->ipif_ill->ill_mtu;
10060 mae.ipAdEntInfo.ae_metric = ipif->ipif_ill->ill_metric;
10061 mae.ipAdEntInfo.ae_broadcast_addr =
10062 ipif->ipif_brd_addr;
10063 mae.ipAdEntInfo.ae_pp_dst_addr =
10064 ipif->ipif_pp_dst_addr;
10065 mae.ipAdEntInfo.ae_flags = ipif->ipif_flags |
10066 ill->ill_flags | ill->ill_phyint->phyint_flags;
10067 mae.ipAdEntRetransmitTime =
10068 ill->ill_reachable_retrans_time;
10069
10070 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10071 (char *)&mae, (int)mae_size)) {
10072 ip1dbg(("ip_snmp_get_mib2_ip_addr: failed to "
10073 "allocate %u bytes\n", (uint_t)mae_size));
10074 }
10075 }
10076 }
10077 rw_exit(&ipst->ips_ill_g_lock);
10078
10079 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10080 ip3dbg(("ip_snmp_get_mib2_ip_addr: level %d, name %d, len %d\n",
10081 (int)optp->level, (int)optp->name, (int)optp->len));
10082 qreply(q, mpctl);
10083 return (mp2ctl);
10084 }
10085
10086 /* IPv6 address information */
10087 static mblk_t *
10088 ip_snmp_get_mib2_ip6_addr(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
10089 boolean_t legacy_req)
10090 {
10091 struct opthdr *optp;
10092 mblk_t *mp2ctl;
10093 mblk_t *mp_tail = NULL;
10094 ill_t *ill;
10095 ipif_t *ipif;
10096 mib2_ipv6AddrEntry_t mae6;
10097 size_t mae6_size;
10098 zoneid_t zoneid;
10099 ill_walk_context_t ctx;
10100
10101 /*
10102 * make a copy of the original message
10103 */
10104 mp2ctl = copymsg(mpctl);
10105
10106 mae6_size = (legacy_req) ?
10107 LEGACY_MIB_SIZE(&mae6, mib2_ipv6AddrEntry_t) :
10108 sizeof (mib2_ipv6AddrEntry_t);
10109
10110 /* ipv6AddrEntryTable */
10111
10112 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10113 optp->level = MIB2_IP6;
10114 optp->name = MIB2_IP6_ADDR;
10115 zoneid = Q_TO_CONN(q)->conn_zoneid;
10116
10117 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10118 ill = ILL_START_WALK_V6(&ctx, ipst);
10119 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10120 for (ipif = ill->ill_ipif; ipif != NULL;
10121 ipif = ipif->ipif_next) {
10122 if (ipif->ipif_zoneid != zoneid &&
10123 ipif->ipif_zoneid != ALL_ZONES)
10124 continue;
10125 /* Sum of count from dead IRE_LO* and our current */
10126 mae6.ipv6AddrInfo.ae_ibcnt = ipif->ipif_ib_pkt_count;
10127 if (ipif->ipif_ire_local != NULL) {
10128 mae6.ipv6AddrInfo.ae_ibcnt +=
10129 ipif->ipif_ire_local->ire_ib_pkt_count;
10130 }
10131 mae6.ipv6AddrInfo.ae_obcnt = 0;
10132 mae6.ipv6AddrInfo.ae_focnt = 0;
10133
10134 ipif_get_name(ipif, mae6.ipv6AddrIfIndex.o_bytes,
10135 OCTET_LENGTH);
10136 mae6.ipv6AddrIfIndex.o_length =
10137 mi_strlen(mae6.ipv6AddrIfIndex.o_bytes);
10138 mae6.ipv6AddrAddress = ipif->ipif_v6lcl_addr;
10139 mae6.ipv6AddrPfxLength =
10140 ip_mask_to_plen_v6(&ipif->ipif_v6net_mask);
10141 mae6.ipv6AddrInfo.ae_subnet = ipif->ipif_v6subnet;
10142 mae6.ipv6AddrInfo.ae_subnet_len =
10143 mae6.ipv6AddrPfxLength;
10144 mae6.ipv6AddrInfo.ae_src_addr = ipif->ipif_v6lcl_addr;
10145
10146 /* Type: stateless(1), stateful(2), unknown(3) */
10147 if (ipif->ipif_flags & IPIF_ADDRCONF)
10148 mae6.ipv6AddrType = 1;
10149 else
10150 mae6.ipv6AddrType = 2;
10151 /* Anycast: true(1), false(2) */
10152 if (ipif->ipif_flags & IPIF_ANYCAST)
10153 mae6.ipv6AddrAnycastFlag = 1;
10154 else
10155 mae6.ipv6AddrAnycastFlag = 2;
10156
10157 /*
10158 * Address status: preferred(1), deprecated(2),
10159 * invalid(3), inaccessible(4), unknown(5)
10160 */
10161 if (ipif->ipif_flags & IPIF_NOLOCAL)
10162 mae6.ipv6AddrStatus = 3;
10163 else if (ipif->ipif_flags & IPIF_DEPRECATED)
10164 mae6.ipv6AddrStatus = 2;
10165 else
10166 mae6.ipv6AddrStatus = 1;
10167 mae6.ipv6AddrInfo.ae_mtu = ipif->ipif_ill->ill_mtu;
10168 mae6.ipv6AddrInfo.ae_metric =
10169 ipif->ipif_ill->ill_metric;
10170 mae6.ipv6AddrInfo.ae_pp_dst_addr =
10171 ipif->ipif_v6pp_dst_addr;
10172 mae6.ipv6AddrInfo.ae_flags = ipif->ipif_flags |
10173 ill->ill_flags | ill->ill_phyint->phyint_flags;
10174 mae6.ipv6AddrReasmMaxSize = IP_MAXPACKET;
10175 mae6.ipv6AddrIdentifier = ill->ill_token;
10176 mae6.ipv6AddrIdentifierLen = ill->ill_token_length;
10177 mae6.ipv6AddrReachableTime = ill->ill_reachable_time;
10178 mae6.ipv6AddrRetransmitTime =
10179 ill->ill_reachable_retrans_time;
10180 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10181 (char *)&mae6, (int)mae6_size)) {
10182 ip1dbg(("ip_snmp_get_mib2_ip6_addr: failed to "
10183 "allocate %u bytes\n",
10184 (uint_t)mae6_size));
10185 }
10186 }
10187 }
10188 rw_exit(&ipst->ips_ill_g_lock);
10189
10190 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10191 ip3dbg(("ip_snmp_get_mib2_ip6_addr: level %d, name %d, len %d\n",
10192 (int)optp->level, (int)optp->name, (int)optp->len));
10193 qreply(q, mpctl);
10194 return (mp2ctl);
10195 }
10196
10197 /* IPv4 multicast group membership. */
10198 static mblk_t *
10199 ip_snmp_get_mib2_ip_group_mem(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10200 {
10201 struct opthdr *optp;
10202 mblk_t *mp2ctl;
10203 ill_t *ill;
10204 ipif_t *ipif;
10205 ilm_t *ilm;
10206 ip_member_t ipm;
10207 mblk_t *mp_tail = NULL;
10208 ill_walk_context_t ctx;
10209 zoneid_t zoneid;
10210
10211 /*
10212 * make a copy of the original message
10213 */
10214 mp2ctl = copymsg(mpctl);
10215 zoneid = Q_TO_CONN(q)->conn_zoneid;
10216
10217 /* ipGroupMember table */
10218 optp = (struct opthdr *)&mpctl->b_rptr[
10219 sizeof (struct T_optmgmt_ack)];
10220 optp->level = MIB2_IP;
10221 optp->name = EXPER_IP_GROUP_MEMBERSHIP;
10222
10223 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10224 ill = ILL_START_WALK_V4(&ctx, ipst);
10225 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10226 /* Make sure the ill isn't going away. */
10227 if (!ill_check_and_refhold(ill))
10228 continue;
10229 rw_exit(&ipst->ips_ill_g_lock);
10230 rw_enter(&ill->ill_mcast_lock, RW_READER);
10231 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10232 if (ilm->ilm_zoneid != zoneid &&
10233 ilm->ilm_zoneid != ALL_ZONES)
10234 continue;
10235
10236 /* Is there an ipif for ilm_ifaddr? */
10237 for (ipif = ill->ill_ipif; ipif != NULL;
10238 ipif = ipif->ipif_next) {
10239 if (!IPIF_IS_CONDEMNED(ipif) &&
10240 ipif->ipif_lcl_addr == ilm->ilm_ifaddr &&
10241 ilm->ilm_ifaddr != INADDR_ANY)
10242 break;
10243 }
10244 if (ipif != NULL) {
10245 ipif_get_name(ipif,
10246 ipm.ipGroupMemberIfIndex.o_bytes,
10247 OCTET_LENGTH);
10248 } else {
10249 ill_get_name(ill,
10250 ipm.ipGroupMemberIfIndex.o_bytes,
10251 OCTET_LENGTH);
10252 }
10253 ipm.ipGroupMemberIfIndex.o_length =
10254 mi_strlen(ipm.ipGroupMemberIfIndex.o_bytes);
10255
10256 ipm.ipGroupMemberAddress = ilm->ilm_addr;
10257 ipm.ipGroupMemberRefCnt = ilm->ilm_refcnt;
10258 ipm.ipGroupMemberFilterMode = ilm->ilm_fmode;
10259 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10260 (char *)&ipm, (int)sizeof (ipm))) {
10261 ip1dbg(("ip_snmp_get_mib2_ip_group: "
10262 "failed to allocate %u bytes\n",
10263 (uint_t)sizeof (ipm)));
10264 }
10265 }
10266 rw_exit(&ill->ill_mcast_lock);
10267 ill_refrele(ill);
10268 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10269 }
10270 rw_exit(&ipst->ips_ill_g_lock);
10271 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10272 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10273 (int)optp->level, (int)optp->name, (int)optp->len));
10274 qreply(q, mpctl);
10275 return (mp2ctl);
10276 }
10277
10278 /* IPv6 multicast group membership. */
10279 static mblk_t *
10280 ip_snmp_get_mib2_ip6_group_mem(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10281 {
10282 struct opthdr *optp;
10283 mblk_t *mp2ctl;
10284 ill_t *ill;
10285 ilm_t *ilm;
10286 ipv6_member_t ipm6;
10287 mblk_t *mp_tail = NULL;
10288 ill_walk_context_t ctx;
10289 zoneid_t zoneid;
10290
10291 /*
10292 * make a copy of the original message
10293 */
10294 mp2ctl = copymsg(mpctl);
10295 zoneid = Q_TO_CONN(q)->conn_zoneid;
10296
10297 /* ip6GroupMember table */
10298 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10299 optp->level = MIB2_IP6;
10300 optp->name = EXPER_IP6_GROUP_MEMBERSHIP;
10301
10302 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10303 ill = ILL_START_WALK_V6(&ctx, ipst);
10304 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10305 /* Make sure the ill isn't going away. */
10306 if (!ill_check_and_refhold(ill))
10307 continue;
10308 rw_exit(&ipst->ips_ill_g_lock);
10309 /*
10310 * Normally we don't have any members on under IPMP interfaces.
10311 * We report them as a debugging aid.
10312 */
10313 rw_enter(&ill->ill_mcast_lock, RW_READER);
10314 ipm6.ipv6GroupMemberIfIndex = ill->ill_phyint->phyint_ifindex;
10315 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10316 if (ilm->ilm_zoneid != zoneid &&
10317 ilm->ilm_zoneid != ALL_ZONES)
10318 continue; /* not this zone */
10319 ipm6.ipv6GroupMemberAddress = ilm->ilm_v6addr;
10320 ipm6.ipv6GroupMemberRefCnt = ilm->ilm_refcnt;
10321 ipm6.ipv6GroupMemberFilterMode = ilm->ilm_fmode;
10322 if (!snmp_append_data2(mpctl->b_cont,
10323 &mp_tail,
10324 (char *)&ipm6, (int)sizeof (ipm6))) {
10325 ip1dbg(("ip_snmp_get_mib2_ip6_group: "
10326 "failed to allocate %u bytes\n",
10327 (uint_t)sizeof (ipm6)));
10328 }
10329 }
10330 rw_exit(&ill->ill_mcast_lock);
10331 ill_refrele(ill);
10332 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10333 }
10334 rw_exit(&ipst->ips_ill_g_lock);
10335
10336 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10337 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10338 (int)optp->level, (int)optp->name, (int)optp->len));
10339 qreply(q, mpctl);
10340 return (mp2ctl);
10341 }
10342
10343 /* IP multicast filtered sources */
10344 static mblk_t *
10345 ip_snmp_get_mib2_ip_group_src(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10346 {
10347 struct opthdr *optp;
10348 mblk_t *mp2ctl;
10349 ill_t *ill;
10350 ipif_t *ipif;
10351 ilm_t *ilm;
10352 ip_grpsrc_t ips;
10353 mblk_t *mp_tail = NULL;
10354 ill_walk_context_t ctx;
10355 zoneid_t zoneid;
10356 int i;
10357 slist_t *sl;
10358
10359 /*
10360 * make a copy of the original message
10361 */
10362 mp2ctl = copymsg(mpctl);
10363 zoneid = Q_TO_CONN(q)->conn_zoneid;
10364
10365 /* ipGroupSource table */
10366 optp = (struct opthdr *)&mpctl->b_rptr[
10367 sizeof (struct T_optmgmt_ack)];
10368 optp->level = MIB2_IP;
10369 optp->name = EXPER_IP_GROUP_SOURCES;
10370
10371 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10372 ill = ILL_START_WALK_V4(&ctx, ipst);
10373 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10374 /* Make sure the ill isn't going away. */
10375 if (!ill_check_and_refhold(ill))
10376 continue;
10377 rw_exit(&ipst->ips_ill_g_lock);
10378 rw_enter(&ill->ill_mcast_lock, RW_READER);
10379 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10380 sl = ilm->ilm_filter;
10381 if (ilm->ilm_zoneid != zoneid &&
10382 ilm->ilm_zoneid != ALL_ZONES)
10383 continue;
10384 if (SLIST_IS_EMPTY(sl))
10385 continue;
10386
10387 /* Is there an ipif for ilm_ifaddr? */
10388 for (ipif = ill->ill_ipif; ipif != NULL;
10389 ipif = ipif->ipif_next) {
10390 if (!IPIF_IS_CONDEMNED(ipif) &&
10391 ipif->ipif_lcl_addr == ilm->ilm_ifaddr &&
10392 ilm->ilm_ifaddr != INADDR_ANY)
10393 break;
10394 }
10395 if (ipif != NULL) {
10396 ipif_get_name(ipif,
10397 ips.ipGroupSourceIfIndex.o_bytes,
10398 OCTET_LENGTH);
10399 } else {
10400 ill_get_name(ill,
10401 ips.ipGroupSourceIfIndex.o_bytes,
10402 OCTET_LENGTH);
10403 }
10404 ips.ipGroupSourceIfIndex.o_length =
10405 mi_strlen(ips.ipGroupSourceIfIndex.o_bytes);
10406
10407 ips.ipGroupSourceGroup = ilm->ilm_addr;
10408 for (i = 0; i < sl->sl_numsrc; i++) {
10409 if (!IN6_IS_ADDR_V4MAPPED(&sl->sl_addr[i]))
10410 continue;
10411 IN6_V4MAPPED_TO_IPADDR(&sl->sl_addr[i],
10412 ips.ipGroupSourceAddress);
10413 if (snmp_append_data2(mpctl->b_cont, &mp_tail,
10414 (char *)&ips, (int)sizeof (ips)) == 0) {
10415 ip1dbg(("ip_snmp_get_mib2_ip_group_src:"
10416 " failed to allocate %u bytes\n",
10417 (uint_t)sizeof (ips)));
10418 }
10419 }
10420 }
10421 rw_exit(&ill->ill_mcast_lock);
10422 ill_refrele(ill);
10423 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10424 }
10425 rw_exit(&ipst->ips_ill_g_lock);
10426 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10427 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10428 (int)optp->level, (int)optp->name, (int)optp->len));
10429 qreply(q, mpctl);
10430 return (mp2ctl);
10431 }
10432
10433 /* IPv6 multicast filtered sources. */
10434 static mblk_t *
10435 ip_snmp_get_mib2_ip6_group_src(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10436 {
10437 struct opthdr *optp;
10438 mblk_t *mp2ctl;
10439 ill_t *ill;
10440 ilm_t *ilm;
10441 ipv6_grpsrc_t ips6;
10442 mblk_t *mp_tail = NULL;
10443 ill_walk_context_t ctx;
10444 zoneid_t zoneid;
10445 int i;
10446 slist_t *sl;
10447
10448 /*
10449 * make a copy of the original message
10450 */
10451 mp2ctl = copymsg(mpctl);
10452 zoneid = Q_TO_CONN(q)->conn_zoneid;
10453
10454 /* ip6GroupMember table */
10455 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10456 optp->level = MIB2_IP6;
10457 optp->name = EXPER_IP6_GROUP_SOURCES;
10458
10459 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10460 ill = ILL_START_WALK_V6(&ctx, ipst);
10461 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10462 /* Make sure the ill isn't going away. */
10463 if (!ill_check_and_refhold(ill))
10464 continue;
10465 rw_exit(&ipst->ips_ill_g_lock);
10466 /*
10467 * Normally we don't have any members on under IPMP interfaces.
10468 * We report them as a debugging aid.
10469 */
10470 rw_enter(&ill->ill_mcast_lock, RW_READER);
10471 ips6.ipv6GroupSourceIfIndex = ill->ill_phyint->phyint_ifindex;
10472 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10473 sl = ilm->ilm_filter;
10474 if (ilm->ilm_zoneid != zoneid &&
10475 ilm->ilm_zoneid != ALL_ZONES)
10476 continue;
10477 if (SLIST_IS_EMPTY(sl))
10478 continue;
10479 ips6.ipv6GroupSourceGroup = ilm->ilm_v6addr;
10480 for (i = 0; i < sl->sl_numsrc; i++) {
10481 ips6.ipv6GroupSourceAddress = sl->sl_addr[i];
10482 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10483 (char *)&ips6, (int)sizeof (ips6))) {
10484 ip1dbg(("ip_snmp_get_mib2_ip6_"
10485 "group_src: failed to allocate "
10486 "%u bytes\n",
10487 (uint_t)sizeof (ips6)));
10488 }
10489 }
10490 }
10491 rw_exit(&ill->ill_mcast_lock);
10492 ill_refrele(ill);
10493 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10494 }
10495 rw_exit(&ipst->ips_ill_g_lock);
10496
10497 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10498 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10499 (int)optp->level, (int)optp->name, (int)optp->len));
10500 qreply(q, mpctl);
10501 return (mp2ctl);
10502 }
10503
10504 /* Multicast routing virtual interface table. */
10505 static mblk_t *
10506 ip_snmp_get_mib2_virt_multi(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10507 {
10508 struct opthdr *optp;
10509 mblk_t *mp2ctl;
10510
10511 /*
10512 * make a copy of the original message
10513 */
10514 mp2ctl = copymsg(mpctl);
10515
10516 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10517 optp->level = EXPER_DVMRP;
10518 optp->name = EXPER_DVMRP_VIF;
10519 if (!ip_mroute_vif(mpctl->b_cont, ipst)) {
10520 ip0dbg(("ip_mroute_vif: failed\n"));
10521 }
10522 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10523 ip3dbg(("ip_snmp_get_mib2_virt_multi: level %d, name %d, len %d\n",
10524 (int)optp->level, (int)optp->name, (int)optp->len));
10525 qreply(q, mpctl);
10526 return (mp2ctl);
10527 }
10528
10529 /* Multicast routing table. */
10530 static mblk_t *
10531 ip_snmp_get_mib2_multi_rtable(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10532 {
10533 struct opthdr *optp;
10534 mblk_t *mp2ctl;
10535
10536 /*
10537 * make a copy of the original message
10538 */
10539 mp2ctl = copymsg(mpctl);
10540
10541 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10542 optp->level = EXPER_DVMRP;
10543 optp->name = EXPER_DVMRP_MRT;
10544 if (!ip_mroute_mrt(mpctl->b_cont, ipst)) {
10545 ip0dbg(("ip_mroute_mrt: failed\n"));
10546 }
10547 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10548 ip3dbg(("ip_snmp_get_mib2_multi_rtable: level %d, name %d, len %d\n",
10549 (int)optp->level, (int)optp->name, (int)optp->len));
10550 qreply(q, mpctl);
10551 return (mp2ctl);
10552 }
10553
10554 /*
10555 * Return ipRouteEntryTable, ipNetToMediaEntryTable, and ipRouteAttributeTable
10556 * in one IRE walk.
10557 */
10558 static mblk_t *
10559 ip_snmp_get_mib2_ip_route_media(queue_t *q, mblk_t *mpctl, int level,
10560 ip_stack_t *ipst)
10561 {
10562 struct opthdr *optp;
10563 mblk_t *mp2ctl; /* Returned */
10564 mblk_t *mp3ctl; /* nettomedia */
10565 mblk_t *mp4ctl; /* routeattrs */
10566 iproutedata_t ird;
10567 zoneid_t zoneid;
10568
10569 /*
10570 * make copies of the original message
10571 * - mp2ctl is returned unchanged to the caller for its use
10572 * - mpctl is sent upstream as ipRouteEntryTable
10573 * - mp3ctl is sent upstream as ipNetToMediaEntryTable
10574 * - mp4ctl is sent upstream as ipRouteAttributeTable
10575 */
10576 mp2ctl = copymsg(mpctl);
10577 mp3ctl = copymsg(mpctl);
10578 mp4ctl = copymsg(mpctl);
10579 if (mp3ctl == NULL || mp4ctl == NULL) {
10580 freemsg(mp4ctl);
10581 freemsg(mp3ctl);
10582 freemsg(mp2ctl);
10583 freemsg(mpctl);
10584 return (NULL);
10585 }
10586
10587 bzero(&ird, sizeof (ird));
10588
10589 ird.ird_route.lp_head = mpctl->b_cont;
10590 ird.ird_netmedia.lp_head = mp3ctl->b_cont;
10591 ird.ird_attrs.lp_head = mp4ctl->b_cont;
10592 /*
10593 * If the level has been set the special EXPER_IP_AND_ALL_IRES value,
10594 * then also include ire_testhidden IREs and IRE_IF_CLONE. This is
10595 * intended a temporary solution until a proper MIB API is provided
10596 * that provides complete filtering/caller-opt-in.
10597 */
10598 if (level == EXPER_IP_AND_ALL_IRES)
10599 ird.ird_flags |= IRD_REPORT_ALL;
10600
10601 zoneid = Q_TO_CONN(q)->conn_zoneid;
10602 ire_walk_v4(ip_snmp_get2_v4, &ird, zoneid, ipst);
10603
10604 /* ipRouteEntryTable in mpctl */
10605 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10606 optp->level = MIB2_IP;
10607 optp->name = MIB2_IP_ROUTE;
10608 optp->len = msgdsize(ird.ird_route.lp_head);
10609 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10610 (int)optp->level, (int)optp->name, (int)optp->len));
10611 qreply(q, mpctl);
10612
10613 /* ipNetToMediaEntryTable in mp3ctl */
10614 ncec_walk(NULL, ip_snmp_get2_v4_media, &ird, ipst);
10615
10616 optp = (struct opthdr *)&mp3ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10617 optp->level = MIB2_IP;
10618 optp->name = MIB2_IP_MEDIA;
10619 optp->len = msgdsize(ird.ird_netmedia.lp_head);
10620 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10621 (int)optp->level, (int)optp->name, (int)optp->len));
10622 qreply(q, mp3ctl);
10623
10624 /* ipRouteAttributeTable in mp4ctl */
10625 optp = (struct opthdr *)&mp4ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10626 optp->level = MIB2_IP;
10627 optp->name = EXPER_IP_RTATTR;
10628 optp->len = msgdsize(ird.ird_attrs.lp_head);
10629 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10630 (int)optp->level, (int)optp->name, (int)optp->len));
10631 if (optp->len == 0)
10632 freemsg(mp4ctl);
10633 else
10634 qreply(q, mp4ctl);
10635
10636 return (mp2ctl);
10637 }
10638
10639 /*
10640 * Return ipv6RouteEntryTable and ipv6RouteAttributeTable in one IRE walk, and
10641 * ipv6NetToMediaEntryTable in an NDP walk.
10642 */
10643 static mblk_t *
10644 ip_snmp_get_mib2_ip6_route_media(queue_t *q, mblk_t *mpctl, int level,
10645 ip_stack_t *ipst)
10646 {
10647 struct opthdr *optp;
10648 mblk_t *mp2ctl; /* Returned */
10649 mblk_t *mp3ctl; /* nettomedia */
10650 mblk_t *mp4ctl; /* routeattrs */
10651 iproutedata_t ird;
10652 zoneid_t zoneid;
10653
10654 /*
10655 * make copies of the original message
10656 * - mp2ctl is returned unchanged to the caller for its use
10657 * - mpctl is sent upstream as ipv6RouteEntryTable
10658 * - mp3ctl is sent upstream as ipv6NetToMediaEntryTable
10659 * - mp4ctl is sent upstream as ipv6RouteAttributeTable
10660 */
10661 mp2ctl = copymsg(mpctl);
10662 mp3ctl = copymsg(mpctl);
10663 mp4ctl = copymsg(mpctl);
10664 if (mp3ctl == NULL || mp4ctl == NULL) {
10665 freemsg(mp4ctl);
10666 freemsg(mp3ctl);
10667 freemsg(mp2ctl);
10668 freemsg(mpctl);
10669 return (NULL);
10670 }
10671
10672 bzero(&ird, sizeof (ird));
10673
10674 ird.ird_route.lp_head = mpctl->b_cont;
10675 ird.ird_netmedia.lp_head = mp3ctl->b_cont;
10676 ird.ird_attrs.lp_head = mp4ctl->b_cont;
10677 /*
10678 * If the level has been set the special EXPER_IP_AND_ALL_IRES value,
10679 * then also include ire_testhidden IREs and IRE_IF_CLONE. This is
10680 * intended a temporary solution until a proper MIB API is provided
10681 * that provides complete filtering/caller-opt-in.
10682 */
10683 if (level == EXPER_IP_AND_ALL_IRES)
10684 ird.ird_flags |= IRD_REPORT_ALL;
10685
10686 zoneid = Q_TO_CONN(q)->conn_zoneid;
10687 ire_walk_v6(ip_snmp_get2_v6_route, &ird, zoneid, ipst);
10688
10689 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10690 optp->level = MIB2_IP6;
10691 optp->name = MIB2_IP6_ROUTE;
10692 optp->len = msgdsize(ird.ird_route.lp_head);
10693 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10694 (int)optp->level, (int)optp->name, (int)optp->len));
10695 qreply(q, mpctl);
10696
10697 /* ipv6NetToMediaEntryTable in mp3ctl */
10698 ncec_walk(NULL, ip_snmp_get2_v6_media, &ird, ipst);
10699
10700 optp = (struct opthdr *)&mp3ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10701 optp->level = MIB2_IP6;
10702 optp->name = MIB2_IP6_MEDIA;
10703 optp->len = msgdsize(ird.ird_netmedia.lp_head);
10704 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10705 (int)optp->level, (int)optp->name, (int)optp->len));
10706 qreply(q, mp3ctl);
10707
10708 /* ipv6RouteAttributeTable in mp4ctl */
10709 optp = (struct opthdr *)&mp4ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10710 optp->level = MIB2_IP6;
10711 optp->name = EXPER_IP_RTATTR;
10712 optp->len = msgdsize(ird.ird_attrs.lp_head);
10713 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10714 (int)optp->level, (int)optp->name, (int)optp->len));
10715 if (optp->len == 0)
10716 freemsg(mp4ctl);
10717 else
10718 qreply(q, mp4ctl);
10719
10720 return (mp2ctl);
10721 }
10722
10723 /*
10724 * IPv6 mib: One per ill
10725 */
10726 static mblk_t *
10727 ip_snmp_get_mib2_ip6(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
10728 boolean_t legacy_req)
10729 {
10730 struct opthdr *optp;
10731 mblk_t *mp2ctl;
10732 ill_t *ill;
10733 ill_walk_context_t ctx;
10734 mblk_t *mp_tail = NULL;
10735 mib2_ipv6AddrEntry_t mae6;
10736 mib2_ipIfStatsEntry_t *ise;
10737 size_t ise_size, iae_size;
10738
10739 /*
10740 * Make a copy of the original message
10741 */
10742 mp2ctl = copymsg(mpctl);
10743
10744 /* fixed length IPv6 structure ... */
10745
10746 if (legacy_req) {
10747 ise_size = LEGACY_MIB_SIZE(&ipst->ips_ip6_mib,
10748 mib2_ipIfStatsEntry_t);
10749 iae_size = LEGACY_MIB_SIZE(&mae6, mib2_ipv6AddrEntry_t);
10750 } else {
10751 ise_size = sizeof (mib2_ipIfStatsEntry_t);
10752 iae_size = sizeof (mib2_ipv6AddrEntry_t);
10753 }
10754
10755 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10756 optp->level = MIB2_IP6;
10757 optp->name = 0;
10758 /* Include "unknown interface" ip6_mib */
10759 ipst->ips_ip6_mib.ipIfStatsIPVersion = MIB2_INETADDRESSTYPE_ipv6;
10760 ipst->ips_ip6_mib.ipIfStatsIfIndex =
10761 MIB2_UNKNOWN_INTERFACE; /* Flag to netstat */
10762 SET_MIB(ipst->ips_ip6_mib.ipIfStatsForwarding,
10763 ipst->ips_ipv6_forwarding ? 1 : 2);
10764 SET_MIB(ipst->ips_ip6_mib.ipIfStatsDefaultHopLimit,
10765 ipst->ips_ipv6_def_hops);
10766 SET_MIB(ipst->ips_ip6_mib.ipIfStatsEntrySize,
10767 sizeof (mib2_ipIfStatsEntry_t));
10768 SET_MIB(ipst->ips_ip6_mib.ipIfStatsAddrEntrySize,
10769 sizeof (mib2_ipv6AddrEntry_t));
10770 SET_MIB(ipst->ips_ip6_mib.ipIfStatsRouteEntrySize,
10771 sizeof (mib2_ipv6RouteEntry_t));
10772 SET_MIB(ipst->ips_ip6_mib.ipIfStatsNetToMediaEntrySize,
10773 sizeof (mib2_ipv6NetToMediaEntry_t));
10774 SET_MIB(ipst->ips_ip6_mib.ipIfStatsMemberEntrySize,
10775 sizeof (ipv6_member_t));
10776 SET_MIB(ipst->ips_ip6_mib.ipIfStatsGroupSourceEntrySize,
10777 sizeof (ipv6_grpsrc_t));
10778
10779 /*
10780 * Synchronize 64- and 32-bit counters
10781 */
10782 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInReceives,
10783 ipIfStatsHCInReceives);
10784 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInDelivers,
10785 ipIfStatsHCInDelivers);
10786 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutRequests,
10787 ipIfStatsHCOutRequests);
10788 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutForwDatagrams,
10789 ipIfStatsHCOutForwDatagrams);
10790 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutMcastPkts,
10791 ipIfStatsHCOutMcastPkts);
10792 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInMcastPkts,
10793 ipIfStatsHCInMcastPkts);
10794
10795 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10796 (char *)&ipst->ips_ip6_mib, (int)ise_size)) {
10797 ip1dbg(("ip_snmp_get_mib2_ip6: failed to allocate %u bytes\n",
10798 (uint_t)ise_size));
10799 } else if (legacy_req) {
10800 /* Adjust the EntrySize fields for legacy requests. */
10801 ise =
10802 (mib2_ipIfStatsEntry_t *)(mp_tail->b_wptr - (int)ise_size);
10803 SET_MIB(ise->ipIfStatsEntrySize, ise_size);
10804 SET_MIB(ise->ipIfStatsAddrEntrySize, iae_size);
10805 }
10806
10807 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10808 ill = ILL_START_WALK_V6(&ctx, ipst);
10809 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10810 ill->ill_ip_mib->ipIfStatsIfIndex =
10811 ill->ill_phyint->phyint_ifindex;
10812 SET_MIB(ill->ill_ip_mib->ipIfStatsForwarding,
10813 ipst->ips_ipv6_forwarding ? 1 : 2);
10814 SET_MIB(ill->ill_ip_mib->ipIfStatsDefaultHopLimit,
10815 ill->ill_max_hops);
10816
10817 /*
10818 * Synchronize 64- and 32-bit counters
10819 */
10820 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInReceives,
10821 ipIfStatsHCInReceives);
10822 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInDelivers,
10823 ipIfStatsHCInDelivers);
10824 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutRequests,
10825 ipIfStatsHCOutRequests);
10826 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutForwDatagrams,
10827 ipIfStatsHCOutForwDatagrams);
10828 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutMcastPkts,
10829 ipIfStatsHCOutMcastPkts);
10830 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInMcastPkts,
10831 ipIfStatsHCInMcastPkts);
10832
10833 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10834 (char *)ill->ill_ip_mib, (int)ise_size)) {
10835 ip1dbg(("ip_snmp_get_mib2_ip6: failed to allocate "
10836 "%u bytes\n", (uint_t)ise_size));
10837 } else if (legacy_req) {
10838 /* Adjust the EntrySize fields for legacy requests. */
10839 ise = (mib2_ipIfStatsEntry_t *)(mp_tail->b_wptr -
10840 (int)ise_size);
10841 SET_MIB(ise->ipIfStatsEntrySize, ise_size);
10842 SET_MIB(ise->ipIfStatsAddrEntrySize, iae_size);
10843 }
10844 }
10845 rw_exit(&ipst->ips_ill_g_lock);
10846
10847 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10848 ip3dbg(("ip_snmp_get_mib2_ip6: level %d, name %d, len %d\n",
10849 (int)optp->level, (int)optp->name, (int)optp->len));
10850 qreply(q, mpctl);
10851 return (mp2ctl);
10852 }
10853
10854 /*
10855 * ICMPv6 mib: One per ill
10856 */
10857 static mblk_t *
10858 ip_snmp_get_mib2_icmp6(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10859 {
10860 struct opthdr *optp;
10861 mblk_t *mp2ctl;
10862 ill_t *ill;
10863 ill_walk_context_t ctx;
10864 mblk_t *mp_tail = NULL;
10865 /*
10866 * Make a copy of the original message
10867 */
10868 mp2ctl = copymsg(mpctl);
10869
10870 /* fixed length ICMPv6 structure ... */
10871
10872 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10873 optp->level = MIB2_ICMP6;
10874 optp->name = 0;
10875 /* Include "unknown interface" icmp6_mib */
10876 ipst->ips_icmp6_mib.ipv6IfIcmpIfIndex =
10877 MIB2_UNKNOWN_INTERFACE; /* netstat flag */
10878 ipst->ips_icmp6_mib.ipv6IfIcmpEntrySize =
10879 sizeof (mib2_ipv6IfIcmpEntry_t);
10880 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10881 (char *)&ipst->ips_icmp6_mib,
10882 (int)sizeof (ipst->ips_icmp6_mib))) {
10883 ip1dbg(("ip_snmp_get_mib2_icmp6: failed to allocate %u bytes\n",
10884 (uint_t)sizeof (ipst->ips_icmp6_mib)));
10885 }
10886
10887 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10888 ill = ILL_START_WALK_V6(&ctx, ipst);
10889 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10890 ill->ill_icmp6_mib->ipv6IfIcmpIfIndex =
10891 ill->ill_phyint->phyint_ifindex;
10892 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10893 (char *)ill->ill_icmp6_mib,
10894 (int)sizeof (*ill->ill_icmp6_mib))) {
10895 ip1dbg(("ip_snmp_get_mib2_icmp6: failed to allocate "
10896 "%u bytes\n",
10897 (uint_t)sizeof (*ill->ill_icmp6_mib)));
10898 }
10899 }
10900 rw_exit(&ipst->ips_ill_g_lock);
10901
10902 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10903 ip3dbg(("ip_snmp_get_mib2_icmp6: level %d, name %d, len %d\n",
10904 (int)optp->level, (int)optp->name, (int)optp->len));
10905 qreply(q, mpctl);
10906 return (mp2ctl);
10907 }
10908
10909 /*
10910 * ire_walk routine to create both ipRouteEntryTable and
10911 * ipRouteAttributeTable in one IRE walk
10912 */
10913 static void
10914 ip_snmp_get2_v4(ire_t *ire, iproutedata_t *ird)
10915 {
10916 ill_t *ill;
10917 mib2_ipRouteEntry_t *re;
10918 mib2_ipAttributeEntry_t iaes;
10919 tsol_ire_gw_secattr_t *attrp;
10920 tsol_gc_t *gc = NULL;
10921 tsol_gcgrp_t *gcgrp = NULL;
10922 ip_stack_t *ipst = ire->ire_ipst;
10923
10924 ASSERT(ire->ire_ipversion == IPV4_VERSION);
10925
10926 if (!(ird->ird_flags & IRD_REPORT_ALL)) {
10927 if (ire->ire_testhidden)
10928 return;
10929 if (ire->ire_type & IRE_IF_CLONE)
10930 return;
10931 }
10932
10933 if ((re = kmem_zalloc(sizeof (*re), KM_NOSLEEP)) == NULL)
10934 return;
10935
10936 if ((attrp = ire->ire_gw_secattr) != NULL) {
10937 mutex_enter(&attrp->igsa_lock);
10938 if ((gc = attrp->igsa_gc) != NULL) {
10939 gcgrp = gc->gc_grp;
10940 ASSERT(gcgrp != NULL);
10941 rw_enter(&gcgrp->gcgrp_rwlock, RW_READER);
10942 }
10943 mutex_exit(&attrp->igsa_lock);
10944 }
10945 /*
10946 * Return all IRE types for route table... let caller pick and choose
10947 */
10948 re->ipRouteDest = ire->ire_addr;
10949 ill = ire->ire_ill;
10950 re->ipRouteIfIndex.o_length = 0;
10951 if (ill != NULL) {
10952 ill_get_name(ill, re->ipRouteIfIndex.o_bytes, OCTET_LENGTH);
10953 re->ipRouteIfIndex.o_length =
10954 mi_strlen(re->ipRouteIfIndex.o_bytes);
10955 }
10956 re->ipRouteMetric1 = -1;
10957 re->ipRouteMetric2 = -1;
10958 re->ipRouteMetric3 = -1;
10959 re->ipRouteMetric4 = -1;
10960
10961 re->ipRouteNextHop = ire->ire_gateway_addr;
10962 /* indirect(4), direct(3), or invalid(2) */
10963 if (ire->ire_flags & (RTF_REJECT | RTF_BLACKHOLE))
10964 re->ipRouteType = 2;
10965 else if (ire->ire_type & IRE_ONLINK)
10966 re->ipRouteType = 3;
10967 else
10968 re->ipRouteType = 4;
10969
10970 re->ipRouteProto = -1;
10971 re->ipRouteAge = gethrestime_sec() - ire->ire_create_time;
10972 re->ipRouteMask = ire->ire_mask;
10973 re->ipRouteMetric5 = -1;
10974 re->ipRouteInfo.re_max_frag = ire->ire_metrics.iulp_mtu;
10975 if (ire->ire_ill != NULL && re->ipRouteInfo.re_max_frag == 0)
10976 re->ipRouteInfo.re_max_frag = ire->ire_ill->ill_mtu;
10977
10978 re->ipRouteInfo.re_frag_flag = 0;
10979 re->ipRouteInfo.re_rtt = 0;
10980 re->ipRouteInfo.re_src_addr = 0;
10981 re->ipRouteInfo.re_ref = ire->ire_refcnt;
10982 re->ipRouteInfo.re_obpkt = ire->ire_ob_pkt_count;
10983 re->ipRouteInfo.re_ibpkt = ire->ire_ib_pkt_count;
10984 re->ipRouteInfo.re_flags = ire->ire_flags;
10985
10986 /* Add the IRE_IF_CLONE's counters to their parent IRE_INTERFACE */
10987 if (ire->ire_type & IRE_INTERFACE) {
10988 ire_t *child;
10989
10990 rw_enter(&ipst->ips_ire_dep_lock, RW_READER);
10991 child = ire->ire_dep_children;
10992 while (child != NULL) {
10993 re->ipRouteInfo.re_obpkt += child->ire_ob_pkt_count;
10994 re->ipRouteInfo.re_ibpkt += child->ire_ib_pkt_count;
10995 child = child->ire_dep_sib_next;
10996 }
10997 rw_exit(&ipst->ips_ire_dep_lock);
10998 }
10999
11000 if (ire->ire_flags & RTF_DYNAMIC) {
11001 re->ipRouteInfo.re_ire_type = IRE_HOST_REDIRECT;
11002 } else {
11003 re->ipRouteInfo.re_ire_type = ire->ire_type;
11004 }
11005
11006 if (!snmp_append_data2(ird->ird_route.lp_head, &ird->ird_route.lp_tail,
11007 (char *)re, (int)sizeof (*re))) {
11008 ip1dbg(("ip_snmp_get2_v4: failed to allocate %u bytes\n",
11009 (uint_t)sizeof (*re)));
11010 }
11011
11012 if (gc != NULL) {
11013 iaes.iae_routeidx = ird->ird_idx;
11014 iaes.iae_doi = gc->gc_db->gcdb_doi;
11015 iaes.iae_slrange = gc->gc_db->gcdb_slrange;
11016
11017 if (!snmp_append_data2(ird->ird_attrs.lp_head,
11018 &ird->ird_attrs.lp_tail, (char *)&iaes, sizeof (iaes))) {
11019 ip1dbg(("ip_snmp_get2_v4: failed to allocate %u "
11020 "bytes\n", (uint_t)sizeof (iaes)));
11021 }
11022 }
11023
11024 /* bump route index for next pass */
11025 ird->ird_idx++;
11026
11027 kmem_free(re, sizeof (*re));
11028 if (gcgrp != NULL)
11029 rw_exit(&gcgrp->gcgrp_rwlock);
11030 }
11031
11032 /*
11033 * ire_walk routine to create ipv6RouteEntryTable and ipRouteEntryTable.
11034 */
11035 static void
11036 ip_snmp_get2_v6_route(ire_t *ire, iproutedata_t *ird)
11037 {
11038 ill_t *ill;
11039 mib2_ipv6RouteEntry_t *re;
11040 mib2_ipAttributeEntry_t iaes;
11041 tsol_ire_gw_secattr_t *attrp;
11042 tsol_gc_t *gc = NULL;
11043 tsol_gcgrp_t *gcgrp = NULL;
11044 ip_stack_t *ipst = ire->ire_ipst;
11045
11046 ASSERT(ire->ire_ipversion == IPV6_VERSION);
11047
11048 if (!(ird->ird_flags & IRD_REPORT_ALL)) {
11049 if (ire->ire_testhidden)
11050 return;
11051 if (ire->ire_type & IRE_IF_CLONE)
11052 return;
11053 }
11054
11055 if ((re = kmem_zalloc(sizeof (*re), KM_NOSLEEP)) == NULL)
11056 return;
11057
11058 if ((attrp = ire->ire_gw_secattr) != NULL) {
11059 mutex_enter(&attrp->igsa_lock);
11060 if ((gc = attrp->igsa_gc) != NULL) {
11061 gcgrp = gc->gc_grp;
11062 ASSERT(gcgrp != NULL);
11063 rw_enter(&gcgrp->gcgrp_rwlock, RW_READER);
11064 }
11065 mutex_exit(&attrp->igsa_lock);
11066 }
11067 /*
11068 * Return all IRE types for route table... let caller pick and choose
11069 */
11070 re->ipv6RouteDest = ire->ire_addr_v6;
11071 re->ipv6RoutePfxLength = ip_mask_to_plen_v6(&ire->ire_mask_v6);
11072 re->ipv6RouteIndex = 0; /* Unique when multiple with same dest/plen */
11073 re->ipv6RouteIfIndex.o_length = 0;
11074 ill = ire->ire_ill;
11075 if (ill != NULL) {
11076 ill_get_name(ill, re->ipv6RouteIfIndex.o_bytes, OCTET_LENGTH);
11077 re->ipv6RouteIfIndex.o_length =
11078 mi_strlen(re->ipv6RouteIfIndex.o_bytes);
11079 }
11080
11081 ASSERT(!(ire->ire_type & IRE_BROADCAST));
11082
11083 mutex_enter(&ire->ire_lock);
11084 re->ipv6RouteNextHop = ire->ire_gateway_addr_v6;
11085 mutex_exit(&ire->ire_lock);
11086
11087 /* remote(4), local(3), or discard(2) */
11088 if (ire->ire_flags & (RTF_REJECT | RTF_BLACKHOLE))
11089 re->ipv6RouteType = 2;
11090 else if (ire->ire_type & IRE_ONLINK)
11091 re->ipv6RouteType = 3;
11092 else
11093 re->ipv6RouteType = 4;
11094
11095 re->ipv6RouteProtocol = -1;
11096 re->ipv6RoutePolicy = 0;
11097 re->ipv6RouteAge = gethrestime_sec() - ire->ire_create_time;
11098 re->ipv6RouteNextHopRDI = 0;
11099 re->ipv6RouteWeight = 0;
11100 re->ipv6RouteMetric = 0;
11101 re->ipv6RouteInfo.re_max_frag = ire->ire_metrics.iulp_mtu;
11102 if (ire->ire_ill != NULL && re->ipv6RouteInfo.re_max_frag == 0)
11103 re->ipv6RouteInfo.re_max_frag = ire->ire_ill->ill_mtu;
11104
11105 re->ipv6RouteInfo.re_frag_flag = 0;
11106 re->ipv6RouteInfo.re_rtt = 0;
11107 re->ipv6RouteInfo.re_src_addr = ipv6_all_zeros;
11108 re->ipv6RouteInfo.re_obpkt = ire->ire_ob_pkt_count;
11109 re->ipv6RouteInfo.re_ibpkt = ire->ire_ib_pkt_count;
11110 re->ipv6RouteInfo.re_ref = ire->ire_refcnt;
11111 re->ipv6RouteInfo.re_flags = ire->ire_flags;
11112
11113 /* Add the IRE_IF_CLONE's counters to their parent IRE_INTERFACE */
11114 if (ire->ire_type & IRE_INTERFACE) {
11115 ire_t *child;
11116
11117 rw_enter(&ipst->ips_ire_dep_lock, RW_READER);
11118 child = ire->ire_dep_children;
11119 while (child != NULL) {
11120 re->ipv6RouteInfo.re_obpkt += child->ire_ob_pkt_count;
11121 re->ipv6RouteInfo.re_ibpkt += child->ire_ib_pkt_count;
11122 child = child->ire_dep_sib_next;
11123 }
11124 rw_exit(&ipst->ips_ire_dep_lock);
11125 }
11126 if (ire->ire_flags & RTF_DYNAMIC) {
11127 re->ipv6RouteInfo.re_ire_type = IRE_HOST_REDIRECT;
11128 } else {
11129 re->ipv6RouteInfo.re_ire_type = ire->ire_type;
11130 }
11131
11132 if (!snmp_append_data2(ird->ird_route.lp_head, &ird->ird_route.lp_tail,
11133 (char *)re, (int)sizeof (*re))) {
11134 ip1dbg(("ip_snmp_get2_v6: failed to allocate %u bytes\n",
11135 (uint_t)sizeof (*re)));
11136 }
11137
11138 if (gc != NULL) {
11139 iaes.iae_routeidx = ird->ird_idx;
11140 iaes.iae_doi = gc->gc_db->gcdb_doi;
11141 iaes.iae_slrange = gc->gc_db->gcdb_slrange;
11142
11143 if (!snmp_append_data2(ird->ird_attrs.lp_head,
11144 &ird->ird_attrs.lp_tail, (char *)&iaes, sizeof (iaes))) {
11145 ip1dbg(("ip_snmp_get2_v6: failed to allocate %u "
11146 "bytes\n", (uint_t)sizeof (iaes)));
11147 }
11148 }
11149
11150 /* bump route index for next pass */
11151 ird->ird_idx++;
11152
11153 kmem_free(re, sizeof (*re));
11154 if (gcgrp != NULL)
11155 rw_exit(&gcgrp->gcgrp_rwlock);
11156 }
11157
11158 /*
11159 * ncec_walk routine to create ipv6NetToMediaEntryTable
11160 */
11161 static int
11162 ip_snmp_get2_v6_media(ncec_t *ncec, iproutedata_t *ird)
11163 {
11164 ill_t *ill;
11165 mib2_ipv6NetToMediaEntry_t ntme;
11166
11167 ill = ncec->ncec_ill;
11168 /* skip arpce entries, and loopback ncec entries */
11169 if (ill->ill_isv6 == B_FALSE || ill->ill_net_type == IRE_LOOPBACK)
11170 return (0);
11171 /*
11172 * Neighbor cache entry attached to IRE with on-link
11173 * destination.
11174 * We report all IPMP groups on ncec_ill which is normally the upper.
11175 */
11176 ntme.ipv6NetToMediaIfIndex = ill->ill_phyint->phyint_ifindex;
11177 ntme.ipv6NetToMediaNetAddress = ncec->ncec_addr;
11178 ntme.ipv6NetToMediaPhysAddress.o_length = ill->ill_phys_addr_length;
11179 if (ncec->ncec_lladdr != NULL) {
11180 bcopy(ncec->ncec_lladdr, ntme.ipv6NetToMediaPhysAddress.o_bytes,
11181 ntme.ipv6NetToMediaPhysAddress.o_length);
11182 }
11183 /*
11184 * Note: Returns ND_* states. Should be:
11185 * reachable(1), stale(2), delay(3), probe(4),
11186 * invalid(5), unknown(6)
11187 */
11188 ntme.ipv6NetToMediaState = ncec->ncec_state;
11189 ntme.ipv6NetToMediaLastUpdated = 0;
11190
11191 /* other(1), dynamic(2), static(3), local(4) */
11192 if (NCE_MYADDR(ncec)) {
11193 ntme.ipv6NetToMediaType = 4;
11194 } else if (ncec->ncec_flags & NCE_F_PUBLISH) {
11195 ntme.ipv6NetToMediaType = 1; /* proxy */
11196 } else if (ncec->ncec_flags & NCE_F_STATIC) {
11197 ntme.ipv6NetToMediaType = 3;
11198 } else if (ncec->ncec_flags & (NCE_F_MCAST|NCE_F_BCAST)) {
11199 ntme.ipv6NetToMediaType = 1;
11200 } else {
11201 ntme.ipv6NetToMediaType = 2;
11202 }
11203
11204 if (!snmp_append_data2(ird->ird_netmedia.lp_head,
11205 &ird->ird_netmedia.lp_tail, (char *)&ntme, sizeof (ntme))) {
11206 ip1dbg(("ip_snmp_get2_v6_media: failed to allocate %u bytes\n",
11207 (uint_t)sizeof (ntme)));
11208 }
11209 return (0);
11210 }
11211
11212 int
11213 nce2ace(ncec_t *ncec)
11214 {
11215 int flags = 0;
11216
11217 if (NCE_ISREACHABLE(ncec))
11218 flags |= ACE_F_RESOLVED;
11219 if (ncec->ncec_flags & NCE_F_AUTHORITY)
11220 flags |= ACE_F_AUTHORITY;
11221 if (ncec->ncec_flags & NCE_F_PUBLISH)
11222 flags |= ACE_F_PUBLISH;
11223 if ((ncec->ncec_flags & NCE_F_NONUD) != 0)
11224 flags |= ACE_F_PERMANENT;
11225 if (NCE_MYADDR(ncec))
11226 flags |= (ACE_F_MYADDR | ACE_F_AUTHORITY);
11227 if (ncec->ncec_flags & NCE_F_UNVERIFIED)
11228 flags |= ACE_F_UNVERIFIED;
11229 if (ncec->ncec_flags & NCE_F_AUTHORITY)
11230 flags |= ACE_F_AUTHORITY;
11231 if (ncec->ncec_flags & NCE_F_DELAYED)
11232 flags |= ACE_F_DELAYED;
11233 return (flags);
11234 }
11235
11236 /*
11237 * ncec_walk routine to create ipNetToMediaEntryTable
11238 */
11239 static int
11240 ip_snmp_get2_v4_media(ncec_t *ncec, iproutedata_t *ird)
11241 {
11242 ill_t *ill;
11243 mib2_ipNetToMediaEntry_t ntme;
11244 const char *name = "unknown";
11245 ipaddr_t ncec_addr;
11246
11247 ill = ncec->ncec_ill;
11248 if (ill->ill_isv6 || (ncec->ncec_flags & NCE_F_BCAST) ||
11249 ill->ill_net_type == IRE_LOOPBACK)
11250 return (0);
11251
11252 /* We report all IPMP groups on ncec_ill which is normally the upper. */
11253 name = ill->ill_name;
11254 /* Based on RFC 4293: other(1), inval(2), dyn(3), stat(4) */
11255 if (NCE_MYADDR(ncec)) {
11256 ntme.ipNetToMediaType = 4;
11257 } else if (ncec->ncec_flags & (NCE_F_MCAST|NCE_F_BCAST|NCE_F_PUBLISH)) {
11258 ntme.ipNetToMediaType = 1;
11259 } else {
11260 ntme.ipNetToMediaType = 3;
11261 }
11262 ntme.ipNetToMediaIfIndex.o_length = MIN(OCTET_LENGTH, strlen(name));
11263 bcopy(name, ntme.ipNetToMediaIfIndex.o_bytes,
11264 ntme.ipNetToMediaIfIndex.o_length);
11265
11266 IN6_V4MAPPED_TO_IPADDR(&ncec->ncec_addr, ncec_addr);
11267 bcopy(&ncec_addr, &ntme.ipNetToMediaNetAddress, sizeof (ncec_addr));
11268
11269 ntme.ipNetToMediaInfo.ntm_mask.o_length = sizeof (ipaddr_t);
11270 ncec_addr = INADDR_BROADCAST;
11271 bcopy(&ncec_addr, ntme.ipNetToMediaInfo.ntm_mask.o_bytes,
11272 sizeof (ncec_addr));
11273 /*
11274 * map all the flags to the ACE counterpart.
11275 */
11276 ntme.ipNetToMediaInfo.ntm_flags = nce2ace(ncec);
11277
11278 ntme.ipNetToMediaPhysAddress.o_length =
11279 MIN(OCTET_LENGTH, ill->ill_phys_addr_length);
11280
11281 if (!NCE_ISREACHABLE(ncec))
11282 ntme.ipNetToMediaPhysAddress.o_length = 0;
11283 else {
11284 if (ncec->ncec_lladdr != NULL) {
11285 bcopy(ncec->ncec_lladdr,
11286 ntme.ipNetToMediaPhysAddress.o_bytes,
11287 ntme.ipNetToMediaPhysAddress.o_length);
11288 }
11289 }
11290
11291 if (!snmp_append_data2(ird->ird_netmedia.lp_head,
11292 &ird->ird_netmedia.lp_tail, (char *)&ntme, sizeof (ntme))) {
11293 ip1dbg(("ip_snmp_get2_v4_media: failed to allocate %u bytes\n",
11294 (uint_t)sizeof (ntme)));
11295 }
11296 return (0);
11297 }
11298
11299 /*
11300 * return (0) if invalid set request, 1 otherwise, including non-tcp requests
11301 */
11302 /* ARGSUSED */
11303 int
11304 ip_snmp_set(queue_t *q, int level, int name, uchar_t *ptr, int len)
11305 {
11306 switch (level) {
11307 case MIB2_IP:
11308 case MIB2_ICMP:
11309 switch (name) {
11310 default:
11311 break;
11312 }
11313 return (1);
11314 default:
11315 return (1);
11316 }
11317 }
11318
11319 /*
11320 * When there exists both a 64- and 32-bit counter of a particular type
11321 * (i.e., InReceives), only the 64-bit counters are added.
11322 */
11323 void
11324 ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *o1, mib2_ipIfStatsEntry_t *o2)
11325 {
11326 UPDATE_MIB(o1, ipIfStatsInHdrErrors, o2->ipIfStatsInHdrErrors);
11327 UPDATE_MIB(o1, ipIfStatsInTooBigErrors, o2->ipIfStatsInTooBigErrors);
11328 UPDATE_MIB(o1, ipIfStatsInNoRoutes, o2->ipIfStatsInNoRoutes);
11329 UPDATE_MIB(o1, ipIfStatsInAddrErrors, o2->ipIfStatsInAddrErrors);
11330 UPDATE_MIB(o1, ipIfStatsInUnknownProtos, o2->ipIfStatsInUnknownProtos);
11331 UPDATE_MIB(o1, ipIfStatsInTruncatedPkts, o2->ipIfStatsInTruncatedPkts);
11332 UPDATE_MIB(o1, ipIfStatsInDiscards, o2->ipIfStatsInDiscards);
11333 UPDATE_MIB(o1, ipIfStatsOutDiscards, o2->ipIfStatsOutDiscards);
11334 UPDATE_MIB(o1, ipIfStatsOutFragOKs, o2->ipIfStatsOutFragOKs);
11335 UPDATE_MIB(o1, ipIfStatsOutFragFails, o2->ipIfStatsOutFragFails);
11336 UPDATE_MIB(o1, ipIfStatsOutFragCreates, o2->ipIfStatsOutFragCreates);
11337 UPDATE_MIB(o1, ipIfStatsReasmReqds, o2->ipIfStatsReasmReqds);
11338 UPDATE_MIB(o1, ipIfStatsReasmOKs, o2->ipIfStatsReasmOKs);
11339 UPDATE_MIB(o1, ipIfStatsReasmFails, o2->ipIfStatsReasmFails);
11340 UPDATE_MIB(o1, ipIfStatsOutNoRoutes, o2->ipIfStatsOutNoRoutes);
11341 UPDATE_MIB(o1, ipIfStatsReasmDuplicates, o2->ipIfStatsReasmDuplicates);
11342 UPDATE_MIB(o1, ipIfStatsReasmPartDups, o2->ipIfStatsReasmPartDups);
11343 UPDATE_MIB(o1, ipIfStatsForwProhibits, o2->ipIfStatsForwProhibits);
11344 UPDATE_MIB(o1, udpInCksumErrs, o2->udpInCksumErrs);
11345 UPDATE_MIB(o1, udpInOverflows, o2->udpInOverflows);
11346 UPDATE_MIB(o1, rawipInOverflows, o2->rawipInOverflows);
11347 UPDATE_MIB(o1, ipIfStatsInWrongIPVersion,
11348 o2->ipIfStatsInWrongIPVersion);
11349 UPDATE_MIB(o1, ipIfStatsOutWrongIPVersion,
11350 o2->ipIfStatsInWrongIPVersion);
11351 UPDATE_MIB(o1, ipIfStatsOutSwitchIPVersion,
11352 o2->ipIfStatsOutSwitchIPVersion);
11353 UPDATE_MIB(o1, ipIfStatsHCInReceives, o2->ipIfStatsHCInReceives);
11354 UPDATE_MIB(o1, ipIfStatsHCInOctets, o2->ipIfStatsHCInOctets);
11355 UPDATE_MIB(o1, ipIfStatsHCInForwDatagrams,
11356 o2->ipIfStatsHCInForwDatagrams);
11357 UPDATE_MIB(o1, ipIfStatsHCInDelivers, o2->ipIfStatsHCInDelivers);
11358 UPDATE_MIB(o1, ipIfStatsHCOutRequests, o2->ipIfStatsHCOutRequests);
11359 UPDATE_MIB(o1, ipIfStatsHCOutForwDatagrams,
11360 o2->ipIfStatsHCOutForwDatagrams);
11361 UPDATE_MIB(o1, ipIfStatsOutFragReqds, o2->ipIfStatsOutFragReqds);
11362 UPDATE_MIB(o1, ipIfStatsHCOutTransmits, o2->ipIfStatsHCOutTransmits);
11363 UPDATE_MIB(o1, ipIfStatsHCOutOctets, o2->ipIfStatsHCOutOctets);
11364 UPDATE_MIB(o1, ipIfStatsHCInMcastPkts, o2->ipIfStatsHCInMcastPkts);
11365 UPDATE_MIB(o1, ipIfStatsHCInMcastOctets, o2->ipIfStatsHCInMcastOctets);
11366 UPDATE_MIB(o1, ipIfStatsHCOutMcastPkts, o2->ipIfStatsHCOutMcastPkts);
11367 UPDATE_MIB(o1, ipIfStatsHCOutMcastOctets,
11368 o2->ipIfStatsHCOutMcastOctets);
11369 UPDATE_MIB(o1, ipIfStatsHCInBcastPkts, o2->ipIfStatsHCInBcastPkts);
11370 UPDATE_MIB(o1, ipIfStatsHCOutBcastPkts, o2->ipIfStatsHCOutBcastPkts);
11371 UPDATE_MIB(o1, ipsecInSucceeded, o2->ipsecInSucceeded);
11372 UPDATE_MIB(o1, ipsecInFailed, o2->ipsecInFailed);
11373 UPDATE_MIB(o1, ipInCksumErrs, o2->ipInCksumErrs);
11374 UPDATE_MIB(o1, tcpInErrs, o2->tcpInErrs);
11375 UPDATE_MIB(o1, udpNoPorts, o2->udpNoPorts);
11376 }
11377
11378 void
11379 ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *o1, mib2_ipv6IfIcmpEntry_t *o2)
11380 {
11381 UPDATE_MIB(o1, ipv6IfIcmpInMsgs, o2->ipv6IfIcmpInMsgs);
11382 UPDATE_MIB(o1, ipv6IfIcmpInErrors, o2->ipv6IfIcmpInErrors);
11383 UPDATE_MIB(o1, ipv6IfIcmpInDestUnreachs, o2->ipv6IfIcmpInDestUnreachs);
11384 UPDATE_MIB(o1, ipv6IfIcmpInAdminProhibs, o2->ipv6IfIcmpInAdminProhibs);
11385 UPDATE_MIB(o1, ipv6IfIcmpInTimeExcds, o2->ipv6IfIcmpInTimeExcds);
11386 UPDATE_MIB(o1, ipv6IfIcmpInParmProblems, o2->ipv6IfIcmpInParmProblems);
11387 UPDATE_MIB(o1, ipv6IfIcmpInPktTooBigs, o2->ipv6IfIcmpInPktTooBigs);
11388 UPDATE_MIB(o1, ipv6IfIcmpInEchos, o2->ipv6IfIcmpInEchos);
11389 UPDATE_MIB(o1, ipv6IfIcmpInEchoReplies, o2->ipv6IfIcmpInEchoReplies);
11390 UPDATE_MIB(o1, ipv6IfIcmpInRouterSolicits,
11391 o2->ipv6IfIcmpInRouterSolicits);
11392 UPDATE_MIB(o1, ipv6IfIcmpInRouterAdvertisements,
11393 o2->ipv6IfIcmpInRouterAdvertisements);
11394 UPDATE_MIB(o1, ipv6IfIcmpInNeighborSolicits,
11395 o2->ipv6IfIcmpInNeighborSolicits);
11396 UPDATE_MIB(o1, ipv6IfIcmpInNeighborAdvertisements,
11397 o2->ipv6IfIcmpInNeighborAdvertisements);
11398 UPDATE_MIB(o1, ipv6IfIcmpInRedirects, o2->ipv6IfIcmpInRedirects);
11399 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembQueries,
11400 o2->ipv6IfIcmpInGroupMembQueries);
11401 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembResponses,
11402 o2->ipv6IfIcmpInGroupMembResponses);
11403 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembReductions,
11404 o2->ipv6IfIcmpInGroupMembReductions);
11405 UPDATE_MIB(o1, ipv6IfIcmpOutMsgs, o2->ipv6IfIcmpOutMsgs);
11406 UPDATE_MIB(o1, ipv6IfIcmpOutErrors, o2->ipv6IfIcmpOutErrors);
11407 UPDATE_MIB(o1, ipv6IfIcmpOutDestUnreachs,
11408 o2->ipv6IfIcmpOutDestUnreachs);
11409 UPDATE_MIB(o1, ipv6IfIcmpOutAdminProhibs,
11410 o2->ipv6IfIcmpOutAdminProhibs);
11411 UPDATE_MIB(o1, ipv6IfIcmpOutTimeExcds, o2->ipv6IfIcmpOutTimeExcds);
11412 UPDATE_MIB(o1, ipv6IfIcmpOutParmProblems,
11413 o2->ipv6IfIcmpOutParmProblems);
11414 UPDATE_MIB(o1, ipv6IfIcmpOutPktTooBigs, o2->ipv6IfIcmpOutPktTooBigs);
11415 UPDATE_MIB(o1, ipv6IfIcmpOutEchos, o2->ipv6IfIcmpOutEchos);
11416 UPDATE_MIB(o1, ipv6IfIcmpOutEchoReplies, o2->ipv6IfIcmpOutEchoReplies);
11417 UPDATE_MIB(o1, ipv6IfIcmpOutRouterSolicits,
11418 o2->ipv6IfIcmpOutRouterSolicits);
11419 UPDATE_MIB(o1, ipv6IfIcmpOutRouterAdvertisements,
11420 o2->ipv6IfIcmpOutRouterAdvertisements);
11421 UPDATE_MIB(o1, ipv6IfIcmpOutNeighborSolicits,
11422 o2->ipv6IfIcmpOutNeighborSolicits);
11423 UPDATE_MIB(o1, ipv6IfIcmpOutNeighborAdvertisements,
11424 o2->ipv6IfIcmpOutNeighborAdvertisements);
11425 UPDATE_MIB(o1, ipv6IfIcmpOutRedirects, o2->ipv6IfIcmpOutRedirects);
11426 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembQueries,
11427 o2->ipv6IfIcmpOutGroupMembQueries);
11428 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembResponses,
11429 o2->ipv6IfIcmpOutGroupMembResponses);
11430 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembReductions,
11431 o2->ipv6IfIcmpOutGroupMembReductions);
11432 UPDATE_MIB(o1, ipv6IfIcmpInOverflows, o2->ipv6IfIcmpInOverflows);
11433 UPDATE_MIB(o1, ipv6IfIcmpBadHoplimit, o2->ipv6IfIcmpBadHoplimit);
11434 UPDATE_MIB(o1, ipv6IfIcmpInBadNeighborAdvertisements,
11435 o2->ipv6IfIcmpInBadNeighborAdvertisements);
11436 UPDATE_MIB(o1, ipv6IfIcmpInBadNeighborSolicitations,
11437 o2->ipv6IfIcmpInBadNeighborSolicitations);
11438 UPDATE_MIB(o1, ipv6IfIcmpInBadRedirects, o2->ipv6IfIcmpInBadRedirects);
11439 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembTotal,
11440 o2->ipv6IfIcmpInGroupMembTotal);
11441 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembBadQueries,
11442 o2->ipv6IfIcmpInGroupMembBadQueries);
11443 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembBadReports,
11444 o2->ipv6IfIcmpInGroupMembBadReports);
11445 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembOurReports,
11446 o2->ipv6IfIcmpInGroupMembOurReports);
11447 }
11448
11449 /*
11450 * Called before the options are updated to check if this packet will
11451 * be source routed from here.
11452 * This routine assumes that the options are well formed i.e. that they
11453 * have already been checked.
11454 */
11455 boolean_t
11456 ip_source_routed(ipha_t *ipha, ip_stack_t *ipst)
11457 {
11458 ipoptp_t opts;
11459 uchar_t *opt;
11460 uint8_t optval;
11461 uint8_t optlen;
11462 ipaddr_t dst;
11463
11464 if (IS_SIMPLE_IPH(ipha)) {
11465 ip2dbg(("not source routed\n"));
11466 return (B_FALSE);
11467 }
11468 dst = ipha->ipha_dst;
11469 for (optval = ipoptp_first(&opts, ipha);
11470 optval != IPOPT_EOL;
11471 optval = ipoptp_next(&opts)) {
11472 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11473 opt = opts.ipoptp_cur;
11474 optlen = opts.ipoptp_len;
11475 ip2dbg(("ip_source_routed: opt %d, len %d\n",
11476 optval, optlen));
11477 switch (optval) {
11478 uint32_t off;
11479 case IPOPT_SSRR:
11480 case IPOPT_LSRR:
11481 /*
11482 * If dst is one of our addresses and there are some
11483 * entries left in the source route return (true).
11484 */
11485 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
11486 ip2dbg(("ip_source_routed: not next"
11487 " source route 0x%x\n",
11488 ntohl(dst)));
11489 return (B_FALSE);
11490 }
11491 off = opt[IPOPT_OFFSET];
11492 off--;
11493 if (optlen < IP_ADDR_LEN ||
11494 off > optlen - IP_ADDR_LEN) {
11495 /* End of source route */
11496 ip1dbg(("ip_source_routed: end of SR\n"));
11497 return (B_FALSE);
11498 }
11499 return (B_TRUE);
11500 }
11501 }
11502 ip2dbg(("not source routed\n"));
11503 return (B_FALSE);
11504 }
11505
11506 /*
11507 * ip_unbind is called by the transports to remove a conn from
11508 * the fanout table.
11509 */
11510 void
11511 ip_unbind(conn_t *connp)
11512 {
11513
11514 ASSERT(!MUTEX_HELD(&connp->conn_lock));
11515
11516 if (is_system_labeled() && connp->conn_anon_port) {
11517 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
11518 connp->conn_mlp_type, connp->conn_proto,
11519 ntohs(connp->conn_lport), B_FALSE);
11520 connp->conn_anon_port = 0;
11521 }
11522 connp->conn_mlp_type = mlptSingle;
11523
11524 ipcl_hash_remove(connp);
11525 }
11526
11527 /*
11528 * Used for deciding the MSS size for the upper layer. Thus
11529 * we need to check the outbound policy values in the conn.
11530 */
11531 int
11532 conn_ipsec_length(conn_t *connp)
11533 {
11534 ipsec_latch_t *ipl;
11535
11536 ipl = connp->conn_latch;
11537 if (ipl == NULL)
11538 return (0);
11539
11540 if (connp->conn_ixa->ixa_ipsec_policy == NULL)
11541 return (0);
11542
11543 return (connp->conn_ixa->ixa_ipsec_policy->ipsp_act->ipa_ovhd);
11544 }
11545
11546 /*
11547 * Returns an estimate of the IPsec headers size. This is used if
11548 * we don't want to call into IPsec to get the exact size.
11549 */
11550 int
11551 ipsec_out_extra_length(ip_xmit_attr_t *ixa)
11552 {
11553 ipsec_action_t *a;
11554
11555 if (!(ixa->ixa_flags & IXAF_IPSEC_SECURE))
11556 return (0);
11557
11558 a = ixa->ixa_ipsec_action;
11559 if (a == NULL) {
11560 ASSERT(ixa->ixa_ipsec_policy != NULL);
11561 a = ixa->ixa_ipsec_policy->ipsp_act;
11562 }
11563 ASSERT(a != NULL);
11564
11565 return (a->ipa_ovhd);
11566 }
11567
11568 /*
11569 * If there are any source route options, return the true final
11570 * destination. Otherwise, return the destination.
11571 */
11572 ipaddr_t
11573 ip_get_dst(ipha_t *ipha)
11574 {
11575 ipoptp_t opts;
11576 uchar_t *opt;
11577 uint8_t optval;
11578 uint8_t optlen;
11579 ipaddr_t dst;
11580 uint32_t off;
11581
11582 dst = ipha->ipha_dst;
11583
11584 if (IS_SIMPLE_IPH(ipha))
11585 return (dst);
11586
11587 for (optval = ipoptp_first(&opts, ipha);
11588 optval != IPOPT_EOL;
11589 optval = ipoptp_next(&opts)) {
11590 opt = opts.ipoptp_cur;
11591 optlen = opts.ipoptp_len;
11592 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11593 switch (optval) {
11594 case IPOPT_SSRR:
11595 case IPOPT_LSRR:
11596 off = opt[IPOPT_OFFSET];
11597 /*
11598 * If one of the conditions is true, it means
11599 * end of options and dst already has the right
11600 * value.
11601 */
11602 if (!(optlen < IP_ADDR_LEN || off > optlen - 3)) {
11603 off = optlen - IP_ADDR_LEN;
11604 bcopy(&opt[off], &dst, IP_ADDR_LEN);
11605 }
11606 return (dst);
11607 default:
11608 break;
11609 }
11610 }
11611
11612 return (dst);
11613 }
11614
11615 /*
11616 * Outbound IP fragmentation routine.
11617 * Assumes the caller has checked whether or not fragmentation should
11618 * be allowed. Here we copy the DF bit from the header to all the generated
11619 * fragments.
11620 */
11621 int
11622 ip_fragment_v4(mblk_t *mp_orig, nce_t *nce, iaflags_t ixaflags,
11623 uint_t pkt_len, uint32_t max_frag, uint32_t xmit_hint, zoneid_t szone,
11624 zoneid_t nolzid, pfirepostfrag_t postfragfn, uintptr_t *ixa_cookie)
11625 {
11626 int i1;
11627 int hdr_len;
11628 mblk_t *hdr_mp;
11629 ipha_t *ipha;
11630 int ip_data_end;
11631 int len;
11632 mblk_t *mp = mp_orig;
11633 int offset;
11634 ill_t *ill = nce->nce_ill;
11635 ip_stack_t *ipst = ill->ill_ipst;
11636 mblk_t *carve_mp;
11637 uint32_t frag_flag;
11638 uint_t priority = mp->b_band;
11639 int error = 0;
11640
11641 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragReqds);
11642
11643 if (pkt_len != msgdsize(mp)) {
11644 ip0dbg(("Packet length mismatch: %d, %ld\n",
11645 pkt_len, msgdsize(mp)));
11646 freemsg(mp);
11647 return (EINVAL);
11648 }
11649
11650 if (max_frag == 0) {
11651 ip1dbg(("ip_fragment_v4: max_frag is zero. Dropping packet\n"));
11652 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11653 ip_drop_output("FragFails: zero max_frag", mp, ill);
11654 freemsg(mp);
11655 return (EINVAL);
11656 }
11657
11658 ASSERT(MBLKL(mp) >= sizeof (ipha_t));
11659 ipha = (ipha_t *)mp->b_rptr;
11660 ASSERT(ntohs(ipha->ipha_length) == pkt_len);
11661 frag_flag = ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_DF;
11662
11663 /*
11664 * Establish the starting offset. May not be zero if we are fragging
11665 * a fragment that is being forwarded.
11666 */
11667 offset = ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_OFFSET;
11668
11669 /* TODO why is this test needed? */
11670 if (((max_frag - ntohs(ipha->ipha_length)) & ~7) < 8) {
11671 /* TODO: notify ulp somehow */
11672 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11673 ip_drop_output("FragFails: bad starting offset", mp, ill);
11674 freemsg(mp);
11675 return (EINVAL);
11676 }
11677
11678 hdr_len = IPH_HDR_LENGTH(ipha);
11679 ipha->ipha_hdr_checksum = 0;
11680
11681 /*
11682 * Establish the number of bytes maximum per frag, after putting
11683 * in the header.
11684 */
11685 len = (max_frag - hdr_len) & ~7;
11686
11687 /* Get a copy of the header for the trailing frags */
11688 hdr_mp = ip_fragment_copyhdr((uchar_t *)ipha, hdr_len, offset, ipst,
11689 mp);
11690 if (hdr_mp == NULL) {
11691 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11692 ip_drop_output("FragFails: no hdr_mp", mp, ill);
11693 freemsg(mp);
11694 return (ENOBUFS);
11695 }
11696
11697 /* Store the starting offset, with the MoreFrags flag. */
11698 i1 = offset | IPH_MF | frag_flag;
11699 ipha->ipha_fragment_offset_and_flags = htons((uint16_t)i1);
11700
11701 /* Establish the ending byte offset, based on the starting offset. */
11702 offset <<= 3;
11703 ip_data_end = offset + ntohs(ipha->ipha_length) - hdr_len;
11704
11705 /* Store the length of the first fragment in the IP header. */
11706 i1 = len + hdr_len;
11707 ASSERT(i1 <= IP_MAXPACKET);
11708 ipha->ipha_length = htons((uint16_t)i1);
11709
11710 /*
11711 * Compute the IP header checksum for the first frag. We have to
11712 * watch out that we stop at the end of the header.
11713 */
11714 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
11715
11716 /*
11717 * Now carve off the first frag. Note that this will include the
11718 * original IP header.
11719 */
11720 if (!(mp = ip_carve_mp(&mp_orig, i1))) {
11721 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11722 ip_drop_output("FragFails: could not carve mp", mp_orig, ill);
11723 freeb(hdr_mp);
11724 freemsg(mp_orig);
11725 return (ENOBUFS);
11726 }
11727
11728 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragCreates);
11729
11730 error = postfragfn(mp, nce, ixaflags, i1, xmit_hint, szone, nolzid,
11731 ixa_cookie);
11732 if (error != 0 && error != EWOULDBLOCK) {
11733 /* No point in sending the other fragments */
11734 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11735 ip_drop_output("FragFails: postfragfn failed", mp_orig, ill);
11736 freeb(hdr_mp);
11737 freemsg(mp_orig);
11738 return (error);
11739 }
11740
11741 /* No need to redo state machine in loop */
11742 ixaflags &= ~IXAF_REACH_CONF;
11743
11744 /* Advance the offset to the second frag starting point. */
11745 offset += len;
11746 /*
11747 * Update hdr_len from the copied header - there might be less options
11748 * in the later fragments.
11749 */
11750 hdr_len = IPH_HDR_LENGTH(hdr_mp->b_rptr);
11751 /* Loop until done. */
11752 for (;;) {
11753 uint16_t offset_and_flags;
11754 uint16_t ip_len;
11755
11756 if (ip_data_end - offset > len) {
11757 /*
11758 * Carve off the appropriate amount from the original
11759 * datagram.
11760 */
11761 if (!(carve_mp = ip_carve_mp(&mp_orig, len))) {
11762 mp = NULL;
11763 break;
11764 }
11765 /*
11766 * More frags after this one. Get another copy
11767 * of the header.
11768 */
11769 if (carve_mp->b_datap->db_ref == 1 &&
11770 hdr_mp->b_wptr - hdr_mp->b_rptr <
11771 carve_mp->b_rptr - carve_mp->b_datap->db_base) {
11772 /* Inline IP header */
11773 carve_mp->b_rptr -= hdr_mp->b_wptr -
11774 hdr_mp->b_rptr;
11775 bcopy(hdr_mp->b_rptr, carve_mp->b_rptr,
11776 hdr_mp->b_wptr - hdr_mp->b_rptr);
11777 mp = carve_mp;
11778 } else {
11779 if (!(mp = copyb(hdr_mp))) {
11780 freemsg(carve_mp);
11781 break;
11782 }
11783 /* Get priority marking, if any. */
11784 mp->b_band = priority;
11785 mp->b_cont = carve_mp;
11786 }
11787 ipha = (ipha_t *)mp->b_rptr;
11788 offset_and_flags = IPH_MF;
11789 } else {
11790 /*
11791 * Last frag. Consume the header. Set len to
11792 * the length of this last piece.
11793 */
11794 len = ip_data_end - offset;
11795
11796 /*
11797 * Carve off the appropriate amount from the original
11798 * datagram.
11799 */
11800 if (!(carve_mp = ip_carve_mp(&mp_orig, len))) {
11801 mp = NULL;
11802 break;
11803 }
11804 if (carve_mp->b_datap->db_ref == 1 &&
11805 hdr_mp->b_wptr - hdr_mp->b_rptr <
11806 carve_mp->b_rptr - carve_mp->b_datap->db_base) {
11807 /* Inline IP header */
11808 carve_mp->b_rptr -= hdr_mp->b_wptr -
11809 hdr_mp->b_rptr;
11810 bcopy(hdr_mp->b_rptr, carve_mp->b_rptr,
11811 hdr_mp->b_wptr - hdr_mp->b_rptr);
11812 mp = carve_mp;
11813 freeb(hdr_mp);
11814 hdr_mp = mp;
11815 } else {
11816 mp = hdr_mp;
11817 /* Get priority marking, if any. */
11818 mp->b_band = priority;
11819 mp->b_cont = carve_mp;
11820 }
11821 ipha = (ipha_t *)mp->b_rptr;
11822 /* A frag of a frag might have IPH_MF non-zero */
11823 offset_and_flags =
11824 ntohs(ipha->ipha_fragment_offset_and_flags) &
11825 IPH_MF;
11826 }
11827 offset_and_flags |= (uint16_t)(offset >> 3);
11828 offset_and_flags |= (uint16_t)frag_flag;
11829 /* Store the offset and flags in the IP header. */
11830 ipha->ipha_fragment_offset_and_flags = htons(offset_and_flags);
11831
11832 /* Store the length in the IP header. */
11833 ip_len = (uint16_t)(len + hdr_len);
11834 ipha->ipha_length = htons(ip_len);
11835
11836 /*
11837 * Set the IP header checksum. Note that mp is just
11838 * the header, so this is easy to pass to ip_csum.
11839 */
11840 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
11841
11842 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragCreates);
11843
11844 error = postfragfn(mp, nce, ixaflags, ip_len, xmit_hint, szone,
11845 nolzid, ixa_cookie);
11846 /* All done if we just consumed the hdr_mp. */
11847 if (mp == hdr_mp) {
11848 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragOKs);
11849 return (error);
11850 }
11851 if (error != 0 && error != EWOULDBLOCK) {
11852 DTRACE_PROBE2(ip__xmit__frag__fail, ill_t *, ill,
11853 mblk_t *, hdr_mp);
11854 /* No point in sending the other fragments */
11855 break;
11856 }
11857
11858 /* Otherwise, advance and loop. */
11859 offset += len;
11860 }
11861 /* Clean up following allocation failure. */
11862 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11863 ip_drop_output("FragFails: loop ended", NULL, ill);
11864 if (mp != hdr_mp)
11865 freeb(hdr_mp);
11866 if (mp != mp_orig)
11867 freemsg(mp_orig);
11868 return (error);
11869 }
11870
11871 /*
11872 * Copy the header plus those options which have the copy bit set
11873 */
11874 static mblk_t *
11875 ip_fragment_copyhdr(uchar_t *rptr, int hdr_len, int offset, ip_stack_t *ipst,
11876 mblk_t *src)
11877 {
11878 mblk_t *mp;
11879 uchar_t *up;
11880
11881 /*
11882 * Quick check if we need to look for options without the copy bit
11883 * set
11884 */
11885 mp = allocb_tmpl(ipst->ips_ip_wroff_extra + hdr_len, src);
11886 if (!mp)
11887 return (mp);
11888 mp->b_rptr += ipst->ips_ip_wroff_extra;
11889 if (hdr_len == IP_SIMPLE_HDR_LENGTH || offset != 0) {
11890 bcopy(rptr, mp->b_rptr, hdr_len);
11891 mp->b_wptr += hdr_len + ipst->ips_ip_wroff_extra;
11892 return (mp);
11893 }
11894 up = mp->b_rptr;
11895 bcopy(rptr, up, IP_SIMPLE_HDR_LENGTH);
11896 up += IP_SIMPLE_HDR_LENGTH;
11897 rptr += IP_SIMPLE_HDR_LENGTH;
11898 hdr_len -= IP_SIMPLE_HDR_LENGTH;
11899 while (hdr_len > 0) {
11900 uint32_t optval;
11901 uint32_t optlen;
11902
11903 optval = *rptr;
11904 if (optval == IPOPT_EOL)
11905 break;
11906 if (optval == IPOPT_NOP)
11907 optlen = 1;
11908 else
11909 optlen = rptr[1];
11910 if (optval & IPOPT_COPY) {
11911 bcopy(rptr, up, optlen);
11912 up += optlen;
11913 }
11914 rptr += optlen;
11915 hdr_len -= optlen;
11916 }
11917 /*
11918 * Make sure that we drop an even number of words by filling
11919 * with EOL to the next word boundary.
11920 */
11921 for (hdr_len = up - (mp->b_rptr + IP_SIMPLE_HDR_LENGTH);
11922 hdr_len & 0x3; hdr_len++)
11923 *up++ = IPOPT_EOL;
11924 mp->b_wptr = up;
11925 /* Update header length */
11926 mp->b_rptr[0] = (uint8_t)((IP_VERSION << 4) | ((up - mp->b_rptr) >> 2));
11927 return (mp);
11928 }
11929
11930 /*
11931 * Update any source route, record route, or timestamp options when
11932 * sending a packet back to ourselves.
11933 * Check that we are at end of strict source route.
11934 * The options have been sanity checked by ip_output_options().
11935 */
11936 void
11937 ip_output_local_options(ipha_t *ipha, ip_stack_t *ipst)
11938 {
11939 ipoptp_t opts;
11940 uchar_t *opt;
11941 uint8_t optval;
11942 uint8_t optlen;
11943 ipaddr_t dst;
11944 uint32_t ts;
11945 timestruc_t now;
11946
11947 for (optval = ipoptp_first(&opts, ipha);
11948 optval != IPOPT_EOL;
11949 optval = ipoptp_next(&opts)) {
11950 opt = opts.ipoptp_cur;
11951 optlen = opts.ipoptp_len;
11952 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11953 switch (optval) {
11954 uint32_t off;
11955 case IPOPT_SSRR:
11956 case IPOPT_LSRR:
11957 off = opt[IPOPT_OFFSET];
11958 off--;
11959 if (optlen < IP_ADDR_LEN ||
11960 off > optlen - IP_ADDR_LEN) {
11961 /* End of source route */
11962 break;
11963 }
11964 /*
11965 * This will only happen if two consecutive entries
11966 * in the source route contains our address or if
11967 * it is a packet with a loose source route which
11968 * reaches us before consuming the whole source route
11969 */
11970
11971 if (optval == IPOPT_SSRR) {
11972 return;
11973 }
11974 /*
11975 * Hack: instead of dropping the packet truncate the
11976 * source route to what has been used by filling the
11977 * rest with IPOPT_NOP.
11978 */
11979 opt[IPOPT_OLEN] = (uint8_t)off;
11980 while (off < optlen) {
11981 opt[off++] = IPOPT_NOP;
11982 }
11983 break;
11984 case IPOPT_RR:
11985 off = opt[IPOPT_OFFSET];
11986 off--;
11987 if (optlen < IP_ADDR_LEN ||
11988 off > optlen - IP_ADDR_LEN) {
11989 /* No more room - ignore */
11990 ip1dbg((
11991 "ip_output_local_options: end of RR\n"));
11992 break;
11993 }
11994 dst = htonl(INADDR_LOOPBACK);
11995 bcopy(&dst, (char *)opt + off, IP_ADDR_LEN);
11996 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
11997 break;
11998 case IPOPT_TS:
11999 /* Insert timestamp if there is romm */
12000 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
12001 case IPOPT_TS_TSONLY:
12002 off = IPOPT_TS_TIMELEN;
12003 break;
12004 case IPOPT_TS_PRESPEC:
12005 case IPOPT_TS_PRESPEC_RFC791:
12006 /* Verify that the address matched */
12007 off = opt[IPOPT_OFFSET] - 1;
12008 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
12009 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
12010 /* Not for us */
12011 break;
12012 }
12013 /* FALLTHRU */
12014 case IPOPT_TS_TSANDADDR:
12015 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
12016 break;
12017 default:
12018 /*
12019 * ip_*put_options should have already
12020 * dropped this packet.
12021 */
12022 cmn_err(CE_PANIC, "ip_output_local_options: "
12023 "unknown IT - bug in ip_output_options?\n");
12024 return; /* Keep "lint" happy */
12025 }
12026 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
12027 /* Increase overflow counter */
12028 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
12029 opt[IPOPT_POS_OV_FLG] = (uint8_t)
12030 (opt[IPOPT_POS_OV_FLG] & 0x0F) |
12031 (off << 4);
12032 break;
12033 }
12034 off = opt[IPOPT_OFFSET] - 1;
12035 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
12036 case IPOPT_TS_PRESPEC:
12037 case IPOPT_TS_PRESPEC_RFC791:
12038 case IPOPT_TS_TSANDADDR:
12039 dst = htonl(INADDR_LOOPBACK);
12040 bcopy(&dst, (char *)opt + off, IP_ADDR_LEN);
12041 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
12042 /* FALLTHRU */
12043 case IPOPT_TS_TSONLY:
12044 off = opt[IPOPT_OFFSET] - 1;
12045 /* Compute # of milliseconds since midnight */
12046 gethrestime(&now);
12047 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
12048 NSEC2MSEC(now.tv_nsec);
12049 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
12050 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
12051 break;
12052 }
12053 break;
12054 }
12055 }
12056 }
12057
12058 /*
12059 * Prepend an M_DATA fastpath header, and if none present prepend a
12060 * DL_UNITDATA_REQ. Frees the mblk on failure.
12061 *
12062 * nce_dlur_mp and nce_fp_mp can not disappear once they have been set.
12063 * If there is a change to them, the nce will be deleted (condemned) and
12064 * a new nce_t will be created when packets are sent. Thus we need no locks
12065 * to access those fields.
12066 *
12067 * We preserve b_band to support IPQoS. If a DL_UNITDATA_REQ is prepended
12068 * we place b_band in dl_priority.dl_max.
12069 */
12070 static mblk_t *
12071 ip_xmit_attach_llhdr(mblk_t *mp, nce_t *nce)
12072 {
12073 uint_t hlen;
12074 mblk_t *mp1;
12075 uint_t priority;
12076 uchar_t *rptr;
12077
12078 rptr = mp->b_rptr;
12079
12080 ASSERT(DB_TYPE(mp) == M_DATA);
12081 priority = mp->b_band;
12082
12083 ASSERT(nce != NULL);
12084 if ((mp1 = nce->nce_fp_mp) != NULL) {
12085 hlen = MBLKL(mp1);
12086 /*
12087 * Check if we have enough room to prepend fastpath
12088 * header
12089 */
12090 if (hlen != 0 && (rptr - mp->b_datap->db_base) >= hlen) {
12091 rptr -= hlen;
12092 bcopy(mp1->b_rptr, rptr, hlen);
12093 /*
12094 * Set the b_rptr to the start of the link layer
12095 * header
12096 */
12097 mp->b_rptr = rptr;
12098 return (mp);
12099 }
12100 mp1 = copyb(mp1);
12101 if (mp1 == NULL) {
12102 ill_t *ill = nce->nce_ill;
12103
12104 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12105 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12106 freemsg(mp);
12107 return (NULL);
12108 }
12109 mp1->b_band = priority;
12110 mp1->b_cont = mp;
12111 DB_CKSUMSTART(mp1) = DB_CKSUMSTART(mp);
12112 DB_CKSUMSTUFF(mp1) = DB_CKSUMSTUFF(mp);
12113 DB_CKSUMEND(mp1) = DB_CKSUMEND(mp);
12114 DB_CKSUMFLAGS(mp1) = DB_CKSUMFLAGS(mp);
12115 DB_LSOMSS(mp1) = DB_LSOMSS(mp);
12116 DTRACE_PROBE1(ip__xmit__copyb, (mblk_t *), mp1);
12117 /*
12118 * XXX disable ICK_VALID and compute checksum
12119 * here; can happen if nce_fp_mp changes and
12120 * it can't be copied now due to insufficient
12121 * space. (unlikely, fp mp can change, but it
12122 * does not increase in length)
12123 */
12124 return (mp1);
12125 }
12126 mp1 = copyb(nce->nce_dlur_mp);
12127
12128 if (mp1 == NULL) {
12129 ill_t *ill = nce->nce_ill;
12130
12131 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12132 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12133 freemsg(mp);
12134 return (NULL);
12135 }
12136 mp1->b_cont = mp;
12137 if (priority != 0) {
12138 mp1->b_band = priority;
12139 ((dl_unitdata_req_t *)(mp1->b_rptr))->dl_priority.dl_max =
12140 priority;
12141 }
12142 return (mp1);
12143 }
12144
12145 /*
12146 * Finish the outbound IPsec processing. This function is called from
12147 * ipsec_out_process() if the IPsec packet was processed
12148 * synchronously, or from {ah,esp}_kcf_callback_outbound() if it was processed
12149 * asynchronously.
12150 *
12151 * This is common to IPv4 and IPv6.
12152 */
12153 int
12154 ip_output_post_ipsec(mblk_t *mp, ip_xmit_attr_t *ixa)
12155 {
12156 iaflags_t ixaflags = ixa->ixa_flags;
12157 uint_t pktlen;
12158
12159
12160 /* AH/ESP don't update ixa_pktlen when they modify the packet */
12161 if (ixaflags & IXAF_IS_IPV4) {
12162 ipha_t *ipha = (ipha_t *)mp->b_rptr;
12163
12164 ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
12165 pktlen = ntohs(ipha->ipha_length);
12166 } else {
12167 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
12168
12169 ASSERT(IPH_HDR_VERSION(mp->b_rptr) == IPV6_VERSION);
12170 pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
12171 }
12172
12173 /*
12174 * We release any hard reference on the SAs here to make
12175 * sure the SAs can be garbage collected. ipsr_sa has a soft reference
12176 * on the SAs.
12177 * If in the future we want the hard latching of the SAs in the
12178 * ip_xmit_attr_t then we should remove this.
12179 */
12180 if (ixa->ixa_ipsec_esp_sa != NULL) {
12181 IPSA_REFRELE(ixa->ixa_ipsec_esp_sa);
12182 ixa->ixa_ipsec_esp_sa = NULL;
12183 }
12184 if (ixa->ixa_ipsec_ah_sa != NULL) {
12185 IPSA_REFRELE(ixa->ixa_ipsec_ah_sa);
12186 ixa->ixa_ipsec_ah_sa = NULL;
12187 }
12188
12189 /* Do we need to fragment? */
12190 if ((ixa->ixa_flags & IXAF_IPV6_ADD_FRAGHDR) ||
12191 pktlen > ixa->ixa_fragsize) {
12192 if (ixaflags & IXAF_IS_IPV4) {
12193 ASSERT(!(ixa->ixa_flags & IXAF_IPV6_ADD_FRAGHDR));
12194 /*
12195 * We check for the DF case in ipsec_out_process
12196 * hence this only handles the non-DF case.
12197 */
12198 return (ip_fragment_v4(mp, ixa->ixa_nce, ixa->ixa_flags,
12199 pktlen, ixa->ixa_fragsize,
12200 ixa->ixa_xmit_hint, ixa->ixa_zoneid,
12201 ixa->ixa_no_loop_zoneid, ixa->ixa_postfragfn,
12202 &ixa->ixa_cookie));
12203 } else {
12204 mp = ip_fraghdr_add_v6(mp, ixa->ixa_ident, ixa);
12205 if (mp == NULL) {
12206 /* MIB and ip_drop_output already done */
12207 return (ENOMEM);
12208 }
12209 pktlen += sizeof (ip6_frag_t);
12210 if (pktlen > ixa->ixa_fragsize) {
12211 return (ip_fragment_v6(mp, ixa->ixa_nce,
12212 ixa->ixa_flags, pktlen,
12213 ixa->ixa_fragsize, ixa->ixa_xmit_hint,
12214 ixa->ixa_zoneid, ixa->ixa_no_loop_zoneid,
12215 ixa->ixa_postfragfn, &ixa->ixa_cookie));
12216 }
12217 }
12218 }
12219 return ((ixa->ixa_postfragfn)(mp, ixa->ixa_nce, ixa->ixa_flags,
12220 pktlen, ixa->ixa_xmit_hint, ixa->ixa_zoneid,
12221 ixa->ixa_no_loop_zoneid, NULL));
12222 }
12223
12224 /*
12225 * Finish the inbound IPsec processing. This function is called from
12226 * ipsec_out_process() if the IPsec packet was processed
12227 * synchronously, or from {ah,esp}_kcf_callback_outbound() if it was processed
12228 * asynchronously.
12229 *
12230 * This is common to IPv4 and IPv6.
12231 */
12232 void
12233 ip_input_post_ipsec(mblk_t *mp, ip_recv_attr_t *ira)
12234 {
12235 iaflags_t iraflags = ira->ira_flags;
12236
12237 /* Length might have changed */
12238 if (iraflags & IRAF_IS_IPV4) {
12239 ipha_t *ipha = (ipha_t *)mp->b_rptr;
12240
12241 ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
12242 ira->ira_pktlen = ntohs(ipha->ipha_length);
12243 ira->ira_ip_hdr_length = IPH_HDR_LENGTH(ipha);
12244 ira->ira_protocol = ipha->ipha_protocol;
12245
12246 ip_fanout_v4(mp, ipha, ira);
12247 } else {
12248 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
12249 uint8_t *nexthdrp;
12250
12251 ASSERT(IPH_HDR_VERSION(mp->b_rptr) == IPV6_VERSION);
12252 ira->ira_pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
12253 if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &ira->ira_ip_hdr_length,
12254 &nexthdrp)) {
12255 /* Malformed packet */
12256 BUMP_MIB(ira->ira_ill->ill_ip_mib, ipIfStatsInDiscards);
12257 ip_drop_input("ipIfStatsInDiscards", mp, ira->ira_ill);
12258 freemsg(mp);
12259 return;
12260 }
12261 ira->ira_protocol = *nexthdrp;
12262 ip_fanout_v6(mp, ip6h, ira);
12263 }
12264 }
12265
12266 /*
12267 * Select which AH & ESP SA's to use (if any) for the outbound packet.
12268 *
12269 * If this function returns B_TRUE, the requested SA's have been filled
12270 * into the ixa_ipsec_*_sa pointers.
12271 *
12272 * If the function returns B_FALSE, the packet has been "consumed", most
12273 * likely by an ACQUIRE sent up via PF_KEY to a key management daemon.
12274 *
12275 * The SA references created by the protocol-specific "select"
12276 * function will be released in ip_output_post_ipsec.
12277 */
12278 static boolean_t
12279 ipsec_out_select_sa(mblk_t *mp, ip_xmit_attr_t *ixa)
12280 {
12281 boolean_t need_ah_acquire = B_FALSE, need_esp_acquire = B_FALSE;
12282 ipsec_policy_t *pp;
12283 ipsec_action_t *ap;
12284
12285 ASSERT(ixa->ixa_flags & IXAF_IPSEC_SECURE);
12286 ASSERT((ixa->ixa_ipsec_policy != NULL) ||
12287 (ixa->ixa_ipsec_action != NULL));
12288
12289 ap = ixa->ixa_ipsec_action;
12290 if (ap == NULL) {
12291 pp = ixa->ixa_ipsec_policy;
12292 ASSERT(pp != NULL);
12293 ap = pp->ipsp_act;
12294 ASSERT(ap != NULL);
12295 }
12296
12297 /*
12298 * We have an action. now, let's select SA's.
12299 * A side effect of setting ixa_ipsec_*_sa is that it will
12300 * be cached in the conn_t.
12301 */
12302 if (ap->ipa_want_esp) {
12303 if (ixa->ixa_ipsec_esp_sa == NULL) {
12304 need_esp_acquire = !ipsec_outbound_sa(mp, ixa,
12305 IPPROTO_ESP);
12306 }
12307 ASSERT(need_esp_acquire || ixa->ixa_ipsec_esp_sa != NULL);
12308 }
12309
12310 if (ap->ipa_want_ah) {
12311 if (ixa->ixa_ipsec_ah_sa == NULL) {
12312 need_ah_acquire = !ipsec_outbound_sa(mp, ixa,
12313 IPPROTO_AH);
12314 }
12315 ASSERT(need_ah_acquire || ixa->ixa_ipsec_ah_sa != NULL);
12316 /*
12317 * The ESP and AH processing order needs to be preserved
12318 * when both protocols are required (ESP should be applied
12319 * before AH for an outbound packet). Force an ESP ACQUIRE
12320 * when both ESP and AH are required, and an AH ACQUIRE
12321 * is needed.
12322 */
12323 if (ap->ipa_want_esp && need_ah_acquire)
12324 need_esp_acquire = B_TRUE;
12325 }
12326
12327 /*
12328 * Send an ACQUIRE (extended, regular, or both) if we need one.
12329 * Release SAs that got referenced, but will not be used until we
12330 * acquire _all_ of the SAs we need.
12331 */
12332 if (need_ah_acquire || need_esp_acquire) {
12333 if (ixa->ixa_ipsec_ah_sa != NULL) {
12334 IPSA_REFRELE(ixa->ixa_ipsec_ah_sa);
12335 ixa->ixa_ipsec_ah_sa = NULL;
12336 }
12337 if (ixa->ixa_ipsec_esp_sa != NULL) {
12338 IPSA_REFRELE(ixa->ixa_ipsec_esp_sa);
12339 ixa->ixa_ipsec_esp_sa = NULL;
12340 }
12341
12342 sadb_acquire(mp, ixa, need_ah_acquire, need_esp_acquire);
12343 return (B_FALSE);
12344 }
12345
12346 return (B_TRUE);
12347 }
12348
12349 /*
12350 * Handle IPsec output processing.
12351 * This function is only entered once for a given packet.
12352 * We try to do things synchronously, but if we need to have user-level
12353 * set up SAs, or ESP or AH uses asynchronous kEF, then the operation
12354 * will be completed
12355 * - when the SAs are added in esp_add_sa_finish/ah_add_sa_finish
12356 * - when asynchronous ESP is done it will do AH
12357 *
12358 * In all cases we come back in ip_output_post_ipsec() to fragment and
12359 * send out the packet.
12360 */
12361 int
12362 ipsec_out_process(mblk_t *mp, ip_xmit_attr_t *ixa)
12363 {
12364 ill_t *ill = ixa->ixa_nce->nce_ill;
12365 ip_stack_t *ipst = ixa->ixa_ipst;
12366 ipsec_stack_t *ipss;
12367 ipsec_policy_t *pp;
12368 ipsec_action_t *ap;
12369
12370 ASSERT(ixa->ixa_flags & IXAF_IPSEC_SECURE);
12371
12372 ASSERT((ixa->ixa_ipsec_policy != NULL) ||
12373 (ixa->ixa_ipsec_action != NULL));
12374
12375 ipss = ipst->ips_netstack->netstack_ipsec;
12376 if (!ipsec_loaded(ipss)) {
12377 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12378 ip_drop_packet(mp, B_TRUE, ill,
12379 DROPPER(ipss, ipds_ip_ipsec_not_loaded),
12380 &ipss->ipsec_dropper);
12381 return (ENOTSUP);
12382 }
12383
12384 ap = ixa->ixa_ipsec_action;
12385 if (ap == NULL) {
12386 pp = ixa->ixa_ipsec_policy;
12387 ASSERT(pp != NULL);
12388 ap = pp->ipsp_act;
12389 ASSERT(ap != NULL);
12390 }
12391
12392 /* Handle explicit drop action and bypass. */
12393 switch (ap->ipa_act.ipa_type) {
12394 case IPSEC_ACT_DISCARD:
12395 case IPSEC_ACT_REJECT:
12396 ip_drop_packet(mp, B_FALSE, ill,
12397 DROPPER(ipss, ipds_spd_explicit), &ipss->ipsec_spd_dropper);
12398 return (EHOSTUNREACH); /* IPsec policy failure */
12399 case IPSEC_ACT_BYPASS:
12400 return (ip_output_post_ipsec(mp, ixa));
12401 }
12402
12403 /*
12404 * The order of processing is first insert a IP header if needed.
12405 * Then insert the ESP header and then the AH header.
12406 */
12407 if ((ixa->ixa_flags & IXAF_IS_IPV4) && ap->ipa_want_se) {
12408 /*
12409 * First get the outer IP header before sending
12410 * it to ESP.
12411 */
12412 ipha_t *oipha, *iipha;
12413 mblk_t *outer_mp, *inner_mp;
12414
12415 if ((outer_mp = allocb(sizeof (ipha_t), BPRI_HI)) == NULL) {
12416 (void) mi_strlog(ill->ill_rq, 0,
12417 SL_ERROR|SL_TRACE|SL_CONSOLE,
12418 "ipsec_out_process: "
12419 "Self-Encapsulation failed: Out of memory\n");
12420 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12421 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12422 freemsg(mp);
12423 return (ENOBUFS);
12424 }
12425 inner_mp = mp;
12426 ASSERT(inner_mp->b_datap->db_type == M_DATA);
12427 oipha = (ipha_t *)outer_mp->b_rptr;
12428 iipha = (ipha_t *)inner_mp->b_rptr;
12429 *oipha = *iipha;
12430 outer_mp->b_wptr += sizeof (ipha_t);
12431 oipha->ipha_length = htons(ntohs(iipha->ipha_length) +
12432 sizeof (ipha_t));
12433 oipha->ipha_protocol = IPPROTO_ENCAP;
12434 oipha->ipha_version_and_hdr_length =
12435 IP_SIMPLE_HDR_VERSION;
12436 oipha->ipha_hdr_checksum = 0;
12437 oipha->ipha_hdr_checksum = ip_csum_hdr(oipha);
12438 outer_mp->b_cont = inner_mp;
12439 mp = outer_mp;
12440
12441 ixa->ixa_flags |= IXAF_IPSEC_TUNNEL;
12442 }
12443
12444 /* If we need to wait for a SA then we can't return any errno */
12445 if (((ap->ipa_want_ah && (ixa->ixa_ipsec_ah_sa == NULL)) ||
12446 (ap->ipa_want_esp && (ixa->ixa_ipsec_esp_sa == NULL))) &&
12447 !ipsec_out_select_sa(mp, ixa))
12448 return (0);
12449
12450 /*
12451 * By now, we know what SA's to use. Toss over to ESP & AH
12452 * to do the heavy lifting.
12453 */
12454 if (ap->ipa_want_esp) {
12455 ASSERT(ixa->ixa_ipsec_esp_sa != NULL);
12456
12457 mp = ixa->ixa_ipsec_esp_sa->ipsa_output_func(mp, ixa);
12458 if (mp == NULL) {
12459 /*
12460 * Either it failed or is pending. In the former case
12461 * ipIfStatsInDiscards was increased.
12462 */
12463 return (0);
12464 }
12465 }
12466
12467 if (ap->ipa_want_ah) {
12468 ASSERT(ixa->ixa_ipsec_ah_sa != NULL);
12469
12470 mp = ixa->ixa_ipsec_ah_sa->ipsa_output_func(mp, ixa);
12471 if (mp == NULL) {
12472 /*
12473 * Either it failed or is pending. In the former case
12474 * ipIfStatsInDiscards was increased.
12475 */
12476 return (0);
12477 }
12478 }
12479 /*
12480 * We are done with IPsec processing. Send it over
12481 * the wire.
12482 */
12483 return (ip_output_post_ipsec(mp, ixa));
12484 }
12485
12486 /*
12487 * ioctls that go through a down/up sequence may need to wait for the down
12488 * to complete. This involves waiting for the ire and ipif refcnts to go down
12489 * to zero. Subsequently the ioctl is restarted from ipif_ill_refrele_tail.
12490 */
12491 /* ARGSUSED */
12492 void
12493 ip_reprocess_ioctl(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
12494 {
12495 struct iocblk *iocp;
12496 mblk_t *mp1;
12497 ip_ioctl_cmd_t *ipip;
12498 int err;
12499 sin_t *sin;
12500 struct lifreq *lifr;
12501 struct ifreq *ifr;
12502
12503 iocp = (struct iocblk *)mp->b_rptr;
12504 ASSERT(ipsq != NULL);
12505 /* Existence of mp1 verified in ip_wput_nondata */
12506 mp1 = mp->b_cont->b_cont;
12507 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12508 if (ipip->ipi_cmd == SIOCSLIFNAME || ipip->ipi_cmd == IF_UNITSEL) {
12509 /*
12510 * Special case where ipx_current_ipif is not set:
12511 * ill_phyint_reinit merged the v4 and v6 into a single ipsq.
12512 * We are here as were not able to complete the operation in
12513 * ipif_set_values because we could not become exclusive on
12514 * the new ipsq.
12515 */
12516 ill_t *ill = q->q_ptr;
12517 ipsq_current_start(ipsq, ill->ill_ipif, ipip->ipi_cmd);
12518 }
12519 ASSERT(ipsq->ipsq_xop->ipx_current_ipif != NULL);
12520
12521 if (ipip->ipi_cmd_type == IF_CMD) {
12522 /* This a old style SIOC[GS]IF* command */
12523 ifr = (struct ifreq *)mp1->b_rptr;
12524 sin = (sin_t *)&ifr->ifr_addr;
12525 } else if (ipip->ipi_cmd_type == LIF_CMD) {
12526 /* This a new style SIOC[GS]LIF* command */
12527 lifr = (struct lifreq *)mp1->b_rptr;
12528 sin = (sin_t *)&lifr->lifr_addr;
12529 } else {
12530 sin = NULL;
12531 }
12532
12533 err = (*ipip->ipi_func_restart)(ipsq->ipsq_xop->ipx_current_ipif, sin,
12534 q, mp, ipip, mp1->b_rptr);
12535
12536 DTRACE_PROBE4(ipif__ioctl, char *, "ip_reprocess_ioctl finish",
12537 int, ipip->ipi_cmd,
12538 ill_t *, ipsq->ipsq_xop->ipx_current_ipif->ipif_ill,
12539 ipif_t *, ipsq->ipsq_xop->ipx_current_ipif);
12540
12541 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), ipsq);
12542 }
12543
12544 /*
12545 * ioctl processing
12546 *
12547 * ioctl processing starts with ip_sioctl_copyin_setup(), which looks up
12548 * the ioctl command in the ioctl tables, determines the copyin data size
12549 * from the ipi_copyin_size field, and does an mi_copyin() of that size.
12550 *
12551 * ioctl processing then continues when the M_IOCDATA makes its way down to
12552 * ip_wput_nondata(). The ioctl is looked up again in the ioctl table, its
12553 * associated 'conn' is refheld till the end of the ioctl and the general
12554 * ioctl processing function ip_process_ioctl() is called to extract the
12555 * arguments and process the ioctl. To simplify extraction, ioctl commands
12556 * are "typed" based on the arguments they take (e.g., LIF_CMD which takes a
12557 * `struct lifreq'), and a common extract function (e.g., ip_extract_lifreq())
12558 * is used to extract the ioctl's arguments.
12559 *
12560 * ip_process_ioctl determines if the ioctl needs to be serialized, and if
12561 * so goes thru the serialization primitive ipsq_try_enter. Then the
12562 * appropriate function to handle the ioctl is called based on the entry in
12563 * the ioctl table. ioctl completion is encapsulated in ip_ioctl_finish
12564 * which also refreleases the 'conn' that was refheld at the start of the
12565 * ioctl. Finally ipsq_exit is called if needed to exit the ipsq.
12566 *
12567 * Many exclusive ioctls go thru an internal down up sequence as part of
12568 * the operation. For example an attempt to change the IP address of an
12569 * ipif entails ipif_down, set address, ipif_up. Bringing down the interface
12570 * does all the cleanup such as deleting all ires that use this address.
12571 * Then we need to wait till all references to the interface go away.
12572 */
12573 void
12574 ip_process_ioctl(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *arg)
12575 {
12576 struct iocblk *iocp = (struct iocblk *)mp->b_rptr;
12577 ip_ioctl_cmd_t *ipip = arg;
12578 ip_extract_func_t *extract_funcp;
12579 ill_t *ill;
12580 cmd_info_t ci;
12581 int err;
12582 boolean_t entered_ipsq = B_FALSE;
12583
12584 ip3dbg(("ip_process_ioctl: ioctl %X\n", iocp->ioc_cmd));
12585
12586 if (ipip == NULL)
12587 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12588
12589 /*
12590 * SIOCLIFADDIF needs to go thru a special path since the
12591 * ill may not exist yet. This happens in the case of lo0
12592 * which is created using this ioctl.
12593 */
12594 if (ipip->ipi_cmd == SIOCLIFADDIF) {
12595 err = ip_sioctl_addif(NULL, NULL, q, mp, NULL, NULL);
12596 DTRACE_PROBE4(ipif__ioctl, char *, "ip_process_ioctl finish",
12597 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12598 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12599 return;
12600 }
12601
12602 ci.ci_ipif = NULL;
12603 switch (ipip->ipi_cmd_type) {
12604 case MISC_CMD:
12605 case MSFILT_CMD:
12606 /*
12607 * All MISC_CMD ioctls come in here -- e.g. SIOCGLIFCONF.
12608 */
12609 if (ipip->ipi_cmd == IF_UNITSEL) {
12610 /* ioctl comes down the ill */
12611 ci.ci_ipif = ((ill_t *)q->q_ptr)->ill_ipif;
12612 ipif_refhold(ci.ci_ipif);
12613 }
12614 err = 0;
12615 ci.ci_sin = NULL;
12616 ci.ci_sin6 = NULL;
12617 ci.ci_lifr = NULL;
12618 extract_funcp = NULL;
12619 break;
12620
12621 case IF_CMD:
12622 case LIF_CMD:
12623 extract_funcp = ip_extract_lifreq;
12624 break;
12625
12626 case ARP_CMD:
12627 case XARP_CMD:
12628 extract_funcp = ip_extract_arpreq;
12629 break;
12630
12631 default:
12632 ASSERT(0);
12633 }
12634
12635 if (extract_funcp != NULL) {
12636 err = (*extract_funcp)(q, mp, ipip, &ci);
12637 if (err != 0) {
12638 DTRACE_PROBE4(ipif__ioctl,
12639 char *, "ip_process_ioctl finish err",
12640 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12641 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12642 return;
12643 }
12644
12645 /*
12646 * All of the extraction functions return a refheld ipif.
12647 */
12648 ASSERT(ci.ci_ipif != NULL);
12649 }
12650
12651 if (!(ipip->ipi_flags & IPI_WR)) {
12652 /*
12653 * A return value of EINPROGRESS means the ioctl is
12654 * either queued and waiting for some reason or has
12655 * already completed.
12656 */
12657 err = (*ipip->ipi_func)(ci.ci_ipif, ci.ci_sin, q, mp, ipip,
12658 ci.ci_lifr);
12659 if (ci.ci_ipif != NULL) {
12660 DTRACE_PROBE4(ipif__ioctl,
12661 char *, "ip_process_ioctl finish RD",
12662 int, ipip->ipi_cmd, ill_t *, ci.ci_ipif->ipif_ill,
12663 ipif_t *, ci.ci_ipif);
12664 ipif_refrele(ci.ci_ipif);
12665 } else {
12666 DTRACE_PROBE4(ipif__ioctl,
12667 char *, "ip_process_ioctl finish RD",
12668 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12669 }
12670 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12671 return;
12672 }
12673
12674 ASSERT(ci.ci_ipif != NULL);
12675
12676 /*
12677 * If ipsq is non-NULL, we are already being called exclusively
12678 */
12679 ASSERT(ipsq == NULL || IAM_WRITER_IPSQ(ipsq));
12680 if (ipsq == NULL) {
12681 ipsq = ipsq_try_enter(ci.ci_ipif, NULL, q, mp, ip_process_ioctl,
12682 NEW_OP, B_TRUE);
12683 if (ipsq == NULL) {
12684 ipif_refrele(ci.ci_ipif);
12685 return;
12686 }
12687 entered_ipsq = B_TRUE;
12688 }
12689 /*
12690 * Release the ipif so that ipif_down and friends that wait for
12691 * references to go away are not misled about the current ipif_refcnt
12692 * values. We are writer so we can access the ipif even after releasing
12693 * the ipif.
12694 */
12695 ipif_refrele(ci.ci_ipif);
12696
12697 ipsq_current_start(ipsq, ci.ci_ipif, ipip->ipi_cmd);
12698
12699 /*
12700 * We need to cache the ill_t that we're going to use as the argument
12701 * to the ipif-ioctl DTrace probe (below) because the ci_ipif can be
12702 * blown away by calling ipi_func.
12703 */
12704 ill = ci.ci_ipif == NULL ? NULL : ci.ci_ipif->ipif_ill;
12705
12706 /*
12707 * A return value of EINPROGRESS means the ioctl is
12708 * either queued and waiting for some reason or has
12709 * already completed.
12710 */
12711 err = (*ipip->ipi_func)(ci.ci_ipif, ci.ci_sin, q, mp, ipip, ci.ci_lifr);
12712
12713 DTRACE_PROBE4(ipif__ioctl, char *, "ip_process_ioctl finish WR",
12714 int, ipip->ipi_cmd, ill_t *, ill, ipif_t *, ci.ci_ipif);
12715 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), ipsq);
12716
12717 if (entered_ipsq)
12718 ipsq_exit(ipsq);
12719 }
12720
12721 /*
12722 * Complete the ioctl. Typically ioctls use the mi package and need to
12723 * do mi_copyout/mi_copy_done.
12724 */
12725 void
12726 ip_ioctl_finish(queue_t *q, mblk_t *mp, int err, int mode, ipsq_t *ipsq)
12727 {
12728 conn_t *connp = NULL;
12729
12730 if (err == EINPROGRESS)
12731 return;
12732
12733 if (CONN_Q(q)) {
12734 connp = Q_TO_CONN(q);
12735 ASSERT(connp->conn_ref >= 2);
12736 }
12737
12738 switch (mode) {
12739 case COPYOUT:
12740 if (err == 0)
12741 mi_copyout(q, mp);
12742 else
12743 mi_copy_done(q, mp, err);
12744 break;
12745
12746 case NO_COPYOUT:
12747 mi_copy_done(q, mp, err);
12748 break;
12749
12750 default:
12751 ASSERT(mode == CONN_CLOSE); /* aborted through CONN_CLOSE */
12752 break;
12753 }
12754
12755 /*
12756 * The conn refhold and ioctlref placed on the conn at the start of the
12757 * ioctl are released here.
12758 */
12759 if (connp != NULL) {
12760 CONN_DEC_IOCTLREF(connp);
12761 CONN_OPER_PENDING_DONE(connp);
12762 }
12763
12764 if (ipsq != NULL)
12765 ipsq_current_finish(ipsq);
12766 }
12767
12768 /* Handles all non data messages */
12769 void
12770 ip_wput_nondata(queue_t *q, mblk_t *mp)
12771 {
12772 mblk_t *mp1;
12773 struct iocblk *iocp;
12774 ip_ioctl_cmd_t *ipip;
12775 conn_t *connp;
12776 cred_t *cr;
12777 char *proto_str;
12778
12779 if (CONN_Q(q))
12780 connp = Q_TO_CONN(q);
12781 else
12782 connp = NULL;
12783
12784 switch (DB_TYPE(mp)) {
12785 case M_IOCTL:
12786 /*
12787 * IOCTL processing begins in ip_sioctl_copyin_setup which
12788 * will arrange to copy in associated control structures.
12789 */
12790 ip_sioctl_copyin_setup(q, mp);
12791 return;
12792 case M_IOCDATA:
12793 /*
12794 * Ensure that this is associated with one of our trans-
12795 * parent ioctls. If it's not ours, discard it if we're
12796 * running as a driver, or pass it on if we're a module.
12797 */
12798 iocp = (struct iocblk *)mp->b_rptr;
12799 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12800 if (ipip == NULL) {
12801 if (q->q_next == NULL) {
12802 goto nak;
12803 } else {
12804 putnext(q, mp);
12805 }
12806 return;
12807 }
12808 if ((q->q_next != NULL) && !(ipip->ipi_flags & IPI_MODOK)) {
12809 /*
12810 * The ioctl is one we recognise, but is not consumed
12811 * by IP as a module and we are a module, so we drop
12812 */
12813 goto nak;
12814 }
12815
12816 /* IOCTL continuation following copyin or copyout. */
12817 if (mi_copy_state(q, mp, NULL) == -1) {
12818 /*
12819 * The copy operation failed. mi_copy_state already
12820 * cleaned up, so we're out of here.
12821 */
12822 return;
12823 }
12824 /*
12825 * If we just completed a copy in, we become writer and
12826 * continue processing in ip_sioctl_copyin_done. If it
12827 * was a copy out, we call mi_copyout again. If there is
12828 * nothing more to copy out, it will complete the IOCTL.
12829 */
12830 if (MI_COPY_DIRECTION(mp) == MI_COPY_IN) {
12831 if (!(mp1 = mp->b_cont) || !(mp1 = mp1->b_cont)) {
12832 mi_copy_done(q, mp, EPROTO);
12833 return;
12834 }
12835 /*
12836 * Check for cases that need more copying. A return
12837 * value of 0 means a second copyin has been started,
12838 * so we return; a return value of 1 means no more
12839 * copying is needed, so we continue.
12840 */
12841 if (ipip->ipi_cmd_type == MSFILT_CMD &&
12842 MI_COPY_COUNT(mp) == 1) {
12843 if (ip_copyin_msfilter(q, mp) == 0)
12844 return;
12845 }
12846 /*
12847 * Refhold the conn, till the ioctl completes. This is
12848 * needed in case the ioctl ends up in the pending mp
12849 * list. Every mp in the ipx_pending_mp list must have
12850 * a refhold on the conn to resume processing. The
12851 * refhold is released when the ioctl completes
12852 * (whether normally or abnormally). An ioctlref is also
12853 * placed on the conn to prevent TCP from removing the
12854 * queue needed to send the ioctl reply back.
12855 * In all cases ip_ioctl_finish is called to finish
12856 * the ioctl and release the refholds.
12857 */
12858 if (connp != NULL) {
12859 /* This is not a reentry */
12860 CONN_INC_REF(connp);
12861 CONN_INC_IOCTLREF(connp);
12862 } else {
12863 if (!(ipip->ipi_flags & IPI_MODOK)) {
12864 mi_copy_done(q, mp, EINVAL);
12865 return;
12866 }
12867 }
12868
12869 ip_process_ioctl(NULL, q, mp, ipip);
12870
12871 } else {
12872 mi_copyout(q, mp);
12873 }
12874 return;
12875
12876 case M_IOCNAK:
12877 /*
12878 * The only way we could get here is if a resolver didn't like
12879 * an IOCTL we sent it. This shouldn't happen.
12880 */
12881 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
12882 "ip_wput_nondata: unexpected M_IOCNAK, ioc_cmd 0x%x",
12883 ((struct iocblk *)mp->b_rptr)->ioc_cmd);
12884 freemsg(mp);
12885 return;
12886 case M_IOCACK:
12887 /* /dev/ip shouldn't see this */
12888 goto nak;
12889 case M_FLUSH:
12890 if (*mp->b_rptr & FLUSHW)
12891 flushq(q, FLUSHALL);
12892 if (q->q_next) {
12893 putnext(q, mp);
12894 return;
12895 }
12896 if (*mp->b_rptr & FLUSHR) {
12897 *mp->b_rptr &= ~FLUSHW;
12898 qreply(q, mp);
12899 return;
12900 }
12901 freemsg(mp);
12902 return;
12903 case M_CTL:
12904 break;
12905 case M_PROTO:
12906 case M_PCPROTO:
12907 /*
12908 * The only PROTO messages we expect are SNMP-related.
12909 */
12910 switch (((union T_primitives *)mp->b_rptr)->type) {
12911 case T_SVR4_OPTMGMT_REQ:
12912 ip2dbg(("ip_wput_nondata: T_SVR4_OPTMGMT_REQ "
12913 "flags %x\n",
12914 ((struct T_optmgmt_req *)mp->b_rptr)->MGMT_flags));
12915
12916 if (connp == NULL) {
12917 proto_str = "T_SVR4_OPTMGMT_REQ";
12918 goto protonak;
12919 }
12920
12921 /*
12922 * All Solaris components should pass a db_credp
12923 * for this TPI message, hence we ASSERT.
12924 * But in case there is some other M_PROTO that looks
12925 * like a TPI message sent by some other kernel
12926 * component, we check and return an error.
12927 */
12928 cr = msg_getcred(mp, NULL);
12929 ASSERT(cr != NULL);
12930 if (cr == NULL) {
12931 mp = mi_tpi_err_ack_alloc(mp, TSYSERR, EINVAL);
12932 if (mp != NULL)
12933 qreply(q, mp);
12934 return;
12935 }
12936
12937 if (!snmpcom_req(q, mp, ip_snmp_set, ip_snmp_get, cr)) {
12938 proto_str = "Bad SNMPCOM request?";
12939 goto protonak;
12940 }
12941 return;
12942 default:
12943 ip1dbg(("ip_wput_nondata: dropping M_PROTO prim %u\n",
12944 (int)*(uint_t *)mp->b_rptr));
12945 freemsg(mp);
12946 return;
12947 }
12948 default:
12949 break;
12950 }
12951 if (q->q_next) {
12952 putnext(q, mp);
12953 } else
12954 freemsg(mp);
12955 return;
12956
12957 nak:
12958 iocp->ioc_error = EINVAL;
12959 mp->b_datap->db_type = M_IOCNAK;
12960 iocp->ioc_count = 0;
12961 qreply(q, mp);
12962 return;
12963
12964 protonak:
12965 cmn_err(CE_NOTE, "IP doesn't process %s as a module", proto_str);
12966 if ((mp = mi_tpi_err_ack_alloc(mp, TPROTO, EINVAL)) != NULL)
12967 qreply(q, mp);
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 void
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 }
13374
13375 /*
13376 * Callback to disable flow control in IP.
13377 *
13378 * This is a mac client callback added when the DLD_CAPAB_DIRECT capability
13379 * is enabled.
13380 *
13381 * When MAC_TX() is not able to send any more packets, dld sets its queue
13382 * to QFULL and enable the STREAMS flow control. Later, when the underlying
13383 * driver is able to continue to send packets, it calls mac_tx_(ring_)update()
13384 * function and wakes up corresponding mac worker threads, which in turn
13385 * calls this callback function, and disables flow control.
13386 */
13387 void
13388 ill_flow_enable(void *arg, ip_mac_tx_cookie_t cookie)
13389 {
13390 ill_t *ill = (ill_t *)arg;
13391 ip_stack_t *ipst = ill->ill_ipst;
13392 idl_tx_list_t *idl_txl;
13393
13394 idl_txl = &ipst->ips_idl_tx_list[IDLHASHINDEX(cookie)];
13395 mutex_enter(&idl_txl->txl_lock);
13396 /* add code to to set a flag to indicate idl_txl is enabled */
13397 conn_walk_drain(ipst, idl_txl);
13398 mutex_exit(&idl_txl->txl_lock);
13399 }
13400
13401 /*
13402 * Flow control has been relieved and STREAMS has backenabled us; drain
13403 * all the conn lists on `tx_list'.
13404 */
13405 static void
13406 conn_walk_drain(ip_stack_t *ipst, idl_tx_list_t *tx_list)
13407 {
13408 int i;
13409 idl_t *idl;
13410
13411 IP_STAT(ipst, ip_conn_walk_drain);
13412
13413 for (i = 0; i < ipst->ips_conn_drain_list_cnt; i++) {
13414 idl = &tx_list->txl_drain_list[i];
13415 mutex_enter(&idl->idl_lock);
13416 conn_drain(idl->idl_conn, B_FALSE);
13417 mutex_exit(&idl->idl_lock);
13418 }
13419 }
13420
13421 /*
13422 * Determine if the ill and multicast aspects of that packets
13423 * "matches" the conn.
13424 */
13425 boolean_t
13426 conn_wantpacket(conn_t *connp, ip_recv_attr_t *ira, ipha_t *ipha)
13427 {
13428 ill_t *ill = ira->ira_rill;
13429 zoneid_t zoneid = ira->ira_zoneid;
13430 uint_t in_ifindex;
13431 ipaddr_t dst, src;
13432
13433 dst = ipha->ipha_dst;
13434 src = ipha->ipha_src;
13435
13436 /*
13437 * conn_incoming_ifindex is set by IP_BOUND_IF which limits
13438 * unicast, broadcast and multicast reception to
13439 * conn_incoming_ifindex.
13440 * conn_wantpacket is called for unicast, broadcast and
13441 * multicast packets.
13442 */
13443 in_ifindex = connp->conn_incoming_ifindex;
13444
13445 /* mpathd can bind to the under IPMP interface, which we allow */
13446 if (in_ifindex != 0 && in_ifindex != ill->ill_phyint->phyint_ifindex) {
13447 if (!IS_UNDER_IPMP(ill))
13448 return (B_FALSE);
13449
13450 if (in_ifindex != ipmp_ill_get_ipmp_ifindex(ill))
13451 return (B_FALSE);
13452 }
13453
13454 if (!IPCL_ZONE_MATCH(connp, zoneid))
13455 return (B_FALSE);
13456
13457 if (!(ira->ira_flags & IRAF_MULTICAST))
13458 return (B_TRUE);
13459
13460 if (connp->conn_multi_router) {
13461 /* multicast packet and multicast router socket: send up */
13462 return (B_TRUE);
13463 }
13464
13465 if (ipha->ipha_protocol == IPPROTO_PIM ||
13466 ipha->ipha_protocol == IPPROTO_RSVP)
13467 return (B_TRUE);
13468
13469 return (conn_hasmembers_ill_withsrc_v4(connp, dst, src, ira->ira_ill));
13470 }
13471
13472 void
13473 conn_setqfull(conn_t *connp, boolean_t *flow_stopped)
13474 {
13475 if (IPCL_IS_NONSTR(connp)) {
13476 (*connp->conn_upcalls->su_txq_full)
13477 (connp->conn_upper_handle, B_TRUE);
13478 if (flow_stopped != NULL)
13479 *flow_stopped = B_TRUE;
13480 } else {
13481 queue_t *q = connp->conn_wq;
13482
13483 ASSERT(q != NULL);
13484 if (!(q->q_flag & QFULL)) {
13485 mutex_enter(QLOCK(q));
13486 if (!(q->q_flag & QFULL)) {
13487 /* still need to set QFULL */
13488 q->q_flag |= QFULL;
13489 /* set flow_stopped to true under QLOCK */
13490 if (flow_stopped != NULL)
13491 *flow_stopped = B_TRUE;
13492 mutex_exit(QLOCK(q));
13493 } else {
13494 /* flow_stopped is left unchanged */
13495 mutex_exit(QLOCK(q));
13496 }
13497 }
13498 }
13499 }
13500
13501 void
13502 conn_clrqfull(conn_t *connp, boolean_t *flow_stopped)
13503 {
13504 if (IPCL_IS_NONSTR(connp)) {
13505 (*connp->conn_upcalls->su_txq_full)
13506 (connp->conn_upper_handle, B_FALSE);
13507 if (flow_stopped != NULL)
13508 *flow_stopped = B_FALSE;
13509 } else {
13510 queue_t *q = connp->conn_wq;
13511
13512 ASSERT(q != NULL);
13513 if (q->q_flag & QFULL) {
13514 mutex_enter(QLOCK(q));
13515 if (q->q_flag & QFULL) {
13516 q->q_flag &= ~QFULL;
13517 /* set flow_stopped to false under QLOCK */
13518 if (flow_stopped != NULL)
13519 *flow_stopped = B_FALSE;
13520 mutex_exit(QLOCK(q));
13521 if (q->q_flag & QWANTW)
13522 qbackenable(q, 0);
13523 } else {
13524 /* flow_stopped is left unchanged */
13525 mutex_exit(QLOCK(q));
13526 }
13527 }
13528 }
13529
13530 mutex_enter(&connp->conn_lock);
13531 connp->conn_blocked = B_FALSE;
13532 mutex_exit(&connp->conn_lock);
13533 }
13534
13535 /*
13536 * Return the length in bytes of the IPv4 headers (base header, label, and
13537 * other IP options) that will be needed based on the
13538 * ip_pkt_t structure passed by the caller.
13539 *
13540 * The returned length does not include the length of the upper level
13541 * protocol (ULP) header.
13542 * The caller needs to check that the length doesn't exceed the max for IPv4.
13543 */
13544 int
13545 ip_total_hdrs_len_v4(const ip_pkt_t *ipp)
13546 {
13547 int len;
13548
13549 len = IP_SIMPLE_HDR_LENGTH;
13550 if (ipp->ipp_fields & IPPF_LABEL_V4) {
13551 ASSERT(ipp->ipp_label_len_v4 != 0);
13552 /* We need to round up here */
13553 len += (ipp->ipp_label_len_v4 + 3) & ~3;
13554 }
13555
13556 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
13557 ASSERT(ipp->ipp_ipv4_options_len != 0);
13558 ASSERT((ipp->ipp_ipv4_options_len & 3) == 0);
13559 len += ipp->ipp_ipv4_options_len;
13560 }
13561 return (len);
13562 }
13563
13564 /*
13565 * All-purpose routine to build an IPv4 header with options based
13566 * on the abstract ip_pkt_t.
13567 *
13568 * The caller has to set the source and destination address as well as
13569 * ipha_length. The caller has to massage any source route and compensate
13570 * for the ULP pseudo-header checksum due to the source route.
13571 */
13572 void
13573 ip_build_hdrs_v4(uchar_t *buf, uint_t buf_len, const ip_pkt_t *ipp,
13574 uint8_t protocol)
13575 {
13576 ipha_t *ipha = (ipha_t *)buf;
13577 uint8_t *cp;
13578
13579 /* Initialize IPv4 header */
13580 ipha->ipha_type_of_service = ipp->ipp_type_of_service;
13581 ipha->ipha_length = 0; /* Caller will set later */
13582 ipha->ipha_ident = 0;
13583 ipha->ipha_fragment_offset_and_flags = 0;
13584 ipha->ipha_ttl = ipp->ipp_unicast_hops;
13585 ipha->ipha_protocol = protocol;
13586 ipha->ipha_hdr_checksum = 0;
13587
13588 if ((ipp->ipp_fields & IPPF_ADDR) &&
13589 IN6_IS_ADDR_V4MAPPED(&ipp->ipp_addr))
13590 ipha->ipha_src = ipp->ipp_addr_v4;
13591
13592 cp = (uint8_t *)&ipha[1];
13593 if (ipp->ipp_fields & IPPF_LABEL_V4) {
13594 ASSERT(ipp->ipp_label_len_v4 != 0);
13595 bcopy(ipp->ipp_label_v4, cp, ipp->ipp_label_len_v4);
13596 cp += ipp->ipp_label_len_v4;
13597 /* We need to round up here */
13598 while ((uintptr_t)cp & 0x3) {
13599 *cp++ = IPOPT_NOP;
13600 }
13601 }
13602
13603 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
13604 ASSERT(ipp->ipp_ipv4_options_len != 0);
13605 ASSERT((ipp->ipp_ipv4_options_len & 3) == 0);
13606 bcopy(ipp->ipp_ipv4_options, cp, ipp->ipp_ipv4_options_len);
13607 cp += ipp->ipp_ipv4_options_len;
13608 }
13609 ipha->ipha_version_and_hdr_length =
13610 (uint8_t)((IP_VERSION << 4) + buf_len / 4);
13611
13612 ASSERT((int)(cp - buf) == buf_len);
13613 }
13614
13615 /* Allocate the private structure */
13616 static int
13617 ip_priv_alloc(void **bufp)
13618 {
13619 void *buf;
13620
13621 if ((buf = kmem_alloc(sizeof (ip_priv_t), KM_NOSLEEP)) == NULL)
13622 return (ENOMEM);
13623
13624 *bufp = buf;
13625 return (0);
13626 }
13627
13628 /* Function to delete the private structure */
13629 void
13630 ip_priv_free(void *buf)
13631 {
13632 ASSERT(buf != NULL);
13633 kmem_free(buf, sizeof (ip_priv_t));
13634 }
13635
13636 /*
13637 * The entry point for IPPF processing.
13638 * If the classifier (IPGPC_CLASSIFY) is not loaded and configured, the
13639 * routine just returns.
13640 *
13641 * When called, ip_process generates an ipp_packet_t structure
13642 * which holds the state information for this packet and invokes the
13643 * the classifier (via ipp_packet_process). The classification, depending on
13644 * configured filters, results in a list of actions for this packet. Invoking
13645 * an action may cause the packet to be dropped, in which case we return NULL.
13646 * proc indicates the callout position for
13647 * this packet and ill is the interface this packet arrived on or will leave
13648 * on (inbound and outbound resp.).
13649 *
13650 * We do the processing on the rill (mapped to the upper if ipmp), but MIB
13651 * on the ill corrsponding to the destination IP address.
13652 */
13653 mblk_t *
13654 ip_process(ip_proc_t proc, mblk_t *mp, ill_t *rill, ill_t *ill)
13655 {
13656 ip_priv_t *priv;
13657 ipp_action_id_t aid;
13658 int rc = 0;
13659 ipp_packet_t *pp;
13660
13661 /* If the classifier is not loaded, return */
13662 if ((aid = ipp_action_lookup(IPGPC_CLASSIFY)) == IPP_ACTION_INVAL) {
13663 return (mp);
13664 }
13665
13666 ASSERT(mp != NULL);
13667
13668 /* Allocate the packet structure */
13669 rc = ipp_packet_alloc(&pp, "ip", aid);
13670 if (rc != 0)
13671 goto drop;
13672
13673 /* Allocate the private structure */
13674 rc = ip_priv_alloc((void **)&priv);
13675 if (rc != 0) {
13676 ipp_packet_free(pp);
13677 goto drop;
13678 }
13679 priv->proc = proc;
13680 priv->ill_index = ill_get_upper_ifindex(rill);
13681
13682 ipp_packet_set_private(pp, priv, ip_priv_free);
13683 ipp_packet_set_data(pp, mp);
13684
13685 /* Invoke the classifier */
13686 rc = ipp_packet_process(&pp);
13687 if (pp != NULL) {
13688 mp = ipp_packet_get_data(pp);
13689 ipp_packet_free(pp);
13690 if (rc != 0)
13691 goto drop;
13692 return (mp);
13693 } else {
13694 /* No mp to trace in ip_drop_input/ip_drop_output */
13695 mp = NULL;
13696 }
13697 drop:
13698 if (proc == IPP_LOCAL_IN || proc == IPP_FWD_IN) {
13699 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
13700 ip_drop_input("ip_process", mp, ill);
13701 } else {
13702 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
13703 ip_drop_output("ip_process", mp, ill);
13704 }
13705 freemsg(mp);
13706 return (NULL);
13707 }
13708
13709 /*
13710 * Propagate a multicast group membership operation (add/drop) on
13711 * all the interfaces crossed by the related multirt routes.
13712 * The call is considered successful if the operation succeeds
13713 * on at least one interface.
13714 *
13715 * This assumes that a set of IRE_HOST/RTF_MULTIRT has been created for the
13716 * multicast addresses with the ire argument being the first one.
13717 * We walk the bucket to find all the of those.
13718 *
13719 * Common to IPv4 and IPv6.
13720 */
13721 static int
13722 ip_multirt_apply_membership(int (*fn)(conn_t *, boolean_t,
13723 const in6_addr_t *, ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *),
13724 ire_t *ire, conn_t *connp, boolean_t checkonly, const in6_addr_t *v6group,
13725 mcast_record_t fmode, const in6_addr_t *v6src)
13726 {
13727 ire_t *ire_gw;
13728 irb_t *irb;
13729 int ifindex;
13730 int error = 0;
13731 int result;
13732 ip_stack_t *ipst = ire->ire_ipst;
13733 ipaddr_t group;
13734 boolean_t isv6;
13735 int match_flags;
13736
13737 if (IN6_IS_ADDR_V4MAPPED(v6group)) {
13738 IN6_V4MAPPED_TO_IPADDR(v6group, group);
13739 isv6 = B_FALSE;
13740 } else {
13741 isv6 = B_TRUE;
13742 }
13743
13744 irb = ire->ire_bucket;
13745 ASSERT(irb != NULL);
13746
13747 result = 0;
13748 irb_refhold(irb);
13749 for (; ire != NULL; ire = ire->ire_next) {
13750 if ((ire->ire_flags & RTF_MULTIRT) == 0)
13751 continue;
13752
13753 /* We handle -ifp routes by matching on the ill if set */
13754 match_flags = MATCH_IRE_TYPE;
13755 if (ire->ire_ill != NULL)
13756 match_flags |= MATCH_IRE_ILL;
13757
13758 if (isv6) {
13759 if (!IN6_ARE_ADDR_EQUAL(&ire->ire_addr_v6, v6group))
13760 continue;
13761
13762 ire_gw = ire_ftable_lookup_v6(&ire->ire_gateway_addr_v6,
13763 0, 0, IRE_INTERFACE, ire->ire_ill, ALL_ZONES, NULL,
13764 match_flags, 0, ipst, NULL);
13765 } else {
13766 if (ire->ire_addr != group)
13767 continue;
13768
13769 ire_gw = ire_ftable_lookup_v4(ire->ire_gateway_addr,
13770 0, 0, IRE_INTERFACE, ire->ire_ill, ALL_ZONES, NULL,
13771 match_flags, 0, ipst, NULL);
13772 }
13773 /* No interface route exists for the gateway; skip this ire. */
13774 if (ire_gw == NULL)
13775 continue;
13776 if (ire_gw->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
13777 ire_refrele(ire_gw);
13778 continue;
13779 }
13780 ASSERT(ire_gw->ire_ill != NULL); /* IRE_INTERFACE */
13781 ifindex = ire_gw->ire_ill->ill_phyint->phyint_ifindex;
13782
13783 /*
13784 * The operation is considered a success if
13785 * it succeeds at least once on any one interface.
13786 */
13787 error = fn(connp, checkonly, v6group, INADDR_ANY, ifindex,
13788 fmode, v6src);
13789 if (error == 0)
13790 result = CGTP_MCAST_SUCCESS;
13791
13792 ire_refrele(ire_gw);
13793 }
13794 irb_refrele(irb);
13795 /*
13796 * Consider the call as successful if we succeeded on at least
13797 * one interface. Otherwise, return the last encountered error.
13798 */
13799 return (result == CGTP_MCAST_SUCCESS ? 0 : error);
13800 }
13801
13802 /*
13803 * Return the expected CGTP hooks version number.
13804 */
13805 int
13806 ip_cgtp_filter_supported(void)
13807 {
13808 return (ip_cgtp_filter_rev);
13809 }
13810
13811 /*
13812 * CGTP hooks can be registered by invoking this function.
13813 * Checks that the version number matches.
13814 */
13815 int
13816 ip_cgtp_filter_register(netstackid_t stackid, cgtp_filter_ops_t *ops)
13817 {
13818 netstack_t *ns;
13819 ip_stack_t *ipst;
13820
13821 if (ops->cfo_filter_rev != CGTP_FILTER_REV)
13822 return (ENOTSUP);
13823
13824 ns = netstack_find_by_stackid(stackid);
13825 if (ns == NULL)
13826 return (EINVAL);
13827 ipst = ns->netstack_ip;
13828 ASSERT(ipst != NULL);
13829
13830 if (ipst->ips_ip_cgtp_filter_ops != NULL) {
13831 netstack_rele(ns);
13832 return (EALREADY);
13833 }
13834
13835 ipst->ips_ip_cgtp_filter_ops = ops;
13836
13837 ill_set_inputfn_all(ipst);
13838
13839 netstack_rele(ns);
13840 return (0);
13841 }
13842
13843 /*
13844 * CGTP hooks can be unregistered by invoking this function.
13845 * Returns ENXIO if there was no registration.
13846 * Returns EBUSY if the ndd variable has not been turned off.
13847 */
13848 int
13849 ip_cgtp_filter_unregister(netstackid_t stackid)
13850 {
13851 netstack_t *ns;
13852 ip_stack_t *ipst;
13853
13854 ns = netstack_find_by_stackid(stackid);
13855 if (ns == NULL)
13856 return (EINVAL);
13857 ipst = ns->netstack_ip;
13858 ASSERT(ipst != NULL);
13859
13860 if (ipst->ips_ip_cgtp_filter) {
13861 netstack_rele(ns);
13862 return (EBUSY);
13863 }
13864
13865 if (ipst->ips_ip_cgtp_filter_ops == NULL) {
13866 netstack_rele(ns);
13867 return (ENXIO);
13868 }
13869 ipst->ips_ip_cgtp_filter_ops = NULL;
13870
13871 ill_set_inputfn_all(ipst);
13872
13873 netstack_rele(ns);
13874 return (0);
13875 }
13876
13877 /*
13878 * Check whether there is a CGTP filter registration.
13879 * Returns non-zero if there is a registration, otherwise returns zero.
13880 * Note: returns zero if bad stackid.
13881 */
13882 int
13883 ip_cgtp_filter_is_registered(netstackid_t stackid)
13884 {
13885 netstack_t *ns;
13886 ip_stack_t *ipst;
13887 int ret;
13888
13889 ns = netstack_find_by_stackid(stackid);
13890 if (ns == NULL)
13891 return (0);
13892 ipst = ns->netstack_ip;
13893 ASSERT(ipst != NULL);
13894
13895 if (ipst->ips_ip_cgtp_filter_ops != NULL)
13896 ret = 1;
13897 else
13898 ret = 0;
13899
13900 netstack_rele(ns);
13901 return (ret);
13902 }
13903
13904 static int
13905 ip_squeue_switch(int val)
13906 {
13907 int rval;
13908
13909 switch (val) {
13910 case IP_SQUEUE_ENTER_NODRAIN:
13911 rval = SQ_NODRAIN;
13912 break;
13913 case IP_SQUEUE_ENTER:
13914 rval = SQ_PROCESS;
13915 break;
13916 case IP_SQUEUE_FILL:
13917 default:
13918 rval = SQ_FILL;
13919 break;
13920 }
13921 return (rval);
13922 }
13923
13924 static void *
13925 ip_kstat2_init(netstackid_t stackid, ip_stat_t *ip_statisticsp)
13926 {
13927 kstat_t *ksp;
13928
13929 ip_stat_t template = {
13930 { "ip_udp_fannorm", KSTAT_DATA_UINT64 },
13931 { "ip_udp_fanmb", KSTAT_DATA_UINT64 },
13932 { "ip_recv_pullup", KSTAT_DATA_UINT64 },
13933 { "ip_db_ref", KSTAT_DATA_UINT64 },
13934 { "ip_notaligned", KSTAT_DATA_UINT64 },
13935 { "ip_multimblk", KSTAT_DATA_UINT64 },
13936 { "ip_opt", KSTAT_DATA_UINT64 },
13937 { "ipsec_proto_ahesp", KSTAT_DATA_UINT64 },
13938 { "ip_conn_flputbq", KSTAT_DATA_UINT64 },
13939 { "ip_conn_walk_drain", KSTAT_DATA_UINT64 },
13940 { "ip_out_sw_cksum", KSTAT_DATA_UINT64 },
13941 { "ip_out_sw_cksum_bytes", KSTAT_DATA_UINT64 },
13942 { "ip_in_sw_cksum", KSTAT_DATA_UINT64 },
13943 { "ip_ire_reclaim_calls", KSTAT_DATA_UINT64 },
13944 { "ip_ire_reclaim_deleted", KSTAT_DATA_UINT64 },
13945 { "ip_nce_reclaim_calls", KSTAT_DATA_UINT64 },
13946 { "ip_nce_reclaim_deleted", KSTAT_DATA_UINT64 },
13947 { "ip_dce_reclaim_calls", KSTAT_DATA_UINT64 },
13948 { "ip_dce_reclaim_deleted", KSTAT_DATA_UINT64 },
13949 { "ip_tcp_in_full_hw_cksum_err", KSTAT_DATA_UINT64 },
13950 { "ip_tcp_in_part_hw_cksum_err", KSTAT_DATA_UINT64 },
13951 { "ip_tcp_in_sw_cksum_err", KSTAT_DATA_UINT64 },
13952 { "ip_udp_in_full_hw_cksum_err", KSTAT_DATA_UINT64 },
13953 { "ip_udp_in_part_hw_cksum_err", KSTAT_DATA_UINT64 },
13954 { "ip_udp_in_sw_cksum_err", KSTAT_DATA_UINT64 },
13955 { "conn_in_recvdstaddr", KSTAT_DATA_UINT64 },
13956 { "conn_in_recvopts", KSTAT_DATA_UINT64 },
13957 { "conn_in_recvif", KSTAT_DATA_UINT64 },
13958 { "conn_in_recvslla", KSTAT_DATA_UINT64 },
13959 { "conn_in_recvucred", KSTAT_DATA_UINT64 },
13960 { "conn_in_recvttl", KSTAT_DATA_UINT64 },
13961 { "conn_in_recvhopopts", KSTAT_DATA_UINT64 },
13962 { "conn_in_recvhoplimit", KSTAT_DATA_UINT64 },
13963 { "conn_in_recvdstopts", KSTAT_DATA_UINT64 },
13964 { "conn_in_recvrthdrdstopts", KSTAT_DATA_UINT64 },
13965 { "conn_in_recvrthdr", KSTAT_DATA_UINT64 },
13966 { "conn_in_recvpktinfo", KSTAT_DATA_UINT64 },
13967 { "conn_in_recvtclass", KSTAT_DATA_UINT64 },
13968 { "conn_in_timestamp", KSTAT_DATA_UINT64 },
13969 };
13970
13971 ksp = kstat_create_netstack("ip", 0, "ipstat", "net",
13972 KSTAT_TYPE_NAMED, sizeof (template) / sizeof (kstat_named_t),
13973 KSTAT_FLAG_VIRTUAL, stackid);
13974
13975 if (ksp == NULL)
13976 return (NULL);
13977
13978 bcopy(&template, ip_statisticsp, sizeof (template));
13979 ksp->ks_data = (void *)ip_statisticsp;
13980 ksp->ks_private = (void *)(uintptr_t)stackid;
13981
13982 kstat_install(ksp);
13983 return (ksp);
13984 }
13985
13986 static void
13987 ip_kstat2_fini(netstackid_t stackid, kstat_t *ksp)
13988 {
13989 if (ksp != NULL) {
13990 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
13991 kstat_delete_netstack(ksp, stackid);
13992 }
13993 }
13994
13995 static void *
13996 ip_kstat_init(netstackid_t stackid, ip_stack_t *ipst)
13997 {
13998 kstat_t *ksp;
13999
14000 ip_named_kstat_t template = {
14001 { "forwarding", KSTAT_DATA_UINT32, 0 },
14002 { "defaultTTL", KSTAT_DATA_UINT32, 0 },
14003 { "inReceives", KSTAT_DATA_UINT64, 0 },
14004 { "inHdrErrors", KSTAT_DATA_UINT32, 0 },
14005 { "inAddrErrors", KSTAT_DATA_UINT32, 0 },
14006 { "forwDatagrams", KSTAT_DATA_UINT64, 0 },
14007 { "inUnknownProtos", KSTAT_DATA_UINT32, 0 },
14008 { "inDiscards", KSTAT_DATA_UINT32, 0 },
14009 { "inDelivers", KSTAT_DATA_UINT64, 0 },
14010 { "outRequests", KSTAT_DATA_UINT64, 0 },
14011 { "outDiscards", KSTAT_DATA_UINT32, 0 },
14012 { "outNoRoutes", KSTAT_DATA_UINT32, 0 },
14013 { "reasmTimeout", KSTAT_DATA_UINT32, 0 },
14014 { "reasmReqds", KSTAT_DATA_UINT32, 0 },
14015 { "reasmOKs", KSTAT_DATA_UINT32, 0 },
14016 { "reasmFails", KSTAT_DATA_UINT32, 0 },
14017 { "fragOKs", KSTAT_DATA_UINT32, 0 },
14018 { "fragFails", KSTAT_DATA_UINT32, 0 },
14019 { "fragCreates", KSTAT_DATA_UINT32, 0 },
14020 { "addrEntrySize", KSTAT_DATA_INT32, 0 },
14021 { "routeEntrySize", KSTAT_DATA_INT32, 0 },
14022 { "netToMediaEntrySize", KSTAT_DATA_INT32, 0 },
14023 { "routingDiscards", KSTAT_DATA_UINT32, 0 },
14024 { "inErrs", KSTAT_DATA_UINT32, 0 },
14025 { "noPorts", KSTAT_DATA_UINT32, 0 },
14026 { "inCksumErrs", KSTAT_DATA_UINT32, 0 },
14027 { "reasmDuplicates", KSTAT_DATA_UINT32, 0 },
14028 { "reasmPartDups", KSTAT_DATA_UINT32, 0 },
14029 { "forwProhibits", KSTAT_DATA_UINT32, 0 },
14030 { "udpInCksumErrs", KSTAT_DATA_UINT32, 0 },
14031 { "udpInOverflows", KSTAT_DATA_UINT32, 0 },
14032 { "rawipInOverflows", KSTAT_DATA_UINT32, 0 },
14033 { "ipsecInSucceeded", KSTAT_DATA_UINT32, 0 },
14034 { "ipsecInFailed", KSTAT_DATA_INT32, 0 },
14035 { "memberEntrySize", KSTAT_DATA_INT32, 0 },
14036 { "inIPv6", KSTAT_DATA_UINT32, 0 },
14037 { "outIPv6", KSTAT_DATA_UINT32, 0 },
14038 { "outSwitchIPv6", KSTAT_DATA_UINT32, 0 },
14039 };
14040
14041 ksp = kstat_create_netstack("ip", 0, "ip", "mib2", KSTAT_TYPE_NAMED,
14042 NUM_OF_FIELDS(ip_named_kstat_t), 0, stackid);
14043 if (ksp == NULL || ksp->ks_data == NULL)
14044 return (NULL);
14045
14046 template.forwarding.value.ui32 = WE_ARE_FORWARDING(ipst) ? 1:2;
14047 template.defaultTTL.value.ui32 = (uint32_t)ipst->ips_ip_def_ttl;
14048 template.reasmTimeout.value.ui32 = ipst->ips_ip_reassembly_timeout;
14049 template.addrEntrySize.value.i32 = sizeof (mib2_ipAddrEntry_t);
14050 template.routeEntrySize.value.i32 = sizeof (mib2_ipRouteEntry_t);
14051
14052 template.netToMediaEntrySize.value.i32 =
14053 sizeof (mib2_ipNetToMediaEntry_t);
14054
14055 template.memberEntrySize.value.i32 = sizeof (ipv6_member_t);
14056
14057 bcopy(&template, ksp->ks_data, sizeof (template));
14058 ksp->ks_update = ip_kstat_update;
14059 ksp->ks_private = (void *)(uintptr_t)stackid;
14060
14061 kstat_install(ksp);
14062 return (ksp);
14063 }
14064
14065 static void
14066 ip_kstat_fini(netstackid_t stackid, kstat_t *ksp)
14067 {
14068 if (ksp != NULL) {
14069 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
14070 kstat_delete_netstack(ksp, stackid);
14071 }
14072 }
14073
14074 static int
14075 ip_kstat_update(kstat_t *kp, int rw)
14076 {
14077 ip_named_kstat_t *ipkp;
14078 mib2_ipIfStatsEntry_t ipmib;
14079 ill_walk_context_t ctx;
14080 ill_t *ill;
14081 netstackid_t stackid = (zoneid_t)(uintptr_t)kp->ks_private;
14082 netstack_t *ns;
14083 ip_stack_t *ipst;
14084
14085 if (kp == NULL || kp->ks_data == NULL)
14086 return (EIO);
14087
14088 if (rw == KSTAT_WRITE)
14089 return (EACCES);
14090
14091 ns = netstack_find_by_stackid(stackid);
14092 if (ns == NULL)
14093 return (-1);
14094 ipst = ns->netstack_ip;
14095 if (ipst == NULL) {
14096 netstack_rele(ns);
14097 return (-1);
14098 }
14099 ipkp = (ip_named_kstat_t *)kp->ks_data;
14100
14101 bcopy(&ipst->ips_ip_mib, &ipmib, sizeof (ipmib));
14102 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
14103 ill = ILL_START_WALK_V4(&ctx, ipst);
14104 for (; ill != NULL; ill = ill_next(&ctx, ill))
14105 ip_mib2_add_ip_stats(&ipmib, ill->ill_ip_mib);
14106 rw_exit(&ipst->ips_ill_g_lock);
14107
14108 ipkp->forwarding.value.ui32 = ipmib.ipIfStatsForwarding;
14109 ipkp->defaultTTL.value.ui32 = ipmib.ipIfStatsDefaultTTL;
14110 ipkp->inReceives.value.ui64 = ipmib.ipIfStatsHCInReceives;
14111 ipkp->inHdrErrors.value.ui32 = ipmib.ipIfStatsInHdrErrors;
14112 ipkp->inAddrErrors.value.ui32 = ipmib.ipIfStatsInAddrErrors;
14113 ipkp->forwDatagrams.value.ui64 = ipmib.ipIfStatsHCOutForwDatagrams;
14114 ipkp->inUnknownProtos.value.ui32 = ipmib.ipIfStatsInUnknownProtos;
14115 ipkp->inDiscards.value.ui32 = ipmib.ipIfStatsInDiscards;
14116 ipkp->inDelivers.value.ui64 = ipmib.ipIfStatsHCInDelivers;
14117 ipkp->outRequests.value.ui64 = ipmib.ipIfStatsHCOutRequests;
14118 ipkp->outDiscards.value.ui32 = ipmib.ipIfStatsOutDiscards;
14119 ipkp->outNoRoutes.value.ui32 = ipmib.ipIfStatsOutNoRoutes;
14120 ipkp->reasmTimeout.value.ui32 = ipst->ips_ip_reassembly_timeout;
14121 ipkp->reasmReqds.value.ui32 = ipmib.ipIfStatsReasmReqds;
14122 ipkp->reasmOKs.value.ui32 = ipmib.ipIfStatsReasmOKs;
14123 ipkp->reasmFails.value.ui32 = ipmib.ipIfStatsReasmFails;
14124 ipkp->fragOKs.value.ui32 = ipmib.ipIfStatsOutFragOKs;
14125 ipkp->fragFails.value.ui32 = ipmib.ipIfStatsOutFragFails;
14126 ipkp->fragCreates.value.ui32 = ipmib.ipIfStatsOutFragCreates;
14127
14128 ipkp->routingDiscards.value.ui32 = 0;
14129 ipkp->inErrs.value.ui32 = ipmib.tcpIfStatsInErrs;
14130 ipkp->noPorts.value.ui32 = ipmib.udpIfStatsNoPorts;
14131 ipkp->inCksumErrs.value.ui32 = ipmib.ipIfStatsInCksumErrs;
14132 ipkp->reasmDuplicates.value.ui32 = ipmib.ipIfStatsReasmDuplicates;
14133 ipkp->reasmPartDups.value.ui32 = ipmib.ipIfStatsReasmPartDups;
14134 ipkp->forwProhibits.value.ui32 = ipmib.ipIfStatsForwProhibits;
14135 ipkp->udpInCksumErrs.value.ui32 = ipmib.udpIfStatsInCksumErrs;
14136 ipkp->udpInOverflows.value.ui32 = ipmib.udpIfStatsInOverflows;
14137 ipkp->rawipInOverflows.value.ui32 = ipmib.rawipIfStatsInOverflows;
14138 ipkp->ipsecInSucceeded.value.ui32 = ipmib.ipsecIfStatsInSucceeded;
14139 ipkp->ipsecInFailed.value.i32 = ipmib.ipsecIfStatsInFailed;
14140
14141 ipkp->inIPv6.value.ui32 = ipmib.ipIfStatsInWrongIPVersion;
14142 ipkp->outIPv6.value.ui32 = ipmib.ipIfStatsOutWrongIPVersion;
14143 ipkp->outSwitchIPv6.value.ui32 = ipmib.ipIfStatsOutSwitchIPVersion;
14144
14145 netstack_rele(ns);
14146
14147 return (0);
14148 }
14149
14150 static void *
14151 icmp_kstat_init(netstackid_t stackid)
14152 {
14153 kstat_t *ksp;
14154
14155 icmp_named_kstat_t template = {
14156 { "inMsgs", KSTAT_DATA_UINT32 },
14157 { "inErrors", KSTAT_DATA_UINT32 },
14158 { "inDestUnreachs", KSTAT_DATA_UINT32 },
14159 { "inTimeExcds", KSTAT_DATA_UINT32 },
14160 { "inParmProbs", KSTAT_DATA_UINT32 },
14161 { "inSrcQuenchs", KSTAT_DATA_UINT32 },
14162 { "inRedirects", KSTAT_DATA_UINT32 },
14163 { "inEchos", KSTAT_DATA_UINT32 },
14164 { "inEchoReps", KSTAT_DATA_UINT32 },
14165 { "inTimestamps", KSTAT_DATA_UINT32 },
14166 { "inTimestampReps", KSTAT_DATA_UINT32 },
14167 { "inAddrMasks", KSTAT_DATA_UINT32 },
14168 { "inAddrMaskReps", KSTAT_DATA_UINT32 },
14169 { "outMsgs", KSTAT_DATA_UINT32 },
14170 { "outErrors", KSTAT_DATA_UINT32 },
14171 { "outDestUnreachs", KSTAT_DATA_UINT32 },
14172 { "outTimeExcds", KSTAT_DATA_UINT32 },
14173 { "outParmProbs", KSTAT_DATA_UINT32 },
14174 { "outSrcQuenchs", KSTAT_DATA_UINT32 },
14175 { "outRedirects", KSTAT_DATA_UINT32 },
14176 { "outEchos", KSTAT_DATA_UINT32 },
14177 { "outEchoReps", KSTAT_DATA_UINT32 },
14178 { "outTimestamps", KSTAT_DATA_UINT32 },
14179 { "outTimestampReps", KSTAT_DATA_UINT32 },
14180 { "outAddrMasks", KSTAT_DATA_UINT32 },
14181 { "outAddrMaskReps", KSTAT_DATA_UINT32 },
14182 { "inChksumErrs", KSTAT_DATA_UINT32 },
14183 { "inUnknowns", KSTAT_DATA_UINT32 },
14184 { "inFragNeeded", KSTAT_DATA_UINT32 },
14185 { "outFragNeeded", KSTAT_DATA_UINT32 },
14186 { "outDrops", KSTAT_DATA_UINT32 },
14187 { "inOverFlows", KSTAT_DATA_UINT32 },
14188 { "inBadRedirects", KSTAT_DATA_UINT32 },
14189 };
14190
14191 ksp = kstat_create_netstack("ip", 0, "icmp", "mib2", KSTAT_TYPE_NAMED,
14192 NUM_OF_FIELDS(icmp_named_kstat_t), 0, stackid);
14193 if (ksp == NULL || ksp->ks_data == NULL)
14194 return (NULL);
14195
14196 bcopy(&template, ksp->ks_data, sizeof (template));
14197
14198 ksp->ks_update = icmp_kstat_update;
14199 ksp->ks_private = (void *)(uintptr_t)stackid;
14200
14201 kstat_install(ksp);
14202 return (ksp);
14203 }
14204
14205 static void
14206 icmp_kstat_fini(netstackid_t stackid, kstat_t *ksp)
14207 {
14208 if (ksp != NULL) {
14209 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
14210 kstat_delete_netstack(ksp, stackid);
14211 }
14212 }
14213
14214 static int
14215 icmp_kstat_update(kstat_t *kp, int rw)
14216 {
14217 icmp_named_kstat_t *icmpkp;
14218 netstackid_t stackid = (zoneid_t)(uintptr_t)kp->ks_private;
14219 netstack_t *ns;
14220 ip_stack_t *ipst;
14221
14222 if ((kp == NULL) || (kp->ks_data == NULL))
14223 return (EIO);
14224
14225 if (rw == KSTAT_WRITE)
14226 return (EACCES);
14227
14228 ns = netstack_find_by_stackid(stackid);
14229 if (ns == NULL)
14230 return (-1);
14231 ipst = ns->netstack_ip;
14232 if (ipst == NULL) {
14233 netstack_rele(ns);
14234 return (-1);
14235 }
14236 icmpkp = (icmp_named_kstat_t *)kp->ks_data;
14237
14238 icmpkp->inMsgs.value.ui32 = ipst->ips_icmp_mib.icmpInMsgs;
14239 icmpkp->inErrors.value.ui32 = ipst->ips_icmp_mib.icmpInErrors;
14240 icmpkp->inDestUnreachs.value.ui32 =
14241 ipst->ips_icmp_mib.icmpInDestUnreachs;
14242 icmpkp->inTimeExcds.value.ui32 = ipst->ips_icmp_mib.icmpInTimeExcds;
14243 icmpkp->inParmProbs.value.ui32 = ipst->ips_icmp_mib.icmpInParmProbs;
14244 icmpkp->inSrcQuenchs.value.ui32 = ipst->ips_icmp_mib.icmpInSrcQuenchs;
14245 icmpkp->inRedirects.value.ui32 = ipst->ips_icmp_mib.icmpInRedirects;
14246 icmpkp->inEchos.value.ui32 = ipst->ips_icmp_mib.icmpInEchos;
14247 icmpkp->inEchoReps.value.ui32 = ipst->ips_icmp_mib.icmpInEchoReps;
14248 icmpkp->inTimestamps.value.ui32 = ipst->ips_icmp_mib.icmpInTimestamps;
14249 icmpkp->inTimestampReps.value.ui32 =
14250 ipst->ips_icmp_mib.icmpInTimestampReps;
14251 icmpkp->inAddrMasks.value.ui32 = ipst->ips_icmp_mib.icmpInAddrMasks;
14252 icmpkp->inAddrMaskReps.value.ui32 =
14253 ipst->ips_icmp_mib.icmpInAddrMaskReps;
14254 icmpkp->outMsgs.value.ui32 = ipst->ips_icmp_mib.icmpOutMsgs;
14255 icmpkp->outErrors.value.ui32 = ipst->ips_icmp_mib.icmpOutErrors;
14256 icmpkp->outDestUnreachs.value.ui32 =
14257 ipst->ips_icmp_mib.icmpOutDestUnreachs;
14258 icmpkp->outTimeExcds.value.ui32 = ipst->ips_icmp_mib.icmpOutTimeExcds;
14259 icmpkp->outParmProbs.value.ui32 = ipst->ips_icmp_mib.icmpOutParmProbs;
14260 icmpkp->outSrcQuenchs.value.ui32 =
14261 ipst->ips_icmp_mib.icmpOutSrcQuenchs;
14262 icmpkp->outRedirects.value.ui32 = ipst->ips_icmp_mib.icmpOutRedirects;
14263 icmpkp->outEchos.value.ui32 = ipst->ips_icmp_mib.icmpOutEchos;
14264 icmpkp->outEchoReps.value.ui32 = ipst->ips_icmp_mib.icmpOutEchoReps;
14265 icmpkp->outTimestamps.value.ui32 =
14266 ipst->ips_icmp_mib.icmpOutTimestamps;
14267 icmpkp->outTimestampReps.value.ui32 =
14268 ipst->ips_icmp_mib.icmpOutTimestampReps;
14269 icmpkp->outAddrMasks.value.ui32 =
14270 ipst->ips_icmp_mib.icmpOutAddrMasks;
14271 icmpkp->outAddrMaskReps.value.ui32 =
14272 ipst->ips_icmp_mib.icmpOutAddrMaskReps;
14273 icmpkp->inCksumErrs.value.ui32 = ipst->ips_icmp_mib.icmpInCksumErrs;
14274 icmpkp->inUnknowns.value.ui32 = ipst->ips_icmp_mib.icmpInUnknowns;
14275 icmpkp->inFragNeeded.value.ui32 = ipst->ips_icmp_mib.icmpInFragNeeded;
14276 icmpkp->outFragNeeded.value.ui32 =
14277 ipst->ips_icmp_mib.icmpOutFragNeeded;
14278 icmpkp->outDrops.value.ui32 = ipst->ips_icmp_mib.icmpOutDrops;
14279 icmpkp->inOverflows.value.ui32 = ipst->ips_icmp_mib.icmpInOverflows;
14280 icmpkp->inBadRedirects.value.ui32 =
14281 ipst->ips_icmp_mib.icmpInBadRedirects;
14282
14283 netstack_rele(ns);
14284 return (0);
14285 }
14286
14287 /*
14288 * This is the fanout function for raw socket opened for SCTP. Note
14289 * that it is called after SCTP checks that there is no socket which
14290 * wants a packet. Then before SCTP handles this out of the blue packet,
14291 * this function is called to see if there is any raw socket for SCTP.
14292 * If there is and it is bound to the correct address, the packet will
14293 * be sent to that socket. Note that only one raw socket can be bound to
14294 * a port. This is assured in ipcl_sctp_hash_insert();
14295 */
14296 void
14297 ip_fanout_sctp_raw(mblk_t *mp, ipha_t *ipha, ip6_t *ip6h, uint32_t ports,
14298 ip_recv_attr_t *ira)
14299 {
14300 conn_t *connp;
14301 queue_t *rq;
14302 boolean_t secure;
14303 ill_t *ill = ira->ira_ill;
14304 ip_stack_t *ipst = ill->ill_ipst;
14305 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
14306 sctp_stack_t *sctps = ipst->ips_netstack->netstack_sctp;
14307 iaflags_t iraflags = ira->ira_flags;
14308 ill_t *rill = ira->ira_rill;
14309
14310 secure = iraflags & IRAF_IPSEC_SECURE;
14311
14312 connp = ipcl_classify_raw(mp, IPPROTO_SCTP, ports, ipha, ip6h,
14313 ira, ipst);
14314 if (connp == NULL) {
14315 /*
14316 * Although raw sctp is not summed, OOB chunks must be.
14317 * Drop the packet here if the sctp checksum failed.
14318 */
14319 if (iraflags & IRAF_SCTP_CSUM_ERR) {
14320 SCTPS_BUMP_MIB(sctps, sctpChecksumError);
14321 freemsg(mp);
14322 return;
14323 }
14324 ira->ira_ill = ira->ira_rill = NULL;
14325 sctp_ootb_input(mp, ira, ipst);
14326 ira->ira_ill = ill;
14327 ira->ira_rill = rill;
14328 return;
14329 }
14330 rq = connp->conn_rq;
14331 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld : !canputnext(rq)) {
14332 CONN_DEC_REF(connp);
14333 BUMP_MIB(ill->ill_ip_mib, rawipIfStatsInOverflows);
14334 freemsg(mp);
14335 return;
14336 }
14337 if (((iraflags & IRAF_IS_IPV4) ?
14338 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
14339 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
14340 secure) {
14341 mp = ipsec_check_inbound_policy(mp, connp, ipha,
14342 ip6h, ira);
14343 if (mp == NULL) {
14344 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
14345 /* Note that mp is NULL */
14346 ip_drop_input("ipIfStatsInDiscards", mp, ill);
14347 CONN_DEC_REF(connp);
14348 return;
14349 }
14350 }
14351
14352 if (iraflags & IRAF_ICMP_ERROR) {
14353 (connp->conn_recvicmp)(connp, mp, NULL, ira);
14354 } else {
14355 ill_t *rill = ira->ira_rill;
14356
14357 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
14358 /* This is the SOCK_RAW, IPPROTO_SCTP case. */
14359 ira->ira_ill = ira->ira_rill = NULL;
14360 (connp->conn_recv)(connp, mp, NULL, ira);
14361 ira->ira_ill = ill;
14362 ira->ira_rill = rill;
14363 }
14364 CONN_DEC_REF(connp);
14365 }
14366
14367 /*
14368 * Free a packet that has the link-layer dl_unitdata_req_t or fast-path
14369 * header before the ip payload.
14370 */
14371 static void
14372 ip_xmit_flowctl_drop(ill_t *ill, mblk_t *mp, boolean_t is_fp_mp, int fp_mp_len)
14373 {
14374 int len = (mp->b_wptr - mp->b_rptr);
14375 mblk_t *ip_mp;
14376
14377 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14378 if (is_fp_mp || len != fp_mp_len) {
14379 if (len > fp_mp_len) {
14380 /*
14381 * fastpath header and ip header in the first mblk
14382 */
14383 mp->b_rptr += fp_mp_len;
14384 } else {
14385 /*
14386 * ip_xmit_attach_llhdr had to prepend an mblk to
14387 * attach the fastpath header before ip header.
14388 */
14389 ip_mp = mp->b_cont;
14390 freeb(mp);
14391 mp = ip_mp;
14392 mp->b_rptr += (fp_mp_len - len);
14393 }
14394 } else {
14395 ip_mp = mp->b_cont;
14396 freeb(mp);
14397 mp = ip_mp;
14398 }
14399 ip_drop_output("ipIfStatsOutDiscards - flow ctl", mp, ill);
14400 freemsg(mp);
14401 }
14402
14403 /*
14404 * Normal post fragmentation function.
14405 *
14406 * Send a packet using the passed in nce. This handles both IPv4 and IPv6
14407 * using the same state machine.
14408 *
14409 * We return an error on failure. In particular we return EWOULDBLOCK
14410 * when the driver flow controls. In that case this ensures that ip_wsrv runs
14411 * (currently by canputnext failure resulting in backenabling from GLD.)
14412 * This allows the callers of conn_ip_output() to use EWOULDBLOCK as an
14413 * indication that they can flow control until ip_wsrv() tells then to restart.
14414 *
14415 * If the nce passed by caller is incomplete, this function
14416 * queues the packet and if necessary, sends ARP request and bails.
14417 * If the Neighbor Cache passed is fully resolved, we simply prepend
14418 * the link-layer header to the packet, do ipsec hw acceleration
14419 * work if necessary, and send the packet out on the wire.
14420 */
14421 /* ARGSUSED6 */
14422 int
14423 ip_xmit(mblk_t *mp, nce_t *nce, iaflags_t ixaflags, uint_t pkt_len,
14424 uint32_t xmit_hint, zoneid_t szone, zoneid_t nolzid, uintptr_t *ixacookie)
14425 {
14426 queue_t *wq;
14427 ill_t *ill = nce->nce_ill;
14428 ip_stack_t *ipst = ill->ill_ipst;
14429 uint64_t delta;
14430 boolean_t isv6 = ill->ill_isv6;
14431 boolean_t fp_mp;
14432 ncec_t *ncec = nce->nce_common;
14433 int64_t now = LBOLT_FASTPATH64;
14434 boolean_t is_probe;
14435
14436 DTRACE_PROBE1(ip__xmit, nce_t *, nce);
14437
14438 ASSERT(mp != NULL);
14439 ASSERT(mp->b_datap->db_type == M_DATA);
14440 ASSERT(pkt_len == msgdsize(mp));
14441
14442 /*
14443 * If we have already been here and are coming back after ARP/ND.
14444 * the IXAF_NO_TRACE flag is set. We skip FW_HOOKS, DTRACE and ipobs
14445 * in that case since they have seen the packet when it came here
14446 * the first time.
14447 */
14448 if (ixaflags & IXAF_NO_TRACE)
14449 goto sendit;
14450
14451 if (ixaflags & IXAF_IS_IPV4) {
14452 ipha_t *ipha = (ipha_t *)mp->b_rptr;
14453
14454 ASSERT(!isv6);
14455 ASSERT(pkt_len == ntohs(((ipha_t *)mp->b_rptr)->ipha_length));
14456 if (HOOKS4_INTERESTED_PHYSICAL_OUT(ipst) &&
14457 !(ixaflags & IXAF_NO_PFHOOK)) {
14458 int error;
14459
14460 FW_HOOKS(ipst->ips_ip4_physical_out_event,
14461 ipst->ips_ipv4firewall_physical_out,
14462 NULL, ill, ipha, mp, mp, 0, ipst, error);
14463 DTRACE_PROBE1(ip4__physical__out__end,
14464 mblk_t *, mp);
14465 if (mp == NULL)
14466 return (error);
14467
14468 /* The length could have changed */
14469 pkt_len = msgdsize(mp);
14470 }
14471 if (ipst->ips_ip4_observe.he_interested) {
14472 /*
14473 * Note that for TX the zoneid is the sending
14474 * zone, whether or not MLP is in play.
14475 * Since the szone argument is the IP zoneid (i.e.,
14476 * zero for exclusive-IP zones) and ipobs wants
14477 * the system zoneid, we map it here.
14478 */
14479 szone = IP_REAL_ZONEID(szone, ipst);
14480
14481 /*
14482 * On the outbound path the destination zone will be
14483 * unknown as we're sending this packet out on the
14484 * wire.
14485 */
14486 ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone, ALL_ZONES,
14487 ill, ipst);
14488 }
14489 DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
14490 void_ip_t *, ipha, __dtrace_ipsr_ill_t *, ill,
14491 ipha_t *, ipha, ip6_t *, NULL, int, 0);
14492 } else {
14493 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
14494
14495 ASSERT(isv6);
14496 ASSERT(pkt_len ==
14497 ntohs(((ip6_t *)mp->b_rptr)->ip6_plen) + IPV6_HDR_LEN);
14498 if (HOOKS6_INTERESTED_PHYSICAL_OUT(ipst) &&
14499 !(ixaflags & IXAF_NO_PFHOOK)) {
14500 int error;
14501
14502 FW_HOOKS6(ipst->ips_ip6_physical_out_event,
14503 ipst->ips_ipv6firewall_physical_out,
14504 NULL, ill, ip6h, mp, mp, 0, ipst, error);
14505 DTRACE_PROBE1(ip6__physical__out__end,
14506 mblk_t *, mp);
14507 if (mp == NULL)
14508 return (error);
14509
14510 /* The length could have changed */
14511 pkt_len = msgdsize(mp);
14512 }
14513 if (ipst->ips_ip6_observe.he_interested) {
14514 /* See above */
14515 szone = IP_REAL_ZONEID(szone, ipst);
14516
14517 ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone, ALL_ZONES,
14518 ill, ipst);
14519 }
14520 DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
14521 void_ip_t *, ip6h, __dtrace_ipsr_ill_t *, ill,
14522 ipha_t *, NULL, ip6_t *, ip6h, int, 0);
14523 }
14524
14525 sendit:
14526 /*
14527 * We check the state without a lock because the state can never
14528 * move "backwards" to initial or incomplete.
14529 */
14530 switch (ncec->ncec_state) {
14531 case ND_REACHABLE:
14532 case ND_STALE:
14533 case ND_DELAY:
14534 case ND_PROBE:
14535 mp = ip_xmit_attach_llhdr(mp, nce);
14536 if (mp == NULL) {
14537 /*
14538 * ip_xmit_attach_llhdr has increased
14539 * ipIfStatsOutDiscards and called ip_drop_output()
14540 */
14541 return (ENOBUFS);
14542 }
14543 /*
14544 * check if nce_fastpath completed and we tagged on a
14545 * copy of nce_fp_mp in ip_xmit_attach_llhdr().
14546 */
14547 fp_mp = (mp->b_datap->db_type == M_DATA);
14548
14549 if (fp_mp &&
14550 (ill->ill_capabilities & ILL_CAPAB_DLD_DIRECT)) {
14551 ill_dld_direct_t *idd;
14552
14553 idd = &ill->ill_dld_capab->idc_direct;
14554 /*
14555 * Send the packet directly to DLD, where it
14556 * may be queued depending on the availability
14557 * of transmit resources at the media layer.
14558 * Return value should be taken into
14559 * account and flow control the TCP.
14560 */
14561 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
14562 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
14563 pkt_len);
14564
14565 if (ixaflags & IXAF_NO_DEV_FLOW_CTL) {
14566 (void) idd->idd_tx_df(idd->idd_tx_dh, mp,
14567 (uintptr_t)xmit_hint, IP_DROP_ON_NO_DESC);
14568 } else {
14569 uintptr_t cookie;
14570
14571 if ((cookie = idd->idd_tx_df(idd->idd_tx_dh,
14572 mp, (uintptr_t)xmit_hint, 0)) != 0) {
14573 if (ixacookie != NULL)
14574 *ixacookie = cookie;
14575 return (EWOULDBLOCK);
14576 }
14577 }
14578 } else {
14579 wq = ill->ill_wq;
14580
14581 if (!(ixaflags & IXAF_NO_DEV_FLOW_CTL) &&
14582 !canputnext(wq)) {
14583 if (ixacookie != NULL)
14584 *ixacookie = 0;
14585 ip_xmit_flowctl_drop(ill, mp, fp_mp,
14586 nce->nce_fp_mp != NULL ?
14587 MBLKL(nce->nce_fp_mp) : 0);
14588 return (EWOULDBLOCK);
14589 }
14590 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
14591 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
14592 pkt_len);
14593 putnext(wq, mp);
14594 }
14595
14596 /*
14597 * The rest of this function implements Neighbor Unreachability
14598 * detection. Determine if the ncec is eligible for NUD.
14599 */
14600 if (ncec->ncec_flags & NCE_F_NONUD)
14601 return (0);
14602
14603 ASSERT(ncec->ncec_state != ND_INCOMPLETE);
14604
14605 /*
14606 * Check for upper layer advice
14607 */
14608 if (ixaflags & IXAF_REACH_CONF) {
14609 timeout_id_t tid;
14610
14611 /*
14612 * It should be o.k. to check the state without
14613 * a lock here, at most we lose an advice.
14614 */
14615 ncec->ncec_last = TICK_TO_MSEC(now);
14616 if (ncec->ncec_state != ND_REACHABLE) {
14617 mutex_enter(&ncec->ncec_lock);
14618 ncec->ncec_state = ND_REACHABLE;
14619 tid = ncec->ncec_timeout_id;
14620 ncec->ncec_timeout_id = 0;
14621 mutex_exit(&ncec->ncec_lock);
14622 (void) untimeout(tid);
14623 if (ip_debug > 2) {
14624 /* ip1dbg */
14625 pr_addr_dbg("ip_xmit: state"
14626 " for %s changed to"
14627 " REACHABLE\n", AF_INET6,
14628 &ncec->ncec_addr);
14629 }
14630 }
14631 return (0);
14632 }
14633
14634 delta = TICK_TO_MSEC(now) - ncec->ncec_last;
14635 ip1dbg(("ip_xmit: delta = %" PRId64
14636 " ill_reachable_time = %d \n", delta,
14637 ill->ill_reachable_time));
14638 if (delta > (uint64_t)ill->ill_reachable_time) {
14639 mutex_enter(&ncec->ncec_lock);
14640 switch (ncec->ncec_state) {
14641 case ND_REACHABLE:
14642 ASSERT((ncec->ncec_flags & NCE_F_NONUD) == 0);
14643 /* FALLTHROUGH */
14644 case ND_STALE:
14645 /*
14646 * ND_REACHABLE is identical to
14647 * ND_STALE in this specific case. If
14648 * reachable time has expired for this
14649 * neighbor (delta is greater than
14650 * reachable time), conceptually, the
14651 * neighbor cache is no longer in
14652 * REACHABLE state, but already in
14653 * STALE state. So the correct
14654 * transition here is to ND_DELAY.
14655 */
14656 ncec->ncec_state = ND_DELAY;
14657 mutex_exit(&ncec->ncec_lock);
14658 nce_restart_timer(ncec,
14659 ipst->ips_delay_first_probe_time);
14660 if (ip_debug > 3) {
14661 /* ip2dbg */
14662 pr_addr_dbg("ip_xmit: state"
14663 " for %s changed to"
14664 " DELAY\n", AF_INET6,
14665 &ncec->ncec_addr);
14666 }
14667 break;
14668 case ND_DELAY:
14669 case ND_PROBE:
14670 mutex_exit(&ncec->ncec_lock);
14671 /* Timers have already started */
14672 break;
14673 case ND_UNREACHABLE:
14674 /*
14675 * nce_timer has detected that this ncec
14676 * is unreachable and initiated deleting
14677 * this ncec.
14678 * This is a harmless race where we found the
14679 * ncec before it was deleted and have
14680 * just sent out a packet using this
14681 * unreachable ncec.
14682 */
14683 mutex_exit(&ncec->ncec_lock);
14684 break;
14685 default:
14686 ASSERT(0);
14687 mutex_exit(&ncec->ncec_lock);
14688 }
14689 }
14690 return (0);
14691
14692 case ND_INCOMPLETE:
14693 /*
14694 * the state could have changed since we didn't hold the lock.
14695 * Re-verify state under lock.
14696 */
14697 is_probe = ipmp_packet_is_probe(mp, nce->nce_ill);
14698 mutex_enter(&ncec->ncec_lock);
14699 if (NCE_ISREACHABLE(ncec)) {
14700 mutex_exit(&ncec->ncec_lock);
14701 goto sendit;
14702 }
14703 /* queue the packet */
14704 nce_queue_mp(ncec, mp, is_probe);
14705 mutex_exit(&ncec->ncec_lock);
14706 DTRACE_PROBE2(ip__xmit__incomplete,
14707 (ncec_t *), ncec, (mblk_t *), mp);
14708 return (0);
14709
14710 case ND_INITIAL:
14711 /*
14712 * State could have changed since we didn't hold the lock, so
14713 * re-verify state.
14714 */
14715 is_probe = ipmp_packet_is_probe(mp, nce->nce_ill);
14716 mutex_enter(&ncec->ncec_lock);
14717 if (NCE_ISREACHABLE(ncec)) {
14718 mutex_exit(&ncec->ncec_lock);
14719 goto sendit;
14720 }
14721 nce_queue_mp(ncec, mp, is_probe);
14722 if (ncec->ncec_state == ND_INITIAL) {
14723 ncec->ncec_state = ND_INCOMPLETE;
14724 mutex_exit(&ncec->ncec_lock);
14725 /*
14726 * figure out the source we want to use
14727 * and resolve it.
14728 */
14729 ip_ndp_resolve(ncec);
14730 } else {
14731 mutex_exit(&ncec->ncec_lock);
14732 }
14733 return (0);
14734
14735 case ND_UNREACHABLE:
14736 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14737 ip_drop_output("ipIfStatsOutDiscards - ND_UNREACHABLE",
14738 mp, ill);
14739 freemsg(mp);
14740 return (0);
14741
14742 default:
14743 ASSERT(0);
14744 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14745 ip_drop_output("ipIfStatsOutDiscards - ND_other",
14746 mp, ill);
14747 freemsg(mp);
14748 return (ENETUNREACH);
14749 }
14750 }
14751
14752 /*
14753 * Return B_TRUE if the buffers differ in length or content.
14754 * This is used for comparing extension header buffers.
14755 * Note that an extension header would be declared different
14756 * even if all that changed was the next header value in that header i.e.
14757 * what really changed is the next extension header.
14758 */
14759 boolean_t
14760 ip_cmpbuf(const void *abuf, uint_t alen, boolean_t b_valid, const void *bbuf,
14761 uint_t blen)
14762 {
14763 if (!b_valid)
14764 blen = 0;
14765
14766 if (alen != blen)
14767 return (B_TRUE);
14768 if (alen == 0)
14769 return (B_FALSE); /* Both zero length */
14770 return (bcmp(abuf, bbuf, alen));
14771 }
14772
14773 /*
14774 * Preallocate memory for ip_savebuf(). Returns B_TRUE if ok.
14775 * Return B_FALSE if memory allocation fails - don't change any state!
14776 */
14777 boolean_t
14778 ip_allocbuf(void **dstp, uint_t *dstlenp, boolean_t src_valid,
14779 const void *src, uint_t srclen)
14780 {
14781 void *dst;
14782
14783 if (!src_valid)
14784 srclen = 0;
14785
14786 ASSERT(*dstlenp == 0);
14787 if (src != NULL && srclen != 0) {
14788 dst = mi_alloc(srclen, BPRI_MED);
14789 if (dst == NULL)
14790 return (B_FALSE);
14791 } else {
14792 dst = NULL;
14793 }
14794 if (*dstp != NULL)
14795 mi_free(*dstp);
14796 *dstp = dst;
14797 *dstlenp = dst == NULL ? 0 : srclen;
14798 return (B_TRUE);
14799 }
14800
14801 /*
14802 * Replace what is in *dst, *dstlen with the source.
14803 * Assumes ip_allocbuf has already been called.
14804 */
14805 void
14806 ip_savebuf(void **dstp, uint_t *dstlenp, boolean_t src_valid,
14807 const void *src, uint_t srclen)
14808 {
14809 if (!src_valid)
14810 srclen = 0;
14811
14812 ASSERT(*dstlenp == srclen);
14813 if (src != NULL && srclen != 0)
14814 bcopy(src, *dstp, srclen);
14815 }
14816
14817 /*
14818 * Free the storage pointed to by the members of an ip_pkt_t.
14819 */
14820 void
14821 ip_pkt_free(ip_pkt_t *ipp)
14822 {
14823 uint_t fields = ipp->ipp_fields;
14824
14825 if (fields & IPPF_HOPOPTS) {
14826 kmem_free(ipp->ipp_hopopts, ipp->ipp_hopoptslen);
14827 ipp->ipp_hopopts = NULL;
14828 ipp->ipp_hopoptslen = 0;
14829 }
14830 if (fields & IPPF_RTHDRDSTOPTS) {
14831 kmem_free(ipp->ipp_rthdrdstopts, ipp->ipp_rthdrdstoptslen);
14832 ipp->ipp_rthdrdstopts = NULL;
14833 ipp->ipp_rthdrdstoptslen = 0;
14834 }
14835 if (fields & IPPF_DSTOPTS) {
14836 kmem_free(ipp->ipp_dstopts, ipp->ipp_dstoptslen);
14837 ipp->ipp_dstopts = NULL;
14838 ipp->ipp_dstoptslen = 0;
14839 }
14840 if (fields & IPPF_RTHDR) {
14841 kmem_free(ipp->ipp_rthdr, ipp->ipp_rthdrlen);
14842 ipp->ipp_rthdr = NULL;
14843 ipp->ipp_rthdrlen = 0;
14844 }
14845 if (fields & IPPF_IPV4_OPTIONS) {
14846 kmem_free(ipp->ipp_ipv4_options, ipp->ipp_ipv4_options_len);
14847 ipp->ipp_ipv4_options = NULL;
14848 ipp->ipp_ipv4_options_len = 0;
14849 }
14850 if (fields & IPPF_LABEL_V4) {
14851 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
14852 ipp->ipp_label_v4 = NULL;
14853 ipp->ipp_label_len_v4 = 0;
14854 }
14855 if (fields & IPPF_LABEL_V6) {
14856 kmem_free(ipp->ipp_label_v6, ipp->ipp_label_len_v6);
14857 ipp->ipp_label_v6 = NULL;
14858 ipp->ipp_label_len_v6 = 0;
14859 }
14860 ipp->ipp_fields &= ~(IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14861 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6);
14862 }
14863
14864 /*
14865 * Copy from src to dst and allocate as needed.
14866 * Returns zero or ENOMEM.
14867 *
14868 * The caller must initialize dst to zero.
14869 */
14870 int
14871 ip_pkt_copy(ip_pkt_t *src, ip_pkt_t *dst, int kmflag)
14872 {
14873 uint_t fields = src->ipp_fields;
14874
14875 /* Start with fields that don't require memory allocation */
14876 dst->ipp_fields = fields &
14877 ~(IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14878 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6);
14879
14880 dst->ipp_addr = src->ipp_addr;
14881 dst->ipp_unicast_hops = src->ipp_unicast_hops;
14882 dst->ipp_hoplimit = src->ipp_hoplimit;
14883 dst->ipp_tclass = src->ipp_tclass;
14884 dst->ipp_type_of_service = src->ipp_type_of_service;
14885
14886 if (!(fields & (IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14887 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6)))
14888 return (0);
14889
14890 if (fields & IPPF_HOPOPTS) {
14891 dst->ipp_hopopts = kmem_alloc(src->ipp_hopoptslen, kmflag);
14892 if (dst->ipp_hopopts == NULL) {
14893 ip_pkt_free(dst);
14894 return (ENOMEM);
14895 }
14896 dst->ipp_fields |= IPPF_HOPOPTS;
14897 bcopy(src->ipp_hopopts, dst->ipp_hopopts,
14898 src->ipp_hopoptslen);
14899 dst->ipp_hopoptslen = src->ipp_hopoptslen;
14900 }
14901 if (fields & IPPF_RTHDRDSTOPTS) {
14902 dst->ipp_rthdrdstopts = kmem_alloc(src->ipp_rthdrdstoptslen,
14903 kmflag);
14904 if (dst->ipp_rthdrdstopts == NULL) {
14905 ip_pkt_free(dst);
14906 return (ENOMEM);
14907 }
14908 dst->ipp_fields |= IPPF_RTHDRDSTOPTS;
14909 bcopy(src->ipp_rthdrdstopts, dst->ipp_rthdrdstopts,
14910 src->ipp_rthdrdstoptslen);
14911 dst->ipp_rthdrdstoptslen = src->ipp_rthdrdstoptslen;
14912 }
14913 if (fields & IPPF_DSTOPTS) {
14914 dst->ipp_dstopts = kmem_alloc(src->ipp_dstoptslen, kmflag);
14915 if (dst->ipp_dstopts == NULL) {
14916 ip_pkt_free(dst);
14917 return (ENOMEM);
14918 }
14919 dst->ipp_fields |= IPPF_DSTOPTS;
14920 bcopy(src->ipp_dstopts, dst->ipp_dstopts,
14921 src->ipp_dstoptslen);
14922 dst->ipp_dstoptslen = src->ipp_dstoptslen;
14923 }
14924 if (fields & IPPF_RTHDR) {
14925 dst->ipp_rthdr = kmem_alloc(src->ipp_rthdrlen, kmflag);
14926 if (dst->ipp_rthdr == NULL) {
14927 ip_pkt_free(dst);
14928 return (ENOMEM);
14929 }
14930 dst->ipp_fields |= IPPF_RTHDR;
14931 bcopy(src->ipp_rthdr, dst->ipp_rthdr,
14932 src->ipp_rthdrlen);
14933 dst->ipp_rthdrlen = src->ipp_rthdrlen;
14934 }
14935 if (fields & IPPF_IPV4_OPTIONS) {
14936 dst->ipp_ipv4_options = kmem_alloc(src->ipp_ipv4_options_len,
14937 kmflag);
14938 if (dst->ipp_ipv4_options == NULL) {
14939 ip_pkt_free(dst);
14940 return (ENOMEM);
14941 }
14942 dst->ipp_fields |= IPPF_IPV4_OPTIONS;
14943 bcopy(src->ipp_ipv4_options, dst->ipp_ipv4_options,
14944 src->ipp_ipv4_options_len);
14945 dst->ipp_ipv4_options_len = src->ipp_ipv4_options_len;
14946 }
14947 if (fields & IPPF_LABEL_V4) {
14948 dst->ipp_label_v4 = kmem_alloc(src->ipp_label_len_v4, kmflag);
14949 if (dst->ipp_label_v4 == NULL) {
14950 ip_pkt_free(dst);
14951 return (ENOMEM);
14952 }
14953 dst->ipp_fields |= IPPF_LABEL_V4;
14954 bcopy(src->ipp_label_v4, dst->ipp_label_v4,
14955 src->ipp_label_len_v4);
14956 dst->ipp_label_len_v4 = src->ipp_label_len_v4;
14957 }
14958 if (fields & IPPF_LABEL_V6) {
14959 dst->ipp_label_v6 = kmem_alloc(src->ipp_label_len_v6, kmflag);
14960 if (dst->ipp_label_v6 == NULL) {
14961 ip_pkt_free(dst);
14962 return (ENOMEM);
14963 }
14964 dst->ipp_fields |= IPPF_LABEL_V6;
14965 bcopy(src->ipp_label_v6, dst->ipp_label_v6,
14966 src->ipp_label_len_v6);
14967 dst->ipp_label_len_v6 = src->ipp_label_len_v6;
14968 }
14969 if (fields & IPPF_FRAGHDR) {
14970 dst->ipp_fraghdr = kmem_alloc(src->ipp_fraghdrlen, kmflag);
14971 if (dst->ipp_fraghdr == NULL) {
14972 ip_pkt_free(dst);
14973 return (ENOMEM);
14974 }
14975 dst->ipp_fields |= IPPF_FRAGHDR;
14976 bcopy(src->ipp_fraghdr, dst->ipp_fraghdr,
14977 src->ipp_fraghdrlen);
14978 dst->ipp_fraghdrlen = src->ipp_fraghdrlen;
14979 }
14980 return (0);
14981 }
14982
14983 /*
14984 * Returns INADDR_ANY if no source route
14985 */
14986 ipaddr_t
14987 ip_pkt_source_route_v4(const ip_pkt_t *ipp)
14988 {
14989 ipaddr_t nexthop = INADDR_ANY;
14990 ipoptp_t opts;
14991 uchar_t *opt;
14992 uint8_t optval;
14993 uint8_t optlen;
14994 uint32_t totallen;
14995
14996 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
14997 return (INADDR_ANY);
14998
14999 totallen = ipp->ipp_ipv4_options_len;
15000 if (totallen & 0x3)
15001 return (INADDR_ANY);
15002
15003 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
15004 optval != IPOPT_EOL;
15005 optval = ipoptp_next(&opts)) {
15006 opt = opts.ipoptp_cur;
15007 switch (optval) {
15008 uint8_t off;
15009 case IPOPT_SSRR:
15010 case IPOPT_LSRR:
15011 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
15012 break;
15013 }
15014 optlen = opts.ipoptp_len;
15015 off = opt[IPOPT_OFFSET];
15016 off--;
15017 if (optlen < IP_ADDR_LEN ||
15018 off > optlen - IP_ADDR_LEN) {
15019 /* End of source route */
15020 break;
15021 }
15022 bcopy((char *)opt + off, &nexthop, IP_ADDR_LEN);
15023 if (nexthop == htonl(INADDR_LOOPBACK)) {
15024 /* Ignore */
15025 nexthop = INADDR_ANY;
15026 break;
15027 }
15028 break;
15029 }
15030 }
15031 return (nexthop);
15032 }
15033
15034 /*
15035 * Reverse a source route.
15036 */
15037 void
15038 ip_pkt_source_route_reverse_v4(ip_pkt_t *ipp)
15039 {
15040 ipaddr_t tmp;
15041 ipoptp_t opts;
15042 uchar_t *opt;
15043 uint8_t optval;
15044 uint32_t totallen;
15045
15046 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
15047 return;
15048
15049 totallen = ipp->ipp_ipv4_options_len;
15050 if (totallen & 0x3)
15051 return;
15052
15053 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
15054 optval != IPOPT_EOL;
15055 optval = ipoptp_next(&opts)) {
15056 uint8_t off1, off2;
15057
15058 opt = opts.ipoptp_cur;
15059 switch (optval) {
15060 case IPOPT_SSRR:
15061 case IPOPT_LSRR:
15062 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
15063 break;
15064 }
15065 off1 = IPOPT_MINOFF_SR - 1;
15066 off2 = opt[IPOPT_OFFSET] - IP_ADDR_LEN - 1;
15067 while (off2 > off1) {
15068 bcopy(opt + off2, &tmp, IP_ADDR_LEN);
15069 bcopy(opt + off1, opt + off2, IP_ADDR_LEN);
15070 bcopy(&tmp, opt + off2, IP_ADDR_LEN);
15071 off2 -= IP_ADDR_LEN;
15072 off1 += IP_ADDR_LEN;
15073 }
15074 opt[IPOPT_OFFSET] = IPOPT_MINOFF_SR;
15075 break;
15076 }
15077 }
15078 }
15079
15080 /*
15081 * Returns NULL if no routing header
15082 */
15083 in6_addr_t *
15084 ip_pkt_source_route_v6(const ip_pkt_t *ipp)
15085 {
15086 in6_addr_t *nexthop = NULL;
15087 ip6_rthdr0_t *rthdr;
15088
15089 if (!(ipp->ipp_fields & IPPF_RTHDR))
15090 return (NULL);
15091
15092 rthdr = (ip6_rthdr0_t *)ipp->ipp_rthdr;
15093 if (rthdr->ip6r0_segleft == 0)
15094 return (NULL);
15095
15096 nexthop = (in6_addr_t *)((char *)rthdr + sizeof (*rthdr));
15097 return (nexthop);
15098 }
15099
15100 zoneid_t
15101 ip_get_zoneid_v4(ipaddr_t addr, mblk_t *mp, ip_recv_attr_t *ira,
15102 zoneid_t lookup_zoneid)
15103 {
15104 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
15105 ire_t *ire;
15106 int ire_flags = MATCH_IRE_TYPE;
15107 zoneid_t zoneid = ALL_ZONES;
15108
15109 if (is_system_labeled() && !tsol_can_accept_raw(mp, ira, B_FALSE))
15110 return (ALL_ZONES);
15111
15112 if (lookup_zoneid != ALL_ZONES)
15113 ire_flags |= MATCH_IRE_ZONEONLY;
15114 ire = ire_ftable_lookup_v4(addr, NULL, NULL, IRE_LOCAL | IRE_LOOPBACK,
15115 NULL, lookup_zoneid, NULL, ire_flags, 0, ipst, NULL);
15116 if (ire != NULL) {
15117 zoneid = IP_REAL_ZONEID(ire->ire_zoneid, ipst);
15118 ire_refrele(ire);
15119 }
15120 return (zoneid);
15121 }
15122
15123 zoneid_t
15124 ip_get_zoneid_v6(in6_addr_t *addr, mblk_t *mp, const ill_t *ill,
15125 ip_recv_attr_t *ira, zoneid_t lookup_zoneid)
15126 {
15127 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
15128 ire_t *ire;
15129 int ire_flags = MATCH_IRE_TYPE;
15130 zoneid_t zoneid = ALL_ZONES;
15131
15132 if (is_system_labeled() && !tsol_can_accept_raw(mp, ira, B_FALSE))
15133 return (ALL_ZONES);
15134
15135 if (IN6_IS_ADDR_LINKLOCAL(addr))
15136 ire_flags |= MATCH_IRE_ILL;
15137
15138 if (lookup_zoneid != ALL_ZONES)
15139 ire_flags |= MATCH_IRE_ZONEONLY;
15140 ire = ire_ftable_lookup_v6(addr, NULL, NULL, IRE_LOCAL | IRE_LOOPBACK,
15141 ill, lookup_zoneid, NULL, ire_flags, 0, ipst, NULL);
15142 if (ire != NULL) {
15143 zoneid = IP_REAL_ZONEID(ire->ire_zoneid, ipst);
15144 ire_refrele(ire);
15145 }
15146 return (zoneid);
15147 }
15148
15149 /*
15150 * IP obserability hook support functions.
15151 */
15152 static void
15153 ipobs_init(ip_stack_t *ipst)
15154 {
15155 netid_t id;
15156
15157 id = net_getnetidbynetstackid(ipst->ips_netstack->netstack_stackid);
15158
15159 ipst->ips_ip4_observe_pr = net_protocol_lookup(id, NHF_INET);
15160 VERIFY(ipst->ips_ip4_observe_pr != NULL);
15161
15162 ipst->ips_ip6_observe_pr = net_protocol_lookup(id, NHF_INET6);
15163 VERIFY(ipst->ips_ip6_observe_pr != NULL);
15164 }
15165
15166 static void
15167 ipobs_fini(ip_stack_t *ipst)
15168 {
15169
15170 VERIFY(net_protocol_release(ipst->ips_ip4_observe_pr) == 0);
15171 VERIFY(net_protocol_release(ipst->ips_ip6_observe_pr) == 0);
15172 }
15173
15174 /*
15175 * hook_pkt_observe_t is composed in network byte order so that the
15176 * entire mblk_t chain handed into hook_run can be used as-is.
15177 * The caveat is that use of the fields, such as the zone fields,
15178 * requires conversion into host byte order first.
15179 */
15180 void
15181 ipobs_hook(mblk_t *mp, int htype, zoneid_t zsrc, zoneid_t zdst,
15182 const ill_t *ill, ip_stack_t *ipst)
15183 {
15184 hook_pkt_observe_t *hdr;
15185 uint64_t grifindex;
15186 mblk_t *imp;
15187
15188 imp = allocb(sizeof (*hdr), BPRI_HI);
15189 if (imp == NULL)
15190 return;
15191
15192 hdr = (hook_pkt_observe_t *)imp->b_rptr;
15193 /*
15194 * b_wptr is set to make the apparent size of the data in the mblk_t
15195 * to exclude the pointers at the end of hook_pkt_observer_t.
15196 */
15197 imp->b_wptr = imp->b_rptr + sizeof (dl_ipnetinfo_t);
15198 imp->b_cont = mp;
15199
15200 ASSERT(DB_TYPE(mp) == M_DATA);
15201
15202 if (IS_UNDER_IPMP(ill))
15203 grifindex = ipmp_ill_get_ipmp_ifindex(ill);
15204 else
15205 grifindex = 0;
15206
15207 hdr->hpo_version = 1;
15208 hdr->hpo_htype = htons(htype);
15209 hdr->hpo_pktlen = htonl((ulong_t)msgdsize(mp));
15210 hdr->hpo_ifindex = htonl(ill->ill_phyint->phyint_ifindex);
15211 hdr->hpo_grifindex = htonl(grifindex);
15212 hdr->hpo_zsrc = htonl(zsrc);
15213 hdr->hpo_zdst = htonl(zdst);
15214 hdr->hpo_pkt = imp;
15215 hdr->hpo_ctx = ipst->ips_netstack;
15216
15217 if (ill->ill_isv6) {
15218 hdr->hpo_family = AF_INET6;
15219 (void) hook_run(ipst->ips_ipv6_net_data->netd_hooks,
15220 ipst->ips_ipv6observing, (hook_data_t)hdr);
15221 } else {
15222 hdr->hpo_family = AF_INET;
15223 (void) hook_run(ipst->ips_ipv4_net_data->netd_hooks,
15224 ipst->ips_ipv4observing, (hook_data_t)hdr);
15225 }
15226
15227 imp->b_cont = NULL;
15228 freemsg(imp);
15229 }
15230
15231 /*
15232 * Utility routine that checks if `v4srcp' is a valid address on underlying
15233 * interface `ill'. If `ipifp' is non-NULL, it's set to a held ipif
15234 * associated with `v4srcp' on success. NOTE: if this is not called from
15235 * inside the IPSQ (ill_g_lock is not held), `ill' may be removed from the
15236 * group during or after this lookup.
15237 */
15238 boolean_t
15239 ipif_lookup_testaddr_v4(ill_t *ill, const in_addr_t *v4srcp, ipif_t **ipifp)
15240 {
15241 ipif_t *ipif;
15242
15243 ipif = ipif_lookup_addr_exact(*v4srcp, ill, ill->ill_ipst);
15244 if (ipif != NULL) {
15245 if (ipifp != NULL)
15246 *ipifp = ipif;
15247 else
15248 ipif_refrele(ipif);
15249 return (B_TRUE);
15250 }
15251
15252 ip1dbg(("ipif_lookup_testaddr_v4: cannot find ipif for src %x\n",
15253 *v4srcp));
15254 return (B_FALSE);
15255 }
15256
15257 /*
15258 * Transport protocol call back function for CPU state change.
15259 */
15260 /* ARGSUSED */
15261 static int
15262 ip_tp_cpu_update(cpu_setup_t what, int id, void *arg)
15263 {
15264 processorid_t cpu_seqid;
15265 netstack_handle_t nh;
15266 netstack_t *ns;
15267
15268 ASSERT(MUTEX_HELD(&cpu_lock));
15269
15270 switch (what) {
15271 case CPU_CONFIG:
15272 case CPU_ON:
15273 case CPU_INIT:
15274 case CPU_CPUPART_IN:
15275 cpu_seqid = cpu[id]->cpu_seqid;
15276 netstack_next_init(&nh);
15277 while ((ns = netstack_next(&nh)) != NULL) {
15278 tcp_stack_cpu_add(ns->netstack_tcp, cpu_seqid);
15279 sctp_stack_cpu_add(ns->netstack_sctp, cpu_seqid);
15280 udp_stack_cpu_add(ns->netstack_udp, cpu_seqid);
15281 netstack_rele(ns);
15282 }
15283 netstack_next_fini(&nh);
15284 break;
15285 case CPU_UNCONFIG:
15286 case CPU_OFF:
15287 case CPU_CPUPART_OUT:
15288 /*
15289 * Nothing to do. We don't remove the per CPU stats from
15290 * the IP stack even when the CPU goes offline.
15291 */
15292 break;
15293 default:
15294 break;
15295 }
15296 return (0);
15297 }