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 */
26
27 #include <sys/types.h>
28 #include <sys/stream.h>
29 #include <sys/dlpi.h>
30 #include <sys/stropts.h>
31 #include <sys/sysmacros.h>
32 #include <sys/strsubr.h>
33 #include <sys/strlog.h>
34 #include <sys/strsun.h>
35 #include <sys/zone.h>
36 #define _SUN_TPI_VERSION 2
37 #include <sys/tihdr.h>
38 #include <sys/xti_inet.h>
39 #include <sys/ddi.h>
40 #include <sys/suntpi.h>
41 #include <sys/cmn_err.h>
42 #include <sys/debug.h>
43 #include <sys/kobj.h>
44 #include <sys/modctl.h>
45 #include <sys/atomic.h>
46 #include <sys/policy.h>
47 #include <sys/priv.h>
48 #include <sys/taskq.h>
49
50 #include <sys/systm.h>
51 #include <sys/param.h>
52 #include <sys/kmem.h>
53 #include <sys/sdt.h>
54 #include <sys/socket.h>
55 #include <sys/vtrace.h>
56 #include <sys/isa_defs.h>
57 #include <sys/mac.h>
58 #include <net/if.h>
59 #include <net/if_arp.h>
60 #include <net/route.h>
61 #include <sys/sockio.h>
62 #include <netinet/in.h>
63 #include <net/if_dl.h>
64
65 #include <inet/common.h>
66 #include <inet/mi.h>
67 #include <inet/mib2.h>
68 #include <inet/nd.h>
69 #include <inet/arp.h>
70 #include <inet/snmpcom.h>
71 #include <inet/optcom.h>
72 #include <inet/kstatcom.h>
73
74 #include <netinet/igmp_var.h>
75 #include <netinet/ip6.h>
76 #include <netinet/icmp6.h>
77 #include <netinet/sctp.h>
78
79 #include <inet/ip.h>
80 #include <inet/ip_impl.h>
81 #include <inet/ip6.h>
82 #include <inet/ip6_asp.h>
83 #include <inet/tcp.h>
84 #include <inet/tcp_impl.h>
85 #include <inet/ip_multi.h>
86 #include <inet/ip_if.h>
87 #include <inet/ip_ire.h>
88 #include <inet/ip_ftable.h>
89 #include <inet/ip_rts.h>
90 #include <inet/ip_ndp.h>
91 #include <inet/ip_listutils.h>
92 #include <netinet/igmp.h>
93 #include <netinet/ip_mroute.h>
94 #include <inet/ipp_common.h>
95
96 #include <net/pfkeyv2.h>
97 #include <inet/sadb.h>
98 #include <inet/ipsec_impl.h>
99 #include <inet/iptun/iptun_impl.h>
100 #include <inet/ipdrop.h>
101 #include <inet/ip_netinfo.h>
102 #include <inet/ilb_ip.h>
103
104 #include <sys/ethernet.h>
105 #include <net/if_types.h>
106 #include <sys/cpuvar.h>
107
108 #include <ipp/ipp.h>
109 #include <ipp/ipp_impl.h>
110 #include <ipp/ipgpc/ipgpc.h>
111
112 #include <sys/pattr.h>
113 #include <inet/ipclassifier.h>
114 #include <inet/sctp_ip.h>
115 #include <inet/sctp/sctp_impl.h>
116 #include <inet/udp_impl.h>
117 #include <inet/rawip_impl.h>
118 #include <inet/rts_impl.h>
119
120 #include <sys/tsol/label.h>
121 #include <sys/tsol/tnet.h>
122
123 #include <sys/squeue_impl.h>
124 #include <inet/ip_arp.h>
125
126 #include <sys/clock_impl.h> /* For LBOLT_FASTPATH{,64} */
127
128 /*
129 * Values for squeue switch:
130 * IP_SQUEUE_ENTER_NODRAIN: SQ_NODRAIN
131 * IP_SQUEUE_ENTER: SQ_PROCESS
132 * IP_SQUEUE_FILL: SQ_FILL
133 */
134 int ip_squeue_enter = IP_SQUEUE_ENTER; /* Setable in /etc/system */
135
136 int ip_squeue_flag;
137
138 /*
139 * Setable in /etc/system
140 */
141 int ip_poll_normal_ms = 100;
142 int ip_poll_normal_ticks = 0;
143 int ip_modclose_ackwait_ms = 3000;
144
145 /*
146 * It would be nice to have these present only in DEBUG systems, but the
147 * current design of the global symbol checking logic requires them to be
148 * unconditionally present.
149 */
150 uint_t ip_thread_data; /* TSD key for debug support */
151 krwlock_t ip_thread_rwlock;
152 list_t ip_thread_list;
153
154 /*
155 * Structure to represent a linked list of msgblks. Used by ip_snmp_ functions.
156 */
157
158 struct listptr_s {
159 mblk_t *lp_head; /* pointer to the head of the list */
160 mblk_t *lp_tail; /* pointer to the tail of the list */
161 };
162
163 typedef struct listptr_s listptr_t;
164
165 /*
166 * This is used by ip_snmp_get_mib2_ip_route_media and
167 * ip_snmp_get_mib2_ip6_route_media to carry the lists of return data.
168 */
169 typedef struct iproutedata_s {
170 uint_t ird_idx;
171 uint_t ird_flags; /* see below */
172 listptr_t ird_route; /* ipRouteEntryTable */
173 listptr_t ird_netmedia; /* ipNetToMediaEntryTable */
174 listptr_t ird_attrs; /* ipRouteAttributeTable */
175 } iproutedata_t;
176
177 /* Include ire_testhidden and IRE_IF_CLONE routes */
178 #define IRD_REPORT_ALL 0x01
179
180 /*
181 * Cluster specific hooks. These should be NULL when booted as a non-cluster
182 */
183
184 /*
185 * Hook functions to enable cluster networking
186 * On non-clustered systems these vectors must always be NULL.
187 *
188 * Hook function to Check ip specified ip address is a shared ip address
189 * in the cluster
190 *
191 */
192 int (*cl_inet_isclusterwide)(netstackid_t stack_id, uint8_t protocol,
193 sa_family_t addr_family, uint8_t *laddrp, void *args) = NULL;
194
195 /*
196 * Hook function to generate cluster wide ip fragment identifier
197 */
198 uint32_t (*cl_inet_ipident)(netstackid_t stack_id, uint8_t protocol,
199 sa_family_t addr_family, uint8_t *laddrp, uint8_t *faddrp,
200 void *args) = NULL;
201
202 /*
203 * Hook function to generate cluster wide SPI.
204 */
205 void (*cl_inet_getspi)(netstackid_t, uint8_t, uint8_t *, size_t,
206 void *) = NULL;
207
208 /*
209 * Hook function to verify if the SPI is already utlized.
210 */
211
212 int (*cl_inet_checkspi)(netstackid_t, uint8_t, uint32_t, void *) = NULL;
213
214 /*
215 * Hook function to delete the SPI from the cluster wide repository.
216 */
217
218 void (*cl_inet_deletespi)(netstackid_t, uint8_t, uint32_t, void *) = NULL;
219
220 /*
221 * Hook function to inform the cluster when packet received on an IDLE SA
222 */
223
224 void (*cl_inet_idlesa)(netstackid_t, uint8_t, uint32_t, sa_family_t,
225 in6_addr_t, in6_addr_t, void *) = NULL;
226
227 /*
228 * Synchronization notes:
229 *
230 * IP is a fully D_MP STREAMS module/driver. Thus it does not depend on any
231 * MT level protection given by STREAMS. IP uses a combination of its own
232 * internal serialization mechanism and standard Solaris locking techniques.
233 * The internal serialization is per phyint. This is used to serialize
234 * plumbing operations, IPMP operations, most set ioctls, etc.
235 *
236 * Plumbing is a long sequence of operations involving message
237 * exchanges between IP, ARP and device drivers. Many set ioctls are typically
238 * involved in plumbing operations. A natural model is to serialize these
239 * ioctls one per ill. For example plumbing of hme0 and qfe0 can go on in
240 * parallel without any interference. But various set ioctls on hme0 are best
241 * serialized, along with IPMP operations and processing of DLPI control
242 * messages received from drivers on a per phyint basis. This serialization is
243 * provided by the ipsq_t and primitives operating on this. Details can
244 * be found in ip_if.c above the core primitives operating on ipsq_t.
245 *
246 * Lookups of an ipif or ill by a thread return a refheld ipif / ill.
247 * Simiarly lookup of an ire by a thread also returns a refheld ire.
248 * In addition ipif's and ill's referenced by the ire are also indirectly
249 * refheld. Thus no ipif or ill can vanish as long as an ipif is refheld
250 * directly or indirectly. For example an SIOCSLIFADDR ioctl that changes the
251 * address of an ipif has to go through the ipsq_t. This ensures that only
252 * one such exclusive operation proceeds at any time on the ipif. It then
253 * waits for all refcnts
254 * associated with this ipif to come down to zero. The address is changed
255 * only after the ipif has been quiesced. Then the ipif is brought up again.
256 * More details are described above the comment in ip_sioctl_flags.
257 *
258 * Packet processing is based mostly on IREs and are fully multi-threaded
259 * using standard Solaris MT techniques.
260 *
261 * There are explicit locks in IP to handle:
262 * - The ip_g_head list maintained by mi_open_link() and friends.
263 *
264 * - The reassembly data structures (one lock per hash bucket)
265 *
266 * - conn_lock is meant to protect conn_t fields. The fields actually
267 * protected by conn_lock are documented in the conn_t definition.
268 *
269 * - ire_lock to protect some of the fields of the ire, IRE tables
270 * (one lock per hash bucket). Refer to ip_ire.c for details.
271 *
272 * - ndp_g_lock and ncec_lock for protecting NCEs.
273 *
274 * - ill_lock protects fields of the ill and ipif. Details in ip.h
275 *
276 * - ill_g_lock: This is a global reader/writer lock. Protects the following
277 * * The AVL tree based global multi list of all ills.
278 * * The linked list of all ipifs of an ill
279 * * The <ipsq-xop> mapping
280 * * <ill-phyint> association
281 * Insertion/deletion of an ill in the system, insertion/deletion of an ipif
282 * into an ill, changing the <ipsq-xop> mapping of an ill, changing the
283 * <ill-phyint> assoc of an ill will all have to hold the ill_g_lock as
284 * writer for the actual duration of the insertion/deletion/change.
285 *
286 * - ill_lock: This is a per ill mutex.
287 * It protects some members of the ill_t struct; see ip.h for details.
288 * It also protects the <ill-phyint> assoc.
289 * It also protects the list of ipifs hanging off the ill.
290 *
291 * - ipsq_lock: This is a per ipsq_t mutex lock.
292 * This protects some members of the ipsq_t struct; see ip.h for details.
293 * It also protects the <ipsq-ipxop> mapping
294 *
295 * - ipx_lock: This is a per ipxop_t mutex lock.
296 * This protects some members of the ipxop_t struct; see ip.h for details.
297 *
298 * - phyint_lock: This is a per phyint mutex lock. Protects just the
299 * phyint_flags
300 *
301 * - ip_addr_avail_lock: This is used to ensure the uniqueness of IP addresses.
302 * This lock is held in ipif_up_done and the ipif is marked IPIF_UP and the
303 * uniqueness check also done atomically.
304 *
305 * - ill_g_usesrc_lock: This readers/writer lock protects the usesrc
306 * group list linked by ill_usesrc_grp_next. It also protects the
307 * ill_usesrc_ifindex field. It is taken as a writer when a member of the
308 * group is being added or deleted. This lock is taken as a reader when
309 * walking the list/group(eg: to get the number of members in a usesrc group).
310 * Note, it is only necessary to take this lock if the ill_usesrc_grp_next
311 * field is changing state i.e from NULL to non-NULL or vice-versa. For
312 * example, it is not necessary to take this lock in the initial portion
313 * of ip_sioctl_slifusesrc or at all in ip_sioctl_flags since these
314 * operations are executed exclusively and that ensures that the "usesrc
315 * group state" cannot change. The "usesrc group state" change can happen
316 * only in the latter part of ip_sioctl_slifusesrc and in ill_delete.
317 *
318 * Changing <ill-phyint>, <ipsq-xop> assocications:
319 *
320 * To change the <ill-phyint> association, the ill_g_lock must be held
321 * as writer, and the ill_locks of both the v4 and v6 instance of the ill
322 * must be held.
323 *
324 * To change the <ipsq-xop> association, the ill_g_lock must be held as
325 * writer, the ipsq_lock must be held, and one must be writer on the ipsq.
326 * This is only done when ills are added or removed from IPMP groups.
327 *
328 * To add or delete an ipif from the list of ipifs hanging off the ill,
329 * ill_g_lock (writer) and ill_lock must be held and the thread must be
330 * a writer on the associated ipsq.
331 *
332 * To add or delete an ill to the system, the ill_g_lock must be held as
333 * writer and the thread must be a writer on the associated ipsq.
334 *
335 * To add or delete an ilm to an ill, the ill_lock must be held and the thread
336 * must be a writer on the associated ipsq.
337 *
338 * Lock hierarchy
339 *
340 * Some lock hierarchy scenarios are listed below.
341 *
342 * ill_g_lock -> conn_lock -> ill_lock -> ipsq_lock -> ipx_lock
343 * ill_g_lock -> ill_lock(s) -> phyint_lock
344 * ill_g_lock -> ndp_g_lock -> ill_lock -> ncec_lock
345 * ill_g_lock -> ip_addr_avail_lock
346 * conn_lock -> irb_lock -> ill_lock -> ire_lock
347 * ill_g_lock -> ip_g_nd_lock
348 * ill_g_lock -> ips_ipmp_lock -> ill_lock -> nce_lock
349 * ill_g_lock -> ndp_g_lock -> ill_lock -> ncec_lock -> nce_lock
350 * arl_lock -> ill_lock
351 * ips_ire_dep_lock -> irb_lock
352 *
353 * When more than 1 ill lock is needed to be held, all ill lock addresses
354 * are sorted on address and locked starting from highest addressed lock
355 * downward.
356 *
357 * Multicast scenarios
358 * ips_ill_g_lock -> ill_mcast_lock
359 * conn_ilg_lock -> ips_ill_g_lock -> ill_lock
360 * ill_mcast_serializer -> ill_mcast_lock -> ips_ipmp_lock -> ill_lock
361 * ill_mcast_serializer -> ill_mcast_lock -> connf_lock -> conn_lock
362 * ill_mcast_serializer -> ill_mcast_lock -> conn_ilg_lock
363 * ill_mcast_serializer -> ill_mcast_lock -> ips_igmp_timer_lock
364 *
365 * IPsec scenarios
366 *
367 * ipsa_lock -> ill_g_lock -> ill_lock
368 * ill_g_usesrc_lock -> ill_g_lock -> ill_lock
369 *
370 * Trusted Solaris scenarios
371 *
372 * igsa_lock -> gcgrp_rwlock -> gcgrp_lock
373 * igsa_lock -> gcdb_lock
374 * gcgrp_rwlock -> ire_lock
375 * gcgrp_rwlock -> gcdb_lock
376 *
377 * squeue(sq_lock), flow related (ft_lock, fe_lock) locking
378 *
379 * cpu_lock --> ill_lock --> sqset_lock --> sq_lock
380 * sq_lock -> conn_lock -> QLOCK(q)
381 * ill_lock -> ft_lock -> fe_lock
382 *
383 * Routing/forwarding table locking notes:
384 *
385 * Lock acquisition order: Radix tree lock, irb_lock.
386 * Requirements:
387 * i. Walker must not hold any locks during the walker callback.
388 * ii Walker must not see a truncated tree during the walk because of any node
389 * deletion.
390 * iii Existing code assumes ire_bucket is valid if it is non-null and is used
391 * in many places in the code to walk the irb list. Thus even if all the
392 * ires in a bucket have been deleted, we still can't free the radix node
393 * until the ires have actually been inactive'd (freed).
394 *
395 * Tree traversal - Need to hold the global tree lock in read mode.
396 * Before dropping the global tree lock, need to either increment the ire_refcnt
397 * to ensure that the radix node can't be deleted.
398 *
399 * Tree add - Need to hold the global tree lock in write mode to add a
400 * radix node. To prevent the node from being deleted, increment the
401 * irb_refcnt, after the node is added to the tree. The ire itself is
402 * added later while holding the irb_lock, but not the tree lock.
403 *
404 * Tree delete - Need to hold the global tree lock and irb_lock in write mode.
405 * All associated ires must be inactive (i.e. freed), and irb_refcnt
406 * must be zero.
407 *
408 * Walker - Increment irb_refcnt before calling the walker callback. Hold the
409 * global tree lock (read mode) for traversal.
410 *
411 * IRE dependencies - In some cases we hold ips_ire_dep_lock across ire_refrele
412 * hence we will acquire irb_lock while holding ips_ire_dep_lock.
413 *
414 * IPsec notes :
415 *
416 * IP interacts with the IPsec code (AH/ESP) by storing IPsec attributes
417 * in the ip_xmit_attr_t ip_recv_attr_t. For outbound datagrams, the
418 * ip_xmit_attr_t has the
419 * information used by the IPsec code for applying the right level of
420 * protection. The information initialized by IP in the ip_xmit_attr_t
421 * is determined by the per-socket policy or global policy in the system.
422 * For inbound datagrams, the ip_recv_attr_t
423 * starts out with nothing in it. It gets filled
424 * with the right information if it goes through the AH/ESP code, which
425 * happens if the incoming packet is secure. The information initialized
426 * by AH/ESP, is later used by IP (during fanouts to ULP) to see whether
427 * the policy requirements needed by per-socket policy or global policy
428 * is met or not.
429 *
430 * For fully connected sockets i.e dst, src [addr, port] is known,
431 * conn_policy_cached is set indicating that policy has been cached.
432 * conn_in_enforce_policy may or may not be set depending on whether
433 * there is a global policy match or per-socket policy match.
434 * Policy inheriting happpens in ip_policy_set once the destination is known.
435 * Once the right policy is set on the conn_t, policy cannot change for
436 * this socket. This makes life simpler for TCP (UDP ?) where
437 * re-transmissions go out with the same policy. For symmetry, policy
438 * is cached for fully connected UDP sockets also. Thus if policy is cached,
439 * it also implies that policy is latched i.e policy cannot change
440 * on these sockets. As we have the right policy on the conn, we don't
441 * have to lookup global policy for every outbound and inbound datagram
442 * and thus serving as an optimization. Note that a global policy change
443 * does not affect fully connected sockets if they have policy. If fully
444 * connected sockets did not have any policy associated with it, global
445 * policy change may affect them.
446 *
447 * IP Flow control notes:
448 * ---------------------
449 * Non-TCP streams are flow controlled by IP. The way this is accomplished
450 * differs when ILL_CAPAB_DLD_DIRECT is enabled for that IP instance. When
451 * ILL_DIRECT_CAPABLE(ill) is TRUE, IP can do direct function calls into
452 * GLDv3. Otherwise packets are sent down to lower layers using STREAMS
453 * functions.
454 *
455 * Per Tx ring udp flow control:
456 * This is applicable only when ILL_CAPAB_DLD_DIRECT capability is set in
457 * the ill (i.e. ILL_DIRECT_CAPABLE(ill) is true).
458 *
459 * The underlying link can expose multiple Tx rings to the GLDv3 mac layer.
460 * To achieve best performance, outgoing traffic need to be fanned out among
461 * these Tx ring. mac_tx() is called (via str_mdata_fastpath_put()) to send
462 * traffic out of the NIC and it takes a fanout hint. UDP connections pass
463 * the address of connp as fanout hint to mac_tx(). Under flow controlled
464 * condition, mac_tx() returns a non-NULL cookie (ip_mac_tx_cookie_t). This
465 * cookie points to a specific Tx ring that is blocked. The cookie is used to
466 * hash into an idl_tx_list[] entry in idl_tx_list[] array. Each idl_tx_list_t
467 * point to drain_lists (idl_t's). These drain list will store the blocked UDP
468 * connp's. The drain list is not a single list but a configurable number of
469 * lists.
470 *
471 * The diagram below shows idl_tx_list_t's and their drain_lists. ip_stack_t
472 * has an array of idl_tx_list_t. The size of the array is TX_FANOUT_SIZE
473 * which is equal to 128. This array in turn contains a pointer to idl_t[],
474 * the ip drain list. The idl_t[] array size is MIN(max_ncpus, 8). The drain
475 * list will point to the list of connp's that are flow controlled.
476 *
477 * --------------- ------- ------- -------
478 * |->|drain_list[0]|-->|connp|-->|connp|-->|connp|-->
479 * | --------------- ------- ------- -------
480 * | --------------- ------- ------- -------
481 * |->|drain_list[1]|-->|connp|-->|connp|-->|connp|-->
482 * ---------------- | --------------- ------- ------- -------
483 * |idl_tx_list[0]|->| --------------- ------- ------- -------
484 * ---------------- |->|drain_list[2]|-->|connp|-->|connp|-->|connp|-->
485 * | --------------- ------- ------- -------
486 * . . . . .
487 * | --------------- ------- ------- -------
488 * |->|drain_list[n]|-->|connp|-->|connp|-->|connp|-->
489 * --------------- ------- ------- -------
490 * --------------- ------- ------- -------
491 * |->|drain_list[0]|-->|connp|-->|connp|-->|connp|-->
492 * | --------------- ------- ------- -------
493 * | --------------- ------- ------- -------
494 * ---------------- |->|drain_list[1]|-->|connp|-->|connp|-->|connp|-->
495 * |idl_tx_list[1]|->| --------------- ------- ------- -------
496 * ---------------- | . . . .
497 * | --------------- ------- ------- -------
498 * |->|drain_list[n]|-->|connp|-->|connp|-->|connp|-->
499 * --------------- ------- ------- -------
500 * .....
501 * ----------------
502 * |idl_tx_list[n]|-> ...
503 * ----------------
504 *
505 * When mac_tx() returns a cookie, the cookie is hashed into an index into
506 * ips_idl_tx_list[], and conn_drain_insert() is called with the idl_tx_list
507 * to insert the conn onto. conn_drain_insert() asserts flow control for the
508 * sockets via su_txq_full() (non-STREAMS) or QFULL on conn_wq (STREAMS).
509 * Further, conn_blocked is set to indicate that the conn is blocked.
510 *
511 * GLDv3 calls ill_flow_enable() when flow control is relieved. The cookie
512 * passed in the call to ill_flow_enable() identifies the blocked Tx ring and
513 * is again hashed to locate the appropriate idl_tx_list, which is then
514 * drained via conn_walk_drain(). conn_walk_drain() goes through each conn in
515 * the drain list and calls conn_drain_remove() to clear flow control (via
516 * calling su_txq_full() or clearing QFULL), and remove the conn from the
517 * drain list.
518 *
519 * Note that the drain list is not a single list but a (configurable) array of
520 * lists (8 elements by default). Synchronization between drain insertion and
521 * flow control wakeup is handled by using idl_txl->txl_lock, and only
522 * conn_drain_insert() and conn_drain_remove() manipulate the drain list.
523 *
524 * Flow control via STREAMS is used when ILL_DIRECT_CAPABLE() returns FALSE.
525 * On the send side, if the packet cannot be sent down to the driver by IP
526 * (canput() fails), ip_xmit() drops the packet and returns EWOULDBLOCK to the
527 * caller, who may then invoke ixa_check_drain_insert() to insert the conn on
528 * the 0'th drain list. When ip_wsrv() runs on the ill_wq because flow
529 * control has been relieved, the blocked conns in the 0'th drain list are
530 * drained as in the non-STREAMS case.
531 *
532 * In both the STREAMS and non-STREAMS cases, the sockfs upcall to set QFULL
533 * is done when the conn is inserted into the drain list (conn_drain_insert())
534 * and cleared when the conn is removed from the it (conn_drain_remove()).
535 *
536 * IPQOS notes:
537 *
538 * IPQoS Policies are applied to packets using IPPF (IP Policy framework)
539 * and IPQoS modules. IPPF includes hooks in IP at different control points
540 * (callout positions) which direct packets to IPQoS modules for policy
541 * processing. Policies, if present, are global.
542 *
543 * The callout positions are located in the following paths:
544 * o local_in (packets destined for this host)
545 * o local_out (packets orginating from this host )
546 * o fwd_in (packets forwarded by this m/c - inbound)
547 * o fwd_out (packets forwarded by this m/c - outbound)
548 * Hooks at these callout points can be enabled/disabled using the ndd variable
549 * ip_policy_mask (a bit mask with the 4 LSB indicating the callout positions).
550 * By default all the callout positions are enabled.
551 *
552 * Outbound (local_out)
553 * Hooks are placed in ire_send_wire_v4 and ire_send_wire_v6.
554 *
555 * Inbound (local_in)
556 * Hooks are placed in ip_fanout_v4 and ip_fanout_v6.
557 *
558 * Forwarding (in and out)
559 * Hooks are placed in ire_recv_forward_v4/v6.
560 *
561 * IP Policy Framework processing (IPPF processing)
562 * Policy processing for a packet is initiated by ip_process, which ascertains
563 * that the classifier (ipgpc) is loaded and configured, failing which the
564 * packet resumes normal processing in IP. If the clasifier is present, the
565 * packet is acted upon by one or more IPQoS modules (action instances), per
566 * filters configured in ipgpc and resumes normal IP processing thereafter.
567 * An action instance can drop a packet in course of its processing.
568 *
569 * Zones notes:
570 *
571 * The partitioning rules for networking are as follows:
572 * 1) Packets coming from a zone must have a source address belonging to that
573 * zone.
574 * 2) Packets coming from a zone can only be sent on a physical interface on
575 * which the zone has an IP address.
576 * 3) Between two zones on the same machine, packet delivery is only allowed if
577 * there's a matching route for the destination and zone in the forwarding
578 * table.
579 * 4) The TCP and UDP port spaces are per-zone; that is, two processes in
580 * different zones can bind to the same port with the wildcard address
581 * (INADDR_ANY).
582 *
583 * The granularity of interface partitioning is at the logical interface level.
584 * Therefore, every zone has its own IP addresses, and incoming packets can be
585 * attributed to a zone unambiguously. A logical interface is placed into a zone
586 * using the SIOCSLIFZONE ioctl; this sets the ipif_zoneid field in the ipif_t
587 * structure. Rule (1) is implemented by modifying the source address selection
588 * algorithm so that the list of eligible addresses is filtered based on the
589 * sending process zone.
590 *
591 * The Internet Routing Entries (IREs) are either exclusive to a zone or shared
592 * across all zones, depending on their type. Here is the break-up:
593 *
594 * IRE type Shared/exclusive
595 * -------- ----------------
596 * IRE_BROADCAST Exclusive
597 * IRE_DEFAULT (default routes) Shared (*)
598 * IRE_LOCAL Exclusive (x)
599 * IRE_LOOPBACK Exclusive
600 * IRE_PREFIX (net routes) Shared (*)
601 * IRE_IF_NORESOLVER (interface routes) Exclusive
602 * IRE_IF_RESOLVER (interface routes) Exclusive
603 * IRE_IF_CLONE (interface routes) Exclusive
604 * IRE_HOST (host routes) Shared (*)
605 *
606 * (*) A zone can only use a default or off-subnet route if the gateway is
607 * directly reachable from the zone, that is, if the gateway's address matches
608 * one of the zone's logical interfaces.
609 *
610 * (x) IRE_LOCAL are handled a bit differently.
611 * When ip_restrict_interzone_loopback is set (the default),
612 * ire_route_recursive restricts loopback using an IRE_LOCAL
613 * between zone to the case when L2 would have conceptually looped the packet
614 * back, i.e. the loopback which is required since neither Ethernet drivers
615 * nor Ethernet hardware loops them back. This is the case when the normal
616 * routes (ignoring IREs with different zoneids) would send out the packet on
617 * the same ill as the ill with which is IRE_LOCAL is associated.
618 *
619 * Multiple zones can share a common broadcast address; typically all zones
620 * share the 255.255.255.255 address. Incoming as well as locally originated
621 * broadcast packets must be dispatched to all the zones on the broadcast
622 * network. For directed broadcasts (e.g. 10.16.72.255) this is not trivial
623 * since some zones may not be on the 10.16.72/24 network. To handle this, each
624 * zone has its own set of IRE_BROADCAST entries; then, broadcast packets are
625 * sent to every zone that has an IRE_BROADCAST entry for the destination
626 * address on the input ill, see ip_input_broadcast().
627 *
628 * Applications in different zones can join the same multicast group address.
629 * The same logic applies for multicast as for broadcast. ip_input_multicast
630 * dispatches packets to all zones that have members on the physical interface.
631 */
632
633 /*
634 * Squeue Fanout flags:
635 * 0: No fanout.
636 * 1: Fanout across all squeues
637 */
638 boolean_t ip_squeue_fanout = 0;
639
640 /*
641 * Maximum dups allowed per packet.
642 */
643 uint_t ip_max_frag_dups = 10;
644
645 static int ip_open(queue_t *q, dev_t *devp, int flag, int sflag,
646 cred_t *credp, boolean_t isv6);
647 static mblk_t *ip_xmit_attach_llhdr(mblk_t *, nce_t *);
648
649 static boolean_t icmp_inbound_verify_v4(mblk_t *, icmph_t *, ip_recv_attr_t *);
650 static void icmp_inbound_too_big_v4(icmph_t *, ip_recv_attr_t *);
651 static void icmp_inbound_error_fanout_v4(mblk_t *, icmph_t *,
652 ip_recv_attr_t *);
653 static void icmp_options_update(ipha_t *);
654 static void icmp_param_problem(mblk_t *, uint8_t, ip_recv_attr_t *);
655 static void icmp_pkt(mblk_t *, void *, size_t, ip_recv_attr_t *);
656 static mblk_t *icmp_pkt_err_ok(mblk_t *, ip_recv_attr_t *);
657 static void icmp_redirect_v4(mblk_t *mp, ipha_t *, icmph_t *,
658 ip_recv_attr_t *);
659 static void icmp_send_redirect(mblk_t *, ipaddr_t, ip_recv_attr_t *);
660 static void icmp_send_reply_v4(mblk_t *, ipha_t *, icmph_t *,
661 ip_recv_attr_t *);
662
663 mblk_t *ip_dlpi_alloc(size_t, t_uscalar_t);
664 char *ip_dot_addr(ipaddr_t, char *);
665 mblk_t *ip_carve_mp(mblk_t **, ssize_t);
666 int ip_close(queue_t *, int);
667 static char *ip_dot_saddr(uchar_t *, char *);
668 static void ip_lrput(queue_t *, mblk_t *);
669 ipaddr_t ip_net_mask(ipaddr_t);
670 char *ip_nv_lookup(nv_t *, int);
671 void ip_rput(queue_t *, mblk_t *);
672 static void ip_rput_dlpi_writer(ipsq_t *dummy_sq, queue_t *q, mblk_t *mp,
673 void *dummy_arg);
674 int ip_snmp_get(queue_t *, mblk_t *, int, boolean_t);
675 static mblk_t *ip_snmp_get_mib2_ip(queue_t *, mblk_t *,
676 mib2_ipIfStatsEntry_t *, ip_stack_t *, boolean_t);
677 static mblk_t *ip_snmp_get_mib2_ip_traffic_stats(queue_t *, mblk_t *,
678 ip_stack_t *, boolean_t);
679 static mblk_t *ip_snmp_get_mib2_ip6(queue_t *, mblk_t *, ip_stack_t *,
680 boolean_t);
681 static mblk_t *ip_snmp_get_mib2_icmp(queue_t *, mblk_t *, ip_stack_t *ipst);
682 static mblk_t *ip_snmp_get_mib2_icmp6(queue_t *, mblk_t *, ip_stack_t *ipst);
683 static mblk_t *ip_snmp_get_mib2_igmp(queue_t *, mblk_t *, ip_stack_t *ipst);
684 static mblk_t *ip_snmp_get_mib2_multi(queue_t *, mblk_t *, ip_stack_t *ipst);
685 static mblk_t *ip_snmp_get_mib2_ip_addr(queue_t *, mblk_t *,
686 ip_stack_t *ipst, boolean_t);
687 static mblk_t *ip_snmp_get_mib2_ip6_addr(queue_t *, mblk_t *,
688 ip_stack_t *ipst, boolean_t);
689 static mblk_t *ip_snmp_get_mib2_ip_group_src(queue_t *, mblk_t *,
690 ip_stack_t *ipst);
691 static mblk_t *ip_snmp_get_mib2_ip6_group_src(queue_t *, mblk_t *,
692 ip_stack_t *ipst);
693 static mblk_t *ip_snmp_get_mib2_ip_group_mem(queue_t *, mblk_t *,
694 ip_stack_t *ipst);
695 static mblk_t *ip_snmp_get_mib2_ip6_group_mem(queue_t *, mblk_t *,
696 ip_stack_t *ipst);
697 static mblk_t *ip_snmp_get_mib2_virt_multi(queue_t *, mblk_t *,
698 ip_stack_t *ipst);
699 static mblk_t *ip_snmp_get_mib2_multi_rtable(queue_t *, mblk_t *,
700 ip_stack_t *ipst);
701 static mblk_t *ip_snmp_get_mib2_ip_route_media(queue_t *, mblk_t *, int,
702 ip_stack_t *ipst);
703 static mblk_t *ip_snmp_get_mib2_ip6_route_media(queue_t *, mblk_t *, int,
704 ip_stack_t *ipst);
705 static void ip_snmp_get2_v4(ire_t *, iproutedata_t *);
706 static void ip_snmp_get2_v6_route(ire_t *, iproutedata_t *);
707 static int ip_snmp_get2_v4_media(ncec_t *, iproutedata_t *);
708 static int ip_snmp_get2_v6_media(ncec_t *, iproutedata_t *);
709 int ip_snmp_set(queue_t *, int, int, uchar_t *, int);
710
711 static mblk_t *ip_fragment_copyhdr(uchar_t *, int, int, ip_stack_t *,
712 mblk_t *);
713
714 static void conn_drain_init(ip_stack_t *);
715 static void conn_drain_fini(ip_stack_t *);
716 static void conn_drain(conn_t *connp, boolean_t closing);
717
718 static void conn_walk_drain(ip_stack_t *, idl_tx_list_t *);
719 static void conn_walk_sctp(pfv_t, void *, zoneid_t, netstack_t *);
720
721 static void *ip_stack_init(netstackid_t stackid, netstack_t *ns);
722 static void ip_stack_shutdown(netstackid_t stackid, void *arg);
723 static void ip_stack_fini(netstackid_t stackid, void *arg);
724
725 static int ip_multirt_apply_membership(int (*fn)(conn_t *, boolean_t,
726 const in6_addr_t *, ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *),
727 ire_t *, conn_t *, boolean_t, const in6_addr_t *, mcast_record_t,
728 const in6_addr_t *);
729
730 static int ip_squeue_switch(int);
731
732 static void *ip_kstat_init(netstackid_t, ip_stack_t *);
733 static void ip_kstat_fini(netstackid_t, kstat_t *);
734 static int ip_kstat_update(kstat_t *kp, int rw);
735 static void *icmp_kstat_init(netstackid_t);
736 static void icmp_kstat_fini(netstackid_t, kstat_t *);
737 static int icmp_kstat_update(kstat_t *kp, int rw);
738 static void *ip_kstat2_init(netstackid_t, ip_stat_t *);
739 static void ip_kstat2_fini(netstackid_t, kstat_t *);
740
741 static void ipobs_init(ip_stack_t *);
742 static void ipobs_fini(ip_stack_t *);
743
744 static int ip_tp_cpu_update(cpu_setup_t, int, void *);
745
746 ipaddr_t ip_g_all_ones = IP_HOST_MASK;
747
748 static long ip_rput_pullups;
749 int dohwcksum = 1; /* use h/w cksum if supported by the hardware */
750
751 vmem_t *ip_minor_arena_sa; /* for minor nos. from INET_MIN_DEV+2 thru 2^^18-1 */
752 vmem_t *ip_minor_arena_la; /* for minor nos. from 2^^18 thru 2^^32-1 */
753
754 int ip_debug;
755
756 /*
757 * Multirouting/CGTP stuff
758 */
759 int ip_cgtp_filter_rev = CGTP_FILTER_REV; /* CGTP hooks version */
760
761 /*
762 * IP tunables related declarations. Definitions are in ip_tunables.c
763 */
764 extern mod_prop_info_t ip_propinfo_tbl[];
765 extern int ip_propinfo_count;
766
767 /*
768 * Table of IP ioctls encoding the various properties of the ioctl and
769 * indexed based on the last byte of the ioctl command. Occasionally there
770 * is a clash, and there is more than 1 ioctl with the same last byte.
771 * In such a case 1 ioctl is encoded in the ndx table and the remaining
772 * ioctls are encoded in the misc table. An entry in the ndx table is
773 * retrieved by indexing on the last byte of the ioctl command and comparing
774 * the ioctl command with the value in the ndx table. In the event of a
775 * mismatch the misc table is then searched sequentially for the desired
776 * ioctl command.
777 *
778 * Entry: <command> <copyin_size> <flags> <cmd_type> <function> <restart_func>
779 */
780 ip_ioctl_cmd_t ip_ndx_ioctl_table[] = {
781 /* 000 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
782 /* 001 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
783 /* 002 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
784 /* 003 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
785 /* 004 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
786 /* 005 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
787 /* 006 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
788 /* 007 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
789 /* 008 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
790 /* 009 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
791
792 /* 010 */ { SIOCADDRT, sizeof (struct rtentry), IPI_PRIV,
793 MISC_CMD, ip_siocaddrt, NULL },
794 /* 011 */ { SIOCDELRT, sizeof (struct rtentry), IPI_PRIV,
795 MISC_CMD, ip_siocdelrt, NULL },
796
797 /* 012 */ { SIOCSIFADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
798 IF_CMD, ip_sioctl_addr, ip_sioctl_addr_restart },
799 /* 013 */ { SIOCGIFADDR, sizeof (struct ifreq), IPI_GET_CMD,
800 IF_CMD, ip_sioctl_get_addr, NULL },
801
802 /* 014 */ { SIOCSIFDSTADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
803 IF_CMD, ip_sioctl_dstaddr, ip_sioctl_dstaddr_restart },
804 /* 015 */ { SIOCGIFDSTADDR, sizeof (struct ifreq),
805 IPI_GET_CMD, IF_CMD, ip_sioctl_get_dstaddr, NULL },
806
807 /* 016 */ { SIOCSIFFLAGS, sizeof (struct ifreq),
808 IPI_PRIV | IPI_WR,
809 IF_CMD, ip_sioctl_flags, ip_sioctl_flags_restart },
810 /* 017 */ { SIOCGIFFLAGS, sizeof (struct ifreq),
811 IPI_MODOK | IPI_GET_CMD,
812 IF_CMD, ip_sioctl_get_flags, NULL },
813
814 /* 018 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
815 /* 019 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
816
817 /* copyin size cannot be coded for SIOCGIFCONF */
818 /* 020 */ { O_SIOCGIFCONF, 0, IPI_GET_CMD,
819 MISC_CMD, ip_sioctl_get_ifconf, NULL },
820
821 /* 021 */ { SIOCSIFMTU, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
822 IF_CMD, ip_sioctl_mtu, NULL },
823 /* 022 */ { SIOCGIFMTU, sizeof (struct ifreq), IPI_GET_CMD,
824 IF_CMD, ip_sioctl_get_mtu, NULL },
825 /* 023 */ { SIOCGIFBRDADDR, sizeof (struct ifreq),
826 IPI_GET_CMD, IF_CMD, ip_sioctl_get_brdaddr, NULL },
827 /* 024 */ { SIOCSIFBRDADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
828 IF_CMD, ip_sioctl_brdaddr, NULL },
829 /* 025 */ { SIOCGIFNETMASK, sizeof (struct ifreq),
830 IPI_GET_CMD, IF_CMD, ip_sioctl_get_netmask, NULL },
831 /* 026 */ { SIOCSIFNETMASK, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
832 IF_CMD, ip_sioctl_netmask, ip_sioctl_netmask_restart },
833 /* 027 */ { SIOCGIFMETRIC, sizeof (struct ifreq),
834 IPI_GET_CMD, IF_CMD, ip_sioctl_get_metric, NULL },
835 /* 028 */ { SIOCSIFMETRIC, sizeof (struct ifreq), IPI_PRIV,
836 IF_CMD, ip_sioctl_metric, NULL },
837 /* 029 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
838
839 /* See 166-168 below for extended SIOC*XARP ioctls */
840 /* 030 */ { SIOCSARP, sizeof (struct arpreq), IPI_PRIV | IPI_WR,
841 ARP_CMD, ip_sioctl_arp, NULL },
842 /* 031 */ { SIOCGARP, sizeof (struct arpreq), IPI_GET_CMD,
843 ARP_CMD, ip_sioctl_arp, NULL },
844 /* 032 */ { SIOCDARP, sizeof (struct arpreq), IPI_PRIV | IPI_WR,
845 ARP_CMD, ip_sioctl_arp, NULL },
846
847 /* 033 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
848 /* 034 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
849 /* 035 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
850 /* 036 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
851 /* 037 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
852 /* 038 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
853 /* 039 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
854 /* 040 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
855 /* 041 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
856 /* 042 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
857 /* 043 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
858 /* 044 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
859 /* 045 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
860 /* 046 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
861 /* 047 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
862 /* 048 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
863 /* 049 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
864 /* 050 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
865 /* 051 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
866 /* 052 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
867 /* 053 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
868
869 /* 054 */ { IF_UNITSEL, sizeof (int), IPI_PRIV | IPI_WR | IPI_MODOK,
870 MISC_CMD, if_unitsel, if_unitsel_restart },
871
872 /* 055 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
873 /* 056 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
874 /* 057 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
875 /* 058 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
876 /* 059 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
877 /* 060 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
878 /* 061 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
879 /* 062 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
880 /* 063 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
881 /* 064 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
882 /* 065 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
883 /* 066 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
884 /* 067 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
885 /* 068 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
886 /* 069 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
887 /* 070 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
888 /* 071 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
889 /* 072 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
890
891 /* 073 */ { SIOCSIFNAME, sizeof (struct ifreq),
892 IPI_PRIV | IPI_WR | IPI_MODOK,
893 IF_CMD, ip_sioctl_sifname, NULL },
894
895 /* 074 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
896 /* 075 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
897 /* 076 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
898 /* 077 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
899 /* 078 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
900 /* 079 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
901 /* 080 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
902 /* 081 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
903 /* 082 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
904 /* 083 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
905 /* 084 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
906 /* 085 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
907 /* 086 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
908
909 /* 087 */ { SIOCGIFNUM, sizeof (int), IPI_GET_CMD,
910 MISC_CMD, ip_sioctl_get_ifnum, NULL },
911 /* 088 */ { SIOCGIFMUXID, sizeof (struct ifreq), IPI_GET_CMD,
912 IF_CMD, ip_sioctl_get_muxid, NULL },
913 /* 089 */ { SIOCSIFMUXID, sizeof (struct ifreq),
914 IPI_PRIV | IPI_WR, IF_CMD, ip_sioctl_muxid, NULL },
915
916 /* Both if and lif variants share same func */
917 /* 090 */ { SIOCGIFINDEX, sizeof (struct ifreq), IPI_GET_CMD,
918 IF_CMD, ip_sioctl_get_lifindex, NULL },
919 /* Both if and lif variants share same func */
920 /* 091 */ { SIOCSIFINDEX, sizeof (struct ifreq),
921 IPI_PRIV | IPI_WR, IF_CMD, ip_sioctl_slifindex, NULL },
922
923 /* copyin size cannot be coded for SIOCGIFCONF */
924 /* 092 */ { SIOCGIFCONF, 0, IPI_GET_CMD,
925 MISC_CMD, ip_sioctl_get_ifconf, NULL },
926 /* 093 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
927 /* 094 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
928 /* 095 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
929 /* 096 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
930 /* 097 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
931 /* 098 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
932 /* 099 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
933 /* 100 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
934 /* 101 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
935 /* 102 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
936 /* 103 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
937 /* 104 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
938 /* 105 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
939 /* 106 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
940 /* 107 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
941 /* 108 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
942 /* 109 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
943
944 /* 110 */ { SIOCLIFREMOVEIF, sizeof (struct lifreq),
945 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_removeif,
946 ip_sioctl_removeif_restart },
947 /* 111 */ { SIOCLIFADDIF, sizeof (struct lifreq),
948 IPI_GET_CMD | IPI_PRIV | IPI_WR,
949 LIF_CMD, ip_sioctl_addif, NULL },
950 #define SIOCLIFADDR_NDX 112
951 /* 112 */ { SIOCSLIFADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
952 LIF_CMD, ip_sioctl_addr, ip_sioctl_addr_restart },
953 /* 113 */ { SIOCGLIFADDR, sizeof (struct lifreq),
954 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_addr, NULL },
955 /* 114 */ { SIOCSLIFDSTADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
956 LIF_CMD, ip_sioctl_dstaddr, ip_sioctl_dstaddr_restart },
957 /* 115 */ { SIOCGLIFDSTADDR, sizeof (struct lifreq),
958 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_dstaddr, NULL },
959 /* 116 */ { SIOCSLIFFLAGS, sizeof (struct lifreq),
960 IPI_PRIV | IPI_WR,
961 LIF_CMD, ip_sioctl_flags, ip_sioctl_flags_restart },
962 /* 117 */ { SIOCGLIFFLAGS, sizeof (struct lifreq),
963 IPI_GET_CMD | IPI_MODOK,
964 LIF_CMD, ip_sioctl_get_flags, NULL },
965
966 /* 118 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
967 /* 119 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
968
969 /* 120 */ { O_SIOCGLIFCONF, 0, IPI_GET_CMD, MISC_CMD,
970 ip_sioctl_get_lifconf, NULL },
971 /* 121 */ { SIOCSLIFMTU, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
972 LIF_CMD, ip_sioctl_mtu, NULL },
973 /* 122 */ { SIOCGLIFMTU, sizeof (struct lifreq), IPI_GET_CMD,
974 LIF_CMD, ip_sioctl_get_mtu, NULL },
975 /* 123 */ { SIOCGLIFBRDADDR, sizeof (struct lifreq),
976 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_brdaddr, NULL },
977 /* 124 */ { SIOCSLIFBRDADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
978 LIF_CMD, ip_sioctl_brdaddr, NULL },
979 /* 125 */ { SIOCGLIFNETMASK, sizeof (struct lifreq),
980 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_netmask, NULL },
981 /* 126 */ { SIOCSLIFNETMASK, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
982 LIF_CMD, ip_sioctl_netmask, ip_sioctl_netmask_restart },
983 /* 127 */ { SIOCGLIFMETRIC, sizeof (struct lifreq),
984 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_metric, NULL },
985 /* 128 */ { SIOCSLIFMETRIC, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
986 LIF_CMD, ip_sioctl_metric, NULL },
987 /* 129 */ { SIOCSLIFNAME, sizeof (struct lifreq),
988 IPI_PRIV | IPI_WR | IPI_MODOK,
989 LIF_CMD, ip_sioctl_slifname,
990 ip_sioctl_slifname_restart },
991
992 /* 130 */ { SIOCGLIFNUM, sizeof (struct lifnum), IPI_GET_CMD,
993 MISC_CMD, ip_sioctl_get_lifnum, NULL },
994 /* 131 */ { SIOCGLIFMUXID, sizeof (struct lifreq),
995 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_muxid, NULL },
996 /* 132 */ { SIOCSLIFMUXID, sizeof (struct lifreq),
997 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_muxid, NULL },
998 /* 133 */ { SIOCGLIFINDEX, sizeof (struct lifreq),
999 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lifindex, 0 },
1000 /* 134 */ { SIOCSLIFINDEX, sizeof (struct lifreq),
1001 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_slifindex, 0 },
1002 /* 135 */ { SIOCSLIFTOKEN, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1003 LIF_CMD, ip_sioctl_token, NULL },
1004 /* 136 */ { SIOCGLIFTOKEN, sizeof (struct lifreq),
1005 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_token, NULL },
1006 /* 137 */ { SIOCSLIFSUBNET, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1007 LIF_CMD, ip_sioctl_subnet, ip_sioctl_subnet_restart },
1008 /* 138 */ { SIOCGLIFSUBNET, sizeof (struct lifreq),
1009 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_subnet, NULL },
1010 /* 139 */ { SIOCSLIFLNKINFO, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1011 LIF_CMD, ip_sioctl_lnkinfo, NULL },
1012
1013 /* 140 */ { SIOCGLIFLNKINFO, sizeof (struct lifreq),
1014 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lnkinfo, NULL },
1015 /* 141 */ { SIOCLIFDELND, sizeof (struct lifreq), IPI_PRIV,
1016 LIF_CMD, ip_siocdelndp_v6, NULL },
1017 /* 142 */ { SIOCLIFGETND, sizeof (struct lifreq), IPI_GET_CMD,
1018 LIF_CMD, ip_siocqueryndp_v6, NULL },
1019 /* 143 */ { SIOCLIFSETND, sizeof (struct lifreq), IPI_PRIV,
1020 LIF_CMD, ip_siocsetndp_v6, NULL },
1021 /* 144 */ { SIOCTMYADDR, sizeof (struct sioc_addrreq), IPI_GET_CMD,
1022 MISC_CMD, ip_sioctl_tmyaddr, NULL },
1023 /* 145 */ { SIOCTONLINK, sizeof (struct sioc_addrreq), IPI_GET_CMD,
1024 MISC_CMD, ip_sioctl_tonlink, NULL },
1025 /* 146 */ { SIOCTMYSITE, sizeof (struct sioc_addrreq), 0,
1026 MISC_CMD, ip_sioctl_tmysite, NULL },
1027 /* 147 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1028 /* 148 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1029 /* IPSECioctls handled in ip_sioctl_copyin_setup itself */
1030 /* 149 */ { SIOCFIPSECONFIG, 0, IPI_PRIV, MISC_CMD, NULL, NULL },
1031 /* 150 */ { SIOCSIPSECONFIG, 0, IPI_PRIV, MISC_CMD, NULL, NULL },
1032 /* 151 */ { SIOCDIPSECONFIG, 0, IPI_PRIV, MISC_CMD, NULL, NULL },
1033 /* 152 */ { SIOCLIPSECONFIG, 0, IPI_PRIV, MISC_CMD, NULL, NULL },
1034
1035 /* 153 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1036
1037 /* 154 */ { SIOCGLIFBINDING, sizeof (struct lifreq), IPI_GET_CMD,
1038 LIF_CMD, ip_sioctl_get_binding, NULL },
1039 /* 155 */ { SIOCSLIFGROUPNAME, sizeof (struct lifreq),
1040 IPI_PRIV | IPI_WR,
1041 LIF_CMD, ip_sioctl_groupname, ip_sioctl_groupname },
1042 /* 156 */ { SIOCGLIFGROUPNAME, sizeof (struct lifreq),
1043 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_groupname, NULL },
1044 /* 157 */ { SIOCGLIFGROUPINFO, sizeof (lifgroupinfo_t),
1045 IPI_GET_CMD, MISC_CMD, ip_sioctl_groupinfo, NULL },
1046
1047 /* Leave 158-160 unused; used to be SIOC*IFARP ioctls */
1048 /* 158 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1049 /* 159 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1050 /* 160 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1051
1052 /* 161 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1053
1054 /* These are handled in ip_sioctl_copyin_setup itself */
1055 /* 162 */ { SIOCGIP6ADDRPOLICY, 0, IPI_NULL_BCONT,
1056 MISC_CMD, NULL, NULL },
1057 /* 163 */ { SIOCSIP6ADDRPOLICY, 0, IPI_PRIV | IPI_NULL_BCONT,
1058 MISC_CMD, NULL, NULL },
1059 /* 164 */ { SIOCGDSTINFO, 0, IPI_GET_CMD, MISC_CMD, NULL, NULL },
1060
1061 /* 165 */ { SIOCGLIFCONF, 0, IPI_GET_CMD, MISC_CMD,
1062 ip_sioctl_get_lifconf, NULL },
1063
1064 /* 166 */ { SIOCSXARP, sizeof (struct xarpreq), IPI_PRIV | IPI_WR,
1065 XARP_CMD, ip_sioctl_arp, NULL },
1066 /* 167 */ { SIOCGXARP, sizeof (struct xarpreq), IPI_GET_CMD,
1067 XARP_CMD, ip_sioctl_arp, NULL },
1068 /* 168 */ { SIOCDXARP, sizeof (struct xarpreq), IPI_PRIV | IPI_WR,
1069 XARP_CMD, ip_sioctl_arp, NULL },
1070
1071 /* SIOCPOPSOCKFS is not handled by IP */
1072 /* 169 */ { IPI_DONTCARE /* SIOCPOPSOCKFS */, 0, 0, 0, NULL, NULL },
1073
1074 /* 170 */ { SIOCGLIFZONE, sizeof (struct lifreq),
1075 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lifzone, NULL },
1076 /* 171 */ { SIOCSLIFZONE, sizeof (struct lifreq),
1077 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_slifzone,
1078 ip_sioctl_slifzone_restart },
1079 /* 172-174 are SCTP ioctls and not handled by IP */
1080 /* 172 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1081 /* 173 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1082 /* 174 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1083 /* 175 */ { SIOCGLIFUSESRC, sizeof (struct lifreq),
1084 IPI_GET_CMD, LIF_CMD,
1085 ip_sioctl_get_lifusesrc, 0 },
1086 /* 176 */ { SIOCSLIFUSESRC, sizeof (struct lifreq),
1087 IPI_PRIV | IPI_WR,
1088 LIF_CMD, ip_sioctl_slifusesrc,
1089 NULL },
1090 /* 177 */ { SIOCGLIFSRCOF, 0, IPI_GET_CMD, MISC_CMD,
1091 ip_sioctl_get_lifsrcof, NULL },
1092 /* 178 */ { SIOCGMSFILTER, sizeof (struct group_filter), IPI_GET_CMD,
1093 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1094 /* 179 */ { SIOCSMSFILTER, sizeof (struct group_filter), 0,
1095 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1096 /* 180 */ { SIOCGIPMSFILTER, sizeof (struct ip_msfilter), IPI_GET_CMD,
1097 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1098 /* 181 */ { SIOCSIPMSFILTER, sizeof (struct ip_msfilter), 0,
1099 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1100 /* 182 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1101 /* SIOCSENABLESDP is handled by SDP */
1102 /* 183 */ { IPI_DONTCARE /* SIOCSENABLESDP */, 0, 0, 0, NULL, NULL },
1103 /* 184 */ { IPI_DONTCARE /* SIOCSQPTR */, 0, 0, 0, NULL, NULL },
1104 /* 185 */ { SIOCGIFHWADDR, sizeof (struct ifreq), IPI_GET_CMD,
1105 IF_CMD, ip_sioctl_get_ifhwaddr, NULL },
1106 /* 186 */ { IPI_DONTCARE /* SIOCGSTAMP */, 0, 0, 0, NULL, NULL },
1107 /* 187 */ { SIOCILB, 0, IPI_PRIV | IPI_GET_CMD, MISC_CMD,
1108 ip_sioctl_ilb_cmd, NULL },
1109 /* 188 */ { SIOCGETPROP, 0, IPI_GET_CMD, 0, NULL, NULL },
1110 /* 189 */ { SIOCSETPROP, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL},
1111 /* 190 */ { SIOCGLIFDADSTATE, sizeof (struct lifreq),
1112 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_dadstate, NULL },
1113 /* 191 */ { SIOCSLIFPREFIX, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1114 LIF_CMD, ip_sioctl_prefix, ip_sioctl_prefix_restart },
1115 /* 192 */ { SIOCGLIFHWADDR, sizeof (struct lifreq), IPI_GET_CMD,
1116 LIF_CMD, ip_sioctl_get_lifhwaddr, NULL }
1117 };
1118
1119 int ip_ndx_ioctl_count = sizeof (ip_ndx_ioctl_table) / sizeof (ip_ioctl_cmd_t);
1120
1121 ip_ioctl_cmd_t ip_misc_ioctl_table[] = {
1122 { I_LINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1123 { I_UNLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1124 { I_PLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1125 { I_PUNLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1126 { ND_GET, 0, 0, 0, NULL, NULL },
1127 { ND_SET, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1128 { IP_IOCTL, 0, 0, 0, NULL, NULL },
1129 { SIOCGETVIFCNT, sizeof (struct sioc_vif_req), IPI_GET_CMD,
1130 MISC_CMD, mrt_ioctl},
1131 { SIOCGETSGCNT, sizeof (struct sioc_sg_req), IPI_GET_CMD,
1132 MISC_CMD, mrt_ioctl},
1133 { SIOCGETLSGCNT, sizeof (struct sioc_lsg_req), IPI_GET_CMD,
1134 MISC_CMD, mrt_ioctl}
1135 };
1136
1137 int ip_misc_ioctl_count =
1138 sizeof (ip_misc_ioctl_table) / sizeof (ip_ioctl_cmd_t);
1139
1140 int conn_drain_nthreads; /* Number of drainers reqd. */
1141 /* Settable in /etc/system */
1142 /* Defined in ip_ire.c */
1143 extern uint32_t ip_ire_max_bucket_cnt, ip6_ire_max_bucket_cnt;
1144 extern uint32_t ip_ire_min_bucket_cnt, ip6_ire_min_bucket_cnt;
1145 extern uint32_t ip_ire_mem_ratio, ip_ire_cpu_ratio;
1146
1147 static nv_t ire_nv_arr[] = {
1148 { IRE_BROADCAST, "BROADCAST" },
1149 { IRE_LOCAL, "LOCAL" },
1150 { IRE_LOOPBACK, "LOOPBACK" },
1151 { IRE_DEFAULT, "DEFAULT" },
1152 { IRE_PREFIX, "PREFIX" },
1153 { IRE_IF_NORESOLVER, "IF_NORESOL" },
1154 { IRE_IF_RESOLVER, "IF_RESOLV" },
1155 { IRE_IF_CLONE, "IF_CLONE" },
1156 { IRE_HOST, "HOST" },
1157 { IRE_MULTICAST, "MULTICAST" },
1158 { IRE_NOROUTE, "NOROUTE" },
1159 { 0 }
1160 };
1161
1162 nv_t *ire_nv_tbl = ire_nv_arr;
1163
1164 /* Simple ICMP IP Header Template */
1165 static ipha_t icmp_ipha = {
1166 IP_SIMPLE_HDR_VERSION, 0, 0, 0, 0, 0, IPPROTO_ICMP
1167 };
1168
1169 struct module_info ip_mod_info = {
1170 IP_MOD_ID, IP_MOD_NAME, IP_MOD_MINPSZ, IP_MOD_MAXPSZ, IP_MOD_HIWAT,
1171 IP_MOD_LOWAT
1172 };
1173
1174 /*
1175 * Duplicate static symbols within a module confuses mdb; so we avoid the
1176 * problem by making the symbols here distinct from those in udp.c.
1177 */
1178
1179 /*
1180 * Entry points for IP as a device and as a module.
1181 * We have separate open functions for the /dev/ip and /dev/ip6 devices.
1182 */
1183 static struct qinit iprinitv4 = {
1184 (pfi_t)ip_rput, NULL, ip_openv4, ip_close, NULL,
1185 &ip_mod_info
1186 };
1187
1188 struct qinit iprinitv6 = {
1189 (pfi_t)ip_rput_v6, NULL, ip_openv6, ip_close, NULL,
1190 &ip_mod_info
1191 };
1192
1193 static struct qinit ipwinit = {
1194 (pfi_t)ip_wput_nondata, (pfi_t)ip_wsrv, NULL, NULL, NULL,
1195 &ip_mod_info
1196 };
1197
1198 static struct qinit iplrinit = {
1199 (pfi_t)ip_lrput, NULL, ip_openv4, ip_close, NULL,
1200 &ip_mod_info
1201 };
1202
1203 static struct qinit iplwinit = {
1204 (pfi_t)ip_lwput, NULL, NULL, NULL, NULL,
1205 &ip_mod_info
1206 };
1207
1208 /* For AF_INET aka /dev/ip */
1209 struct streamtab ipinfov4 = {
1210 &iprinitv4, &ipwinit, &iplrinit, &iplwinit
1211 };
1212
1213 /* For AF_INET6 aka /dev/ip6 */
1214 struct streamtab ipinfov6 = {
1215 &iprinitv6, &ipwinit, &iplrinit, &iplwinit
1216 };
1217
1218 #ifdef DEBUG
1219 boolean_t skip_sctp_cksum = B_FALSE;
1220 #endif
1221
1222 /*
1223 * Generate an ICMP fragmentation needed message.
1224 * When called from ip_output side a minimal ip_recv_attr_t needs to be
1225 * constructed by the caller.
1226 */
1227 void
1228 icmp_frag_needed(mblk_t *mp, int mtu, ip_recv_attr_t *ira)
1229 {
1230 icmph_t icmph;
1231 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
1232
1233 mp = icmp_pkt_err_ok(mp, ira);
1234 if (mp == NULL)
1235 return;
1236
1237 bzero(&icmph, sizeof (icmph_t));
1238 icmph.icmph_type = ICMP_DEST_UNREACHABLE;
1239 icmph.icmph_code = ICMP_FRAGMENTATION_NEEDED;
1240 icmph.icmph_du_mtu = htons((uint16_t)mtu);
1241 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutFragNeeded);
1242 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDestUnreachs);
1243
1244 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
1245 }
1246
1247 /*
1248 * icmp_inbound_v4 deals with ICMP messages that are handled by IP.
1249 * If the ICMP message is consumed by IP, i.e., it should not be delivered
1250 * to any IPPROTO_ICMP raw sockets, then it returns NULL.
1251 * Likewise, if the ICMP error is misformed (too short, etc), then it
1252 * returns NULL. The caller uses this to determine whether or not to send
1253 * to raw sockets.
1254 *
1255 * All error messages are passed to the matching transport stream.
1256 *
1257 * The following cases are handled by icmp_inbound:
1258 * 1) It needs to send a reply back and possibly delivering it
1259 * to the "interested" upper clients.
1260 * 2) Return the mblk so that the caller can pass it to the RAW socket clients.
1261 * 3) It needs to change some values in IP only.
1262 * 4) It needs to change some values in IP and upper layers e.g TCP
1263 * by delivering an error to the upper layers.
1264 *
1265 * We handle the above three cases in the context of IPsec in the
1266 * following way :
1267 *
1268 * 1) Send the reply back in the same way as the request came in.
1269 * If it came in encrypted, it goes out encrypted. If it came in
1270 * clear, it goes out in clear. Thus, this will prevent chosen
1271 * plain text attack.
1272 * 2) The client may or may not expect things to come in secure.
1273 * If it comes in secure, the policy constraints are checked
1274 * before delivering it to the upper layers. If it comes in
1275 * clear, ipsec_inbound_accept_clear will decide whether to
1276 * accept this in clear or not. In both the cases, if the returned
1277 * message (IP header + 8 bytes) that caused the icmp message has
1278 * AH/ESP headers, it is sent up to AH/ESP for validation before
1279 * sending up. If there are only 8 bytes of returned message, then
1280 * upper client will not be notified.
1281 * 3) Check with global policy to see whether it matches the constaints.
1282 * But this will be done only if icmp_accept_messages_in_clear is
1283 * zero.
1284 * 4) If we need to change both in IP and ULP, then the decision taken
1285 * while affecting the values in IP and while delivering up to TCP
1286 * should be the same.
1287 *
1288 * There are two cases.
1289 *
1290 * a) If we reject data at the IP layer (ipsec_check_global_policy()
1291 * failed), we will not deliver it to the ULP, even though they
1292 * are *willing* to accept in *clear*. This is fine as our global
1293 * disposition to icmp messages asks us reject the datagram.
1294 *
1295 * b) If we accept data at the IP layer (ipsec_check_global_policy()
1296 * succeeded or icmp_accept_messages_in_clear is 1), and not able
1297 * to deliver it to ULP (policy failed), it can lead to
1298 * consistency problems. The cases known at this time are
1299 * ICMP_DESTINATION_UNREACHABLE messages with following code
1300 * values :
1301 *
1302 * - ICMP_FRAGMENTATION_NEEDED : IP adapts to the new value
1303 * and Upper layer rejects. Then the communication will
1304 * come to a stop. This is solved by making similar decisions
1305 * at both levels. Currently, when we are unable to deliver
1306 * to the Upper Layer (due to policy failures) while IP has
1307 * adjusted dce_pmtu, the next outbound datagram would
1308 * generate a local ICMP_FRAGMENTATION_NEEDED message - which
1309 * will be with the right level of protection. Thus the right
1310 * value will be communicated even if we are not able to
1311 * communicate when we get from the wire initially. But this
1312 * assumes there would be at least one outbound datagram after
1313 * IP has adjusted its dce_pmtu value. To make things
1314 * simpler, we accept in clear after the validation of
1315 * AH/ESP headers.
1316 *
1317 * - Other ICMP ERRORS : We may not be able to deliver it to the
1318 * upper layer depending on the level of protection the upper
1319 * layer expects and the disposition in ipsec_inbound_accept_clear().
1320 * ipsec_inbound_accept_clear() decides whether a given ICMP error
1321 * should be accepted in clear when the Upper layer expects secure.
1322 * Thus the communication may get aborted by some bad ICMP
1323 * packets.
1324 */
1325 mblk_t *
1326 icmp_inbound_v4(mblk_t *mp, ip_recv_attr_t *ira)
1327 {
1328 icmph_t *icmph;
1329 ipha_t *ipha; /* Outer header */
1330 int ip_hdr_length; /* Outer header length */
1331 boolean_t interested;
1332 ipif_t *ipif;
1333 uint32_t ts;
1334 uint32_t *tsp;
1335 timestruc_t now;
1336 ill_t *ill = ira->ira_ill;
1337 ip_stack_t *ipst = ill->ill_ipst;
1338 zoneid_t zoneid = ira->ira_zoneid;
1339 int len_needed;
1340 mblk_t *mp_ret = NULL;
1341
1342 ipha = (ipha_t *)mp->b_rptr;
1343
1344 BUMP_MIB(&ipst->ips_icmp_mib, icmpInMsgs);
1345
1346 ip_hdr_length = ira->ira_ip_hdr_length;
1347 if ((mp->b_wptr - mp->b_rptr) < (ip_hdr_length + ICMPH_SIZE)) {
1348 if (ira->ira_pktlen < (ip_hdr_length + ICMPH_SIZE)) {
1349 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
1350 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
1351 freemsg(mp);
1352 return (NULL);
1353 }
1354 /* Last chance to get real. */
1355 ipha = ip_pullup(mp, ip_hdr_length + ICMPH_SIZE, ira);
1356 if (ipha == NULL) {
1357 BUMP_MIB(&ipst->ips_icmp_mib, icmpInErrors);
1358 freemsg(mp);
1359 return (NULL);
1360 }
1361 }
1362
1363 /* The IP header will always be a multiple of four bytes */
1364 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1365 ip2dbg(("icmp_inbound_v4: type %d code %d\n", icmph->icmph_type,
1366 icmph->icmph_code));
1367
1368 /*
1369 * We will set "interested" to "true" if we should pass a copy to
1370 * the transport or if we handle the packet locally.
1371 */
1372 interested = B_FALSE;
1373 switch (icmph->icmph_type) {
1374 case ICMP_ECHO_REPLY:
1375 BUMP_MIB(&ipst->ips_icmp_mib, icmpInEchoReps);
1376 break;
1377 case ICMP_DEST_UNREACHABLE:
1378 if (icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED)
1379 BUMP_MIB(&ipst->ips_icmp_mib, icmpInFragNeeded);
1380 interested = B_TRUE; /* Pass up to transport */
1381 BUMP_MIB(&ipst->ips_icmp_mib, icmpInDestUnreachs);
1382 break;
1383 case ICMP_SOURCE_QUENCH:
1384 interested = B_TRUE; /* Pass up to transport */
1385 BUMP_MIB(&ipst->ips_icmp_mib, icmpInSrcQuenchs);
1386 break;
1387 case ICMP_REDIRECT:
1388 if (!ipst->ips_ip_ignore_redirect)
1389 interested = B_TRUE;
1390 BUMP_MIB(&ipst->ips_icmp_mib, icmpInRedirects);
1391 break;
1392 case ICMP_ECHO_REQUEST:
1393 /*
1394 * Whether to respond to echo requests that come in as IP
1395 * broadcasts or as IP multicast is subject to debate
1396 * (what isn't?). We aim to please, you pick it.
1397 * Default is do it.
1398 */
1399 if (ira->ira_flags & IRAF_MULTICAST) {
1400 /* multicast: respond based on tunable */
1401 interested = ipst->ips_ip_g_resp_to_echo_mcast;
1402 } else if (ira->ira_flags & IRAF_BROADCAST) {
1403 /* broadcast: respond based on tunable */
1404 interested = ipst->ips_ip_g_resp_to_echo_bcast;
1405 } else {
1406 /* unicast: always respond */
1407 interested = B_TRUE;
1408 }
1409 BUMP_MIB(&ipst->ips_icmp_mib, icmpInEchos);
1410 if (!interested) {
1411 /* We never pass these to RAW sockets */
1412 freemsg(mp);
1413 return (NULL);
1414 }
1415
1416 /* Check db_ref to make sure we can modify the packet. */
1417 if (mp->b_datap->db_ref > 1) {
1418 mblk_t *mp1;
1419
1420 mp1 = copymsg(mp);
1421 freemsg(mp);
1422 if (!mp1) {
1423 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1424 return (NULL);
1425 }
1426 mp = mp1;
1427 ipha = (ipha_t *)mp->b_rptr;
1428 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1429 }
1430 icmph->icmph_type = ICMP_ECHO_REPLY;
1431 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutEchoReps);
1432 icmp_send_reply_v4(mp, ipha, icmph, ira);
1433 return (NULL);
1434
1435 case ICMP_ROUTER_ADVERTISEMENT:
1436 case ICMP_ROUTER_SOLICITATION:
1437 break;
1438 case ICMP_TIME_EXCEEDED:
1439 interested = B_TRUE; /* Pass up to transport */
1440 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimeExcds);
1441 break;
1442 case ICMP_PARAM_PROBLEM:
1443 interested = B_TRUE; /* Pass up to transport */
1444 BUMP_MIB(&ipst->ips_icmp_mib, icmpInParmProbs);
1445 break;
1446 case ICMP_TIME_STAMP_REQUEST:
1447 /* Response to Time Stamp Requests is local policy. */
1448 if (ipst->ips_ip_g_resp_to_timestamp) {
1449 if (ira->ira_flags & IRAF_MULTIBROADCAST)
1450 interested =
1451 ipst->ips_ip_g_resp_to_timestamp_bcast;
1452 else
1453 interested = B_TRUE;
1454 }
1455 if (!interested) {
1456 /* We never pass these to RAW sockets */
1457 freemsg(mp);
1458 return (NULL);
1459 }
1460
1461 /* Make sure we have enough of the packet */
1462 len_needed = ip_hdr_length + ICMPH_SIZE +
1463 3 * sizeof (uint32_t);
1464
1465 if (mp->b_wptr - mp->b_rptr < len_needed) {
1466 ipha = ip_pullup(mp, len_needed, ira);
1467 if (ipha == NULL) {
1468 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1469 ip_drop_input("ipIfStatsInDiscards - ip_pullup",
1470 mp, ill);
1471 freemsg(mp);
1472 return (NULL);
1473 }
1474 /* Refresh following the pullup. */
1475 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1476 }
1477 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimestamps);
1478 /* Check db_ref to make sure we can modify the packet. */
1479 if (mp->b_datap->db_ref > 1) {
1480 mblk_t *mp1;
1481
1482 mp1 = copymsg(mp);
1483 freemsg(mp);
1484 if (!mp1) {
1485 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1486 return (NULL);
1487 }
1488 mp = mp1;
1489 ipha = (ipha_t *)mp->b_rptr;
1490 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1491 }
1492 icmph->icmph_type = ICMP_TIME_STAMP_REPLY;
1493 tsp = (uint32_t *)&icmph[1];
1494 tsp++; /* Skip past 'originate time' */
1495 /* Compute # of milliseconds since midnight */
1496 gethrestime(&now);
1497 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
1498 now.tv_nsec / (NANOSEC / MILLISEC);
1499 *tsp++ = htonl(ts); /* Lay in 'receive time' */
1500 *tsp++ = htonl(ts); /* Lay in 'send time' */
1501 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutTimestampReps);
1502 icmp_send_reply_v4(mp, ipha, icmph, ira);
1503 return (NULL);
1504
1505 case ICMP_TIME_STAMP_REPLY:
1506 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimestampReps);
1507 break;
1508 case ICMP_INFO_REQUEST:
1509 /* Per RFC 1122 3.2.2.7, ignore this. */
1510 case ICMP_INFO_REPLY:
1511 break;
1512 case ICMP_ADDRESS_MASK_REQUEST:
1513 if (ira->ira_flags & IRAF_MULTIBROADCAST) {
1514 interested =
1515 ipst->ips_ip_respond_to_address_mask_broadcast;
1516 } else {
1517 interested = B_TRUE;
1518 }
1519 if (!interested) {
1520 /* We never pass these to RAW sockets */
1521 freemsg(mp);
1522 return (NULL);
1523 }
1524 len_needed = ip_hdr_length + ICMPH_SIZE + IP_ADDR_LEN;
1525 if (mp->b_wptr - mp->b_rptr < len_needed) {
1526 ipha = ip_pullup(mp, len_needed, ira);
1527 if (ipha == NULL) {
1528 BUMP_MIB(ill->ill_ip_mib,
1529 ipIfStatsInTruncatedPkts);
1530 ip_drop_input("ipIfStatsInTruncatedPkts", mp,
1531 ill);
1532 freemsg(mp);
1533 return (NULL);
1534 }
1535 /* Refresh following the pullup. */
1536 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1537 }
1538 BUMP_MIB(&ipst->ips_icmp_mib, icmpInAddrMasks);
1539 /* Check db_ref to make sure we can modify the packet. */
1540 if (mp->b_datap->db_ref > 1) {
1541 mblk_t *mp1;
1542
1543 mp1 = copymsg(mp);
1544 freemsg(mp);
1545 if (!mp1) {
1546 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1547 return (NULL);
1548 }
1549 mp = mp1;
1550 ipha = (ipha_t *)mp->b_rptr;
1551 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1552 }
1553 /*
1554 * Need the ipif with the mask be the same as the source
1555 * address of the mask reply. For unicast we have a specific
1556 * ipif. For multicast/broadcast we only handle onlink
1557 * senders, and use the source address to pick an ipif.
1558 */
1559 ipif = ipif_lookup_addr(ipha->ipha_dst, ill, zoneid, ipst);
1560 if (ipif == NULL) {
1561 /* Broadcast or multicast */
1562 ipif = ipif_lookup_remote(ill, ipha->ipha_src, zoneid);
1563 if (ipif == NULL) {
1564 freemsg(mp);
1565 return (NULL);
1566 }
1567 }
1568 icmph->icmph_type = ICMP_ADDRESS_MASK_REPLY;
1569 bcopy(&ipif->ipif_net_mask, &icmph[1], IP_ADDR_LEN);
1570 ipif_refrele(ipif);
1571 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutAddrMaskReps);
1572 icmp_send_reply_v4(mp, ipha, icmph, ira);
1573 return (NULL);
1574
1575 case ICMP_ADDRESS_MASK_REPLY:
1576 BUMP_MIB(&ipst->ips_icmp_mib, icmpInAddrMaskReps);
1577 break;
1578 default:
1579 interested = B_TRUE; /* Pass up to transport */
1580 BUMP_MIB(&ipst->ips_icmp_mib, icmpInUnknowns);
1581 break;
1582 }
1583 /*
1584 * See if there is an ICMP client to avoid an extra copymsg/freemsg
1585 * if there isn't one.
1586 */
1587 if (ipst->ips_ipcl_proto_fanout_v4[IPPROTO_ICMP].connf_head != NULL) {
1588 /* If there is an ICMP client and we want one too, copy it. */
1589
1590 if (!interested) {
1591 /* Caller will deliver to RAW sockets */
1592 return (mp);
1593 }
1594 mp_ret = copymsg(mp);
1595 if (mp_ret == NULL) {
1596 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1597 ip_drop_input("ipIfStatsInDiscards - copymsg", mp, ill);
1598 }
1599 } else if (!interested) {
1600 /* Neither we nor raw sockets are interested. Drop packet now */
1601 freemsg(mp);
1602 return (NULL);
1603 }
1604
1605 /*
1606 * ICMP error or redirect packet. Make sure we have enough of
1607 * the header and that db_ref == 1 since we might end up modifying
1608 * the packet.
1609 */
1610 if (mp->b_cont != NULL) {
1611 if (ip_pullup(mp, -1, ira) == NULL) {
1612 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1613 ip_drop_input("ipIfStatsInDiscards - ip_pullup",
1614 mp, ill);
1615 freemsg(mp);
1616 return (mp_ret);
1617 }
1618 }
1619
1620 if (mp->b_datap->db_ref > 1) {
1621 mblk_t *mp1;
1622
1623 mp1 = copymsg(mp);
1624 if (mp1 == NULL) {
1625 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1626 ip_drop_input("ipIfStatsInDiscards - copymsg", mp, ill);
1627 freemsg(mp);
1628 return (mp_ret);
1629 }
1630 freemsg(mp);
1631 mp = mp1;
1632 }
1633
1634 /*
1635 * In case mp has changed, verify the message before any further
1636 * processes.
1637 */
1638 ipha = (ipha_t *)mp->b_rptr;
1639 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1640 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
1641 freemsg(mp);
1642 return (mp_ret);
1643 }
1644
1645 switch (icmph->icmph_type) {
1646 case ICMP_REDIRECT:
1647 icmp_redirect_v4(mp, ipha, icmph, ira);
1648 break;
1649 case ICMP_DEST_UNREACHABLE:
1650 if (icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED) {
1651 /* Update DCE and adjust MTU is icmp header if needed */
1652 icmp_inbound_too_big_v4(icmph, ira);
1653 }
1654 /* FALLTHRU */
1655 default:
1656 icmp_inbound_error_fanout_v4(mp, icmph, ira);
1657 break;
1658 }
1659 return (mp_ret);
1660 }
1661
1662 /*
1663 * Send an ICMP echo, timestamp or address mask reply.
1664 * The caller has already updated the payload part of the packet.
1665 * We handle the ICMP checksum, IP source address selection and feed
1666 * the packet into ip_output_simple.
1667 */
1668 static void
1669 icmp_send_reply_v4(mblk_t *mp, ipha_t *ipha, icmph_t *icmph,
1670 ip_recv_attr_t *ira)
1671 {
1672 uint_t ip_hdr_length = ira->ira_ip_hdr_length;
1673 ill_t *ill = ira->ira_ill;
1674 ip_stack_t *ipst = ill->ill_ipst;
1675 ip_xmit_attr_t ixas;
1676
1677 /* Send out an ICMP packet */
1678 icmph->icmph_checksum = 0;
1679 icmph->icmph_checksum = IP_CSUM(mp, ip_hdr_length, 0);
1680 /* Reset time to live. */
1681 ipha->ipha_ttl = ipst->ips_ip_def_ttl;
1682 {
1683 /* Swap source and destination addresses */
1684 ipaddr_t tmp;
1685
1686 tmp = ipha->ipha_src;
1687 ipha->ipha_src = ipha->ipha_dst;
1688 ipha->ipha_dst = tmp;
1689 }
1690 ipha->ipha_ident = 0;
1691 if (!IS_SIMPLE_IPH(ipha))
1692 icmp_options_update(ipha);
1693
1694 bzero(&ixas, sizeof (ixas));
1695 ixas.ixa_flags = IXAF_BASIC_SIMPLE_V4;
1696 ixas.ixa_zoneid = ira->ira_zoneid;
1697 ixas.ixa_cred = kcred;
1698 ixas.ixa_cpid = NOPID;
1699 ixas.ixa_tsl = ira->ira_tsl; /* Behave as a multi-level responder */
1700 ixas.ixa_ifindex = 0;
1701 ixas.ixa_ipst = ipst;
1702 ixas.ixa_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1703
1704 if (!(ira->ira_flags & IRAF_IPSEC_SECURE)) {
1705 /*
1706 * This packet should go out the same way as it
1707 * came in i.e in clear, independent of the IPsec policy
1708 * for transmitting packets.
1709 */
1710 ixas.ixa_flags |= IXAF_NO_IPSEC;
1711 } else {
1712 if (!ipsec_in_to_out(ira, &ixas, mp, ipha, NULL)) {
1713 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1714 /* Note: mp already consumed and ip_drop_packet done */
1715 return;
1716 }
1717 }
1718 if (ira->ira_flags & IRAF_MULTIBROADCAST) {
1719 /*
1720 * Not one or our addresses (IRE_LOCALs), thus we let
1721 * ip_output_simple pick the source.
1722 */
1723 ipha->ipha_src = INADDR_ANY;
1724 ixas.ixa_flags |= IXAF_SET_SOURCE;
1725 }
1726 /* Should we send with DF and use dce_pmtu? */
1727 if (ipst->ips_ipv4_icmp_return_pmtu) {
1728 ixas.ixa_flags |= IXAF_PMTU_DISCOVERY;
1729 ipha->ipha_fragment_offset_and_flags |= IPH_DF_HTONS;
1730 }
1731
1732 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutMsgs);
1733
1734 (void) ip_output_simple(mp, &ixas);
1735 ixa_cleanup(&ixas);
1736 }
1737
1738 /*
1739 * Verify the ICMP messages for either for ICMP error or redirect packet.
1740 * The caller should have fully pulled up the message. If it's a redirect
1741 * packet, only basic checks on IP header will be done; otherwise, verify
1742 * the packet by looking at the included ULP header.
1743 *
1744 * Called before icmp_inbound_error_fanout_v4 is called.
1745 */
1746 static boolean_t
1747 icmp_inbound_verify_v4(mblk_t *mp, icmph_t *icmph, ip_recv_attr_t *ira)
1748 {
1749 ill_t *ill = ira->ira_ill;
1750 int hdr_length;
1751 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
1752 conn_t *connp;
1753 ipha_t *ipha; /* Inner IP header */
1754
1755 ipha = (ipha_t *)&icmph[1];
1756 if ((uchar_t *)ipha + IP_SIMPLE_HDR_LENGTH > mp->b_wptr)
1757 goto truncated;
1758
1759 hdr_length = IPH_HDR_LENGTH(ipha);
1760
1761 if ((IPH_HDR_VERSION(ipha) != IPV4_VERSION))
1762 goto discard_pkt;
1763
1764 if (hdr_length < sizeof (ipha_t))
1765 goto truncated;
1766
1767 if ((uchar_t *)ipha + hdr_length > mp->b_wptr)
1768 goto truncated;
1769
1770 /*
1771 * Stop here for ICMP_REDIRECT.
1772 */
1773 if (icmph->icmph_type == ICMP_REDIRECT)
1774 return (B_TRUE);
1775
1776 /*
1777 * ICMP errors only.
1778 */
1779 switch (ipha->ipha_protocol) {
1780 case IPPROTO_UDP:
1781 /*
1782 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1783 * transport header.
1784 */
1785 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1786 mp->b_wptr)
1787 goto truncated;
1788 break;
1789 case IPPROTO_TCP: {
1790 tcpha_t *tcpha;
1791
1792 /*
1793 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1794 * transport header.
1795 */
1796 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1797 mp->b_wptr)
1798 goto truncated;
1799
1800 tcpha = (tcpha_t *)((uchar_t *)ipha + hdr_length);
1801 connp = ipcl_tcp_lookup_reversed_ipv4(ipha, tcpha, TCPS_LISTEN,
1802 ipst);
1803 if (connp == NULL)
1804 goto discard_pkt;
1805
1806 if ((connp->conn_verifyicmp != NULL) &&
1807 !connp->conn_verifyicmp(connp, tcpha, icmph, NULL, ira)) {
1808 CONN_DEC_REF(connp);
1809 goto discard_pkt;
1810 }
1811 CONN_DEC_REF(connp);
1812 break;
1813 }
1814 case IPPROTO_SCTP:
1815 /*
1816 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1817 * transport header.
1818 */
1819 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1820 mp->b_wptr)
1821 goto truncated;
1822 break;
1823 case IPPROTO_ESP:
1824 case IPPROTO_AH:
1825 break;
1826 case IPPROTO_ENCAP:
1827 if ((uchar_t *)ipha + hdr_length + sizeof (ipha_t) >
1828 mp->b_wptr)
1829 goto truncated;
1830 break;
1831 default:
1832 break;
1833 }
1834
1835 return (B_TRUE);
1836
1837 discard_pkt:
1838 /* Bogus ICMP error. */
1839 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1840 return (B_FALSE);
1841
1842 truncated:
1843 /* We pulled up everthing already. Must be truncated */
1844 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
1845 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
1846 return (B_FALSE);
1847 }
1848
1849 /* Table from RFC 1191 */
1850 static int icmp_frag_size_table[] =
1851 { 32000, 17914, 8166, 4352, 2002, 1496, 1006, 508, 296, 68 };
1852
1853 /*
1854 * Process received ICMP Packet too big.
1855 * Just handles the DCE create/update, including using the above table of
1856 * PMTU guesses. The caller is responsible for validating the packet before
1857 * passing it in and also to fanout the ICMP error to any matching transport
1858 * conns. Assumes the message has been fully pulled up and verified.
1859 *
1860 * Before getting here, the caller has called icmp_inbound_verify_v4()
1861 * that should have verified with ULP to prevent undoing the changes we're
1862 * going to make to DCE. For example, TCP might have verified that the packet
1863 * which generated error is in the send window.
1864 *
1865 * In some cases modified this MTU in the ICMP header packet; the caller
1866 * should pass to the matching ULP after this returns.
1867 */
1868 static void
1869 icmp_inbound_too_big_v4(icmph_t *icmph, ip_recv_attr_t *ira)
1870 {
1871 dce_t *dce;
1872 int old_mtu;
1873 int mtu, orig_mtu;
1874 ipaddr_t dst;
1875 boolean_t disable_pmtud;
1876 ill_t *ill = ira->ira_ill;
1877 ip_stack_t *ipst = ill->ill_ipst;
1878 uint_t hdr_length;
1879 ipha_t *ipha;
1880
1881 /* Caller already pulled up everything. */
1882 ipha = (ipha_t *)&icmph[1];
1883 ASSERT(icmph->icmph_type == ICMP_DEST_UNREACHABLE &&
1884 icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED);
1885 ASSERT(ill != NULL);
1886
1887 hdr_length = IPH_HDR_LENGTH(ipha);
1888
1889 /*
1890 * We handle path MTU for source routed packets since the DCE
1891 * is looked up using the final destination.
1892 */
1893 dst = ip_get_dst(ipha);
1894
1895 dce = dce_lookup_and_add_v4(dst, ipst);
1896 if (dce == NULL) {
1897 /* Couldn't add a unique one - ENOMEM */
1898 ip1dbg(("icmp_inbound_too_big_v4: no dce for 0x%x\n",
1899 ntohl(dst)));
1900 return;
1901 }
1902
1903 /* Check for MTU discovery advice as described in RFC 1191 */
1904 mtu = ntohs(icmph->icmph_du_mtu);
1905 orig_mtu = mtu;
1906 disable_pmtud = B_FALSE;
1907
1908 mutex_enter(&dce->dce_lock);
1909 if (dce->dce_flags & DCEF_PMTU)
1910 old_mtu = dce->dce_pmtu;
1911 else
1912 old_mtu = ill->ill_mtu;
1913
1914 if (icmph->icmph_du_zero != 0 || mtu < ipst->ips_ip_pmtu_min) {
1915 uint32_t length;
1916 int i;
1917
1918 /*
1919 * Use the table from RFC 1191 to figure out
1920 * the next "plateau" based on the length in
1921 * the original IP packet.
1922 */
1923 length = ntohs(ipha->ipha_length);
1924 DTRACE_PROBE2(ip4__pmtu__guess, dce_t *, dce,
1925 uint32_t, length);
1926 if (old_mtu <= length &&
1927 old_mtu >= length - hdr_length) {
1928 /*
1929 * Handle broken BSD 4.2 systems that
1930 * return the wrong ipha_length in ICMP
1931 * errors.
1932 */
1933 ip1dbg(("Wrong mtu: sent %d, dce %d\n",
1934 length, old_mtu));
1935 length -= hdr_length;
1936 }
1937 for (i = 0; i < A_CNT(icmp_frag_size_table); i++) {
1938 if (length > icmp_frag_size_table[i])
1939 break;
1940 }
1941 if (i == A_CNT(icmp_frag_size_table)) {
1942 /* Smaller than IP_MIN_MTU! */
1943 ip1dbg(("Too big for packet size %d\n",
1944 length));
1945 disable_pmtud = B_TRUE;
1946 mtu = ipst->ips_ip_pmtu_min;
1947 } else {
1948 mtu = icmp_frag_size_table[i];
1949 ip1dbg(("Calculated mtu %d, packet size %d, "
1950 "before %d\n", mtu, length, old_mtu));
1951 if (mtu < ipst->ips_ip_pmtu_min) {
1952 mtu = ipst->ips_ip_pmtu_min;
1953 disable_pmtud = B_TRUE;
1954 }
1955 }
1956 }
1957 if (disable_pmtud)
1958 dce->dce_flags |= DCEF_TOO_SMALL_PMTU;
1959 else
1960 dce->dce_flags &= ~DCEF_TOO_SMALL_PMTU;
1961
1962 dce->dce_pmtu = MIN(old_mtu, mtu);
1963 /* Prepare to send the new max frag size for the ULP. */
1964 icmph->icmph_du_zero = 0;
1965 icmph->icmph_du_mtu = htons((uint16_t)dce->dce_pmtu);
1966 DTRACE_PROBE4(ip4__pmtu__change, icmph_t *, icmph, dce_t *,
1967 dce, int, orig_mtu, int, mtu);
1968
1969 /* We now have a PMTU for sure */
1970 dce->dce_flags |= DCEF_PMTU;
1971 dce->dce_last_change_time = TICK_TO_SEC(ddi_get_lbolt64());
1972 mutex_exit(&dce->dce_lock);
1973 /*
1974 * After dropping the lock the new value is visible to everyone.
1975 * Then we bump the generation number so any cached values reinspect
1976 * the dce_t.
1977 */
1978 dce_increment_generation(dce);
1979 dce_refrele(dce);
1980 }
1981
1982 /*
1983 * If the packet in error is Self-Encapsulated, icmp_inbound_error_fanout_v4
1984 * calls this function.
1985 */
1986 static mblk_t *
1987 icmp_inbound_self_encap_error_v4(mblk_t *mp, ipha_t *ipha, ipha_t *in_ipha)
1988 {
1989 int length;
1990
1991 ASSERT(mp->b_datap->db_type == M_DATA);
1992
1993 /* icmp_inbound_v4 has already pulled up the whole error packet */
1994 ASSERT(mp->b_cont == NULL);
1995
1996 /*
1997 * The length that we want to overlay is the inner header
1998 * and what follows it.
1999 */
2000 length = msgdsize(mp) - ((uchar_t *)in_ipha - mp->b_rptr);
2001
2002 /*
2003 * Overlay the inner header and whatever follows it over the
2004 * outer header.
2005 */
2006 bcopy((uchar_t *)in_ipha, (uchar_t *)ipha, length);
2007
2008 /* Adjust for what we removed */
2009 mp->b_wptr -= (uchar_t *)in_ipha - (uchar_t *)ipha;
2010 return (mp);
2011 }
2012
2013 /*
2014 * Try to pass the ICMP message upstream in case the ULP cares.
2015 *
2016 * If the packet that caused the ICMP error is secure, we send
2017 * it to AH/ESP to make sure that the attached packet has a
2018 * valid association. ipha in the code below points to the
2019 * IP header of the packet that caused the error.
2020 *
2021 * For IPsec cases, we let the next-layer-up (which has access to
2022 * cached policy on the conn_t, or can query the SPD directly)
2023 * subtract out any IPsec overhead if they must. We therefore make no
2024 * adjustments here for IPsec overhead.
2025 *
2026 * IFN could have been generated locally or by some router.
2027 *
2028 * LOCAL : ire_send_wire (before calling ipsec_out_process) can call
2029 * icmp_frag_needed/icmp_pkt2big_v6 to generated a local IFN.
2030 * This happens because IP adjusted its value of MTU on an
2031 * earlier IFN message and could not tell the upper layer,
2032 * the new adjusted value of MTU e.g. Packet was encrypted
2033 * or there was not enough information to fanout to upper
2034 * layers. Thus on the next outbound datagram, ire_send_wire
2035 * generates the IFN, where IPsec processing has *not* been
2036 * done.
2037 *
2038 * Note that we retain ixa_fragsize across IPsec thus once
2039 * we have picking ixa_fragsize and entered ipsec_out_process we do
2040 * no change the fragsize even if the path MTU changes before
2041 * we reach ip_output_post_ipsec.
2042 *
2043 * In the local case, IRAF_LOOPBACK will be set indicating
2044 * that IFN was generated locally.
2045 *
2046 * ROUTER : IFN could be secure or non-secure.
2047 *
2048 * * SECURE : We use the IPSEC_IN to fanout to AH/ESP if the
2049 * packet in error has AH/ESP headers to validate the AH/ESP
2050 * headers. AH/ESP will verify whether there is a valid SA or
2051 * not and send it back. We will fanout again if we have more
2052 * data in the packet.
2053 *
2054 * If the packet in error does not have AH/ESP, we handle it
2055 * like any other case.
2056 *
2057 * * NON_SECURE : If the packet in error has AH/ESP headers, we send it
2058 * up to AH/ESP for validation. AH/ESP will verify whether there is a
2059 * valid SA or not and send it back. We will fanout again if
2060 * we have more data in the packet.
2061 *
2062 * If the packet in error does not have AH/ESP, we handle it
2063 * like any other case.
2064 *
2065 * The caller must have called icmp_inbound_verify_v4.
2066 */
2067 static void
2068 icmp_inbound_error_fanout_v4(mblk_t *mp, icmph_t *icmph, ip_recv_attr_t *ira)
2069 {
2070 uint16_t *up; /* Pointer to ports in ULP header */
2071 uint32_t ports; /* reversed ports for fanout */
2072 ipha_t ripha; /* With reversed addresses */
2073 ipha_t *ipha; /* Inner IP header */
2074 uint_t hdr_length; /* Inner IP header length */
2075 tcpha_t *tcpha;
2076 conn_t *connp;
2077 ill_t *ill = ira->ira_ill;
2078 ip_stack_t *ipst = ill->ill_ipst;
2079 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
2080 ill_t *rill = ira->ira_rill;
2081
2082 /* Caller already pulled up everything. */
2083 ipha = (ipha_t *)&icmph[1];
2084 ASSERT((uchar_t *)&ipha[1] <= mp->b_wptr);
2085 ASSERT(mp->b_cont == NULL);
2086
2087 hdr_length = IPH_HDR_LENGTH(ipha);
2088 ira->ira_protocol = ipha->ipha_protocol;
2089
2090 /*
2091 * We need a separate IP header with the source and destination
2092 * addresses reversed to do fanout/classification because the ipha in
2093 * the ICMP error is in the form we sent it out.
2094 */
2095 ripha.ipha_src = ipha->ipha_dst;
2096 ripha.ipha_dst = ipha->ipha_src;
2097 ripha.ipha_protocol = ipha->ipha_protocol;
2098 ripha.ipha_version_and_hdr_length = ipha->ipha_version_and_hdr_length;
2099
2100 ip2dbg(("icmp_inbound_error_v4: proto %d %x to %x: %d/%d\n",
2101 ripha.ipha_protocol, ntohl(ipha->ipha_src),
2102 ntohl(ipha->ipha_dst),
2103 icmph->icmph_type, icmph->icmph_code));
2104
2105 switch (ipha->ipha_protocol) {
2106 case IPPROTO_UDP:
2107 up = (uint16_t *)((uchar_t *)ipha + hdr_length);
2108
2109 /* Attempt to find a client stream based on port. */
2110 ip2dbg(("icmp_inbound_error_v4: UDP ports %d to %d\n",
2111 ntohs(up[0]), ntohs(up[1])));
2112
2113 /* Note that we send error to all matches. */
2114 ira->ira_flags |= IRAF_ICMP_ERROR;
2115 ip_fanout_udp_multi_v4(mp, &ripha, up[0], up[1], ira);
2116 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2117 return;
2118
2119 case IPPROTO_TCP:
2120 /*
2121 * Find a TCP client stream for this packet.
2122 * Note that we do a reverse lookup since the header is
2123 * in the form we sent it out.
2124 */
2125 tcpha = (tcpha_t *)((uchar_t *)ipha + hdr_length);
2126 connp = ipcl_tcp_lookup_reversed_ipv4(ipha, tcpha, TCPS_LISTEN,
2127 ipst);
2128 if (connp == NULL)
2129 goto discard_pkt;
2130
2131 if (CONN_INBOUND_POLICY_PRESENT(connp, ipss) ||
2132 (ira->ira_flags & IRAF_IPSEC_SECURE)) {
2133 mp = ipsec_check_inbound_policy(mp, connp,
2134 ipha, NULL, ira);
2135 if (mp == NULL) {
2136 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2137 /* Note that mp is NULL */
2138 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2139 CONN_DEC_REF(connp);
2140 return;
2141 }
2142 }
2143
2144 ira->ira_flags |= IRAF_ICMP_ERROR;
2145 ira->ira_ill = ira->ira_rill = NULL;
2146 if (IPCL_IS_TCP(connp)) {
2147 SQUEUE_ENTER_ONE(connp->conn_sqp, mp,
2148 connp->conn_recvicmp, connp, ira, SQ_FILL,
2149 SQTAG_TCP_INPUT_ICMP_ERR);
2150 } else {
2151 /* Not TCP; must be SOCK_RAW, IPPROTO_TCP */
2152 (connp->conn_recv)(connp, mp, NULL, ira);
2153 CONN_DEC_REF(connp);
2154 }
2155 ira->ira_ill = ill;
2156 ira->ira_rill = rill;
2157 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2158 return;
2159
2160 case IPPROTO_SCTP:
2161 up = (uint16_t *)((uchar_t *)ipha + hdr_length);
2162 /* Find a SCTP client stream for this packet. */
2163 ((uint16_t *)&ports)[0] = up[1];
2164 ((uint16_t *)&ports)[1] = up[0];
2165
2166 ira->ira_flags |= IRAF_ICMP_ERROR;
2167 ip_fanout_sctp(mp, &ripha, NULL, ports, ira);
2168 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2169 return;
2170
2171 case IPPROTO_ESP:
2172 case IPPROTO_AH:
2173 if (!ipsec_loaded(ipss)) {
2174 ip_proto_not_sup(mp, ira);
2175 return;
2176 }
2177
2178 if (ipha->ipha_protocol == IPPROTO_ESP)
2179 mp = ipsecesp_icmp_error(mp, ira);
2180 else
2181 mp = ipsecah_icmp_error(mp, ira);
2182 if (mp == NULL)
2183 return;
2184
2185 /* Just in case ipsec didn't preserve the NULL b_cont */
2186 if (mp->b_cont != NULL) {
2187 if (!pullupmsg(mp, -1))
2188 goto discard_pkt;
2189 }
2190
2191 /*
2192 * Note that ira_pktlen and ira_ip_hdr_length are no longer
2193 * correct, but we don't use them any more here.
2194 *
2195 * If succesful, the mp has been modified to not include
2196 * the ESP/AH header so we can fanout to the ULP's icmp
2197 * error handler.
2198 */
2199 if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH)
2200 goto truncated;
2201
2202 /* Verify the modified message before any further processes. */
2203 ipha = (ipha_t *)mp->b_rptr;
2204 hdr_length = IPH_HDR_LENGTH(ipha);
2205 icmph = (icmph_t *)&mp->b_rptr[hdr_length];
2206 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
2207 freemsg(mp);
2208 return;
2209 }
2210
2211 icmp_inbound_error_fanout_v4(mp, icmph, ira);
2212 return;
2213
2214 case IPPROTO_ENCAP: {
2215 /* Look for self-encapsulated packets that caused an error */
2216 ipha_t *in_ipha;
2217
2218 /*
2219 * Caller has verified that length has to be
2220 * at least the size of IP header.
2221 */
2222 ASSERT(hdr_length >= sizeof (ipha_t));
2223 /*
2224 * Check the sanity of the inner IP header like
2225 * we did for the outer header.
2226 */
2227 in_ipha = (ipha_t *)((uchar_t *)ipha + hdr_length);
2228 if ((IPH_HDR_VERSION(in_ipha) != IPV4_VERSION)) {
2229 goto discard_pkt;
2230 }
2231 if (IPH_HDR_LENGTH(in_ipha) < sizeof (ipha_t)) {
2232 goto discard_pkt;
2233 }
2234 /* Check for Self-encapsulated tunnels */
2235 if (in_ipha->ipha_src == ipha->ipha_src &&
2236 in_ipha->ipha_dst == ipha->ipha_dst) {
2237
2238 mp = icmp_inbound_self_encap_error_v4(mp, ipha,
2239 in_ipha);
2240 if (mp == NULL)
2241 goto discard_pkt;
2242
2243 /*
2244 * Just in case self_encap didn't preserve the NULL
2245 * b_cont
2246 */
2247 if (mp->b_cont != NULL) {
2248 if (!pullupmsg(mp, -1))
2249 goto discard_pkt;
2250 }
2251 /*
2252 * Note that ira_pktlen and ira_ip_hdr_length are no
2253 * longer correct, but we don't use them any more here.
2254 */
2255 if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH)
2256 goto truncated;
2257
2258 /*
2259 * Verify the modified message before any further
2260 * processes.
2261 */
2262 ipha = (ipha_t *)mp->b_rptr;
2263 hdr_length = IPH_HDR_LENGTH(ipha);
2264 icmph = (icmph_t *)&mp->b_rptr[hdr_length];
2265 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
2266 freemsg(mp);
2267 return;
2268 }
2269
2270 /*
2271 * The packet in error is self-encapsualted.
2272 * And we are finding it further encapsulated
2273 * which we could not have possibly generated.
2274 */
2275 if (ipha->ipha_protocol == IPPROTO_ENCAP) {
2276 goto discard_pkt;
2277 }
2278 icmp_inbound_error_fanout_v4(mp, icmph, ira);
2279 return;
2280 }
2281 /* No self-encapsulated */
2282 /* FALLTHRU */
2283 }
2284 case IPPROTO_IPV6:
2285 if ((connp = ipcl_iptun_classify_v4(&ripha.ipha_src,
2286 &ripha.ipha_dst, ipst)) != NULL) {
2287 ira->ira_flags |= IRAF_ICMP_ERROR;
2288 connp->conn_recvicmp(connp, mp, NULL, ira);
2289 CONN_DEC_REF(connp);
2290 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2291 return;
2292 }
2293 /*
2294 * No IP tunnel is interested, fallthrough and see
2295 * if a raw socket will want it.
2296 */
2297 /* FALLTHRU */
2298 default:
2299 ira->ira_flags |= IRAF_ICMP_ERROR;
2300 ip_fanout_proto_v4(mp, &ripha, ira);
2301 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2302 return;
2303 }
2304 /* NOTREACHED */
2305 discard_pkt:
2306 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2307 ip1dbg(("icmp_inbound_error_fanout_v4: drop pkt\n"));
2308 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2309 freemsg(mp);
2310 return;
2311
2312 truncated:
2313 /* We pulled up everthing already. Must be truncated */
2314 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
2315 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
2316 freemsg(mp);
2317 }
2318
2319 /*
2320 * Common IP options parser.
2321 *
2322 * Setup routine: fill in *optp with options-parsing state, then
2323 * tail-call ipoptp_next to return the first option.
2324 */
2325 uint8_t
2326 ipoptp_first(ipoptp_t *optp, ipha_t *ipha)
2327 {
2328 uint32_t totallen; /* total length of all options */
2329
2330 totallen = ipha->ipha_version_and_hdr_length -
2331 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
2332 totallen <<= 2;
2333 optp->ipoptp_next = (uint8_t *)(&ipha[1]);
2334 optp->ipoptp_end = optp->ipoptp_next + totallen;
2335 optp->ipoptp_flags = 0;
2336 return (ipoptp_next(optp));
2337 }
2338
2339 /* Like above but without an ipha_t */
2340 uint8_t
2341 ipoptp_first2(ipoptp_t *optp, uint32_t totallen, uint8_t *opt)
2342 {
2343 optp->ipoptp_next = opt;
2344 optp->ipoptp_end = optp->ipoptp_next + totallen;
2345 optp->ipoptp_flags = 0;
2346 return (ipoptp_next(optp));
2347 }
2348
2349 /*
2350 * Common IP options parser: extract next option.
2351 */
2352 uint8_t
2353 ipoptp_next(ipoptp_t *optp)
2354 {
2355 uint8_t *end = optp->ipoptp_end;
2356 uint8_t *cur = optp->ipoptp_next;
2357 uint8_t opt, len, pointer;
2358
2359 /*
2360 * If cur > end already, then the ipoptp_end or ipoptp_next pointer
2361 * has been corrupted.
2362 */
2363 ASSERT(cur <= end);
2364
2365 if (cur == end)
2366 return (IPOPT_EOL);
2367
2368 opt = cur[IPOPT_OPTVAL];
2369
2370 /*
2371 * Skip any NOP options.
2372 */
2373 while (opt == IPOPT_NOP) {
2374 cur++;
2375 if (cur == end)
2376 return (IPOPT_EOL);
2377 opt = cur[IPOPT_OPTVAL];
2378 }
2379
2380 if (opt == IPOPT_EOL)
2381 return (IPOPT_EOL);
2382
2383 /*
2384 * Option requiring a length.
2385 */
2386 if ((cur + 1) >= end) {
2387 optp->ipoptp_flags |= IPOPTP_ERROR;
2388 return (IPOPT_EOL);
2389 }
2390 len = cur[IPOPT_OLEN];
2391 if (len < 2) {
2392 optp->ipoptp_flags |= IPOPTP_ERROR;
2393 return (IPOPT_EOL);
2394 }
2395 optp->ipoptp_cur = cur;
2396 optp->ipoptp_len = len;
2397 optp->ipoptp_next = cur + len;
2398 if (cur + len > end) {
2399 optp->ipoptp_flags |= IPOPTP_ERROR;
2400 return (IPOPT_EOL);
2401 }
2402
2403 /*
2404 * For the options which require a pointer field, make sure
2405 * its there, and make sure it points to either something
2406 * inside this option, or the end of the option.
2407 */
2408 switch (opt) {
2409 case IPOPT_RR:
2410 case IPOPT_TS:
2411 case IPOPT_LSRR:
2412 case IPOPT_SSRR:
2413 if (len <= IPOPT_OFFSET) {
2414 optp->ipoptp_flags |= IPOPTP_ERROR;
2415 return (opt);
2416 }
2417 pointer = cur[IPOPT_OFFSET];
2418 if (pointer - 1 > len) {
2419 optp->ipoptp_flags |= IPOPTP_ERROR;
2420 return (opt);
2421 }
2422 break;
2423 }
2424
2425 /*
2426 * Sanity check the pointer field based on the type of the
2427 * option.
2428 */
2429 switch (opt) {
2430 case IPOPT_RR:
2431 case IPOPT_SSRR:
2432 case IPOPT_LSRR:
2433 if (pointer < IPOPT_MINOFF_SR)
2434 optp->ipoptp_flags |= IPOPTP_ERROR;
2435 break;
2436 case IPOPT_TS:
2437 if (pointer < IPOPT_MINOFF_IT)
2438 optp->ipoptp_flags |= IPOPTP_ERROR;
2439 /*
2440 * Note that the Internet Timestamp option also
2441 * contains two four bit fields (the Overflow field,
2442 * and the Flag field), which follow the pointer
2443 * field. We don't need to check that these fields
2444 * fall within the length of the option because this
2445 * was implicitely done above. We've checked that the
2446 * pointer value is at least IPOPT_MINOFF_IT, and that
2447 * it falls within the option. Since IPOPT_MINOFF_IT >
2448 * IPOPT_POS_OV_FLG, we don't need the explicit check.
2449 */
2450 ASSERT(len > IPOPT_POS_OV_FLG);
2451 break;
2452 }
2453
2454 return (opt);
2455 }
2456
2457 /*
2458 * Use the outgoing IP header to create an IP_OPTIONS option the way
2459 * it was passed down from the application.
2460 *
2461 * This is compatible with BSD in that it returns
2462 * the reverse source route with the final destination
2463 * as the last entry. The first 4 bytes of the option
2464 * will contain the final destination.
2465 */
2466 int
2467 ip_opt_get_user(conn_t *connp, uchar_t *buf)
2468 {
2469 ipoptp_t opts;
2470 uchar_t *opt;
2471 uint8_t optval;
2472 uint8_t optlen;
2473 uint32_t len = 0;
2474 uchar_t *buf1 = buf;
2475 uint32_t totallen;
2476 ipaddr_t dst;
2477 ip_pkt_t *ipp = &connp->conn_xmit_ipp;
2478
2479 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
2480 return (0);
2481
2482 totallen = ipp->ipp_ipv4_options_len;
2483 if (totallen & 0x3)
2484 return (0);
2485
2486 buf += IP_ADDR_LEN; /* Leave room for final destination */
2487 len += IP_ADDR_LEN;
2488 bzero(buf1, IP_ADDR_LEN);
2489
2490 dst = connp->conn_faddr_v4;
2491
2492 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
2493 optval != IPOPT_EOL;
2494 optval = ipoptp_next(&opts)) {
2495 int off;
2496
2497 opt = opts.ipoptp_cur;
2498 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
2499 break;
2500 }
2501 optlen = opts.ipoptp_len;
2502
2503 switch (optval) {
2504 case IPOPT_SSRR:
2505 case IPOPT_LSRR:
2506
2507 /*
2508 * Insert destination as the first entry in the source
2509 * route and move down the entries on step.
2510 * The last entry gets placed at buf1.
2511 */
2512 buf[IPOPT_OPTVAL] = optval;
2513 buf[IPOPT_OLEN] = optlen;
2514 buf[IPOPT_OFFSET] = optlen;
2515
2516 off = optlen - IP_ADDR_LEN;
2517 if (off < 0) {
2518 /* No entries in source route */
2519 break;
2520 }
2521 /* Last entry in source route if not already set */
2522 if (dst == INADDR_ANY)
2523 bcopy(opt + off, buf1, IP_ADDR_LEN);
2524 off -= IP_ADDR_LEN;
2525
2526 while (off > 0) {
2527 bcopy(opt + off,
2528 buf + off + IP_ADDR_LEN,
2529 IP_ADDR_LEN);
2530 off -= IP_ADDR_LEN;
2531 }
2532 /* ipha_dst into first slot */
2533 bcopy(&dst, buf + off + IP_ADDR_LEN,
2534 IP_ADDR_LEN);
2535 buf += optlen;
2536 len += optlen;
2537 break;
2538
2539 default:
2540 bcopy(opt, buf, optlen);
2541 buf += optlen;
2542 len += optlen;
2543 break;
2544 }
2545 }
2546 done:
2547 /* Pad the resulting options */
2548 while (len & 0x3) {
2549 *buf++ = IPOPT_EOL;
2550 len++;
2551 }
2552 return (len);
2553 }
2554
2555 /*
2556 * Update any record route or timestamp options to include this host.
2557 * Reverse any source route option.
2558 * This routine assumes that the options are well formed i.e. that they
2559 * have already been checked.
2560 */
2561 static void
2562 icmp_options_update(ipha_t *ipha)
2563 {
2564 ipoptp_t opts;
2565 uchar_t *opt;
2566 uint8_t optval;
2567 ipaddr_t src; /* Our local address */
2568 ipaddr_t dst;
2569
2570 ip2dbg(("icmp_options_update\n"));
2571 src = ipha->ipha_src;
2572 dst = ipha->ipha_dst;
2573
2574 for (optval = ipoptp_first(&opts, ipha);
2575 optval != IPOPT_EOL;
2576 optval = ipoptp_next(&opts)) {
2577 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
2578 opt = opts.ipoptp_cur;
2579 ip2dbg(("icmp_options_update: opt %d, len %d\n",
2580 optval, opts.ipoptp_len));
2581 switch (optval) {
2582 int off1, off2;
2583 case IPOPT_SSRR:
2584 case IPOPT_LSRR:
2585 /*
2586 * Reverse the source route. The first entry
2587 * should be the next to last one in the current
2588 * source route (the last entry is our address).
2589 * The last entry should be the final destination.
2590 */
2591 off1 = IPOPT_MINOFF_SR - 1;
2592 off2 = opt[IPOPT_OFFSET] - IP_ADDR_LEN - 1;
2593 if (off2 < 0) {
2594 /* No entries in source route */
2595 ip1dbg((
2596 "icmp_options_update: bad src route\n"));
2597 break;
2598 }
2599 bcopy((char *)opt + off2, &dst, IP_ADDR_LEN);
2600 bcopy(&ipha->ipha_dst, (char *)opt + off2, IP_ADDR_LEN);
2601 bcopy(&dst, &ipha->ipha_dst, IP_ADDR_LEN);
2602 off2 -= IP_ADDR_LEN;
2603
2604 while (off1 < off2) {
2605 bcopy((char *)opt + off1, &src, IP_ADDR_LEN);
2606 bcopy((char *)opt + off2, (char *)opt + off1,
2607 IP_ADDR_LEN);
2608 bcopy(&src, (char *)opt + off2, IP_ADDR_LEN);
2609 off1 += IP_ADDR_LEN;
2610 off2 -= IP_ADDR_LEN;
2611 }
2612 opt[IPOPT_OFFSET] = IPOPT_MINOFF_SR;
2613 break;
2614 }
2615 }
2616 }
2617
2618 /*
2619 * Process received ICMP Redirect messages.
2620 * Assumes the caller has verified that the headers are in the pulled up mblk.
2621 * Consumes mp.
2622 */
2623 static void
2624 icmp_redirect_v4(mblk_t *mp, ipha_t *ipha, icmph_t *icmph, ip_recv_attr_t *ira)
2625 {
2626 ire_t *ire, *nire;
2627 ire_t *prev_ire;
2628 ipaddr_t src, dst, gateway;
2629 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2630 ipha_t *inner_ipha; /* Inner IP header */
2631
2632 /* Caller already pulled up everything. */
2633 inner_ipha = (ipha_t *)&icmph[1];
2634 src = ipha->ipha_src;
2635 dst = inner_ipha->ipha_dst;
2636 gateway = icmph->icmph_rd_gateway;
2637 /* Make sure the new gateway is reachable somehow. */
2638 ire = ire_ftable_lookup_v4(gateway, 0, 0, IRE_ONLINK, NULL,
2639 ALL_ZONES, NULL, MATCH_IRE_TYPE, 0, ipst, NULL);
2640 /*
2641 * Make sure we had a route for the dest in question and that
2642 * that route was pointing to the old gateway (the source of the
2643 * redirect packet.)
2644 * We do longest match and then compare ire_gateway_addr below.
2645 */
2646 prev_ire = ire_ftable_lookup_v4(dst, 0, 0, 0, NULL, ALL_ZONES,
2647 NULL, MATCH_IRE_DSTONLY, 0, ipst, NULL);
2648 /*
2649 * Check that
2650 * the redirect was not from ourselves
2651 * the new gateway and the old gateway are directly reachable
2652 */
2653 if (prev_ire == NULL || ire == NULL ||
2654 (prev_ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK)) ||
2655 (prev_ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) ||
2656 !(ire->ire_type & IRE_IF_ALL) ||
2657 prev_ire->ire_gateway_addr != src) {
2658 BUMP_MIB(&ipst->ips_icmp_mib, icmpInBadRedirects);
2659 ip_drop_input("icmpInBadRedirects - ire", mp, ira->ira_ill);
2660 freemsg(mp);
2661 if (ire != NULL)
2662 ire_refrele(ire);
2663 if (prev_ire != NULL)
2664 ire_refrele(prev_ire);
2665 return;
2666 }
2667
2668 ire_refrele(prev_ire);
2669 ire_refrele(ire);
2670
2671 /*
2672 * TODO: more precise handling for cases 0, 2, 3, the latter two
2673 * require TOS routing
2674 */
2675 switch (icmph->icmph_code) {
2676 case 0:
2677 case 1:
2678 /* TODO: TOS specificity for cases 2 and 3 */
2679 case 2:
2680 case 3:
2681 break;
2682 default:
2683 BUMP_MIB(&ipst->ips_icmp_mib, icmpInBadRedirects);
2684 ip_drop_input("icmpInBadRedirects - code", mp, ira->ira_ill);
2685 freemsg(mp);
2686 return;
2687 }
2688 /*
2689 * Create a Route Association. This will allow us to remember that
2690 * someone we believe told us to use the particular gateway.
2691 */
2692 ire = ire_create(
2693 (uchar_t *)&dst, /* dest addr */
2694 (uchar_t *)&ip_g_all_ones, /* mask */
2695 (uchar_t *)&gateway, /* gateway addr */
2696 IRE_HOST,
2697 NULL, /* ill */
2698 ALL_ZONES,
2699 (RTF_DYNAMIC | RTF_GATEWAY | RTF_HOST),
2700 NULL, /* tsol_gc_t */
2701 ipst);
2702
2703 if (ire == NULL) {
2704 freemsg(mp);
2705 return;
2706 }
2707 nire = ire_add(ire);
2708 /* Check if it was a duplicate entry */
2709 if (nire != NULL && nire != ire) {
2710 ASSERT(nire->ire_identical_ref > 1);
2711 ire_delete(nire);
2712 ire_refrele(nire);
2713 nire = NULL;
2714 }
2715 ire = nire;
2716 if (ire != NULL) {
2717 ire_refrele(ire); /* Held in ire_add */
2718
2719 /* tell routing sockets that we received a redirect */
2720 ip_rts_change(RTM_REDIRECT, dst, gateway, IP_HOST_MASK, 0, src,
2721 (RTF_DYNAMIC | RTF_GATEWAY | RTF_HOST), 0,
2722 (RTA_DST | RTA_GATEWAY | RTA_NETMASK | RTA_AUTHOR), ipst);
2723 }
2724
2725 /*
2726 * Delete any existing IRE_HOST type redirect ires for this destination.
2727 * This together with the added IRE has the effect of
2728 * modifying an existing redirect.
2729 */
2730 prev_ire = ire_ftable_lookup_v4(dst, 0, src, IRE_HOST, NULL,
2731 ALL_ZONES, NULL, (MATCH_IRE_GW | MATCH_IRE_TYPE), 0, ipst, NULL);
2732 if (prev_ire != NULL) {
2733 if (prev_ire ->ire_flags & RTF_DYNAMIC)
2734 ire_delete(prev_ire);
2735 ire_refrele(prev_ire);
2736 }
2737
2738 freemsg(mp);
2739 }
2740
2741 /*
2742 * Generate an ICMP parameter problem message.
2743 * When called from ip_output side a minimal ip_recv_attr_t needs to be
2744 * constructed by the caller.
2745 */
2746 static void
2747 icmp_param_problem(mblk_t *mp, uint8_t ptr, ip_recv_attr_t *ira)
2748 {
2749 icmph_t icmph;
2750 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2751
2752 mp = icmp_pkt_err_ok(mp, ira);
2753 if (mp == NULL)
2754 return;
2755
2756 bzero(&icmph, sizeof (icmph_t));
2757 icmph.icmph_type = ICMP_PARAM_PROBLEM;
2758 icmph.icmph_pp_ptr = ptr;
2759 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutParmProbs);
2760 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
2761 }
2762
2763 /*
2764 * Build and ship an IPv4 ICMP message using the packet data in mp, and
2765 * the ICMP header pointed to by "stuff". (May be called as writer.)
2766 * Note: assumes that icmp_pkt_err_ok has been called to verify that
2767 * an icmp error packet can be sent.
2768 * Assigns an appropriate source address to the packet. If ipha_dst is
2769 * one of our addresses use it for source. Otherwise let ip_output_simple
2770 * pick the source address.
2771 */
2772 static void
2773 icmp_pkt(mblk_t *mp, void *stuff, size_t len, ip_recv_attr_t *ira)
2774 {
2775 ipaddr_t dst;
2776 icmph_t *icmph;
2777 ipha_t *ipha;
2778 uint_t len_needed;
2779 size_t msg_len;
2780 mblk_t *mp1;
2781 ipaddr_t src;
2782 ire_t *ire;
2783 ip_xmit_attr_t ixas;
2784 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2785
2786 ipha = (ipha_t *)mp->b_rptr;
2787
2788 bzero(&ixas, sizeof (ixas));
2789 ixas.ixa_flags = IXAF_BASIC_SIMPLE_V4;
2790 ixas.ixa_zoneid = ira->ira_zoneid;
2791 ixas.ixa_ifindex = 0;
2792 ixas.ixa_ipst = ipst;
2793 ixas.ixa_cred = kcred;
2794 ixas.ixa_cpid = NOPID;
2795 ixas.ixa_tsl = ira->ira_tsl; /* Behave as a multi-level responder */
2796 ixas.ixa_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
2797
2798 if (ira->ira_flags & IRAF_IPSEC_SECURE) {
2799 /*
2800 * Apply IPsec based on how IPsec was applied to
2801 * the packet that had the error.
2802 *
2803 * If it was an outbound packet that caused the ICMP
2804 * error, then the caller will have setup the IRA
2805 * appropriately.
2806 */
2807 if (!ipsec_in_to_out(ira, &ixas, mp, ipha, NULL)) {
2808 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
2809 /* Note: mp already consumed and ip_drop_packet done */
2810 return;
2811 }
2812 } else {
2813 /*
2814 * This is in clear. The icmp message we are building
2815 * here should go out in clear, independent of our policy.
2816 */
2817 ixas.ixa_flags |= IXAF_NO_IPSEC;
2818 }
2819
2820 /* Remember our eventual destination */
2821 dst = ipha->ipha_src;
2822
2823 /*
2824 * If the packet was for one of our unicast addresses, make
2825 * sure we respond with that as the source. Otherwise
2826 * have ip_output_simple pick the source address.
2827 */
2828 ire = ire_ftable_lookup_v4(ipha->ipha_dst, 0, 0,
2829 (IRE_LOCAL|IRE_LOOPBACK), NULL, ira->ira_zoneid, NULL,
2830 MATCH_IRE_TYPE|MATCH_IRE_ZONEONLY, 0, ipst, NULL);
2831 if (ire != NULL) {
2832 ire_refrele(ire);
2833 src = ipha->ipha_dst;
2834 } else {
2835 src = INADDR_ANY;
2836 ixas.ixa_flags |= IXAF_SET_SOURCE;
2837 }
2838
2839 /*
2840 * Check if we can send back more then 8 bytes in addition to
2841 * the IP header. We try to send 64 bytes of data and the internal
2842 * header in the special cases of ipv4 encapsulated ipv4 or ipv6.
2843 */
2844 len_needed = IPH_HDR_LENGTH(ipha);
2845 if (ipha->ipha_protocol == IPPROTO_ENCAP ||
2846 ipha->ipha_protocol == IPPROTO_IPV6) {
2847 if (!pullupmsg(mp, -1)) {
2848 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
2849 ip_drop_output("ipIfStatsOutDiscards", mp, NULL);
2850 freemsg(mp);
2851 return;
2852 }
2853 ipha = (ipha_t *)mp->b_rptr;
2854
2855 if (ipha->ipha_protocol == IPPROTO_ENCAP) {
2856 len_needed += IPH_HDR_LENGTH(((uchar_t *)ipha +
2857 len_needed));
2858 } else {
2859 ip6_t *ip6h = (ip6_t *)((uchar_t *)ipha + len_needed);
2860
2861 ASSERT(ipha->ipha_protocol == IPPROTO_IPV6);
2862 len_needed += ip_hdr_length_v6(mp, ip6h);
2863 }
2864 }
2865 len_needed += ipst->ips_ip_icmp_return;
2866 msg_len = msgdsize(mp);
2867 if (msg_len > len_needed) {
2868 (void) adjmsg(mp, len_needed - msg_len);
2869 msg_len = len_needed;
2870 }
2871 mp1 = allocb(sizeof (icmp_ipha) + len, BPRI_MED);
2872 if (mp1 == NULL) {
2873 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutErrors);
2874 freemsg(mp);
2875 return;
2876 }
2877 mp1->b_cont = mp;
2878 mp = mp1;
2879
2880 /*
2881 * Set IXAF_TRUSTED_ICMP so we can let the ICMP messages this
2882 * node generates be accepted in peace by all on-host destinations.
2883 * If we do NOT assume that all on-host destinations trust
2884 * self-generated ICMP messages, then rework here, ip6.c, and spd.c.
2885 * (Look for IXAF_TRUSTED_ICMP).
2886 */
2887 ixas.ixa_flags |= IXAF_TRUSTED_ICMP;
2888
2889 ipha = (ipha_t *)mp->b_rptr;
2890 mp1->b_wptr = (uchar_t *)ipha + (sizeof (icmp_ipha) + len);
2891 *ipha = icmp_ipha;
2892 ipha->ipha_src = src;
2893 ipha->ipha_dst = dst;
2894 ipha->ipha_ttl = ipst->ips_ip_def_ttl;
2895 msg_len += sizeof (icmp_ipha) + len;
2896 if (msg_len > IP_MAXPACKET) {
2897 (void) adjmsg(mp, IP_MAXPACKET - msg_len);
2898 msg_len = IP_MAXPACKET;
2899 }
2900 ipha->ipha_length = htons((uint16_t)msg_len);
2901 icmph = (icmph_t *)&ipha[1];
2902 bcopy(stuff, icmph, len);
2903 icmph->icmph_checksum = 0;
2904 icmph->icmph_checksum = IP_CSUM(mp, (int32_t)sizeof (ipha_t), 0);
2905 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutMsgs);
2906
2907 (void) ip_output_simple(mp, &ixas);
2908 ixa_cleanup(&ixas);
2909 }
2910
2911 /*
2912 * Determine if an ICMP error packet can be sent given the rate limit.
2913 * The limit consists of an average frequency (icmp_pkt_err_interval measured
2914 * in milliseconds) and a burst size. Burst size number of packets can
2915 * be sent arbitrarely closely spaced.
2916 * The state is tracked using two variables to implement an approximate
2917 * token bucket filter:
2918 * icmp_pkt_err_last - lbolt value when the last burst started
2919 * icmp_pkt_err_sent - number of packets sent in current burst
2920 */
2921 boolean_t
2922 icmp_err_rate_limit(ip_stack_t *ipst)
2923 {
2924 clock_t now = TICK_TO_MSEC(ddi_get_lbolt());
2925 uint_t refilled; /* Number of packets refilled in tbf since last */
2926 /* Guard against changes by loading into local variable */
2927 uint_t err_interval = ipst->ips_ip_icmp_err_interval;
2928
2929 if (err_interval == 0)
2930 return (B_FALSE);
2931
2932 if (ipst->ips_icmp_pkt_err_last > now) {
2933 /* 100HZ lbolt in ms for 32bit arch wraps every 49.7 days */
2934 ipst->ips_icmp_pkt_err_last = 0;
2935 ipst->ips_icmp_pkt_err_sent = 0;
2936 }
2937 /*
2938 * If we are in a burst update the token bucket filter.
2939 * Update the "last" time to be close to "now" but make sure
2940 * we don't loose precision.
2941 */
2942 if (ipst->ips_icmp_pkt_err_sent != 0) {
2943 refilled = (now - ipst->ips_icmp_pkt_err_last)/err_interval;
2944 if (refilled > ipst->ips_icmp_pkt_err_sent) {
2945 ipst->ips_icmp_pkt_err_sent = 0;
2946 } else {
2947 ipst->ips_icmp_pkt_err_sent -= refilled;
2948 ipst->ips_icmp_pkt_err_last += refilled * err_interval;
2949 }
2950 }
2951 if (ipst->ips_icmp_pkt_err_sent == 0) {
2952 /* Start of new burst */
2953 ipst->ips_icmp_pkt_err_last = now;
2954 }
2955 if (ipst->ips_icmp_pkt_err_sent < ipst->ips_ip_icmp_err_burst) {
2956 ipst->ips_icmp_pkt_err_sent++;
2957 ip1dbg(("icmp_err_rate_limit: %d sent in burst\n",
2958 ipst->ips_icmp_pkt_err_sent));
2959 return (B_FALSE);
2960 }
2961 ip1dbg(("icmp_err_rate_limit: dropped\n"));
2962 return (B_TRUE);
2963 }
2964
2965 /*
2966 * Check if it is ok to send an IPv4 ICMP error packet in
2967 * response to the IPv4 packet in mp.
2968 * Free the message and return null if no
2969 * ICMP error packet should be sent.
2970 */
2971 static mblk_t *
2972 icmp_pkt_err_ok(mblk_t *mp, ip_recv_attr_t *ira)
2973 {
2974 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2975 icmph_t *icmph;
2976 ipha_t *ipha;
2977 uint_t len_needed;
2978
2979 if (!mp)
2980 return (NULL);
2981 ipha = (ipha_t *)mp->b_rptr;
2982 if (ip_csum_hdr(ipha)) {
2983 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsInCksumErrs);
2984 ip_drop_input("ipIfStatsInCksumErrs", mp, NULL);
2985 freemsg(mp);
2986 return (NULL);
2987 }
2988 if (ip_type_v4(ipha->ipha_dst, ipst) == IRE_BROADCAST ||
2989 ip_type_v4(ipha->ipha_src, ipst) == IRE_BROADCAST ||
2990 CLASSD(ipha->ipha_dst) ||
2991 CLASSD(ipha->ipha_src) ||
2992 (ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_OFFSET)) {
2993 /* Note: only errors to the fragment with offset 0 */
2994 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
2995 freemsg(mp);
2996 return (NULL);
2997 }
2998 if (ipha->ipha_protocol == IPPROTO_ICMP) {
2999 /*
3000 * Check the ICMP type. RFC 1122 sez: don't send ICMP
3001 * errors in response to any ICMP errors.
3002 */
3003 len_needed = IPH_HDR_LENGTH(ipha) + ICMPH_SIZE;
3004 if (mp->b_wptr - mp->b_rptr < len_needed) {
3005 if (!pullupmsg(mp, len_needed)) {
3006 BUMP_MIB(&ipst->ips_icmp_mib, icmpInErrors);
3007 freemsg(mp);
3008 return (NULL);
3009 }
3010 ipha = (ipha_t *)mp->b_rptr;
3011 }
3012 icmph = (icmph_t *)
3013 (&((char *)ipha)[IPH_HDR_LENGTH(ipha)]);
3014 switch (icmph->icmph_type) {
3015 case ICMP_DEST_UNREACHABLE:
3016 case ICMP_SOURCE_QUENCH:
3017 case ICMP_TIME_EXCEEDED:
3018 case ICMP_PARAM_PROBLEM:
3019 case ICMP_REDIRECT:
3020 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
3021 freemsg(mp);
3022 return (NULL);
3023 default:
3024 break;
3025 }
3026 }
3027 /*
3028 * If this is a labeled system, then check to see if we're allowed to
3029 * send a response to this particular sender. If not, then just drop.
3030 */
3031 if (is_system_labeled() && !tsol_can_reply_error(mp, ira)) {
3032 ip2dbg(("icmp_pkt_err_ok: can't respond to packet\n"));
3033 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
3034 freemsg(mp);
3035 return (NULL);
3036 }
3037 if (icmp_err_rate_limit(ipst)) {
3038 /*
3039 * Only send ICMP error packets every so often.
3040 * This should be done on a per port/source basis,
3041 * but for now this will suffice.
3042 */
3043 freemsg(mp);
3044 return (NULL);
3045 }
3046 return (mp);
3047 }
3048
3049 /*
3050 * Called when a packet was sent out the same link that it arrived on.
3051 * Check if it is ok to send a redirect and then send it.
3052 */
3053 void
3054 ip_send_potential_redirect_v4(mblk_t *mp, ipha_t *ipha, ire_t *ire,
3055 ip_recv_attr_t *ira)
3056 {
3057 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3058 ipaddr_t src, nhop;
3059 mblk_t *mp1;
3060 ire_t *nhop_ire;
3061
3062 /*
3063 * Check the source address to see if it originated
3064 * on the same logical subnet it is going back out on.
3065 * If so, we should be able to send it a redirect.
3066 * Avoid sending a redirect if the destination
3067 * is directly connected (i.e., we matched an IRE_ONLINK),
3068 * or if the packet was source routed out this interface.
3069 *
3070 * We avoid sending a redirect if the
3071 * destination is directly connected
3072 * because it is possible that multiple
3073 * IP subnets may have been configured on
3074 * the link, and the source may not
3075 * be on the same subnet as ip destination,
3076 * even though they are on the same
3077 * physical link.
3078 */
3079 if ((ire->ire_type & IRE_ONLINK) ||
3080 ip_source_routed(ipha, ipst))
3081 return;
3082
3083 nhop_ire = ire_nexthop(ire);
3084 if (nhop_ire == NULL)
3085 return;
3086
3087 nhop = nhop_ire->ire_addr;
3088
3089 if (nhop_ire->ire_type & IRE_IF_CLONE) {
3090 ire_t *ire2;
3091
3092 /* Follow ire_dep_parent to find non-clone IRE_INTERFACE */
3093 mutex_enter(&nhop_ire->ire_lock);
3094 ire2 = nhop_ire->ire_dep_parent;
3095 if (ire2 != NULL)
3096 ire_refhold(ire2);
3097 mutex_exit(&nhop_ire->ire_lock);
3098 ire_refrele(nhop_ire);
3099 nhop_ire = ire2;
3100 }
3101 if (nhop_ire == NULL)
3102 return;
3103
3104 ASSERT(!(nhop_ire->ire_type & IRE_IF_CLONE));
3105
3106 src = ipha->ipha_src;
3107
3108 /*
3109 * We look at the interface ire for the nexthop,
3110 * to see if ipha_src is in the same subnet
3111 * as the nexthop.
3112 */
3113 if ((src & nhop_ire->ire_mask) == (nhop & nhop_ire->ire_mask)) {
3114 /*
3115 * The source is directly connected.
3116 */
3117 mp1 = copymsg(mp);
3118 if (mp1 != NULL) {
3119 icmp_send_redirect(mp1, nhop, ira);
3120 }
3121 }
3122 ire_refrele(nhop_ire);
3123 }
3124
3125 /*
3126 * Generate an ICMP redirect message.
3127 */
3128 static void
3129 icmp_send_redirect(mblk_t *mp, ipaddr_t gateway, ip_recv_attr_t *ira)
3130 {
3131 icmph_t icmph;
3132 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3133
3134 mp = icmp_pkt_err_ok(mp, ira);
3135 if (mp == NULL)
3136 return;
3137
3138 bzero(&icmph, sizeof (icmph_t));
3139 icmph.icmph_type = ICMP_REDIRECT;
3140 icmph.icmph_code = 1;
3141 icmph.icmph_rd_gateway = gateway;
3142 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutRedirects);
3143 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3144 }
3145
3146 /*
3147 * Generate an ICMP time exceeded message.
3148 */
3149 void
3150 icmp_time_exceeded(mblk_t *mp, uint8_t code, ip_recv_attr_t *ira)
3151 {
3152 icmph_t icmph;
3153 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3154
3155 mp = icmp_pkt_err_ok(mp, ira);
3156 if (mp == NULL)
3157 return;
3158
3159 bzero(&icmph, sizeof (icmph_t));
3160 icmph.icmph_type = ICMP_TIME_EXCEEDED;
3161 icmph.icmph_code = code;
3162 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutTimeExcds);
3163 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3164 }
3165
3166 /*
3167 * Generate an ICMP unreachable message.
3168 * When called from ip_output side a minimal ip_recv_attr_t needs to be
3169 * constructed by the caller.
3170 */
3171 void
3172 icmp_unreachable(mblk_t *mp, uint8_t code, ip_recv_attr_t *ira)
3173 {
3174 icmph_t icmph;
3175 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3176
3177 mp = icmp_pkt_err_ok(mp, ira);
3178 if (mp == NULL)
3179 return;
3180
3181 bzero(&icmph, sizeof (icmph_t));
3182 icmph.icmph_type = ICMP_DEST_UNREACHABLE;
3183 icmph.icmph_code = code;
3184 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDestUnreachs);
3185 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3186 }
3187
3188 /*
3189 * Latch in the IPsec state for a stream based the policy in the listener
3190 * and the actions in the ip_recv_attr_t.
3191 * Called directly from TCP and SCTP.
3192 */
3193 boolean_t
3194 ip_ipsec_policy_inherit(conn_t *connp, conn_t *lconnp, ip_recv_attr_t *ira)
3195 {
3196 ASSERT(lconnp->conn_policy != NULL);
3197 ASSERT(connp->conn_policy == NULL);
3198
3199 IPPH_REFHOLD(lconnp->conn_policy);
3200 connp->conn_policy = lconnp->conn_policy;
3201
3202 if (ira->ira_ipsec_action != NULL) {
3203 if (connp->conn_latch == NULL) {
3204 connp->conn_latch = iplatch_create();
3205 if (connp->conn_latch == NULL)
3206 return (B_FALSE);
3207 }
3208 ipsec_latch_inbound(connp, ira);
3209 }
3210 return (B_TRUE);
3211 }
3212
3213 /*
3214 * Verify whether or not the IP address is a valid local address.
3215 * Could be a unicast, including one for a down interface.
3216 * If allow_mcbc then a multicast or broadcast address is also
3217 * acceptable.
3218 *
3219 * In the case of a broadcast/multicast address, however, the
3220 * upper protocol is expected to reset the src address
3221 * to zero when we return IPVL_MCAST/IPVL_BCAST so that
3222 * no packets are emitted with broadcast/multicast address as
3223 * source address (that violates hosts requirements RFC 1122)
3224 * The addresses valid for bind are:
3225 * (1) - INADDR_ANY (0)
3226 * (2) - IP address of an UP interface
3227 * (3) - IP address of a DOWN interface
3228 * (4) - valid local IP broadcast addresses. In this case
3229 * the conn will only receive packets destined to
3230 * the specified broadcast address.
3231 * (5) - a multicast address. In this case
3232 * the conn will only receive packets destined to
3233 * the specified multicast address. Note: the
3234 * application still has to issue an
3235 * IP_ADD_MEMBERSHIP socket option.
3236 *
3237 * In all the above cases, the bound address must be valid in the current zone.
3238 * When the address is loopback, multicast or broadcast, there might be many
3239 * matching IREs so bind has to look up based on the zone.
3240 */
3241 ip_laddr_t
3242 ip_laddr_verify_v4(ipaddr_t src_addr, zoneid_t zoneid,
3243 ip_stack_t *ipst, boolean_t allow_mcbc)
3244 {
3245 ire_t *src_ire;
3246
3247 ASSERT(src_addr != INADDR_ANY);
3248
3249 src_ire = ire_ftable_lookup_v4(src_addr, 0, 0, 0,
3250 NULL, zoneid, NULL, MATCH_IRE_ZONEONLY, 0, ipst, NULL);
3251
3252 /*
3253 * If an address other than in6addr_any is requested,
3254 * we verify that it is a valid address for bind
3255 * Note: Following code is in if-else-if form for
3256 * readability compared to a condition check.
3257 */
3258 if (src_ire != NULL && (src_ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK))) {
3259 /*
3260 * (2) Bind to address of local UP interface
3261 */
3262 ire_refrele(src_ire);
3263 return (IPVL_UNICAST_UP);
3264 } else if (src_ire != NULL && src_ire->ire_type & IRE_BROADCAST) {
3265 /*
3266 * (4) Bind to broadcast address
3267 */
3268 ire_refrele(src_ire);
3269 if (allow_mcbc)
3270 return (IPVL_BCAST);
3271 else
3272 return (IPVL_BAD);
3273 } else if (CLASSD(src_addr)) {
3274 /* (5) bind to multicast address. */
3275 if (src_ire != NULL)
3276 ire_refrele(src_ire);
3277
3278 if (allow_mcbc)
3279 return (IPVL_MCAST);
3280 else
3281 return (IPVL_BAD);
3282 } else {
3283 ipif_t *ipif;
3284
3285 /*
3286 * (3) Bind to address of local DOWN interface?
3287 * (ipif_lookup_addr() looks up all interfaces
3288 * but we do not get here for UP interfaces
3289 * - case (2) above)
3290 */
3291 if (src_ire != NULL)
3292 ire_refrele(src_ire);
3293
3294 ipif = ipif_lookup_addr(src_addr, NULL, zoneid, ipst);
3295 if (ipif == NULL)
3296 return (IPVL_BAD);
3297
3298 /* Not a useful source? */
3299 if (ipif->ipif_flags & (IPIF_NOLOCAL | IPIF_ANYCAST)) {
3300 ipif_refrele(ipif);
3301 return (IPVL_BAD);
3302 }
3303 ipif_refrele(ipif);
3304 return (IPVL_UNICAST_DOWN);
3305 }
3306 }
3307
3308 /*
3309 * Insert in the bind fanout for IPv4 and IPv6.
3310 * The caller should already have used ip_laddr_verify_v*() before calling
3311 * this.
3312 */
3313 int
3314 ip_laddr_fanout_insert(conn_t *connp)
3315 {
3316 int error;
3317
3318 /*
3319 * Allow setting new policies. For example, disconnects result
3320 * in us being called. As we would have set conn_policy_cached
3321 * to B_TRUE before, we should set it to B_FALSE, so that policy
3322 * can change after the disconnect.
3323 */
3324 connp->conn_policy_cached = B_FALSE;
3325
3326 error = ipcl_bind_insert(connp);
3327 if (error != 0) {
3328 if (connp->conn_anon_port) {
3329 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
3330 connp->conn_mlp_type, connp->conn_proto,
3331 ntohs(connp->conn_lport), B_FALSE);
3332 }
3333 connp->conn_mlp_type = mlptSingle;
3334 }
3335 return (error);
3336 }
3337
3338 /*
3339 * Verify that both the source and destination addresses are valid. If
3340 * IPDF_VERIFY_DST is not set, then the destination address may be unreachable,
3341 * i.e. have no route to it. Protocols like TCP want to verify destination
3342 * reachability, while tunnels do not.
3343 *
3344 * Determine the route, the interface, and (optionally) the source address
3345 * to use to reach a given destination.
3346 * Note that we allow connect to broadcast and multicast addresses when
3347 * IPDF_ALLOW_MCBC is set.
3348 * first_hop and dst_addr are normally the same, but if source routing
3349 * they will differ; in that case the first_hop is what we'll use for the
3350 * routing lookup but the dce and label checks will be done on dst_addr,
3351 *
3352 * If uinfo is set, then we fill in the best available information
3353 * we have for the destination. This is based on (in priority order) any
3354 * metrics and path MTU stored in a dce_t, route metrics, and finally the
3355 * ill_mtu/ill_mc_mtu.
3356 *
3357 * Tsol note: If we have a source route then dst_addr != firsthop. But we
3358 * always do the label check on dst_addr.
3359 */
3360 int
3361 ip_set_destination_v4(ipaddr_t *src_addrp, ipaddr_t dst_addr, ipaddr_t firsthop,
3362 ip_xmit_attr_t *ixa, iulp_t *uinfo, uint32_t flags, uint_t mac_mode)
3363 {
3364 ire_t *ire = NULL;
3365 int error = 0;
3366 ipaddr_t setsrc; /* RTF_SETSRC */
3367 zoneid_t zoneid = ixa->ixa_zoneid; /* Honors SO_ALLZONES */
3368 ip_stack_t *ipst = ixa->ixa_ipst;
3369 dce_t *dce;
3370 uint_t pmtu;
3371 uint_t generation;
3372 nce_t *nce;
3373 ill_t *ill = NULL;
3374 boolean_t multirt = B_FALSE;
3375
3376 ASSERT(ixa->ixa_flags & IXAF_IS_IPV4);
3377
3378 /*
3379 * We never send to zero; the ULPs map it to the loopback address.
3380 * We can't allow it since we use zero to mean unitialized in some
3381 * places.
3382 */
3383 ASSERT(dst_addr != INADDR_ANY);
3384
3385 if (is_system_labeled()) {
3386 ts_label_t *tsl = NULL;
3387
3388 error = tsol_check_dest(ixa->ixa_tsl, &dst_addr, IPV4_VERSION,
3389 mac_mode, (flags & IPDF_ZONE_IS_GLOBAL) != 0, &tsl);
3390 if (error != 0)
3391 return (error);
3392 if (tsl != NULL) {
3393 /* Update the label */
3394 ip_xmit_attr_replace_tsl(ixa, tsl);
3395 }
3396 }
3397
3398 setsrc = INADDR_ANY;
3399 /*
3400 * Select a route; For IPMP interfaces, we would only select
3401 * a "hidden" route (i.e., going through a specific under_ill)
3402 * if ixa_ifindex has been specified.
3403 */
3404 ire = ip_select_route_v4(firsthop, *src_addrp, ixa,
3405 &generation, &setsrc, &error, &multirt);
3406 ASSERT(ire != NULL); /* IRE_NOROUTE if none found */
3407 if (error != 0)
3408 goto bad_addr;
3409
3410 /*
3411 * ire can't be a broadcast or multicast unless IPDF_ALLOW_MCBC is set.
3412 * If IPDF_VERIFY_DST is set, the destination must be reachable;
3413 * Otherwise the destination needn't be reachable.
3414 *
3415 * If we match on a reject or black hole, then we've got a
3416 * local failure. May as well fail out the connect() attempt,
3417 * since it's never going to succeed.
3418 */
3419 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
3420 /*
3421 * If we're verifying destination reachability, we always want
3422 * to complain here.
3423 *
3424 * If we're not verifying destination reachability but the
3425 * destination has a route, we still want to fail on the
3426 * temporary address and broadcast address tests.
3427 *
3428 * In both cases do we let the code continue so some reasonable
3429 * information is returned to the caller. That enables the
3430 * caller to use (and even cache) the IRE. conn_ip_ouput will
3431 * use the generation mismatch path to check for the unreachable
3432 * case thereby avoiding any specific check in the main path.
3433 */
3434 ASSERT(generation == IRE_GENERATION_VERIFY);
3435 if (flags & IPDF_VERIFY_DST) {
3436 /*
3437 * Set errno but continue to set up ixa_ire to be
3438 * the RTF_REJECT|RTF_BLACKHOLE IRE.
3439 * That allows callers to use ip_output to get an
3440 * ICMP error back.
3441 */
3442 if (!(ire->ire_type & IRE_HOST))
3443 error = ENETUNREACH;
3444 else
3445 error = EHOSTUNREACH;
3446 }
3447 }
3448
3449 if ((ire->ire_type & (IRE_BROADCAST|IRE_MULTICAST)) &&
3450 !(flags & IPDF_ALLOW_MCBC)) {
3451 ire_refrele(ire);
3452 ire = ire_reject(ipst, B_FALSE);
3453 generation = IRE_GENERATION_VERIFY;
3454 error = ENETUNREACH;
3455 }
3456
3457 /* Cache things */
3458 if (ixa->ixa_ire != NULL)
3459 ire_refrele_notr(ixa->ixa_ire);
3460 #ifdef DEBUG
3461 ire_refhold_notr(ire);
3462 ire_refrele(ire);
3463 #endif
3464 ixa->ixa_ire = ire;
3465 ixa->ixa_ire_generation = generation;
3466
3467 /*
3468 * Ensure that ixa_dce is always set any time that ixa_ire is set,
3469 * since some callers will send a packet to conn_ip_output() even if
3470 * there's an error.
3471 */
3472 if (flags & IPDF_UNIQUE_DCE) {
3473 /* Fallback to the default dce if allocation fails */
3474 dce = dce_lookup_and_add_v4(dst_addr, ipst);
3475 if (dce != NULL)
3476 generation = dce->dce_generation;
3477 else
3478 dce = dce_lookup_v4(dst_addr, ipst, &generation);
3479 } else {
3480 dce = dce_lookup_v4(dst_addr, ipst, &generation);
3481 }
3482 ASSERT(dce != NULL);
3483 if (ixa->ixa_dce != NULL)
3484 dce_refrele_notr(ixa->ixa_dce);
3485 #ifdef DEBUG
3486 dce_refhold_notr(dce);
3487 dce_refrele(dce);
3488 #endif
3489 ixa->ixa_dce = dce;
3490 ixa->ixa_dce_generation = generation;
3491
3492 /*
3493 * For multicast with multirt we have a flag passed back from
3494 * ire_lookup_multi_ill_v4 since we don't have an IRE for each
3495 * possible multicast address.
3496 * We also need a flag for multicast since we can't check
3497 * whether RTF_MULTIRT is set in ixa_ire for multicast.
3498 */
3499 if (multirt) {
3500 ixa->ixa_postfragfn = ip_postfrag_multirt_v4;
3501 ixa->ixa_flags |= IXAF_MULTIRT_MULTICAST;
3502 } else {
3503 ixa->ixa_postfragfn = ire->ire_postfragfn;
3504 ixa->ixa_flags &= ~IXAF_MULTIRT_MULTICAST;
3505 }
3506 if (!(ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))) {
3507 /* Get an nce to cache. */
3508 nce = ire_to_nce(ire, firsthop, NULL);
3509 if (nce == NULL) {
3510 /* Allocation failure? */
3511 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3512 } else {
3513 if (ixa->ixa_nce != NULL)
3514 nce_refrele(ixa->ixa_nce);
3515 ixa->ixa_nce = nce;
3516 }
3517 }
3518
3519 /*
3520 * If the source address is a loopback address, the
3521 * destination had best be local or multicast.
3522 * If we are sending to an IRE_LOCAL using a loopback source then
3523 * it had better be the same zoneid.
3524 */
3525 if (*src_addrp == htonl(INADDR_LOOPBACK)) {
3526 if ((ire->ire_type & IRE_LOCAL) && ire->ire_zoneid != zoneid) {
3527 ire = NULL; /* Stored in ixa_ire */
3528 error = EADDRNOTAVAIL;
3529 goto bad_addr;
3530 }
3531 if (!(ire->ire_type & (IRE_LOOPBACK|IRE_LOCAL|IRE_MULTICAST))) {
3532 ire = NULL; /* Stored in ixa_ire */
3533 error = EADDRNOTAVAIL;
3534 goto bad_addr;
3535 }
3536 }
3537 if (ire->ire_type & IRE_BROADCAST) {
3538 /*
3539 * If the ULP didn't have a specified source, then we
3540 * make sure we reselect the source when sending
3541 * broadcasts out different interfaces.
3542 */
3543 if (flags & IPDF_SELECT_SRC)
3544 ixa->ixa_flags |= IXAF_SET_SOURCE;
3545 else
3546 ixa->ixa_flags &= ~IXAF_SET_SOURCE;
3547 }
3548
3549 /*
3550 * Does the caller want us to pick a source address?
3551 */
3552 if (flags & IPDF_SELECT_SRC) {
3553 ipaddr_t src_addr;
3554
3555 /*
3556 * We use use ire_nexthop_ill to avoid the under ipmp
3557 * interface for source address selection. Note that for ipmp
3558 * probe packets, ixa_ifindex would have been specified, and
3559 * the ip_select_route() invocation would have picked an ire
3560 * will ire_ill pointing at an under interface.
3561 */
3562 ill = ire_nexthop_ill(ire);
3563
3564 /* If unreachable we have no ill but need some source */
3565 if (ill == NULL) {
3566 src_addr = htonl(INADDR_LOOPBACK);
3567 /* Make sure we look for a better source address */
3568 generation = SRC_GENERATION_VERIFY;
3569 } else {
3570 error = ip_select_source_v4(ill, setsrc, dst_addr,
3571 ixa->ixa_multicast_ifaddr, zoneid,
3572 ipst, &src_addr, &generation, NULL);
3573 if (error != 0) {
3574 ire = NULL; /* Stored in ixa_ire */
3575 goto bad_addr;
3576 }
3577 }
3578
3579 /*
3580 * We allow the source address to to down.
3581 * However, we check that we don't use the loopback address
3582 * as a source when sending out on the wire.
3583 */
3584 if ((src_addr == htonl(INADDR_LOOPBACK)) &&
3585 !(ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK|IRE_MULTICAST)) &&
3586 !(ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))) {
3587 ire = NULL; /* Stored in ixa_ire */
3588 error = EADDRNOTAVAIL;
3589 goto bad_addr;
3590 }
3591
3592 *src_addrp = src_addr;
3593 ixa->ixa_src_generation = generation;
3594 }
3595
3596 /*
3597 * Make sure we don't leave an unreachable ixa_nce in place
3598 * since ip_select_route is used when we unplumb i.e., remove
3599 * references on ixa_ire, ixa_nce, and ixa_dce.
3600 */
3601 nce = ixa->ixa_nce;
3602 if (nce != NULL && nce->nce_is_condemned) {
3603 nce_refrele(nce);
3604 ixa->ixa_nce = NULL;
3605 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3606 }
3607
3608 /*
3609 * The caller has set IXAF_PMTU_DISCOVERY if path MTU is desired.
3610 * However, we can't do it for IPv4 multicast or broadcast.
3611 */
3612 if (ire->ire_type & (IRE_BROADCAST|IRE_MULTICAST))
3613 ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
3614
3615 /*
3616 * Set initial value for fragmentation limit. Either conn_ip_output
3617 * or ULP might updates it when there are routing changes.
3618 * Handles a NULL ixa_ire->ire_ill or a NULL ixa_nce for RTF_REJECT.
3619 */
3620 pmtu = ip_get_pmtu(ixa);
3621 ixa->ixa_fragsize = pmtu;
3622 /* Make sure ixa_fragsize and ixa_pmtu remain identical */
3623 if (ixa->ixa_flags & IXAF_VERIFY_PMTU)
3624 ixa->ixa_pmtu = pmtu;
3625
3626 /*
3627 * Extract information useful for some transports.
3628 * First we look for DCE metrics. Then we take what we have in
3629 * the metrics in the route, where the offlink is used if we have
3630 * one.
3631 */
3632 if (uinfo != NULL) {
3633 bzero(uinfo, sizeof (*uinfo));
3634
3635 if (dce->dce_flags & DCEF_UINFO)
3636 *uinfo = dce->dce_uinfo;
3637
3638 rts_merge_metrics(uinfo, &ire->ire_metrics);
3639
3640 /* Allow ire_metrics to decrease the path MTU from above */
3641 if (uinfo->iulp_mtu == 0 || uinfo->iulp_mtu > pmtu)
3642 uinfo->iulp_mtu = pmtu;
3643
3644 uinfo->iulp_localnet = (ire->ire_type & IRE_ONLINK) != 0;
3645 uinfo->iulp_loopback = (ire->ire_type & IRE_LOOPBACK) != 0;
3646 uinfo->iulp_local = (ire->ire_type & IRE_LOCAL) != 0;
3647 }
3648
3649 if (ill != NULL)
3650 ill_refrele(ill);
3651
3652 return (error);
3653
3654 bad_addr:
3655 if (ire != NULL)
3656 ire_refrele(ire);
3657
3658 if (ill != NULL)
3659 ill_refrele(ill);
3660
3661 /*
3662 * Make sure we don't leave an unreachable ixa_nce in place
3663 * since ip_select_route is used when we unplumb i.e., remove
3664 * references on ixa_ire, ixa_nce, and ixa_dce.
3665 */
3666 nce = ixa->ixa_nce;
3667 if (nce != NULL && nce->nce_is_condemned) {
3668 nce_refrele(nce);
3669 ixa->ixa_nce = NULL;
3670 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3671 }
3672
3673 return (error);
3674 }
3675
3676
3677 /*
3678 * Get the base MTU for the case when path MTU discovery is not used.
3679 * Takes the MTU of the IRE into account.
3680 */
3681 uint_t
3682 ip_get_base_mtu(ill_t *ill, ire_t *ire)
3683 {
3684 uint_t mtu;
3685 uint_t iremtu = ire->ire_metrics.iulp_mtu;
3686
3687 if (ire->ire_type & (IRE_MULTICAST|IRE_BROADCAST))
3688 mtu = ill->ill_mc_mtu;
3689 else
3690 mtu = ill->ill_mtu;
3691
3692 if (iremtu != 0 && iremtu < mtu)
3693 mtu = iremtu;
3694
3695 return (mtu);
3696 }
3697
3698 /*
3699 * Get the PMTU for the attributes. Handles both IPv4 and IPv6.
3700 * Assumes that ixa_ire, dce, and nce have already been set up.
3701 *
3702 * The caller has set IXAF_PMTU_DISCOVERY if path MTU discovery is desired.
3703 * We avoid path MTU discovery if it is disabled with ndd.
3704 * Furtermore, if the path MTU is too small, then we don't set DF for IPv4.
3705 *
3706 * NOTE: We also used to turn it off for source routed packets. That
3707 * is no longer required since the dce is per final destination.
3708 */
3709 uint_t
3710 ip_get_pmtu(ip_xmit_attr_t *ixa)
3711 {
3712 ip_stack_t *ipst = ixa->ixa_ipst;
3713 dce_t *dce;
3714 nce_t *nce;
3715 ire_t *ire;
3716 uint_t pmtu;
3717
3718 ire = ixa->ixa_ire;
3719 dce = ixa->ixa_dce;
3720 nce = ixa->ixa_nce;
3721
3722 /*
3723 * If path MTU discovery has been turned off by ndd, then we ignore
3724 * any dce_pmtu and for IPv4 we will not set DF.
3725 */
3726 if (!ipst->ips_ip_path_mtu_discovery)
3727 ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
3728
3729 pmtu = IP_MAXPACKET;
3730 /*
3731 * Decide whether whether IPv4 sets DF
3732 * For IPv6 "no DF" means to use the 1280 mtu
3733 */
3734 if (ixa->ixa_flags & IXAF_PMTU_DISCOVERY) {
3735 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3736 } else {
3737 ixa->ixa_flags &= ~IXAF_PMTU_IPV4_DF;
3738 if (!(ixa->ixa_flags & IXAF_IS_IPV4))
3739 pmtu = IPV6_MIN_MTU;
3740 }
3741
3742 /* Check if the PMTU is to old before we use it */
3743 if ((dce->dce_flags & DCEF_PMTU) &&
3744 TICK_TO_SEC(ddi_get_lbolt64()) - dce->dce_last_change_time >
3745 ipst->ips_ip_pathmtu_interval) {
3746 /*
3747 * Older than 20 minutes. Drop the path MTU information.
3748 */
3749 mutex_enter(&dce->dce_lock);
3750 dce->dce_flags &= ~(DCEF_PMTU|DCEF_TOO_SMALL_PMTU);
3751 dce->dce_last_change_time = TICK_TO_SEC(ddi_get_lbolt64());
3752 mutex_exit(&dce->dce_lock);
3753 dce_increment_generation(dce);
3754 }
3755
3756 /* The metrics on the route can lower the path MTU */
3757 if (ire->ire_metrics.iulp_mtu != 0 &&
3758 ire->ire_metrics.iulp_mtu < pmtu)
3759 pmtu = ire->ire_metrics.iulp_mtu;
3760
3761 /*
3762 * If the path MTU is smaller than some minimum, we still use dce_pmtu
3763 * above (would be 576 for IPv4 and 1280 for IPv6), but we clear
3764 * IXAF_PMTU_IPV4_DF so that we avoid setting DF for IPv4.
3765 */
3766 if (ixa->ixa_flags & IXAF_PMTU_DISCOVERY) {
3767 if (dce->dce_flags & DCEF_PMTU) {
3768 if (dce->dce_pmtu < pmtu)
3769 pmtu = dce->dce_pmtu;
3770
3771 if (dce->dce_flags & DCEF_TOO_SMALL_PMTU) {
3772 ixa->ixa_flags |= IXAF_PMTU_TOO_SMALL;
3773 ixa->ixa_flags &= ~IXAF_PMTU_IPV4_DF;
3774 } else {
3775 ixa->ixa_flags &= ~IXAF_PMTU_TOO_SMALL;
3776 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3777 }
3778 } else {
3779 ixa->ixa_flags &= ~IXAF_PMTU_TOO_SMALL;
3780 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3781 }
3782 }
3783
3784 /*
3785 * If we have an IRE_LOCAL we use the loopback mtu instead of
3786 * the ill for going out the wire i.e., IRE_LOCAL gets the same
3787 * mtu as IRE_LOOPBACK.
3788 */
3789 if (ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK)) {
3790 uint_t loopback_mtu;
3791
3792 loopback_mtu = (ire->ire_ipversion == IPV6_VERSION) ?
3793 ip_loopback_mtu_v6plus : ip_loopback_mtuplus;
3794
3795 if (loopback_mtu < pmtu)
3796 pmtu = loopback_mtu;
3797 } else if (nce != NULL) {
3798 /*
3799 * Make sure we don't exceed the interface MTU.
3800 * In the case of RTF_REJECT or RTF_BLACKHOLE we might not have
3801 * an ill. We'd use the above IP_MAXPACKET in that case just
3802 * to tell the transport something larger than zero.
3803 */
3804 if (ire->ire_type & (IRE_MULTICAST|IRE_BROADCAST)) {
3805 if (nce->nce_common->ncec_ill->ill_mc_mtu < pmtu)
3806 pmtu = nce->nce_common->ncec_ill->ill_mc_mtu;
3807 if (nce->nce_common->ncec_ill != nce->nce_ill &&
3808 nce->nce_ill->ill_mc_mtu < pmtu) {
3809 /*
3810 * for interfaces in an IPMP group, the mtu of
3811 * the nce_ill (under_ill) could be different
3812 * from the mtu of the ncec_ill, so we take the
3813 * min of the two.
3814 */
3815 pmtu = nce->nce_ill->ill_mc_mtu;
3816 }
3817 } else {
3818 if (nce->nce_common->ncec_ill->ill_mtu < pmtu)
3819 pmtu = nce->nce_common->ncec_ill->ill_mtu;
3820 if (nce->nce_common->ncec_ill != nce->nce_ill &&
3821 nce->nce_ill->ill_mtu < pmtu) {
3822 /*
3823 * for interfaces in an IPMP group, the mtu of
3824 * the nce_ill (under_ill) could be different
3825 * from the mtu of the ncec_ill, so we take the
3826 * min of the two.
3827 */
3828 pmtu = nce->nce_ill->ill_mtu;
3829 }
3830 }
3831 }
3832
3833 /*
3834 * Handle the IPV6_USE_MIN_MTU socket option or ancillary data.
3835 * Only applies to IPv6.
3836 */
3837 if (!(ixa->ixa_flags & IXAF_IS_IPV4)) {
3838 if (ixa->ixa_flags & IXAF_USE_MIN_MTU) {
3839 switch (ixa->ixa_use_min_mtu) {
3840 case IPV6_USE_MIN_MTU_MULTICAST:
3841 if (ire->ire_type & IRE_MULTICAST)
3842 pmtu = IPV6_MIN_MTU;
3843 break;
3844 case IPV6_USE_MIN_MTU_ALWAYS:
3845 pmtu = IPV6_MIN_MTU;
3846 break;
3847 case IPV6_USE_MIN_MTU_NEVER:
3848 break;
3849 }
3850 } else {
3851 /* Default is IPV6_USE_MIN_MTU_MULTICAST */
3852 if (ire->ire_type & IRE_MULTICAST)
3853 pmtu = IPV6_MIN_MTU;
3854 }
3855 }
3856
3857 /*
3858 * After receiving an ICMPv6 "packet too big" message with a
3859 * MTU < 1280, and for multirouted IPv6 packets, the IP layer
3860 * will insert a 8-byte fragment header in every packet. We compensate
3861 * for those cases by returning a smaller path MTU to the ULP.
3862 *
3863 * In the case of CGTP then ip_output will add a fragment header.
3864 * Make sure there is room for it by telling a smaller number
3865 * to the transport.
3866 *
3867 * When IXAF_IPV6_ADDR_FRAGHDR we subtract the frag hdr here
3868 * so the ULPs consistently see a iulp_pmtu and ip_get_pmtu()
3869 * which is the size of the packets it can send.
3870 */
3871 if (!(ixa->ixa_flags & IXAF_IS_IPV4)) {
3872 if ((dce->dce_flags & DCEF_TOO_SMALL_PMTU) ||
3873 (ire->ire_flags & RTF_MULTIRT) ||
3874 (ixa->ixa_flags & IXAF_MULTIRT_MULTICAST)) {
3875 pmtu -= sizeof (ip6_frag_t);
3876 ixa->ixa_flags |= IXAF_IPV6_ADD_FRAGHDR;
3877 }
3878 }
3879
3880 return (pmtu);
3881 }
3882
3883 /*
3884 * Carve "len" bytes out of an mblk chain, consuming any we empty, and duping
3885 * the final piece where we don't. Return a pointer to the first mblk in the
3886 * result, and update the pointer to the next mblk to chew on. If anything
3887 * goes wrong (i.e., dupb fails), we waste everything in sight and return a
3888 * NULL pointer.
3889 */
3890 mblk_t *
3891 ip_carve_mp(mblk_t **mpp, ssize_t len)
3892 {
3893 mblk_t *mp0;
3894 mblk_t *mp1;
3895 mblk_t *mp2;
3896
3897 if (!len || !mpp || !(mp0 = *mpp))
3898 return (NULL);
3899 /* If we aren't going to consume the first mblk, we need a dup. */
3900 if (mp0->b_wptr - mp0->b_rptr > len) {
3901 mp1 = dupb(mp0);
3902 if (mp1) {
3903 /* Partition the data between the two mblks. */
3904 mp1->b_wptr = mp1->b_rptr + len;
3905 mp0->b_rptr = mp1->b_wptr;
3906 /*
3907 * after adjustments if mblk not consumed is now
3908 * unaligned, try to align it. If this fails free
3909 * all messages and let upper layer recover.
3910 */
3911 if (!OK_32PTR(mp0->b_rptr)) {
3912 if (!pullupmsg(mp0, -1)) {
3913 freemsg(mp0);
3914 freemsg(mp1);
3915 *mpp = NULL;
3916 return (NULL);
3917 }
3918 }
3919 }
3920 return (mp1);
3921 }
3922 /* Eat through as many mblks as we need to get len bytes. */
3923 len -= mp0->b_wptr - mp0->b_rptr;
3924 for (mp2 = mp1 = mp0; (mp2 = mp2->b_cont) != 0 && len; mp1 = mp2) {
3925 if (mp2->b_wptr - mp2->b_rptr > len) {
3926 /*
3927 * We won't consume the entire last mblk. Like
3928 * above, dup and partition it.
3929 */
3930 mp1->b_cont = dupb(mp2);
3931 mp1 = mp1->b_cont;
3932 if (!mp1) {
3933 /*
3934 * Trouble. Rather than go to a lot of
3935 * trouble to clean up, we free the messages.
3936 * This won't be any worse than losing it on
3937 * the wire.
3938 */
3939 freemsg(mp0);
3940 freemsg(mp2);
3941 *mpp = NULL;
3942 return (NULL);
3943 }
3944 mp1->b_wptr = mp1->b_rptr + len;
3945 mp2->b_rptr = mp1->b_wptr;
3946 /*
3947 * after adjustments if mblk not consumed is now
3948 * unaligned, try to align it. If this fails free
3949 * all messages and let upper layer recover.
3950 */
3951 if (!OK_32PTR(mp2->b_rptr)) {
3952 if (!pullupmsg(mp2, -1)) {
3953 freemsg(mp0);
3954 freemsg(mp2);
3955 *mpp = NULL;
3956 return (NULL);
3957 }
3958 }
3959 *mpp = mp2;
3960 return (mp0);
3961 }
3962 /* Decrement len by the amount we just got. */
3963 len -= mp2->b_wptr - mp2->b_rptr;
3964 }
3965 /*
3966 * len should be reduced to zero now. If not our caller has
3967 * screwed up.
3968 */
3969 if (len) {
3970 /* Shouldn't happen! */
3971 freemsg(mp0);
3972 *mpp = NULL;
3973 return (NULL);
3974 }
3975 /*
3976 * We consumed up to exactly the end of an mblk. Detach the part
3977 * we are returning from the rest of the chain.
3978 */
3979 mp1->b_cont = NULL;
3980 *mpp = mp2;
3981 return (mp0);
3982 }
3983
3984 /* The ill stream is being unplumbed. Called from ip_close */
3985 int
3986 ip_modclose(ill_t *ill)
3987 {
3988 boolean_t success;
3989 ipsq_t *ipsq;
3990 ipif_t *ipif;
3991 queue_t *q = ill->ill_rq;
3992 ip_stack_t *ipst = ill->ill_ipst;
3993 int i;
3994 arl_ill_common_t *ai = ill->ill_common;
3995
3996 /*
3997 * The punlink prior to this may have initiated a capability
3998 * negotiation. But ipsq_enter will block until that finishes or
3999 * times out.
4000 */
4001 success = ipsq_enter(ill, B_FALSE, NEW_OP);
4002
4003 /*
4004 * Open/close/push/pop is guaranteed to be single threaded
4005 * per stream by STREAMS. FS guarantees that all references
4006 * from top are gone before close is called. So there can't
4007 * be another close thread that has set CONDEMNED on this ill.
4008 * and cause ipsq_enter to return failure.
4009 */
4010 ASSERT(success);
4011 ipsq = ill->ill_phyint->phyint_ipsq;
4012
4013 /*
4014 * Mark it condemned. No new reference will be made to this ill.
4015 * Lookup functions will return an error. Threads that try to
4016 * increment the refcnt must check for ILL_CAN_LOOKUP. This ensures
4017 * that the refcnt will drop down to zero.
4018 */
4019 mutex_enter(&ill->ill_lock);
4020 ill->ill_state_flags |= ILL_CONDEMNED;
4021 for (ipif = ill->ill_ipif; ipif != NULL;
4022 ipif = ipif->ipif_next) {
4023 ipif->ipif_state_flags |= IPIF_CONDEMNED;
4024 }
4025 /*
4026 * Wake up anybody waiting to enter the ipsq. ipsq_enter
4027 * returns error if ILL_CONDEMNED is set
4028 */
4029 cv_broadcast(&ill->ill_cv);
4030 mutex_exit(&ill->ill_lock);
4031
4032 /*
4033 * Send all the deferred DLPI messages downstream which came in
4034 * during the small window right before ipsq_enter(). We do this
4035 * without waiting for the ACKs because all the ACKs for M_PROTO
4036 * messages are ignored in ip_rput() when ILL_CONDEMNED is set.
4037 */
4038 ill_dlpi_send_deferred(ill);
4039
4040 /*
4041 * Shut down fragmentation reassembly.
4042 * ill_frag_timer won't start a timer again.
4043 * Now cancel any existing timer
4044 */
4045 (void) untimeout(ill->ill_frag_timer_id);
4046 (void) ill_frag_timeout(ill, 0);
4047
4048 /*
4049 * Call ill_delete to bring down the ipifs, ilms and ill on
4050 * this ill. Then wait for the refcnts to drop to zero.
4051 * ill_is_freeable checks whether the ill is really quiescent.
4052 * Then make sure that threads that are waiting to enter the
4053 * ipsq have seen the error returned by ipsq_enter and have
4054 * gone away. Then we call ill_delete_tail which does the
4055 * DL_UNBIND_REQ with the driver and then qprocsoff.
4056 */
4057 ill_delete(ill);
4058 mutex_enter(&ill->ill_lock);
4059 while (!ill_is_freeable(ill))
4060 cv_wait(&ill->ill_cv, &ill->ill_lock);
4061
4062 while (ill->ill_waiters)
4063 cv_wait(&ill->ill_cv, &ill->ill_lock);
4064
4065 mutex_exit(&ill->ill_lock);
4066
4067 /*
4068 * ill_delete_tail drops reference on ill_ipst, but we need to keep
4069 * it held until the end of the function since the cleanup
4070 * below needs to be able to use the ip_stack_t.
4071 */
4072 netstack_hold(ipst->ips_netstack);
4073
4074 /* qprocsoff is done via ill_delete_tail */
4075 ill_delete_tail(ill);
4076 /*
4077 * synchronously wait for arp stream to unbind. After this, we
4078 * cannot get any data packets up from the driver.
4079 */
4080 arp_unbind_complete(ill);
4081 ASSERT(ill->ill_ipst == NULL);
4082
4083 /*
4084 * Walk through all conns and qenable those that have queued data.
4085 * Close synchronization needs this to
4086 * be done to ensure that all upper layers blocked
4087 * due to flow control to the closing device
4088 * get unblocked.
4089 */
4090 ip1dbg(("ip_wsrv: walking\n"));
4091 for (i = 0; i < TX_FANOUT_SIZE; i++) {
4092 conn_walk_drain(ipst, &ipst->ips_idl_tx_list[i]);
4093 }
4094
4095 /*
4096 * ai can be null if this is an IPv6 ill, or if the IPv4
4097 * stream is being torn down before ARP was plumbed (e.g.,
4098 * /sbin/ifconfig plumbing a stream twice, and encountering
4099 * an error
4100 */
4101 if (ai != NULL) {
4102 ASSERT(!ill->ill_isv6);
4103 mutex_enter(&ai->ai_lock);
4104 ai->ai_ill = NULL;
4105 if (ai->ai_arl == NULL) {
4106 mutex_destroy(&ai->ai_lock);
4107 kmem_free(ai, sizeof (*ai));
4108 } else {
4109 cv_signal(&ai->ai_ill_unplumb_done);
4110 mutex_exit(&ai->ai_lock);
4111 }
4112 }
4113
4114 mutex_enter(&ipst->ips_ip_mi_lock);
4115 mi_close_unlink(&ipst->ips_ip_g_head, (IDP)ill);
4116 mutex_exit(&ipst->ips_ip_mi_lock);
4117
4118 /*
4119 * credp could be null if the open didn't succeed and ip_modopen
4120 * itself calls ip_close.
4121 */
4122 if (ill->ill_credp != NULL)
4123 crfree(ill->ill_credp);
4124
4125 mutex_destroy(&ill->ill_saved_ire_lock);
4126 mutex_destroy(&ill->ill_lock);
4127 rw_destroy(&ill->ill_mcast_lock);
4128 mutex_destroy(&ill->ill_mcast_serializer);
4129 list_destroy(&ill->ill_nce);
4130
4131 /*
4132 * Now we are done with the module close pieces that
4133 * need the netstack_t.
4134 */
4135 netstack_rele(ipst->ips_netstack);
4136
4137 mi_close_free((IDP)ill);
4138 q->q_ptr = WR(q)->q_ptr = NULL;
4139
4140 ipsq_exit(ipsq);
4141
4142 return (0);
4143 }
4144
4145 /*
4146 * This is called as part of close() for IP, UDP, ICMP, and RTS
4147 * in order to quiesce the conn.
4148 */
4149 void
4150 ip_quiesce_conn(conn_t *connp)
4151 {
4152 boolean_t drain_cleanup_reqd = B_FALSE;
4153 boolean_t conn_ioctl_cleanup_reqd = B_FALSE;
4154 boolean_t ilg_cleanup_reqd = B_FALSE;
4155 ip_stack_t *ipst;
4156
4157 ASSERT(!IPCL_IS_TCP(connp));
4158 ipst = connp->conn_netstack->netstack_ip;
4159
4160 /*
4161 * Mark the conn as closing, and this conn must not be
4162 * inserted in future into any list. Eg. conn_drain_insert(),
4163 * won't insert this conn into the conn_drain_list.
4164 *
4165 * conn_idl, and conn_ilg cannot get set henceforth.
4166 */
4167 mutex_enter(&connp->conn_lock);
4168 ASSERT(!(connp->conn_state_flags & CONN_QUIESCED));
4169 connp->conn_state_flags |= CONN_CLOSING;
4170 if (connp->conn_idl != NULL)
4171 drain_cleanup_reqd = B_TRUE;
4172 if (connp->conn_oper_pending_ill != NULL)
4173 conn_ioctl_cleanup_reqd = B_TRUE;
4174 if (connp->conn_dhcpinit_ill != NULL) {
4175 ASSERT(connp->conn_dhcpinit_ill->ill_dhcpinit != 0);
4176 atomic_dec_32(&connp->conn_dhcpinit_ill->ill_dhcpinit);
4177 ill_set_inputfn(connp->conn_dhcpinit_ill);
4178 connp->conn_dhcpinit_ill = NULL;
4179 }
4180 if (connp->conn_ilg != NULL)
4181 ilg_cleanup_reqd = B_TRUE;
4182 mutex_exit(&connp->conn_lock);
4183
4184 if (conn_ioctl_cleanup_reqd)
4185 conn_ioctl_cleanup(connp);
4186
4187 if (is_system_labeled() && connp->conn_anon_port) {
4188 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
4189 connp->conn_mlp_type, connp->conn_proto,
4190 ntohs(connp->conn_lport), B_FALSE);
4191 connp->conn_anon_port = 0;
4192 }
4193 connp->conn_mlp_type = mlptSingle;
4194
4195 /*
4196 * Remove this conn from any fanout list it is on.
4197 * and then wait for any threads currently operating
4198 * on this endpoint to finish
4199 */
4200 ipcl_hash_remove(connp);
4201
4202 /*
4203 * Remove this conn from the drain list, and do any other cleanup that
4204 * may be required. (TCP conns are never flow controlled, and
4205 * conn_idl will be NULL.)
4206 */
4207 if (drain_cleanup_reqd && connp->conn_idl != NULL) {
4208 idl_t *idl = connp->conn_idl;
4209
4210 mutex_enter(&idl->idl_lock);
4211 conn_drain(connp, B_TRUE);
4212 mutex_exit(&idl->idl_lock);
4213 }
4214
4215 if (connp == ipst->ips_ip_g_mrouter)
4216 (void) ip_mrouter_done(ipst);
4217
4218 if (ilg_cleanup_reqd)
4219 ilg_delete_all(connp);
4220
4221 /*
4222 * Now conn refcnt can increase only thru CONN_INC_REF_LOCKED.
4223 * callers from write side can't be there now because close
4224 * is in progress. The only other caller is ipcl_walk
4225 * which checks for the condemned flag.
4226 */
4227 mutex_enter(&connp->conn_lock);
4228 connp->conn_state_flags |= CONN_CONDEMNED;
4229 while (connp->conn_ref != 1)
4230 cv_wait(&connp->conn_cv, &connp->conn_lock);
4231 connp->conn_state_flags |= CONN_QUIESCED;
4232 mutex_exit(&connp->conn_lock);
4233 }
4234
4235 /* ARGSUSED */
4236 int
4237 ip_close(queue_t *q, int flags)
4238 {
4239 conn_t *connp;
4240
4241 /*
4242 * Call the appropriate delete routine depending on whether this is
4243 * a module or device.
4244 */
4245 if (WR(q)->q_next != NULL) {
4246 /* This is a module close */
4247 return (ip_modclose((ill_t *)q->q_ptr));
4248 }
4249
4250 connp = q->q_ptr;
4251 ip_quiesce_conn(connp);
4252
4253 qprocsoff(q);
4254
4255 /*
4256 * Now we are truly single threaded on this stream, and can
4257 * delete the things hanging off the connp, and finally the connp.
4258 * We removed this connp from the fanout list, it cannot be
4259 * accessed thru the fanouts, and we already waited for the
4260 * conn_ref to drop to 0. We are already in close, so
4261 * there cannot be any other thread from the top. qprocsoff
4262 * has completed, and service has completed or won't run in
4263 * future.
4264 */
4265 ASSERT(connp->conn_ref == 1);
4266
4267 inet_minor_free(connp->conn_minor_arena, connp->conn_dev);
4268
4269 connp->conn_ref--;
4270 ipcl_conn_destroy(connp);
4271
4272 q->q_ptr = WR(q)->q_ptr = NULL;
4273 return (0);
4274 }
4275
4276 /*
4277 * Wapper around putnext() so that ip_rts_request can merely use
4278 * conn_recv.
4279 */
4280 /*ARGSUSED2*/
4281 static void
4282 ip_conn_input(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira)
4283 {
4284 conn_t *connp = (conn_t *)arg1;
4285
4286 putnext(connp->conn_rq, mp);
4287 }
4288
4289 /* Dummy in case ICMP error delivery is attempted to a /dev/ip instance */
4290 /* ARGSUSED */
4291 static void
4292 ip_conn_input_icmp(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira)
4293 {
4294 freemsg(mp);
4295 }
4296
4297 /*
4298 * Called when the module is about to be unloaded
4299 */
4300 void
4301 ip_ddi_destroy(void)
4302 {
4303 /* This needs to be called before destroying any transports. */
4304 mutex_enter(&cpu_lock);
4305 unregister_cpu_setup_func(ip_tp_cpu_update, NULL);
4306 mutex_exit(&cpu_lock);
4307
4308 tnet_fini();
4309
4310 icmp_ddi_g_destroy();
4311 rts_ddi_g_destroy();
4312 udp_ddi_g_destroy();
4313 sctp_ddi_g_destroy();
4314 tcp_ddi_g_destroy();
4315 ilb_ddi_g_destroy();
4316 dce_g_destroy();
4317 ipsec_policy_g_destroy();
4318 ipcl_g_destroy();
4319 ip_net_g_destroy();
4320 ip_ire_g_fini();
4321 inet_minor_destroy(ip_minor_arena_sa);
4322 #if defined(_LP64)
4323 inet_minor_destroy(ip_minor_arena_la);
4324 #endif
4325
4326 #ifdef DEBUG
4327 list_destroy(&ip_thread_list);
4328 rw_destroy(&ip_thread_rwlock);
4329 tsd_destroy(&ip_thread_data);
4330 #endif
4331
4332 netstack_unregister(NS_IP);
4333 }
4334
4335 /*
4336 * First step in cleanup.
4337 */
4338 /* ARGSUSED */
4339 static void
4340 ip_stack_shutdown(netstackid_t stackid, void *arg)
4341 {
4342 ip_stack_t *ipst = (ip_stack_t *)arg;
4343
4344 #ifdef NS_DEBUG
4345 printf("ip_stack_shutdown(%p, stack %d)\n", (void *)ipst, stackid);
4346 #endif
4347
4348 /*
4349 * Perform cleanup for special interfaces (loopback and IPMP).
4350 */
4351 ip_interface_cleanup(ipst);
4352
4353 /*
4354 * The *_hook_shutdown()s start the process of notifying any
4355 * consumers that things are going away.... nothing is destroyed.
4356 */
4357 ipv4_hook_shutdown(ipst);
4358 ipv6_hook_shutdown(ipst);
4359 arp_hook_shutdown(ipst);
4360
4361 mutex_enter(&ipst->ips_capab_taskq_lock);
4362 ipst->ips_capab_taskq_quit = B_TRUE;
4363 cv_signal(&ipst->ips_capab_taskq_cv);
4364 mutex_exit(&ipst->ips_capab_taskq_lock);
4365 }
4366
4367 /*
4368 * Free the IP stack instance.
4369 */
4370 static void
4371 ip_stack_fini(netstackid_t stackid, void *arg)
4372 {
4373 ip_stack_t *ipst = (ip_stack_t *)arg;
4374 int ret;
4375
4376 #ifdef NS_DEBUG
4377 printf("ip_stack_fini(%p, stack %d)\n", (void *)ipst, stackid);
4378 #endif
4379 /*
4380 * At this point, all of the notifications that the events and
4381 * protocols are going away have been run, meaning that we can
4382 * now set about starting to clean things up.
4383 */
4384 ipobs_fini(ipst);
4385 ipv4_hook_destroy(ipst);
4386 ipv6_hook_destroy(ipst);
4387 arp_hook_destroy(ipst);
4388 ip_net_destroy(ipst);
4389
4390 ipmp_destroy(ipst);
4391
4392 ip_kstat_fini(stackid, ipst->ips_ip_mibkp);
4393 ipst->ips_ip_mibkp = NULL;
4394 icmp_kstat_fini(stackid, ipst->ips_icmp_mibkp);
4395 ipst->ips_icmp_mibkp = NULL;
4396 ip_kstat2_fini(stackid, ipst->ips_ip_kstat);
4397 ipst->ips_ip_kstat = NULL;
4398 bzero(&ipst->ips_ip_statistics, sizeof (ipst->ips_ip_statistics));
4399 ip6_kstat_fini(stackid, ipst->ips_ip6_kstat);
4400 ipst->ips_ip6_kstat = NULL;
4401 bzero(&ipst->ips_ip6_statistics, sizeof (ipst->ips_ip6_statistics));
4402
4403 kmem_free(ipst->ips_propinfo_tbl,
4404 ip_propinfo_count * sizeof (mod_prop_info_t));
4405 ipst->ips_propinfo_tbl = NULL;
4406
4407 dce_stack_destroy(ipst);
4408 ip_mrouter_stack_destroy(ipst);
4409
4410 ret = untimeout(ipst->ips_igmp_timeout_id);
4411 if (ret == -1) {
4412 ASSERT(ipst->ips_igmp_timeout_id == 0);
4413 } else {
4414 ASSERT(ipst->ips_igmp_timeout_id != 0);
4415 ipst->ips_igmp_timeout_id = 0;
4416 }
4417 ret = untimeout(ipst->ips_igmp_slowtimeout_id);
4418 if (ret == -1) {
4419 ASSERT(ipst->ips_igmp_slowtimeout_id == 0);
4420 } else {
4421 ASSERT(ipst->ips_igmp_slowtimeout_id != 0);
4422 ipst->ips_igmp_slowtimeout_id = 0;
4423 }
4424 ret = untimeout(ipst->ips_mld_timeout_id);
4425 if (ret == -1) {
4426 ASSERT(ipst->ips_mld_timeout_id == 0);
4427 } else {
4428 ASSERT(ipst->ips_mld_timeout_id != 0);
4429 ipst->ips_mld_timeout_id = 0;
4430 }
4431 ret = untimeout(ipst->ips_mld_slowtimeout_id);
4432 if (ret == -1) {
4433 ASSERT(ipst->ips_mld_slowtimeout_id == 0);
4434 } else {
4435 ASSERT(ipst->ips_mld_slowtimeout_id != 0);
4436 ipst->ips_mld_slowtimeout_id = 0;
4437 }
4438
4439 ip_ire_fini(ipst);
4440 ip6_asp_free(ipst);
4441 conn_drain_fini(ipst);
4442 ipcl_destroy(ipst);
4443
4444 mutex_destroy(&ipst->ips_ndp4->ndp_g_lock);
4445 mutex_destroy(&ipst->ips_ndp6->ndp_g_lock);
4446 kmem_free(ipst->ips_ndp4, sizeof (ndp_g_t));
4447 ipst->ips_ndp4 = NULL;
4448 kmem_free(ipst->ips_ndp6, sizeof (ndp_g_t));
4449 ipst->ips_ndp6 = NULL;
4450
4451 if (ipst->ips_loopback_ksp != NULL) {
4452 kstat_delete_netstack(ipst->ips_loopback_ksp, stackid);
4453 ipst->ips_loopback_ksp = NULL;
4454 }
4455
4456 mutex_destroy(&ipst->ips_capab_taskq_lock);
4457 cv_destroy(&ipst->ips_capab_taskq_cv);
4458
4459 rw_destroy(&ipst->ips_srcid_lock);
4460
4461 mutex_destroy(&ipst->ips_ip_mi_lock);
4462 rw_destroy(&ipst->ips_ill_g_usesrc_lock);
4463
4464 mutex_destroy(&ipst->ips_igmp_timer_lock);
4465 mutex_destroy(&ipst->ips_mld_timer_lock);
4466 mutex_destroy(&ipst->ips_igmp_slowtimeout_lock);
4467 mutex_destroy(&ipst->ips_mld_slowtimeout_lock);
4468 mutex_destroy(&ipst->ips_ip_addr_avail_lock);
4469 rw_destroy(&ipst->ips_ill_g_lock);
4470
4471 kmem_free(ipst->ips_phyint_g_list, sizeof (phyint_list_t));
4472 ipst->ips_phyint_g_list = NULL;
4473 kmem_free(ipst->ips_ill_g_heads, sizeof (ill_g_head_t) * MAX_G_HEADS);
4474 ipst->ips_ill_g_heads = NULL;
4475
4476 ldi_ident_release(ipst->ips_ldi_ident);
4477 kmem_free(ipst, sizeof (*ipst));
4478 }
4479
4480 /*
4481 * This function is called from the TSD destructor, and is used to debug
4482 * reference count issues in IP. See block comment in <inet/ip_if.h> for
4483 * details.
4484 */
4485 static void
4486 ip_thread_exit(void *phash)
4487 {
4488 th_hash_t *thh = phash;
4489
4490 rw_enter(&ip_thread_rwlock, RW_WRITER);
4491 list_remove(&ip_thread_list, thh);
4492 rw_exit(&ip_thread_rwlock);
4493 mod_hash_destroy_hash(thh->thh_hash);
4494 kmem_free(thh, sizeof (*thh));
4495 }
4496
4497 /*
4498 * Called when the IP kernel module is loaded into the kernel
4499 */
4500 void
4501 ip_ddi_init(void)
4502 {
4503 ip_squeue_flag = ip_squeue_switch(ip_squeue_enter);
4504
4505 /*
4506 * For IP and TCP the minor numbers should start from 2 since we have 4
4507 * initial devices: ip, ip6, tcp, tcp6.
4508 */
4509 /*
4510 * If this is a 64-bit kernel, then create two separate arenas -
4511 * one for TLIs in the range of INET_MIN_DEV+2 through 2^^18-1, and the
4512 * other for socket apps in the range 2^^18 through 2^^32-1.
4513 */
4514 ip_minor_arena_la = NULL;
4515 ip_minor_arena_sa = NULL;
4516 #if defined(_LP64)
4517 if ((ip_minor_arena_sa = inet_minor_create("ip_minor_arena_sa",
4518 INET_MIN_DEV + 2, MAXMIN32, KM_SLEEP)) == NULL) {
4519 cmn_err(CE_PANIC,
4520 "ip_ddi_init: ip_minor_arena_sa creation failed\n");
4521 }
4522 if ((ip_minor_arena_la = inet_minor_create("ip_minor_arena_la",
4523 MAXMIN32 + 1, MAXMIN64, KM_SLEEP)) == NULL) {
4524 cmn_err(CE_PANIC,
4525 "ip_ddi_init: ip_minor_arena_la creation failed\n");
4526 }
4527 #else
4528 if ((ip_minor_arena_sa = inet_minor_create("ip_minor_arena_sa",
4529 INET_MIN_DEV + 2, MAXMIN, KM_SLEEP)) == NULL) {
4530 cmn_err(CE_PANIC,
4531 "ip_ddi_init: ip_minor_arena_sa creation failed\n");
4532 }
4533 #endif
4534 ip_poll_normal_ticks = MSEC_TO_TICK_ROUNDUP(ip_poll_normal_ms);
4535
4536 ipcl_g_init();
4537 ip_ire_g_init();
4538 ip_net_g_init();
4539
4540 #ifdef DEBUG
4541 tsd_create(&ip_thread_data, ip_thread_exit);
4542 rw_init(&ip_thread_rwlock, NULL, RW_DEFAULT, NULL);
4543 list_create(&ip_thread_list, sizeof (th_hash_t),
4544 offsetof(th_hash_t, thh_link));
4545 #endif
4546 ipsec_policy_g_init();
4547 tcp_ddi_g_init();
4548 sctp_ddi_g_init();
4549 dce_g_init();
4550
4551 /*
4552 * We want to be informed each time a stack is created or
4553 * destroyed in the kernel, so we can maintain the
4554 * set of udp_stack_t's.
4555 */
4556 netstack_register(NS_IP, ip_stack_init, ip_stack_shutdown,
4557 ip_stack_fini);
4558
4559 tnet_init();
4560
4561 udp_ddi_g_init();
4562 rts_ddi_g_init();
4563 icmp_ddi_g_init();
4564 ilb_ddi_g_init();
4565
4566 /* This needs to be called after all transports are initialized. */
4567 mutex_enter(&cpu_lock);
4568 register_cpu_setup_func(ip_tp_cpu_update, NULL);
4569 mutex_exit(&cpu_lock);
4570 }
4571
4572 /*
4573 * Initialize the IP stack instance.
4574 */
4575 static void *
4576 ip_stack_init(netstackid_t stackid, netstack_t *ns)
4577 {
4578 ip_stack_t *ipst;
4579 size_t arrsz;
4580 major_t major;
4581
4582 #ifdef NS_DEBUG
4583 printf("ip_stack_init(stack %d)\n", stackid);
4584 #endif
4585
4586 ipst = (ip_stack_t *)kmem_zalloc(sizeof (*ipst), KM_SLEEP);
4587 ipst->ips_netstack = ns;
4588
4589 ipst->ips_ill_g_heads = kmem_zalloc(sizeof (ill_g_head_t) * MAX_G_HEADS,
4590 KM_SLEEP);
4591 ipst->ips_phyint_g_list = kmem_zalloc(sizeof (phyint_list_t),
4592 KM_SLEEP);
4593 ipst->ips_ndp4 = kmem_zalloc(sizeof (ndp_g_t), KM_SLEEP);
4594 ipst->ips_ndp6 = kmem_zalloc(sizeof (ndp_g_t), KM_SLEEP);
4595 mutex_init(&ipst->ips_ndp4->ndp_g_lock, NULL, MUTEX_DEFAULT, NULL);
4596 mutex_init(&ipst->ips_ndp6->ndp_g_lock, NULL, MUTEX_DEFAULT, NULL);
4597
4598 mutex_init(&ipst->ips_igmp_timer_lock, NULL, MUTEX_DEFAULT, NULL);
4599 ipst->ips_igmp_deferred_next = INFINITY;
4600 mutex_init(&ipst->ips_mld_timer_lock, NULL, MUTEX_DEFAULT, NULL);
4601 ipst->ips_mld_deferred_next = INFINITY;
4602 mutex_init(&ipst->ips_igmp_slowtimeout_lock, NULL, MUTEX_DEFAULT, NULL);
4603 mutex_init(&ipst->ips_mld_slowtimeout_lock, NULL, MUTEX_DEFAULT, NULL);
4604 mutex_init(&ipst->ips_ip_mi_lock, NULL, MUTEX_DEFAULT, NULL);
4605 mutex_init(&ipst->ips_ip_addr_avail_lock, NULL, MUTEX_DEFAULT, NULL);
4606 rw_init(&ipst->ips_ill_g_lock, NULL, RW_DEFAULT, NULL);
4607 rw_init(&ipst->ips_ill_g_usesrc_lock, NULL, RW_DEFAULT, NULL);
4608
4609 ipcl_init(ipst);
4610 ip_ire_init(ipst);
4611 ip6_asp_init(ipst);
4612 ipif_init(ipst);
4613 conn_drain_init(ipst);
4614 ip_mrouter_stack_init(ipst);
4615 dce_stack_init(ipst);
4616
4617 ipst->ips_ip_multirt_log_interval = 1000;
4618
4619 ipst->ips_ill_index = 1;
4620
4621 ipst->ips_saved_ip_forwarding = -1;
4622 ipst->ips_reg_vif_num = ALL_VIFS; /* Index to Register vif */
4623
4624 arrsz = ip_propinfo_count * sizeof (mod_prop_info_t);
4625 ipst->ips_propinfo_tbl = (mod_prop_info_t *)kmem_alloc(arrsz, KM_SLEEP);
4626 bcopy(ip_propinfo_tbl, ipst->ips_propinfo_tbl, arrsz);
4627
4628 ipst->ips_ip_mibkp = ip_kstat_init(stackid, ipst);
4629 ipst->ips_icmp_mibkp = icmp_kstat_init(stackid);
4630 ipst->ips_ip_kstat = ip_kstat2_init(stackid, &ipst->ips_ip_statistics);
4631 ipst->ips_ip6_kstat =
4632 ip6_kstat_init(stackid, &ipst->ips_ip6_statistics);
4633
4634 ipst->ips_ip_src_id = 1;
4635 rw_init(&ipst->ips_srcid_lock, NULL, RW_DEFAULT, NULL);
4636
4637 ipst->ips_src_generation = SRC_GENERATION_INITIAL;
4638
4639 ip_net_init(ipst, ns);
4640 ipv4_hook_init(ipst);
4641 ipv6_hook_init(ipst);
4642 arp_hook_init(ipst);
4643 ipmp_init(ipst);
4644 ipobs_init(ipst);
4645
4646 /*
4647 * Create the taskq dispatcher thread and initialize related stuff.
4648 */
4649 ipst->ips_capab_taskq_thread = thread_create(NULL, 0,
4650 ill_taskq_dispatch, ipst, 0, &p0, TS_RUN, minclsyspri);
4651 mutex_init(&ipst->ips_capab_taskq_lock, NULL, MUTEX_DEFAULT, NULL);
4652 cv_init(&ipst->ips_capab_taskq_cv, NULL, CV_DEFAULT, NULL);
4653
4654 major = mod_name_to_major(INET_NAME);
4655 (void) ldi_ident_from_major(major, &ipst->ips_ldi_ident);
4656 return (ipst);
4657 }
4658
4659 /*
4660 * Allocate and initialize a DLPI template of the specified length. (May be
4661 * called as writer.)
4662 */
4663 mblk_t *
4664 ip_dlpi_alloc(size_t len, t_uscalar_t prim)
4665 {
4666 mblk_t *mp;
4667
4668 mp = allocb(len, BPRI_MED);
4669 if (!mp)
4670 return (NULL);
4671
4672 /*
4673 * DLPIv2 says that DL_INFO_REQ and DL_TOKEN_REQ (the latter
4674 * of which we don't seem to use) are sent with M_PCPROTO, and
4675 * that other DLPI are M_PROTO.
4676 */
4677 if (prim == DL_INFO_REQ) {
4678 mp->b_datap->db_type = M_PCPROTO;
4679 } else {
4680 mp->b_datap->db_type = M_PROTO;
4681 }
4682
4683 mp->b_wptr = mp->b_rptr + len;
4684 bzero(mp->b_rptr, len);
4685 ((dl_unitdata_req_t *)mp->b_rptr)->dl_primitive = prim;
4686 return (mp);
4687 }
4688
4689 /*
4690 * Allocate and initialize a DLPI notification. (May be called as writer.)
4691 */
4692 mblk_t *
4693 ip_dlnotify_alloc(uint_t notification, uint_t data)
4694 {
4695 dl_notify_ind_t *notifyp;
4696 mblk_t *mp;
4697
4698 if ((mp = ip_dlpi_alloc(DL_NOTIFY_IND_SIZE, DL_NOTIFY_IND)) == NULL)
4699 return (NULL);
4700
4701 notifyp = (dl_notify_ind_t *)mp->b_rptr;
4702 notifyp->dl_notification = notification;
4703 notifyp->dl_data = data;
4704 return (mp);
4705 }
4706
4707 mblk_t *
4708 ip_dlnotify_alloc2(uint_t notification, uint_t data1, uint_t data2)
4709 {
4710 dl_notify_ind_t *notifyp;
4711 mblk_t *mp;
4712
4713 if ((mp = ip_dlpi_alloc(DL_NOTIFY_IND_SIZE, DL_NOTIFY_IND)) == NULL)
4714 return (NULL);
4715
4716 notifyp = (dl_notify_ind_t *)mp->b_rptr;
4717 notifyp->dl_notification = notification;
4718 notifyp->dl_data1 = data1;
4719 notifyp->dl_data2 = data2;
4720 return (mp);
4721 }
4722
4723 /*
4724 * Debug formatting routine. Returns a character string representation of the
4725 * addr in buf, of the form xxx.xxx.xxx.xxx. This routine takes the address
4726 * in the form of a ipaddr_t and calls ip_dot_saddr with a pointer.
4727 *
4728 * Once the ndd table-printing interfaces are removed, this can be changed to
4729 * standard dotted-decimal form.
4730 */
4731 char *
4732 ip_dot_addr(ipaddr_t addr, char *buf)
4733 {
4734 uint8_t *ap = (uint8_t *)&addr;
4735
4736 (void) mi_sprintf(buf, "%03d.%03d.%03d.%03d",
4737 ap[0] & 0xFF, ap[1] & 0xFF, ap[2] & 0xFF, ap[3] & 0xFF);
4738 return (buf);
4739 }
4740
4741 /*
4742 * Write the given MAC address as a printable string in the usual colon-
4743 * separated format.
4744 */
4745 const char *
4746 mac_colon_addr(const uint8_t *addr, size_t alen, char *buf, size_t buflen)
4747 {
4748 char *bp;
4749
4750 if (alen == 0 || buflen < 4)
4751 return ("?");
4752 bp = buf;
4753 for (;;) {
4754 /*
4755 * If there are more MAC address bytes available, but we won't
4756 * have any room to print them, then add "..." to the string
4757 * instead. See below for the 'magic number' explanation.
4758 */
4759 if ((alen == 2 && buflen < 6) || (alen > 2 && buflen < 7)) {
4760 (void) strcpy(bp, "...");
4761 break;
4762 }
4763 (void) sprintf(bp, "%02x", *addr++);
4764 bp += 2;
4765 if (--alen == 0)
4766 break;
4767 *bp++ = ':';
4768 buflen -= 3;
4769 /*
4770 * At this point, based on the first 'if' statement above,
4771 * either alen == 1 and buflen >= 3, or alen > 1 and
4772 * buflen >= 4. The first case leaves room for the final "xx"
4773 * number and trailing NUL byte. The second leaves room for at
4774 * least "...". Thus the apparently 'magic' numbers chosen for
4775 * that statement.
4776 */
4777 }
4778 return (buf);
4779 }
4780
4781 /*
4782 * Called when it is conceptually a ULP that would sent the packet
4783 * e.g., port unreachable and protocol unreachable. Check that the packet
4784 * would have passed the IPsec global policy before sending the error.
4785 *
4786 * Send an ICMP error after patching up the packet appropriately.
4787 * Uses ip_drop_input and bumps the appropriate MIB.
4788 */
4789 void
4790 ip_fanout_send_icmp_v4(mblk_t *mp, uint_t icmp_type, uint_t icmp_code,
4791 ip_recv_attr_t *ira)
4792 {
4793 ipha_t *ipha;
4794 boolean_t secure;
4795 ill_t *ill = ira->ira_ill;
4796 ip_stack_t *ipst = ill->ill_ipst;
4797 netstack_t *ns = ipst->ips_netstack;
4798 ipsec_stack_t *ipss = ns->netstack_ipsec;
4799
4800 secure = ira->ira_flags & IRAF_IPSEC_SECURE;
4801
4802 /*
4803 * We are generating an icmp error for some inbound packet.
4804 * Called from all ip_fanout_(udp, tcp, proto) functions.
4805 * Before we generate an error, check with global policy
4806 * to see whether this is allowed to enter the system. As
4807 * there is no "conn", we are checking with global policy.
4808 */
4809 ipha = (ipha_t *)mp->b_rptr;
4810 if (secure || ipss->ipsec_inbound_v4_policy_present) {
4811 mp = ipsec_check_global_policy(mp, NULL, ipha, NULL, ira, ns);
4812 if (mp == NULL)
4813 return;
4814 }
4815
4816 /* We never send errors for protocols that we do implement */
4817 if (ira->ira_protocol == IPPROTO_ICMP ||
4818 ira->ira_protocol == IPPROTO_IGMP) {
4819 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
4820 ip_drop_input("ip_fanout_send_icmp_v4", mp, ill);
4821 freemsg(mp);
4822 return;
4823 }
4824 /*
4825 * Have to correct checksum since
4826 * the packet might have been
4827 * fragmented and the reassembly code in ip_rput
4828 * does not restore the IP checksum.
4829 */
4830 ipha->ipha_hdr_checksum = 0;
4831 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
4832
4833 switch (icmp_type) {
4834 case ICMP_DEST_UNREACHABLE:
4835 switch (icmp_code) {
4836 case ICMP_PROTOCOL_UNREACHABLE:
4837 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInUnknownProtos);
4838 ip_drop_input("ipIfStatsInUnknownProtos", mp, ill);
4839 break;
4840 case ICMP_PORT_UNREACHABLE:
4841 BUMP_MIB(ill->ill_ip_mib, udpIfStatsNoPorts);
4842 ip_drop_input("ipIfStatsNoPorts", mp, ill);
4843 break;
4844 }
4845
4846 icmp_unreachable(mp, icmp_code, ira);
4847 break;
4848 default:
4849 #ifdef DEBUG
4850 panic("ip_fanout_send_icmp_v4: wrong type");
4851 /*NOTREACHED*/
4852 #else
4853 freemsg(mp);
4854 break;
4855 #endif
4856 }
4857 }
4858
4859 /*
4860 * Used to send an ICMP error message when a packet is received for
4861 * a protocol that is not supported. The mblk passed as argument
4862 * is consumed by this function.
4863 */
4864 void
4865 ip_proto_not_sup(mblk_t *mp, ip_recv_attr_t *ira)
4866 {
4867 ipha_t *ipha;
4868
4869 ipha = (ipha_t *)mp->b_rptr;
4870 if (ira->ira_flags & IRAF_IS_IPV4) {
4871 ASSERT(IPH_HDR_VERSION(ipha) == IP_VERSION);
4872 ip_fanout_send_icmp_v4(mp, ICMP_DEST_UNREACHABLE,
4873 ICMP_PROTOCOL_UNREACHABLE, ira);
4874 } else {
4875 ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
4876 ip_fanout_send_icmp_v6(mp, ICMP6_PARAM_PROB,
4877 ICMP6_PARAMPROB_NEXTHEADER, ira);
4878 }
4879 }
4880
4881 /*
4882 * Deliver a rawip packet to the given conn, possibly applying ipsec policy.
4883 * Handles IPv4 and IPv6.
4884 * We are responsible for disposing of mp, such as by freemsg() or putnext()
4885 * Caller is responsible for dropping references to the conn.
4886 */
4887 void
4888 ip_fanout_proto_conn(conn_t *connp, mblk_t *mp, ipha_t *ipha, ip6_t *ip6h,
4889 ip_recv_attr_t *ira)
4890 {
4891 ill_t *ill = ira->ira_ill;
4892 ip_stack_t *ipst = ill->ill_ipst;
4893 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
4894 boolean_t secure;
4895 uint_t protocol = ira->ira_protocol;
4896 iaflags_t iraflags = ira->ira_flags;
4897 queue_t *rq;
4898
4899 secure = iraflags & IRAF_IPSEC_SECURE;
4900
4901 rq = connp->conn_rq;
4902 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld : !canputnext(rq)) {
4903 switch (protocol) {
4904 case IPPROTO_ICMPV6:
4905 BUMP_MIB(ill->ill_icmp6_mib, ipv6IfIcmpInOverflows);
4906 break;
4907 case IPPROTO_ICMP:
4908 BUMP_MIB(&ipst->ips_icmp_mib, icmpInOverflows);
4909 break;
4910 default:
4911 BUMP_MIB(ill->ill_ip_mib, rawipIfStatsInOverflows);
4912 break;
4913 }
4914 freemsg(mp);
4915 return;
4916 }
4917
4918 ASSERT(!(IPCL_IS_IPTUN(connp)));
4919
4920 if (((iraflags & IRAF_IS_IPV4) ?
4921 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
4922 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
4923 secure) {
4924 mp = ipsec_check_inbound_policy(mp, connp, ipha,
4925 ip6h, ira);
4926 if (mp == NULL) {
4927 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
4928 /* Note that mp is NULL */
4929 ip_drop_input("ipIfStatsInDiscards", mp, ill);
4930 return;
4931 }
4932 }
4933
4934 if (iraflags & IRAF_ICMP_ERROR) {
4935 (connp->conn_recvicmp)(connp, mp, NULL, ira);
4936 } else {
4937 ill_t *rill = ira->ira_rill;
4938
4939 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
4940 ira->ira_ill = ira->ira_rill = NULL;
4941 /* Send it upstream */
4942 (connp->conn_recv)(connp, mp, NULL, ira);
4943 ira->ira_ill = ill;
4944 ira->ira_rill = rill;
4945 }
4946 }
4947
4948 /*
4949 * Handle protocols with which IP is less intimate. There
4950 * can be more than one stream bound to a particular
4951 * protocol. When this is the case, normally each one gets a copy
4952 * of any incoming packets.
4953 *
4954 * IPsec NOTE :
4955 *
4956 * Don't allow a secure packet going up a non-secure connection.
4957 * We don't allow this because
4958 *
4959 * 1) Reply might go out in clear which will be dropped at
4960 * the sending side.
4961 * 2) If the reply goes out in clear it will give the
4962 * adversary enough information for getting the key in
4963 * most of the cases.
4964 *
4965 * Moreover getting a secure packet when we expect clear
4966 * implies that SA's were added without checking for
4967 * policy on both ends. This should not happen once ISAKMP
4968 * is used to negotiate SAs as SAs will be added only after
4969 * verifying the policy.
4970 *
4971 * Zones notes:
4972 * Earlier in ip_input on a system with multiple shared-IP zones we
4973 * duplicate the multicast and broadcast packets and send them up
4974 * with each explicit zoneid that exists on that ill.
4975 * This means that here we can match the zoneid with SO_ALLZONES being special.
4976 */
4977 void
4978 ip_fanout_proto_v4(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
4979 {
4980 mblk_t *mp1;
4981 ipaddr_t laddr;
4982 conn_t *connp, *first_connp, *next_connp;
4983 connf_t *connfp;
4984 ill_t *ill = ira->ira_ill;
4985 ip_stack_t *ipst = ill->ill_ipst;
4986
4987 laddr = ipha->ipha_dst;
4988
4989 connfp = &ipst->ips_ipcl_proto_fanout_v4[ira->ira_protocol];
4990 mutex_enter(&connfp->connf_lock);
4991 connp = connfp->connf_head;
4992 for (connp = connfp->connf_head; connp != NULL;
4993 connp = connp->conn_next) {
4994 /* Note: IPCL_PROTO_MATCH includes conn_wantpacket */
4995 if (IPCL_PROTO_MATCH(connp, ira, ipha) &&
4996 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
4997 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp))) {
4998 break;
4999 }
5000 }
5001
5002 if (connp == NULL) {
5003 /*
5004 * No one bound to these addresses. Is
5005 * there a client that wants all
5006 * unclaimed datagrams?
5007 */
5008 mutex_exit(&connfp->connf_lock);
5009 ip_fanout_send_icmp_v4(mp, ICMP_DEST_UNREACHABLE,
5010 ICMP_PROTOCOL_UNREACHABLE, ira);
5011 return;
5012 }
5013
5014 ASSERT(IPCL_IS_NONSTR(connp) || connp->conn_rq != NULL);
5015
5016 CONN_INC_REF(connp);
5017 first_connp = connp;
5018 connp = connp->conn_next;
5019
5020 for (;;) {
5021 while (connp != NULL) {
5022 /* Note: IPCL_PROTO_MATCH includes conn_wantpacket */
5023 if (IPCL_PROTO_MATCH(connp, ira, ipha) &&
5024 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5025 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5026 ira, connp)))
5027 break;
5028 connp = connp->conn_next;
5029 }
5030
5031 if (connp == NULL) {
5032 /* No more interested clients */
5033 connp = first_connp;
5034 break;
5035 }
5036 if (((mp1 = dupmsg(mp)) == NULL) &&
5037 ((mp1 = copymsg(mp)) == NULL)) {
5038 /* Memory allocation failed */
5039 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5040 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5041 connp = first_connp;
5042 break;
5043 }
5044
5045 CONN_INC_REF(connp);
5046 mutex_exit(&connfp->connf_lock);
5047
5048 ip_fanout_proto_conn(connp, mp1, (ipha_t *)mp1->b_rptr, NULL,
5049 ira);
5050
5051 mutex_enter(&connfp->connf_lock);
5052 /* Follow the next pointer before releasing the conn. */
5053 next_connp = connp->conn_next;
5054 CONN_DEC_REF(connp);
5055 connp = next_connp;
5056 }
5057
5058 /* Last one. Send it upstream. */
5059 mutex_exit(&connfp->connf_lock);
5060
5061 ip_fanout_proto_conn(connp, mp, ipha, NULL, ira);
5062
5063 CONN_DEC_REF(connp);
5064 }
5065
5066 /*
5067 * If we have a IPsec NAT-Traversal packet, strip the zero-SPI or
5068 * pass it along to ESP if the SPI is non-zero. Returns the mblk if the mblk
5069 * is not consumed.
5070 *
5071 * One of three things can happen, all of which affect the passed-in mblk:
5072 *
5073 * 1.) The packet is stock UDP and gets its zero-SPI stripped. Return mblk..
5074 *
5075 * 2.) The packet is ESP-in-UDP, gets transformed into an equivalent
5076 * ESP packet, and is passed along to ESP for consumption. Return NULL.
5077 *
5078 * 3.) The packet is an ESP-in-UDP Keepalive. Drop it and return NULL.
5079 */
5080 mblk_t *
5081 zero_spi_check(mblk_t *mp, ip_recv_attr_t *ira)
5082 {
5083 int shift, plen, iph_len;
5084 ipha_t *ipha;
5085 udpha_t *udpha;
5086 uint32_t *spi;
5087 uint32_t esp_ports;
5088 uint8_t *orptr;
5089 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
5090 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
5091
5092 ipha = (ipha_t *)mp->b_rptr;
5093 iph_len = ira->ira_ip_hdr_length;
5094 plen = ira->ira_pktlen;
5095
5096 if (plen - iph_len - sizeof (udpha_t) < sizeof (uint32_t)) {
5097 /*
5098 * Most likely a keepalive for the benefit of an intervening
5099 * NAT. These aren't for us, per se, so drop it.
5100 *
5101 * RFC 3947/8 doesn't say for sure what to do for 2-3
5102 * byte packets (keepalives are 1-byte), but we'll drop them
5103 * also.
5104 */
5105 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
5106 DROPPER(ipss, ipds_esp_nat_t_ka), &ipss->ipsec_dropper);
5107 return (NULL);
5108 }
5109
5110 if (MBLKL(mp) < iph_len + sizeof (udpha_t) + sizeof (*spi)) {
5111 /* might as well pull it all up - it might be ESP. */
5112 if (!pullupmsg(mp, -1)) {
5113 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
5114 DROPPER(ipss, ipds_esp_nomem),
5115 &ipss->ipsec_dropper);
5116 return (NULL);
5117 }
5118
5119 ipha = (ipha_t *)mp->b_rptr;
5120 }
5121 spi = (uint32_t *)(mp->b_rptr + iph_len + sizeof (udpha_t));
5122 if (*spi == 0) {
5123 /* UDP packet - remove 0-spi. */
5124 shift = sizeof (uint32_t);
5125 } else {
5126 /* ESP-in-UDP packet - reduce to ESP. */
5127 ipha->ipha_protocol = IPPROTO_ESP;
5128 shift = sizeof (udpha_t);
5129 }
5130
5131 /* Fix IP header */
5132 ira->ira_pktlen = (plen - shift);
5133 ipha->ipha_length = htons(ira->ira_pktlen);
5134 ipha->ipha_hdr_checksum = 0;
5135
5136 orptr = mp->b_rptr;
5137 mp->b_rptr += shift;
5138
5139 udpha = (udpha_t *)(orptr + iph_len);
5140 if (*spi == 0) {
5141 ASSERT((uint8_t *)ipha == orptr);
5142 udpha->uha_length = htons(plen - shift - iph_len);
5143 iph_len += sizeof (udpha_t); /* For the call to ovbcopy(). */
5144 esp_ports = 0;
5145 } else {
5146 esp_ports = *((uint32_t *)udpha);
5147 ASSERT(esp_ports != 0);
5148 }
5149 ovbcopy(orptr, orptr + shift, iph_len);
5150 if (esp_ports != 0) /* Punt up for ESP processing. */ {
5151 ipha = (ipha_t *)(orptr + shift);
5152
5153 ira->ira_flags |= IRAF_ESP_UDP_PORTS;
5154 ira->ira_esp_udp_ports = esp_ports;
5155 ip_fanout_v4(mp, ipha, ira);
5156 return (NULL);
5157 }
5158 return (mp);
5159 }
5160
5161 /*
5162 * Deliver a udp packet to the given conn, possibly applying ipsec policy.
5163 * Handles IPv4 and IPv6.
5164 * We are responsible for disposing of mp, such as by freemsg() or putnext()
5165 * Caller is responsible for dropping references to the conn.
5166 */
5167 void
5168 ip_fanout_udp_conn(conn_t *connp, mblk_t *mp, ipha_t *ipha, ip6_t *ip6h,
5169 ip_recv_attr_t *ira)
5170 {
5171 ill_t *ill = ira->ira_ill;
5172 ip_stack_t *ipst = ill->ill_ipst;
5173 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
5174 boolean_t secure;
5175 iaflags_t iraflags = ira->ira_flags;
5176
5177 secure = iraflags & IRAF_IPSEC_SECURE;
5178
5179 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld :
5180 !canputnext(connp->conn_rq)) {
5181 BUMP_MIB(ill->ill_ip_mib, udpIfStatsInOverflows);
5182 freemsg(mp);
5183 return;
5184 }
5185
5186 if (((iraflags & IRAF_IS_IPV4) ?
5187 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
5188 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
5189 secure) {
5190 mp = ipsec_check_inbound_policy(mp, connp, ipha,
5191 ip6h, ira);
5192 if (mp == NULL) {
5193 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5194 /* Note that mp is NULL */
5195 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5196 return;
5197 }
5198 }
5199
5200 /*
5201 * Since this code is not used for UDP unicast we don't need a NAT_T
5202 * check. Only ip_fanout_v4 has that check.
5203 */
5204 if (ira->ira_flags & IRAF_ICMP_ERROR) {
5205 (connp->conn_recvicmp)(connp, mp, NULL, ira);
5206 } else {
5207 ill_t *rill = ira->ira_rill;
5208
5209 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
5210 ira->ira_ill = ira->ira_rill = NULL;
5211 /* Send it upstream */
5212 (connp->conn_recv)(connp, mp, NULL, ira);
5213 ira->ira_ill = ill;
5214 ira->ira_rill = rill;
5215 }
5216 }
5217
5218 /*
5219 * Fanout for UDP packets that are multicast or broadcast, and ICMP errors.
5220 * (Unicast fanout is handled in ip_input_v4.)
5221 *
5222 * If SO_REUSEADDR is set all multicast and broadcast packets
5223 * will be delivered to all conns bound to the same port.
5224 *
5225 * If there is at least one matching AF_INET receiver, then we will
5226 * ignore any AF_INET6 receivers.
5227 * In the special case where an AF_INET socket binds to 0.0.0.0/<port> and an
5228 * AF_INET6 socket binds to ::/<port>, only the AF_INET socket receives the IPv4
5229 * packets.
5230 *
5231 * Zones notes:
5232 * Earlier in ip_input on a system with multiple shared-IP zones we
5233 * duplicate the multicast and broadcast packets and send them up
5234 * with each explicit zoneid that exists on that ill.
5235 * This means that here we can match the zoneid with SO_ALLZONES being special.
5236 */
5237 void
5238 ip_fanout_udp_multi_v4(mblk_t *mp, ipha_t *ipha, uint16_t lport, uint16_t fport,
5239 ip_recv_attr_t *ira)
5240 {
5241 ipaddr_t laddr;
5242 in6_addr_t v6faddr;
5243 conn_t *connp;
5244 connf_t *connfp;
5245 ipaddr_t faddr;
5246 ill_t *ill = ira->ira_ill;
5247 ip_stack_t *ipst = ill->ill_ipst;
5248
5249 ASSERT(ira->ira_flags & (IRAF_MULTIBROADCAST|IRAF_ICMP_ERROR));
5250
5251 laddr = ipha->ipha_dst;
5252 faddr = ipha->ipha_src;
5253
5254 connfp = &ipst->ips_ipcl_udp_fanout[IPCL_UDP_HASH(lport, ipst)];
5255 mutex_enter(&connfp->connf_lock);
5256 connp = connfp->connf_head;
5257
5258 /*
5259 * If SO_REUSEADDR has been set on the first we send the
5260 * packet to all clients that have joined the group and
5261 * match the port.
5262 */
5263 while (connp != NULL) {
5264 if ((IPCL_UDP_MATCH(connp, lport, laddr, fport, faddr)) &&
5265 conn_wantpacket(connp, ira, ipha) &&
5266 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5267 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp)))
5268 break;
5269 connp = connp->conn_next;
5270 }
5271
5272 if (connp == NULL)
5273 goto notfound;
5274
5275 CONN_INC_REF(connp);
5276
5277 if (connp->conn_reuseaddr) {
5278 conn_t *first_connp = connp;
5279 conn_t *next_connp;
5280 mblk_t *mp1;
5281
5282 connp = connp->conn_next;
5283 for (;;) {
5284 while (connp != NULL) {
5285 if (IPCL_UDP_MATCH(connp, lport, laddr,
5286 fport, faddr) &&
5287 conn_wantpacket(connp, ira, ipha) &&
5288 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5289 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5290 ira, connp)))
5291 break;
5292 connp = connp->conn_next;
5293 }
5294 if (connp == NULL) {
5295 /* No more interested clients */
5296 connp = first_connp;
5297 break;
5298 }
5299 if (((mp1 = dupmsg(mp)) == NULL) &&
5300 ((mp1 = copymsg(mp)) == NULL)) {
5301 /* Memory allocation failed */
5302 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5303 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5304 connp = first_connp;
5305 break;
5306 }
5307 CONN_INC_REF(connp);
5308 mutex_exit(&connfp->connf_lock);
5309
5310 IP_STAT(ipst, ip_udp_fanmb);
5311 ip_fanout_udp_conn(connp, mp1, (ipha_t *)mp1->b_rptr,
5312 NULL, ira);
5313 mutex_enter(&connfp->connf_lock);
5314 /* Follow the next pointer before releasing the conn */
5315 next_connp = connp->conn_next;
5316 CONN_DEC_REF(connp);
5317 connp = next_connp;
5318 }
5319 }
5320
5321 /* Last one. Send it upstream. */
5322 mutex_exit(&connfp->connf_lock);
5323 IP_STAT(ipst, ip_udp_fanmb);
5324 ip_fanout_udp_conn(connp, mp, ipha, NULL, ira);
5325 CONN_DEC_REF(connp);
5326 return;
5327
5328 notfound:
5329 mutex_exit(&connfp->connf_lock);
5330 /*
5331 * IPv6 endpoints bound to multicast IPv4-mapped addresses
5332 * have already been matched above, since they live in the IPv4
5333 * fanout tables. This implies we only need to
5334 * check for IPv6 in6addr_any endpoints here.
5335 * Thus we compare using ipv6_all_zeros instead of the destination
5336 * address, except for the multicast group membership lookup which
5337 * uses the IPv4 destination.
5338 */
5339 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_src, &v6faddr);
5340 connfp = &ipst->ips_ipcl_udp_fanout[IPCL_UDP_HASH(lport, ipst)];
5341 mutex_enter(&connfp->connf_lock);
5342 connp = connfp->connf_head;
5343 /*
5344 * IPv4 multicast packet being delivered to an AF_INET6
5345 * in6addr_any endpoint.
5346 * Need to check conn_wantpacket(). Note that we use conn_wantpacket()
5347 * and not conn_wantpacket_v6() since any multicast membership is
5348 * for an IPv4-mapped multicast address.
5349 */
5350 while (connp != NULL) {
5351 if (IPCL_UDP_MATCH_V6(connp, lport, ipv6_all_zeros,
5352 fport, v6faddr) &&
5353 conn_wantpacket(connp, ira, ipha) &&
5354 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5355 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp)))
5356 break;
5357 connp = connp->conn_next;
5358 }
5359
5360 if (connp == NULL) {
5361 /*
5362 * No one bound to this port. Is
5363 * there a client that wants all
5364 * unclaimed datagrams?
5365 */
5366 mutex_exit(&connfp->connf_lock);
5367
5368 if (ipst->ips_ipcl_proto_fanout_v4[IPPROTO_UDP].connf_head !=
5369 NULL) {
5370 ASSERT(ira->ira_protocol == IPPROTO_UDP);
5371 ip_fanout_proto_v4(mp, ipha, ira);
5372 } else {
5373 /*
5374 * We used to attempt to send an icmp error here, but
5375 * since this is known to be a multicast packet
5376 * and we don't send icmp errors in response to
5377 * multicast, just drop the packet and give up sooner.
5378 */
5379 BUMP_MIB(ill->ill_ip_mib, udpIfStatsNoPorts);
5380 freemsg(mp);
5381 }
5382 return;
5383 }
5384 ASSERT(IPCL_IS_NONSTR(connp) || connp->conn_rq != NULL);
5385
5386 /*
5387 * If SO_REUSEADDR has been set on the first we send the
5388 * packet to all clients that have joined the group and
5389 * match the port.
5390 */
5391 if (connp->conn_reuseaddr) {
5392 conn_t *first_connp = connp;
5393 conn_t *next_connp;
5394 mblk_t *mp1;
5395
5396 CONN_INC_REF(connp);
5397 connp = connp->conn_next;
5398 for (;;) {
5399 while (connp != NULL) {
5400 if (IPCL_UDP_MATCH_V6(connp, lport,
5401 ipv6_all_zeros, fport, v6faddr) &&
5402 conn_wantpacket(connp, ira, ipha) &&
5403 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5404 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5405 ira, connp)))
5406 break;
5407 connp = connp->conn_next;
5408 }
5409 if (connp == NULL) {
5410 /* No more interested clients */
5411 connp = first_connp;
5412 break;
5413 }
5414 if (((mp1 = dupmsg(mp)) == NULL) &&
5415 ((mp1 = copymsg(mp)) == NULL)) {
5416 /* Memory allocation failed */
5417 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5418 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5419 connp = first_connp;
5420 break;
5421 }
5422 CONN_INC_REF(connp);
5423 mutex_exit(&connfp->connf_lock);
5424
5425 IP_STAT(ipst, ip_udp_fanmb);
5426 ip_fanout_udp_conn(connp, mp1, (ipha_t *)mp1->b_rptr,
5427 NULL, ira);
5428 mutex_enter(&connfp->connf_lock);
5429 /* Follow the next pointer before releasing the conn */
5430 next_connp = connp->conn_next;
5431 CONN_DEC_REF(connp);
5432 connp = next_connp;
5433 }
5434 }
5435
5436 /* Last one. Send it upstream. */
5437 mutex_exit(&connfp->connf_lock);
5438 IP_STAT(ipst, ip_udp_fanmb);
5439 ip_fanout_udp_conn(connp, mp, ipha, NULL, ira);
5440 CONN_DEC_REF(connp);
5441 }
5442
5443 /*
5444 * Split an incoming packet's IPv4 options into the label and the other options.
5445 * If 'allocate' is set it does memory allocation for the ip_pkt_t, including
5446 * clearing out any leftover label or options.
5447 * Otherwise it just makes ipp point into the packet.
5448 *
5449 * Returns zero if ok; ENOMEM if the buffer couldn't be allocated.
5450 */
5451 int
5452 ip_find_hdr_v4(ipha_t *ipha, ip_pkt_t *ipp, boolean_t allocate)
5453 {
5454 uchar_t *opt;
5455 uint32_t totallen;
5456 uint32_t optval;
5457 uint32_t optlen;
5458
5459 ipp->ipp_fields |= IPPF_HOPLIMIT | IPPF_TCLASS | IPPF_ADDR;
5460 ipp->ipp_hoplimit = ipha->ipha_ttl;
5461 ipp->ipp_type_of_service = ipha->ipha_type_of_service;
5462 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &ipp->ipp_addr);
5463
5464 /*
5465 * Get length (in 4 byte octets) of IP header options.
5466 */
5467 totallen = ipha->ipha_version_and_hdr_length -
5468 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
5469
5470 if (totallen == 0) {
5471 if (!allocate)
5472 return (0);
5473
5474 /* Clear out anything from a previous packet */
5475 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
5476 kmem_free(ipp->ipp_ipv4_options,
5477 ipp->ipp_ipv4_options_len);
5478 ipp->ipp_ipv4_options = NULL;
5479 ipp->ipp_ipv4_options_len = 0;
5480 ipp->ipp_fields &= ~IPPF_IPV4_OPTIONS;
5481 }
5482 if (ipp->ipp_fields & IPPF_LABEL_V4) {
5483 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
5484 ipp->ipp_label_v4 = NULL;
5485 ipp->ipp_label_len_v4 = 0;
5486 ipp->ipp_fields &= ~IPPF_LABEL_V4;
5487 }
5488 return (0);
5489 }
5490
5491 totallen <<= 2;
5492 opt = (uchar_t *)&ipha[1];
5493 if (!is_system_labeled()) {
5494
5495 copyall:
5496 if (!allocate) {
5497 if (totallen != 0) {
5498 ipp->ipp_ipv4_options = opt;
5499 ipp->ipp_ipv4_options_len = totallen;
5500 ipp->ipp_fields |= IPPF_IPV4_OPTIONS;
5501 }
5502 return (0);
5503 }
5504 /* Just copy all of options */
5505 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
5506 if (totallen == ipp->ipp_ipv4_options_len) {
5507 bcopy(opt, ipp->ipp_ipv4_options, totallen);
5508 return (0);
5509 }
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 (totallen == 0)
5517 return (0);
5518
5519 ipp->ipp_ipv4_options = kmem_alloc(totallen, KM_NOSLEEP);
5520 if (ipp->ipp_ipv4_options == NULL)
5521 return (ENOMEM);
5522 ipp->ipp_ipv4_options_len = totallen;
5523 ipp->ipp_fields |= IPPF_IPV4_OPTIONS;
5524 bcopy(opt, ipp->ipp_ipv4_options, totallen);
5525 return (0);
5526 }
5527
5528 if (allocate && (ipp->ipp_fields & IPPF_LABEL_V4)) {
5529 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
5530 ipp->ipp_label_v4 = NULL;
5531 ipp->ipp_label_len_v4 = 0;
5532 ipp->ipp_fields &= ~IPPF_LABEL_V4;
5533 }
5534
5535 /*
5536 * Search for CIPSO option.
5537 * We assume CIPSO is first in options if it is present.
5538 * If it isn't, then ipp_opt_ipv4_options will not include the options
5539 * prior to the CIPSO option.
5540 */
5541 while (totallen != 0) {
5542 switch (optval = opt[IPOPT_OPTVAL]) {
5543 case IPOPT_EOL:
5544 return (0);
5545 case IPOPT_NOP:
5546 optlen = 1;
5547 break;
5548 default:
5549 if (totallen <= IPOPT_OLEN)
5550 return (EINVAL);
5551 optlen = opt[IPOPT_OLEN];
5552 if (optlen < 2)
5553 return (EINVAL);
5554 }
5555 if (optlen > totallen)
5556 return (EINVAL);
5557
5558 switch (optval) {
5559 case IPOPT_COMSEC:
5560 if (!allocate) {
5561 ipp->ipp_label_v4 = opt;
5562 ipp->ipp_label_len_v4 = optlen;
5563 ipp->ipp_fields |= IPPF_LABEL_V4;
5564 } else {
5565 ipp->ipp_label_v4 = kmem_alloc(optlen,
5566 KM_NOSLEEP);
5567 if (ipp->ipp_label_v4 == NULL)
5568 return (ENOMEM);
5569 ipp->ipp_label_len_v4 = optlen;
5570 ipp->ipp_fields |= IPPF_LABEL_V4;
5571 bcopy(opt, ipp->ipp_label_v4, optlen);
5572 }
5573 totallen -= optlen;
5574 opt += optlen;
5575
5576 /* Skip padding bytes until we get to a multiple of 4 */
5577 while ((totallen & 3) != 0 && opt[0] == IPOPT_NOP) {
5578 totallen--;
5579 opt++;
5580 }
5581 /* Remaining as ipp_ipv4_options */
5582 goto copyall;
5583 }
5584 totallen -= optlen;
5585 opt += optlen;
5586 }
5587 /* No CIPSO found; return everything as ipp_ipv4_options */
5588 totallen = ipha->ipha_version_and_hdr_length -
5589 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
5590 totallen <<= 2;
5591 opt = (uchar_t *)&ipha[1];
5592 goto copyall;
5593 }
5594
5595 /*
5596 * Efficient versions of lookup for an IRE when we only
5597 * match the address.
5598 * For RTF_REJECT or BLACKHOLE we return IRE_NOROUTE.
5599 * Does not handle multicast addresses.
5600 */
5601 uint_t
5602 ip_type_v4(ipaddr_t addr, ip_stack_t *ipst)
5603 {
5604 ire_t *ire;
5605 uint_t result;
5606
5607 ire = ire_ftable_lookup_simple_v4(addr, 0, ipst, NULL);
5608 ASSERT(ire != NULL);
5609 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))
5610 result = IRE_NOROUTE;
5611 else
5612 result = ire->ire_type;
5613 ire_refrele(ire);
5614 return (result);
5615 }
5616
5617 /*
5618 * Efficient versions of lookup for an IRE when we only
5619 * match the address.
5620 * For RTF_REJECT or BLACKHOLE we return IRE_NOROUTE.
5621 * Does not handle multicast addresses.
5622 */
5623 uint_t
5624 ip_type_v6(const in6_addr_t *addr, ip_stack_t *ipst)
5625 {
5626 ire_t *ire;
5627 uint_t result;
5628
5629 ire = ire_ftable_lookup_simple_v6(addr, 0, ipst, NULL);
5630 ASSERT(ire != NULL);
5631 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))
5632 result = IRE_NOROUTE;
5633 else
5634 result = ire->ire_type;
5635 ire_refrele(ire);
5636 return (result);
5637 }
5638
5639 /*
5640 * Nobody should be sending
5641 * packets up this stream
5642 */
5643 static void
5644 ip_lrput(queue_t *q, mblk_t *mp)
5645 {
5646 switch (mp->b_datap->db_type) {
5647 case M_FLUSH:
5648 /* Turn around */
5649 if (*mp->b_rptr & FLUSHW) {
5650 *mp->b_rptr &= ~FLUSHR;
5651 qreply(q, mp);
5652 return;
5653 }
5654 break;
5655 }
5656 freemsg(mp);
5657 }
5658
5659 /* Nobody should be sending packets down this stream */
5660 /* ARGSUSED */
5661 void
5662 ip_lwput(queue_t *q, mblk_t *mp)
5663 {
5664 freemsg(mp);
5665 }
5666
5667 /*
5668 * Move the first hop in any source route to ipha_dst and remove that part of
5669 * the source route. Called by other protocols. Errors in option formatting
5670 * are ignored - will be handled by ip_output_options. Return the final
5671 * destination (either ipha_dst or the last entry in a source route.)
5672 */
5673 ipaddr_t
5674 ip_massage_options(ipha_t *ipha, netstack_t *ns)
5675 {
5676 ipoptp_t opts;
5677 uchar_t *opt;
5678 uint8_t optval;
5679 uint8_t optlen;
5680 ipaddr_t dst;
5681 int i;
5682 ip_stack_t *ipst = ns->netstack_ip;
5683
5684 ip2dbg(("ip_massage_options\n"));
5685 dst = ipha->ipha_dst;
5686 for (optval = ipoptp_first(&opts, ipha);
5687 optval != IPOPT_EOL;
5688 optval = ipoptp_next(&opts)) {
5689 opt = opts.ipoptp_cur;
5690 switch (optval) {
5691 uint8_t off;
5692 case IPOPT_SSRR:
5693 case IPOPT_LSRR:
5694 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
5695 ip1dbg(("ip_massage_options: bad src route\n"));
5696 break;
5697 }
5698 optlen = opts.ipoptp_len;
5699 off = opt[IPOPT_OFFSET];
5700 off--;
5701 redo_srr:
5702 if (optlen < IP_ADDR_LEN ||
5703 off > optlen - IP_ADDR_LEN) {
5704 /* End of source route */
5705 ip1dbg(("ip_massage_options: end of SR\n"));
5706 break;
5707 }
5708 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
5709 ip1dbg(("ip_massage_options: next hop 0x%x\n",
5710 ntohl(dst)));
5711 /*
5712 * Check if our address is present more than
5713 * once as consecutive hops in source route.
5714 * XXX verify per-interface ip_forwarding
5715 * for source route?
5716 */
5717 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
5718 off += IP_ADDR_LEN;
5719 goto redo_srr;
5720 }
5721 if (dst == htonl(INADDR_LOOPBACK)) {
5722 ip1dbg(("ip_massage_options: loopback addr in "
5723 "source route!\n"));
5724 break;
5725 }
5726 /*
5727 * Update ipha_dst to be the first hop and remove the
5728 * first hop from the source route (by overwriting
5729 * part of the option with NOP options).
5730 */
5731 ipha->ipha_dst = dst;
5732 /* Put the last entry in dst */
5733 off = ((optlen - IP_ADDR_LEN - 3) & ~(IP_ADDR_LEN-1)) +
5734 3;
5735 bcopy(&opt[off], &dst, IP_ADDR_LEN);
5736
5737 ip1dbg(("ip_massage_options: last hop 0x%x\n",
5738 ntohl(dst)));
5739 /* Move down and overwrite */
5740 opt[IP_ADDR_LEN] = opt[0];
5741 opt[IP_ADDR_LEN+1] = opt[IPOPT_OLEN] - IP_ADDR_LEN;
5742 opt[IP_ADDR_LEN+2] = opt[IPOPT_OFFSET];
5743 for (i = 0; i < IP_ADDR_LEN; i++)
5744 opt[i] = IPOPT_NOP;
5745 break;
5746 }
5747 }
5748 return (dst);
5749 }
5750
5751 /*
5752 * Return the network mask
5753 * associated with the specified address.
5754 */
5755 ipaddr_t
5756 ip_net_mask(ipaddr_t addr)
5757 {
5758 uchar_t *up = (uchar_t *)&addr;
5759 ipaddr_t mask = 0;
5760 uchar_t *maskp = (uchar_t *)&mask;
5761
5762 #if defined(__i386) || defined(__amd64)
5763 #define TOTALLY_BRAIN_DAMAGED_C_COMPILER
5764 #endif
5765 #ifdef TOTALLY_BRAIN_DAMAGED_C_COMPILER
5766 maskp[0] = maskp[1] = maskp[2] = maskp[3] = 0;
5767 #endif
5768 if (CLASSD(addr)) {
5769 maskp[0] = 0xF0;
5770 return (mask);
5771 }
5772
5773 /* We assume Class E default netmask to be 32 */
5774 if (CLASSE(addr))
5775 return (0xffffffffU);
5776
5777 if (addr == 0)
5778 return (0);
5779 maskp[0] = 0xFF;
5780 if ((up[0] & 0x80) == 0)
5781 return (mask);
5782
5783 maskp[1] = 0xFF;
5784 if ((up[0] & 0xC0) == 0x80)
5785 return (mask);
5786
5787 maskp[2] = 0xFF;
5788 if ((up[0] & 0xE0) == 0xC0)
5789 return (mask);
5790
5791 /* Otherwise return no mask */
5792 return ((ipaddr_t)0);
5793 }
5794
5795 /* Name/Value Table Lookup Routine */
5796 char *
5797 ip_nv_lookup(nv_t *nv, int value)
5798 {
5799 if (!nv)
5800 return (NULL);
5801 for (; nv->nv_name; nv++) {
5802 if (nv->nv_value == value)
5803 return (nv->nv_name);
5804 }
5805 return ("unknown");
5806 }
5807
5808 static int
5809 ip_wait_for_info_ack(ill_t *ill)
5810 {
5811 int err;
5812
5813 mutex_enter(&ill->ill_lock);
5814 while (ill->ill_state_flags & ILL_LL_SUBNET_PENDING) {
5815 /*
5816 * Return value of 0 indicates a pending signal.
5817 */
5818 err = cv_wait_sig(&ill->ill_cv, &ill->ill_lock);
5819 if (err == 0) {
5820 mutex_exit(&ill->ill_lock);
5821 return (EINTR);
5822 }
5823 }
5824 mutex_exit(&ill->ill_lock);
5825 /*
5826 * ip_rput_other could have set an error in ill_error on
5827 * receipt of M_ERROR.
5828 */
5829 return (ill->ill_error);
5830 }
5831
5832 /*
5833 * This is a module open, i.e. this is a control stream for access
5834 * to a DLPI device. We allocate an ill_t as the instance data in
5835 * this case.
5836 */
5837 static int
5838 ip_modopen(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5839 {
5840 ill_t *ill;
5841 int err;
5842 zoneid_t zoneid;
5843 netstack_t *ns;
5844 ip_stack_t *ipst;
5845
5846 /*
5847 * Prevent unprivileged processes from pushing IP so that
5848 * they can't send raw IP.
5849 */
5850 if (secpolicy_net_rawaccess(credp) != 0)
5851 return (EPERM);
5852
5853 ns = netstack_find_by_cred(credp);
5854 ASSERT(ns != NULL);
5855 ipst = ns->netstack_ip;
5856 ASSERT(ipst != NULL);
5857
5858 /*
5859 * For exclusive stacks we set the zoneid to zero
5860 * to make IP operate as if in the global zone.
5861 */
5862 if (ipst->ips_netstack->netstack_stackid != GLOBAL_NETSTACKID)
5863 zoneid = GLOBAL_ZONEID;
5864 else
5865 zoneid = crgetzoneid(credp);
5866
5867 ill = (ill_t *)mi_open_alloc_sleep(sizeof (ill_t));
5868 q->q_ptr = WR(q)->q_ptr = ill;
5869 ill->ill_ipst = ipst;
5870 ill->ill_zoneid = zoneid;
5871
5872 /*
5873 * ill_init initializes the ill fields and then sends down
5874 * down a DL_INFO_REQ after calling qprocson.
5875 */
5876 err = ill_init(q, ill);
5877
5878 if (err != 0) {
5879 mi_free(ill);
5880 netstack_rele(ipst->ips_netstack);
5881 q->q_ptr = NULL;
5882 WR(q)->q_ptr = NULL;
5883 return (err);
5884 }
5885
5886 /*
5887 * Wait for the DL_INFO_ACK if a DL_INFO_REQ was sent.
5888 *
5889 * ill_init initializes the ipsq marking this thread as
5890 * writer
5891 */
5892 ipsq_exit(ill->ill_phyint->phyint_ipsq);
5893 err = ip_wait_for_info_ack(ill);
5894 if (err == 0)
5895 ill->ill_credp = credp;
5896 else
5897 goto fail;
5898
5899 crhold(credp);
5900
5901 mutex_enter(&ipst->ips_ip_mi_lock);
5902 err = mi_open_link(&ipst->ips_ip_g_head, (IDP)q->q_ptr, devp, flag,
5903 sflag, credp);
5904 mutex_exit(&ipst->ips_ip_mi_lock);
5905 fail:
5906 if (err) {
5907 (void) ip_close(q, 0);
5908 return (err);
5909 }
5910 return (0);
5911 }
5912
5913 /* For /dev/ip aka AF_INET open */
5914 int
5915 ip_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5916 {
5917 return (ip_open(q, devp, flag, sflag, credp, B_FALSE));
5918 }
5919
5920 /* For /dev/ip6 aka AF_INET6 open */
5921 int
5922 ip_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5923 {
5924 return (ip_open(q, devp, flag, sflag, credp, B_TRUE));
5925 }
5926
5927 /* IP open routine. */
5928 int
5929 ip_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp,
5930 boolean_t isv6)
5931 {
5932 conn_t *connp;
5933 major_t maj;
5934 zoneid_t zoneid;
5935 netstack_t *ns;
5936 ip_stack_t *ipst;
5937
5938 /* Allow reopen. */
5939 if (q->q_ptr != NULL)
5940 return (0);
5941
5942 if (sflag & MODOPEN) {
5943 /* This is a module open */
5944 return (ip_modopen(q, devp, flag, sflag, credp));
5945 }
5946
5947 if ((flag & ~(FKLYR)) == IP_HELPER_STR) {
5948 /*
5949 * Non streams based socket looking for a stream
5950 * to access IP
5951 */
5952 return (ip_helper_stream_setup(q, devp, flag, sflag,
5953 credp, isv6));
5954 }
5955
5956 ns = netstack_find_by_cred(credp);
5957 ASSERT(ns != NULL);
5958 ipst = ns->netstack_ip;
5959 ASSERT(ipst != NULL);
5960
5961 /*
5962 * For exclusive stacks we set the zoneid to zero
5963 * to make IP operate as if in the global zone.
5964 */
5965 if (ipst->ips_netstack->netstack_stackid != GLOBAL_NETSTACKID)
5966 zoneid = GLOBAL_ZONEID;
5967 else
5968 zoneid = crgetzoneid(credp);
5969
5970 /*
5971 * We are opening as a device. This is an IP client stream, and we
5972 * allocate an conn_t as the instance data.
5973 */
5974 connp = ipcl_conn_create(IPCL_IPCCONN, KM_SLEEP, ipst->ips_netstack);
5975
5976 /*
5977 * ipcl_conn_create did a netstack_hold. Undo the hold that was
5978 * done by netstack_find_by_cred()
5979 */
5980 netstack_rele(ipst->ips_netstack);
5981
5982 connp->conn_ixa->ixa_flags |= IXAF_MULTICAST_LOOP | IXAF_SET_ULP_CKSUM;
5983 /* conn_allzones can not be set this early, hence no IPCL_ZONEID */
5984 connp->conn_ixa->ixa_zoneid = zoneid;
5985 connp->conn_zoneid = zoneid;
5986
5987 connp->conn_rq = q;
5988 q->q_ptr = WR(q)->q_ptr = connp;
5989
5990 /* Minor tells us which /dev entry was opened */
5991 if (isv6) {
5992 connp->conn_family = AF_INET6;
5993 connp->conn_ipversion = IPV6_VERSION;
5994 connp->conn_ixa->ixa_flags &= ~IXAF_IS_IPV4;
5995 connp->conn_ixa->ixa_src_preferences = IPV6_PREFER_SRC_DEFAULT;
5996 } else {
5997 connp->conn_family = AF_INET;
5998 connp->conn_ipversion = IPV4_VERSION;
5999 connp->conn_ixa->ixa_flags |= IXAF_IS_IPV4;
6000 }
6001
6002 if ((ip_minor_arena_la != NULL) && (flag & SO_SOCKSTR) &&
6003 ((connp->conn_dev = inet_minor_alloc(ip_minor_arena_la)) != 0)) {
6004 connp->conn_minor_arena = ip_minor_arena_la;
6005 } else {
6006 /*
6007 * Either minor numbers in the large arena were exhausted
6008 * or a non socket application is doing the open.
6009 * Try to allocate from the small arena.
6010 */
6011 if ((connp->conn_dev =
6012 inet_minor_alloc(ip_minor_arena_sa)) == 0) {
6013 /* CONN_DEC_REF takes care of netstack_rele() */
6014 q->q_ptr = WR(q)->q_ptr = NULL;
6015 CONN_DEC_REF(connp);
6016 return (EBUSY);
6017 }
6018 connp->conn_minor_arena = ip_minor_arena_sa;
6019 }
6020
6021 maj = getemajor(*devp);
6022 *devp = makedevice(maj, (minor_t)connp->conn_dev);
6023
6024 /*
6025 * connp->conn_cred is crfree()ed in ipcl_conn_destroy()
6026 */
6027 connp->conn_cred = credp;
6028 connp->conn_cpid = curproc->p_pid;
6029 /* Cache things in ixa without an extra refhold */
6030 ASSERT(!(connp->conn_ixa->ixa_free_flags & IXA_FREE_CRED));
6031 connp->conn_ixa->ixa_cred = connp->conn_cred;
6032 connp->conn_ixa->ixa_cpid = connp->conn_cpid;
6033 if (is_system_labeled())
6034 connp->conn_ixa->ixa_tsl = crgetlabel(connp->conn_cred);
6035
6036 /*
6037 * Handle IP_IOC_RTS_REQUEST and other ioctls which use conn_recv
6038 */
6039 connp->conn_recv = ip_conn_input;
6040 connp->conn_recvicmp = ip_conn_input_icmp;
6041
6042 crhold(connp->conn_cred);
6043
6044 /*
6045 * If the caller has the process-wide flag set, then default to MAC
6046 * exempt mode. This allows read-down to unlabeled hosts.
6047 */
6048 if (getpflags(NET_MAC_AWARE, credp) != 0)
6049 connp->conn_mac_mode = CONN_MAC_AWARE;
6050
6051 connp->conn_zone_is_global = (crgetzoneid(credp) == GLOBAL_ZONEID);
6052
6053 connp->conn_rq = q;
6054 connp->conn_wq = WR(q);
6055
6056 /* Non-zero default values */
6057 connp->conn_ixa->ixa_flags |= IXAF_MULTICAST_LOOP;
6058
6059 /*
6060 * Make the conn globally visible to walkers
6061 */
6062 ASSERT(connp->conn_ref == 1);
6063 mutex_enter(&connp->conn_lock);
6064 connp->conn_state_flags &= ~CONN_INCIPIENT;
6065 mutex_exit(&connp->conn_lock);
6066
6067 qprocson(q);
6068
6069 return (0);
6070 }
6071
6072 /*
6073 * Set IPsec policy from an ipsec_req_t. If the req is not "zero" and valid,
6074 * all of them are copied to the conn_t. If the req is "zero", the policy is
6075 * zeroed out. A "zero" policy has zero ipsr_{ah,req,self_encap}_req
6076 * fields.
6077 * We keep only the latest setting of the policy and thus policy setting
6078 * is not incremental/cumulative.
6079 *
6080 * Requests to set policies with multiple alternative actions will
6081 * go through a different API.
6082 */
6083 int
6084 ipsec_set_req(cred_t *cr, conn_t *connp, ipsec_req_t *req)
6085 {
6086 uint_t ah_req = 0;
6087 uint_t esp_req = 0;
6088 uint_t se_req = 0;
6089 ipsec_act_t *actp = NULL;
6090 uint_t nact;
6091 ipsec_policy_head_t *ph;
6092 boolean_t is_pol_reset, is_pol_inserted = B_FALSE;
6093 int error = 0;
6094 netstack_t *ns = connp->conn_netstack;
6095 ip_stack_t *ipst = ns->netstack_ip;
6096 ipsec_stack_t *ipss = ns->netstack_ipsec;
6097
6098 #define REQ_MASK (IPSEC_PREF_REQUIRED|IPSEC_PREF_NEVER)
6099
6100 /*
6101 * The IP_SEC_OPT option does not allow variable length parameters,
6102 * hence a request cannot be NULL.
6103 */
6104 if (req == NULL)
6105 return (EINVAL);
6106
6107 ah_req = req->ipsr_ah_req;
6108 esp_req = req->ipsr_esp_req;
6109 se_req = req->ipsr_self_encap_req;
6110
6111 /* Don't allow setting self-encap without one or more of AH/ESP. */
6112 if (se_req != 0 && esp_req == 0 && ah_req == 0)
6113 return (EINVAL);
6114
6115 /*
6116 * Are we dealing with a request to reset the policy (i.e.
6117 * zero requests).
6118 */
6119 is_pol_reset = ((ah_req & REQ_MASK) == 0 &&
6120 (esp_req & REQ_MASK) == 0 &&
6121 (se_req & REQ_MASK) == 0);
6122
6123 if (!is_pol_reset) {
6124 /*
6125 * If we couldn't load IPsec, fail with "protocol
6126 * not supported".
6127 * IPsec may not have been loaded for a request with zero
6128 * policies, so we don't fail in this case.
6129 */
6130 mutex_enter(&ipss->ipsec_loader_lock);
6131 if (ipss->ipsec_loader_state != IPSEC_LOADER_SUCCEEDED) {
6132 mutex_exit(&ipss->ipsec_loader_lock);
6133 return (EPROTONOSUPPORT);
6134 }
6135 mutex_exit(&ipss->ipsec_loader_lock);
6136
6137 /*
6138 * Test for valid requests. Invalid algorithms
6139 * need to be tested by IPsec code because new
6140 * algorithms can be added dynamically.
6141 */
6142 if ((ah_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0 ||
6143 (esp_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0 ||
6144 (se_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0) {
6145 return (EINVAL);
6146 }
6147
6148 /*
6149 * Only privileged users can issue these
6150 * requests.
6151 */
6152 if (((ah_req & IPSEC_PREF_NEVER) ||
6153 (esp_req & IPSEC_PREF_NEVER) ||
6154 (se_req & IPSEC_PREF_NEVER)) &&
6155 secpolicy_ip_config(cr, B_FALSE) != 0) {
6156 return (EPERM);
6157 }
6158
6159 /*
6160 * The IPSEC_PREF_REQUIRED and IPSEC_PREF_NEVER
6161 * are mutually exclusive.
6162 */
6163 if (((ah_req & REQ_MASK) == REQ_MASK) ||
6164 ((esp_req & REQ_MASK) == REQ_MASK) ||
6165 ((se_req & REQ_MASK) == REQ_MASK)) {
6166 /* Both of them are set */
6167 return (EINVAL);
6168 }
6169 }
6170
6171 ASSERT(MUTEX_HELD(&connp->conn_lock));
6172
6173 /*
6174 * If we have already cached policies in conn_connect(), don't
6175 * let them change now. We cache policies for connections
6176 * whose src,dst [addr, port] is known.
6177 */
6178 if (connp->conn_policy_cached) {
6179 return (EINVAL);
6180 }
6181
6182 /*
6183 * We have a zero policies, reset the connection policy if already
6184 * set. This will cause the connection to inherit the
6185 * global policy, if any.
6186 */
6187 if (is_pol_reset) {
6188 if (connp->conn_policy != NULL) {
6189 IPPH_REFRELE(connp->conn_policy, ipst->ips_netstack);
6190 connp->conn_policy = NULL;
6191 }
6192 connp->conn_in_enforce_policy = B_FALSE;
6193 connp->conn_out_enforce_policy = B_FALSE;
6194 return (0);
6195 }
6196
6197 ph = connp->conn_policy = ipsec_polhead_split(connp->conn_policy,
6198 ipst->ips_netstack);
6199 if (ph == NULL)
6200 goto enomem;
6201
6202 ipsec_actvec_from_req(req, &actp, &nact, ipst->ips_netstack);
6203 if (actp == NULL)
6204 goto enomem;
6205
6206 /*
6207 * Always insert IPv4 policy entries, since they can also apply to
6208 * ipv6 sockets being used in ipv4-compat mode.
6209 */
6210 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V4,
6211 IPSEC_TYPE_INBOUND, ns))
6212 goto enomem;
6213 is_pol_inserted = B_TRUE;
6214 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V4,
6215 IPSEC_TYPE_OUTBOUND, ns))
6216 goto enomem;
6217
6218 /*
6219 * We're looking at a v6 socket, also insert the v6-specific
6220 * entries.
6221 */
6222 if (connp->conn_family == AF_INET6) {
6223 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V6,
6224 IPSEC_TYPE_INBOUND, ns))
6225 goto enomem;
6226 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V6,
6227 IPSEC_TYPE_OUTBOUND, ns))
6228 goto enomem;
6229 }
6230
6231 ipsec_actvec_free(actp, nact);
6232
6233 /*
6234 * If the requests need security, set enforce_policy.
6235 * If the requests are IPSEC_PREF_NEVER, one should
6236 * still set conn_out_enforce_policy so that ip_set_destination
6237 * marks the ip_xmit_attr_t appropriatly. This is needed so that
6238 * for connections that we don't cache policy in at connect time,
6239 * if global policy matches in ip_output_attach_policy, we
6240 * don't wrongly inherit global policy. Similarly, we need
6241 * to set conn_in_enforce_policy also so that we don't verify
6242 * policy wrongly.
6243 */
6244 if ((ah_req & REQ_MASK) != 0 ||
6245 (esp_req & REQ_MASK) != 0 ||
6246 (se_req & REQ_MASK) != 0) {
6247 connp->conn_in_enforce_policy = B_TRUE;
6248 connp->conn_out_enforce_policy = B_TRUE;
6249 }
6250
6251 return (error);
6252 #undef REQ_MASK
6253
6254 /*
6255 * Common memory-allocation-failure exit path.
6256 */
6257 enomem:
6258 if (actp != NULL)
6259 ipsec_actvec_free(actp, nact);
6260 if (is_pol_inserted)
6261 ipsec_polhead_flush(ph, ns);
6262 return (ENOMEM);
6263 }
6264
6265 /*
6266 * Set socket options for joining and leaving multicast groups.
6267 * Common to IPv4 and IPv6; inet6 indicates the type of socket.
6268 * The caller has already check that the option name is consistent with
6269 * the address family of the socket.
6270 */
6271 int
6272 ip_opt_set_multicast_group(conn_t *connp, t_scalar_t name,
6273 uchar_t *invalp, boolean_t inet6, boolean_t checkonly)
6274 {
6275 int *i1 = (int *)invalp;
6276 int error = 0;
6277 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
6278 struct ip_mreq *v4_mreqp;
6279 struct ipv6_mreq *v6_mreqp;
6280 struct group_req *greqp;
6281 ire_t *ire;
6282 boolean_t done = B_FALSE;
6283 ipaddr_t ifaddr;
6284 in6_addr_t v6group;
6285 uint_t ifindex;
6286 boolean_t mcast_opt = B_TRUE;
6287 mcast_record_t fmode;
6288 int (*optfn)(conn_t *, boolean_t, const in6_addr_t *,
6289 ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *);
6290
6291 switch (name) {
6292 case IP_ADD_MEMBERSHIP:
6293 case IPV6_JOIN_GROUP:
6294 mcast_opt = B_FALSE;
6295 /* FALLTHRU */
6296 case MCAST_JOIN_GROUP:
6297 fmode = MODE_IS_EXCLUDE;
6298 optfn = ip_opt_add_group;
6299 break;
6300
6301 case IP_DROP_MEMBERSHIP:
6302 case IPV6_LEAVE_GROUP:
6303 mcast_opt = B_FALSE;
6304 /* FALLTHRU */
6305 case MCAST_LEAVE_GROUP:
6306 fmode = MODE_IS_INCLUDE;
6307 optfn = ip_opt_delete_group;
6308 break;
6309 default:
6310 ASSERT(0);
6311 }
6312
6313 if (mcast_opt) {
6314 struct sockaddr_in *sin;
6315 struct sockaddr_in6 *sin6;
6316
6317 greqp = (struct group_req *)i1;
6318 if (greqp->gr_group.ss_family == AF_INET) {
6319 sin = (struct sockaddr_in *)&(greqp->gr_group);
6320 IN6_INADDR_TO_V4MAPPED(&sin->sin_addr, &v6group);
6321 } else {
6322 if (!inet6)
6323 return (EINVAL); /* Not on INET socket */
6324
6325 sin6 = (struct sockaddr_in6 *)&(greqp->gr_group);
6326 v6group = sin6->sin6_addr;
6327 }
6328 ifaddr = INADDR_ANY;
6329 ifindex = greqp->gr_interface;
6330 } else if (inet6) {
6331 v6_mreqp = (struct ipv6_mreq *)i1;
6332 v6group = v6_mreqp->ipv6mr_multiaddr;
6333 ifaddr = INADDR_ANY;
6334 ifindex = v6_mreqp->ipv6mr_interface;
6335 } else {
6336 v4_mreqp = (struct ip_mreq *)i1;
6337 IN6_INADDR_TO_V4MAPPED(&v4_mreqp->imr_multiaddr, &v6group);
6338 ifaddr = (ipaddr_t)v4_mreqp->imr_interface.s_addr;
6339 ifindex = 0;
6340 }
6341
6342 /*
6343 * In the multirouting case, we need to replicate
6344 * the request on all interfaces that will take part
6345 * in replication. We do so because multirouting is
6346 * reflective, thus we will probably receive multi-
6347 * casts on those interfaces.
6348 * The ip_multirt_apply_membership() succeeds if
6349 * the operation succeeds on at least one interface.
6350 */
6351 if (IN6_IS_ADDR_V4MAPPED(&v6group)) {
6352 ipaddr_t group;
6353
6354 IN6_V4MAPPED_TO_IPADDR(&v6group, group);
6355
6356 ire = ire_ftable_lookup_v4(group, IP_HOST_MASK, 0,
6357 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6358 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6359 } else {
6360 ire = ire_ftable_lookup_v6(&v6group, &ipv6_all_ones, 0,
6361 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6362 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6363 }
6364 if (ire != NULL) {
6365 if (ire->ire_flags & RTF_MULTIRT) {
6366 error = ip_multirt_apply_membership(optfn, ire, connp,
6367 checkonly, &v6group, fmode, &ipv6_all_zeros);
6368 done = B_TRUE;
6369 }
6370 ire_refrele(ire);
6371 }
6372
6373 if (!done) {
6374 error = optfn(connp, checkonly, &v6group, ifaddr, ifindex,
6375 fmode, &ipv6_all_zeros);
6376 }
6377 return (error);
6378 }
6379
6380 /*
6381 * Set socket options for joining and leaving multicast groups
6382 * for specific sources.
6383 * Common to IPv4 and IPv6; inet6 indicates the type of socket.
6384 * The caller has already check that the option name is consistent with
6385 * the address family of the socket.
6386 */
6387 int
6388 ip_opt_set_multicast_sources(conn_t *connp, t_scalar_t name,
6389 uchar_t *invalp, boolean_t inet6, boolean_t checkonly)
6390 {
6391 int *i1 = (int *)invalp;
6392 int error = 0;
6393 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
6394 struct ip_mreq_source *imreqp;
6395 struct group_source_req *gsreqp;
6396 in6_addr_t v6group, v6src;
6397 uint32_t ifindex;
6398 ipaddr_t ifaddr;
6399 boolean_t mcast_opt = B_TRUE;
6400 mcast_record_t fmode;
6401 ire_t *ire;
6402 boolean_t done = B_FALSE;
6403 int (*optfn)(conn_t *, boolean_t, const in6_addr_t *,
6404 ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *);
6405
6406 switch (name) {
6407 case IP_BLOCK_SOURCE:
6408 mcast_opt = B_FALSE;
6409 /* FALLTHRU */
6410 case MCAST_BLOCK_SOURCE:
6411 fmode = MODE_IS_EXCLUDE;
6412 optfn = ip_opt_add_group;
6413 break;
6414
6415 case IP_UNBLOCK_SOURCE:
6416 mcast_opt = B_FALSE;
6417 /* FALLTHRU */
6418 case MCAST_UNBLOCK_SOURCE:
6419 fmode = MODE_IS_EXCLUDE;
6420 optfn = ip_opt_delete_group;
6421 break;
6422
6423 case IP_ADD_SOURCE_MEMBERSHIP:
6424 mcast_opt = B_FALSE;
6425 /* FALLTHRU */
6426 case MCAST_JOIN_SOURCE_GROUP:
6427 fmode = MODE_IS_INCLUDE;
6428 optfn = ip_opt_add_group;
6429 break;
6430
6431 case IP_DROP_SOURCE_MEMBERSHIP:
6432 mcast_opt = B_FALSE;
6433 /* FALLTHRU */
6434 case MCAST_LEAVE_SOURCE_GROUP:
6435 fmode = MODE_IS_INCLUDE;
6436 optfn = ip_opt_delete_group;
6437 break;
6438 default:
6439 ASSERT(0);
6440 }
6441
6442 if (mcast_opt) {
6443 gsreqp = (struct group_source_req *)i1;
6444 ifindex = gsreqp->gsr_interface;
6445 if (gsreqp->gsr_group.ss_family == AF_INET) {
6446 struct sockaddr_in *s;
6447 s = (struct sockaddr_in *)&gsreqp->gsr_group;
6448 IN6_INADDR_TO_V4MAPPED(&s->sin_addr, &v6group);
6449 s = (struct sockaddr_in *)&gsreqp->gsr_source;
6450 IN6_INADDR_TO_V4MAPPED(&s->sin_addr, &v6src);
6451 } else {
6452 struct sockaddr_in6 *s6;
6453
6454 if (!inet6)
6455 return (EINVAL); /* Not on INET socket */
6456
6457 s6 = (struct sockaddr_in6 *)&gsreqp->gsr_group;
6458 v6group = s6->sin6_addr;
6459 s6 = (struct sockaddr_in6 *)&gsreqp->gsr_source;
6460 v6src = s6->sin6_addr;
6461 }
6462 ifaddr = INADDR_ANY;
6463 } else {
6464 imreqp = (struct ip_mreq_source *)i1;
6465 IN6_INADDR_TO_V4MAPPED(&imreqp->imr_multiaddr, &v6group);
6466 IN6_INADDR_TO_V4MAPPED(&imreqp->imr_sourceaddr, &v6src);
6467 ifaddr = (ipaddr_t)imreqp->imr_interface.s_addr;
6468 ifindex = 0;
6469 }
6470
6471 /*
6472 * Handle src being mapped INADDR_ANY by changing it to unspecified.
6473 */
6474 if (IN6_IS_ADDR_V4MAPPED_ANY(&v6src))
6475 v6src = ipv6_all_zeros;
6476
6477 /*
6478 * In the multirouting case, we need to replicate
6479 * the request as noted in the mcast cases above.
6480 */
6481 if (IN6_IS_ADDR_V4MAPPED(&v6group)) {
6482 ipaddr_t group;
6483
6484 IN6_V4MAPPED_TO_IPADDR(&v6group, group);
6485
6486 ire = ire_ftable_lookup_v4(group, IP_HOST_MASK, 0,
6487 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6488 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6489 } else {
6490 ire = ire_ftable_lookup_v6(&v6group, &ipv6_all_ones, 0,
6491 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6492 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6493 }
6494 if (ire != NULL) {
6495 if (ire->ire_flags & RTF_MULTIRT) {
6496 error = ip_multirt_apply_membership(optfn, ire, connp,
6497 checkonly, &v6group, fmode, &v6src);
6498 done = B_TRUE;
6499 }
6500 ire_refrele(ire);
6501 }
6502 if (!done) {
6503 error = optfn(connp, checkonly, &v6group, ifaddr, ifindex,
6504 fmode, &v6src);
6505 }
6506 return (error);
6507 }
6508
6509 /*
6510 * Given a destination address and a pointer to where to put the information
6511 * this routine fills in the mtuinfo.
6512 * The socket must be connected.
6513 * For sctp conn_faddr is the primary address.
6514 */
6515 int
6516 ip_fill_mtuinfo(conn_t *connp, ip_xmit_attr_t *ixa, struct ip6_mtuinfo *mtuinfo)
6517 {
6518 uint32_t pmtu = IP_MAXPACKET;
6519 uint_t scopeid;
6520
6521 if (IN6_IS_ADDR_UNSPECIFIED(&connp->conn_faddr_v6))
6522 return (-1);
6523
6524 /* In case we never sent or called ip_set_destination_v4/v6 */
6525 if (ixa->ixa_ire != NULL)
6526 pmtu = ip_get_pmtu(ixa);
6527
6528 if (ixa->ixa_flags & IXAF_SCOPEID_SET)
6529 scopeid = ixa->ixa_scopeid;
6530 else
6531 scopeid = 0;
6532
6533 bzero(mtuinfo, sizeof (*mtuinfo));
6534 mtuinfo->ip6m_addr.sin6_family = AF_INET6;
6535 mtuinfo->ip6m_addr.sin6_port = connp->conn_fport;
6536 mtuinfo->ip6m_addr.sin6_addr = connp->conn_faddr_v6;
6537 mtuinfo->ip6m_addr.sin6_scope_id = scopeid;
6538 mtuinfo->ip6m_mtu = pmtu;
6539
6540 return (sizeof (struct ip6_mtuinfo));
6541 }
6542
6543 /*
6544 * When the src multihoming is changed from weak to [strong, preferred]
6545 * ip_ire_rebind_walker is called to walk the list of all ire_t entries
6546 * and identify routes that were created by user-applications in the
6547 * unbound state (i.e., without RTA_IFP), and for which an ire_ill is not
6548 * currently defined. These routes are then 'rebound', i.e., their ire_ill
6549 * is selected by finding an interface route for the gateway.
6550 */
6551 /* ARGSUSED */
6552 void
6553 ip_ire_rebind_walker(ire_t *ire, void *notused)
6554 {
6555 if (!ire->ire_unbound || ire->ire_ill != NULL)
6556 return;
6557 ire_rebind(ire);
6558 ire_delete(ire);
6559 }
6560
6561 /*
6562 * When the src multihoming is changed from [strong, preferred] to weak,
6563 * ip_ire_unbind_walker is called to walk the list of all ire_t entries, and
6564 * set any entries that were created by user-applications in the unbound state
6565 * (i.e., without RTA_IFP) back to having a NULL ire_ill.
6566 */
6567 /* ARGSUSED */
6568 void
6569 ip_ire_unbind_walker(ire_t *ire, void *notused)
6570 {
6571 ire_t *new_ire;
6572
6573 if (!ire->ire_unbound || ire->ire_ill == NULL)
6574 return;
6575 if (ire->ire_ipversion == IPV6_VERSION) {
6576 new_ire = ire_create_v6(&ire->ire_addr_v6, &ire->ire_mask_v6,
6577 &ire->ire_gateway_addr_v6, ire->ire_type, NULL,
6578 ire->ire_zoneid, ire->ire_flags, NULL, ire->ire_ipst);
6579 } else {
6580 new_ire = ire_create((uchar_t *)&ire->ire_addr,
6581 (uchar_t *)&ire->ire_mask,
6582 (uchar_t *)&ire->ire_gateway_addr, ire->ire_type, NULL,
6583 ire->ire_zoneid, ire->ire_flags, NULL, ire->ire_ipst);
6584 }
6585 if (new_ire == NULL)
6586 return;
6587 new_ire->ire_unbound = B_TRUE;
6588 /*
6589 * The bound ire must first be deleted so that we don't return
6590 * the existing one on the attempt to add the unbound new_ire.
6591 */
6592 ire_delete(ire);
6593 new_ire = ire_add(new_ire);
6594 if (new_ire != NULL)
6595 ire_refrele(new_ire);
6596 }
6597
6598 /*
6599 * When the settings of ip*_strict_src_multihoming tunables are changed,
6600 * all cached routes need to be recomputed. This recomputation needs to be
6601 * done when going from weaker to stronger modes so that the cached ire
6602 * for the connection does not violate the current ip*_strict_src_multihoming
6603 * setting. It also needs to be done when going from stronger to weaker modes,
6604 * so that we fall back to matching on the longest-matching-route (as opposed
6605 * to a shorter match that may have been selected in the strong mode
6606 * to satisfy src_multihoming settings).
6607 *
6608 * The cached ixa_ire entires for all conn_t entries are marked as
6609 * "verify" so that they will be recomputed for the next packet.
6610 */
6611 void
6612 conn_ire_revalidate(conn_t *connp, void *arg)
6613 {
6614 boolean_t isv6 = (boolean_t)arg;
6615
6616 if ((isv6 && connp->conn_ipversion != IPV6_VERSION) ||
6617 (!isv6 && connp->conn_ipversion != IPV4_VERSION))
6618 return;
6619 connp->conn_ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
6620 }
6621
6622 /*
6623 * Handles both IPv4 and IPv6 reassembly - doing the out-of-order cases,
6624 * When an ipf is passed here for the first time, if
6625 * we already have in-order fragments on the queue, we convert from the fast-
6626 * path reassembly scheme to the hard-case scheme. From then on, additional
6627 * fragments are reassembled here. We keep track of the start and end offsets
6628 * of each piece, and the number of holes in the chain. When the hole count
6629 * goes to zero, we are done!
6630 *
6631 * The ipf_count will be updated to account for any mblk(s) added (pointed to
6632 * by mp) or subtracted (freeb()ed dups), upon return the caller must update
6633 * ipfb_count and ill_frag_count by the difference of ipf_count before and
6634 * after the call to ip_reassemble().
6635 */
6636 int
6637 ip_reassemble(mblk_t *mp, ipf_t *ipf, uint_t start, boolean_t more, ill_t *ill,
6638 size_t msg_len)
6639 {
6640 uint_t end;
6641 mblk_t *next_mp;
6642 mblk_t *mp1;
6643 uint_t offset;
6644 boolean_t incr_dups = B_TRUE;
6645 boolean_t offset_zero_seen = B_FALSE;
6646 boolean_t pkt_boundary_checked = B_FALSE;
6647
6648 /* If start == 0 then ipf_nf_hdr_len has to be set. */
6649 ASSERT(start != 0 || ipf->ipf_nf_hdr_len != 0);
6650
6651 /* Add in byte count */
6652 ipf->ipf_count += msg_len;
6653 if (ipf->ipf_end) {
6654 /*
6655 * We were part way through in-order reassembly, but now there
6656 * is a hole. We walk through messages already queued, and
6657 * mark them for hard case reassembly. We know that up till
6658 * now they were in order starting from offset zero.
6659 */
6660 offset = 0;
6661 for (mp1 = ipf->ipf_mp->b_cont; mp1; mp1 = mp1->b_cont) {
6662 IP_REASS_SET_START(mp1, offset);
6663 if (offset == 0) {
6664 ASSERT(ipf->ipf_nf_hdr_len != 0);
6665 offset = -ipf->ipf_nf_hdr_len;
6666 }
6667 offset += mp1->b_wptr - mp1->b_rptr;
6668 IP_REASS_SET_END(mp1, offset);
6669 }
6670 /* One hole at the end. */
6671 ipf->ipf_hole_cnt = 1;
6672 /* Brand it as a hard case, forever. */
6673 ipf->ipf_end = 0;
6674 }
6675 /* Walk through all the new pieces. */
6676 do {
6677 end = start + (mp->b_wptr - mp->b_rptr);
6678 /*
6679 * If start is 0, decrease 'end' only for the first mblk of
6680 * the fragment. Otherwise 'end' can get wrong value in the
6681 * second pass of the loop if first mblk is exactly the
6682 * size of ipf_nf_hdr_len.
6683 */
6684 if (start == 0 && !offset_zero_seen) {
6685 /* First segment */
6686 ASSERT(ipf->ipf_nf_hdr_len != 0);
6687 end -= ipf->ipf_nf_hdr_len;
6688 offset_zero_seen = B_TRUE;
6689 }
6690 next_mp = mp->b_cont;
6691 /*
6692 * We are checking to see if there is any interesing data
6693 * to process. If there isn't and the mblk isn't the
6694 * one which carries the unfragmentable header then we
6695 * drop it. It's possible to have just the unfragmentable
6696 * header come through without any data. That needs to be
6697 * saved.
6698 *
6699 * If the assert at the top of this function holds then the
6700 * term "ipf->ipf_nf_hdr_len != 0" isn't needed. This code
6701 * is infrequently traveled enough that the test is left in
6702 * to protect against future code changes which break that
6703 * invariant.
6704 */
6705 if (start == end && start != 0 && ipf->ipf_nf_hdr_len != 0) {
6706 /* Empty. Blast it. */
6707 IP_REASS_SET_START(mp, 0);
6708 IP_REASS_SET_END(mp, 0);
6709 /*
6710 * If the ipf points to the mblk we are about to free,
6711 * update ipf to point to the next mblk (or NULL
6712 * if none).
6713 */
6714 if (ipf->ipf_mp->b_cont == mp)
6715 ipf->ipf_mp->b_cont = next_mp;
6716 freeb(mp);
6717 continue;
6718 }
6719 mp->b_cont = NULL;
6720 IP_REASS_SET_START(mp, start);
6721 IP_REASS_SET_END(mp, end);
6722 if (!ipf->ipf_tail_mp) {
6723 ipf->ipf_tail_mp = mp;
6724 ipf->ipf_mp->b_cont = mp;
6725 if (start == 0 || !more) {
6726 ipf->ipf_hole_cnt = 1;
6727 /*
6728 * if the first fragment comes in more than one
6729 * mblk, this loop will be executed for each
6730 * mblk. Need to adjust hole count so exiting
6731 * this routine will leave hole count at 1.
6732 */
6733 if (next_mp)
6734 ipf->ipf_hole_cnt++;
6735 } else
6736 ipf->ipf_hole_cnt = 2;
6737 continue;
6738 } else if (ipf->ipf_last_frag_seen && !more &&
6739 !pkt_boundary_checked) {
6740 /*
6741 * We check datagram boundary only if this fragment
6742 * claims to be the last fragment and we have seen a
6743 * last fragment in the past too. We do this only
6744 * once for a given fragment.
6745 *
6746 * start cannot be 0 here as fragments with start=0
6747 * and MF=0 gets handled as a complete packet. These
6748 * fragments should not reach here.
6749 */
6750
6751 if (start + msgdsize(mp) !=
6752 IP_REASS_END(ipf->ipf_tail_mp)) {
6753 /*
6754 * We have two fragments both of which claim
6755 * to be the last fragment but gives conflicting
6756 * information about the whole datagram size.
6757 * Something fishy is going on. Drop the
6758 * fragment and free up the reassembly list.
6759 */
6760 return (IP_REASS_FAILED);
6761 }
6762
6763 /*
6764 * We shouldn't come to this code block again for this
6765 * particular fragment.
6766 */
6767 pkt_boundary_checked = B_TRUE;
6768 }
6769
6770 /* New stuff at or beyond tail? */
6771 offset = IP_REASS_END(ipf->ipf_tail_mp);
6772 if (start >= offset) {
6773 if (ipf->ipf_last_frag_seen) {
6774 /* current fragment is beyond last fragment */
6775 return (IP_REASS_FAILED);
6776 }
6777 /* Link it on end. */
6778 ipf->ipf_tail_mp->b_cont = mp;
6779 ipf->ipf_tail_mp = mp;
6780 if (more) {
6781 if (start != offset)
6782 ipf->ipf_hole_cnt++;
6783 } else if (start == offset && next_mp == NULL)
6784 ipf->ipf_hole_cnt--;
6785 continue;
6786 }
6787 mp1 = ipf->ipf_mp->b_cont;
6788 offset = IP_REASS_START(mp1);
6789 /* New stuff at the front? */
6790 if (start < offset) {
6791 if (start == 0) {
6792 if (end >= offset) {
6793 /* Nailed the hole at the begining. */
6794 ipf->ipf_hole_cnt--;
6795 }
6796 } else if (end < offset) {
6797 /*
6798 * A hole, stuff, and a hole where there used
6799 * to be just a hole.
6800 */
6801 ipf->ipf_hole_cnt++;
6802 }
6803 mp->b_cont = mp1;
6804 /* Check for overlap. */
6805 while (end > offset) {
6806 if (end < IP_REASS_END(mp1)) {
6807 mp->b_wptr -= end - offset;
6808 IP_REASS_SET_END(mp, offset);
6809 BUMP_MIB(ill->ill_ip_mib,
6810 ipIfStatsReasmPartDups);
6811 break;
6812 }
6813 /* Did we cover another hole? */
6814 if ((mp1->b_cont &&
6815 IP_REASS_END(mp1) !=
6816 IP_REASS_START(mp1->b_cont) &&
6817 end >= IP_REASS_START(mp1->b_cont)) ||
6818 (!ipf->ipf_last_frag_seen && !more)) {
6819 ipf->ipf_hole_cnt--;
6820 }
6821 /* Clip out mp1. */
6822 if ((mp->b_cont = mp1->b_cont) == NULL) {
6823 /*
6824 * After clipping out mp1, this guy
6825 * is now hanging off the end.
6826 */
6827 ipf->ipf_tail_mp = mp;
6828 }
6829 IP_REASS_SET_START(mp1, 0);
6830 IP_REASS_SET_END(mp1, 0);
6831 /* Subtract byte count */
6832 ipf->ipf_count -= mp1->b_datap->db_lim -
6833 mp1->b_datap->db_base;
6834 freeb(mp1);
6835 BUMP_MIB(ill->ill_ip_mib,
6836 ipIfStatsReasmPartDups);
6837 mp1 = mp->b_cont;
6838 if (!mp1)
6839 break;
6840 offset = IP_REASS_START(mp1);
6841 }
6842 ipf->ipf_mp->b_cont = mp;
6843 continue;
6844 }
6845 /*
6846 * The new piece starts somewhere between the start of the head
6847 * and before the end of the tail.
6848 */
6849 for (; mp1; mp1 = mp1->b_cont) {
6850 offset = IP_REASS_END(mp1);
6851 if (start < offset) {
6852 if (end <= offset) {
6853 /* Nothing new. */
6854 IP_REASS_SET_START(mp, 0);
6855 IP_REASS_SET_END(mp, 0);
6856 /* Subtract byte count */
6857 ipf->ipf_count -= mp->b_datap->db_lim -
6858 mp->b_datap->db_base;
6859 if (incr_dups) {
6860 ipf->ipf_num_dups++;
6861 incr_dups = B_FALSE;
6862 }
6863 freeb(mp);
6864 BUMP_MIB(ill->ill_ip_mib,
6865 ipIfStatsReasmDuplicates);
6866 break;
6867 }
6868 /*
6869 * Trim redundant stuff off beginning of new
6870 * piece.
6871 */
6872 IP_REASS_SET_START(mp, offset);
6873 mp->b_rptr += offset - start;
6874 BUMP_MIB(ill->ill_ip_mib,
6875 ipIfStatsReasmPartDups);
6876 start = offset;
6877 if (!mp1->b_cont) {
6878 /*
6879 * After trimming, this guy is now
6880 * hanging off the end.
6881 */
6882 mp1->b_cont = mp;
6883 ipf->ipf_tail_mp = mp;
6884 if (!more) {
6885 ipf->ipf_hole_cnt--;
6886 }
6887 break;
6888 }
6889 }
6890 if (start >= IP_REASS_START(mp1->b_cont))
6891 continue;
6892 /* Fill a hole */
6893 if (start > offset)
6894 ipf->ipf_hole_cnt++;
6895 mp->b_cont = mp1->b_cont;
6896 mp1->b_cont = mp;
6897 mp1 = mp->b_cont;
6898 offset = IP_REASS_START(mp1);
6899 if (end >= offset) {
6900 ipf->ipf_hole_cnt--;
6901 /* Check for overlap. */
6902 while (end > offset) {
6903 if (end < IP_REASS_END(mp1)) {
6904 mp->b_wptr -= end - offset;
6905 IP_REASS_SET_END(mp, offset);
6906 /*
6907 * TODO we might bump
6908 * this up twice if there is
6909 * overlap at both ends.
6910 */
6911 BUMP_MIB(ill->ill_ip_mib,
6912 ipIfStatsReasmPartDups);
6913 break;
6914 }
6915 /* Did we cover another hole? */
6916 if ((mp1->b_cont &&
6917 IP_REASS_END(mp1)
6918 != IP_REASS_START(mp1->b_cont) &&
6919 end >=
6920 IP_REASS_START(mp1->b_cont)) ||
6921 (!ipf->ipf_last_frag_seen &&
6922 !more)) {
6923 ipf->ipf_hole_cnt--;
6924 }
6925 /* Clip out mp1. */
6926 if ((mp->b_cont = mp1->b_cont) ==
6927 NULL) {
6928 /*
6929 * After clipping out mp1,
6930 * this guy is now hanging
6931 * off the end.
6932 */
6933 ipf->ipf_tail_mp = mp;
6934 }
6935 IP_REASS_SET_START(mp1, 0);
6936 IP_REASS_SET_END(mp1, 0);
6937 /* Subtract byte count */
6938 ipf->ipf_count -=
6939 mp1->b_datap->db_lim -
6940 mp1->b_datap->db_base;
6941 freeb(mp1);
6942 BUMP_MIB(ill->ill_ip_mib,
6943 ipIfStatsReasmPartDups);
6944 mp1 = mp->b_cont;
6945 if (!mp1)
6946 break;
6947 offset = IP_REASS_START(mp1);
6948 }
6949 }
6950 break;
6951 }
6952 } while (start = end, mp = next_mp);
6953
6954 /* Fragment just processed could be the last one. Remember this fact */
6955 if (!more)
6956 ipf->ipf_last_frag_seen = B_TRUE;
6957
6958 /* Still got holes? */
6959 if (ipf->ipf_hole_cnt)
6960 return (IP_REASS_PARTIAL);
6961 /* Clean up overloaded fields to avoid upstream disasters. */
6962 for (mp1 = ipf->ipf_mp->b_cont; mp1; mp1 = mp1->b_cont) {
6963 IP_REASS_SET_START(mp1, 0);
6964 IP_REASS_SET_END(mp1, 0);
6965 }
6966 return (IP_REASS_COMPLETE);
6967 }
6968
6969 /*
6970 * Fragmentation reassembly. Each ILL has a hash table for
6971 * queuing packets undergoing reassembly for all IPIFs
6972 * associated with the ILL. The hash is based on the packet
6973 * IP ident field. The ILL frag hash table was allocated
6974 * as a timer block at the time the ILL was created. Whenever
6975 * there is anything on the reassembly queue, the timer will
6976 * be running. Returns the reassembled packet if reassembly completes.
6977 */
6978 mblk_t *
6979 ip_input_fragment(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
6980 {
6981 uint32_t frag_offset_flags;
6982 mblk_t *t_mp;
6983 ipaddr_t dst;
6984 uint8_t proto = ipha->ipha_protocol;
6985 uint32_t sum_val;
6986 uint16_t sum_flags;
6987 ipf_t *ipf;
6988 ipf_t **ipfp;
6989 ipfb_t *ipfb;
6990 uint16_t ident;
6991 uint32_t offset;
6992 ipaddr_t src;
6993 uint_t hdr_length;
6994 uint32_t end;
6995 mblk_t *mp1;
6996 mblk_t *tail_mp;
6997 size_t count;
6998 size_t msg_len;
6999 uint8_t ecn_info = 0;
7000 uint32_t packet_size;
7001 boolean_t pruned = B_FALSE;
7002 ill_t *ill = ira->ira_ill;
7003 ip_stack_t *ipst = ill->ill_ipst;
7004
7005 /*
7006 * Drop the fragmented as early as possible, if
7007 * we don't have resource(s) to re-assemble.
7008 */
7009 if (ipst->ips_ip_reass_queue_bytes == 0) {
7010 freemsg(mp);
7011 return (NULL);
7012 }
7013
7014 /* Check for fragmentation offset; return if there's none */
7015 if ((frag_offset_flags = ntohs(ipha->ipha_fragment_offset_and_flags) &
7016 (IPH_MF | IPH_OFFSET)) == 0)
7017 return (mp);
7018
7019 /*
7020 * We utilize hardware computed checksum info only for UDP since
7021 * IP fragmentation is a normal occurrence for the protocol. In
7022 * addition, checksum offload support for IP fragments carrying
7023 * UDP payload is commonly implemented across network adapters.
7024 */
7025 ASSERT(ira->ira_rill != NULL);
7026 if (proto == IPPROTO_UDP && dohwcksum &&
7027 ILL_HCKSUM_CAPABLE(ira->ira_rill) &&
7028 (DB_CKSUMFLAGS(mp) & (HCK_FULLCKSUM | HCK_PARTIALCKSUM))) {
7029 mblk_t *mp1 = mp->b_cont;
7030 int32_t len;
7031
7032 /* Record checksum information from the packet */
7033 sum_val = (uint32_t)DB_CKSUM16(mp);
7034 sum_flags = DB_CKSUMFLAGS(mp);
7035
7036 /* IP payload offset from beginning of mblk */
7037 offset = ((uchar_t *)ipha + IPH_HDR_LENGTH(ipha)) - mp->b_rptr;
7038
7039 if ((sum_flags & HCK_PARTIALCKSUM) &&
7040 (mp1 == NULL || mp1->b_cont == NULL) &&
7041 offset >= DB_CKSUMSTART(mp) &&
7042 ((len = offset - DB_CKSUMSTART(mp)) & 1) == 0) {
7043 uint32_t adj;
7044 /*
7045 * Partial checksum has been calculated by hardware
7046 * and attached to the packet; in addition, any
7047 * prepended extraneous data is even byte aligned.
7048 * If any such data exists, we adjust the checksum;
7049 * this would also handle any postpended data.
7050 */
7051 IP_ADJCKSUM_PARTIAL(mp->b_rptr + DB_CKSUMSTART(mp),
7052 mp, mp1, len, adj);
7053
7054 /* One's complement subtract extraneous checksum */
7055 if (adj >= sum_val)
7056 sum_val = ~(adj - sum_val) & 0xFFFF;
7057 else
7058 sum_val -= adj;
7059 }
7060 } else {
7061 sum_val = 0;
7062 sum_flags = 0;
7063 }
7064
7065 /* Clear hardware checksumming flag */
7066 DB_CKSUMFLAGS(mp) = 0;
7067
7068 ident = ipha->ipha_ident;
7069 offset = (frag_offset_flags << 3) & 0xFFFF;
7070 src = ipha->ipha_src;
7071 dst = ipha->ipha_dst;
7072 hdr_length = IPH_HDR_LENGTH(ipha);
7073 end = ntohs(ipha->ipha_length) - hdr_length;
7074
7075 /* If end == 0 then we have a packet with no data, so just free it */
7076 if (end == 0) {
7077 freemsg(mp);
7078 return (NULL);
7079 }
7080
7081 /* Record the ECN field info. */
7082 ecn_info = (ipha->ipha_type_of_service & 0x3);
7083 if (offset != 0) {
7084 /*
7085 * If this isn't the first piece, strip the header, and
7086 * add the offset to the end value.
7087 */
7088 mp->b_rptr += hdr_length;
7089 end += offset;
7090 }
7091
7092 /* Handle vnic loopback of fragments */
7093 if (mp->b_datap->db_ref > 2)
7094 msg_len = 0;
7095 else
7096 msg_len = MBLKSIZE(mp);
7097
7098 tail_mp = mp;
7099 while (tail_mp->b_cont != NULL) {
7100 tail_mp = tail_mp->b_cont;
7101 if (tail_mp->b_datap->db_ref <= 2)
7102 msg_len += MBLKSIZE(tail_mp);
7103 }
7104
7105 /* If the reassembly list for this ILL will get too big, prune it */
7106 if ((msg_len + sizeof (*ipf) + ill->ill_frag_count) >=
7107 ipst->ips_ip_reass_queue_bytes) {
7108 DTRACE_PROBE3(ip_reass_queue_bytes, uint_t, msg_len,
7109 uint_t, ill->ill_frag_count,
7110 uint_t, ipst->ips_ip_reass_queue_bytes);
7111 ill_frag_prune(ill,
7112 (ipst->ips_ip_reass_queue_bytes < msg_len) ? 0 :
7113 (ipst->ips_ip_reass_queue_bytes - msg_len));
7114 pruned = B_TRUE;
7115 }
7116
7117 ipfb = &ill->ill_frag_hash_tbl[ILL_FRAG_HASH(src, ident)];
7118 mutex_enter(&ipfb->ipfb_lock);
7119
7120 ipfp = &ipfb->ipfb_ipf;
7121 /* Try to find an existing fragment queue for this packet. */
7122 for (;;) {
7123 ipf = ipfp[0];
7124 if (ipf != NULL) {
7125 /*
7126 * It has to match on ident and src/dst address.
7127 */
7128 if (ipf->ipf_ident == ident &&
7129 ipf->ipf_src == src &&
7130 ipf->ipf_dst == dst &&
7131 ipf->ipf_protocol == proto) {
7132 /*
7133 * If we have received too many
7134 * duplicate fragments for this packet
7135 * free it.
7136 */
7137 if (ipf->ipf_num_dups > ip_max_frag_dups) {
7138 ill_frag_free_pkts(ill, ipfb, ipf, 1);
7139 freemsg(mp);
7140 mutex_exit(&ipfb->ipfb_lock);
7141 return (NULL);
7142 }
7143 /* Found it. */
7144 break;
7145 }
7146 ipfp = &ipf->ipf_hash_next;
7147 continue;
7148 }
7149
7150 /*
7151 * If we pruned the list, do we want to store this new
7152 * fragment?. We apply an optimization here based on the
7153 * fact that most fragments will be received in order.
7154 * So if the offset of this incoming fragment is zero,
7155 * it is the first fragment of a new packet. We will
7156 * keep it. Otherwise drop the fragment, as we have
7157 * probably pruned the packet already (since the
7158 * packet cannot be found).
7159 */
7160 if (pruned && offset != 0) {
7161 mutex_exit(&ipfb->ipfb_lock);
7162 freemsg(mp);
7163 return (NULL);
7164 }
7165
7166 if (ipfb->ipfb_frag_pkts >= MAX_FRAG_PKTS(ipst)) {
7167 /*
7168 * Too many fragmented packets in this hash
7169 * bucket. Free the oldest.
7170 */
7171 ill_frag_free_pkts(ill, ipfb, ipfb->ipfb_ipf, 1);
7172 }
7173
7174 /* New guy. Allocate a frag message. */
7175 mp1 = allocb(sizeof (*ipf), BPRI_MED);
7176 if (mp1 == NULL) {
7177 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7178 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7179 freemsg(mp);
7180 reass_done:
7181 mutex_exit(&ipfb->ipfb_lock);
7182 return (NULL);
7183 }
7184
7185 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmReqds);
7186 mp1->b_cont = mp;
7187
7188 /* Initialize the fragment header. */
7189 ipf = (ipf_t *)mp1->b_rptr;
7190 ipf->ipf_mp = mp1;
7191 ipf->ipf_ptphn = ipfp;
7192 ipfp[0] = ipf;
7193 ipf->ipf_hash_next = NULL;
7194 ipf->ipf_ident = ident;
7195 ipf->ipf_protocol = proto;
7196 ipf->ipf_src = src;
7197 ipf->ipf_dst = dst;
7198 ipf->ipf_nf_hdr_len = 0;
7199 /* Record reassembly start time. */
7200 ipf->ipf_timestamp = gethrestime_sec();
7201 /* Record ipf generation and account for frag header */
7202 ipf->ipf_gen = ill->ill_ipf_gen++;
7203 ipf->ipf_count = MBLKSIZE(mp1);
7204 ipf->ipf_last_frag_seen = B_FALSE;
7205 ipf->ipf_ecn = ecn_info;
7206 ipf->ipf_num_dups = 0;
7207 ipfb->ipfb_frag_pkts++;
7208 ipf->ipf_checksum = 0;
7209 ipf->ipf_checksum_flags = 0;
7210
7211 /* Store checksum value in fragment header */
7212 if (sum_flags != 0) {
7213 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7214 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7215 ipf->ipf_checksum = sum_val;
7216 ipf->ipf_checksum_flags = sum_flags;
7217 }
7218
7219 /*
7220 * We handle reassembly two ways. In the easy case,
7221 * where all the fragments show up in order, we do
7222 * minimal bookkeeping, and just clip new pieces on
7223 * the end. If we ever see a hole, then we go off
7224 * to ip_reassemble which has to mark the pieces and
7225 * keep track of the number of holes, etc. Obviously,
7226 * the point of having both mechanisms is so we can
7227 * handle the easy case as efficiently as possible.
7228 */
7229 if (offset == 0) {
7230 /* Easy case, in-order reassembly so far. */
7231 ipf->ipf_count += msg_len;
7232 ipf->ipf_tail_mp = tail_mp;
7233 /*
7234 * Keep track of next expected offset in
7235 * ipf_end.
7236 */
7237 ipf->ipf_end = end;
7238 ipf->ipf_nf_hdr_len = hdr_length;
7239 } else {
7240 /* Hard case, hole at the beginning. */
7241 ipf->ipf_tail_mp = NULL;
7242 /*
7243 * ipf_end == 0 means that we have given up
7244 * on easy reassembly.
7245 */
7246 ipf->ipf_end = 0;
7247
7248 /* Forget checksum offload from now on */
7249 ipf->ipf_checksum_flags = 0;
7250
7251 /*
7252 * ipf_hole_cnt is set by ip_reassemble.
7253 * ipf_count is updated by ip_reassemble.
7254 * No need to check for return value here
7255 * as we don't expect reassembly to complete
7256 * or fail for the first fragment itself.
7257 */
7258 (void) ip_reassemble(mp, ipf,
7259 (frag_offset_flags & IPH_OFFSET) << 3,
7260 (frag_offset_flags & IPH_MF), ill, msg_len);
7261 }
7262 /* Update per ipfb and ill byte counts */
7263 ipfb->ipfb_count += ipf->ipf_count;
7264 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7265 atomic_add_32(&ill->ill_frag_count, ipf->ipf_count);
7266 /* If the frag timer wasn't already going, start it. */
7267 mutex_enter(&ill->ill_lock);
7268 ill_frag_timer_start(ill);
7269 mutex_exit(&ill->ill_lock);
7270 goto reass_done;
7271 }
7272
7273 /*
7274 * If the packet's flag has changed (it could be coming up
7275 * from an interface different than the previous, therefore
7276 * possibly different checksum capability), then forget about
7277 * any stored checksum states. Otherwise add the value to
7278 * the existing one stored in the fragment header.
7279 */
7280 if (sum_flags != 0 && sum_flags == ipf->ipf_checksum_flags) {
7281 sum_val += ipf->ipf_checksum;
7282 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7283 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7284 ipf->ipf_checksum = sum_val;
7285 } else if (ipf->ipf_checksum_flags != 0) {
7286 /* Forget checksum offload from now on */
7287 ipf->ipf_checksum_flags = 0;
7288 }
7289
7290 /*
7291 * We have a new piece of a datagram which is already being
7292 * reassembled. Update the ECN info if all IP fragments
7293 * are ECN capable. If there is one which is not, clear
7294 * all the info. If there is at least one which has CE
7295 * code point, IP needs to report that up to transport.
7296 */
7297 if (ecn_info != IPH_ECN_NECT && ipf->ipf_ecn != IPH_ECN_NECT) {
7298 if (ecn_info == IPH_ECN_CE)
7299 ipf->ipf_ecn = IPH_ECN_CE;
7300 } else {
7301 ipf->ipf_ecn = IPH_ECN_NECT;
7302 }
7303 if (offset && ipf->ipf_end == offset) {
7304 /* The new fragment fits at the end */
7305 ipf->ipf_tail_mp->b_cont = mp;
7306 /* Update the byte count */
7307 ipf->ipf_count += msg_len;
7308 /* Update per ipfb and ill byte counts */
7309 ipfb->ipfb_count += msg_len;
7310 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7311 atomic_add_32(&ill->ill_frag_count, msg_len);
7312 if (frag_offset_flags & IPH_MF) {
7313 /* More to come. */
7314 ipf->ipf_end = end;
7315 ipf->ipf_tail_mp = tail_mp;
7316 goto reass_done;
7317 }
7318 } else {
7319 /* Go do the hard cases. */
7320 int ret;
7321
7322 if (offset == 0)
7323 ipf->ipf_nf_hdr_len = hdr_length;
7324
7325 /* Save current byte count */
7326 count = ipf->ipf_count;
7327 ret = ip_reassemble(mp, ipf,
7328 (frag_offset_flags & IPH_OFFSET) << 3,
7329 (frag_offset_flags & IPH_MF), ill, msg_len);
7330 /* Count of bytes added and subtracted (freeb()ed) */
7331 count = ipf->ipf_count - count;
7332 if (count) {
7333 /* Update per ipfb and ill byte counts */
7334 ipfb->ipfb_count += count;
7335 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7336 atomic_add_32(&ill->ill_frag_count, count);
7337 }
7338 if (ret == IP_REASS_PARTIAL) {
7339 goto reass_done;
7340 } else if (ret == IP_REASS_FAILED) {
7341 /* Reassembly failed. Free up all resources */
7342 ill_frag_free_pkts(ill, ipfb, ipf, 1);
7343 for (t_mp = mp; t_mp != NULL; t_mp = t_mp->b_cont) {
7344 IP_REASS_SET_START(t_mp, 0);
7345 IP_REASS_SET_END(t_mp, 0);
7346 }
7347 freemsg(mp);
7348 goto reass_done;
7349 }
7350 /* We will reach here iff 'ret' is IP_REASS_COMPLETE */
7351 }
7352 /*
7353 * We have completed reassembly. Unhook the frag header from
7354 * the reassembly list.
7355 *
7356 * Before we free the frag header, record the ECN info
7357 * to report back to the transport.
7358 */
7359 ecn_info = ipf->ipf_ecn;
7360 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmOKs);
7361 ipfp = ipf->ipf_ptphn;
7362
7363 /* We need to supply these to caller */
7364 if ((sum_flags = ipf->ipf_checksum_flags) != 0)
7365 sum_val = ipf->ipf_checksum;
7366 else
7367 sum_val = 0;
7368
7369 mp1 = ipf->ipf_mp;
7370 count = ipf->ipf_count;
7371 ipf = ipf->ipf_hash_next;
7372 if (ipf != NULL)
7373 ipf->ipf_ptphn = ipfp;
7374 ipfp[0] = ipf;
7375 atomic_add_32(&ill->ill_frag_count, -count);
7376 ASSERT(ipfb->ipfb_count >= count);
7377 ipfb->ipfb_count -= count;
7378 ipfb->ipfb_frag_pkts--;
7379 mutex_exit(&ipfb->ipfb_lock);
7380 /* Ditch the frag header. */
7381 mp = mp1->b_cont;
7382
7383 freeb(mp1);
7384
7385 /* Restore original IP length in header. */
7386 packet_size = (uint32_t)msgdsize(mp);
7387 if (packet_size > IP_MAXPACKET) {
7388 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7389 ip_drop_input("Reassembled packet too large", mp, ill);
7390 freemsg(mp);
7391 return (NULL);
7392 }
7393
7394 if (DB_REF(mp) > 1) {
7395 mblk_t *mp2 = copymsg(mp);
7396
7397 if (mp2 == NULL) {
7398 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7399 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7400 freemsg(mp);
7401 return (NULL);
7402 }
7403 freemsg(mp);
7404 mp = mp2;
7405 }
7406 ipha = (ipha_t *)mp->b_rptr;
7407
7408 ipha->ipha_length = htons((uint16_t)packet_size);
7409 /* We're now complete, zip the frag state */
7410 ipha->ipha_fragment_offset_and_flags = 0;
7411 /* Record the ECN info. */
7412 ipha->ipha_type_of_service &= 0xFC;
7413 ipha->ipha_type_of_service |= ecn_info;
7414
7415 /* Update the receive attributes */
7416 ira->ira_pktlen = packet_size;
7417 ira->ira_ip_hdr_length = IPH_HDR_LENGTH(ipha);
7418
7419 /* Reassembly is successful; set checksum information in packet */
7420 DB_CKSUM16(mp) = (uint16_t)sum_val;
7421 DB_CKSUMFLAGS(mp) = sum_flags;
7422 DB_CKSUMSTART(mp) = ira->ira_ip_hdr_length;
7423
7424 return (mp);
7425 }
7426
7427 /*
7428 * Pullup function that should be used for IP input in order to
7429 * ensure we do not loose the L2 source address; we need the l2 source
7430 * address for IP_RECVSLLA and for ndp_input.
7431 *
7432 * We return either NULL or b_rptr.
7433 */
7434 void *
7435 ip_pullup(mblk_t *mp, ssize_t len, ip_recv_attr_t *ira)
7436 {
7437 ill_t *ill = ira->ira_ill;
7438
7439 if (ip_rput_pullups++ == 0) {
7440 (void) mi_strlog(ill->ill_rq, 1, SL_ERROR|SL_TRACE,
7441 "ip_pullup: %s forced us to "
7442 " pullup pkt, hdr len %ld, hdr addr %p",
7443 ill->ill_name, len, (void *)mp->b_rptr);
7444 }
7445 if (!(ira->ira_flags & IRAF_L2SRC_SET))
7446 ip_setl2src(mp, ira, ira->ira_rill);
7447 ASSERT(ira->ira_flags & IRAF_L2SRC_SET);
7448 if (!pullupmsg(mp, len))
7449 return (NULL);
7450 else
7451 return (mp->b_rptr);
7452 }
7453
7454 /*
7455 * Make sure ira_l2src has an address. If we don't have one fill with zeros.
7456 * When called from the ULP ira_rill will be NULL hence the caller has to
7457 * pass in the ill.
7458 */
7459 /* ARGSUSED */
7460 void
7461 ip_setl2src(mblk_t *mp, ip_recv_attr_t *ira, ill_t *ill)
7462 {
7463 const uchar_t *addr;
7464 int alen;
7465
7466 if (ira->ira_flags & IRAF_L2SRC_SET)
7467 return;
7468
7469 ASSERT(ill != NULL);
7470 alen = ill->ill_phys_addr_length;
7471 ASSERT(alen <= sizeof (ira->ira_l2src));
7472 if (ira->ira_mhip != NULL &&
7473 (addr = ira->ira_mhip->mhi_saddr) != NULL) {
7474 bcopy(addr, ira->ira_l2src, alen);
7475 } else if ((ira->ira_flags & IRAF_L2SRC_LOOPBACK) &&
7476 (addr = ill->ill_phys_addr) != NULL) {
7477 bcopy(addr, ira->ira_l2src, alen);
7478 } else {
7479 bzero(ira->ira_l2src, alen);
7480 }
7481 ira->ira_flags |= IRAF_L2SRC_SET;
7482 }
7483
7484 /*
7485 * check ip header length and align it.
7486 */
7487 mblk_t *
7488 ip_check_and_align_header(mblk_t *mp, uint_t min_size, ip_recv_attr_t *ira)
7489 {
7490 ill_t *ill = ira->ira_ill;
7491 ssize_t len;
7492
7493 len = MBLKL(mp);
7494
7495 if (!OK_32PTR(mp->b_rptr))
7496 IP_STAT(ill->ill_ipst, ip_notaligned);
7497 else
7498 IP_STAT(ill->ill_ipst, ip_recv_pullup);
7499
7500 /* Guard against bogus device drivers */
7501 if (len < 0) {
7502 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7503 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7504 freemsg(mp);
7505 return (NULL);
7506 }
7507
7508 if (len == 0) {
7509 /* GLD sometimes sends up mblk with b_rptr == b_wptr! */
7510 mblk_t *mp1 = mp->b_cont;
7511
7512 if (!(ira->ira_flags & IRAF_L2SRC_SET))
7513 ip_setl2src(mp, ira, ira->ira_rill);
7514 ASSERT(ira->ira_flags & IRAF_L2SRC_SET);
7515
7516 freeb(mp);
7517 mp = mp1;
7518 if (mp == NULL)
7519 return (NULL);
7520
7521 if (OK_32PTR(mp->b_rptr) && MBLKL(mp) >= min_size)
7522 return (mp);
7523 }
7524 if (ip_pullup(mp, min_size, ira) == NULL) {
7525 if (msgdsize(mp) < min_size) {
7526 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7527 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7528 } else {
7529 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7530 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7531 }
7532 freemsg(mp);
7533 return (NULL);
7534 }
7535 return (mp);
7536 }
7537
7538 /*
7539 * Common code for IPv4 and IPv6 to check and pullup multi-mblks
7540 */
7541 mblk_t *
7542 ip_check_length(mblk_t *mp, uchar_t *rptr, ssize_t len, uint_t pkt_len,
7543 uint_t min_size, ip_recv_attr_t *ira)
7544 {
7545 ill_t *ill = ira->ira_ill;
7546
7547 /*
7548 * Make sure we have data length consistent
7549 * with the IP header.
7550 */
7551 if (mp->b_cont == NULL) {
7552 /* pkt_len is based on ipha_len, not the mblk length */
7553 if (pkt_len < min_size) {
7554 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7555 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7556 freemsg(mp);
7557 return (NULL);
7558 }
7559 if (len < 0) {
7560 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
7561 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
7562 freemsg(mp);
7563 return (NULL);
7564 }
7565 /* Drop any pad */
7566 mp->b_wptr = rptr + pkt_len;
7567 } else if ((len += msgdsize(mp->b_cont)) != 0) {
7568 ASSERT(pkt_len >= min_size);
7569 if (pkt_len < min_size) {
7570 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7571 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7572 freemsg(mp);
7573 return (NULL);
7574 }
7575 if (len < 0) {
7576 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
7577 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
7578 freemsg(mp);
7579 return (NULL);
7580 }
7581 /* Drop any pad */
7582 (void) adjmsg(mp, -len);
7583 /*
7584 * adjmsg may have freed an mblk from the chain, hence
7585 * invalidate any hw checksum here. This will force IP to
7586 * calculate the checksum in sw, but only for this packet.
7587 */
7588 DB_CKSUMFLAGS(mp) = 0;
7589 IP_STAT(ill->ill_ipst, ip_multimblk);
7590 }
7591 return (mp);
7592 }
7593
7594 /*
7595 * Check that the IPv4 opt_len is consistent with the packet and pullup
7596 * the options.
7597 */
7598 mblk_t *
7599 ip_check_optlen(mblk_t *mp, ipha_t *ipha, uint_t opt_len, uint_t pkt_len,
7600 ip_recv_attr_t *ira)
7601 {
7602 ill_t *ill = ira->ira_ill;
7603 ssize_t len;
7604
7605 /* Assume no IPv6 packets arrive over the IPv4 queue */
7606 if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
7607 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7608 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInWrongIPVersion);
7609 ip_drop_input("IPvN packet on IPv4 ill", mp, ill);
7610 freemsg(mp);
7611 return (NULL);
7612 }
7613
7614 if (opt_len > (15 - IP_SIMPLE_HDR_LENGTH_IN_WORDS)) {
7615 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7616 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7617 freemsg(mp);
7618 return (NULL);
7619 }
7620 /*
7621 * Recompute complete header length and make sure we
7622 * have access to all of it.
7623 */
7624 len = ((size_t)opt_len + IP_SIMPLE_HDR_LENGTH_IN_WORDS) << 2;
7625 if (len > (mp->b_wptr - mp->b_rptr)) {
7626 if (len > pkt_len) {
7627 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7628 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7629 freemsg(mp);
7630 return (NULL);
7631 }
7632 if (ip_pullup(mp, len, ira) == NULL) {
7633 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7634 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7635 freemsg(mp);
7636 return (NULL);
7637 }
7638 }
7639 return (mp);
7640 }
7641
7642 /*
7643 * Returns a new ire, or the same ire, or NULL.
7644 * If a different IRE is returned, then it is held; the caller
7645 * needs to release it.
7646 * In no case is there any hold/release on the ire argument.
7647 */
7648 ire_t *
7649 ip_check_multihome(void *addr, ire_t *ire, ill_t *ill)
7650 {
7651 ire_t *new_ire;
7652 ill_t *ire_ill;
7653 uint_t ifindex;
7654 ip_stack_t *ipst = ill->ill_ipst;
7655 boolean_t strict_check = B_FALSE;
7656
7657 /*
7658 * IPMP common case: if IRE and ILL are in the same group, there's no
7659 * issue (e.g. packet received on an underlying interface matched an
7660 * IRE_LOCAL on its associated group interface).
7661 */
7662 ASSERT(ire->ire_ill != NULL);
7663 if (IS_IN_SAME_ILLGRP(ill, ire->ire_ill))
7664 return (ire);
7665
7666 /*
7667 * Do another ire lookup here, using the ingress ill, to see if the
7668 * interface is in a usesrc group.
7669 * As long as the ills belong to the same group, we don't consider
7670 * them to be arriving on the wrong interface. Thus, if the switch
7671 * is doing inbound load spreading, we won't drop packets when the
7672 * ip*_strict_dst_multihoming switch is on.
7673 * We also need to check for IPIF_UNNUMBERED point2point interfaces
7674 * where the local address may not be unique. In this case we were
7675 * at the mercy of the initial ire lookup and the IRE_LOCAL it
7676 * actually returned. The new lookup, which is more specific, should
7677 * only find the IRE_LOCAL associated with the ingress ill if one
7678 * exists.
7679 */
7680 if (ire->ire_ipversion == IPV4_VERSION) {
7681 if (ipst->ips_ip_strict_dst_multihoming)
7682 strict_check = B_TRUE;
7683 new_ire = ire_ftable_lookup_v4(*((ipaddr_t *)addr), 0, 0,
7684 IRE_LOCAL, ill, ALL_ZONES, NULL,
7685 (MATCH_IRE_TYPE|MATCH_IRE_ILL), 0, ipst, NULL);
7686 } else {
7687 ASSERT(!IN6_IS_ADDR_MULTICAST((in6_addr_t *)addr));
7688 if (ipst->ips_ipv6_strict_dst_multihoming)
7689 strict_check = B_TRUE;
7690 new_ire = ire_ftable_lookup_v6((in6_addr_t *)addr, NULL, NULL,
7691 IRE_LOCAL, ill, ALL_ZONES, NULL,
7692 (MATCH_IRE_TYPE|MATCH_IRE_ILL), 0, ipst, NULL);
7693 }
7694 /*
7695 * If the same ire that was returned in ip_input() is found then this
7696 * is an indication that usesrc groups are in use. The packet
7697 * arrived on a different ill in the group than the one associated with
7698 * the destination address. If a different ire was found then the same
7699 * IP address must be hosted on multiple ills. This is possible with
7700 * unnumbered point2point interfaces. We switch to use this new ire in
7701 * order to have accurate interface statistics.
7702 */
7703 if (new_ire != NULL) {
7704 /* Note: held in one case but not the other? Caller handles */
7705 if (new_ire != ire)
7706 return (new_ire);
7707 /* Unchanged */
7708 ire_refrele(new_ire);
7709 return (ire);
7710 }
7711
7712 /*
7713 * Chase pointers once and store locally.
7714 */
7715 ASSERT(ire->ire_ill != NULL);
7716 ire_ill = ire->ire_ill;
7717 ifindex = ill->ill_usesrc_ifindex;
7718
7719 /*
7720 * Check if it's a legal address on the 'usesrc' interface.
7721 * For IPMP data addresses the IRE_LOCAL is the upper, hence we
7722 * can just check phyint_ifindex.
7723 */
7724 if (ifindex != 0 && ifindex == ire_ill->ill_phyint->phyint_ifindex) {
7725 return (ire);
7726 }
7727
7728 /*
7729 * If the ip*_strict_dst_multihoming switch is on then we can
7730 * only accept this packet if the interface is marked as routing.
7731 */
7732 if (!(strict_check))
7733 return (ire);
7734
7735 if ((ill->ill_flags & ire->ire_ill->ill_flags & ILLF_ROUTER) != 0) {
7736 return (ire);
7737 }
7738 return (NULL);
7739 }
7740
7741 /*
7742 * This function is used to construct a mac_header_info_s from a
7743 * DL_UNITDATA_IND message.
7744 * The address fields in the mhi structure points into the message,
7745 * thus the caller can't use those fields after freeing the message.
7746 *
7747 * We determine whether the packet received is a non-unicast packet
7748 * and in doing so, determine whether or not it is broadcast vs multicast.
7749 * For it to be a broadcast packet, we must have the appropriate mblk_t
7750 * hanging off the ill_t. If this is either not present or doesn't match
7751 * the destination mac address in the DL_UNITDATA_IND, the packet is deemed
7752 * to be multicast. Thus NICs that have no broadcast address (or no
7753 * capability for one, such as point to point links) cannot return as
7754 * the packet being broadcast.
7755 */
7756 void
7757 ip_dlur_to_mhi(ill_t *ill, mblk_t *mb, struct mac_header_info_s *mhip)
7758 {
7759 dl_unitdata_ind_t *ind = (dl_unitdata_ind_t *)mb->b_rptr;
7760 mblk_t *bmp;
7761 uint_t extra_offset;
7762
7763 bzero(mhip, sizeof (struct mac_header_info_s));
7764
7765 mhip->mhi_dsttype = MAC_ADDRTYPE_UNICAST;
7766
7767 if (ill->ill_sap_length < 0)
7768 extra_offset = 0;
7769 else
7770 extra_offset = ill->ill_sap_length;
7771
7772 mhip->mhi_daddr = (uchar_t *)ind + ind->dl_dest_addr_offset +
7773 extra_offset;
7774 mhip->mhi_saddr = (uchar_t *)ind + ind->dl_src_addr_offset +
7775 extra_offset;
7776
7777 if (!ind->dl_group_address)
7778 return;
7779
7780 /* Multicast or broadcast */
7781 mhip->mhi_dsttype = MAC_ADDRTYPE_MULTICAST;
7782
7783 if (ind->dl_dest_addr_offset > sizeof (*ind) &&
7784 ind->dl_dest_addr_offset + ind->dl_dest_addr_length < MBLKL(mb) &&
7785 (bmp = ill->ill_bcast_mp) != NULL) {
7786 dl_unitdata_req_t *dlur;
7787 uint8_t *bphys_addr;
7788
7789 dlur = (dl_unitdata_req_t *)bmp->b_rptr;
7790 bphys_addr = (uchar_t *)dlur + dlur->dl_dest_addr_offset +
7791 extra_offset;
7792
7793 if (bcmp(mhip->mhi_daddr, bphys_addr,
7794 ind->dl_dest_addr_length) == 0)
7795 mhip->mhi_dsttype = MAC_ADDRTYPE_BROADCAST;
7796 }
7797 }
7798
7799 /*
7800 * This function is used to construct a mac_header_info_s from a
7801 * M_DATA fastpath message from a DLPI driver.
7802 * The address fields in the mhi structure points into the message,
7803 * thus the caller can't use those fields after freeing the message.
7804 *
7805 * We determine whether the packet received is a non-unicast packet
7806 * and in doing so, determine whether or not it is broadcast vs multicast.
7807 * For it to be a broadcast packet, we must have the appropriate mblk_t
7808 * hanging off the ill_t. If this is either not present or doesn't match
7809 * the destination mac address in the DL_UNITDATA_IND, the packet is deemed
7810 * to be multicast. Thus NICs that have no broadcast address (or no
7811 * capability for one, such as point to point links) cannot return as
7812 * the packet being broadcast.
7813 */
7814 void
7815 ip_mdata_to_mhi(ill_t *ill, mblk_t *mp, struct mac_header_info_s *mhip)
7816 {
7817 mblk_t *bmp;
7818 struct ether_header *pether;
7819
7820 bzero(mhip, sizeof (struct mac_header_info_s));
7821
7822 mhip->mhi_dsttype = MAC_ADDRTYPE_UNICAST;
7823
7824 pether = (struct ether_header *)((char *)mp->b_rptr
7825 - sizeof (struct ether_header));
7826
7827 /*
7828 * Make sure the interface is an ethernet type, since we don't
7829 * know the header format for anything but Ethernet. Also make
7830 * sure we are pointing correctly above db_base.
7831 */
7832 if (ill->ill_type != IFT_ETHER)
7833 return;
7834
7835 retry:
7836 if ((uchar_t *)pether < mp->b_datap->db_base)
7837 return;
7838
7839 /* Is there a VLAN tag? */
7840 if (ill->ill_isv6) {
7841 if (pether->ether_type != htons(ETHERTYPE_IPV6)) {
7842 pether = (struct ether_header *)((char *)pether - 4);
7843 goto retry;
7844 }
7845 } else {
7846 if (pether->ether_type != htons(ETHERTYPE_IP)) {
7847 pether = (struct ether_header *)((char *)pether - 4);
7848 goto retry;
7849 }
7850 }
7851 mhip->mhi_daddr = (uchar_t *)&pether->ether_dhost;
7852 mhip->mhi_saddr = (uchar_t *)&pether->ether_shost;
7853
7854 if (!(mhip->mhi_daddr[0] & 0x01))
7855 return;
7856
7857 /* Multicast or broadcast */
7858 mhip->mhi_dsttype = MAC_ADDRTYPE_MULTICAST;
7859
7860 if ((bmp = ill->ill_bcast_mp) != NULL) {
7861 dl_unitdata_req_t *dlur;
7862 uint8_t *bphys_addr;
7863 uint_t addrlen;
7864
7865 dlur = (dl_unitdata_req_t *)bmp->b_rptr;
7866 addrlen = dlur->dl_dest_addr_length;
7867 if (ill->ill_sap_length < 0) {
7868 bphys_addr = (uchar_t *)dlur +
7869 dlur->dl_dest_addr_offset;
7870 addrlen += ill->ill_sap_length;
7871 } else {
7872 bphys_addr = (uchar_t *)dlur +
7873 dlur->dl_dest_addr_offset +
7874 ill->ill_sap_length;
7875 addrlen -= ill->ill_sap_length;
7876 }
7877 if (bcmp(mhip->mhi_daddr, bphys_addr, addrlen) == 0)
7878 mhip->mhi_dsttype = MAC_ADDRTYPE_BROADCAST;
7879 }
7880 }
7881
7882 /*
7883 * Handle anything but M_DATA messages
7884 * We see the DL_UNITDATA_IND which are part
7885 * of the data path, and also the other messages from the driver.
7886 */
7887 void
7888 ip_rput_notdata(ill_t *ill, mblk_t *mp)
7889 {
7890 mblk_t *first_mp;
7891 struct iocblk *iocp;
7892 struct mac_header_info_s mhi;
7893
7894 switch (DB_TYPE(mp)) {
7895 case M_PROTO:
7896 case M_PCPROTO: {
7897 if (((dl_unitdata_ind_t *)mp->b_rptr)->dl_primitive !=
7898 DL_UNITDATA_IND) {
7899 /* Go handle anything other than data elsewhere. */
7900 ip_rput_dlpi(ill, mp);
7901 return;
7902 }
7903
7904 first_mp = mp;
7905 mp = first_mp->b_cont;
7906 first_mp->b_cont = NULL;
7907
7908 if (mp == NULL) {
7909 freeb(first_mp);
7910 return;
7911 }
7912 ip_dlur_to_mhi(ill, first_mp, &mhi);
7913 if (ill->ill_isv6)
7914 ip_input_v6(ill, NULL, mp, &mhi);
7915 else
7916 ip_input(ill, NULL, mp, &mhi);
7917
7918 /* Ditch the DLPI header. */
7919 freeb(first_mp);
7920 return;
7921 }
7922 case M_IOCACK:
7923 iocp = (struct iocblk *)mp->b_rptr;
7924 switch (iocp->ioc_cmd) {
7925 case DL_IOC_HDR_INFO:
7926 ill_fastpath_ack(ill, mp);
7927 return;
7928 default:
7929 putnext(ill->ill_rq, mp);
7930 return;
7931 }
7932 /* FALLTHRU */
7933 case M_ERROR:
7934 case M_HANGUP:
7935 mutex_enter(&ill->ill_lock);
7936 if (ill->ill_state_flags & ILL_CONDEMNED) {
7937 mutex_exit(&ill->ill_lock);
7938 freemsg(mp);
7939 return;
7940 }
7941 ill_refhold_locked(ill);
7942 mutex_exit(&ill->ill_lock);
7943 qwriter_ip(ill, ill->ill_rq, mp, ip_rput_other, CUR_OP,
7944 B_FALSE);
7945 return;
7946 case M_CTL:
7947 putnext(ill->ill_rq, mp);
7948 return;
7949 case M_IOCNAK:
7950 ip1dbg(("got iocnak "));
7951 iocp = (struct iocblk *)mp->b_rptr;
7952 switch (iocp->ioc_cmd) {
7953 case DL_IOC_HDR_INFO:
7954 ip_rput_other(NULL, ill->ill_rq, mp, NULL);
7955 return;
7956 default:
7957 break;
7958 }
7959 /* FALLTHRU */
7960 default:
7961 putnext(ill->ill_rq, mp);
7962 return;
7963 }
7964 }
7965
7966 /* Read side put procedure. Packets coming from the wire arrive here. */
7967 void
7968 ip_rput(queue_t *q, mblk_t *mp)
7969 {
7970 ill_t *ill;
7971 union DL_primitives *dl;
7972
7973 ill = (ill_t *)q->q_ptr;
7974
7975 if (ill->ill_state_flags & (ILL_CONDEMNED | ILL_LL_SUBNET_PENDING)) {
7976 /*
7977 * If things are opening or closing, only accept high-priority
7978 * DLPI messages. (On open ill->ill_ipif has not yet been
7979 * created; on close, things hanging off the ill may have been
7980 * freed already.)
7981 */
7982 dl = (union DL_primitives *)mp->b_rptr;
7983 if (DB_TYPE(mp) != M_PCPROTO ||
7984 dl->dl_primitive == DL_UNITDATA_IND) {
7985 inet_freemsg(mp);
7986 return;
7987 }
7988 }
7989 if (DB_TYPE(mp) == M_DATA) {
7990 struct mac_header_info_s mhi;
7991
7992 ip_mdata_to_mhi(ill, mp, &mhi);
7993 ip_input(ill, NULL, mp, &mhi);
7994 } else {
7995 ip_rput_notdata(ill, mp);
7996 }
7997 }
7998
7999 /*
8000 * Move the information to a copy.
8001 */
8002 mblk_t *
8003 ip_fix_dbref(mblk_t *mp, ip_recv_attr_t *ira)
8004 {
8005 mblk_t *mp1;
8006 ill_t *ill = ira->ira_ill;
8007 ip_stack_t *ipst = ill->ill_ipst;
8008
8009 IP_STAT(ipst, ip_db_ref);
8010
8011 /* Make sure we have ira_l2src before we loose the original mblk */
8012 if (!(ira->ira_flags & IRAF_L2SRC_SET))
8013 ip_setl2src(mp, ira, ira->ira_rill);
8014
8015 mp1 = copymsg(mp);
8016 if (mp1 == NULL) {
8017 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
8018 ip_drop_input("ipIfStatsInDiscards", mp, ill);
8019 freemsg(mp);
8020 return (NULL);
8021 }
8022 /* preserve the hardware checksum flags and data, if present */
8023 if (DB_CKSUMFLAGS(mp) != 0) {
8024 DB_CKSUMFLAGS(mp1) = DB_CKSUMFLAGS(mp);
8025 DB_CKSUMSTART(mp1) = DB_CKSUMSTART(mp);
8026 DB_CKSUMSTUFF(mp1) = DB_CKSUMSTUFF(mp);
8027 DB_CKSUMEND(mp1) = DB_CKSUMEND(mp);
8028 DB_CKSUM16(mp1) = DB_CKSUM16(mp);
8029 }
8030 freemsg(mp);
8031 return (mp1);
8032 }
8033
8034 static void
8035 ip_dlpi_error(ill_t *ill, t_uscalar_t prim, t_uscalar_t dl_err,
8036 t_uscalar_t err)
8037 {
8038 if (dl_err == DL_SYSERR) {
8039 (void) mi_strlog(ill->ill_rq, 1, SL_CONSOLE|SL_ERROR|SL_TRACE,
8040 "%s: %s failed: DL_SYSERR (errno %u)\n",
8041 ill->ill_name, dl_primstr(prim), err);
8042 return;
8043 }
8044
8045 (void) mi_strlog(ill->ill_rq, 1, SL_CONSOLE|SL_ERROR|SL_TRACE,
8046 "%s: %s failed: %s\n", ill->ill_name, dl_primstr(prim),
8047 dl_errstr(dl_err));
8048 }
8049
8050 /*
8051 * ip_rput_dlpi is called by ip_rput to handle all DLPI messages other
8052 * than DL_UNITDATA_IND messages. If we need to process this message
8053 * exclusively, we call qwriter_ip, in which case we also need to call
8054 * ill_refhold before that, since qwriter_ip does an ill_refrele.
8055 */
8056 void
8057 ip_rput_dlpi(ill_t *ill, mblk_t *mp)
8058 {
8059 dl_ok_ack_t *dloa = (dl_ok_ack_t *)mp->b_rptr;
8060 dl_error_ack_t *dlea = (dl_error_ack_t *)dloa;
8061 queue_t *q = ill->ill_rq;
8062 t_uscalar_t prim = dloa->dl_primitive;
8063 t_uscalar_t reqprim = DL_PRIM_INVAL;
8064
8065 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi",
8066 char *, dl_primstr(prim), ill_t *, ill);
8067 ip1dbg(("ip_rput_dlpi"));
8068
8069 /*
8070 * If we received an ACK but didn't send a request for it, then it
8071 * can't be part of any pending operation; discard up-front.
8072 */
8073 switch (prim) {
8074 case DL_ERROR_ACK:
8075 reqprim = dlea->dl_error_primitive;
8076 ip2dbg(("ip_rput_dlpi(%s): DL_ERROR_ACK for %s (0x%x): %s "
8077 "(0x%x), unix %u\n", ill->ill_name, dl_primstr(reqprim),
8078 reqprim, dl_errstr(dlea->dl_errno), dlea->dl_errno,
8079 dlea->dl_unix_errno));
8080 break;
8081 case DL_OK_ACK:
8082 reqprim = dloa->dl_correct_primitive;
8083 break;
8084 case DL_INFO_ACK:
8085 reqprim = DL_INFO_REQ;
8086 break;
8087 case DL_BIND_ACK:
8088 reqprim = DL_BIND_REQ;
8089 break;
8090 case DL_PHYS_ADDR_ACK:
8091 reqprim = DL_PHYS_ADDR_REQ;
8092 break;
8093 case DL_NOTIFY_ACK:
8094 reqprim = DL_NOTIFY_REQ;
8095 break;
8096 case DL_CAPABILITY_ACK:
8097 reqprim = DL_CAPABILITY_REQ;
8098 break;
8099 }
8100
8101 if (prim != DL_NOTIFY_IND) {
8102 if (reqprim == DL_PRIM_INVAL ||
8103 !ill_dlpi_pending(ill, reqprim)) {
8104 /* Not a DLPI message we support or expected */
8105 freemsg(mp);
8106 return;
8107 }
8108 ip1dbg(("ip_rput: received %s for %s\n", dl_primstr(prim),
8109 dl_primstr(reqprim)));
8110 }
8111
8112 switch (reqprim) {
8113 case DL_UNBIND_REQ:
8114 /*
8115 * NOTE: we mark the unbind as complete even if we got a
8116 * DL_ERROR_ACK, since there's not much else we can do.
8117 */
8118 mutex_enter(&ill->ill_lock);
8119 ill->ill_state_flags &= ~ILL_DL_UNBIND_IN_PROGRESS;
8120 cv_signal(&ill->ill_cv);
8121 mutex_exit(&ill->ill_lock);
8122 break;
8123
8124 case DL_ENABMULTI_REQ:
8125 if (prim == DL_OK_ACK) {
8126 if (ill->ill_dlpi_multicast_state == IDS_INPROGRESS)
8127 ill->ill_dlpi_multicast_state = IDS_OK;
8128 }
8129 break;
8130 }
8131
8132 /*
8133 * The message is one we're waiting for (or DL_NOTIFY_IND), but we
8134 * need to become writer to continue to process it. Because an
8135 * exclusive operation doesn't complete until replies to all queued
8136 * DLPI messages have been received, we know we're in the middle of an
8137 * exclusive operation and pass CUR_OP (except for DL_NOTIFY_IND).
8138 *
8139 * As required by qwriter_ip(), we refhold the ill; it will refrele.
8140 * Since this is on the ill stream we unconditionally bump up the
8141 * refcount without doing ILL_CAN_LOOKUP().
8142 */
8143 ill_refhold(ill);
8144 if (prim == DL_NOTIFY_IND)
8145 qwriter_ip(ill, q, mp, ip_rput_dlpi_writer, NEW_OP, B_FALSE);
8146 else
8147 qwriter_ip(ill, q, mp, ip_rput_dlpi_writer, CUR_OP, B_FALSE);
8148 }
8149
8150 /*
8151 * Handling of DLPI messages that require exclusive access to the ipsq.
8152 *
8153 * Need to do ipsq_pending_mp_get on ioctl completion, which could
8154 * happen here. (along with mi_copy_done)
8155 */
8156 /* ARGSUSED */
8157 static void
8158 ip_rput_dlpi_writer(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
8159 {
8160 dl_ok_ack_t *dloa = (dl_ok_ack_t *)mp->b_rptr;
8161 dl_error_ack_t *dlea = (dl_error_ack_t *)dloa;
8162 int err = 0;
8163 ill_t *ill = (ill_t *)q->q_ptr;
8164 ipif_t *ipif = NULL;
8165 mblk_t *mp1 = NULL;
8166 conn_t *connp = NULL;
8167 t_uscalar_t paddrreq;
8168 mblk_t *mp_hw;
8169 boolean_t success;
8170 boolean_t ioctl_aborted = B_FALSE;
8171 boolean_t log = B_TRUE;
8172
8173 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer",
8174 char *, dl_primstr(dloa->dl_primitive), ill_t *, ill);
8175
8176 ip1dbg(("ip_rput_dlpi_writer .."));
8177 ASSERT(ipsq->ipsq_xop == ill->ill_phyint->phyint_ipsq->ipsq_xop);
8178 ASSERT(IAM_WRITER_ILL(ill));
8179
8180 ipif = ipsq->ipsq_xop->ipx_pending_ipif;
8181 /*
8182 * The current ioctl could have been aborted by the user and a new
8183 * ioctl to bring up another ill could have started. We could still
8184 * get a response from the driver later.
8185 */
8186 if (ipif != NULL && ipif->ipif_ill != ill)
8187 ioctl_aborted = B_TRUE;
8188
8189 switch (dloa->dl_primitive) {
8190 case DL_ERROR_ACK:
8191 ip1dbg(("ip_rput_dlpi_writer: got DL_ERROR_ACK for %s\n",
8192 dl_primstr(dlea->dl_error_primitive)));
8193
8194 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer error",
8195 char *, dl_primstr(dlea->dl_error_primitive),
8196 ill_t *, ill);
8197
8198 switch (dlea->dl_error_primitive) {
8199 case DL_DISABMULTI_REQ:
8200 ill_dlpi_done(ill, dlea->dl_error_primitive);
8201 break;
8202 case DL_PROMISCON_REQ:
8203 case DL_PROMISCOFF_REQ:
8204 case DL_UNBIND_REQ:
8205 case DL_ATTACH_REQ:
8206 case DL_INFO_REQ:
8207 ill_dlpi_done(ill, dlea->dl_error_primitive);
8208 break;
8209 case DL_NOTIFY_REQ:
8210 ill_dlpi_done(ill, DL_NOTIFY_REQ);
8211 log = B_FALSE;
8212 break;
8213 case DL_PHYS_ADDR_REQ:
8214 /*
8215 * For IPv6 only, there are two additional
8216 * phys_addr_req's sent to the driver to get the
8217 * IPv6 token and lla. This allows IP to acquire
8218 * the hardware address format for a given interface
8219 * without having built in knowledge of the hardware
8220 * address. ill_phys_addr_pend keeps track of the last
8221 * DL_PAR sent so we know which response we are
8222 * dealing with. ill_dlpi_done will update
8223 * ill_phys_addr_pend when it sends the next req.
8224 * We don't complete the IOCTL until all three DL_PARs
8225 * have been attempted, so set *_len to 0 and break.
8226 */
8227 paddrreq = ill->ill_phys_addr_pend;
8228 ill_dlpi_done(ill, DL_PHYS_ADDR_REQ);
8229 if (paddrreq == DL_IPV6_TOKEN) {
8230 ill->ill_token_length = 0;
8231 log = B_FALSE;
8232 break;
8233 } else if (paddrreq == DL_IPV6_LINK_LAYER_ADDR) {
8234 ill->ill_nd_lla_len = 0;
8235 log = B_FALSE;
8236 break;
8237 }
8238 /*
8239 * Something went wrong with the DL_PHYS_ADDR_REQ.
8240 * We presumably have an IOCTL hanging out waiting
8241 * for completion. Find it and complete the IOCTL
8242 * with the error noted.
8243 * However, ill_dl_phys was called on an ill queue
8244 * (from SIOCSLIFNAME), thus conn_pending_ill is not
8245 * set. But the ioctl is known to be pending on ill_wq.
8246 */
8247 if (!ill->ill_ifname_pending)
8248 break;
8249 ill->ill_ifname_pending = 0;
8250 if (!ioctl_aborted)
8251 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8252 if (mp1 != NULL) {
8253 /*
8254 * This operation (SIOCSLIFNAME) must have
8255 * happened on the ill. Assert there is no conn
8256 */
8257 ASSERT(connp == NULL);
8258 q = ill->ill_wq;
8259 }
8260 break;
8261 case DL_BIND_REQ:
8262 ill_dlpi_done(ill, DL_BIND_REQ);
8263 if (ill->ill_ifname_pending)
8264 break;
8265 mutex_enter(&ill->ill_lock);
8266 ill->ill_state_flags &= ~ILL_DOWN_IN_PROGRESS;
8267 mutex_exit(&ill->ill_lock);
8268 /*
8269 * Something went wrong with the bind. We presumably
8270 * have an IOCTL hanging out waiting for completion.
8271 * Find it, take down the interface that was coming
8272 * up, and complete the IOCTL with the error noted.
8273 */
8274 if (!ioctl_aborted)
8275 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8276 if (mp1 != NULL) {
8277 /*
8278 * This might be a result of a DL_NOTE_REPLUMB
8279 * notification. In that case, connp is NULL.
8280 */
8281 if (connp != NULL)
8282 q = CONNP_TO_WQ(connp);
8283
8284 (void) ipif_down(ipif, NULL, NULL);
8285 /* error is set below the switch */
8286 }
8287 break;
8288 case DL_ENABMULTI_REQ:
8289 ill_dlpi_done(ill, DL_ENABMULTI_REQ);
8290
8291 if (ill->ill_dlpi_multicast_state == IDS_INPROGRESS)
8292 ill->ill_dlpi_multicast_state = IDS_FAILED;
8293 if (ill->ill_dlpi_multicast_state == IDS_FAILED) {
8294
8295 printf("ip: joining multicasts failed (%d)"
8296 " on %s - will use link layer "
8297 "broadcasts for multicast\n",
8298 dlea->dl_errno, ill->ill_name);
8299
8300 /*
8301 * Set up for multi_bcast; We are the
8302 * writer, so ok to access ill->ill_ipif
8303 * without any lock.
8304 */
8305 mutex_enter(&ill->ill_phyint->phyint_lock);
8306 ill->ill_phyint->phyint_flags |=
8307 PHYI_MULTI_BCAST;
8308 mutex_exit(&ill->ill_phyint->phyint_lock);
8309
8310 }
8311 freemsg(mp); /* Don't want to pass this up */
8312 return;
8313 case DL_CAPABILITY_REQ:
8314 ip1dbg(("ip_rput_dlpi_writer: got DL_ERROR_ACK for "
8315 "DL_CAPABILITY REQ\n"));
8316 if (ill->ill_dlpi_capab_state == IDCS_PROBE_SENT)
8317 ill->ill_dlpi_capab_state = IDCS_FAILED;
8318 ill_capability_done(ill);
8319 freemsg(mp);
8320 return;
8321 }
8322 /*
8323 * Note the error for IOCTL completion (mp1 is set when
8324 * ready to complete ioctl). If ill_ifname_pending_err is
8325 * set, an error occured during plumbing (ill_ifname_pending),
8326 * so we want to report that error.
8327 *
8328 * NOTE: there are two addtional DL_PHYS_ADDR_REQ's
8329 * (DL_IPV6_TOKEN and DL_IPV6_LINK_LAYER_ADDR) that are
8330 * expected to get errack'd if the driver doesn't support
8331 * these flags (e.g. ethernet). log will be set to B_FALSE
8332 * if these error conditions are encountered.
8333 */
8334 if (mp1 != NULL) {
8335 if (ill->ill_ifname_pending_err != 0) {
8336 err = ill->ill_ifname_pending_err;
8337 ill->ill_ifname_pending_err = 0;
8338 } else {
8339 err = dlea->dl_unix_errno ?
8340 dlea->dl_unix_errno : ENXIO;
8341 }
8342 /*
8343 * If we're plumbing an interface and an error hasn't already
8344 * been saved, set ill_ifname_pending_err to the error passed
8345 * up. Ignore the error if log is B_FALSE (see comment above).
8346 */
8347 } else if (log && ill->ill_ifname_pending &&
8348 ill->ill_ifname_pending_err == 0) {
8349 ill->ill_ifname_pending_err = dlea->dl_unix_errno ?
8350 dlea->dl_unix_errno : ENXIO;
8351 }
8352
8353 if (log)
8354 ip_dlpi_error(ill, dlea->dl_error_primitive,
8355 dlea->dl_errno, dlea->dl_unix_errno);
8356 break;
8357 case DL_CAPABILITY_ACK:
8358 ill_capability_ack(ill, mp);
8359 /*
8360 * The message has been handed off to ill_capability_ack
8361 * and must not be freed below
8362 */
8363 mp = NULL;
8364 break;
8365
8366 case DL_INFO_ACK:
8367 /* Call a routine to handle this one. */
8368 ill_dlpi_done(ill, DL_INFO_REQ);
8369 ip_ll_subnet_defaults(ill, mp);
8370 ASSERT(!MUTEX_HELD(&ill->ill_phyint->phyint_ipsq->ipsq_lock));
8371 return;
8372 case DL_BIND_ACK:
8373 /*
8374 * We should have an IOCTL waiting on this unless
8375 * sent by ill_dl_phys, in which case just return
8376 */
8377 ill_dlpi_done(ill, DL_BIND_REQ);
8378
8379 if (ill->ill_ifname_pending) {
8380 DTRACE_PROBE2(ip__rput__dlpi__ifname__pending,
8381 ill_t *, ill, mblk_t *, mp);
8382 break;
8383 }
8384 mutex_enter(&ill->ill_lock);
8385 ill->ill_dl_up = 1;
8386 ill->ill_state_flags &= ~ILL_DOWN_IN_PROGRESS;
8387 mutex_exit(&ill->ill_lock);
8388
8389 if (!ioctl_aborted)
8390 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8391 if (mp1 == NULL) {
8392 DTRACE_PROBE1(ip__rput__dlpi__no__mblk, ill_t *, ill);
8393 break;
8394 }
8395 /*
8396 * mp1 was added by ill_dl_up(). if that is a result of
8397 * a DL_NOTE_REPLUMB notification, connp could be NULL.
8398 */
8399 if (connp != NULL)
8400 q = CONNP_TO_WQ(connp);
8401 /*
8402 * We are exclusive. So nothing can change even after
8403 * we get the pending mp.
8404 */
8405 ip1dbg(("ip_rput_dlpi: bind_ack %s\n", ill->ill_name));
8406 DTRACE_PROBE1(ip__rput__dlpi__bind__ack, ill_t *, ill);
8407 ill_nic_event_dispatch(ill, 0, NE_UP, NULL, 0);
8408
8409 /*
8410 * Now bring up the resolver; when that is complete, we'll
8411 * create IREs. Note that we intentionally mirror what
8412 * ipif_up() would have done, because we got here by way of
8413 * ill_dl_up(), which stopped ipif_up()'s processing.
8414 */
8415 if (ill->ill_isv6) {
8416 /*
8417 * v6 interfaces.
8418 * Unlike ARP which has to do another bind
8419 * and attach, once we get here we are
8420 * done with NDP
8421 */
8422 (void) ipif_resolver_up(ipif, Res_act_initial);
8423 if ((err = ipif_ndp_up(ipif, B_TRUE)) == 0)
8424 err = ipif_up_done_v6(ipif);
8425 } else if (ill->ill_net_type == IRE_IF_RESOLVER) {
8426 /*
8427 * ARP and other v4 external resolvers.
8428 * Leave the pending mblk intact so that
8429 * the ioctl completes in ip_rput().
8430 */
8431 if (connp != NULL)
8432 mutex_enter(&connp->conn_lock);
8433 mutex_enter(&ill->ill_lock);
8434 success = ipsq_pending_mp_add(connp, ipif, q, mp1, 0);
8435 mutex_exit(&ill->ill_lock);
8436 if (connp != NULL)
8437 mutex_exit(&connp->conn_lock);
8438 if (success) {
8439 err = ipif_resolver_up(ipif, Res_act_initial);
8440 if (err == EINPROGRESS) {
8441 freemsg(mp);
8442 return;
8443 }
8444 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8445 } else {
8446 /* The conn has started closing */
8447 err = EINTR;
8448 }
8449 } else {
8450 /*
8451 * This one is complete. Reply to pending ioctl.
8452 */
8453 (void) ipif_resolver_up(ipif, Res_act_initial);
8454 err = ipif_up_done(ipif);
8455 }
8456
8457 if ((err == 0) && (ill->ill_up_ipifs)) {
8458 err = ill_up_ipifs(ill, q, mp1);
8459 if (err == EINPROGRESS) {
8460 freemsg(mp);
8461 return;
8462 }
8463 }
8464
8465 /*
8466 * If we have a moved ipif to bring up, and everything has
8467 * succeeded to this point, bring it up on the IPMP ill.
8468 * Otherwise, leave it down -- the admin can try to bring it
8469 * up by hand if need be.
8470 */
8471 if (ill->ill_move_ipif != NULL) {
8472 if (err != 0) {
8473 ill->ill_move_ipif = NULL;
8474 } else {
8475 ipif = ill->ill_move_ipif;
8476 ill->ill_move_ipif = NULL;
8477 err = ipif_up(ipif, q, mp1);
8478 if (err == EINPROGRESS) {
8479 freemsg(mp);
8480 return;
8481 }
8482 }
8483 }
8484 break;
8485
8486 case DL_NOTIFY_IND: {
8487 dl_notify_ind_t *notify = (dl_notify_ind_t *)mp->b_rptr;
8488 uint_t orig_mtu, orig_mc_mtu;
8489
8490 switch (notify->dl_notification) {
8491 case DL_NOTE_PHYS_ADDR:
8492 err = ill_set_phys_addr(ill, mp);
8493 break;
8494
8495 case DL_NOTE_REPLUMB:
8496 /*
8497 * Directly return after calling ill_replumb().
8498 * Note that we should not free mp as it is reused
8499 * in the ill_replumb() function.
8500 */
8501 err = ill_replumb(ill, mp);
8502 return;
8503
8504 case DL_NOTE_FASTPATH_FLUSH:
8505 nce_flush(ill, B_FALSE);
8506 break;
8507
8508 case DL_NOTE_SDU_SIZE:
8509 case DL_NOTE_SDU_SIZE2:
8510 /*
8511 * The dce and fragmentation code can cope with
8512 * this changing while packets are being sent.
8513 * When packets are sent ip_output will discover
8514 * a change.
8515 *
8516 * Change the MTU size of the interface.
8517 */
8518 mutex_enter(&ill->ill_lock);
8519 orig_mtu = ill->ill_mtu;
8520 orig_mc_mtu = ill->ill_mc_mtu;
8521 switch (notify->dl_notification) {
8522 case DL_NOTE_SDU_SIZE:
8523 ill->ill_current_frag =
8524 (uint_t)notify->dl_data;
8525 ill->ill_mc_mtu = (uint_t)notify->dl_data;
8526 break;
8527 case DL_NOTE_SDU_SIZE2:
8528 ill->ill_current_frag =
8529 (uint_t)notify->dl_data1;
8530 ill->ill_mc_mtu = (uint_t)notify->dl_data2;
8531 break;
8532 }
8533 if (ill->ill_current_frag > ill->ill_max_frag)
8534 ill->ill_max_frag = ill->ill_current_frag;
8535
8536 if (!(ill->ill_flags & ILLF_FIXEDMTU)) {
8537 ill->ill_mtu = ill->ill_current_frag;
8538
8539 /*
8540 * If ill_user_mtu was set (via
8541 * SIOCSLIFLNKINFO), clamp ill_mtu at it.
8542 */
8543 if (ill->ill_user_mtu != 0 &&
8544 ill->ill_user_mtu < ill->ill_mtu)
8545 ill->ill_mtu = ill->ill_user_mtu;
8546
8547 if (ill->ill_user_mtu != 0 &&
8548 ill->ill_user_mtu < ill->ill_mc_mtu)
8549 ill->ill_mc_mtu = ill->ill_user_mtu;
8550
8551 if (ill->ill_isv6) {
8552 if (ill->ill_mtu < IPV6_MIN_MTU)
8553 ill->ill_mtu = IPV6_MIN_MTU;
8554 if (ill->ill_mc_mtu < IPV6_MIN_MTU)
8555 ill->ill_mc_mtu = IPV6_MIN_MTU;
8556 } else {
8557 if (ill->ill_mtu < IP_MIN_MTU)
8558 ill->ill_mtu = IP_MIN_MTU;
8559 if (ill->ill_mc_mtu < IP_MIN_MTU)
8560 ill->ill_mc_mtu = IP_MIN_MTU;
8561 }
8562 } else if (ill->ill_mc_mtu > ill->ill_mtu) {
8563 ill->ill_mc_mtu = ill->ill_mtu;
8564 }
8565
8566 mutex_exit(&ill->ill_lock);
8567 /*
8568 * Make sure all dce_generation checks find out
8569 * that ill_mtu/ill_mc_mtu has changed.
8570 */
8571 if (orig_mtu != ill->ill_mtu ||
8572 orig_mc_mtu != ill->ill_mc_mtu) {
8573 dce_increment_all_generations(ill->ill_isv6,
8574 ill->ill_ipst);
8575 }
8576
8577 /*
8578 * Refresh IPMP meta-interface MTU if necessary.
8579 */
8580 if (IS_UNDER_IPMP(ill))
8581 ipmp_illgrp_refresh_mtu(ill->ill_grp);
8582 break;
8583
8584 case DL_NOTE_LINK_UP:
8585 case DL_NOTE_LINK_DOWN: {
8586 /*
8587 * We are writer. ill / phyint / ipsq assocs stable.
8588 * The RUNNING flag reflects the state of the link.
8589 */
8590 phyint_t *phyint = ill->ill_phyint;
8591 uint64_t new_phyint_flags;
8592 boolean_t changed = B_FALSE;
8593 boolean_t went_up;
8594
8595 went_up = notify->dl_notification == DL_NOTE_LINK_UP;
8596 mutex_enter(&phyint->phyint_lock);
8597
8598 new_phyint_flags = went_up ?
8599 phyint->phyint_flags | PHYI_RUNNING :
8600 phyint->phyint_flags & ~PHYI_RUNNING;
8601
8602 if (IS_IPMP(ill)) {
8603 new_phyint_flags = went_up ?
8604 new_phyint_flags & ~PHYI_FAILED :
8605 new_phyint_flags | PHYI_FAILED;
8606 }
8607
8608 if (new_phyint_flags != phyint->phyint_flags) {
8609 phyint->phyint_flags = new_phyint_flags;
8610 changed = B_TRUE;
8611 }
8612 mutex_exit(&phyint->phyint_lock);
8613 /*
8614 * ill_restart_dad handles the DAD restart and routing
8615 * socket notification logic.
8616 */
8617 if (changed) {
8618 ill_restart_dad(phyint->phyint_illv4, went_up);
8619 ill_restart_dad(phyint->phyint_illv6, went_up);
8620 }
8621 break;
8622 }
8623 case DL_NOTE_PROMISC_ON_PHYS: {
8624 phyint_t *phyint = ill->ill_phyint;
8625
8626 mutex_enter(&phyint->phyint_lock);
8627 phyint->phyint_flags |= PHYI_PROMISC;
8628 mutex_exit(&phyint->phyint_lock);
8629 break;
8630 }
8631 case DL_NOTE_PROMISC_OFF_PHYS: {
8632 phyint_t *phyint = ill->ill_phyint;
8633
8634 mutex_enter(&phyint->phyint_lock);
8635 phyint->phyint_flags &= ~PHYI_PROMISC;
8636 mutex_exit(&phyint->phyint_lock);
8637 break;
8638 }
8639 case DL_NOTE_CAPAB_RENEG:
8640 /*
8641 * Something changed on the driver side.
8642 * It wants us to renegotiate the capabilities
8643 * on this ill. One possible cause is the aggregation
8644 * interface under us where a port got added or
8645 * went away.
8646 *
8647 * If the capability negotiation is already done
8648 * or is in progress, reset the capabilities and
8649 * mark the ill's ill_capab_reneg to be B_TRUE,
8650 * so that when the ack comes back, we can start
8651 * the renegotiation process.
8652 *
8653 * Note that if ill_capab_reneg is already B_TRUE
8654 * (ill_dlpi_capab_state is IDS_UNKNOWN in this case),
8655 * the capability resetting request has been sent
8656 * and the renegotiation has not been started yet;
8657 * nothing needs to be done in this case.
8658 */
8659 ipsq_current_start(ipsq, ill->ill_ipif, 0);
8660 ill_capability_reset(ill, B_TRUE);
8661 ipsq_current_finish(ipsq);
8662 break;
8663
8664 case DL_NOTE_ALLOWED_IPS:
8665 ill_set_allowed_ips(ill, mp);
8666 break;
8667 default:
8668 ip0dbg(("ip_rput_dlpi_writer: unknown notification "
8669 "type 0x%x for DL_NOTIFY_IND\n",
8670 notify->dl_notification));
8671 break;
8672 }
8673
8674 /*
8675 * As this is an asynchronous operation, we
8676 * should not call ill_dlpi_done
8677 */
8678 break;
8679 }
8680 case DL_NOTIFY_ACK: {
8681 dl_notify_ack_t *noteack = (dl_notify_ack_t *)mp->b_rptr;
8682
8683 if (noteack->dl_notifications & DL_NOTE_LINK_UP)
8684 ill->ill_note_link = 1;
8685 ill_dlpi_done(ill, DL_NOTIFY_REQ);
8686 break;
8687 }
8688 case DL_PHYS_ADDR_ACK: {
8689 /*
8690 * As part of plumbing the interface via SIOCSLIFNAME,
8691 * ill_dl_phys() will queue a series of DL_PHYS_ADDR_REQs,
8692 * whose answers we receive here. As each answer is received,
8693 * we call ill_dlpi_done() to dispatch the next request as
8694 * we're processing the current one. Once all answers have
8695 * been received, we use ipsq_pending_mp_get() to dequeue the
8696 * outstanding IOCTL and reply to it. (Because ill_dl_phys()
8697 * is invoked from an ill queue, conn_oper_pending_ill is not
8698 * available, but we know the ioctl is pending on ill_wq.)
8699 */
8700 uint_t paddrlen, paddroff;
8701 uint8_t *addr;
8702
8703 paddrreq = ill->ill_phys_addr_pend;
8704 paddrlen = ((dl_phys_addr_ack_t *)mp->b_rptr)->dl_addr_length;
8705 paddroff = ((dl_phys_addr_ack_t *)mp->b_rptr)->dl_addr_offset;
8706 addr = mp->b_rptr + paddroff;
8707
8708 ill_dlpi_done(ill, DL_PHYS_ADDR_REQ);
8709 if (paddrreq == DL_IPV6_TOKEN) {
8710 /*
8711 * bcopy to low-order bits of ill_token
8712 *
8713 * XXX Temporary hack - currently, all known tokens
8714 * are 64 bits, so I'll cheat for the moment.
8715 */
8716 bcopy(addr, &ill->ill_token.s6_addr32[2], paddrlen);
8717 ill->ill_token_length = paddrlen;
8718 break;
8719 } else if (paddrreq == DL_IPV6_LINK_LAYER_ADDR) {
8720 ASSERT(ill->ill_nd_lla_mp == NULL);
8721 ill_set_ndmp(ill, mp, paddroff, paddrlen);
8722 mp = NULL;
8723 break;
8724 } else if (paddrreq == DL_CURR_DEST_ADDR) {
8725 ASSERT(ill->ill_dest_addr_mp == NULL);
8726 ill->ill_dest_addr_mp = mp;
8727 ill->ill_dest_addr = addr;
8728 mp = NULL;
8729 if (ill->ill_isv6) {
8730 ill_setdesttoken(ill);
8731 ipif_setdestlinklocal(ill->ill_ipif);
8732 }
8733 break;
8734 }
8735
8736 ASSERT(paddrreq == DL_CURR_PHYS_ADDR);
8737 ASSERT(ill->ill_phys_addr_mp == NULL);
8738 if (!ill->ill_ifname_pending)
8739 break;
8740 ill->ill_ifname_pending = 0;
8741 if (!ioctl_aborted)
8742 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8743 if (mp1 != NULL) {
8744 ASSERT(connp == NULL);
8745 q = ill->ill_wq;
8746 }
8747 /*
8748 * If any error acks received during the plumbing sequence,
8749 * ill_ifname_pending_err will be set. Break out and send up
8750 * the error to the pending ioctl.
8751 */
8752 if (ill->ill_ifname_pending_err != 0) {
8753 err = ill->ill_ifname_pending_err;
8754 ill->ill_ifname_pending_err = 0;
8755 break;
8756 }
8757
8758 ill->ill_phys_addr_mp = mp;
8759 ill->ill_phys_addr = (paddrlen == 0 ? NULL : addr);
8760 mp = NULL;
8761
8762 /*
8763 * If paddrlen or ill_phys_addr_length is zero, the DLPI
8764 * provider doesn't support physical addresses. We check both
8765 * paddrlen and ill_phys_addr_length because sppp (PPP) does
8766 * not have physical addresses, but historically adversises a
8767 * physical address length of 0 in its DL_INFO_ACK, but 6 in
8768 * its DL_PHYS_ADDR_ACK.
8769 */
8770 if (paddrlen == 0 || ill->ill_phys_addr_length == 0) {
8771 ill->ill_phys_addr = NULL;
8772 } else if (paddrlen != ill->ill_phys_addr_length) {
8773 ip0dbg(("DL_PHYS_ADDR_ACK: got addrlen %d, expected %d",
8774 paddrlen, ill->ill_phys_addr_length));
8775 err = EINVAL;
8776 break;
8777 }
8778
8779 if (ill->ill_nd_lla_mp == NULL) {
8780 if ((mp_hw = copyb(ill->ill_phys_addr_mp)) == NULL) {
8781 err = ENOMEM;
8782 break;
8783 }
8784 ill_set_ndmp(ill, mp_hw, paddroff, paddrlen);
8785 }
8786
8787 if (ill->ill_isv6) {
8788 ill_setdefaulttoken(ill);
8789 ipif_setlinklocal(ill->ill_ipif);
8790 }
8791 break;
8792 }
8793 case DL_OK_ACK:
8794 ip2dbg(("DL_OK_ACK %s (0x%x)\n",
8795 dl_primstr((int)dloa->dl_correct_primitive),
8796 dloa->dl_correct_primitive));
8797 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer ok",
8798 char *, dl_primstr(dloa->dl_correct_primitive),
8799 ill_t *, ill);
8800
8801 switch (dloa->dl_correct_primitive) {
8802 case DL_ENABMULTI_REQ:
8803 case DL_DISABMULTI_REQ:
8804 ill_dlpi_done(ill, dloa->dl_correct_primitive);
8805 break;
8806 case DL_PROMISCON_REQ:
8807 case DL_PROMISCOFF_REQ:
8808 case DL_UNBIND_REQ:
8809 case DL_ATTACH_REQ:
8810 ill_dlpi_done(ill, dloa->dl_correct_primitive);
8811 break;
8812 }
8813 break;
8814 default:
8815 break;
8816 }
8817
8818 freemsg(mp);
8819 if (mp1 == NULL)
8820 return;
8821
8822 /*
8823 * The operation must complete without EINPROGRESS since
8824 * ipsq_pending_mp_get() has removed the mblk (mp1). Otherwise,
8825 * the operation will be stuck forever inside the IPSQ.
8826 */
8827 ASSERT(err != EINPROGRESS);
8828
8829 DTRACE_PROBE4(ipif__ioctl, char *, "ip_rput_dlpi_writer finish",
8830 int, ipsq->ipsq_xop->ipx_current_ioctl, ill_t *, ill,
8831 ipif_t *, NULL);
8832
8833 switch (ipsq->ipsq_xop->ipx_current_ioctl) {
8834 case 0:
8835 ipsq_current_finish(ipsq);
8836 break;
8837
8838 case SIOCSLIFNAME:
8839 case IF_UNITSEL: {
8840 ill_t *ill_other = ILL_OTHER(ill);
8841
8842 /*
8843 * If SIOCSLIFNAME or IF_UNITSEL is about to succeed, and the
8844 * ill has a peer which is in an IPMP group, then place ill
8845 * into the same group. One catch: although ifconfig plumbs
8846 * the appropriate IPMP meta-interface prior to plumbing this
8847 * ill, it is possible for multiple ifconfig applications to
8848 * race (or for another application to adjust plumbing), in
8849 * which case the IPMP meta-interface we need will be missing.
8850 * If so, kick the phyint out of the group.
8851 */
8852 if (err == 0 && ill_other != NULL && IS_UNDER_IPMP(ill_other)) {
8853 ipmp_grp_t *grp = ill->ill_phyint->phyint_grp;
8854 ipmp_illgrp_t *illg;
8855
8856 illg = ill->ill_isv6 ? grp->gr_v6 : grp->gr_v4;
8857 if (illg == NULL)
8858 ipmp_phyint_leave_grp(ill->ill_phyint);
8859 else
8860 ipmp_ill_join_illgrp(ill, illg);
8861 }
8862
8863 if (ipsq->ipsq_xop->ipx_current_ioctl == IF_UNITSEL)
8864 ip_ioctl_finish(q, mp1, err, NO_COPYOUT, ipsq);
8865 else
8866 ip_ioctl_finish(q, mp1, err, COPYOUT, ipsq);
8867 break;
8868 }
8869 case SIOCLIFADDIF:
8870 ip_ioctl_finish(q, mp1, err, COPYOUT, ipsq);
8871 break;
8872
8873 default:
8874 ip_ioctl_finish(q, mp1, err, NO_COPYOUT, ipsq);
8875 break;
8876 }
8877 }
8878
8879 /*
8880 * ip_rput_other is called by ip_rput to handle messages modifying the global
8881 * state in IP. If 'ipsq' is non-NULL, caller is writer on it.
8882 */
8883 /* ARGSUSED */
8884 void
8885 ip_rput_other(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
8886 {
8887 ill_t *ill = q->q_ptr;
8888 struct iocblk *iocp;
8889
8890 ip1dbg(("ip_rput_other "));
8891 if (ipsq != NULL) {
8892 ASSERT(IAM_WRITER_IPSQ(ipsq));
8893 ASSERT(ipsq->ipsq_xop ==
8894 ill->ill_phyint->phyint_ipsq->ipsq_xop);
8895 }
8896
8897 switch (mp->b_datap->db_type) {
8898 case M_ERROR:
8899 case M_HANGUP:
8900 /*
8901 * The device has a problem. We force the ILL down. It can
8902 * be brought up again manually using SIOCSIFFLAGS (via
8903 * ifconfig or equivalent).
8904 */
8905 ASSERT(ipsq != NULL);
8906 if (mp->b_rptr < mp->b_wptr)
8907 ill->ill_error = (int)(*mp->b_rptr & 0xFF);
8908 if (ill->ill_error == 0)
8909 ill->ill_error = ENXIO;
8910 if (!ill_down_start(q, mp))
8911 return;
8912 ipif_all_down_tail(ipsq, q, mp, NULL);
8913 break;
8914 case M_IOCNAK: {
8915 iocp = (struct iocblk *)mp->b_rptr;
8916
8917 ASSERT(iocp->ioc_cmd == DL_IOC_HDR_INFO);
8918 /*
8919 * If this was the first attempt, turn off the fastpath
8920 * probing.
8921 */
8922 mutex_enter(&ill->ill_lock);
8923 if (ill->ill_dlpi_fastpath_state == IDS_INPROGRESS) {
8924 ill->ill_dlpi_fastpath_state = IDS_FAILED;
8925 mutex_exit(&ill->ill_lock);
8926 /*
8927 * don't flush the nce_t entries: we use them
8928 * as an index to the ncec itself.
8929 */
8930 ip1dbg(("ip_rput: DLPI fastpath off on interface %s\n",
8931 ill->ill_name));
8932 } else {
8933 mutex_exit(&ill->ill_lock);
8934 }
8935 freemsg(mp);
8936 break;
8937 }
8938 default:
8939 ASSERT(0);
8940 break;
8941 }
8942 }
8943
8944 /*
8945 * Update any source route, record route or timestamp options
8946 * When it fails it has consumed the message and BUMPed the MIB.
8947 */
8948 boolean_t
8949 ip_forward_options(mblk_t *mp, ipha_t *ipha, ill_t *dst_ill,
8950 ip_recv_attr_t *ira)
8951 {
8952 ipoptp_t opts;
8953 uchar_t *opt;
8954 uint8_t optval;
8955 uint8_t optlen;
8956 ipaddr_t dst;
8957 ipaddr_t ifaddr;
8958 uint32_t ts;
8959 timestruc_t now;
8960 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
8961
8962 ip2dbg(("ip_forward_options\n"));
8963 dst = ipha->ipha_dst;
8964 for (optval = ipoptp_first(&opts, ipha);
8965 optval != IPOPT_EOL;
8966 optval = ipoptp_next(&opts)) {
8967 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
8968 opt = opts.ipoptp_cur;
8969 optlen = opts.ipoptp_len;
8970 ip2dbg(("ip_forward_options: opt %d, len %d\n",
8971 optval, opts.ipoptp_len));
8972 switch (optval) {
8973 uint32_t off;
8974 case IPOPT_SSRR:
8975 case IPOPT_LSRR:
8976 /* Check if adminstratively disabled */
8977 if (!ipst->ips_ip_forward_src_routed) {
8978 BUMP_MIB(dst_ill->ill_ip_mib,
8979 ipIfStatsForwProhibits);
8980 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED",
8981 mp, dst_ill);
8982 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED,
8983 ira);
8984 return (B_FALSE);
8985 }
8986 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
8987 /*
8988 * Must be partial since ip_input_options
8989 * checked for strict.
8990 */
8991 break;
8992 }
8993 off = opt[IPOPT_OFFSET];
8994 off--;
8995 redo_srr:
8996 if (optlen < IP_ADDR_LEN ||
8997 off > optlen - IP_ADDR_LEN) {
8998 /* End of source route */
8999 ip1dbg((
9000 "ip_forward_options: end of SR\n"));
9001 break;
9002 }
9003 /* Pick a reasonable address on the outbound if */
9004 ASSERT(dst_ill != NULL);
9005 if (ip_select_source_v4(dst_ill, INADDR_ANY, dst,
9006 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9007 NULL) != 0) {
9008 /* No source! Shouldn't happen */
9009 ifaddr = INADDR_ANY;
9010 }
9011 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9012 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9013 ip1dbg(("ip_forward_options: next hop 0x%x\n",
9014 ntohl(dst)));
9015
9016 /*
9017 * Check if our address is present more than
9018 * once as consecutive hops in source route.
9019 */
9020 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
9021 off += IP_ADDR_LEN;
9022 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9023 goto redo_srr;
9024 }
9025 ipha->ipha_dst = dst;
9026 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9027 break;
9028 case IPOPT_RR:
9029 off = opt[IPOPT_OFFSET];
9030 off--;
9031 if (optlen < IP_ADDR_LEN ||
9032 off > optlen - IP_ADDR_LEN) {
9033 /* No more room - ignore */
9034 ip1dbg((
9035 "ip_forward_options: end of RR\n"));
9036 break;
9037 }
9038 /* Pick a reasonable address on the outbound if */
9039 ASSERT(dst_ill != NULL);
9040 if (ip_select_source_v4(dst_ill, INADDR_ANY, dst,
9041 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9042 NULL) != 0) {
9043 /* No source! Shouldn't happen */
9044 ifaddr = INADDR_ANY;
9045 }
9046 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9047 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9048 break;
9049 case IPOPT_TS:
9050 /* Insert timestamp if there is room */
9051 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9052 case IPOPT_TS_TSONLY:
9053 off = IPOPT_TS_TIMELEN;
9054 break;
9055 case IPOPT_TS_PRESPEC:
9056 case IPOPT_TS_PRESPEC_RFC791:
9057 /* Verify that the address matched */
9058 off = opt[IPOPT_OFFSET] - 1;
9059 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9060 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9061 /* Not for us */
9062 break;
9063 }
9064 /* FALLTHRU */
9065 case IPOPT_TS_TSANDADDR:
9066 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9067 break;
9068 default:
9069 /*
9070 * ip_*put_options should have already
9071 * dropped this packet.
9072 */
9073 cmn_err(CE_PANIC, "ip_forward_options: "
9074 "unknown IT - bug in ip_input_options?\n");
9075 return (B_TRUE); /* Keep "lint" happy */
9076 }
9077 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
9078 /* Increase overflow counter */
9079 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
9080 opt[IPOPT_POS_OV_FLG] =
9081 (uint8_t)((opt[IPOPT_POS_OV_FLG] & 0x0F) |
9082 (off << 4));
9083 break;
9084 }
9085 off = opt[IPOPT_OFFSET] - 1;
9086 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9087 case IPOPT_TS_PRESPEC:
9088 case IPOPT_TS_PRESPEC_RFC791:
9089 case IPOPT_TS_TSANDADDR:
9090 /* Pick a reasonable addr on the outbound if */
9091 ASSERT(dst_ill != NULL);
9092 if (ip_select_source_v4(dst_ill, INADDR_ANY,
9093 dst, INADDR_ANY, ALL_ZONES, ipst, &ifaddr,
9094 NULL, NULL) != 0) {
9095 /* No source! Shouldn't happen */
9096 ifaddr = INADDR_ANY;
9097 }
9098 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9099 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9100 /* FALLTHRU */
9101 case IPOPT_TS_TSONLY:
9102 off = opt[IPOPT_OFFSET] - 1;
9103 /* Compute # of milliseconds since midnight */
9104 gethrestime(&now);
9105 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
9106 now.tv_nsec / (NANOSEC / MILLISEC);
9107 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
9108 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
9109 break;
9110 }
9111 break;
9112 }
9113 }
9114 return (B_TRUE);
9115 }
9116
9117 /*
9118 * Call ill_frag_timeout to do garbage collection. ill_frag_timeout
9119 * returns 'true' if there are still fragments left on the queue, in
9120 * which case we restart the timer.
9121 */
9122 void
9123 ill_frag_timer(void *arg)
9124 {
9125 ill_t *ill = (ill_t *)arg;
9126 boolean_t frag_pending;
9127 ip_stack_t *ipst = ill->ill_ipst;
9128 time_t timeout;
9129
9130 mutex_enter(&ill->ill_lock);
9131 ASSERT(!ill->ill_fragtimer_executing);
9132 if (ill->ill_state_flags & ILL_CONDEMNED) {
9133 ill->ill_frag_timer_id = 0;
9134 mutex_exit(&ill->ill_lock);
9135 return;
9136 }
9137 ill->ill_fragtimer_executing = 1;
9138 mutex_exit(&ill->ill_lock);
9139
9140 timeout = (ill->ill_isv6 ? ipst->ips_ipv6_reassembly_timeout :
9141 ipst->ips_ip_reassembly_timeout);
9142
9143 frag_pending = ill_frag_timeout(ill, timeout);
9144
9145 /*
9146 * Restart the timer, if we have fragments pending or if someone
9147 * wanted us to be scheduled again.
9148 */
9149 mutex_enter(&ill->ill_lock);
9150 ill->ill_fragtimer_executing = 0;
9151 ill->ill_frag_timer_id = 0;
9152 if (frag_pending || ill->ill_fragtimer_needrestart)
9153 ill_frag_timer_start(ill);
9154 mutex_exit(&ill->ill_lock);
9155 }
9156
9157 void
9158 ill_frag_timer_start(ill_t *ill)
9159 {
9160 ip_stack_t *ipst = ill->ill_ipst;
9161 clock_t timeo_ms;
9162
9163 ASSERT(MUTEX_HELD(&ill->ill_lock));
9164
9165 /* If the ill is closing or opening don't proceed */
9166 if (ill->ill_state_flags & ILL_CONDEMNED)
9167 return;
9168
9169 if (ill->ill_fragtimer_executing) {
9170 /*
9171 * ill_frag_timer is currently executing. Just record the
9172 * the fact that we want the timer to be restarted.
9173 * ill_frag_timer will post a timeout before it returns,
9174 * ensuring it will be called again.
9175 */
9176 ill->ill_fragtimer_needrestart = 1;
9177 return;
9178 }
9179
9180 if (ill->ill_frag_timer_id == 0) {
9181 timeo_ms = (ill->ill_isv6 ? ipst->ips_ipv6_reassembly_timeout :
9182 ipst->ips_ip_reassembly_timeout) * SECONDS;
9183
9184 /*
9185 * The timer is neither running nor is the timeout handler
9186 * executing. Post a timeout so that ill_frag_timer will be
9187 * called
9188 */
9189 ill->ill_frag_timer_id = timeout(ill_frag_timer, ill,
9190 MSEC_TO_TICK(timeo_ms >> 1));
9191 ill->ill_fragtimer_needrestart = 0;
9192 }
9193 }
9194
9195 /*
9196 * Update any source route, record route or timestamp options.
9197 * Check that we are at end of strict source route.
9198 * The options have already been checked for sanity in ip_input_options().
9199 */
9200 boolean_t
9201 ip_input_local_options(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
9202 {
9203 ipoptp_t opts;
9204 uchar_t *opt;
9205 uint8_t optval;
9206 uint8_t optlen;
9207 ipaddr_t dst;
9208 ipaddr_t ifaddr;
9209 uint32_t ts;
9210 timestruc_t now;
9211 ill_t *ill = ira->ira_ill;
9212 ip_stack_t *ipst = ill->ill_ipst;
9213
9214 ip2dbg(("ip_input_local_options\n"));
9215
9216 for (optval = ipoptp_first(&opts, ipha);
9217 optval != IPOPT_EOL;
9218 optval = ipoptp_next(&opts)) {
9219 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
9220 opt = opts.ipoptp_cur;
9221 optlen = opts.ipoptp_len;
9222 ip2dbg(("ip_input_local_options: opt %d, len %d\n",
9223 optval, optlen));
9224 switch (optval) {
9225 uint32_t off;
9226 case IPOPT_SSRR:
9227 case IPOPT_LSRR:
9228 off = opt[IPOPT_OFFSET];
9229 off--;
9230 if (optlen < IP_ADDR_LEN ||
9231 off > optlen - IP_ADDR_LEN) {
9232 /* End of source route */
9233 ip1dbg(("ip_input_local_options: end of SR\n"));
9234 break;
9235 }
9236 /*
9237 * This will only happen if two consecutive entries
9238 * in the source route contains our address or if
9239 * it is a packet with a loose source route which
9240 * reaches us before consuming the whole source route
9241 */
9242 ip1dbg(("ip_input_local_options: not end of SR\n"));
9243 if (optval == IPOPT_SSRR) {
9244 goto bad_src_route;
9245 }
9246 /*
9247 * Hack: instead of dropping the packet truncate the
9248 * source route to what has been used by filling the
9249 * rest with IPOPT_NOP.
9250 */
9251 opt[IPOPT_OLEN] = (uint8_t)off;
9252 while (off < optlen) {
9253 opt[off++] = IPOPT_NOP;
9254 }
9255 break;
9256 case IPOPT_RR:
9257 off = opt[IPOPT_OFFSET];
9258 off--;
9259 if (optlen < IP_ADDR_LEN ||
9260 off > optlen - IP_ADDR_LEN) {
9261 /* No more room - ignore */
9262 ip1dbg((
9263 "ip_input_local_options: end of RR\n"));
9264 break;
9265 }
9266 /* Pick a reasonable address on the outbound if */
9267 if (ip_select_source_v4(ill, INADDR_ANY, ipha->ipha_dst,
9268 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9269 NULL) != 0) {
9270 /* No source! Shouldn't happen */
9271 ifaddr = INADDR_ANY;
9272 }
9273 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9274 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9275 break;
9276 case IPOPT_TS:
9277 /* Insert timestamp if there is romm */
9278 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9279 case IPOPT_TS_TSONLY:
9280 off = IPOPT_TS_TIMELEN;
9281 break;
9282 case IPOPT_TS_PRESPEC:
9283 case IPOPT_TS_PRESPEC_RFC791:
9284 /* Verify that the address matched */
9285 off = opt[IPOPT_OFFSET] - 1;
9286 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9287 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9288 /* Not for us */
9289 break;
9290 }
9291 /* FALLTHRU */
9292 case IPOPT_TS_TSANDADDR:
9293 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9294 break;
9295 default:
9296 /*
9297 * ip_*put_options should have already
9298 * dropped this packet.
9299 */
9300 cmn_err(CE_PANIC, "ip_input_local_options: "
9301 "unknown IT - bug in ip_input_options?\n");
9302 return (B_TRUE); /* Keep "lint" happy */
9303 }
9304 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
9305 /* Increase overflow counter */
9306 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
9307 opt[IPOPT_POS_OV_FLG] =
9308 (uint8_t)((opt[IPOPT_POS_OV_FLG] & 0x0F) |
9309 (off << 4));
9310 break;
9311 }
9312 off = opt[IPOPT_OFFSET] - 1;
9313 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9314 case IPOPT_TS_PRESPEC:
9315 case IPOPT_TS_PRESPEC_RFC791:
9316 case IPOPT_TS_TSANDADDR:
9317 /* Pick a reasonable addr on the outbound if */
9318 if (ip_select_source_v4(ill, INADDR_ANY,
9319 ipha->ipha_dst, INADDR_ANY, ALL_ZONES, ipst,
9320 &ifaddr, NULL, NULL) != 0) {
9321 /* No source! Shouldn't happen */
9322 ifaddr = INADDR_ANY;
9323 }
9324 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9325 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9326 /* FALLTHRU */
9327 case IPOPT_TS_TSONLY:
9328 off = opt[IPOPT_OFFSET] - 1;
9329 /* Compute # of milliseconds since midnight */
9330 gethrestime(&now);
9331 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
9332 now.tv_nsec / (NANOSEC / MILLISEC);
9333 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
9334 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
9335 break;
9336 }
9337 break;
9338 }
9339 }
9340 return (B_TRUE);
9341
9342 bad_src_route:
9343 /* make sure we clear any indication of a hardware checksum */
9344 DB_CKSUMFLAGS(mp) = 0;
9345 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ill);
9346 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
9347 return (B_FALSE);
9348
9349 }
9350
9351 /*
9352 * Process IP options in an inbound packet. Always returns the nexthop.
9353 * Normally this is the passed in nexthop, but if there is an option
9354 * that effects the nexthop (such as a source route) that will be returned.
9355 * Sets *errorp if there is an error, in which case an ICMP error has been sent
9356 * and mp freed.
9357 */
9358 ipaddr_t
9359 ip_input_options(ipha_t *ipha, ipaddr_t dst, mblk_t *mp,
9360 ip_recv_attr_t *ira, int *errorp)
9361 {
9362 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
9363 ipoptp_t opts;
9364 uchar_t *opt;
9365 uint8_t optval;
9366 uint8_t optlen;
9367 intptr_t code = 0;
9368 ire_t *ire;
9369
9370 ip2dbg(("ip_input_options\n"));
9371 *errorp = 0;
9372 for (optval = ipoptp_first(&opts, ipha);
9373 optval != IPOPT_EOL;
9374 optval = ipoptp_next(&opts)) {
9375 opt = opts.ipoptp_cur;
9376 optlen = opts.ipoptp_len;
9377 ip2dbg(("ip_input_options: opt %d, len %d\n",
9378 optval, optlen));
9379 /*
9380 * Note: we need to verify the checksum before we
9381 * modify anything thus this routine only extracts the next
9382 * hop dst from any source route.
9383 */
9384 switch (optval) {
9385 uint32_t off;
9386 case IPOPT_SSRR:
9387 case IPOPT_LSRR:
9388 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9389 if (optval == IPOPT_SSRR) {
9390 ip1dbg(("ip_input_options: not next"
9391 " strict source route 0x%x\n",
9392 ntohl(dst)));
9393 code = (char *)&ipha->ipha_dst -
9394 (char *)ipha;
9395 goto param_prob; /* RouterReq's */
9396 }
9397 ip2dbg(("ip_input_options: "
9398 "not next source route 0x%x\n",
9399 ntohl(dst)));
9400 break;
9401 }
9402
9403 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9404 ip1dbg((
9405 "ip_input_options: bad option offset\n"));
9406 code = (char *)&opt[IPOPT_OLEN] -
9407 (char *)ipha;
9408 goto param_prob;
9409 }
9410 off = opt[IPOPT_OFFSET];
9411 off--;
9412 redo_srr:
9413 if (optlen < IP_ADDR_LEN ||
9414 off > optlen - IP_ADDR_LEN) {
9415 /* End of source route */
9416 ip1dbg(("ip_input_options: end of SR\n"));
9417 break;
9418 }
9419 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9420 ip1dbg(("ip_input_options: next hop 0x%x\n",
9421 ntohl(dst)));
9422
9423 /*
9424 * Check if our address is present more than
9425 * once as consecutive hops in source route.
9426 * XXX verify per-interface ip_forwarding
9427 * for source route?
9428 */
9429 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
9430 off += IP_ADDR_LEN;
9431 goto redo_srr;
9432 }
9433
9434 if (dst == htonl(INADDR_LOOPBACK)) {
9435 ip1dbg(("ip_input_options: loopback addr in "
9436 "source route!\n"));
9437 goto bad_src_route;
9438 }
9439 /*
9440 * For strict: verify that dst is directly
9441 * reachable.
9442 */
9443 if (optval == IPOPT_SSRR) {
9444 ire = ire_ftable_lookup_v4(dst, 0, 0,
9445 IRE_INTERFACE, NULL, ALL_ZONES,
9446 ira->ira_tsl,
9447 MATCH_IRE_TYPE | MATCH_IRE_SECATTR, 0, ipst,
9448 NULL);
9449 if (ire == NULL) {
9450 ip1dbg(("ip_input_options: SSRR not "
9451 "directly reachable: 0x%x\n",
9452 ntohl(dst)));
9453 goto bad_src_route;
9454 }
9455 ire_refrele(ire);
9456 }
9457 /*
9458 * Defer update of the offset and the record route
9459 * until the packet is forwarded.
9460 */
9461 break;
9462 case IPOPT_RR:
9463 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9464 ip1dbg((
9465 "ip_input_options: bad option offset\n"));
9466 code = (char *)&opt[IPOPT_OLEN] -
9467 (char *)ipha;
9468 goto param_prob;
9469 }
9470 break;
9471 case IPOPT_TS:
9472 /*
9473 * Verify that length >= 5 and that there is either
9474 * room for another timestamp or that the overflow
9475 * counter is not maxed out.
9476 */
9477 code = (char *)&opt[IPOPT_OLEN] - (char *)ipha;
9478 if (optlen < IPOPT_MINLEN_IT) {
9479 goto param_prob;
9480 }
9481 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9482 ip1dbg((
9483 "ip_input_options: bad option offset\n"));
9484 code = (char *)&opt[IPOPT_OFFSET] -
9485 (char *)ipha;
9486 goto param_prob;
9487 }
9488 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9489 case IPOPT_TS_TSONLY:
9490 off = IPOPT_TS_TIMELEN;
9491 break;
9492 case IPOPT_TS_TSANDADDR:
9493 case IPOPT_TS_PRESPEC:
9494 case IPOPT_TS_PRESPEC_RFC791:
9495 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9496 break;
9497 default:
9498 code = (char *)&opt[IPOPT_POS_OV_FLG] -
9499 (char *)ipha;
9500 goto param_prob;
9501 }
9502 if (opt[IPOPT_OFFSET] - 1 + off > optlen &&
9503 (opt[IPOPT_POS_OV_FLG] & 0xF0) == 0xF0) {
9504 /*
9505 * No room and the overflow counter is 15
9506 * already.
9507 */
9508 goto param_prob;
9509 }
9510 break;
9511 }
9512 }
9513
9514 if ((opts.ipoptp_flags & IPOPTP_ERROR) == 0) {
9515 return (dst);
9516 }
9517
9518 ip1dbg(("ip_input_options: error processing IP options."));
9519 code = (char *)&opt[IPOPT_OFFSET] - (char *)ipha;
9520
9521 param_prob:
9522 /* make sure we clear any indication of a hardware checksum */
9523 DB_CKSUMFLAGS(mp) = 0;
9524 ip_drop_input("ICMP_PARAM_PROBLEM", mp, ira->ira_ill);
9525 icmp_param_problem(mp, (uint8_t)code, ira);
9526 *errorp = -1;
9527 return (dst);
9528
9529 bad_src_route:
9530 /* make sure we clear any indication of a hardware checksum */
9531 DB_CKSUMFLAGS(mp) = 0;
9532 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ira->ira_ill);
9533 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
9534 *errorp = -1;
9535 return (dst);
9536 }
9537
9538 /*
9539 * IP & ICMP info in >=14 msg's ...
9540 * - ip fixed part (mib2_ip_t)
9541 * - icmp fixed part (mib2_icmp_t)
9542 * - ipAddrEntryTable (ip 20) all IPv4 ipifs
9543 * - ipRouteEntryTable (ip 21) all IPv4 IREs
9544 * - ipNetToMediaEntryTable (ip 22) all IPv4 Neighbor Cache entries
9545 * - ipRouteAttributeTable (ip 102) labeled routes
9546 * - ip multicast membership (ip_member_t)
9547 * - ip multicast source filtering (ip_grpsrc_t)
9548 * - igmp fixed part (struct igmpstat)
9549 * - multicast routing stats (struct mrtstat)
9550 * - multicast routing vifs (array of struct vifctl)
9551 * - multicast routing routes (array of struct mfcctl)
9552 * - ip6 fixed part (mib2_ipv6IfStatsEntry_t)
9553 * One per ill plus one generic
9554 * - icmp6 fixed part (mib2_ipv6IfIcmpEntry_t)
9555 * One per ill plus one generic
9556 * - ipv6RouteEntry all IPv6 IREs
9557 * - ipv6RouteAttributeTable (ip6 102) labeled routes
9558 * - ipv6NetToMediaEntry all IPv6 Neighbor Cache entries
9559 * - ipv6AddrEntry all IPv6 ipifs
9560 * - ipv6 multicast membership (ipv6_member_t)
9561 * - ipv6 multicast source filtering (ipv6_grpsrc_t)
9562 *
9563 * NOTE: original mpctl is copied for msg's 2..N, since its ctl part is
9564 * already filled in by the caller.
9565 * If legacy_req is true then MIB structures needs to be truncated to their
9566 * legacy sizes before being returned.
9567 * Return value of 0 indicates that no messages were sent and caller
9568 * should free mpctl.
9569 */
9570 int
9571 ip_snmp_get(queue_t *q, mblk_t *mpctl, int level, boolean_t legacy_req)
9572 {
9573 ip_stack_t *ipst;
9574 sctp_stack_t *sctps;
9575
9576 if (q->q_next != NULL) {
9577 ipst = ILLQ_TO_IPST(q);
9578 } else {
9579 ipst = CONNQ_TO_IPST(q);
9580 }
9581 ASSERT(ipst != NULL);
9582 sctps = ipst->ips_netstack->netstack_sctp;
9583
9584 if (mpctl == NULL || mpctl->b_cont == NULL) {
9585 return (0);
9586 }
9587
9588 /*
9589 * For the purposes of the (broken) packet shell use
9590 * of the level we make sure MIB2_TCP/MIB2_UDP can be used
9591 * to make TCP and UDP appear first in the list of mib items.
9592 * TBD: We could expand this and use it in netstat so that
9593 * the kernel doesn't have to produce large tables (connections,
9594 * routes, etc) when netstat only wants the statistics or a particular
9595 * table.
9596 */
9597 if (!(level == MIB2_TCP || level == MIB2_UDP)) {
9598 if ((mpctl = icmp_snmp_get(q, mpctl)) == NULL) {
9599 return (1);
9600 }
9601 }
9602
9603 if (level != MIB2_TCP) {
9604 if ((mpctl = udp_snmp_get(q, mpctl, legacy_req)) == NULL) {
9605 return (1);
9606 }
9607 }
9608
9609 if (level != MIB2_UDP) {
9610 if ((mpctl = tcp_snmp_get(q, mpctl, legacy_req)) == NULL) {
9611 return (1);
9612 }
9613 }
9614
9615 if ((mpctl = ip_snmp_get_mib2_ip_traffic_stats(q, mpctl,
9616 ipst, legacy_req)) == NULL) {
9617 return (1);
9618 }
9619
9620 if ((mpctl = ip_snmp_get_mib2_ip6(q, mpctl, ipst,
9621 legacy_req)) == NULL) {
9622 return (1);
9623 }
9624
9625 if ((mpctl = ip_snmp_get_mib2_icmp(q, mpctl, ipst)) == NULL) {
9626 return (1);
9627 }
9628
9629 if ((mpctl = ip_snmp_get_mib2_icmp6(q, mpctl, ipst)) == NULL) {
9630 return (1);
9631 }
9632
9633 if ((mpctl = ip_snmp_get_mib2_igmp(q, mpctl, ipst)) == NULL) {
9634 return (1);
9635 }
9636
9637 if ((mpctl = ip_snmp_get_mib2_multi(q, mpctl, ipst)) == NULL) {
9638 return (1);
9639 }
9640
9641 if ((mpctl = ip_snmp_get_mib2_ip_addr(q, mpctl, ipst,
9642 legacy_req)) == NULL) {
9643 return (1);
9644 }
9645
9646 if ((mpctl = ip_snmp_get_mib2_ip6_addr(q, mpctl, ipst,
9647 legacy_req)) == NULL) {
9648 return (1);
9649 }
9650
9651 if ((mpctl = ip_snmp_get_mib2_ip_group_mem(q, mpctl, ipst)) == NULL) {
9652 return (1);
9653 }
9654
9655 if ((mpctl = ip_snmp_get_mib2_ip6_group_mem(q, mpctl, ipst)) == NULL) {
9656 return (1);
9657 }
9658
9659 if ((mpctl = ip_snmp_get_mib2_ip_group_src(q, mpctl, ipst)) == NULL) {
9660 return (1);
9661 }
9662
9663 if ((mpctl = ip_snmp_get_mib2_ip6_group_src(q, mpctl, ipst)) == NULL) {
9664 return (1);
9665 }
9666
9667 if ((mpctl = ip_snmp_get_mib2_virt_multi(q, mpctl, ipst)) == NULL) {
9668 return (1);
9669 }
9670
9671 if ((mpctl = ip_snmp_get_mib2_multi_rtable(q, mpctl, ipst)) == NULL) {
9672 return (1);
9673 }
9674
9675 mpctl = ip_snmp_get_mib2_ip_route_media(q, mpctl, level, ipst);
9676 if (mpctl == NULL)
9677 return (1);
9678
9679 mpctl = ip_snmp_get_mib2_ip6_route_media(q, mpctl, level, ipst);
9680 if (mpctl == NULL)
9681 return (1);
9682
9683 if ((mpctl = sctp_snmp_get_mib2(q, mpctl, sctps)) == NULL) {
9684 return (1);
9685 }
9686 if ((mpctl = ip_snmp_get_mib2_ip_dce(q, mpctl, ipst)) == NULL) {
9687 return (1);
9688 }
9689 freemsg(mpctl);
9690 return (1);
9691 }
9692
9693 /* Get global (legacy) IPv4 statistics */
9694 static mblk_t *
9695 ip_snmp_get_mib2_ip(queue_t *q, mblk_t *mpctl, mib2_ipIfStatsEntry_t *ipmib,
9696 ip_stack_t *ipst, boolean_t legacy_req)
9697 {
9698 mib2_ip_t old_ip_mib;
9699 struct opthdr *optp;
9700 mblk_t *mp2ctl;
9701 mib2_ipAddrEntry_t mae;
9702
9703 /*
9704 * make a copy of the original message
9705 */
9706 mp2ctl = copymsg(mpctl);
9707
9708 /* fixed length IP structure... */
9709 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9710 optp->level = MIB2_IP;
9711 optp->name = 0;
9712 SET_MIB(old_ip_mib.ipForwarding,
9713 (WE_ARE_FORWARDING(ipst) ? 1 : 2));
9714 SET_MIB(old_ip_mib.ipDefaultTTL,
9715 (uint32_t)ipst->ips_ip_def_ttl);
9716 SET_MIB(old_ip_mib.ipReasmTimeout,
9717 ipst->ips_ip_reassembly_timeout);
9718 SET_MIB(old_ip_mib.ipAddrEntrySize,
9719 (legacy_req) ? LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t) :
9720 sizeof (mib2_ipAddrEntry_t));
9721 SET_MIB(old_ip_mib.ipRouteEntrySize,
9722 sizeof (mib2_ipRouteEntry_t));
9723 SET_MIB(old_ip_mib.ipNetToMediaEntrySize,
9724 sizeof (mib2_ipNetToMediaEntry_t));
9725 SET_MIB(old_ip_mib.ipMemberEntrySize, sizeof (ip_member_t));
9726 SET_MIB(old_ip_mib.ipGroupSourceEntrySize, sizeof (ip_grpsrc_t));
9727 SET_MIB(old_ip_mib.ipRouteAttributeSize,
9728 sizeof (mib2_ipAttributeEntry_t));
9729 SET_MIB(old_ip_mib.transportMLPSize, sizeof (mib2_transportMLPEntry_t));
9730 SET_MIB(old_ip_mib.ipDestEntrySize, sizeof (dest_cache_entry_t));
9731
9732 /*
9733 * Grab the statistics from the new IP MIB
9734 */
9735 SET_MIB(old_ip_mib.ipInReceives,
9736 (uint32_t)ipmib->ipIfStatsHCInReceives);
9737 SET_MIB(old_ip_mib.ipInHdrErrors, ipmib->ipIfStatsInHdrErrors);
9738 SET_MIB(old_ip_mib.ipInAddrErrors, ipmib->ipIfStatsInAddrErrors);
9739 SET_MIB(old_ip_mib.ipForwDatagrams,
9740 (uint32_t)ipmib->ipIfStatsHCOutForwDatagrams);
9741 SET_MIB(old_ip_mib.ipInUnknownProtos,
9742 ipmib->ipIfStatsInUnknownProtos);
9743 SET_MIB(old_ip_mib.ipInDiscards, ipmib->ipIfStatsInDiscards);
9744 SET_MIB(old_ip_mib.ipInDelivers,
9745 (uint32_t)ipmib->ipIfStatsHCInDelivers);
9746 SET_MIB(old_ip_mib.ipOutRequests,
9747 (uint32_t)ipmib->ipIfStatsHCOutRequests);
9748 SET_MIB(old_ip_mib.ipOutDiscards, ipmib->ipIfStatsOutDiscards);
9749 SET_MIB(old_ip_mib.ipOutNoRoutes, ipmib->ipIfStatsOutNoRoutes);
9750 SET_MIB(old_ip_mib.ipReasmReqds, ipmib->ipIfStatsReasmReqds);
9751 SET_MIB(old_ip_mib.ipReasmOKs, ipmib->ipIfStatsReasmOKs);
9752 SET_MIB(old_ip_mib.ipReasmFails, ipmib->ipIfStatsReasmFails);
9753 SET_MIB(old_ip_mib.ipFragOKs, ipmib->ipIfStatsOutFragOKs);
9754 SET_MIB(old_ip_mib.ipFragFails, ipmib->ipIfStatsOutFragFails);
9755 SET_MIB(old_ip_mib.ipFragCreates, ipmib->ipIfStatsOutFragCreates);
9756
9757 /* ipRoutingDiscards is not being used */
9758 SET_MIB(old_ip_mib.ipRoutingDiscards, 0);
9759 SET_MIB(old_ip_mib.tcpInErrs, ipmib->tcpIfStatsInErrs);
9760 SET_MIB(old_ip_mib.udpNoPorts, ipmib->udpIfStatsNoPorts);
9761 SET_MIB(old_ip_mib.ipInCksumErrs, ipmib->ipIfStatsInCksumErrs);
9762 SET_MIB(old_ip_mib.ipReasmDuplicates,
9763 ipmib->ipIfStatsReasmDuplicates);
9764 SET_MIB(old_ip_mib.ipReasmPartDups, ipmib->ipIfStatsReasmPartDups);
9765 SET_MIB(old_ip_mib.ipForwProhibits, ipmib->ipIfStatsForwProhibits);
9766 SET_MIB(old_ip_mib.udpInCksumErrs, ipmib->udpIfStatsInCksumErrs);
9767 SET_MIB(old_ip_mib.udpInOverflows, ipmib->udpIfStatsInOverflows);
9768 SET_MIB(old_ip_mib.rawipInOverflows,
9769 ipmib->rawipIfStatsInOverflows);
9770
9771 SET_MIB(old_ip_mib.ipsecInSucceeded, ipmib->ipsecIfStatsInSucceeded);
9772 SET_MIB(old_ip_mib.ipsecInFailed, ipmib->ipsecIfStatsInFailed);
9773 SET_MIB(old_ip_mib.ipInIPv6, ipmib->ipIfStatsInWrongIPVersion);
9774 SET_MIB(old_ip_mib.ipOutIPv6, ipmib->ipIfStatsOutWrongIPVersion);
9775 SET_MIB(old_ip_mib.ipOutSwitchIPv6,
9776 ipmib->ipIfStatsOutSwitchIPVersion);
9777
9778 if (!snmp_append_data(mpctl->b_cont, (char *)&old_ip_mib,
9779 (int)sizeof (old_ip_mib))) {
9780 ip1dbg(("ip_snmp_get_mib2_ip: failed to allocate %u bytes\n",
9781 (uint_t)sizeof (old_ip_mib)));
9782 }
9783
9784 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9785 ip3dbg(("ip_snmp_get_mib2_ip: level %d, name %d, len %d\n",
9786 (int)optp->level, (int)optp->name, (int)optp->len));
9787 qreply(q, mpctl);
9788 return (mp2ctl);
9789 }
9790
9791 /* Per interface IPv4 statistics */
9792 static mblk_t *
9793 ip_snmp_get_mib2_ip_traffic_stats(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
9794 boolean_t legacy_req)
9795 {
9796 struct opthdr *optp;
9797 mblk_t *mp2ctl;
9798 ill_t *ill;
9799 ill_walk_context_t ctx;
9800 mblk_t *mp_tail = NULL;
9801 mib2_ipIfStatsEntry_t global_ip_mib;
9802 mib2_ipAddrEntry_t mae;
9803
9804 /*
9805 * Make a copy of the original message
9806 */
9807 mp2ctl = copymsg(mpctl);
9808
9809 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9810 optp->level = MIB2_IP;
9811 optp->name = MIB2_IP_TRAFFIC_STATS;
9812 /* Include "unknown interface" ip_mib */
9813 ipst->ips_ip_mib.ipIfStatsIPVersion = MIB2_INETADDRESSTYPE_ipv4;
9814 ipst->ips_ip_mib.ipIfStatsIfIndex =
9815 MIB2_UNKNOWN_INTERFACE; /* Flag to netstat */
9816 SET_MIB(ipst->ips_ip_mib.ipIfStatsForwarding,
9817 (ipst->ips_ip_forwarding ? 1 : 2));
9818 SET_MIB(ipst->ips_ip_mib.ipIfStatsDefaultTTL,
9819 (uint32_t)ipst->ips_ip_def_ttl);
9820 SET_MIB(ipst->ips_ip_mib.ipIfStatsEntrySize,
9821 sizeof (mib2_ipIfStatsEntry_t));
9822 SET_MIB(ipst->ips_ip_mib.ipIfStatsAddrEntrySize,
9823 sizeof (mib2_ipAddrEntry_t));
9824 SET_MIB(ipst->ips_ip_mib.ipIfStatsRouteEntrySize,
9825 sizeof (mib2_ipRouteEntry_t));
9826 SET_MIB(ipst->ips_ip_mib.ipIfStatsNetToMediaEntrySize,
9827 sizeof (mib2_ipNetToMediaEntry_t));
9828 SET_MIB(ipst->ips_ip_mib.ipIfStatsMemberEntrySize,
9829 sizeof (ip_member_t));
9830 SET_MIB(ipst->ips_ip_mib.ipIfStatsGroupSourceEntrySize,
9831 sizeof (ip_grpsrc_t));
9832
9833 bcopy(&ipst->ips_ip_mib, &global_ip_mib, sizeof (global_ip_mib));
9834
9835 if (legacy_req) {
9836 SET_MIB(global_ip_mib.ipIfStatsAddrEntrySize,
9837 LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t));
9838 }
9839
9840 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
9841 (char *)&global_ip_mib, (int)sizeof (global_ip_mib))) {
9842 ip1dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9843 "failed to allocate %u bytes\n",
9844 (uint_t)sizeof (global_ip_mib)));
9845 }
9846
9847 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
9848 ill = ILL_START_WALK_V4(&ctx, ipst);
9849 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
9850 ill->ill_ip_mib->ipIfStatsIfIndex =
9851 ill->ill_phyint->phyint_ifindex;
9852 SET_MIB(ill->ill_ip_mib->ipIfStatsForwarding,
9853 (ipst->ips_ip_forwarding ? 1 : 2));
9854 SET_MIB(ill->ill_ip_mib->ipIfStatsDefaultTTL,
9855 (uint32_t)ipst->ips_ip_def_ttl);
9856
9857 ip_mib2_add_ip_stats(&global_ip_mib, ill->ill_ip_mib);
9858 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
9859 (char *)ill->ill_ip_mib,
9860 (int)sizeof (*ill->ill_ip_mib))) {
9861 ip1dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9862 "failed to allocate %u bytes\n",
9863 (uint_t)sizeof (*ill->ill_ip_mib)));
9864 }
9865 }
9866 rw_exit(&ipst->ips_ill_g_lock);
9867
9868 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9869 ip3dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9870 "level %d, name %d, len %d\n",
9871 (int)optp->level, (int)optp->name, (int)optp->len));
9872 qreply(q, mpctl);
9873
9874 if (mp2ctl == NULL)
9875 return (NULL);
9876
9877 return (ip_snmp_get_mib2_ip(q, mp2ctl, &global_ip_mib, ipst,
9878 legacy_req));
9879 }
9880
9881 /* Global IPv4 ICMP statistics */
9882 static mblk_t *
9883 ip_snmp_get_mib2_icmp(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9884 {
9885 struct opthdr *optp;
9886 mblk_t *mp2ctl;
9887
9888 /*
9889 * Make a copy of the original message
9890 */
9891 mp2ctl = copymsg(mpctl);
9892
9893 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9894 optp->level = MIB2_ICMP;
9895 optp->name = 0;
9896 if (!snmp_append_data(mpctl->b_cont, (char *)&ipst->ips_icmp_mib,
9897 (int)sizeof (ipst->ips_icmp_mib))) {
9898 ip1dbg(("ip_snmp_get_mib2_icmp: failed to allocate %u bytes\n",
9899 (uint_t)sizeof (ipst->ips_icmp_mib)));
9900 }
9901 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9902 ip3dbg(("ip_snmp_get_mib2_icmp: level %d, name %d, len %d\n",
9903 (int)optp->level, (int)optp->name, (int)optp->len));
9904 qreply(q, mpctl);
9905 return (mp2ctl);
9906 }
9907
9908 /* Global IPv4 IGMP statistics */
9909 static mblk_t *
9910 ip_snmp_get_mib2_igmp(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9911 {
9912 struct opthdr *optp;
9913 mblk_t *mp2ctl;
9914
9915 /*
9916 * make a copy of the original message
9917 */
9918 mp2ctl = copymsg(mpctl);
9919
9920 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9921 optp->level = EXPER_IGMP;
9922 optp->name = 0;
9923 if (!snmp_append_data(mpctl->b_cont, (char *)&ipst->ips_igmpstat,
9924 (int)sizeof (ipst->ips_igmpstat))) {
9925 ip1dbg(("ip_snmp_get_mib2_igmp: failed to allocate %u bytes\n",
9926 (uint_t)sizeof (ipst->ips_igmpstat)));
9927 }
9928 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9929 ip3dbg(("ip_snmp_get_mib2_igmp: level %d, name %d, len %d\n",
9930 (int)optp->level, (int)optp->name, (int)optp->len));
9931 qreply(q, mpctl);
9932 return (mp2ctl);
9933 }
9934
9935 /* Global IPv4 Multicast Routing statistics */
9936 static mblk_t *
9937 ip_snmp_get_mib2_multi(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9938 {
9939 struct opthdr *optp;
9940 mblk_t *mp2ctl;
9941
9942 /*
9943 * make a copy of the original message
9944 */
9945 mp2ctl = copymsg(mpctl);
9946
9947 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9948 optp->level = EXPER_DVMRP;
9949 optp->name = 0;
9950 if (!ip_mroute_stats(mpctl->b_cont, ipst)) {
9951 ip0dbg(("ip_mroute_stats: failed\n"));
9952 }
9953 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9954 ip3dbg(("ip_snmp_get_mib2_multi: level %d, name %d, len %d\n",
9955 (int)optp->level, (int)optp->name, (int)optp->len));
9956 qreply(q, mpctl);
9957 return (mp2ctl);
9958 }
9959
9960 /* IPv4 address information */
9961 static mblk_t *
9962 ip_snmp_get_mib2_ip_addr(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
9963 boolean_t legacy_req)
9964 {
9965 struct opthdr *optp;
9966 mblk_t *mp2ctl;
9967 mblk_t *mp_tail = NULL;
9968 ill_t *ill;
9969 ipif_t *ipif;
9970 uint_t bitval;
9971 mib2_ipAddrEntry_t mae;
9972 size_t mae_size;
9973 zoneid_t zoneid;
9974 ill_walk_context_t ctx;
9975
9976 /*
9977 * make a copy of the original message
9978 */
9979 mp2ctl = copymsg(mpctl);
9980
9981 mae_size = (legacy_req) ? LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t) :
9982 sizeof (mib2_ipAddrEntry_t);
9983
9984 /* ipAddrEntryTable */
9985
9986 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9987 optp->level = MIB2_IP;
9988 optp->name = MIB2_IP_ADDR;
9989 zoneid = Q_TO_CONN(q)->conn_zoneid;
9990
9991 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
9992 ill = ILL_START_WALK_V4(&ctx, ipst);
9993 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
9994 for (ipif = ill->ill_ipif; ipif != NULL;
9995 ipif = ipif->ipif_next) {
9996 if (ipif->ipif_zoneid != zoneid &&
9997 ipif->ipif_zoneid != ALL_ZONES)
9998 continue;
9999 /* Sum of count from dead IRE_LO* and our current */
10000 mae.ipAdEntInfo.ae_ibcnt = ipif->ipif_ib_pkt_count;
10001 if (ipif->ipif_ire_local != NULL) {
10002 mae.ipAdEntInfo.ae_ibcnt +=
10003 ipif->ipif_ire_local->ire_ib_pkt_count;
10004 }
10005 mae.ipAdEntInfo.ae_obcnt = 0;
10006 mae.ipAdEntInfo.ae_focnt = 0;
10007
10008 ipif_get_name(ipif, mae.ipAdEntIfIndex.o_bytes,
10009 OCTET_LENGTH);
10010 mae.ipAdEntIfIndex.o_length =
10011 mi_strlen(mae.ipAdEntIfIndex.o_bytes);
10012 mae.ipAdEntAddr = ipif->ipif_lcl_addr;
10013 mae.ipAdEntNetMask = ipif->ipif_net_mask;
10014 mae.ipAdEntInfo.ae_subnet = ipif->ipif_subnet;
10015 mae.ipAdEntInfo.ae_subnet_len =
10016 ip_mask_to_plen(ipif->ipif_net_mask);
10017 mae.ipAdEntInfo.ae_src_addr = ipif->ipif_lcl_addr;
10018 for (bitval = 1;
10019 bitval &&
10020 !(bitval & ipif->ipif_brd_addr);
10021 bitval <<= 1)
10022 noop;
10023 mae.ipAdEntBcastAddr = bitval;
10024 mae.ipAdEntReasmMaxSize = IP_MAXPACKET;
10025 mae.ipAdEntInfo.ae_mtu = ipif->ipif_ill->ill_mtu;
10026 mae.ipAdEntInfo.ae_metric = ipif->ipif_ill->ill_metric;
10027 mae.ipAdEntInfo.ae_broadcast_addr =
10028 ipif->ipif_brd_addr;
10029 mae.ipAdEntInfo.ae_pp_dst_addr =
10030 ipif->ipif_pp_dst_addr;
10031 mae.ipAdEntInfo.ae_flags = ipif->ipif_flags |
10032 ill->ill_flags | ill->ill_phyint->phyint_flags;
10033 mae.ipAdEntRetransmitTime =
10034 ill->ill_reachable_retrans_time;
10035
10036 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10037 (char *)&mae, (int)mae_size)) {
10038 ip1dbg(("ip_snmp_get_mib2_ip_addr: failed to "
10039 "allocate %u bytes\n", (uint_t)mae_size));
10040 }
10041 }
10042 }
10043 rw_exit(&ipst->ips_ill_g_lock);
10044
10045 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10046 ip3dbg(("ip_snmp_get_mib2_ip_addr: level %d, name %d, len %d\n",
10047 (int)optp->level, (int)optp->name, (int)optp->len));
10048 qreply(q, mpctl);
10049 return (mp2ctl);
10050 }
10051
10052 /* IPv6 address information */
10053 static mblk_t *
10054 ip_snmp_get_mib2_ip6_addr(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
10055 boolean_t legacy_req)
10056 {
10057 struct opthdr *optp;
10058 mblk_t *mp2ctl;
10059 mblk_t *mp_tail = NULL;
10060 ill_t *ill;
10061 ipif_t *ipif;
10062 mib2_ipv6AddrEntry_t mae6;
10063 size_t mae6_size;
10064 zoneid_t zoneid;
10065 ill_walk_context_t ctx;
10066
10067 /*
10068 * make a copy of the original message
10069 */
10070 mp2ctl = copymsg(mpctl);
10071
10072 mae6_size = (legacy_req) ?
10073 LEGACY_MIB_SIZE(&mae6, mib2_ipv6AddrEntry_t) :
10074 sizeof (mib2_ipv6AddrEntry_t);
10075
10076 /* ipv6AddrEntryTable */
10077
10078 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10079 optp->level = MIB2_IP6;
10080 optp->name = MIB2_IP6_ADDR;
10081 zoneid = Q_TO_CONN(q)->conn_zoneid;
10082
10083 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10084 ill = ILL_START_WALK_V6(&ctx, ipst);
10085 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10086 for (ipif = ill->ill_ipif; ipif != NULL;
10087 ipif = ipif->ipif_next) {
10088 if (ipif->ipif_zoneid != zoneid &&
10089 ipif->ipif_zoneid != ALL_ZONES)
10090 continue;
10091 /* Sum of count from dead IRE_LO* and our current */
10092 mae6.ipv6AddrInfo.ae_ibcnt = ipif->ipif_ib_pkt_count;
10093 if (ipif->ipif_ire_local != NULL) {
10094 mae6.ipv6AddrInfo.ae_ibcnt +=
10095 ipif->ipif_ire_local->ire_ib_pkt_count;
10096 }
10097 mae6.ipv6AddrInfo.ae_obcnt = 0;
10098 mae6.ipv6AddrInfo.ae_focnt = 0;
10099
10100 ipif_get_name(ipif, mae6.ipv6AddrIfIndex.o_bytes,
10101 OCTET_LENGTH);
10102 mae6.ipv6AddrIfIndex.o_length =
10103 mi_strlen(mae6.ipv6AddrIfIndex.o_bytes);
10104 mae6.ipv6AddrAddress = ipif->ipif_v6lcl_addr;
10105 mae6.ipv6AddrPfxLength =
10106 ip_mask_to_plen_v6(&ipif->ipif_v6net_mask);
10107 mae6.ipv6AddrInfo.ae_subnet = ipif->ipif_v6subnet;
10108 mae6.ipv6AddrInfo.ae_subnet_len =
10109 mae6.ipv6AddrPfxLength;
10110 mae6.ipv6AddrInfo.ae_src_addr = ipif->ipif_v6lcl_addr;
10111
10112 /* Type: stateless(1), stateful(2), unknown(3) */
10113 if (ipif->ipif_flags & IPIF_ADDRCONF)
10114 mae6.ipv6AddrType = 1;
10115 else
10116 mae6.ipv6AddrType = 2;
10117 /* Anycast: true(1), false(2) */
10118 if (ipif->ipif_flags & IPIF_ANYCAST)
10119 mae6.ipv6AddrAnycastFlag = 1;
10120 else
10121 mae6.ipv6AddrAnycastFlag = 2;
10122
10123 /*
10124 * Address status: preferred(1), deprecated(2),
10125 * invalid(3), inaccessible(4), unknown(5)
10126 */
10127 if (ipif->ipif_flags & IPIF_NOLOCAL)
10128 mae6.ipv6AddrStatus = 3;
10129 else if (ipif->ipif_flags & IPIF_DEPRECATED)
10130 mae6.ipv6AddrStatus = 2;
10131 else
10132 mae6.ipv6AddrStatus = 1;
10133 mae6.ipv6AddrInfo.ae_mtu = ipif->ipif_ill->ill_mtu;
10134 mae6.ipv6AddrInfo.ae_metric =
10135 ipif->ipif_ill->ill_metric;
10136 mae6.ipv6AddrInfo.ae_pp_dst_addr =
10137 ipif->ipif_v6pp_dst_addr;
10138 mae6.ipv6AddrInfo.ae_flags = ipif->ipif_flags |
10139 ill->ill_flags | ill->ill_phyint->phyint_flags;
10140 mae6.ipv6AddrReasmMaxSize = IP_MAXPACKET;
10141 mae6.ipv6AddrIdentifier = ill->ill_token;
10142 mae6.ipv6AddrIdentifierLen = ill->ill_token_length;
10143 mae6.ipv6AddrReachableTime = ill->ill_reachable_time;
10144 mae6.ipv6AddrRetransmitTime =
10145 ill->ill_reachable_retrans_time;
10146 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10147 (char *)&mae6, (int)mae6_size)) {
10148 ip1dbg(("ip_snmp_get_mib2_ip6_addr: failed to "
10149 "allocate %u bytes\n",
10150 (uint_t)mae6_size));
10151 }
10152 }
10153 }
10154 rw_exit(&ipst->ips_ill_g_lock);
10155
10156 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10157 ip3dbg(("ip_snmp_get_mib2_ip6_addr: level %d, name %d, len %d\n",
10158 (int)optp->level, (int)optp->name, (int)optp->len));
10159 qreply(q, mpctl);
10160 return (mp2ctl);
10161 }
10162
10163 /* IPv4 multicast group membership. */
10164 static mblk_t *
10165 ip_snmp_get_mib2_ip_group_mem(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10166 {
10167 struct opthdr *optp;
10168 mblk_t *mp2ctl;
10169 ill_t *ill;
10170 ipif_t *ipif;
10171 ilm_t *ilm;
10172 ip_member_t ipm;
10173 mblk_t *mp_tail = NULL;
10174 ill_walk_context_t ctx;
10175 zoneid_t zoneid;
10176
10177 /*
10178 * make a copy of the original message
10179 */
10180 mp2ctl = copymsg(mpctl);
10181 zoneid = Q_TO_CONN(q)->conn_zoneid;
10182
10183 /* ipGroupMember table */
10184 optp = (struct opthdr *)&mpctl->b_rptr[
10185 sizeof (struct T_optmgmt_ack)];
10186 optp->level = MIB2_IP;
10187 optp->name = EXPER_IP_GROUP_MEMBERSHIP;
10188
10189 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10190 ill = ILL_START_WALK_V4(&ctx, ipst);
10191 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10192 /* Make sure the ill isn't going away. */
10193 if (!ill_check_and_refhold(ill))
10194 continue;
10195 rw_exit(&ipst->ips_ill_g_lock);
10196 rw_enter(&ill->ill_mcast_lock, RW_READER);
10197 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10198 if (ilm->ilm_zoneid != zoneid &&
10199 ilm->ilm_zoneid != ALL_ZONES)
10200 continue;
10201
10202 /* Is there an ipif for ilm_ifaddr? */
10203 for (ipif = ill->ill_ipif; ipif != NULL;
10204 ipif = ipif->ipif_next) {
10205 if (!IPIF_IS_CONDEMNED(ipif) &&
10206 ipif->ipif_lcl_addr == ilm->ilm_ifaddr &&
10207 ilm->ilm_ifaddr != INADDR_ANY)
10208 break;
10209 }
10210 if (ipif != NULL) {
10211 ipif_get_name(ipif,
10212 ipm.ipGroupMemberIfIndex.o_bytes,
10213 OCTET_LENGTH);
10214 } else {
10215 ill_get_name(ill,
10216 ipm.ipGroupMemberIfIndex.o_bytes,
10217 OCTET_LENGTH);
10218 }
10219 ipm.ipGroupMemberIfIndex.o_length =
10220 mi_strlen(ipm.ipGroupMemberIfIndex.o_bytes);
10221
10222 ipm.ipGroupMemberAddress = ilm->ilm_addr;
10223 ipm.ipGroupMemberRefCnt = ilm->ilm_refcnt;
10224 ipm.ipGroupMemberFilterMode = ilm->ilm_fmode;
10225 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10226 (char *)&ipm, (int)sizeof (ipm))) {
10227 ip1dbg(("ip_snmp_get_mib2_ip_group: "
10228 "failed to allocate %u bytes\n",
10229 (uint_t)sizeof (ipm)));
10230 }
10231 }
10232 rw_exit(&ill->ill_mcast_lock);
10233 ill_refrele(ill);
10234 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10235 }
10236 rw_exit(&ipst->ips_ill_g_lock);
10237 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10238 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10239 (int)optp->level, (int)optp->name, (int)optp->len));
10240 qreply(q, mpctl);
10241 return (mp2ctl);
10242 }
10243
10244 /* IPv6 multicast group membership. */
10245 static mblk_t *
10246 ip_snmp_get_mib2_ip6_group_mem(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10247 {
10248 struct opthdr *optp;
10249 mblk_t *mp2ctl;
10250 ill_t *ill;
10251 ilm_t *ilm;
10252 ipv6_member_t ipm6;
10253 mblk_t *mp_tail = NULL;
10254 ill_walk_context_t ctx;
10255 zoneid_t zoneid;
10256
10257 /*
10258 * make a copy of the original message
10259 */
10260 mp2ctl = copymsg(mpctl);
10261 zoneid = Q_TO_CONN(q)->conn_zoneid;
10262
10263 /* ip6GroupMember table */
10264 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10265 optp->level = MIB2_IP6;
10266 optp->name = EXPER_IP6_GROUP_MEMBERSHIP;
10267
10268 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10269 ill = ILL_START_WALK_V6(&ctx, ipst);
10270 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10271 /* Make sure the ill isn't going away. */
10272 if (!ill_check_and_refhold(ill))
10273 continue;
10274 rw_exit(&ipst->ips_ill_g_lock);
10275 /*
10276 * Normally we don't have any members on under IPMP interfaces.
10277 * We report them as a debugging aid.
10278 */
10279 rw_enter(&ill->ill_mcast_lock, RW_READER);
10280 ipm6.ipv6GroupMemberIfIndex = ill->ill_phyint->phyint_ifindex;
10281 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10282 if (ilm->ilm_zoneid != zoneid &&
10283 ilm->ilm_zoneid != ALL_ZONES)
10284 continue; /* not this zone */
10285 ipm6.ipv6GroupMemberAddress = ilm->ilm_v6addr;
10286 ipm6.ipv6GroupMemberRefCnt = ilm->ilm_refcnt;
10287 ipm6.ipv6GroupMemberFilterMode = ilm->ilm_fmode;
10288 if (!snmp_append_data2(mpctl->b_cont,
10289 &mp_tail,
10290 (char *)&ipm6, (int)sizeof (ipm6))) {
10291 ip1dbg(("ip_snmp_get_mib2_ip6_group: "
10292 "failed to allocate %u bytes\n",
10293 (uint_t)sizeof (ipm6)));
10294 }
10295 }
10296 rw_exit(&ill->ill_mcast_lock);
10297 ill_refrele(ill);
10298 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10299 }
10300 rw_exit(&ipst->ips_ill_g_lock);
10301
10302 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10303 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10304 (int)optp->level, (int)optp->name, (int)optp->len));
10305 qreply(q, mpctl);
10306 return (mp2ctl);
10307 }
10308
10309 /* IP multicast filtered sources */
10310 static mblk_t *
10311 ip_snmp_get_mib2_ip_group_src(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10312 {
10313 struct opthdr *optp;
10314 mblk_t *mp2ctl;
10315 ill_t *ill;
10316 ipif_t *ipif;
10317 ilm_t *ilm;
10318 ip_grpsrc_t ips;
10319 mblk_t *mp_tail = NULL;
10320 ill_walk_context_t ctx;
10321 zoneid_t zoneid;
10322 int i;
10323 slist_t *sl;
10324
10325 /*
10326 * make a copy of the original message
10327 */
10328 mp2ctl = copymsg(mpctl);
10329 zoneid = Q_TO_CONN(q)->conn_zoneid;
10330
10331 /* ipGroupSource table */
10332 optp = (struct opthdr *)&mpctl->b_rptr[
10333 sizeof (struct T_optmgmt_ack)];
10334 optp->level = MIB2_IP;
10335 optp->name = EXPER_IP_GROUP_SOURCES;
10336
10337 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10338 ill = ILL_START_WALK_V4(&ctx, ipst);
10339 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10340 /* Make sure the ill isn't going away. */
10341 if (!ill_check_and_refhold(ill))
10342 continue;
10343 rw_exit(&ipst->ips_ill_g_lock);
10344 rw_enter(&ill->ill_mcast_lock, RW_READER);
10345 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10346 sl = ilm->ilm_filter;
10347 if (ilm->ilm_zoneid != zoneid &&
10348 ilm->ilm_zoneid != ALL_ZONES)
10349 continue;
10350 if (SLIST_IS_EMPTY(sl))
10351 continue;
10352
10353 /* Is there an ipif for ilm_ifaddr? */
10354 for (ipif = ill->ill_ipif; ipif != NULL;
10355 ipif = ipif->ipif_next) {
10356 if (!IPIF_IS_CONDEMNED(ipif) &&
10357 ipif->ipif_lcl_addr == ilm->ilm_ifaddr &&
10358 ilm->ilm_ifaddr != INADDR_ANY)
10359 break;
10360 }
10361 if (ipif != NULL) {
10362 ipif_get_name(ipif,
10363 ips.ipGroupSourceIfIndex.o_bytes,
10364 OCTET_LENGTH);
10365 } else {
10366 ill_get_name(ill,
10367 ips.ipGroupSourceIfIndex.o_bytes,
10368 OCTET_LENGTH);
10369 }
10370 ips.ipGroupSourceIfIndex.o_length =
10371 mi_strlen(ips.ipGroupSourceIfIndex.o_bytes);
10372
10373 ips.ipGroupSourceGroup = ilm->ilm_addr;
10374 for (i = 0; i < sl->sl_numsrc; i++) {
10375 if (!IN6_IS_ADDR_V4MAPPED(&sl->sl_addr[i]))
10376 continue;
10377 IN6_V4MAPPED_TO_IPADDR(&sl->sl_addr[i],
10378 ips.ipGroupSourceAddress);
10379 if (snmp_append_data2(mpctl->b_cont, &mp_tail,
10380 (char *)&ips, (int)sizeof (ips)) == 0) {
10381 ip1dbg(("ip_snmp_get_mib2_ip_group_src:"
10382 " failed to allocate %u bytes\n",
10383 (uint_t)sizeof (ips)));
10384 }
10385 }
10386 }
10387 rw_exit(&ill->ill_mcast_lock);
10388 ill_refrele(ill);
10389 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10390 }
10391 rw_exit(&ipst->ips_ill_g_lock);
10392 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10393 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10394 (int)optp->level, (int)optp->name, (int)optp->len));
10395 qreply(q, mpctl);
10396 return (mp2ctl);
10397 }
10398
10399 /* IPv6 multicast filtered sources. */
10400 static mblk_t *
10401 ip_snmp_get_mib2_ip6_group_src(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10402 {
10403 struct opthdr *optp;
10404 mblk_t *mp2ctl;
10405 ill_t *ill;
10406 ilm_t *ilm;
10407 ipv6_grpsrc_t ips6;
10408 mblk_t *mp_tail = NULL;
10409 ill_walk_context_t ctx;
10410 zoneid_t zoneid;
10411 int i;
10412 slist_t *sl;
10413
10414 /*
10415 * make a copy of the original message
10416 */
10417 mp2ctl = copymsg(mpctl);
10418 zoneid = Q_TO_CONN(q)->conn_zoneid;
10419
10420 /* ip6GroupMember table */
10421 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10422 optp->level = MIB2_IP6;
10423 optp->name = EXPER_IP6_GROUP_SOURCES;
10424
10425 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10426 ill = ILL_START_WALK_V6(&ctx, ipst);
10427 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10428 /* Make sure the ill isn't going away. */
10429 if (!ill_check_and_refhold(ill))
10430 continue;
10431 rw_exit(&ipst->ips_ill_g_lock);
10432 /*
10433 * Normally we don't have any members on under IPMP interfaces.
10434 * We report them as a debugging aid.
10435 */
10436 rw_enter(&ill->ill_mcast_lock, RW_READER);
10437 ips6.ipv6GroupSourceIfIndex = ill->ill_phyint->phyint_ifindex;
10438 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10439 sl = ilm->ilm_filter;
10440 if (ilm->ilm_zoneid != zoneid &&
10441 ilm->ilm_zoneid != ALL_ZONES)
10442 continue;
10443 if (SLIST_IS_EMPTY(sl))
10444 continue;
10445 ips6.ipv6GroupSourceGroup = ilm->ilm_v6addr;
10446 for (i = 0; i < sl->sl_numsrc; i++) {
10447 ips6.ipv6GroupSourceAddress = sl->sl_addr[i];
10448 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10449 (char *)&ips6, (int)sizeof (ips6))) {
10450 ip1dbg(("ip_snmp_get_mib2_ip6_"
10451 "group_src: failed to allocate "
10452 "%u bytes\n",
10453 (uint_t)sizeof (ips6)));
10454 }
10455 }
10456 }
10457 rw_exit(&ill->ill_mcast_lock);
10458 ill_refrele(ill);
10459 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10460 }
10461 rw_exit(&ipst->ips_ill_g_lock);
10462
10463 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10464 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10465 (int)optp->level, (int)optp->name, (int)optp->len));
10466 qreply(q, mpctl);
10467 return (mp2ctl);
10468 }
10469
10470 /* Multicast routing virtual interface table. */
10471 static mblk_t *
10472 ip_snmp_get_mib2_virt_multi(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10473 {
10474 struct opthdr *optp;
10475 mblk_t *mp2ctl;
10476
10477 /*
10478 * make a copy of the original message
10479 */
10480 mp2ctl = copymsg(mpctl);
10481
10482 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10483 optp->level = EXPER_DVMRP;
10484 optp->name = EXPER_DVMRP_VIF;
10485 if (!ip_mroute_vif(mpctl->b_cont, ipst)) {
10486 ip0dbg(("ip_mroute_vif: failed\n"));
10487 }
10488 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10489 ip3dbg(("ip_snmp_get_mib2_virt_multi: level %d, name %d, len %d\n",
10490 (int)optp->level, (int)optp->name, (int)optp->len));
10491 qreply(q, mpctl);
10492 return (mp2ctl);
10493 }
10494
10495 /* Multicast routing table. */
10496 static mblk_t *
10497 ip_snmp_get_mib2_multi_rtable(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10498 {
10499 struct opthdr *optp;
10500 mblk_t *mp2ctl;
10501
10502 /*
10503 * make a copy of the original message
10504 */
10505 mp2ctl = copymsg(mpctl);
10506
10507 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10508 optp->level = EXPER_DVMRP;
10509 optp->name = EXPER_DVMRP_MRT;
10510 if (!ip_mroute_mrt(mpctl->b_cont, ipst)) {
10511 ip0dbg(("ip_mroute_mrt: failed\n"));
10512 }
10513 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10514 ip3dbg(("ip_snmp_get_mib2_multi_rtable: level %d, name %d, len %d\n",
10515 (int)optp->level, (int)optp->name, (int)optp->len));
10516 qreply(q, mpctl);
10517 return (mp2ctl);
10518 }
10519
10520 /*
10521 * Return ipRouteEntryTable, ipNetToMediaEntryTable, and ipRouteAttributeTable
10522 * in one IRE walk.
10523 */
10524 static mblk_t *
10525 ip_snmp_get_mib2_ip_route_media(queue_t *q, mblk_t *mpctl, int level,
10526 ip_stack_t *ipst)
10527 {
10528 struct opthdr *optp;
10529 mblk_t *mp2ctl; /* Returned */
10530 mblk_t *mp3ctl; /* nettomedia */
10531 mblk_t *mp4ctl; /* routeattrs */
10532 iproutedata_t ird;
10533 zoneid_t zoneid;
10534
10535 /*
10536 * make copies of the original message
10537 * - mp2ctl is returned unchanged to the caller for his use
10538 * - mpctl is sent upstream as ipRouteEntryTable
10539 * - mp3ctl is sent upstream as ipNetToMediaEntryTable
10540 * - mp4ctl is sent upstream as ipRouteAttributeTable
10541 */
10542 mp2ctl = copymsg(mpctl);
10543 mp3ctl = copymsg(mpctl);
10544 mp4ctl = copymsg(mpctl);
10545 if (mp3ctl == NULL || mp4ctl == NULL) {
10546 freemsg(mp4ctl);
10547 freemsg(mp3ctl);
10548 freemsg(mp2ctl);
10549 freemsg(mpctl);
10550 return (NULL);
10551 }
10552
10553 bzero(&ird, sizeof (ird));
10554
10555 ird.ird_route.lp_head = mpctl->b_cont;
10556 ird.ird_netmedia.lp_head = mp3ctl->b_cont;
10557 ird.ird_attrs.lp_head = mp4ctl->b_cont;
10558 /*
10559 * If the level has been set the special EXPER_IP_AND_ALL_IRES value,
10560 * then also include ire_testhidden IREs and IRE_IF_CLONE. This is
10561 * intended a temporary solution until a proper MIB API is provided
10562 * that provides complete filtering/caller-opt-in.
10563 */
10564 if (level == EXPER_IP_AND_ALL_IRES)
10565 ird.ird_flags |= IRD_REPORT_ALL;
10566
10567 zoneid = Q_TO_CONN(q)->conn_zoneid;
10568 ire_walk_v4(ip_snmp_get2_v4, &ird, zoneid, ipst);
10569
10570 /* ipRouteEntryTable in mpctl */
10571 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10572 optp->level = MIB2_IP;
10573 optp->name = MIB2_IP_ROUTE;
10574 optp->len = msgdsize(ird.ird_route.lp_head);
10575 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10576 (int)optp->level, (int)optp->name, (int)optp->len));
10577 qreply(q, mpctl);
10578
10579 /* ipNetToMediaEntryTable in mp3ctl */
10580 ncec_walk(NULL, ip_snmp_get2_v4_media, &ird, ipst);
10581
10582 optp = (struct opthdr *)&mp3ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10583 optp->level = MIB2_IP;
10584 optp->name = MIB2_IP_MEDIA;
10585 optp->len = msgdsize(ird.ird_netmedia.lp_head);
10586 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10587 (int)optp->level, (int)optp->name, (int)optp->len));
10588 qreply(q, mp3ctl);
10589
10590 /* ipRouteAttributeTable in mp4ctl */
10591 optp = (struct opthdr *)&mp4ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10592 optp->level = MIB2_IP;
10593 optp->name = EXPER_IP_RTATTR;
10594 optp->len = msgdsize(ird.ird_attrs.lp_head);
10595 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10596 (int)optp->level, (int)optp->name, (int)optp->len));
10597 if (optp->len == 0)
10598 freemsg(mp4ctl);
10599 else
10600 qreply(q, mp4ctl);
10601
10602 return (mp2ctl);
10603 }
10604
10605 /*
10606 * Return ipv6RouteEntryTable and ipv6RouteAttributeTable in one IRE walk, and
10607 * ipv6NetToMediaEntryTable in an NDP walk.
10608 */
10609 static mblk_t *
10610 ip_snmp_get_mib2_ip6_route_media(queue_t *q, mblk_t *mpctl, int level,
10611 ip_stack_t *ipst)
10612 {
10613 struct opthdr *optp;
10614 mblk_t *mp2ctl; /* Returned */
10615 mblk_t *mp3ctl; /* nettomedia */
10616 mblk_t *mp4ctl; /* routeattrs */
10617 iproutedata_t ird;
10618 zoneid_t zoneid;
10619
10620 /*
10621 * make copies of the original message
10622 * - mp2ctl is returned unchanged to the caller for his use
10623 * - mpctl is sent upstream as ipv6RouteEntryTable
10624 * - mp3ctl is sent upstream as ipv6NetToMediaEntryTable
10625 * - mp4ctl is sent upstream as ipv6RouteAttributeTable
10626 */
10627 mp2ctl = copymsg(mpctl);
10628 mp3ctl = copymsg(mpctl);
10629 mp4ctl = copymsg(mpctl);
10630 if (mp3ctl == NULL || mp4ctl == NULL) {
10631 freemsg(mp4ctl);
10632 freemsg(mp3ctl);
10633 freemsg(mp2ctl);
10634 freemsg(mpctl);
10635 return (NULL);
10636 }
10637
10638 bzero(&ird, sizeof (ird));
10639
10640 ird.ird_route.lp_head = mpctl->b_cont;
10641 ird.ird_netmedia.lp_head = mp3ctl->b_cont;
10642 ird.ird_attrs.lp_head = mp4ctl->b_cont;
10643 /*
10644 * If the level has been set the special EXPER_IP_AND_ALL_IRES value,
10645 * then also include ire_testhidden IREs and IRE_IF_CLONE. This is
10646 * intended a temporary solution until a proper MIB API is provided
10647 * that provides complete filtering/caller-opt-in.
10648 */
10649 if (level == EXPER_IP_AND_ALL_IRES)
10650 ird.ird_flags |= IRD_REPORT_ALL;
10651
10652 zoneid = Q_TO_CONN(q)->conn_zoneid;
10653 ire_walk_v6(ip_snmp_get2_v6_route, &ird, zoneid, ipst);
10654
10655 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10656 optp->level = MIB2_IP6;
10657 optp->name = MIB2_IP6_ROUTE;
10658 optp->len = msgdsize(ird.ird_route.lp_head);
10659 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10660 (int)optp->level, (int)optp->name, (int)optp->len));
10661 qreply(q, mpctl);
10662
10663 /* ipv6NetToMediaEntryTable in mp3ctl */
10664 ncec_walk(NULL, ip_snmp_get2_v6_media, &ird, ipst);
10665
10666 optp = (struct opthdr *)&mp3ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10667 optp->level = MIB2_IP6;
10668 optp->name = MIB2_IP6_MEDIA;
10669 optp->len = msgdsize(ird.ird_netmedia.lp_head);
10670 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10671 (int)optp->level, (int)optp->name, (int)optp->len));
10672 qreply(q, mp3ctl);
10673
10674 /* ipv6RouteAttributeTable in mp4ctl */
10675 optp = (struct opthdr *)&mp4ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10676 optp->level = MIB2_IP6;
10677 optp->name = EXPER_IP_RTATTR;
10678 optp->len = msgdsize(ird.ird_attrs.lp_head);
10679 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10680 (int)optp->level, (int)optp->name, (int)optp->len));
10681 if (optp->len == 0)
10682 freemsg(mp4ctl);
10683 else
10684 qreply(q, mp4ctl);
10685
10686 return (mp2ctl);
10687 }
10688
10689 /*
10690 * IPv6 mib: One per ill
10691 */
10692 static mblk_t *
10693 ip_snmp_get_mib2_ip6(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
10694 boolean_t legacy_req)
10695 {
10696 struct opthdr *optp;
10697 mblk_t *mp2ctl;
10698 ill_t *ill;
10699 ill_walk_context_t ctx;
10700 mblk_t *mp_tail = NULL;
10701 mib2_ipv6AddrEntry_t mae6;
10702 mib2_ipIfStatsEntry_t *ise;
10703 size_t ise_size, iae_size;
10704
10705 /*
10706 * Make a copy of the original message
10707 */
10708 mp2ctl = copymsg(mpctl);
10709
10710 /* fixed length IPv6 structure ... */
10711
10712 if (legacy_req) {
10713 ise_size = LEGACY_MIB_SIZE(&ipst->ips_ip6_mib,
10714 mib2_ipIfStatsEntry_t);
10715 iae_size = LEGACY_MIB_SIZE(&mae6, mib2_ipv6AddrEntry_t);
10716 } else {
10717 ise_size = sizeof (mib2_ipIfStatsEntry_t);
10718 iae_size = sizeof (mib2_ipv6AddrEntry_t);
10719 }
10720
10721 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10722 optp->level = MIB2_IP6;
10723 optp->name = 0;
10724 /* Include "unknown interface" ip6_mib */
10725 ipst->ips_ip6_mib.ipIfStatsIPVersion = MIB2_INETADDRESSTYPE_ipv6;
10726 ipst->ips_ip6_mib.ipIfStatsIfIndex =
10727 MIB2_UNKNOWN_INTERFACE; /* Flag to netstat */
10728 SET_MIB(ipst->ips_ip6_mib.ipIfStatsForwarding,
10729 ipst->ips_ipv6_forwarding ? 1 : 2);
10730 SET_MIB(ipst->ips_ip6_mib.ipIfStatsDefaultHopLimit,
10731 ipst->ips_ipv6_def_hops);
10732 SET_MIB(ipst->ips_ip6_mib.ipIfStatsEntrySize,
10733 sizeof (mib2_ipIfStatsEntry_t));
10734 SET_MIB(ipst->ips_ip6_mib.ipIfStatsAddrEntrySize,
10735 sizeof (mib2_ipv6AddrEntry_t));
10736 SET_MIB(ipst->ips_ip6_mib.ipIfStatsRouteEntrySize,
10737 sizeof (mib2_ipv6RouteEntry_t));
10738 SET_MIB(ipst->ips_ip6_mib.ipIfStatsNetToMediaEntrySize,
10739 sizeof (mib2_ipv6NetToMediaEntry_t));
10740 SET_MIB(ipst->ips_ip6_mib.ipIfStatsMemberEntrySize,
10741 sizeof (ipv6_member_t));
10742 SET_MIB(ipst->ips_ip6_mib.ipIfStatsGroupSourceEntrySize,
10743 sizeof (ipv6_grpsrc_t));
10744
10745 /*
10746 * Synchronize 64- and 32-bit counters
10747 */
10748 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInReceives,
10749 ipIfStatsHCInReceives);
10750 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInDelivers,
10751 ipIfStatsHCInDelivers);
10752 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutRequests,
10753 ipIfStatsHCOutRequests);
10754 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutForwDatagrams,
10755 ipIfStatsHCOutForwDatagrams);
10756 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutMcastPkts,
10757 ipIfStatsHCOutMcastPkts);
10758 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInMcastPkts,
10759 ipIfStatsHCInMcastPkts);
10760
10761 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10762 (char *)&ipst->ips_ip6_mib, (int)ise_size)) {
10763 ip1dbg(("ip_snmp_get_mib2_ip6: failed to allocate %u bytes\n",
10764 (uint_t)ise_size));
10765 } else if (legacy_req) {
10766 /* Adjust the EntrySize fields for legacy requests. */
10767 ise =
10768 (mib2_ipIfStatsEntry_t *)(mp_tail->b_wptr - (int)ise_size);
10769 SET_MIB(ise->ipIfStatsEntrySize, ise_size);
10770 SET_MIB(ise->ipIfStatsAddrEntrySize, iae_size);
10771 }
10772
10773 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10774 ill = ILL_START_WALK_V6(&ctx, ipst);
10775 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10776 ill->ill_ip_mib->ipIfStatsIfIndex =
10777 ill->ill_phyint->phyint_ifindex;
10778 SET_MIB(ill->ill_ip_mib->ipIfStatsForwarding,
10779 ipst->ips_ipv6_forwarding ? 1 : 2);
10780 SET_MIB(ill->ill_ip_mib->ipIfStatsDefaultHopLimit,
10781 ill->ill_max_hops);
10782
10783 /*
10784 * Synchronize 64- and 32-bit counters
10785 */
10786 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInReceives,
10787 ipIfStatsHCInReceives);
10788 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInDelivers,
10789 ipIfStatsHCInDelivers);
10790 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutRequests,
10791 ipIfStatsHCOutRequests);
10792 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutForwDatagrams,
10793 ipIfStatsHCOutForwDatagrams);
10794 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutMcastPkts,
10795 ipIfStatsHCOutMcastPkts);
10796 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInMcastPkts,
10797 ipIfStatsHCInMcastPkts);
10798
10799 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10800 (char *)ill->ill_ip_mib, (int)ise_size)) {
10801 ip1dbg(("ip_snmp_get_mib2_ip6: failed to allocate "
10802 "%u bytes\n", (uint_t)ise_size));
10803 } else if (legacy_req) {
10804 /* Adjust the EntrySize fields for legacy requests. */
10805 ise = (mib2_ipIfStatsEntry_t *)(mp_tail->b_wptr -
10806 (int)ise_size);
10807 SET_MIB(ise->ipIfStatsEntrySize, ise_size);
10808 SET_MIB(ise->ipIfStatsAddrEntrySize, iae_size);
10809 }
10810 }
10811 rw_exit(&ipst->ips_ill_g_lock);
10812
10813 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10814 ip3dbg(("ip_snmp_get_mib2_ip6: level %d, name %d, len %d\n",
10815 (int)optp->level, (int)optp->name, (int)optp->len));
10816 qreply(q, mpctl);
10817 return (mp2ctl);
10818 }
10819
10820 /*
10821 * ICMPv6 mib: One per ill
10822 */
10823 static mblk_t *
10824 ip_snmp_get_mib2_icmp6(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10825 {
10826 struct opthdr *optp;
10827 mblk_t *mp2ctl;
10828 ill_t *ill;
10829 ill_walk_context_t ctx;
10830 mblk_t *mp_tail = NULL;
10831 /*
10832 * Make a copy of the original message
10833 */
10834 mp2ctl = copymsg(mpctl);
10835
10836 /* fixed length ICMPv6 structure ... */
10837
10838 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10839 optp->level = MIB2_ICMP6;
10840 optp->name = 0;
10841 /* Include "unknown interface" icmp6_mib */
10842 ipst->ips_icmp6_mib.ipv6IfIcmpIfIndex =
10843 MIB2_UNKNOWN_INTERFACE; /* netstat flag */
10844 ipst->ips_icmp6_mib.ipv6IfIcmpEntrySize =
10845 sizeof (mib2_ipv6IfIcmpEntry_t);
10846 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10847 (char *)&ipst->ips_icmp6_mib,
10848 (int)sizeof (ipst->ips_icmp6_mib))) {
10849 ip1dbg(("ip_snmp_get_mib2_icmp6: failed to allocate %u bytes\n",
10850 (uint_t)sizeof (ipst->ips_icmp6_mib)));
10851 }
10852
10853 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10854 ill = ILL_START_WALK_V6(&ctx, ipst);
10855 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10856 ill->ill_icmp6_mib->ipv6IfIcmpIfIndex =
10857 ill->ill_phyint->phyint_ifindex;
10858 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10859 (char *)ill->ill_icmp6_mib,
10860 (int)sizeof (*ill->ill_icmp6_mib))) {
10861 ip1dbg(("ip_snmp_get_mib2_icmp6: failed to allocate "
10862 "%u bytes\n",
10863 (uint_t)sizeof (*ill->ill_icmp6_mib)));
10864 }
10865 }
10866 rw_exit(&ipst->ips_ill_g_lock);
10867
10868 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10869 ip3dbg(("ip_snmp_get_mib2_icmp6: level %d, name %d, len %d\n",
10870 (int)optp->level, (int)optp->name, (int)optp->len));
10871 qreply(q, mpctl);
10872 return (mp2ctl);
10873 }
10874
10875 /*
10876 * ire_walk routine to create both ipRouteEntryTable and
10877 * ipRouteAttributeTable in one IRE walk
10878 */
10879 static void
10880 ip_snmp_get2_v4(ire_t *ire, iproutedata_t *ird)
10881 {
10882 ill_t *ill;
10883 mib2_ipRouteEntry_t *re;
10884 mib2_ipAttributeEntry_t iaes;
10885 tsol_ire_gw_secattr_t *attrp;
10886 tsol_gc_t *gc = NULL;
10887 tsol_gcgrp_t *gcgrp = NULL;
10888 ip_stack_t *ipst = ire->ire_ipst;
10889
10890 ASSERT(ire->ire_ipversion == IPV4_VERSION);
10891
10892 if (!(ird->ird_flags & IRD_REPORT_ALL)) {
10893 if (ire->ire_testhidden)
10894 return;
10895 if (ire->ire_type & IRE_IF_CLONE)
10896 return;
10897 }
10898
10899 if ((re = kmem_zalloc(sizeof (*re), KM_NOSLEEP)) == NULL)
10900 return;
10901
10902 if ((attrp = ire->ire_gw_secattr) != NULL) {
10903 mutex_enter(&attrp->igsa_lock);
10904 if ((gc = attrp->igsa_gc) != NULL) {
10905 gcgrp = gc->gc_grp;
10906 ASSERT(gcgrp != NULL);
10907 rw_enter(&gcgrp->gcgrp_rwlock, RW_READER);
10908 }
10909 mutex_exit(&attrp->igsa_lock);
10910 }
10911 /*
10912 * Return all IRE types for route table... let caller pick and choose
10913 */
10914 re->ipRouteDest = ire->ire_addr;
10915 ill = ire->ire_ill;
10916 re->ipRouteIfIndex.o_length = 0;
10917 if (ill != NULL) {
10918 ill_get_name(ill, re->ipRouteIfIndex.o_bytes, OCTET_LENGTH);
10919 re->ipRouteIfIndex.o_length =
10920 mi_strlen(re->ipRouteIfIndex.o_bytes);
10921 }
10922 re->ipRouteMetric1 = -1;
10923 re->ipRouteMetric2 = -1;
10924 re->ipRouteMetric3 = -1;
10925 re->ipRouteMetric4 = -1;
10926
10927 re->ipRouteNextHop = ire->ire_gateway_addr;
10928 /* indirect(4), direct(3), or invalid(2) */
10929 if (ire->ire_flags & (RTF_REJECT | RTF_BLACKHOLE))
10930 re->ipRouteType = 2;
10931 else if (ire->ire_type & IRE_ONLINK)
10932 re->ipRouteType = 3;
10933 else
10934 re->ipRouteType = 4;
10935
10936 re->ipRouteProto = -1;
10937 re->ipRouteAge = gethrestime_sec() - ire->ire_create_time;
10938 re->ipRouteMask = ire->ire_mask;
10939 re->ipRouteMetric5 = -1;
10940 re->ipRouteInfo.re_max_frag = ire->ire_metrics.iulp_mtu;
10941 if (ire->ire_ill != NULL && re->ipRouteInfo.re_max_frag == 0)
10942 re->ipRouteInfo.re_max_frag = ire->ire_ill->ill_mtu;
10943
10944 re->ipRouteInfo.re_frag_flag = 0;
10945 re->ipRouteInfo.re_rtt = 0;
10946 re->ipRouteInfo.re_src_addr = 0;
10947 re->ipRouteInfo.re_ref = ire->ire_refcnt;
10948 re->ipRouteInfo.re_obpkt = ire->ire_ob_pkt_count;
10949 re->ipRouteInfo.re_ibpkt = ire->ire_ib_pkt_count;
10950 re->ipRouteInfo.re_flags = ire->ire_flags;
10951
10952 /* Add the IRE_IF_CLONE's counters to their parent IRE_INTERFACE */
10953 if (ire->ire_type & IRE_INTERFACE) {
10954 ire_t *child;
10955
10956 rw_enter(&ipst->ips_ire_dep_lock, RW_READER);
10957 child = ire->ire_dep_children;
10958 while (child != NULL) {
10959 re->ipRouteInfo.re_obpkt += child->ire_ob_pkt_count;
10960 re->ipRouteInfo.re_ibpkt += child->ire_ib_pkt_count;
10961 child = child->ire_dep_sib_next;
10962 }
10963 rw_exit(&ipst->ips_ire_dep_lock);
10964 }
10965
10966 if (ire->ire_flags & RTF_DYNAMIC) {
10967 re->ipRouteInfo.re_ire_type = IRE_HOST_REDIRECT;
10968 } else {
10969 re->ipRouteInfo.re_ire_type = ire->ire_type;
10970 }
10971
10972 if (!snmp_append_data2(ird->ird_route.lp_head, &ird->ird_route.lp_tail,
10973 (char *)re, (int)sizeof (*re))) {
10974 ip1dbg(("ip_snmp_get2_v4: failed to allocate %u bytes\n",
10975 (uint_t)sizeof (*re)));
10976 }
10977
10978 if (gc != NULL) {
10979 iaes.iae_routeidx = ird->ird_idx;
10980 iaes.iae_doi = gc->gc_db->gcdb_doi;
10981 iaes.iae_slrange = gc->gc_db->gcdb_slrange;
10982
10983 if (!snmp_append_data2(ird->ird_attrs.lp_head,
10984 &ird->ird_attrs.lp_tail, (char *)&iaes, sizeof (iaes))) {
10985 ip1dbg(("ip_snmp_get2_v4: failed to allocate %u "
10986 "bytes\n", (uint_t)sizeof (iaes)));
10987 }
10988 }
10989
10990 /* bump route index for next pass */
10991 ird->ird_idx++;
10992
10993 kmem_free(re, sizeof (*re));
10994 if (gcgrp != NULL)
10995 rw_exit(&gcgrp->gcgrp_rwlock);
10996 }
10997
10998 /*
10999 * ire_walk routine to create ipv6RouteEntryTable and ipRouteEntryTable.
11000 */
11001 static void
11002 ip_snmp_get2_v6_route(ire_t *ire, iproutedata_t *ird)
11003 {
11004 ill_t *ill;
11005 mib2_ipv6RouteEntry_t *re;
11006 mib2_ipAttributeEntry_t iaes;
11007 tsol_ire_gw_secattr_t *attrp;
11008 tsol_gc_t *gc = NULL;
11009 tsol_gcgrp_t *gcgrp = NULL;
11010 ip_stack_t *ipst = ire->ire_ipst;
11011
11012 ASSERT(ire->ire_ipversion == IPV6_VERSION);
11013
11014 if (!(ird->ird_flags & IRD_REPORT_ALL)) {
11015 if (ire->ire_testhidden)
11016 return;
11017 if (ire->ire_type & IRE_IF_CLONE)
11018 return;
11019 }
11020
11021 if ((re = kmem_zalloc(sizeof (*re), KM_NOSLEEP)) == NULL)
11022 return;
11023
11024 if ((attrp = ire->ire_gw_secattr) != NULL) {
11025 mutex_enter(&attrp->igsa_lock);
11026 if ((gc = attrp->igsa_gc) != NULL) {
11027 gcgrp = gc->gc_grp;
11028 ASSERT(gcgrp != NULL);
11029 rw_enter(&gcgrp->gcgrp_rwlock, RW_READER);
11030 }
11031 mutex_exit(&attrp->igsa_lock);
11032 }
11033 /*
11034 * Return all IRE types for route table... let caller pick and choose
11035 */
11036 re->ipv6RouteDest = ire->ire_addr_v6;
11037 re->ipv6RoutePfxLength = ip_mask_to_plen_v6(&ire->ire_mask_v6);
11038 re->ipv6RouteIndex = 0; /* Unique when multiple with same dest/plen */
11039 re->ipv6RouteIfIndex.o_length = 0;
11040 ill = ire->ire_ill;
11041 if (ill != NULL) {
11042 ill_get_name(ill, re->ipv6RouteIfIndex.o_bytes, OCTET_LENGTH);
11043 re->ipv6RouteIfIndex.o_length =
11044 mi_strlen(re->ipv6RouteIfIndex.o_bytes);
11045 }
11046
11047 ASSERT(!(ire->ire_type & IRE_BROADCAST));
11048
11049 mutex_enter(&ire->ire_lock);
11050 re->ipv6RouteNextHop = ire->ire_gateway_addr_v6;
11051 mutex_exit(&ire->ire_lock);
11052
11053 /* remote(4), local(3), or discard(2) */
11054 if (ire->ire_flags & (RTF_REJECT | RTF_BLACKHOLE))
11055 re->ipv6RouteType = 2;
11056 else if (ire->ire_type & IRE_ONLINK)
11057 re->ipv6RouteType = 3;
11058 else
11059 re->ipv6RouteType = 4;
11060
11061 re->ipv6RouteProtocol = -1;
11062 re->ipv6RoutePolicy = 0;
11063 re->ipv6RouteAge = gethrestime_sec() - ire->ire_create_time;
11064 re->ipv6RouteNextHopRDI = 0;
11065 re->ipv6RouteWeight = 0;
11066 re->ipv6RouteMetric = 0;
11067 re->ipv6RouteInfo.re_max_frag = ire->ire_metrics.iulp_mtu;
11068 if (ire->ire_ill != NULL && re->ipv6RouteInfo.re_max_frag == 0)
11069 re->ipv6RouteInfo.re_max_frag = ire->ire_ill->ill_mtu;
11070
11071 re->ipv6RouteInfo.re_frag_flag = 0;
11072 re->ipv6RouteInfo.re_rtt = 0;
11073 re->ipv6RouteInfo.re_src_addr = ipv6_all_zeros;
11074 re->ipv6RouteInfo.re_obpkt = ire->ire_ob_pkt_count;
11075 re->ipv6RouteInfo.re_ibpkt = ire->ire_ib_pkt_count;
11076 re->ipv6RouteInfo.re_ref = ire->ire_refcnt;
11077 re->ipv6RouteInfo.re_flags = ire->ire_flags;
11078
11079 /* Add the IRE_IF_CLONE's counters to their parent IRE_INTERFACE */
11080 if (ire->ire_type & IRE_INTERFACE) {
11081 ire_t *child;
11082
11083 rw_enter(&ipst->ips_ire_dep_lock, RW_READER);
11084 child = ire->ire_dep_children;
11085 while (child != NULL) {
11086 re->ipv6RouteInfo.re_obpkt += child->ire_ob_pkt_count;
11087 re->ipv6RouteInfo.re_ibpkt += child->ire_ib_pkt_count;
11088 child = child->ire_dep_sib_next;
11089 }
11090 rw_exit(&ipst->ips_ire_dep_lock);
11091 }
11092 if (ire->ire_flags & RTF_DYNAMIC) {
11093 re->ipv6RouteInfo.re_ire_type = IRE_HOST_REDIRECT;
11094 } else {
11095 re->ipv6RouteInfo.re_ire_type = ire->ire_type;
11096 }
11097
11098 if (!snmp_append_data2(ird->ird_route.lp_head, &ird->ird_route.lp_tail,
11099 (char *)re, (int)sizeof (*re))) {
11100 ip1dbg(("ip_snmp_get2_v6: failed to allocate %u bytes\n",
11101 (uint_t)sizeof (*re)));
11102 }
11103
11104 if (gc != NULL) {
11105 iaes.iae_routeidx = ird->ird_idx;
11106 iaes.iae_doi = gc->gc_db->gcdb_doi;
11107 iaes.iae_slrange = gc->gc_db->gcdb_slrange;
11108
11109 if (!snmp_append_data2(ird->ird_attrs.lp_head,
11110 &ird->ird_attrs.lp_tail, (char *)&iaes, sizeof (iaes))) {
11111 ip1dbg(("ip_snmp_get2_v6: failed to allocate %u "
11112 "bytes\n", (uint_t)sizeof (iaes)));
11113 }
11114 }
11115
11116 /* bump route index for next pass */
11117 ird->ird_idx++;
11118
11119 kmem_free(re, sizeof (*re));
11120 if (gcgrp != NULL)
11121 rw_exit(&gcgrp->gcgrp_rwlock);
11122 }
11123
11124 /*
11125 * ncec_walk routine to create ipv6NetToMediaEntryTable
11126 */
11127 static int
11128 ip_snmp_get2_v6_media(ncec_t *ncec, iproutedata_t *ird)
11129 {
11130 ill_t *ill;
11131 mib2_ipv6NetToMediaEntry_t ntme;
11132
11133 ill = ncec->ncec_ill;
11134 /* skip arpce entries, and loopback ncec entries */
11135 if (ill->ill_isv6 == B_FALSE || ill->ill_net_type == IRE_LOOPBACK)
11136 return (0);
11137 /*
11138 * Neighbor cache entry attached to IRE with on-link
11139 * destination.
11140 * We report all IPMP groups on ncec_ill which is normally the upper.
11141 */
11142 ntme.ipv6NetToMediaIfIndex = ill->ill_phyint->phyint_ifindex;
11143 ntme.ipv6NetToMediaNetAddress = ncec->ncec_addr;
11144 ntme.ipv6NetToMediaPhysAddress.o_length = ill->ill_phys_addr_length;
11145 if (ncec->ncec_lladdr != NULL) {
11146 bcopy(ncec->ncec_lladdr, ntme.ipv6NetToMediaPhysAddress.o_bytes,
11147 ntme.ipv6NetToMediaPhysAddress.o_length);
11148 }
11149 /*
11150 * Note: Returns ND_* states. Should be:
11151 * reachable(1), stale(2), delay(3), probe(4),
11152 * invalid(5), unknown(6)
11153 */
11154 ntme.ipv6NetToMediaState = ncec->ncec_state;
11155 ntme.ipv6NetToMediaLastUpdated = 0;
11156
11157 /* other(1), dynamic(2), static(3), local(4) */
11158 if (NCE_MYADDR(ncec)) {
11159 ntme.ipv6NetToMediaType = 4;
11160 } else if (ncec->ncec_flags & NCE_F_PUBLISH) {
11161 ntme.ipv6NetToMediaType = 1; /* proxy */
11162 } else if (ncec->ncec_flags & NCE_F_STATIC) {
11163 ntme.ipv6NetToMediaType = 3;
11164 } else if (ncec->ncec_flags & (NCE_F_MCAST|NCE_F_BCAST)) {
11165 ntme.ipv6NetToMediaType = 1;
11166 } else {
11167 ntme.ipv6NetToMediaType = 2;
11168 }
11169
11170 if (!snmp_append_data2(ird->ird_netmedia.lp_head,
11171 &ird->ird_netmedia.lp_tail, (char *)&ntme, sizeof (ntme))) {
11172 ip1dbg(("ip_snmp_get2_v6_media: failed to allocate %u bytes\n",
11173 (uint_t)sizeof (ntme)));
11174 }
11175 return (0);
11176 }
11177
11178 int
11179 nce2ace(ncec_t *ncec)
11180 {
11181 int flags = 0;
11182
11183 if (NCE_ISREACHABLE(ncec))
11184 flags |= ACE_F_RESOLVED;
11185 if (ncec->ncec_flags & NCE_F_AUTHORITY)
11186 flags |= ACE_F_AUTHORITY;
11187 if (ncec->ncec_flags & NCE_F_PUBLISH)
11188 flags |= ACE_F_PUBLISH;
11189 if ((ncec->ncec_flags & NCE_F_NONUD) != 0)
11190 flags |= ACE_F_PERMANENT;
11191 if (NCE_MYADDR(ncec))
11192 flags |= (ACE_F_MYADDR | ACE_F_AUTHORITY);
11193 if (ncec->ncec_flags & NCE_F_UNVERIFIED)
11194 flags |= ACE_F_UNVERIFIED;
11195 if (ncec->ncec_flags & NCE_F_AUTHORITY)
11196 flags |= ACE_F_AUTHORITY;
11197 if (ncec->ncec_flags & NCE_F_DELAYED)
11198 flags |= ACE_F_DELAYED;
11199 return (flags);
11200 }
11201
11202 /*
11203 * ncec_walk routine to create ipNetToMediaEntryTable
11204 */
11205 static int
11206 ip_snmp_get2_v4_media(ncec_t *ncec, iproutedata_t *ird)
11207 {
11208 ill_t *ill;
11209 mib2_ipNetToMediaEntry_t ntme;
11210 const char *name = "unknown";
11211 ipaddr_t ncec_addr;
11212
11213 ill = ncec->ncec_ill;
11214 if (ill->ill_isv6 || (ncec->ncec_flags & NCE_F_BCAST) ||
11215 ill->ill_net_type == IRE_LOOPBACK)
11216 return (0);
11217
11218 /* We report all IPMP groups on ncec_ill which is normally the upper. */
11219 name = ill->ill_name;
11220 /* Based on RFC 4293: other(1), inval(2), dyn(3), stat(4) */
11221 if (NCE_MYADDR(ncec)) {
11222 ntme.ipNetToMediaType = 4;
11223 } else if (ncec->ncec_flags & (NCE_F_MCAST|NCE_F_BCAST|NCE_F_PUBLISH)) {
11224 ntme.ipNetToMediaType = 1;
11225 } else {
11226 ntme.ipNetToMediaType = 3;
11227 }
11228 ntme.ipNetToMediaIfIndex.o_length = MIN(OCTET_LENGTH, strlen(name));
11229 bcopy(name, ntme.ipNetToMediaIfIndex.o_bytes,
11230 ntme.ipNetToMediaIfIndex.o_length);
11231
11232 IN6_V4MAPPED_TO_IPADDR(&ncec->ncec_addr, ncec_addr);
11233 bcopy(&ncec_addr, &ntme.ipNetToMediaNetAddress, sizeof (ncec_addr));
11234
11235 ntme.ipNetToMediaInfo.ntm_mask.o_length = sizeof (ipaddr_t);
11236 ncec_addr = INADDR_BROADCAST;
11237 bcopy(&ncec_addr, ntme.ipNetToMediaInfo.ntm_mask.o_bytes,
11238 sizeof (ncec_addr));
11239 /*
11240 * map all the flags to the ACE counterpart.
11241 */
11242 ntme.ipNetToMediaInfo.ntm_flags = nce2ace(ncec);
11243
11244 ntme.ipNetToMediaPhysAddress.o_length =
11245 MIN(OCTET_LENGTH, ill->ill_phys_addr_length);
11246
11247 if (!NCE_ISREACHABLE(ncec))
11248 ntme.ipNetToMediaPhysAddress.o_length = 0;
11249 else {
11250 if (ncec->ncec_lladdr != NULL) {
11251 bcopy(ncec->ncec_lladdr,
11252 ntme.ipNetToMediaPhysAddress.o_bytes,
11253 ntme.ipNetToMediaPhysAddress.o_length);
11254 }
11255 }
11256
11257 if (!snmp_append_data2(ird->ird_netmedia.lp_head,
11258 &ird->ird_netmedia.lp_tail, (char *)&ntme, sizeof (ntme))) {
11259 ip1dbg(("ip_snmp_get2_v4_media: failed to allocate %u bytes\n",
11260 (uint_t)sizeof (ntme)));
11261 }
11262 return (0);
11263 }
11264
11265 /*
11266 * return (0) if invalid set request, 1 otherwise, including non-tcp requests
11267 */
11268 /* ARGSUSED */
11269 int
11270 ip_snmp_set(queue_t *q, int level, int name, uchar_t *ptr, int len)
11271 {
11272 switch (level) {
11273 case MIB2_IP:
11274 case MIB2_ICMP:
11275 switch (name) {
11276 default:
11277 break;
11278 }
11279 return (1);
11280 default:
11281 return (1);
11282 }
11283 }
11284
11285 /*
11286 * When there exists both a 64- and 32-bit counter of a particular type
11287 * (i.e., InReceives), only the 64-bit counters are added.
11288 */
11289 void
11290 ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *o1, mib2_ipIfStatsEntry_t *o2)
11291 {
11292 UPDATE_MIB(o1, ipIfStatsInHdrErrors, o2->ipIfStatsInHdrErrors);
11293 UPDATE_MIB(o1, ipIfStatsInTooBigErrors, o2->ipIfStatsInTooBigErrors);
11294 UPDATE_MIB(o1, ipIfStatsInNoRoutes, o2->ipIfStatsInNoRoutes);
11295 UPDATE_MIB(o1, ipIfStatsInAddrErrors, o2->ipIfStatsInAddrErrors);
11296 UPDATE_MIB(o1, ipIfStatsInUnknownProtos, o2->ipIfStatsInUnknownProtos);
11297 UPDATE_MIB(o1, ipIfStatsInTruncatedPkts, o2->ipIfStatsInTruncatedPkts);
11298 UPDATE_MIB(o1, ipIfStatsInDiscards, o2->ipIfStatsInDiscards);
11299 UPDATE_MIB(o1, ipIfStatsOutDiscards, o2->ipIfStatsOutDiscards);
11300 UPDATE_MIB(o1, ipIfStatsOutFragOKs, o2->ipIfStatsOutFragOKs);
11301 UPDATE_MIB(o1, ipIfStatsOutFragFails, o2->ipIfStatsOutFragFails);
11302 UPDATE_MIB(o1, ipIfStatsOutFragCreates, o2->ipIfStatsOutFragCreates);
11303 UPDATE_MIB(o1, ipIfStatsReasmReqds, o2->ipIfStatsReasmReqds);
11304 UPDATE_MIB(o1, ipIfStatsReasmOKs, o2->ipIfStatsReasmOKs);
11305 UPDATE_MIB(o1, ipIfStatsReasmFails, o2->ipIfStatsReasmFails);
11306 UPDATE_MIB(o1, ipIfStatsOutNoRoutes, o2->ipIfStatsOutNoRoutes);
11307 UPDATE_MIB(o1, ipIfStatsReasmDuplicates, o2->ipIfStatsReasmDuplicates);
11308 UPDATE_MIB(o1, ipIfStatsReasmPartDups, o2->ipIfStatsReasmPartDups);
11309 UPDATE_MIB(o1, ipIfStatsForwProhibits, o2->ipIfStatsForwProhibits);
11310 UPDATE_MIB(o1, udpInCksumErrs, o2->udpInCksumErrs);
11311 UPDATE_MIB(o1, udpInOverflows, o2->udpInOverflows);
11312 UPDATE_MIB(o1, rawipInOverflows, o2->rawipInOverflows);
11313 UPDATE_MIB(o1, ipIfStatsInWrongIPVersion,
11314 o2->ipIfStatsInWrongIPVersion);
11315 UPDATE_MIB(o1, ipIfStatsOutWrongIPVersion,
11316 o2->ipIfStatsInWrongIPVersion);
11317 UPDATE_MIB(o1, ipIfStatsOutSwitchIPVersion,
11318 o2->ipIfStatsOutSwitchIPVersion);
11319 UPDATE_MIB(o1, ipIfStatsHCInReceives, o2->ipIfStatsHCInReceives);
11320 UPDATE_MIB(o1, ipIfStatsHCInOctets, o2->ipIfStatsHCInOctets);
11321 UPDATE_MIB(o1, ipIfStatsHCInForwDatagrams,
11322 o2->ipIfStatsHCInForwDatagrams);
11323 UPDATE_MIB(o1, ipIfStatsHCInDelivers, o2->ipIfStatsHCInDelivers);
11324 UPDATE_MIB(o1, ipIfStatsHCOutRequests, o2->ipIfStatsHCOutRequests);
11325 UPDATE_MIB(o1, ipIfStatsHCOutForwDatagrams,
11326 o2->ipIfStatsHCOutForwDatagrams);
11327 UPDATE_MIB(o1, ipIfStatsOutFragReqds, o2->ipIfStatsOutFragReqds);
11328 UPDATE_MIB(o1, ipIfStatsHCOutTransmits, o2->ipIfStatsHCOutTransmits);
11329 UPDATE_MIB(o1, ipIfStatsHCOutOctets, o2->ipIfStatsHCOutOctets);
11330 UPDATE_MIB(o1, ipIfStatsHCInMcastPkts, o2->ipIfStatsHCInMcastPkts);
11331 UPDATE_MIB(o1, ipIfStatsHCInMcastOctets, o2->ipIfStatsHCInMcastOctets);
11332 UPDATE_MIB(o1, ipIfStatsHCOutMcastPkts, o2->ipIfStatsHCOutMcastPkts);
11333 UPDATE_MIB(o1, ipIfStatsHCOutMcastOctets,
11334 o2->ipIfStatsHCOutMcastOctets);
11335 UPDATE_MIB(o1, ipIfStatsHCInBcastPkts, o2->ipIfStatsHCInBcastPkts);
11336 UPDATE_MIB(o1, ipIfStatsHCOutBcastPkts, o2->ipIfStatsHCOutBcastPkts);
11337 UPDATE_MIB(o1, ipsecInSucceeded, o2->ipsecInSucceeded);
11338 UPDATE_MIB(o1, ipsecInFailed, o2->ipsecInFailed);
11339 UPDATE_MIB(o1, ipInCksumErrs, o2->ipInCksumErrs);
11340 UPDATE_MIB(o1, tcpInErrs, o2->tcpInErrs);
11341 UPDATE_MIB(o1, udpNoPorts, o2->udpNoPorts);
11342 }
11343
11344 void
11345 ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *o1, mib2_ipv6IfIcmpEntry_t *o2)
11346 {
11347 UPDATE_MIB(o1, ipv6IfIcmpInMsgs, o2->ipv6IfIcmpInMsgs);
11348 UPDATE_MIB(o1, ipv6IfIcmpInErrors, o2->ipv6IfIcmpInErrors);
11349 UPDATE_MIB(o1, ipv6IfIcmpInDestUnreachs, o2->ipv6IfIcmpInDestUnreachs);
11350 UPDATE_MIB(o1, ipv6IfIcmpInAdminProhibs, o2->ipv6IfIcmpInAdminProhibs);
11351 UPDATE_MIB(o1, ipv6IfIcmpInTimeExcds, o2->ipv6IfIcmpInTimeExcds);
11352 UPDATE_MIB(o1, ipv6IfIcmpInParmProblems, o2->ipv6IfIcmpInParmProblems);
11353 UPDATE_MIB(o1, ipv6IfIcmpInPktTooBigs, o2->ipv6IfIcmpInPktTooBigs);
11354 UPDATE_MIB(o1, ipv6IfIcmpInEchos, o2->ipv6IfIcmpInEchos);
11355 UPDATE_MIB(o1, ipv6IfIcmpInEchoReplies, o2->ipv6IfIcmpInEchoReplies);
11356 UPDATE_MIB(o1, ipv6IfIcmpInRouterSolicits,
11357 o2->ipv6IfIcmpInRouterSolicits);
11358 UPDATE_MIB(o1, ipv6IfIcmpInRouterAdvertisements,
11359 o2->ipv6IfIcmpInRouterAdvertisements);
11360 UPDATE_MIB(o1, ipv6IfIcmpInNeighborSolicits,
11361 o2->ipv6IfIcmpInNeighborSolicits);
11362 UPDATE_MIB(o1, ipv6IfIcmpInNeighborAdvertisements,
11363 o2->ipv6IfIcmpInNeighborAdvertisements);
11364 UPDATE_MIB(o1, ipv6IfIcmpInRedirects, o2->ipv6IfIcmpInRedirects);
11365 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembQueries,
11366 o2->ipv6IfIcmpInGroupMembQueries);
11367 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembResponses,
11368 o2->ipv6IfIcmpInGroupMembResponses);
11369 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembReductions,
11370 o2->ipv6IfIcmpInGroupMembReductions);
11371 UPDATE_MIB(o1, ipv6IfIcmpOutMsgs, o2->ipv6IfIcmpOutMsgs);
11372 UPDATE_MIB(o1, ipv6IfIcmpOutErrors, o2->ipv6IfIcmpOutErrors);
11373 UPDATE_MIB(o1, ipv6IfIcmpOutDestUnreachs,
11374 o2->ipv6IfIcmpOutDestUnreachs);
11375 UPDATE_MIB(o1, ipv6IfIcmpOutAdminProhibs,
11376 o2->ipv6IfIcmpOutAdminProhibs);
11377 UPDATE_MIB(o1, ipv6IfIcmpOutTimeExcds, o2->ipv6IfIcmpOutTimeExcds);
11378 UPDATE_MIB(o1, ipv6IfIcmpOutParmProblems,
11379 o2->ipv6IfIcmpOutParmProblems);
11380 UPDATE_MIB(o1, ipv6IfIcmpOutPktTooBigs, o2->ipv6IfIcmpOutPktTooBigs);
11381 UPDATE_MIB(o1, ipv6IfIcmpOutEchos, o2->ipv6IfIcmpOutEchos);
11382 UPDATE_MIB(o1, ipv6IfIcmpOutEchoReplies, o2->ipv6IfIcmpOutEchoReplies);
11383 UPDATE_MIB(o1, ipv6IfIcmpOutRouterSolicits,
11384 o2->ipv6IfIcmpOutRouterSolicits);
11385 UPDATE_MIB(o1, ipv6IfIcmpOutRouterAdvertisements,
11386 o2->ipv6IfIcmpOutRouterAdvertisements);
11387 UPDATE_MIB(o1, ipv6IfIcmpOutNeighborSolicits,
11388 o2->ipv6IfIcmpOutNeighborSolicits);
11389 UPDATE_MIB(o1, ipv6IfIcmpOutNeighborAdvertisements,
11390 o2->ipv6IfIcmpOutNeighborAdvertisements);
11391 UPDATE_MIB(o1, ipv6IfIcmpOutRedirects, o2->ipv6IfIcmpOutRedirects);
11392 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembQueries,
11393 o2->ipv6IfIcmpOutGroupMembQueries);
11394 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembResponses,
11395 o2->ipv6IfIcmpOutGroupMembResponses);
11396 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembReductions,
11397 o2->ipv6IfIcmpOutGroupMembReductions);
11398 UPDATE_MIB(o1, ipv6IfIcmpInOverflows, o2->ipv6IfIcmpInOverflows);
11399 UPDATE_MIB(o1, ipv6IfIcmpBadHoplimit, o2->ipv6IfIcmpBadHoplimit);
11400 UPDATE_MIB(o1, ipv6IfIcmpInBadNeighborAdvertisements,
11401 o2->ipv6IfIcmpInBadNeighborAdvertisements);
11402 UPDATE_MIB(o1, ipv6IfIcmpInBadNeighborSolicitations,
11403 o2->ipv6IfIcmpInBadNeighborSolicitations);
11404 UPDATE_MIB(o1, ipv6IfIcmpInBadRedirects, o2->ipv6IfIcmpInBadRedirects);
11405 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembTotal,
11406 o2->ipv6IfIcmpInGroupMembTotal);
11407 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembBadQueries,
11408 o2->ipv6IfIcmpInGroupMembBadQueries);
11409 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembBadReports,
11410 o2->ipv6IfIcmpInGroupMembBadReports);
11411 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembOurReports,
11412 o2->ipv6IfIcmpInGroupMembOurReports);
11413 }
11414
11415 /*
11416 * Called before the options are updated to check if this packet will
11417 * be source routed from here.
11418 * This routine assumes that the options are well formed i.e. that they
11419 * have already been checked.
11420 */
11421 boolean_t
11422 ip_source_routed(ipha_t *ipha, ip_stack_t *ipst)
11423 {
11424 ipoptp_t opts;
11425 uchar_t *opt;
11426 uint8_t optval;
11427 uint8_t optlen;
11428 ipaddr_t dst;
11429
11430 if (IS_SIMPLE_IPH(ipha)) {
11431 ip2dbg(("not source routed\n"));
11432 return (B_FALSE);
11433 }
11434 dst = ipha->ipha_dst;
11435 for (optval = ipoptp_first(&opts, ipha);
11436 optval != IPOPT_EOL;
11437 optval = ipoptp_next(&opts)) {
11438 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11439 opt = opts.ipoptp_cur;
11440 optlen = opts.ipoptp_len;
11441 ip2dbg(("ip_source_routed: opt %d, len %d\n",
11442 optval, optlen));
11443 switch (optval) {
11444 uint32_t off;
11445 case IPOPT_SSRR:
11446 case IPOPT_LSRR:
11447 /*
11448 * If dst is one of our addresses and there are some
11449 * entries left in the source route return (true).
11450 */
11451 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
11452 ip2dbg(("ip_source_routed: not next"
11453 " source route 0x%x\n",
11454 ntohl(dst)));
11455 return (B_FALSE);
11456 }
11457 off = opt[IPOPT_OFFSET];
11458 off--;
11459 if (optlen < IP_ADDR_LEN ||
11460 off > optlen - IP_ADDR_LEN) {
11461 /* End of source route */
11462 ip1dbg(("ip_source_routed: end of SR\n"));
11463 return (B_FALSE);
11464 }
11465 return (B_TRUE);
11466 }
11467 }
11468 ip2dbg(("not source routed\n"));
11469 return (B_FALSE);
11470 }
11471
11472 /*
11473 * ip_unbind is called by the transports to remove a conn from
11474 * the fanout table.
11475 */
11476 void
11477 ip_unbind(conn_t *connp)
11478 {
11479
11480 ASSERT(!MUTEX_HELD(&connp->conn_lock));
11481
11482 if (is_system_labeled() && connp->conn_anon_port) {
11483 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
11484 connp->conn_mlp_type, connp->conn_proto,
11485 ntohs(connp->conn_lport), B_FALSE);
11486 connp->conn_anon_port = 0;
11487 }
11488 connp->conn_mlp_type = mlptSingle;
11489
11490 ipcl_hash_remove(connp);
11491 }
11492
11493 /*
11494 * Used for deciding the MSS size for the upper layer. Thus
11495 * we need to check the outbound policy values in the conn.
11496 */
11497 int
11498 conn_ipsec_length(conn_t *connp)
11499 {
11500 ipsec_latch_t *ipl;
11501
11502 ipl = connp->conn_latch;
11503 if (ipl == NULL)
11504 return (0);
11505
11506 if (connp->conn_ixa->ixa_ipsec_policy == NULL)
11507 return (0);
11508
11509 return (connp->conn_ixa->ixa_ipsec_policy->ipsp_act->ipa_ovhd);
11510 }
11511
11512 /*
11513 * Returns an estimate of the IPsec headers size. This is used if
11514 * we don't want to call into IPsec to get the exact size.
11515 */
11516 int
11517 ipsec_out_extra_length(ip_xmit_attr_t *ixa)
11518 {
11519 ipsec_action_t *a;
11520
11521 if (!(ixa->ixa_flags & IXAF_IPSEC_SECURE))
11522 return (0);
11523
11524 a = ixa->ixa_ipsec_action;
11525 if (a == NULL) {
11526 ASSERT(ixa->ixa_ipsec_policy != NULL);
11527 a = ixa->ixa_ipsec_policy->ipsp_act;
11528 }
11529 ASSERT(a != NULL);
11530
11531 return (a->ipa_ovhd);
11532 }
11533
11534 /*
11535 * If there are any source route options, return the true final
11536 * destination. Otherwise, return the destination.
11537 */
11538 ipaddr_t
11539 ip_get_dst(ipha_t *ipha)
11540 {
11541 ipoptp_t opts;
11542 uchar_t *opt;
11543 uint8_t optval;
11544 uint8_t optlen;
11545 ipaddr_t dst;
11546 uint32_t off;
11547
11548 dst = ipha->ipha_dst;
11549
11550 if (IS_SIMPLE_IPH(ipha))
11551 return (dst);
11552
11553 for (optval = ipoptp_first(&opts, ipha);
11554 optval != IPOPT_EOL;
11555 optval = ipoptp_next(&opts)) {
11556 opt = opts.ipoptp_cur;
11557 optlen = opts.ipoptp_len;
11558 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11559 switch (optval) {
11560 case IPOPT_SSRR:
11561 case IPOPT_LSRR:
11562 off = opt[IPOPT_OFFSET];
11563 /*
11564 * If one of the conditions is true, it means
11565 * end of options and dst already has the right
11566 * value.
11567 */
11568 if (!(optlen < IP_ADDR_LEN || off > optlen - 3)) {
11569 off = optlen - IP_ADDR_LEN;
11570 bcopy(&opt[off], &dst, IP_ADDR_LEN);
11571 }
11572 return (dst);
11573 default:
11574 break;
11575 }
11576 }
11577
11578 return (dst);
11579 }
11580
11581 /*
11582 * Outbound IP fragmentation routine.
11583 * Assumes the caller has checked whether or not fragmentation should
11584 * be allowed. Here we copy the DF bit from the header to all the generated
11585 * fragments.
11586 */
11587 int
11588 ip_fragment_v4(mblk_t *mp_orig, nce_t *nce, iaflags_t ixaflags,
11589 uint_t pkt_len, uint32_t max_frag, uint32_t xmit_hint, zoneid_t szone,
11590 zoneid_t nolzid, pfirepostfrag_t postfragfn, uintptr_t *ixa_cookie)
11591 {
11592 int i1;
11593 int hdr_len;
11594 mblk_t *hdr_mp;
11595 ipha_t *ipha;
11596 int ip_data_end;
11597 int len;
11598 mblk_t *mp = mp_orig;
11599 int offset;
11600 ill_t *ill = nce->nce_ill;
11601 ip_stack_t *ipst = ill->ill_ipst;
11602 mblk_t *carve_mp;
11603 uint32_t frag_flag;
11604 uint_t priority = mp->b_band;
11605 int error = 0;
11606
11607 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragReqds);
11608
11609 if (pkt_len != msgdsize(mp)) {
11610 ip0dbg(("Packet length mismatch: %d, %ld\n",
11611 pkt_len, msgdsize(mp)));
11612 freemsg(mp);
11613 return (EINVAL);
11614 }
11615
11616 if (max_frag == 0) {
11617 ip1dbg(("ip_fragment_v4: max_frag is zero. Dropping packet\n"));
11618 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11619 ip_drop_output("FragFails: zero max_frag", mp, ill);
11620 freemsg(mp);
11621 return (EINVAL);
11622 }
11623
11624 ASSERT(MBLKL(mp) >= sizeof (ipha_t));
11625 ipha = (ipha_t *)mp->b_rptr;
11626 ASSERT(ntohs(ipha->ipha_length) == pkt_len);
11627 frag_flag = ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_DF;
11628
11629 /*
11630 * Establish the starting offset. May not be zero if we are fragging
11631 * a fragment that is being forwarded.
11632 */
11633 offset = ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_OFFSET;
11634
11635 /* TODO why is this test needed? */
11636 if (((max_frag - ntohs(ipha->ipha_length)) & ~7) < 8) {
11637 /* TODO: notify ulp somehow */
11638 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11639 ip_drop_output("FragFails: bad starting offset", mp, ill);
11640 freemsg(mp);
11641 return (EINVAL);
11642 }
11643
11644 hdr_len = IPH_HDR_LENGTH(ipha);
11645 ipha->ipha_hdr_checksum = 0;
11646
11647 /*
11648 * Establish the number of bytes maximum per frag, after putting
11649 * in the header.
11650 */
11651 len = (max_frag - hdr_len) & ~7;
11652
11653 /* Get a copy of the header for the trailing frags */
11654 hdr_mp = ip_fragment_copyhdr((uchar_t *)ipha, hdr_len, offset, ipst,
11655 mp);
11656 if (hdr_mp == NULL) {
11657 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11658 ip_drop_output("FragFails: no hdr_mp", mp, ill);
11659 freemsg(mp);
11660 return (ENOBUFS);
11661 }
11662
11663 /* Store the starting offset, with the MoreFrags flag. */
11664 i1 = offset | IPH_MF | frag_flag;
11665 ipha->ipha_fragment_offset_and_flags = htons((uint16_t)i1);
11666
11667 /* Establish the ending byte offset, based on the starting offset. */
11668 offset <<= 3;
11669 ip_data_end = offset + ntohs(ipha->ipha_length) - hdr_len;
11670
11671 /* Store the length of the first fragment in the IP header. */
11672 i1 = len + hdr_len;
11673 ASSERT(i1 <= IP_MAXPACKET);
11674 ipha->ipha_length = htons((uint16_t)i1);
11675
11676 /*
11677 * Compute the IP header checksum for the first frag. We have to
11678 * watch out that we stop at the end of the header.
11679 */
11680 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
11681
11682 /*
11683 * Now carve off the first frag. Note that this will include the
11684 * original IP header.
11685 */
11686 if (!(mp = ip_carve_mp(&mp_orig, i1))) {
11687 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11688 ip_drop_output("FragFails: could not carve mp", mp_orig, ill);
11689 freeb(hdr_mp);
11690 freemsg(mp_orig);
11691 return (ENOBUFS);
11692 }
11693
11694 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragCreates);
11695
11696 error = postfragfn(mp, nce, ixaflags, i1, xmit_hint, szone, nolzid,
11697 ixa_cookie);
11698 if (error != 0 && error != EWOULDBLOCK) {
11699 /* No point in sending the other fragments */
11700 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11701 ip_drop_output("FragFails: postfragfn failed", mp_orig, ill);
11702 freeb(hdr_mp);
11703 freemsg(mp_orig);
11704 return (error);
11705 }
11706
11707 /* No need to redo state machine in loop */
11708 ixaflags &= ~IXAF_REACH_CONF;
11709
11710 /* Advance the offset to the second frag starting point. */
11711 offset += len;
11712 /*
11713 * Update hdr_len from the copied header - there might be less options
11714 * in the later fragments.
11715 */
11716 hdr_len = IPH_HDR_LENGTH(hdr_mp->b_rptr);
11717 /* Loop until done. */
11718 for (;;) {
11719 uint16_t offset_and_flags;
11720 uint16_t ip_len;
11721
11722 if (ip_data_end - offset > len) {
11723 /*
11724 * Carve off the appropriate amount from the original
11725 * datagram.
11726 */
11727 if (!(carve_mp = ip_carve_mp(&mp_orig, len))) {
11728 mp = NULL;
11729 break;
11730 }
11731 /*
11732 * More frags after this one. Get another copy
11733 * of the header.
11734 */
11735 if (carve_mp->b_datap->db_ref == 1 &&
11736 hdr_mp->b_wptr - hdr_mp->b_rptr <
11737 carve_mp->b_rptr - carve_mp->b_datap->db_base) {
11738 /* Inline IP header */
11739 carve_mp->b_rptr -= hdr_mp->b_wptr -
11740 hdr_mp->b_rptr;
11741 bcopy(hdr_mp->b_rptr, carve_mp->b_rptr,
11742 hdr_mp->b_wptr - hdr_mp->b_rptr);
11743 mp = carve_mp;
11744 } else {
11745 if (!(mp = copyb(hdr_mp))) {
11746 freemsg(carve_mp);
11747 break;
11748 }
11749 /* Get priority marking, if any. */
11750 mp->b_band = priority;
11751 mp->b_cont = carve_mp;
11752 }
11753 ipha = (ipha_t *)mp->b_rptr;
11754 offset_and_flags = IPH_MF;
11755 } else {
11756 /*
11757 * Last frag. Consume the header. Set len to
11758 * the length of this last piece.
11759 */
11760 len = ip_data_end - offset;
11761
11762 /*
11763 * Carve off the appropriate amount from the original
11764 * datagram.
11765 */
11766 if (!(carve_mp = ip_carve_mp(&mp_orig, len))) {
11767 mp = NULL;
11768 break;
11769 }
11770 if (carve_mp->b_datap->db_ref == 1 &&
11771 hdr_mp->b_wptr - hdr_mp->b_rptr <
11772 carve_mp->b_rptr - carve_mp->b_datap->db_base) {
11773 /* Inline IP header */
11774 carve_mp->b_rptr -= hdr_mp->b_wptr -
11775 hdr_mp->b_rptr;
11776 bcopy(hdr_mp->b_rptr, carve_mp->b_rptr,
11777 hdr_mp->b_wptr - hdr_mp->b_rptr);
11778 mp = carve_mp;
11779 freeb(hdr_mp);
11780 hdr_mp = mp;
11781 } else {
11782 mp = hdr_mp;
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 /* A frag of a frag might have IPH_MF non-zero */
11789 offset_and_flags =
11790 ntohs(ipha->ipha_fragment_offset_and_flags) &
11791 IPH_MF;
11792 }
11793 offset_and_flags |= (uint16_t)(offset >> 3);
11794 offset_and_flags |= (uint16_t)frag_flag;
11795 /* Store the offset and flags in the IP header. */
11796 ipha->ipha_fragment_offset_and_flags = htons(offset_and_flags);
11797
11798 /* Store the length in the IP header. */
11799 ip_len = (uint16_t)(len + hdr_len);
11800 ipha->ipha_length = htons(ip_len);
11801
11802 /*
11803 * Set the IP header checksum. Note that mp is just
11804 * the header, so this is easy to pass to ip_csum.
11805 */
11806 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
11807
11808 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragCreates);
11809
11810 error = postfragfn(mp, nce, ixaflags, ip_len, xmit_hint, szone,
11811 nolzid, ixa_cookie);
11812 /* All done if we just consumed the hdr_mp. */
11813 if (mp == hdr_mp) {
11814 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragOKs);
11815 return (error);
11816 }
11817 if (error != 0 && error != EWOULDBLOCK) {
11818 DTRACE_PROBE2(ip__xmit__frag__fail, ill_t *, ill,
11819 mblk_t *, hdr_mp);
11820 /* No point in sending the other fragments */
11821 break;
11822 }
11823
11824 /* Otherwise, advance and loop. */
11825 offset += len;
11826 }
11827 /* Clean up following allocation failure. */
11828 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11829 ip_drop_output("FragFails: loop ended", NULL, ill);
11830 if (mp != hdr_mp)
11831 freeb(hdr_mp);
11832 if (mp != mp_orig)
11833 freemsg(mp_orig);
11834 return (error);
11835 }
11836
11837 /*
11838 * Copy the header plus those options which have the copy bit set
11839 */
11840 static mblk_t *
11841 ip_fragment_copyhdr(uchar_t *rptr, int hdr_len, int offset, ip_stack_t *ipst,
11842 mblk_t *src)
11843 {
11844 mblk_t *mp;
11845 uchar_t *up;
11846
11847 /*
11848 * Quick check if we need to look for options without the copy bit
11849 * set
11850 */
11851 mp = allocb_tmpl(ipst->ips_ip_wroff_extra + hdr_len, src);
11852 if (!mp)
11853 return (mp);
11854 mp->b_rptr += ipst->ips_ip_wroff_extra;
11855 if (hdr_len == IP_SIMPLE_HDR_LENGTH || offset != 0) {
11856 bcopy(rptr, mp->b_rptr, hdr_len);
11857 mp->b_wptr += hdr_len + ipst->ips_ip_wroff_extra;
11858 return (mp);
11859 }
11860 up = mp->b_rptr;
11861 bcopy(rptr, up, IP_SIMPLE_HDR_LENGTH);
11862 up += IP_SIMPLE_HDR_LENGTH;
11863 rptr += IP_SIMPLE_HDR_LENGTH;
11864 hdr_len -= IP_SIMPLE_HDR_LENGTH;
11865 while (hdr_len > 0) {
11866 uint32_t optval;
11867 uint32_t optlen;
11868
11869 optval = *rptr;
11870 if (optval == IPOPT_EOL)
11871 break;
11872 if (optval == IPOPT_NOP)
11873 optlen = 1;
11874 else
11875 optlen = rptr[1];
11876 if (optval & IPOPT_COPY) {
11877 bcopy(rptr, up, optlen);
11878 up += optlen;
11879 }
11880 rptr += optlen;
11881 hdr_len -= optlen;
11882 }
11883 /*
11884 * Make sure that we drop an even number of words by filling
11885 * with EOL to the next word boundary.
11886 */
11887 for (hdr_len = up - (mp->b_rptr + IP_SIMPLE_HDR_LENGTH);
11888 hdr_len & 0x3; hdr_len++)
11889 *up++ = IPOPT_EOL;
11890 mp->b_wptr = up;
11891 /* Update header length */
11892 mp->b_rptr[0] = (uint8_t)((IP_VERSION << 4) | ((up - mp->b_rptr) >> 2));
11893 return (mp);
11894 }
11895
11896 /*
11897 * Update any source route, record route, or timestamp options when
11898 * sending a packet back to ourselves.
11899 * Check that we are at end of strict source route.
11900 * The options have been sanity checked by ip_output_options().
11901 */
11902 void
11903 ip_output_local_options(ipha_t *ipha, ip_stack_t *ipst)
11904 {
11905 ipoptp_t opts;
11906 uchar_t *opt;
11907 uint8_t optval;
11908 uint8_t optlen;
11909 ipaddr_t dst;
11910 uint32_t ts;
11911 timestruc_t now;
11912
11913 for (optval = ipoptp_first(&opts, ipha);
11914 optval != IPOPT_EOL;
11915 optval = ipoptp_next(&opts)) {
11916 opt = opts.ipoptp_cur;
11917 optlen = opts.ipoptp_len;
11918 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11919 switch (optval) {
11920 uint32_t off;
11921 case IPOPT_SSRR:
11922 case IPOPT_LSRR:
11923 off = opt[IPOPT_OFFSET];
11924 off--;
11925 if (optlen < IP_ADDR_LEN ||
11926 off > optlen - IP_ADDR_LEN) {
11927 /* End of source route */
11928 break;
11929 }
11930 /*
11931 * This will only happen if two consecutive entries
11932 * in the source route contains our address or if
11933 * it is a packet with a loose source route which
11934 * reaches us before consuming the whole source route
11935 */
11936
11937 if (optval == IPOPT_SSRR) {
11938 return;
11939 }
11940 /*
11941 * Hack: instead of dropping the packet truncate the
11942 * source route to what has been used by filling the
11943 * rest with IPOPT_NOP.
11944 */
11945 opt[IPOPT_OLEN] = (uint8_t)off;
11946 while (off < optlen) {
11947 opt[off++] = IPOPT_NOP;
11948 }
11949 break;
11950 case IPOPT_RR:
11951 off = opt[IPOPT_OFFSET];
11952 off--;
11953 if (optlen < IP_ADDR_LEN ||
11954 off > optlen - IP_ADDR_LEN) {
11955 /* No more room - ignore */
11956 ip1dbg((
11957 "ip_output_local_options: end of RR\n"));
11958 break;
11959 }
11960 dst = htonl(INADDR_LOOPBACK);
11961 bcopy(&dst, (char *)opt + off, IP_ADDR_LEN);
11962 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
11963 break;
11964 case IPOPT_TS:
11965 /* Insert timestamp if there is romm */
11966 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
11967 case IPOPT_TS_TSONLY:
11968 off = IPOPT_TS_TIMELEN;
11969 break;
11970 case IPOPT_TS_PRESPEC:
11971 case IPOPT_TS_PRESPEC_RFC791:
11972 /* Verify that the address matched */
11973 off = opt[IPOPT_OFFSET] - 1;
11974 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
11975 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
11976 /* Not for us */
11977 break;
11978 }
11979 /* FALLTHRU */
11980 case IPOPT_TS_TSANDADDR:
11981 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
11982 break;
11983 default:
11984 /*
11985 * ip_*put_options should have already
11986 * dropped this packet.
11987 */
11988 cmn_err(CE_PANIC, "ip_output_local_options: "
11989 "unknown IT - bug in ip_output_options?\n");
11990 return; /* Keep "lint" happy */
11991 }
11992 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
11993 /* Increase overflow counter */
11994 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
11995 opt[IPOPT_POS_OV_FLG] = (uint8_t)
11996 (opt[IPOPT_POS_OV_FLG] & 0x0F) |
11997 (off << 4);
11998 break;
11999 }
12000 off = opt[IPOPT_OFFSET] - 1;
12001 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
12002 case IPOPT_TS_PRESPEC:
12003 case IPOPT_TS_PRESPEC_RFC791:
12004 case IPOPT_TS_TSANDADDR:
12005 dst = htonl(INADDR_LOOPBACK);
12006 bcopy(&dst, (char *)opt + off, IP_ADDR_LEN);
12007 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
12008 /* FALLTHRU */
12009 case IPOPT_TS_TSONLY:
12010 off = opt[IPOPT_OFFSET] - 1;
12011 /* Compute # of milliseconds since midnight */
12012 gethrestime(&now);
12013 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
12014 now.tv_nsec / (NANOSEC / MILLISEC);
12015 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
12016 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
12017 break;
12018 }
12019 break;
12020 }
12021 }
12022 }
12023
12024 /*
12025 * Prepend an M_DATA fastpath header, and if none present prepend a
12026 * DL_UNITDATA_REQ. Frees the mblk on failure.
12027 *
12028 * nce_dlur_mp and nce_fp_mp can not disappear once they have been set.
12029 * If there is a change to them, the nce will be deleted (condemned) and
12030 * a new nce_t will be created when packets are sent. Thus we need no locks
12031 * to access those fields.
12032 *
12033 * We preserve b_band to support IPQoS. If a DL_UNITDATA_REQ is prepended
12034 * we place b_band in dl_priority.dl_max.
12035 */
12036 static mblk_t *
12037 ip_xmit_attach_llhdr(mblk_t *mp, nce_t *nce)
12038 {
12039 uint_t hlen;
12040 mblk_t *mp1;
12041 uint_t priority;
12042 uchar_t *rptr;
12043
12044 rptr = mp->b_rptr;
12045
12046 ASSERT(DB_TYPE(mp) == M_DATA);
12047 priority = mp->b_band;
12048
12049 ASSERT(nce != NULL);
12050 if ((mp1 = nce->nce_fp_mp) != NULL) {
12051 hlen = MBLKL(mp1);
12052 /*
12053 * Check if we have enough room to prepend fastpath
12054 * header
12055 */
12056 if (hlen != 0 && (rptr - mp->b_datap->db_base) >= hlen) {
12057 rptr -= hlen;
12058 bcopy(mp1->b_rptr, rptr, hlen);
12059 /*
12060 * Set the b_rptr to the start of the link layer
12061 * header
12062 */
12063 mp->b_rptr = rptr;
12064 return (mp);
12065 }
12066 mp1 = copyb(mp1);
12067 if (mp1 == NULL) {
12068 ill_t *ill = nce->nce_ill;
12069
12070 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12071 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12072 freemsg(mp);
12073 return (NULL);
12074 }
12075 mp1->b_band = priority;
12076 mp1->b_cont = mp;
12077 DB_CKSUMSTART(mp1) = DB_CKSUMSTART(mp);
12078 DB_CKSUMSTUFF(mp1) = DB_CKSUMSTUFF(mp);
12079 DB_CKSUMEND(mp1) = DB_CKSUMEND(mp);
12080 DB_CKSUMFLAGS(mp1) = DB_CKSUMFLAGS(mp);
12081 DB_LSOMSS(mp1) = DB_LSOMSS(mp);
12082 DTRACE_PROBE1(ip__xmit__copyb, (mblk_t *), mp1);
12083 /*
12084 * XXX disable ICK_VALID and compute checksum
12085 * here; can happen if nce_fp_mp changes and
12086 * it can't be copied now due to insufficient
12087 * space. (unlikely, fp mp can change, but it
12088 * does not increase in length)
12089 */
12090 return (mp1);
12091 }
12092 mp1 = copyb(nce->nce_dlur_mp);
12093
12094 if (mp1 == NULL) {
12095 ill_t *ill = nce->nce_ill;
12096
12097 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12098 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12099 freemsg(mp);
12100 return (NULL);
12101 }
12102 mp1->b_cont = mp;
12103 if (priority != 0) {
12104 mp1->b_band = priority;
12105 ((dl_unitdata_req_t *)(mp1->b_rptr))->dl_priority.dl_max =
12106 priority;
12107 }
12108 return (mp1);
12109 #undef rptr
12110 }
12111
12112 /*
12113 * Finish the outbound IPsec processing. This function is called from
12114 * ipsec_out_process() if the IPsec packet was processed
12115 * synchronously, or from {ah,esp}_kcf_callback_outbound() if it was processed
12116 * asynchronously.
12117 *
12118 * This is common to IPv4 and IPv6.
12119 */
12120 int
12121 ip_output_post_ipsec(mblk_t *mp, ip_xmit_attr_t *ixa)
12122 {
12123 iaflags_t ixaflags = ixa->ixa_flags;
12124 uint_t pktlen;
12125
12126
12127 /* AH/ESP don't update ixa_pktlen when they modify the packet */
12128 if (ixaflags & IXAF_IS_IPV4) {
12129 ipha_t *ipha = (ipha_t *)mp->b_rptr;
12130
12131 ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
12132 pktlen = ntohs(ipha->ipha_length);
12133 } else {
12134 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
12135
12136 ASSERT(IPH_HDR_VERSION(mp->b_rptr) == IPV6_VERSION);
12137 pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
12138 }
12139
12140 /*
12141 * We release any hard reference on the SAs here to make
12142 * sure the SAs can be garbage collected. ipsr_sa has a soft reference
12143 * on the SAs.
12144 * If in the future we want the hard latching of the SAs in the
12145 * ip_xmit_attr_t then we should remove this.
12146 */
12147 if (ixa->ixa_ipsec_esp_sa != NULL) {
12148 IPSA_REFRELE(ixa->ixa_ipsec_esp_sa);
12149 ixa->ixa_ipsec_esp_sa = NULL;
12150 }
12151 if (ixa->ixa_ipsec_ah_sa != NULL) {
12152 IPSA_REFRELE(ixa->ixa_ipsec_ah_sa);
12153 ixa->ixa_ipsec_ah_sa = NULL;
12154 }
12155
12156 /* Do we need to fragment? */
12157 if ((ixa->ixa_flags & IXAF_IPV6_ADD_FRAGHDR) ||
12158 pktlen > ixa->ixa_fragsize) {
12159 if (ixaflags & IXAF_IS_IPV4) {
12160 ASSERT(!(ixa->ixa_flags & IXAF_IPV6_ADD_FRAGHDR));
12161 /*
12162 * We check for the DF case in ipsec_out_process
12163 * hence this only handles the non-DF case.
12164 */
12165 return (ip_fragment_v4(mp, ixa->ixa_nce, ixa->ixa_flags,
12166 pktlen, ixa->ixa_fragsize,
12167 ixa->ixa_xmit_hint, ixa->ixa_zoneid,
12168 ixa->ixa_no_loop_zoneid, ixa->ixa_postfragfn,
12169 &ixa->ixa_cookie));
12170 } else {
12171 mp = ip_fraghdr_add_v6(mp, ixa->ixa_ident, ixa);
12172 if (mp == NULL) {
12173 /* MIB and ip_drop_output already done */
12174 return (ENOMEM);
12175 }
12176 pktlen += sizeof (ip6_frag_t);
12177 if (pktlen > ixa->ixa_fragsize) {
12178 return (ip_fragment_v6(mp, ixa->ixa_nce,
12179 ixa->ixa_flags, pktlen,
12180 ixa->ixa_fragsize, ixa->ixa_xmit_hint,
12181 ixa->ixa_zoneid, ixa->ixa_no_loop_zoneid,
12182 ixa->ixa_postfragfn, &ixa->ixa_cookie));
12183 }
12184 }
12185 }
12186 return ((ixa->ixa_postfragfn)(mp, ixa->ixa_nce, ixa->ixa_flags,
12187 pktlen, ixa->ixa_xmit_hint, ixa->ixa_zoneid,
12188 ixa->ixa_no_loop_zoneid, NULL));
12189 }
12190
12191 /*
12192 * Finish the inbound IPsec processing. This function is called from
12193 * ipsec_out_process() if the IPsec packet was processed
12194 * synchronously, or from {ah,esp}_kcf_callback_outbound() if it was processed
12195 * asynchronously.
12196 *
12197 * This is common to IPv4 and IPv6.
12198 */
12199 void
12200 ip_input_post_ipsec(mblk_t *mp, ip_recv_attr_t *ira)
12201 {
12202 iaflags_t iraflags = ira->ira_flags;
12203
12204 /* Length might have changed */
12205 if (iraflags & IRAF_IS_IPV4) {
12206 ipha_t *ipha = (ipha_t *)mp->b_rptr;
12207
12208 ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
12209 ira->ira_pktlen = ntohs(ipha->ipha_length);
12210 ira->ira_ip_hdr_length = IPH_HDR_LENGTH(ipha);
12211 ira->ira_protocol = ipha->ipha_protocol;
12212
12213 ip_fanout_v4(mp, ipha, ira);
12214 } else {
12215 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
12216 uint8_t *nexthdrp;
12217
12218 ASSERT(IPH_HDR_VERSION(mp->b_rptr) == IPV6_VERSION);
12219 ira->ira_pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
12220 if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &ira->ira_ip_hdr_length,
12221 &nexthdrp)) {
12222 /* Malformed packet */
12223 BUMP_MIB(ira->ira_ill->ill_ip_mib, ipIfStatsInDiscards);
12224 ip_drop_input("ipIfStatsInDiscards", mp, ira->ira_ill);
12225 freemsg(mp);
12226 return;
12227 }
12228 ira->ira_protocol = *nexthdrp;
12229 ip_fanout_v6(mp, ip6h, ira);
12230 }
12231 }
12232
12233 /*
12234 * Select which AH & ESP SA's to use (if any) for the outbound packet.
12235 *
12236 * If this function returns B_TRUE, the requested SA's have been filled
12237 * into the ixa_ipsec_*_sa pointers.
12238 *
12239 * If the function returns B_FALSE, the packet has been "consumed", most
12240 * likely by an ACQUIRE sent up via PF_KEY to a key management daemon.
12241 *
12242 * The SA references created by the protocol-specific "select"
12243 * function will be released in ip_output_post_ipsec.
12244 */
12245 static boolean_t
12246 ipsec_out_select_sa(mblk_t *mp, ip_xmit_attr_t *ixa)
12247 {
12248 boolean_t need_ah_acquire = B_FALSE, need_esp_acquire = B_FALSE;
12249 ipsec_policy_t *pp;
12250 ipsec_action_t *ap;
12251
12252 ASSERT(ixa->ixa_flags & IXAF_IPSEC_SECURE);
12253 ASSERT((ixa->ixa_ipsec_policy != NULL) ||
12254 (ixa->ixa_ipsec_action != NULL));
12255
12256 ap = ixa->ixa_ipsec_action;
12257 if (ap == NULL) {
12258 pp = ixa->ixa_ipsec_policy;
12259 ASSERT(pp != NULL);
12260 ap = pp->ipsp_act;
12261 ASSERT(ap != NULL);
12262 }
12263
12264 /*
12265 * We have an action. now, let's select SA's.
12266 * A side effect of setting ixa_ipsec_*_sa is that it will
12267 * be cached in the conn_t.
12268 */
12269 if (ap->ipa_want_esp) {
12270 if (ixa->ixa_ipsec_esp_sa == NULL) {
12271 need_esp_acquire = !ipsec_outbound_sa(mp, ixa,
12272 IPPROTO_ESP);
12273 }
12274 ASSERT(need_esp_acquire || ixa->ixa_ipsec_esp_sa != NULL);
12275 }
12276
12277 if (ap->ipa_want_ah) {
12278 if (ixa->ixa_ipsec_ah_sa == NULL) {
12279 need_ah_acquire = !ipsec_outbound_sa(mp, ixa,
12280 IPPROTO_AH);
12281 }
12282 ASSERT(need_ah_acquire || ixa->ixa_ipsec_ah_sa != NULL);
12283 /*
12284 * The ESP and AH processing order needs to be preserved
12285 * when both protocols are required (ESP should be applied
12286 * before AH for an outbound packet). Force an ESP ACQUIRE
12287 * when both ESP and AH are required, and an AH ACQUIRE
12288 * is needed.
12289 */
12290 if (ap->ipa_want_esp && need_ah_acquire)
12291 need_esp_acquire = B_TRUE;
12292 }
12293
12294 /*
12295 * Send an ACQUIRE (extended, regular, or both) if we need one.
12296 * Release SAs that got referenced, but will not be used until we
12297 * acquire _all_ of the SAs we need.
12298 */
12299 if (need_ah_acquire || need_esp_acquire) {
12300 if (ixa->ixa_ipsec_ah_sa != NULL) {
12301 IPSA_REFRELE(ixa->ixa_ipsec_ah_sa);
12302 ixa->ixa_ipsec_ah_sa = NULL;
12303 }
12304 if (ixa->ixa_ipsec_esp_sa != NULL) {
12305 IPSA_REFRELE(ixa->ixa_ipsec_esp_sa);
12306 ixa->ixa_ipsec_esp_sa = NULL;
12307 }
12308
12309 sadb_acquire(mp, ixa, need_ah_acquire, need_esp_acquire);
12310 return (B_FALSE);
12311 }
12312
12313 return (B_TRUE);
12314 }
12315
12316 /*
12317 * Handle IPsec output processing.
12318 * This function is only entered once for a given packet.
12319 * We try to do things synchronously, but if we need to have user-level
12320 * set up SAs, or ESP or AH uses asynchronous kEF, then the operation
12321 * will be completed
12322 * - when the SAs are added in esp_add_sa_finish/ah_add_sa_finish
12323 * - when asynchronous ESP is done it will do AH
12324 *
12325 * In all cases we come back in ip_output_post_ipsec() to fragment and
12326 * send out the packet.
12327 */
12328 int
12329 ipsec_out_process(mblk_t *mp, ip_xmit_attr_t *ixa)
12330 {
12331 ill_t *ill = ixa->ixa_nce->nce_ill;
12332 ip_stack_t *ipst = ixa->ixa_ipst;
12333 ipsec_stack_t *ipss;
12334 ipsec_policy_t *pp;
12335 ipsec_action_t *ap;
12336
12337 ASSERT(ixa->ixa_flags & IXAF_IPSEC_SECURE);
12338
12339 ASSERT((ixa->ixa_ipsec_policy != NULL) ||
12340 (ixa->ixa_ipsec_action != NULL));
12341
12342 ipss = ipst->ips_netstack->netstack_ipsec;
12343 if (!ipsec_loaded(ipss)) {
12344 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12345 ip_drop_packet(mp, B_TRUE, ill,
12346 DROPPER(ipss, ipds_ip_ipsec_not_loaded),
12347 &ipss->ipsec_dropper);
12348 return (ENOTSUP);
12349 }
12350
12351 ap = ixa->ixa_ipsec_action;
12352 if (ap == NULL) {
12353 pp = ixa->ixa_ipsec_policy;
12354 ASSERT(pp != NULL);
12355 ap = pp->ipsp_act;
12356 ASSERT(ap != NULL);
12357 }
12358
12359 /* Handle explicit drop action and bypass. */
12360 switch (ap->ipa_act.ipa_type) {
12361 case IPSEC_ACT_DISCARD:
12362 case IPSEC_ACT_REJECT:
12363 ip_drop_packet(mp, B_FALSE, ill,
12364 DROPPER(ipss, ipds_spd_explicit), &ipss->ipsec_spd_dropper);
12365 return (EHOSTUNREACH); /* IPsec policy failure */
12366 case IPSEC_ACT_BYPASS:
12367 return (ip_output_post_ipsec(mp, ixa));
12368 }
12369
12370 /*
12371 * The order of processing is first insert a IP header if needed.
12372 * Then insert the ESP header and then the AH header.
12373 */
12374 if ((ixa->ixa_flags & IXAF_IS_IPV4) && ap->ipa_want_se) {
12375 /*
12376 * First get the outer IP header before sending
12377 * it to ESP.
12378 */
12379 ipha_t *oipha, *iipha;
12380 mblk_t *outer_mp, *inner_mp;
12381
12382 if ((outer_mp = allocb(sizeof (ipha_t), BPRI_HI)) == NULL) {
12383 (void) mi_strlog(ill->ill_rq, 0,
12384 SL_ERROR|SL_TRACE|SL_CONSOLE,
12385 "ipsec_out_process: "
12386 "Self-Encapsulation failed: Out of memory\n");
12387 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12388 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12389 freemsg(mp);
12390 return (ENOBUFS);
12391 }
12392 inner_mp = mp;
12393 ASSERT(inner_mp->b_datap->db_type == M_DATA);
12394 oipha = (ipha_t *)outer_mp->b_rptr;
12395 iipha = (ipha_t *)inner_mp->b_rptr;
12396 *oipha = *iipha;
12397 outer_mp->b_wptr += sizeof (ipha_t);
12398 oipha->ipha_length = htons(ntohs(iipha->ipha_length) +
12399 sizeof (ipha_t));
12400 oipha->ipha_protocol = IPPROTO_ENCAP;
12401 oipha->ipha_version_and_hdr_length =
12402 IP_SIMPLE_HDR_VERSION;
12403 oipha->ipha_hdr_checksum = 0;
12404 oipha->ipha_hdr_checksum = ip_csum_hdr(oipha);
12405 outer_mp->b_cont = inner_mp;
12406 mp = outer_mp;
12407
12408 ixa->ixa_flags |= IXAF_IPSEC_TUNNEL;
12409 }
12410
12411 /* If we need to wait for a SA then we can't return any errno */
12412 if (((ap->ipa_want_ah && (ixa->ixa_ipsec_ah_sa == NULL)) ||
12413 (ap->ipa_want_esp && (ixa->ixa_ipsec_esp_sa == NULL))) &&
12414 !ipsec_out_select_sa(mp, ixa))
12415 return (0);
12416
12417 /*
12418 * By now, we know what SA's to use. Toss over to ESP & AH
12419 * to do the heavy lifting.
12420 */
12421 if (ap->ipa_want_esp) {
12422 ASSERT(ixa->ixa_ipsec_esp_sa != NULL);
12423
12424 mp = ixa->ixa_ipsec_esp_sa->ipsa_output_func(mp, ixa);
12425 if (mp == NULL) {
12426 /*
12427 * Either it failed or is pending. In the former case
12428 * ipIfStatsInDiscards was increased.
12429 */
12430 return (0);
12431 }
12432 }
12433
12434 if (ap->ipa_want_ah) {
12435 ASSERT(ixa->ixa_ipsec_ah_sa != NULL);
12436
12437 mp = ixa->ixa_ipsec_ah_sa->ipsa_output_func(mp, ixa);
12438 if (mp == NULL) {
12439 /*
12440 * Either it failed or is pending. In the former case
12441 * ipIfStatsInDiscards was increased.
12442 */
12443 return (0);
12444 }
12445 }
12446 /*
12447 * We are done with IPsec processing. Send it over
12448 * the wire.
12449 */
12450 return (ip_output_post_ipsec(mp, ixa));
12451 }
12452
12453 /*
12454 * ioctls that go through a down/up sequence may need to wait for the down
12455 * to complete. This involves waiting for the ire and ipif refcnts to go down
12456 * to zero. Subsequently the ioctl is restarted from ipif_ill_refrele_tail.
12457 */
12458 /* ARGSUSED */
12459 void
12460 ip_reprocess_ioctl(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
12461 {
12462 struct iocblk *iocp;
12463 mblk_t *mp1;
12464 ip_ioctl_cmd_t *ipip;
12465 int err;
12466 sin_t *sin;
12467 struct lifreq *lifr;
12468 struct ifreq *ifr;
12469
12470 iocp = (struct iocblk *)mp->b_rptr;
12471 ASSERT(ipsq != NULL);
12472 /* Existence of mp1 verified in ip_wput_nondata */
12473 mp1 = mp->b_cont->b_cont;
12474 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12475 if (ipip->ipi_cmd == SIOCSLIFNAME || ipip->ipi_cmd == IF_UNITSEL) {
12476 /*
12477 * Special case where ipx_current_ipif is not set:
12478 * ill_phyint_reinit merged the v4 and v6 into a single ipsq.
12479 * We are here as were not able to complete the operation in
12480 * ipif_set_values because we could not become exclusive on
12481 * the new ipsq.
12482 */
12483 ill_t *ill = q->q_ptr;
12484 ipsq_current_start(ipsq, ill->ill_ipif, ipip->ipi_cmd);
12485 }
12486 ASSERT(ipsq->ipsq_xop->ipx_current_ipif != NULL);
12487
12488 if (ipip->ipi_cmd_type == IF_CMD) {
12489 /* This a old style SIOC[GS]IF* command */
12490 ifr = (struct ifreq *)mp1->b_rptr;
12491 sin = (sin_t *)&ifr->ifr_addr;
12492 } else if (ipip->ipi_cmd_type == LIF_CMD) {
12493 /* This a new style SIOC[GS]LIF* command */
12494 lifr = (struct lifreq *)mp1->b_rptr;
12495 sin = (sin_t *)&lifr->lifr_addr;
12496 } else {
12497 sin = NULL;
12498 }
12499
12500 err = (*ipip->ipi_func_restart)(ipsq->ipsq_xop->ipx_current_ipif, sin,
12501 q, mp, ipip, mp1->b_rptr);
12502
12503 DTRACE_PROBE4(ipif__ioctl, char *, "ip_reprocess_ioctl finish",
12504 int, ipip->ipi_cmd,
12505 ill_t *, ipsq->ipsq_xop->ipx_current_ipif->ipif_ill,
12506 ipif_t *, ipsq->ipsq_xop->ipx_current_ipif);
12507
12508 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), ipsq);
12509 }
12510
12511 /*
12512 * ioctl processing
12513 *
12514 * ioctl processing starts with ip_sioctl_copyin_setup(), which looks up
12515 * the ioctl command in the ioctl tables, determines the copyin data size
12516 * from the ipi_copyin_size field, and does an mi_copyin() of that size.
12517 *
12518 * ioctl processing then continues when the M_IOCDATA makes its way down to
12519 * ip_wput_nondata(). The ioctl is looked up again in the ioctl table, its
12520 * associated 'conn' is refheld till the end of the ioctl and the general
12521 * ioctl processing function ip_process_ioctl() is called to extract the
12522 * arguments and process the ioctl. To simplify extraction, ioctl commands
12523 * are "typed" based on the arguments they take (e.g., LIF_CMD which takes a
12524 * `struct lifreq'), and a common extract function (e.g., ip_extract_lifreq())
12525 * is used to extract the ioctl's arguments.
12526 *
12527 * ip_process_ioctl determines if the ioctl needs to be serialized, and if
12528 * so goes thru the serialization primitive ipsq_try_enter. Then the
12529 * appropriate function to handle the ioctl is called based on the entry in
12530 * the ioctl table. ioctl completion is encapsulated in ip_ioctl_finish
12531 * which also refreleases the 'conn' that was refheld at the start of the
12532 * ioctl. Finally ipsq_exit is called if needed to exit the ipsq.
12533 *
12534 * Many exclusive ioctls go thru an internal down up sequence as part of
12535 * the operation. For example an attempt to change the IP address of an
12536 * ipif entails ipif_down, set address, ipif_up. Bringing down the interface
12537 * does all the cleanup such as deleting all ires that use this address.
12538 * Then we need to wait till all references to the interface go away.
12539 */
12540 void
12541 ip_process_ioctl(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *arg)
12542 {
12543 struct iocblk *iocp = (struct iocblk *)mp->b_rptr;
12544 ip_ioctl_cmd_t *ipip = arg;
12545 ip_extract_func_t *extract_funcp;
12546 cmd_info_t ci;
12547 int err;
12548 boolean_t entered_ipsq = B_FALSE;
12549
12550 ip3dbg(("ip_process_ioctl: ioctl %X\n", iocp->ioc_cmd));
12551
12552 if (ipip == NULL)
12553 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12554
12555 /*
12556 * SIOCLIFADDIF needs to go thru a special path since the
12557 * ill may not exist yet. This happens in the case of lo0
12558 * which is created using this ioctl.
12559 */
12560 if (ipip->ipi_cmd == SIOCLIFADDIF) {
12561 err = ip_sioctl_addif(NULL, NULL, q, mp, NULL, NULL);
12562 DTRACE_PROBE4(ipif__ioctl, char *, "ip_process_ioctl finish",
12563 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12564 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12565 return;
12566 }
12567
12568 ci.ci_ipif = NULL;
12569 switch (ipip->ipi_cmd_type) {
12570 case MISC_CMD:
12571 case MSFILT_CMD:
12572 /*
12573 * All MISC_CMD ioctls come in here -- e.g. SIOCGLIFCONF.
12574 */
12575 if (ipip->ipi_cmd == IF_UNITSEL) {
12576 /* ioctl comes down the ill */
12577 ci.ci_ipif = ((ill_t *)q->q_ptr)->ill_ipif;
12578 ipif_refhold(ci.ci_ipif);
12579 }
12580 err = 0;
12581 ci.ci_sin = NULL;
12582 ci.ci_sin6 = NULL;
12583 ci.ci_lifr = NULL;
12584 extract_funcp = NULL;
12585 break;
12586
12587 case IF_CMD:
12588 case LIF_CMD:
12589 extract_funcp = ip_extract_lifreq;
12590 break;
12591
12592 case ARP_CMD:
12593 case XARP_CMD:
12594 extract_funcp = ip_extract_arpreq;
12595 break;
12596
12597 default:
12598 ASSERT(0);
12599 }
12600
12601 if (extract_funcp != NULL) {
12602 err = (*extract_funcp)(q, mp, ipip, &ci);
12603 if (err != 0) {
12604 DTRACE_PROBE4(ipif__ioctl,
12605 char *, "ip_process_ioctl finish err",
12606 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12607 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12608 return;
12609 }
12610
12611 /*
12612 * All of the extraction functions return a refheld ipif.
12613 */
12614 ASSERT(ci.ci_ipif != NULL);
12615 }
12616
12617 if (!(ipip->ipi_flags & IPI_WR)) {
12618 /*
12619 * A return value of EINPROGRESS means the ioctl is
12620 * either queued and waiting for some reason or has
12621 * already completed.
12622 */
12623 err = (*ipip->ipi_func)(ci.ci_ipif, ci.ci_sin, q, mp, ipip,
12624 ci.ci_lifr);
12625 if (ci.ci_ipif != NULL) {
12626 DTRACE_PROBE4(ipif__ioctl,
12627 char *, "ip_process_ioctl finish RD",
12628 int, ipip->ipi_cmd, ill_t *, ci.ci_ipif->ipif_ill,
12629 ipif_t *, ci.ci_ipif);
12630 ipif_refrele(ci.ci_ipif);
12631 } else {
12632 DTRACE_PROBE4(ipif__ioctl,
12633 char *, "ip_process_ioctl finish RD",
12634 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12635 }
12636 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12637 return;
12638 }
12639
12640 ASSERT(ci.ci_ipif != NULL);
12641
12642 /*
12643 * If ipsq is non-NULL, we are already being called exclusively
12644 */
12645 ASSERT(ipsq == NULL || IAM_WRITER_IPSQ(ipsq));
12646 if (ipsq == NULL) {
12647 ipsq = ipsq_try_enter(ci.ci_ipif, NULL, q, mp, ip_process_ioctl,
12648 NEW_OP, B_TRUE);
12649 if (ipsq == NULL) {
12650 ipif_refrele(ci.ci_ipif);
12651 return;
12652 }
12653 entered_ipsq = B_TRUE;
12654 }
12655 /*
12656 * Release the ipif so that ipif_down and friends that wait for
12657 * references to go away are not misled about the current ipif_refcnt
12658 * values. We are writer so we can access the ipif even after releasing
12659 * the ipif.
12660 */
12661 ipif_refrele(ci.ci_ipif);
12662
12663 ipsq_current_start(ipsq, ci.ci_ipif, ipip->ipi_cmd);
12664
12665 /*
12666 * A return value of EINPROGRESS means the ioctl is
12667 * either queued and waiting for some reason or has
12668 * already completed.
12669 */
12670 err = (*ipip->ipi_func)(ci.ci_ipif, ci.ci_sin, q, mp, ipip, ci.ci_lifr);
12671
12672 DTRACE_PROBE4(ipif__ioctl, char *, "ip_process_ioctl finish WR",
12673 int, ipip->ipi_cmd,
12674 ill_t *, ci.ci_ipif == NULL ? NULL : ci.ci_ipif->ipif_ill,
12675 ipif_t *, ci.ci_ipif);
12676 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), ipsq);
12677
12678 if (entered_ipsq)
12679 ipsq_exit(ipsq);
12680 }
12681
12682 /*
12683 * Complete the ioctl. Typically ioctls use the mi package and need to
12684 * do mi_copyout/mi_copy_done.
12685 */
12686 void
12687 ip_ioctl_finish(queue_t *q, mblk_t *mp, int err, int mode, ipsq_t *ipsq)
12688 {
12689 conn_t *connp = NULL;
12690
12691 if (err == EINPROGRESS)
12692 return;
12693
12694 if (CONN_Q(q)) {
12695 connp = Q_TO_CONN(q);
12696 ASSERT(connp->conn_ref >= 2);
12697 }
12698
12699 switch (mode) {
12700 case COPYOUT:
12701 if (err == 0)
12702 mi_copyout(q, mp);
12703 else
12704 mi_copy_done(q, mp, err);
12705 break;
12706
12707 case NO_COPYOUT:
12708 mi_copy_done(q, mp, err);
12709 break;
12710
12711 default:
12712 ASSERT(mode == CONN_CLOSE); /* aborted through CONN_CLOSE */
12713 break;
12714 }
12715
12716 /*
12717 * The conn refhold and ioctlref placed on the conn at the start of the
12718 * ioctl are released here.
12719 */
12720 if (connp != NULL) {
12721 CONN_DEC_IOCTLREF(connp);
12722 CONN_OPER_PENDING_DONE(connp);
12723 }
12724
12725 if (ipsq != NULL)
12726 ipsq_current_finish(ipsq);
12727 }
12728
12729 /* Handles all non data messages */
12730 void
12731 ip_wput_nondata(queue_t *q, mblk_t *mp)
12732 {
12733 mblk_t *mp1;
12734 struct iocblk *iocp;
12735 ip_ioctl_cmd_t *ipip;
12736 conn_t *connp;
12737 cred_t *cr;
12738 char *proto_str;
12739
12740 if (CONN_Q(q))
12741 connp = Q_TO_CONN(q);
12742 else
12743 connp = NULL;
12744
12745 switch (DB_TYPE(mp)) {
12746 case M_IOCTL:
12747 /*
12748 * IOCTL processing begins in ip_sioctl_copyin_setup which
12749 * will arrange to copy in associated control structures.
12750 */
12751 ip_sioctl_copyin_setup(q, mp);
12752 return;
12753 case M_IOCDATA:
12754 /*
12755 * Ensure that this is associated with one of our trans-
12756 * parent ioctls. If it's not ours, discard it if we're
12757 * running as a driver, or pass it on if we're a module.
12758 */
12759 iocp = (struct iocblk *)mp->b_rptr;
12760 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12761 if (ipip == NULL) {
12762 if (q->q_next == NULL) {
12763 goto nak;
12764 } else {
12765 putnext(q, mp);
12766 }
12767 return;
12768 }
12769 if ((q->q_next != NULL) && !(ipip->ipi_flags & IPI_MODOK)) {
12770 /*
12771 * The ioctl is one we recognise, but is not consumed
12772 * by IP as a module and we are a module, so we drop
12773 */
12774 goto nak;
12775 }
12776
12777 /* IOCTL continuation following copyin or copyout. */
12778 if (mi_copy_state(q, mp, NULL) == -1) {
12779 /*
12780 * The copy operation failed. mi_copy_state already
12781 * cleaned up, so we're out of here.
12782 */
12783 return;
12784 }
12785 /*
12786 * If we just completed a copy in, we become writer and
12787 * continue processing in ip_sioctl_copyin_done. If it
12788 * was a copy out, we call mi_copyout again. If there is
12789 * nothing more to copy out, it will complete the IOCTL.
12790 */
12791 if (MI_COPY_DIRECTION(mp) == MI_COPY_IN) {
12792 if (!(mp1 = mp->b_cont) || !(mp1 = mp1->b_cont)) {
12793 mi_copy_done(q, mp, EPROTO);
12794 return;
12795 }
12796 /*
12797 * Check for cases that need more copying. A return
12798 * value of 0 means a second copyin has been started,
12799 * so we return; a return value of 1 means no more
12800 * copying is needed, so we continue.
12801 */
12802 if (ipip->ipi_cmd_type == MSFILT_CMD &&
12803 MI_COPY_COUNT(mp) == 1) {
12804 if (ip_copyin_msfilter(q, mp) == 0)
12805 return;
12806 }
12807 /*
12808 * Refhold the conn, till the ioctl completes. This is
12809 * needed in case the ioctl ends up in the pending mp
12810 * list. Every mp in the ipx_pending_mp list must have
12811 * a refhold on the conn to resume processing. The
12812 * refhold is released when the ioctl completes
12813 * (whether normally or abnormally). An ioctlref is also
12814 * placed on the conn to prevent TCP from removing the
12815 * queue needed to send the ioctl reply back.
12816 * In all cases ip_ioctl_finish is called to finish
12817 * the ioctl and release the refholds.
12818 */
12819 if (connp != NULL) {
12820 /* This is not a reentry */
12821 CONN_INC_REF(connp);
12822 CONN_INC_IOCTLREF(connp);
12823 } else {
12824 if (!(ipip->ipi_flags & IPI_MODOK)) {
12825 mi_copy_done(q, mp, EINVAL);
12826 return;
12827 }
12828 }
12829
12830 ip_process_ioctl(NULL, q, mp, ipip);
12831
12832 } else {
12833 mi_copyout(q, mp);
12834 }
12835 return;
12836
12837 case M_IOCNAK:
12838 /*
12839 * The only way we could get here is if a resolver didn't like
12840 * an IOCTL we sent it. This shouldn't happen.
12841 */
12842 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
12843 "ip_wput_nondata: unexpected M_IOCNAK, ioc_cmd 0x%x",
12844 ((struct iocblk *)mp->b_rptr)->ioc_cmd);
12845 freemsg(mp);
12846 return;
12847 case M_IOCACK:
12848 /* /dev/ip shouldn't see this */
12849 goto nak;
12850 case M_FLUSH:
12851 if (*mp->b_rptr & FLUSHW)
12852 flushq(q, FLUSHALL);
12853 if (q->q_next) {
12854 putnext(q, mp);
12855 return;
12856 }
12857 if (*mp->b_rptr & FLUSHR) {
12858 *mp->b_rptr &= ~FLUSHW;
12859 qreply(q, mp);
12860 return;
12861 }
12862 freemsg(mp);
12863 return;
12864 case M_CTL:
12865 break;
12866 case M_PROTO:
12867 case M_PCPROTO:
12868 /*
12869 * The only PROTO messages we expect are SNMP-related.
12870 */
12871 switch (((union T_primitives *)mp->b_rptr)->type) {
12872 case T_SVR4_OPTMGMT_REQ:
12873 ip2dbg(("ip_wput_nondata: T_SVR4_OPTMGMT_REQ "
12874 "flags %x\n",
12875 ((struct T_optmgmt_req *)mp->b_rptr)->MGMT_flags));
12876
12877 if (connp == NULL) {
12878 proto_str = "T_SVR4_OPTMGMT_REQ";
12879 goto protonak;
12880 }
12881
12882 /*
12883 * All Solaris components should pass a db_credp
12884 * for this TPI message, hence we ASSERT.
12885 * But in case there is some other M_PROTO that looks
12886 * like a TPI message sent by some other kernel
12887 * component, we check and return an error.
12888 */
12889 cr = msg_getcred(mp, NULL);
12890 ASSERT(cr != NULL);
12891 if (cr == NULL) {
12892 mp = mi_tpi_err_ack_alloc(mp, TSYSERR, EINVAL);
12893 if (mp != NULL)
12894 qreply(q, mp);
12895 return;
12896 }
12897
12898 if (!snmpcom_req(q, mp, ip_snmp_set, ip_snmp_get, cr)) {
12899 proto_str = "Bad SNMPCOM request?";
12900 goto protonak;
12901 }
12902 return;
12903 default:
12904 ip1dbg(("ip_wput_nondata: dropping M_PROTO prim %u\n",
12905 (int)*(uint_t *)mp->b_rptr));
12906 freemsg(mp);
12907 return;
12908 }
12909 default:
12910 break;
12911 }
12912 if (q->q_next) {
12913 putnext(q, mp);
12914 } else
12915 freemsg(mp);
12916 return;
12917
12918 nak:
12919 iocp->ioc_error = EINVAL;
12920 mp->b_datap->db_type = M_IOCNAK;
12921 iocp->ioc_count = 0;
12922 qreply(q, mp);
12923 return;
12924
12925 protonak:
12926 cmn_err(CE_NOTE, "IP doesn't process %s as a module", proto_str);
12927 if ((mp = mi_tpi_err_ack_alloc(mp, TPROTO, EINVAL)) != NULL)
12928 qreply(q, mp);
12929 }
12930
12931 /*
12932 * Process IP options in an outbound packet. Verify that the nexthop in a
12933 * strict source route is onlink.
12934 * Returns non-zero if something fails in which case an ICMP error has been
12935 * sent and mp freed.
12936 *
12937 * Assumes the ULP has called ip_massage_options to move nexthop into ipha_dst.
12938 */
12939 int
12940 ip_output_options(mblk_t *mp, ipha_t *ipha, ip_xmit_attr_t *ixa, ill_t *ill)
12941 {
12942 ipoptp_t opts;
12943 uchar_t *opt;
12944 uint8_t optval;
12945 uint8_t optlen;
12946 ipaddr_t dst;
12947 intptr_t code = 0;
12948 ire_t *ire;
12949 ip_stack_t *ipst = ixa->ixa_ipst;
12950 ip_recv_attr_t iras;
12951
12952 ip2dbg(("ip_output_options\n"));
12953
12954 dst = ipha->ipha_dst;
12955 for (optval = ipoptp_first(&opts, ipha);
12956 optval != IPOPT_EOL;
12957 optval = ipoptp_next(&opts)) {
12958 opt = opts.ipoptp_cur;
12959 optlen = opts.ipoptp_len;
12960 ip2dbg(("ip_output_options: opt %d, len %d\n",
12961 optval, optlen));
12962 switch (optval) {
12963 uint32_t off;
12964 case IPOPT_SSRR:
12965 case IPOPT_LSRR:
12966 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
12967 ip1dbg((
12968 "ip_output_options: bad option offset\n"));
12969 code = (char *)&opt[IPOPT_OLEN] -
12970 (char *)ipha;
12971 goto param_prob;
12972 }
12973 off = opt[IPOPT_OFFSET];
12974 ip1dbg(("ip_output_options: next hop 0x%x\n",
12975 ntohl(dst)));
12976 /*
12977 * For strict: verify that dst is directly
12978 * reachable.
12979 */
12980 if (optval == IPOPT_SSRR) {
12981 ire = ire_ftable_lookup_v4(dst, 0, 0,
12982 IRE_INTERFACE, NULL, ALL_ZONES,
12983 ixa->ixa_tsl,
12984 MATCH_IRE_TYPE | MATCH_IRE_SECATTR, 0, ipst,
12985 NULL);
12986 if (ire == NULL) {
12987 ip1dbg(("ip_output_options: SSRR not"
12988 " directly reachable: 0x%x\n",
12989 ntohl(dst)));
12990 goto bad_src_route;
12991 }
12992 ire_refrele(ire);
12993 }
12994 break;
12995 case IPOPT_RR:
12996 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
12997 ip1dbg((
12998 "ip_output_options: bad option offset\n"));
12999 code = (char *)&opt[IPOPT_OLEN] -
13000 (char *)ipha;
13001 goto param_prob;
13002 }
13003 break;
13004 case IPOPT_TS:
13005 /*
13006 * Verify that length >=5 and that there is either
13007 * room for another timestamp or that the overflow
13008 * counter is not maxed out.
13009 */
13010 code = (char *)&opt[IPOPT_OLEN] - (char *)ipha;
13011 if (optlen < IPOPT_MINLEN_IT) {
13012 goto param_prob;
13013 }
13014 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
13015 ip1dbg((
13016 "ip_output_options: bad option offset\n"));
13017 code = (char *)&opt[IPOPT_OFFSET] -
13018 (char *)ipha;
13019 goto param_prob;
13020 }
13021 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
13022 case IPOPT_TS_TSONLY:
13023 off = IPOPT_TS_TIMELEN;
13024 break;
13025 case IPOPT_TS_TSANDADDR:
13026 case IPOPT_TS_PRESPEC:
13027 case IPOPT_TS_PRESPEC_RFC791:
13028 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
13029 break;
13030 default:
13031 code = (char *)&opt[IPOPT_POS_OV_FLG] -
13032 (char *)ipha;
13033 goto param_prob;
13034 }
13035 if (opt[IPOPT_OFFSET] - 1 + off > optlen &&
13036 (opt[IPOPT_POS_OV_FLG] & 0xF0) == 0xF0) {
13037 /*
13038 * No room and the overflow counter is 15
13039 * already.
13040 */
13041 goto param_prob;
13042 }
13043 break;
13044 }
13045 }
13046
13047 if ((opts.ipoptp_flags & IPOPTP_ERROR) == 0)
13048 return (0);
13049
13050 ip1dbg(("ip_output_options: error processing IP options."));
13051 code = (char *)&opt[IPOPT_OFFSET] - (char *)ipha;
13052
13053 param_prob:
13054 bzero(&iras, sizeof (iras));
13055 iras.ira_ill = iras.ira_rill = ill;
13056 iras.ira_ruifindex = ill->ill_phyint->phyint_ifindex;
13057 iras.ira_rifindex = iras.ira_ruifindex;
13058 iras.ira_flags = IRAF_IS_IPV4;
13059
13060 ip_drop_output("ip_output_options", mp, ill);
13061 icmp_param_problem(mp, (uint8_t)code, &iras);
13062 ASSERT(!(iras.ira_flags & IRAF_IPSEC_SECURE));
13063 return (-1);
13064
13065 bad_src_route:
13066 bzero(&iras, sizeof (iras));
13067 iras.ira_ill = iras.ira_rill = ill;
13068 iras.ira_ruifindex = ill->ill_phyint->phyint_ifindex;
13069 iras.ira_rifindex = iras.ira_ruifindex;
13070 iras.ira_flags = IRAF_IS_IPV4;
13071
13072 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ill);
13073 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, &iras);
13074 ASSERT(!(iras.ira_flags & IRAF_IPSEC_SECURE));
13075 return (-1);
13076 }
13077
13078 /*
13079 * The maximum value of conn_drain_list_cnt is CONN_MAXDRAINCNT.
13080 * conn_drain_list_cnt can be changed by setting conn_drain_nthreads
13081 * thru /etc/system.
13082 */
13083 #define CONN_MAXDRAINCNT 64
13084
13085 static void
13086 conn_drain_init(ip_stack_t *ipst)
13087 {
13088 int i, j;
13089 idl_tx_list_t *itl_tx;
13090
13091 ipst->ips_conn_drain_list_cnt = conn_drain_nthreads;
13092
13093 if ((ipst->ips_conn_drain_list_cnt == 0) ||
13094 (ipst->ips_conn_drain_list_cnt > CONN_MAXDRAINCNT)) {
13095 /*
13096 * Default value of the number of drainers is the
13097 * number of cpus, subject to maximum of 8 drainers.
13098 */
13099 if (boot_max_ncpus != -1)
13100 ipst->ips_conn_drain_list_cnt = MIN(boot_max_ncpus, 8);
13101 else
13102 ipst->ips_conn_drain_list_cnt = MIN(max_ncpus, 8);
13103 }
13104
13105 ipst->ips_idl_tx_list =
13106 kmem_zalloc(TX_FANOUT_SIZE * sizeof (idl_tx_list_t), KM_SLEEP);
13107 for (i = 0; i < TX_FANOUT_SIZE; i++) {
13108 itl_tx = &ipst->ips_idl_tx_list[i];
13109 itl_tx->txl_drain_list =
13110 kmem_zalloc(ipst->ips_conn_drain_list_cnt *
13111 sizeof (idl_t), KM_SLEEP);
13112 mutex_init(&itl_tx->txl_lock, NULL, MUTEX_DEFAULT, NULL);
13113 for (j = 0; j < ipst->ips_conn_drain_list_cnt; j++) {
13114 mutex_init(&itl_tx->txl_drain_list[j].idl_lock, NULL,
13115 MUTEX_DEFAULT, NULL);
13116 itl_tx->txl_drain_list[j].idl_itl = itl_tx;
13117 }
13118 }
13119 }
13120
13121 static void
13122 conn_drain_fini(ip_stack_t *ipst)
13123 {
13124 int i;
13125 idl_tx_list_t *itl_tx;
13126
13127 for (i = 0; i < TX_FANOUT_SIZE; i++) {
13128 itl_tx = &ipst->ips_idl_tx_list[i];
13129 kmem_free(itl_tx->txl_drain_list,
13130 ipst->ips_conn_drain_list_cnt * sizeof (idl_t));
13131 }
13132 kmem_free(ipst->ips_idl_tx_list,
13133 TX_FANOUT_SIZE * sizeof (idl_tx_list_t));
13134 ipst->ips_idl_tx_list = NULL;
13135 }
13136
13137 /*
13138 * Flow control has blocked us from proceeding. Insert the given conn in one
13139 * of the conn drain lists. When flow control is unblocked, either ip_wsrv()
13140 * (STREAMS) or ill_flow_enable() (direct) will be called back, which in turn
13141 * will call conn_walk_drain(). See the flow control notes at the top of this
13142 * file for more details.
13143 */
13144 void
13145 conn_drain_insert(conn_t *connp, idl_tx_list_t *tx_list)
13146 {
13147 idl_t *idl = tx_list->txl_drain_list;
13148 uint_t index;
13149 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
13150
13151 mutex_enter(&connp->conn_lock);
13152 if (connp->conn_state_flags & CONN_CLOSING) {
13153 /*
13154 * The conn is closing as a result of which CONN_CLOSING
13155 * is set. Return.
13156 */
13157 mutex_exit(&connp->conn_lock);
13158 return;
13159 } else if (connp->conn_idl == NULL) {
13160 /*
13161 * Assign the next drain list round robin. We dont' use
13162 * a lock, and thus it may not be strictly round robin.
13163 * Atomicity of load/stores is enough to make sure that
13164 * conn_drain_list_index is always within bounds.
13165 */
13166 index = tx_list->txl_drain_index;
13167 ASSERT(index < ipst->ips_conn_drain_list_cnt);
13168 connp->conn_idl = &tx_list->txl_drain_list[index];
13169 index++;
13170 if (index == ipst->ips_conn_drain_list_cnt)
13171 index = 0;
13172 tx_list->txl_drain_index = index;
13173 } else {
13174 ASSERT(connp->conn_idl->idl_itl == tx_list);
13175 }
13176 mutex_exit(&connp->conn_lock);
13177
13178 idl = connp->conn_idl;
13179 mutex_enter(&idl->idl_lock);
13180 if ((connp->conn_drain_prev != NULL) ||
13181 (connp->conn_state_flags & CONN_CLOSING)) {
13182 /*
13183 * The conn is either already in the drain list or closing.
13184 * (We needed to check for CONN_CLOSING again since close can
13185 * sneak in between dropping conn_lock and acquiring idl_lock.)
13186 */
13187 mutex_exit(&idl->idl_lock);
13188 return;
13189 }
13190
13191 /*
13192 * The conn is not in the drain list. Insert it at the
13193 * tail of the drain list. The drain list is circular
13194 * and doubly linked. idl_conn points to the 1st element
13195 * in the list.
13196 */
13197 if (idl->idl_conn == NULL) {
13198 idl->idl_conn = connp;
13199 connp->conn_drain_next = connp;
13200 connp->conn_drain_prev = connp;
13201 } else {
13202 conn_t *head = idl->idl_conn;
13203
13204 connp->conn_drain_next = head;
13205 connp->conn_drain_prev = head->conn_drain_prev;
13206 head->conn_drain_prev->conn_drain_next = connp;
13207 head->conn_drain_prev = connp;
13208 }
13209 /*
13210 * For non streams based sockets assert flow control.
13211 */
13212 conn_setqfull(connp, NULL);
13213 mutex_exit(&idl->idl_lock);
13214 }
13215
13216 static void
13217 conn_drain_remove(conn_t *connp)
13218 {
13219 idl_t *idl = connp->conn_idl;
13220
13221 if (idl != NULL) {
13222 /*
13223 * Remove ourself from the drain list.
13224 */
13225 if (connp->conn_drain_next == connp) {
13226 /* Singleton in the list */
13227 ASSERT(connp->conn_drain_prev == connp);
13228 idl->idl_conn = NULL;
13229 } else {
13230 connp->conn_drain_prev->conn_drain_next =
13231 connp->conn_drain_next;
13232 connp->conn_drain_next->conn_drain_prev =
13233 connp->conn_drain_prev;
13234 if (idl->idl_conn == connp)
13235 idl->idl_conn = connp->conn_drain_next;
13236 }
13237
13238 /*
13239 * NOTE: because conn_idl is associated with a specific drain
13240 * list which in turn is tied to the index the TX ring
13241 * (txl_cookie) hashes to, and because the TX ring can change
13242 * over the lifetime of the conn_t, we must clear conn_idl so
13243 * a subsequent conn_drain_insert() will set conn_idl again
13244 * based on the latest txl_cookie.
13245 */
13246 connp->conn_idl = NULL;
13247 }
13248 connp->conn_drain_next = NULL;
13249 connp->conn_drain_prev = NULL;
13250
13251 conn_clrqfull(connp, NULL);
13252 /*
13253 * For streams based sockets open up flow control.
13254 */
13255 if (!IPCL_IS_NONSTR(connp))
13256 enableok(connp->conn_wq);
13257 }
13258
13259 /*
13260 * This conn is closing, and we are called from ip_close. OR
13261 * this conn is draining because flow-control on the ill has been relieved.
13262 *
13263 * We must also need to remove conn's on this idl from the list, and also
13264 * inform the sockfs upcalls about the change in flow-control.
13265 */
13266 static void
13267 conn_drain(conn_t *connp, boolean_t closing)
13268 {
13269 idl_t *idl;
13270 conn_t *next_connp;
13271
13272 /*
13273 * connp->conn_idl is stable at this point, and no lock is needed
13274 * to check it. If we are called from ip_close, close has already
13275 * set CONN_CLOSING, thus freezing the value of conn_idl, and
13276 * called us only because conn_idl is non-null. If we are called thru
13277 * service, conn_idl could be null, but it cannot change because
13278 * service is single-threaded per queue, and there cannot be another
13279 * instance of service trying to call conn_drain_insert on this conn
13280 * now.
13281 */
13282 ASSERT(!closing || connp == NULL || connp->conn_idl != NULL);
13283
13284 /*
13285 * If the conn doesn't exist or is not on a drain list, bail.
13286 */
13287 if (connp == NULL || connp->conn_idl == NULL ||
13288 connp->conn_drain_prev == NULL) {
13289 return;
13290 }
13291
13292 idl = connp->conn_idl;
13293 ASSERT(MUTEX_HELD(&idl->idl_lock));
13294
13295 if (!closing) {
13296 next_connp = connp->conn_drain_next;
13297 while (next_connp != connp) {
13298 conn_t *delconnp = next_connp;
13299
13300 next_connp = next_connp->conn_drain_next;
13301 conn_drain_remove(delconnp);
13302 }
13303 ASSERT(connp->conn_drain_next == idl->idl_conn);
13304 }
13305 conn_drain_remove(connp);
13306 }
13307
13308 /*
13309 * Write service routine. Shared perimeter entry point.
13310 * The device queue's messages has fallen below the low water mark and STREAMS
13311 * has backenabled the ill_wq. Send sockfs notification about flow-control on
13312 * each waiting conn.
13313 */
13314 void
13315 ip_wsrv(queue_t *q)
13316 {
13317 ill_t *ill;
13318
13319 ill = (ill_t *)q->q_ptr;
13320 if (ill->ill_state_flags == 0) {
13321 ip_stack_t *ipst = ill->ill_ipst;
13322
13323 /*
13324 * The device flow control has opened up.
13325 * Walk through conn drain lists and qenable the
13326 * first conn in each list. This makes sense only
13327 * if the stream is fully plumbed and setup.
13328 * Hence the ill_state_flags check above.
13329 */
13330 ip1dbg(("ip_wsrv: walking\n"));
13331 conn_walk_drain(ipst, &ipst->ips_idl_tx_list[0]);
13332 enableok(ill->ill_wq);
13333 }
13334 }
13335
13336 /*
13337 * Callback to disable flow control in IP.
13338 *
13339 * This is a mac client callback added when the DLD_CAPAB_DIRECT capability
13340 * is enabled.
13341 *
13342 * When MAC_TX() is not able to send any more packets, dld sets its queue
13343 * to QFULL and enable the STREAMS flow control. Later, when the underlying
13344 * driver is able to continue to send packets, it calls mac_tx_(ring_)update()
13345 * function and wakes up corresponding mac worker threads, which in turn
13346 * calls this callback function, and disables flow control.
13347 */
13348 void
13349 ill_flow_enable(void *arg, ip_mac_tx_cookie_t cookie)
13350 {
13351 ill_t *ill = (ill_t *)arg;
13352 ip_stack_t *ipst = ill->ill_ipst;
13353 idl_tx_list_t *idl_txl;
13354
13355 idl_txl = &ipst->ips_idl_tx_list[IDLHASHINDEX(cookie)];
13356 mutex_enter(&idl_txl->txl_lock);
13357 /* add code to to set a flag to indicate idl_txl is enabled */
13358 conn_walk_drain(ipst, idl_txl);
13359 mutex_exit(&idl_txl->txl_lock);
13360 }
13361
13362 /*
13363 * Flow control has been relieved and STREAMS has backenabled us; drain
13364 * all the conn lists on `tx_list'.
13365 */
13366 static void
13367 conn_walk_drain(ip_stack_t *ipst, idl_tx_list_t *tx_list)
13368 {
13369 int i;
13370 idl_t *idl;
13371
13372 IP_STAT(ipst, ip_conn_walk_drain);
13373
13374 for (i = 0; i < ipst->ips_conn_drain_list_cnt; i++) {
13375 idl = &tx_list->txl_drain_list[i];
13376 mutex_enter(&idl->idl_lock);
13377 conn_drain(idl->idl_conn, B_FALSE);
13378 mutex_exit(&idl->idl_lock);
13379 }
13380 }
13381
13382 /*
13383 * Determine if the ill and multicast aspects of that packets
13384 * "matches" the conn.
13385 */
13386 boolean_t
13387 conn_wantpacket(conn_t *connp, ip_recv_attr_t *ira, ipha_t *ipha)
13388 {
13389 ill_t *ill = ira->ira_rill;
13390 zoneid_t zoneid = ira->ira_zoneid;
13391 uint_t in_ifindex;
13392 ipaddr_t dst, src;
13393
13394 dst = ipha->ipha_dst;
13395 src = ipha->ipha_src;
13396
13397 /*
13398 * conn_incoming_ifindex is set by IP_BOUND_IF which limits
13399 * unicast, broadcast and multicast reception to
13400 * conn_incoming_ifindex.
13401 * conn_wantpacket is called for unicast, broadcast and
13402 * multicast packets.
13403 */
13404 in_ifindex = connp->conn_incoming_ifindex;
13405
13406 /* mpathd can bind to the under IPMP interface, which we allow */
13407 if (in_ifindex != 0 && in_ifindex != ill->ill_phyint->phyint_ifindex) {
13408 if (!IS_UNDER_IPMP(ill))
13409 return (B_FALSE);
13410
13411 if (in_ifindex != ipmp_ill_get_ipmp_ifindex(ill))
13412 return (B_FALSE);
13413 }
13414
13415 if (!IPCL_ZONE_MATCH(connp, zoneid))
13416 return (B_FALSE);
13417
13418 if (!(ira->ira_flags & IRAF_MULTICAST))
13419 return (B_TRUE);
13420
13421 if (connp->conn_multi_router) {
13422 /* multicast packet and multicast router socket: send up */
13423 return (B_TRUE);
13424 }
13425
13426 if (ipha->ipha_protocol == IPPROTO_PIM ||
13427 ipha->ipha_protocol == IPPROTO_RSVP)
13428 return (B_TRUE);
13429
13430 return (conn_hasmembers_ill_withsrc_v4(connp, dst, src, ira->ira_ill));
13431 }
13432
13433 void
13434 conn_setqfull(conn_t *connp, boolean_t *flow_stopped)
13435 {
13436 if (IPCL_IS_NONSTR(connp)) {
13437 (*connp->conn_upcalls->su_txq_full)
13438 (connp->conn_upper_handle, B_TRUE);
13439 if (flow_stopped != NULL)
13440 *flow_stopped = B_TRUE;
13441 } else {
13442 queue_t *q = connp->conn_wq;
13443
13444 ASSERT(q != NULL);
13445 if (!(q->q_flag & QFULL)) {
13446 mutex_enter(QLOCK(q));
13447 if (!(q->q_flag & QFULL)) {
13448 /* still need to set QFULL */
13449 q->q_flag |= QFULL;
13450 /* set flow_stopped to true under QLOCK */
13451 if (flow_stopped != NULL)
13452 *flow_stopped = B_TRUE;
13453 mutex_exit(QLOCK(q));
13454 } else {
13455 /* flow_stopped is left unchanged */
13456 mutex_exit(QLOCK(q));
13457 }
13458 }
13459 }
13460 }
13461
13462 void
13463 conn_clrqfull(conn_t *connp, boolean_t *flow_stopped)
13464 {
13465 if (IPCL_IS_NONSTR(connp)) {
13466 (*connp->conn_upcalls->su_txq_full)
13467 (connp->conn_upper_handle, B_FALSE);
13468 if (flow_stopped != NULL)
13469 *flow_stopped = B_FALSE;
13470 } else {
13471 queue_t *q = connp->conn_wq;
13472
13473 ASSERT(q != NULL);
13474 if (q->q_flag & QFULL) {
13475 mutex_enter(QLOCK(q));
13476 if (q->q_flag & QFULL) {
13477 q->q_flag &= ~QFULL;
13478 /* set flow_stopped to false under QLOCK */
13479 if (flow_stopped != NULL)
13480 *flow_stopped = B_FALSE;
13481 mutex_exit(QLOCK(q));
13482 if (q->q_flag & QWANTW)
13483 qbackenable(q, 0);
13484 } else {
13485 /* flow_stopped is left unchanged */
13486 mutex_exit(QLOCK(q));
13487 }
13488 }
13489 }
13490
13491 mutex_enter(&connp->conn_lock);
13492 connp->conn_blocked = B_FALSE;
13493 mutex_exit(&connp->conn_lock);
13494 }
13495
13496 /*
13497 * Return the length in bytes of the IPv4 headers (base header, label, and
13498 * other IP options) that will be needed based on the
13499 * ip_pkt_t structure passed by the caller.
13500 *
13501 * The returned length does not include the length of the upper level
13502 * protocol (ULP) header.
13503 * The caller needs to check that the length doesn't exceed the max for IPv4.
13504 */
13505 int
13506 ip_total_hdrs_len_v4(const ip_pkt_t *ipp)
13507 {
13508 int len;
13509
13510 len = IP_SIMPLE_HDR_LENGTH;
13511 if (ipp->ipp_fields & IPPF_LABEL_V4) {
13512 ASSERT(ipp->ipp_label_len_v4 != 0);
13513 /* We need to round up here */
13514 len += (ipp->ipp_label_len_v4 + 3) & ~3;
13515 }
13516
13517 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
13518 ASSERT(ipp->ipp_ipv4_options_len != 0);
13519 ASSERT((ipp->ipp_ipv4_options_len & 3) == 0);
13520 len += ipp->ipp_ipv4_options_len;
13521 }
13522 return (len);
13523 }
13524
13525 /*
13526 * All-purpose routine to build an IPv4 header with options based
13527 * on the abstract ip_pkt_t.
13528 *
13529 * The caller has to set the source and destination address as well as
13530 * ipha_length. The caller has to massage any source route and compensate
13531 * for the ULP pseudo-header checksum due to the source route.
13532 */
13533 void
13534 ip_build_hdrs_v4(uchar_t *buf, uint_t buf_len, const ip_pkt_t *ipp,
13535 uint8_t protocol)
13536 {
13537 ipha_t *ipha = (ipha_t *)buf;
13538 uint8_t *cp;
13539
13540 /* Initialize IPv4 header */
13541 ipha->ipha_type_of_service = ipp->ipp_type_of_service;
13542 ipha->ipha_length = 0; /* Caller will set later */
13543 ipha->ipha_ident = 0;
13544 ipha->ipha_fragment_offset_and_flags = 0;
13545 ipha->ipha_ttl = ipp->ipp_unicast_hops;
13546 ipha->ipha_protocol = protocol;
13547 ipha->ipha_hdr_checksum = 0;
13548
13549 if ((ipp->ipp_fields & IPPF_ADDR) &&
13550 IN6_IS_ADDR_V4MAPPED(&ipp->ipp_addr))
13551 ipha->ipha_src = ipp->ipp_addr_v4;
13552
13553 cp = (uint8_t *)&ipha[1];
13554 if (ipp->ipp_fields & IPPF_LABEL_V4) {
13555 ASSERT(ipp->ipp_label_len_v4 != 0);
13556 bcopy(ipp->ipp_label_v4, cp, ipp->ipp_label_len_v4);
13557 cp += ipp->ipp_label_len_v4;
13558 /* We need to round up here */
13559 while ((uintptr_t)cp & 0x3) {
13560 *cp++ = IPOPT_NOP;
13561 }
13562 }
13563
13564 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
13565 ASSERT(ipp->ipp_ipv4_options_len != 0);
13566 ASSERT((ipp->ipp_ipv4_options_len & 3) == 0);
13567 bcopy(ipp->ipp_ipv4_options, cp, ipp->ipp_ipv4_options_len);
13568 cp += ipp->ipp_ipv4_options_len;
13569 }
13570 ipha->ipha_version_and_hdr_length =
13571 (uint8_t)((IP_VERSION << 4) + buf_len / 4);
13572
13573 ASSERT((int)(cp - buf) == buf_len);
13574 }
13575
13576 /* Allocate the private structure */
13577 static int
13578 ip_priv_alloc(void **bufp)
13579 {
13580 void *buf;
13581
13582 if ((buf = kmem_alloc(sizeof (ip_priv_t), KM_NOSLEEP)) == NULL)
13583 return (ENOMEM);
13584
13585 *bufp = buf;
13586 return (0);
13587 }
13588
13589 /* Function to delete the private structure */
13590 void
13591 ip_priv_free(void *buf)
13592 {
13593 ASSERT(buf != NULL);
13594 kmem_free(buf, sizeof (ip_priv_t));
13595 }
13596
13597 /*
13598 * The entry point for IPPF processing.
13599 * If the classifier (IPGPC_CLASSIFY) is not loaded and configured, the
13600 * routine just returns.
13601 *
13602 * When called, ip_process generates an ipp_packet_t structure
13603 * which holds the state information for this packet and invokes the
13604 * the classifier (via ipp_packet_process). The classification, depending on
13605 * configured filters, results in a list of actions for this packet. Invoking
13606 * an action may cause the packet to be dropped, in which case we return NULL.
13607 * proc indicates the callout position for
13608 * this packet and ill is the interface this packet arrived on or will leave
13609 * on (inbound and outbound resp.).
13610 *
13611 * We do the processing on the rill (mapped to the upper if ipmp), but MIB
13612 * on the ill corrsponding to the destination IP address.
13613 */
13614 mblk_t *
13615 ip_process(ip_proc_t proc, mblk_t *mp, ill_t *rill, ill_t *ill)
13616 {
13617 ip_priv_t *priv;
13618 ipp_action_id_t aid;
13619 int rc = 0;
13620 ipp_packet_t *pp;
13621
13622 /* If the classifier is not loaded, return */
13623 if ((aid = ipp_action_lookup(IPGPC_CLASSIFY)) == IPP_ACTION_INVAL) {
13624 return (mp);
13625 }
13626
13627 ASSERT(mp != NULL);
13628
13629 /* Allocate the packet structure */
13630 rc = ipp_packet_alloc(&pp, "ip", aid);
13631 if (rc != 0)
13632 goto drop;
13633
13634 /* Allocate the private structure */
13635 rc = ip_priv_alloc((void **)&priv);
13636 if (rc != 0) {
13637 ipp_packet_free(pp);
13638 goto drop;
13639 }
13640 priv->proc = proc;
13641 priv->ill_index = ill_get_upper_ifindex(rill);
13642
13643 ipp_packet_set_private(pp, priv, ip_priv_free);
13644 ipp_packet_set_data(pp, mp);
13645
13646 /* Invoke the classifier */
13647 rc = ipp_packet_process(&pp);
13648 if (pp != NULL) {
13649 mp = ipp_packet_get_data(pp);
13650 ipp_packet_free(pp);
13651 if (rc != 0)
13652 goto drop;
13653 return (mp);
13654 } else {
13655 /* No mp to trace in ip_drop_input/ip_drop_output */
13656 mp = NULL;
13657 }
13658 drop:
13659 if (proc == IPP_LOCAL_IN || proc == IPP_FWD_IN) {
13660 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
13661 ip_drop_input("ip_process", mp, ill);
13662 } else {
13663 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
13664 ip_drop_output("ip_process", mp, ill);
13665 }
13666 freemsg(mp);
13667 return (NULL);
13668 }
13669
13670 /*
13671 * Propagate a multicast group membership operation (add/drop) on
13672 * all the interfaces crossed by the related multirt routes.
13673 * The call is considered successful if the operation succeeds
13674 * on at least one interface.
13675 *
13676 * This assumes that a set of IRE_HOST/RTF_MULTIRT has been created for the
13677 * multicast addresses with the ire argument being the first one.
13678 * We walk the bucket to find all the of those.
13679 *
13680 * Common to IPv4 and IPv6.
13681 */
13682 static int
13683 ip_multirt_apply_membership(int (*fn)(conn_t *, boolean_t,
13684 const in6_addr_t *, ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *),
13685 ire_t *ire, conn_t *connp, boolean_t checkonly, const in6_addr_t *v6group,
13686 mcast_record_t fmode, const in6_addr_t *v6src)
13687 {
13688 ire_t *ire_gw;
13689 irb_t *irb;
13690 int ifindex;
13691 int error = 0;
13692 int result;
13693 ip_stack_t *ipst = ire->ire_ipst;
13694 ipaddr_t group;
13695 boolean_t isv6;
13696 int match_flags;
13697
13698 if (IN6_IS_ADDR_V4MAPPED(v6group)) {
13699 IN6_V4MAPPED_TO_IPADDR(v6group, group);
13700 isv6 = B_FALSE;
13701 } else {
13702 isv6 = B_TRUE;
13703 }
13704
13705 irb = ire->ire_bucket;
13706 ASSERT(irb != NULL);
13707
13708 result = 0;
13709 irb_refhold(irb);
13710 for (; ire != NULL; ire = ire->ire_next) {
13711 if ((ire->ire_flags & RTF_MULTIRT) == 0)
13712 continue;
13713
13714 /* We handle -ifp routes by matching on the ill if set */
13715 match_flags = MATCH_IRE_TYPE;
13716 if (ire->ire_ill != NULL)
13717 match_flags |= MATCH_IRE_ILL;
13718
13719 if (isv6) {
13720 if (!IN6_ARE_ADDR_EQUAL(&ire->ire_addr_v6, v6group))
13721 continue;
13722
13723 ire_gw = ire_ftable_lookup_v6(&ire->ire_gateway_addr_v6,
13724 0, 0, IRE_INTERFACE, ire->ire_ill, ALL_ZONES, NULL,
13725 match_flags, 0, ipst, NULL);
13726 } else {
13727 if (ire->ire_addr != group)
13728 continue;
13729
13730 ire_gw = ire_ftable_lookup_v4(ire->ire_gateway_addr,
13731 0, 0, IRE_INTERFACE, ire->ire_ill, ALL_ZONES, NULL,
13732 match_flags, 0, ipst, NULL);
13733 }
13734 /* No interface route exists for the gateway; skip this ire. */
13735 if (ire_gw == NULL)
13736 continue;
13737 if (ire_gw->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
13738 ire_refrele(ire_gw);
13739 continue;
13740 }
13741 ASSERT(ire_gw->ire_ill != NULL); /* IRE_INTERFACE */
13742 ifindex = ire_gw->ire_ill->ill_phyint->phyint_ifindex;
13743
13744 /*
13745 * The operation is considered a success if
13746 * it succeeds at least once on any one interface.
13747 */
13748 error = fn(connp, checkonly, v6group, INADDR_ANY, ifindex,
13749 fmode, v6src);
13750 if (error == 0)
13751 result = CGTP_MCAST_SUCCESS;
13752
13753 ire_refrele(ire_gw);
13754 }
13755 irb_refrele(irb);
13756 /*
13757 * Consider the call as successful if we succeeded on at least
13758 * one interface. Otherwise, return the last encountered error.
13759 */
13760 return (result == CGTP_MCAST_SUCCESS ? 0 : error);
13761 }
13762
13763 /*
13764 * Return the expected CGTP hooks version number.
13765 */
13766 int
13767 ip_cgtp_filter_supported(void)
13768 {
13769 return (ip_cgtp_filter_rev);
13770 }
13771
13772 /*
13773 * CGTP hooks can be registered by invoking this function.
13774 * Checks that the version number matches.
13775 */
13776 int
13777 ip_cgtp_filter_register(netstackid_t stackid, cgtp_filter_ops_t *ops)
13778 {
13779 netstack_t *ns;
13780 ip_stack_t *ipst;
13781
13782 if (ops->cfo_filter_rev != CGTP_FILTER_REV)
13783 return (ENOTSUP);
13784
13785 ns = netstack_find_by_stackid(stackid);
13786 if (ns == NULL)
13787 return (EINVAL);
13788 ipst = ns->netstack_ip;
13789 ASSERT(ipst != NULL);
13790
13791 if (ipst->ips_ip_cgtp_filter_ops != NULL) {
13792 netstack_rele(ns);
13793 return (EALREADY);
13794 }
13795
13796 ipst->ips_ip_cgtp_filter_ops = ops;
13797
13798 ill_set_inputfn_all(ipst);
13799
13800 netstack_rele(ns);
13801 return (0);
13802 }
13803
13804 /*
13805 * CGTP hooks can be unregistered by invoking this function.
13806 * Returns ENXIO if there was no registration.
13807 * Returns EBUSY if the ndd variable has not been turned off.
13808 */
13809 int
13810 ip_cgtp_filter_unregister(netstackid_t stackid)
13811 {
13812 netstack_t *ns;
13813 ip_stack_t *ipst;
13814
13815 ns = netstack_find_by_stackid(stackid);
13816 if (ns == NULL)
13817 return (EINVAL);
13818 ipst = ns->netstack_ip;
13819 ASSERT(ipst != NULL);
13820
13821 if (ipst->ips_ip_cgtp_filter) {
13822 netstack_rele(ns);
13823 return (EBUSY);
13824 }
13825
13826 if (ipst->ips_ip_cgtp_filter_ops == NULL) {
13827 netstack_rele(ns);
13828 return (ENXIO);
13829 }
13830 ipst->ips_ip_cgtp_filter_ops = NULL;
13831
13832 ill_set_inputfn_all(ipst);
13833
13834 netstack_rele(ns);
13835 return (0);
13836 }
13837
13838 /*
13839 * Check whether there is a CGTP filter registration.
13840 * Returns non-zero if there is a registration, otherwise returns zero.
13841 * Note: returns zero if bad stackid.
13842 */
13843 int
13844 ip_cgtp_filter_is_registered(netstackid_t stackid)
13845 {
13846 netstack_t *ns;
13847 ip_stack_t *ipst;
13848 int ret;
13849
13850 ns = netstack_find_by_stackid(stackid);
13851 if (ns == NULL)
13852 return (0);
13853 ipst = ns->netstack_ip;
13854 ASSERT(ipst != NULL);
13855
13856 if (ipst->ips_ip_cgtp_filter_ops != NULL)
13857 ret = 1;
13858 else
13859 ret = 0;
13860
13861 netstack_rele(ns);
13862 return (ret);
13863 }
13864
13865 static int
13866 ip_squeue_switch(int val)
13867 {
13868 int rval;
13869
13870 switch (val) {
13871 case IP_SQUEUE_ENTER_NODRAIN:
13872 rval = SQ_NODRAIN;
13873 break;
13874 case IP_SQUEUE_ENTER:
13875 rval = SQ_PROCESS;
13876 break;
13877 case IP_SQUEUE_FILL:
13878 default:
13879 rval = SQ_FILL;
13880 break;
13881 }
13882 return (rval);
13883 }
13884
13885 static void *
13886 ip_kstat2_init(netstackid_t stackid, ip_stat_t *ip_statisticsp)
13887 {
13888 kstat_t *ksp;
13889
13890 ip_stat_t template = {
13891 { "ip_udp_fannorm", KSTAT_DATA_UINT64 },
13892 { "ip_udp_fanmb", KSTAT_DATA_UINT64 },
13893 { "ip_recv_pullup", KSTAT_DATA_UINT64 },
13894 { "ip_db_ref", KSTAT_DATA_UINT64 },
13895 { "ip_notaligned", KSTAT_DATA_UINT64 },
13896 { "ip_multimblk", KSTAT_DATA_UINT64 },
13897 { "ip_opt", KSTAT_DATA_UINT64 },
13898 { "ipsec_proto_ahesp", KSTAT_DATA_UINT64 },
13899 { "ip_conn_flputbq", KSTAT_DATA_UINT64 },
13900 { "ip_conn_walk_drain", KSTAT_DATA_UINT64 },
13901 { "ip_out_sw_cksum", KSTAT_DATA_UINT64 },
13902 { "ip_out_sw_cksum_bytes", KSTAT_DATA_UINT64 },
13903 { "ip_in_sw_cksum", KSTAT_DATA_UINT64 },
13904 { "ip_ire_reclaim_calls", KSTAT_DATA_UINT64 },
13905 { "ip_ire_reclaim_deleted", KSTAT_DATA_UINT64 },
13906 { "ip_nce_reclaim_calls", KSTAT_DATA_UINT64 },
13907 { "ip_nce_reclaim_deleted", KSTAT_DATA_UINT64 },
13908 { "ip_dce_reclaim_calls", KSTAT_DATA_UINT64 },
13909 { "ip_dce_reclaim_deleted", KSTAT_DATA_UINT64 },
13910 { "ip_tcp_in_full_hw_cksum_err", KSTAT_DATA_UINT64 },
13911 { "ip_tcp_in_part_hw_cksum_err", KSTAT_DATA_UINT64 },
13912 { "ip_tcp_in_sw_cksum_err", KSTAT_DATA_UINT64 },
13913 { "ip_udp_in_full_hw_cksum_err", KSTAT_DATA_UINT64 },
13914 { "ip_udp_in_part_hw_cksum_err", KSTAT_DATA_UINT64 },
13915 { "ip_udp_in_sw_cksum_err", KSTAT_DATA_UINT64 },
13916 { "conn_in_recvdstaddr", KSTAT_DATA_UINT64 },
13917 { "conn_in_recvopts", KSTAT_DATA_UINT64 },
13918 { "conn_in_recvif", KSTAT_DATA_UINT64 },
13919 { "conn_in_recvslla", KSTAT_DATA_UINT64 },
13920 { "conn_in_recvucred", KSTAT_DATA_UINT64 },
13921 { "conn_in_recvttl", KSTAT_DATA_UINT64 },
13922 { "conn_in_recvhopopts", KSTAT_DATA_UINT64 },
13923 { "conn_in_recvhoplimit", KSTAT_DATA_UINT64 },
13924 { "conn_in_recvdstopts", KSTAT_DATA_UINT64 },
13925 { "conn_in_recvrthdrdstopts", KSTAT_DATA_UINT64 },
13926 { "conn_in_recvrthdr", KSTAT_DATA_UINT64 },
13927 { "conn_in_recvpktinfo", KSTAT_DATA_UINT64 },
13928 { "conn_in_recvtclass", KSTAT_DATA_UINT64 },
13929 { "conn_in_timestamp", KSTAT_DATA_UINT64 },
13930 };
13931
13932 ksp = kstat_create_netstack("ip", 0, "ipstat", "net",
13933 KSTAT_TYPE_NAMED, sizeof (template) / sizeof (kstat_named_t),
13934 KSTAT_FLAG_VIRTUAL, stackid);
13935
13936 if (ksp == NULL)
13937 return (NULL);
13938
13939 bcopy(&template, ip_statisticsp, sizeof (template));
13940 ksp->ks_data = (void *)ip_statisticsp;
13941 ksp->ks_private = (void *)(uintptr_t)stackid;
13942
13943 kstat_install(ksp);
13944 return (ksp);
13945 }
13946
13947 static void
13948 ip_kstat2_fini(netstackid_t stackid, kstat_t *ksp)
13949 {
13950 if (ksp != NULL) {
13951 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
13952 kstat_delete_netstack(ksp, stackid);
13953 }
13954 }
13955
13956 static void *
13957 ip_kstat_init(netstackid_t stackid, ip_stack_t *ipst)
13958 {
13959 kstat_t *ksp;
13960
13961 ip_named_kstat_t template = {
13962 { "forwarding", KSTAT_DATA_UINT32, 0 },
13963 { "defaultTTL", KSTAT_DATA_UINT32, 0 },
13964 { "inReceives", KSTAT_DATA_UINT64, 0 },
13965 { "inHdrErrors", KSTAT_DATA_UINT32, 0 },
13966 { "inAddrErrors", KSTAT_DATA_UINT32, 0 },
13967 { "forwDatagrams", KSTAT_DATA_UINT64, 0 },
13968 { "inUnknownProtos", KSTAT_DATA_UINT32, 0 },
13969 { "inDiscards", KSTAT_DATA_UINT32, 0 },
13970 { "inDelivers", KSTAT_DATA_UINT64, 0 },
13971 { "outRequests", KSTAT_DATA_UINT64, 0 },
13972 { "outDiscards", KSTAT_DATA_UINT32, 0 },
13973 { "outNoRoutes", KSTAT_DATA_UINT32, 0 },
13974 { "reasmTimeout", KSTAT_DATA_UINT32, 0 },
13975 { "reasmReqds", KSTAT_DATA_UINT32, 0 },
13976 { "reasmOKs", KSTAT_DATA_UINT32, 0 },
13977 { "reasmFails", KSTAT_DATA_UINT32, 0 },
13978 { "fragOKs", KSTAT_DATA_UINT32, 0 },
13979 { "fragFails", KSTAT_DATA_UINT32, 0 },
13980 { "fragCreates", KSTAT_DATA_UINT32, 0 },
13981 { "addrEntrySize", KSTAT_DATA_INT32, 0 },
13982 { "routeEntrySize", KSTAT_DATA_INT32, 0 },
13983 { "netToMediaEntrySize", KSTAT_DATA_INT32, 0 },
13984 { "routingDiscards", KSTAT_DATA_UINT32, 0 },
13985 { "inErrs", KSTAT_DATA_UINT32, 0 },
13986 { "noPorts", KSTAT_DATA_UINT32, 0 },
13987 { "inCksumErrs", KSTAT_DATA_UINT32, 0 },
13988 { "reasmDuplicates", KSTAT_DATA_UINT32, 0 },
13989 { "reasmPartDups", KSTAT_DATA_UINT32, 0 },
13990 { "forwProhibits", KSTAT_DATA_UINT32, 0 },
13991 { "udpInCksumErrs", KSTAT_DATA_UINT32, 0 },
13992 { "udpInOverflows", KSTAT_DATA_UINT32, 0 },
13993 { "rawipInOverflows", KSTAT_DATA_UINT32, 0 },
13994 { "ipsecInSucceeded", KSTAT_DATA_UINT32, 0 },
13995 { "ipsecInFailed", KSTAT_DATA_INT32, 0 },
13996 { "memberEntrySize", KSTAT_DATA_INT32, 0 },
13997 { "inIPv6", KSTAT_DATA_UINT32, 0 },
13998 { "outIPv6", KSTAT_DATA_UINT32, 0 },
13999 { "outSwitchIPv6", KSTAT_DATA_UINT32, 0 },
14000 };
14001
14002 ksp = kstat_create_netstack("ip", 0, "ip", "mib2", KSTAT_TYPE_NAMED,
14003 NUM_OF_FIELDS(ip_named_kstat_t), 0, stackid);
14004 if (ksp == NULL || ksp->ks_data == NULL)
14005 return (NULL);
14006
14007 template.forwarding.value.ui32 = WE_ARE_FORWARDING(ipst) ? 1:2;
14008 template.defaultTTL.value.ui32 = (uint32_t)ipst->ips_ip_def_ttl;
14009 template.reasmTimeout.value.ui32 = ipst->ips_ip_reassembly_timeout;
14010 template.addrEntrySize.value.i32 = sizeof (mib2_ipAddrEntry_t);
14011 template.routeEntrySize.value.i32 = sizeof (mib2_ipRouteEntry_t);
14012
14013 template.netToMediaEntrySize.value.i32 =
14014 sizeof (mib2_ipNetToMediaEntry_t);
14015
14016 template.memberEntrySize.value.i32 = sizeof (ipv6_member_t);
14017
14018 bcopy(&template, ksp->ks_data, sizeof (template));
14019 ksp->ks_update = ip_kstat_update;
14020 ksp->ks_private = (void *)(uintptr_t)stackid;
14021
14022 kstat_install(ksp);
14023 return (ksp);
14024 }
14025
14026 static void
14027 ip_kstat_fini(netstackid_t stackid, kstat_t *ksp)
14028 {
14029 if (ksp != NULL) {
14030 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
14031 kstat_delete_netstack(ksp, stackid);
14032 }
14033 }
14034
14035 static int
14036 ip_kstat_update(kstat_t *kp, int rw)
14037 {
14038 ip_named_kstat_t *ipkp;
14039 mib2_ipIfStatsEntry_t ipmib;
14040 ill_walk_context_t ctx;
14041 ill_t *ill;
14042 netstackid_t stackid = (zoneid_t)(uintptr_t)kp->ks_private;
14043 netstack_t *ns;
14044 ip_stack_t *ipst;
14045
14046 if (kp == NULL || kp->ks_data == NULL)
14047 return (EIO);
14048
14049 if (rw == KSTAT_WRITE)
14050 return (EACCES);
14051
14052 ns = netstack_find_by_stackid(stackid);
14053 if (ns == NULL)
14054 return (-1);
14055 ipst = ns->netstack_ip;
14056 if (ipst == NULL) {
14057 netstack_rele(ns);
14058 return (-1);
14059 }
14060 ipkp = (ip_named_kstat_t *)kp->ks_data;
14061
14062 bcopy(&ipst->ips_ip_mib, &ipmib, sizeof (ipmib));
14063 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
14064 ill = ILL_START_WALK_V4(&ctx, ipst);
14065 for (; ill != NULL; ill = ill_next(&ctx, ill))
14066 ip_mib2_add_ip_stats(&ipmib, ill->ill_ip_mib);
14067 rw_exit(&ipst->ips_ill_g_lock);
14068
14069 ipkp->forwarding.value.ui32 = ipmib.ipIfStatsForwarding;
14070 ipkp->defaultTTL.value.ui32 = ipmib.ipIfStatsDefaultTTL;
14071 ipkp->inReceives.value.ui64 = ipmib.ipIfStatsHCInReceives;
14072 ipkp->inHdrErrors.value.ui32 = ipmib.ipIfStatsInHdrErrors;
14073 ipkp->inAddrErrors.value.ui32 = ipmib.ipIfStatsInAddrErrors;
14074 ipkp->forwDatagrams.value.ui64 = ipmib.ipIfStatsHCOutForwDatagrams;
14075 ipkp->inUnknownProtos.value.ui32 = ipmib.ipIfStatsInUnknownProtos;
14076 ipkp->inDiscards.value.ui32 = ipmib.ipIfStatsInDiscards;
14077 ipkp->inDelivers.value.ui64 = ipmib.ipIfStatsHCInDelivers;
14078 ipkp->outRequests.value.ui64 = ipmib.ipIfStatsHCOutRequests;
14079 ipkp->outDiscards.value.ui32 = ipmib.ipIfStatsOutDiscards;
14080 ipkp->outNoRoutes.value.ui32 = ipmib.ipIfStatsOutNoRoutes;
14081 ipkp->reasmTimeout.value.ui32 = ipst->ips_ip_reassembly_timeout;
14082 ipkp->reasmReqds.value.ui32 = ipmib.ipIfStatsReasmReqds;
14083 ipkp->reasmOKs.value.ui32 = ipmib.ipIfStatsReasmOKs;
14084 ipkp->reasmFails.value.ui32 = ipmib.ipIfStatsReasmFails;
14085 ipkp->fragOKs.value.ui32 = ipmib.ipIfStatsOutFragOKs;
14086 ipkp->fragFails.value.ui32 = ipmib.ipIfStatsOutFragFails;
14087 ipkp->fragCreates.value.ui32 = ipmib.ipIfStatsOutFragCreates;
14088
14089 ipkp->routingDiscards.value.ui32 = 0;
14090 ipkp->inErrs.value.ui32 = ipmib.tcpIfStatsInErrs;
14091 ipkp->noPorts.value.ui32 = ipmib.udpIfStatsNoPorts;
14092 ipkp->inCksumErrs.value.ui32 = ipmib.ipIfStatsInCksumErrs;
14093 ipkp->reasmDuplicates.value.ui32 = ipmib.ipIfStatsReasmDuplicates;
14094 ipkp->reasmPartDups.value.ui32 = ipmib.ipIfStatsReasmPartDups;
14095 ipkp->forwProhibits.value.ui32 = ipmib.ipIfStatsForwProhibits;
14096 ipkp->udpInCksumErrs.value.ui32 = ipmib.udpIfStatsInCksumErrs;
14097 ipkp->udpInOverflows.value.ui32 = ipmib.udpIfStatsInOverflows;
14098 ipkp->rawipInOverflows.value.ui32 = ipmib.rawipIfStatsInOverflows;
14099 ipkp->ipsecInSucceeded.value.ui32 = ipmib.ipsecIfStatsInSucceeded;
14100 ipkp->ipsecInFailed.value.i32 = ipmib.ipsecIfStatsInFailed;
14101
14102 ipkp->inIPv6.value.ui32 = ipmib.ipIfStatsInWrongIPVersion;
14103 ipkp->outIPv6.value.ui32 = ipmib.ipIfStatsOutWrongIPVersion;
14104 ipkp->outSwitchIPv6.value.ui32 = ipmib.ipIfStatsOutSwitchIPVersion;
14105
14106 netstack_rele(ns);
14107
14108 return (0);
14109 }
14110
14111 static void *
14112 icmp_kstat_init(netstackid_t stackid)
14113 {
14114 kstat_t *ksp;
14115
14116 icmp_named_kstat_t template = {
14117 { "inMsgs", KSTAT_DATA_UINT32 },
14118 { "inErrors", KSTAT_DATA_UINT32 },
14119 { "inDestUnreachs", KSTAT_DATA_UINT32 },
14120 { "inTimeExcds", KSTAT_DATA_UINT32 },
14121 { "inParmProbs", KSTAT_DATA_UINT32 },
14122 { "inSrcQuenchs", KSTAT_DATA_UINT32 },
14123 { "inRedirects", KSTAT_DATA_UINT32 },
14124 { "inEchos", KSTAT_DATA_UINT32 },
14125 { "inEchoReps", KSTAT_DATA_UINT32 },
14126 { "inTimestamps", KSTAT_DATA_UINT32 },
14127 { "inTimestampReps", KSTAT_DATA_UINT32 },
14128 { "inAddrMasks", KSTAT_DATA_UINT32 },
14129 { "inAddrMaskReps", KSTAT_DATA_UINT32 },
14130 { "outMsgs", KSTAT_DATA_UINT32 },
14131 { "outErrors", KSTAT_DATA_UINT32 },
14132 { "outDestUnreachs", KSTAT_DATA_UINT32 },
14133 { "outTimeExcds", KSTAT_DATA_UINT32 },
14134 { "outParmProbs", KSTAT_DATA_UINT32 },
14135 { "outSrcQuenchs", KSTAT_DATA_UINT32 },
14136 { "outRedirects", KSTAT_DATA_UINT32 },
14137 { "outEchos", KSTAT_DATA_UINT32 },
14138 { "outEchoReps", KSTAT_DATA_UINT32 },
14139 { "outTimestamps", KSTAT_DATA_UINT32 },
14140 { "outTimestampReps", KSTAT_DATA_UINT32 },
14141 { "outAddrMasks", KSTAT_DATA_UINT32 },
14142 { "outAddrMaskReps", KSTAT_DATA_UINT32 },
14143 { "inChksumErrs", KSTAT_DATA_UINT32 },
14144 { "inUnknowns", KSTAT_DATA_UINT32 },
14145 { "inFragNeeded", KSTAT_DATA_UINT32 },
14146 { "outFragNeeded", KSTAT_DATA_UINT32 },
14147 { "outDrops", KSTAT_DATA_UINT32 },
14148 { "inOverFlows", KSTAT_DATA_UINT32 },
14149 { "inBadRedirects", KSTAT_DATA_UINT32 },
14150 };
14151
14152 ksp = kstat_create_netstack("ip", 0, "icmp", "mib2", KSTAT_TYPE_NAMED,
14153 NUM_OF_FIELDS(icmp_named_kstat_t), 0, stackid);
14154 if (ksp == NULL || ksp->ks_data == NULL)
14155 return (NULL);
14156
14157 bcopy(&template, ksp->ks_data, sizeof (template));
14158
14159 ksp->ks_update = icmp_kstat_update;
14160 ksp->ks_private = (void *)(uintptr_t)stackid;
14161
14162 kstat_install(ksp);
14163 return (ksp);
14164 }
14165
14166 static void
14167 icmp_kstat_fini(netstackid_t stackid, kstat_t *ksp)
14168 {
14169 if (ksp != NULL) {
14170 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
14171 kstat_delete_netstack(ksp, stackid);
14172 }
14173 }
14174
14175 static int
14176 icmp_kstat_update(kstat_t *kp, int rw)
14177 {
14178 icmp_named_kstat_t *icmpkp;
14179 netstackid_t stackid = (zoneid_t)(uintptr_t)kp->ks_private;
14180 netstack_t *ns;
14181 ip_stack_t *ipst;
14182
14183 if ((kp == NULL) || (kp->ks_data == NULL))
14184 return (EIO);
14185
14186 if (rw == KSTAT_WRITE)
14187 return (EACCES);
14188
14189 ns = netstack_find_by_stackid(stackid);
14190 if (ns == NULL)
14191 return (-1);
14192 ipst = ns->netstack_ip;
14193 if (ipst == NULL) {
14194 netstack_rele(ns);
14195 return (-1);
14196 }
14197 icmpkp = (icmp_named_kstat_t *)kp->ks_data;
14198
14199 icmpkp->inMsgs.value.ui32 = ipst->ips_icmp_mib.icmpInMsgs;
14200 icmpkp->inErrors.value.ui32 = ipst->ips_icmp_mib.icmpInErrors;
14201 icmpkp->inDestUnreachs.value.ui32 =
14202 ipst->ips_icmp_mib.icmpInDestUnreachs;
14203 icmpkp->inTimeExcds.value.ui32 = ipst->ips_icmp_mib.icmpInTimeExcds;
14204 icmpkp->inParmProbs.value.ui32 = ipst->ips_icmp_mib.icmpInParmProbs;
14205 icmpkp->inSrcQuenchs.value.ui32 = ipst->ips_icmp_mib.icmpInSrcQuenchs;
14206 icmpkp->inRedirects.value.ui32 = ipst->ips_icmp_mib.icmpInRedirects;
14207 icmpkp->inEchos.value.ui32 = ipst->ips_icmp_mib.icmpInEchos;
14208 icmpkp->inEchoReps.value.ui32 = ipst->ips_icmp_mib.icmpInEchoReps;
14209 icmpkp->inTimestamps.value.ui32 = ipst->ips_icmp_mib.icmpInTimestamps;
14210 icmpkp->inTimestampReps.value.ui32 =
14211 ipst->ips_icmp_mib.icmpInTimestampReps;
14212 icmpkp->inAddrMasks.value.ui32 = ipst->ips_icmp_mib.icmpInAddrMasks;
14213 icmpkp->inAddrMaskReps.value.ui32 =
14214 ipst->ips_icmp_mib.icmpInAddrMaskReps;
14215 icmpkp->outMsgs.value.ui32 = ipst->ips_icmp_mib.icmpOutMsgs;
14216 icmpkp->outErrors.value.ui32 = ipst->ips_icmp_mib.icmpOutErrors;
14217 icmpkp->outDestUnreachs.value.ui32 =
14218 ipst->ips_icmp_mib.icmpOutDestUnreachs;
14219 icmpkp->outTimeExcds.value.ui32 = ipst->ips_icmp_mib.icmpOutTimeExcds;
14220 icmpkp->outParmProbs.value.ui32 = ipst->ips_icmp_mib.icmpOutParmProbs;
14221 icmpkp->outSrcQuenchs.value.ui32 =
14222 ipst->ips_icmp_mib.icmpOutSrcQuenchs;
14223 icmpkp->outRedirects.value.ui32 = ipst->ips_icmp_mib.icmpOutRedirects;
14224 icmpkp->outEchos.value.ui32 = ipst->ips_icmp_mib.icmpOutEchos;
14225 icmpkp->outEchoReps.value.ui32 = ipst->ips_icmp_mib.icmpOutEchoReps;
14226 icmpkp->outTimestamps.value.ui32 =
14227 ipst->ips_icmp_mib.icmpOutTimestamps;
14228 icmpkp->outTimestampReps.value.ui32 =
14229 ipst->ips_icmp_mib.icmpOutTimestampReps;
14230 icmpkp->outAddrMasks.value.ui32 =
14231 ipst->ips_icmp_mib.icmpOutAddrMasks;
14232 icmpkp->outAddrMaskReps.value.ui32 =
14233 ipst->ips_icmp_mib.icmpOutAddrMaskReps;
14234 icmpkp->inCksumErrs.value.ui32 = ipst->ips_icmp_mib.icmpInCksumErrs;
14235 icmpkp->inUnknowns.value.ui32 = ipst->ips_icmp_mib.icmpInUnknowns;
14236 icmpkp->inFragNeeded.value.ui32 = ipst->ips_icmp_mib.icmpInFragNeeded;
14237 icmpkp->outFragNeeded.value.ui32 =
14238 ipst->ips_icmp_mib.icmpOutFragNeeded;
14239 icmpkp->outDrops.value.ui32 = ipst->ips_icmp_mib.icmpOutDrops;
14240 icmpkp->inOverflows.value.ui32 = ipst->ips_icmp_mib.icmpInOverflows;
14241 icmpkp->inBadRedirects.value.ui32 =
14242 ipst->ips_icmp_mib.icmpInBadRedirects;
14243
14244 netstack_rele(ns);
14245 return (0);
14246 }
14247
14248 /*
14249 * This is the fanout function for raw socket opened for SCTP. Note
14250 * that it is called after SCTP checks that there is no socket which
14251 * wants a packet. Then before SCTP handles this out of the blue packet,
14252 * this function is called to see if there is any raw socket for SCTP.
14253 * If there is and it is bound to the correct address, the packet will
14254 * be sent to that socket. Note that only one raw socket can be bound to
14255 * a port. This is assured in ipcl_sctp_hash_insert();
14256 */
14257 void
14258 ip_fanout_sctp_raw(mblk_t *mp, ipha_t *ipha, ip6_t *ip6h, uint32_t ports,
14259 ip_recv_attr_t *ira)
14260 {
14261 conn_t *connp;
14262 queue_t *rq;
14263 boolean_t secure;
14264 ill_t *ill = ira->ira_ill;
14265 ip_stack_t *ipst = ill->ill_ipst;
14266 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
14267 sctp_stack_t *sctps = ipst->ips_netstack->netstack_sctp;
14268 iaflags_t iraflags = ira->ira_flags;
14269 ill_t *rill = ira->ira_rill;
14270
14271 secure = iraflags & IRAF_IPSEC_SECURE;
14272
14273 connp = ipcl_classify_raw(mp, IPPROTO_SCTP, ports, ipha, ip6h,
14274 ira, ipst);
14275 if (connp == NULL) {
14276 /*
14277 * Although raw sctp is not summed, OOB chunks must be.
14278 * Drop the packet here if the sctp checksum failed.
14279 */
14280 if (iraflags & IRAF_SCTP_CSUM_ERR) {
14281 SCTPS_BUMP_MIB(sctps, sctpChecksumError);
14282 freemsg(mp);
14283 return;
14284 }
14285 ira->ira_ill = ira->ira_rill = NULL;
14286 sctp_ootb_input(mp, ira, ipst);
14287 ira->ira_ill = ill;
14288 ira->ira_rill = rill;
14289 return;
14290 }
14291 rq = connp->conn_rq;
14292 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld : !canputnext(rq)) {
14293 CONN_DEC_REF(connp);
14294 BUMP_MIB(ill->ill_ip_mib, rawipIfStatsInOverflows);
14295 freemsg(mp);
14296 return;
14297 }
14298 if (((iraflags & IRAF_IS_IPV4) ?
14299 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
14300 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
14301 secure) {
14302 mp = ipsec_check_inbound_policy(mp, connp, ipha,
14303 ip6h, ira);
14304 if (mp == NULL) {
14305 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
14306 /* Note that mp is NULL */
14307 ip_drop_input("ipIfStatsInDiscards", mp, ill);
14308 CONN_DEC_REF(connp);
14309 return;
14310 }
14311 }
14312
14313 if (iraflags & IRAF_ICMP_ERROR) {
14314 (connp->conn_recvicmp)(connp, mp, NULL, ira);
14315 } else {
14316 ill_t *rill = ira->ira_rill;
14317
14318 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
14319 /* This is the SOCK_RAW, IPPROTO_SCTP case. */
14320 ira->ira_ill = ira->ira_rill = NULL;
14321 (connp->conn_recv)(connp, mp, NULL, ira);
14322 ira->ira_ill = ill;
14323 ira->ira_rill = rill;
14324 }
14325 CONN_DEC_REF(connp);
14326 }
14327
14328 /*
14329 * Free a packet that has the link-layer dl_unitdata_req_t or fast-path
14330 * header before the ip payload.
14331 */
14332 static void
14333 ip_xmit_flowctl_drop(ill_t *ill, mblk_t *mp, boolean_t is_fp_mp, int fp_mp_len)
14334 {
14335 int len = (mp->b_wptr - mp->b_rptr);
14336 mblk_t *ip_mp;
14337
14338 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14339 if (is_fp_mp || len != fp_mp_len) {
14340 if (len > fp_mp_len) {
14341 /*
14342 * fastpath header and ip header in the first mblk
14343 */
14344 mp->b_rptr += fp_mp_len;
14345 } else {
14346 /*
14347 * ip_xmit_attach_llhdr had to prepend an mblk to
14348 * attach the fastpath header before ip header.
14349 */
14350 ip_mp = mp->b_cont;
14351 freeb(mp);
14352 mp = ip_mp;
14353 mp->b_rptr += (fp_mp_len - len);
14354 }
14355 } else {
14356 ip_mp = mp->b_cont;
14357 freeb(mp);
14358 mp = ip_mp;
14359 }
14360 ip_drop_output("ipIfStatsOutDiscards - flow ctl", mp, ill);
14361 freemsg(mp);
14362 }
14363
14364 /*
14365 * Normal post fragmentation function.
14366 *
14367 * Send a packet using the passed in nce. This handles both IPv4 and IPv6
14368 * using the same state machine.
14369 *
14370 * We return an error on failure. In particular we return EWOULDBLOCK
14371 * when the driver flow controls. In that case this ensures that ip_wsrv runs
14372 * (currently by canputnext failure resulting in backenabling from GLD.)
14373 * This allows the callers of conn_ip_output() to use EWOULDBLOCK as an
14374 * indication that they can flow control until ip_wsrv() tells then to restart.
14375 *
14376 * If the nce passed by caller is incomplete, this function
14377 * queues the packet and if necessary, sends ARP request and bails.
14378 * If the Neighbor Cache passed is fully resolved, we simply prepend
14379 * the link-layer header to the packet, do ipsec hw acceleration
14380 * work if necessary, and send the packet out on the wire.
14381 */
14382 /* ARGSUSED6 */
14383 int
14384 ip_xmit(mblk_t *mp, nce_t *nce, iaflags_t ixaflags, uint_t pkt_len,
14385 uint32_t xmit_hint, zoneid_t szone, zoneid_t nolzid, uintptr_t *ixacookie)
14386 {
14387 queue_t *wq;
14388 ill_t *ill = nce->nce_ill;
14389 ip_stack_t *ipst = ill->ill_ipst;
14390 uint64_t delta;
14391 boolean_t isv6 = ill->ill_isv6;
14392 boolean_t fp_mp;
14393 ncec_t *ncec = nce->nce_common;
14394 int64_t now = LBOLT_FASTPATH64;
14395 boolean_t is_probe;
14396
14397 DTRACE_PROBE1(ip__xmit, nce_t *, nce);
14398
14399 ASSERT(mp != NULL);
14400 ASSERT(mp->b_datap->db_type == M_DATA);
14401 ASSERT(pkt_len == msgdsize(mp));
14402
14403 /*
14404 * If we have already been here and are coming back after ARP/ND.
14405 * the IXAF_NO_TRACE flag is set. We skip FW_HOOKS, DTRACE and ipobs
14406 * in that case since they have seen the packet when it came here
14407 * the first time.
14408 */
14409 if (ixaflags & IXAF_NO_TRACE)
14410 goto sendit;
14411
14412 if (ixaflags & IXAF_IS_IPV4) {
14413 ipha_t *ipha = (ipha_t *)mp->b_rptr;
14414
14415 ASSERT(!isv6);
14416 ASSERT(pkt_len == ntohs(((ipha_t *)mp->b_rptr)->ipha_length));
14417 if (HOOKS4_INTERESTED_PHYSICAL_OUT(ipst) &&
14418 !(ixaflags & IXAF_NO_PFHOOK)) {
14419 int error;
14420
14421 FW_HOOKS(ipst->ips_ip4_physical_out_event,
14422 ipst->ips_ipv4firewall_physical_out,
14423 NULL, ill, ipha, mp, mp, 0, ipst, error);
14424 DTRACE_PROBE1(ip4__physical__out__end,
14425 mblk_t *, mp);
14426 if (mp == NULL)
14427 return (error);
14428
14429 /* The length could have changed */
14430 pkt_len = msgdsize(mp);
14431 }
14432 if (ipst->ips_ip4_observe.he_interested) {
14433 /*
14434 * Note that for TX the zoneid is the sending
14435 * zone, whether or not MLP is in play.
14436 * Since the szone argument is the IP zoneid (i.e.,
14437 * zero for exclusive-IP zones) and ipobs wants
14438 * the system zoneid, we map it here.
14439 */
14440 szone = IP_REAL_ZONEID(szone, ipst);
14441
14442 /*
14443 * On the outbound path the destination zone will be
14444 * unknown as we're sending this packet out on the
14445 * wire.
14446 */
14447 ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone, ALL_ZONES,
14448 ill, ipst);
14449 }
14450 DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
14451 void_ip_t *, ipha, __dtrace_ipsr_ill_t *, ill,
14452 ipha_t *, ipha, ip6_t *, NULL, int, 0);
14453 } else {
14454 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
14455
14456 ASSERT(isv6);
14457 ASSERT(pkt_len ==
14458 ntohs(((ip6_t *)mp->b_rptr)->ip6_plen) + IPV6_HDR_LEN);
14459 if (HOOKS6_INTERESTED_PHYSICAL_OUT(ipst) &&
14460 !(ixaflags & IXAF_NO_PFHOOK)) {
14461 int error;
14462
14463 FW_HOOKS6(ipst->ips_ip6_physical_out_event,
14464 ipst->ips_ipv6firewall_physical_out,
14465 NULL, ill, ip6h, mp, mp, 0, ipst, error);
14466 DTRACE_PROBE1(ip6__physical__out__end,
14467 mblk_t *, mp);
14468 if (mp == NULL)
14469 return (error);
14470
14471 /* The length could have changed */
14472 pkt_len = msgdsize(mp);
14473 }
14474 if (ipst->ips_ip6_observe.he_interested) {
14475 /* See above */
14476 szone = IP_REAL_ZONEID(szone, ipst);
14477
14478 ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone, ALL_ZONES,
14479 ill, ipst);
14480 }
14481 DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
14482 void_ip_t *, ip6h, __dtrace_ipsr_ill_t *, ill,
14483 ipha_t *, NULL, ip6_t *, ip6h, int, 0);
14484 }
14485
14486 sendit:
14487 /*
14488 * We check the state without a lock because the state can never
14489 * move "backwards" to initial or incomplete.
14490 */
14491 switch (ncec->ncec_state) {
14492 case ND_REACHABLE:
14493 case ND_STALE:
14494 case ND_DELAY:
14495 case ND_PROBE:
14496 mp = ip_xmit_attach_llhdr(mp, nce);
14497 if (mp == NULL) {
14498 /*
14499 * ip_xmit_attach_llhdr has increased
14500 * ipIfStatsOutDiscards and called ip_drop_output()
14501 */
14502 return (ENOBUFS);
14503 }
14504 /*
14505 * check if nce_fastpath completed and we tagged on a
14506 * copy of nce_fp_mp in ip_xmit_attach_llhdr().
14507 */
14508 fp_mp = (mp->b_datap->db_type == M_DATA);
14509
14510 if (fp_mp &&
14511 (ill->ill_capabilities & ILL_CAPAB_DLD_DIRECT)) {
14512 ill_dld_direct_t *idd;
14513
14514 idd = &ill->ill_dld_capab->idc_direct;
14515 /*
14516 * Send the packet directly to DLD, where it
14517 * may be queued depending on the availability
14518 * of transmit resources at the media layer.
14519 * Return value should be taken into
14520 * account and flow control the TCP.
14521 */
14522 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
14523 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
14524 pkt_len);
14525
14526 if (ixaflags & IXAF_NO_DEV_FLOW_CTL) {
14527 (void) idd->idd_tx_df(idd->idd_tx_dh, mp,
14528 (uintptr_t)xmit_hint, IP_DROP_ON_NO_DESC);
14529 } else {
14530 uintptr_t cookie;
14531
14532 if ((cookie = idd->idd_tx_df(idd->idd_tx_dh,
14533 mp, (uintptr_t)xmit_hint, 0)) != 0) {
14534 if (ixacookie != NULL)
14535 *ixacookie = cookie;
14536 return (EWOULDBLOCK);
14537 }
14538 }
14539 } else {
14540 wq = ill->ill_wq;
14541
14542 if (!(ixaflags & IXAF_NO_DEV_FLOW_CTL) &&
14543 !canputnext(wq)) {
14544 if (ixacookie != NULL)
14545 *ixacookie = 0;
14546 ip_xmit_flowctl_drop(ill, mp, fp_mp,
14547 nce->nce_fp_mp != NULL ?
14548 MBLKL(nce->nce_fp_mp) : 0);
14549 return (EWOULDBLOCK);
14550 }
14551 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
14552 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
14553 pkt_len);
14554 putnext(wq, mp);
14555 }
14556
14557 /*
14558 * The rest of this function implements Neighbor Unreachability
14559 * detection. Determine if the ncec is eligible for NUD.
14560 */
14561 if (ncec->ncec_flags & NCE_F_NONUD)
14562 return (0);
14563
14564 ASSERT(ncec->ncec_state != ND_INCOMPLETE);
14565
14566 /*
14567 * Check for upper layer advice
14568 */
14569 if (ixaflags & IXAF_REACH_CONF) {
14570 timeout_id_t tid;
14571
14572 /*
14573 * It should be o.k. to check the state without
14574 * a lock here, at most we lose an advice.
14575 */
14576 ncec->ncec_last = TICK_TO_MSEC(now);
14577 if (ncec->ncec_state != ND_REACHABLE) {
14578 mutex_enter(&ncec->ncec_lock);
14579 ncec->ncec_state = ND_REACHABLE;
14580 tid = ncec->ncec_timeout_id;
14581 ncec->ncec_timeout_id = 0;
14582 mutex_exit(&ncec->ncec_lock);
14583 (void) untimeout(tid);
14584 if (ip_debug > 2) {
14585 /* ip1dbg */
14586 pr_addr_dbg("ip_xmit: state"
14587 " for %s changed to"
14588 " REACHABLE\n", AF_INET6,
14589 &ncec->ncec_addr);
14590 }
14591 }
14592 return (0);
14593 }
14594
14595 delta = TICK_TO_MSEC(now) - ncec->ncec_last;
14596 ip1dbg(("ip_xmit: delta = %" PRId64
14597 " ill_reachable_time = %d \n", delta,
14598 ill->ill_reachable_time));
14599 if (delta > (uint64_t)ill->ill_reachable_time) {
14600 mutex_enter(&ncec->ncec_lock);
14601 switch (ncec->ncec_state) {
14602 case ND_REACHABLE:
14603 ASSERT((ncec->ncec_flags & NCE_F_NONUD) == 0);
14604 /* FALLTHROUGH */
14605 case ND_STALE:
14606 /*
14607 * ND_REACHABLE is identical to
14608 * ND_STALE in this specific case. If
14609 * reachable time has expired for this
14610 * neighbor (delta is greater than
14611 * reachable time), conceptually, the
14612 * neighbor cache is no longer in
14613 * REACHABLE state, but already in
14614 * STALE state. So the correct
14615 * transition here is to ND_DELAY.
14616 */
14617 ncec->ncec_state = ND_DELAY;
14618 mutex_exit(&ncec->ncec_lock);
14619 nce_restart_timer(ncec,
14620 ipst->ips_delay_first_probe_time);
14621 if (ip_debug > 3) {
14622 /* ip2dbg */
14623 pr_addr_dbg("ip_xmit: state"
14624 " for %s changed to"
14625 " DELAY\n", AF_INET6,
14626 &ncec->ncec_addr);
14627 }
14628 break;
14629 case ND_DELAY:
14630 case ND_PROBE:
14631 mutex_exit(&ncec->ncec_lock);
14632 /* Timers have already started */
14633 break;
14634 case ND_UNREACHABLE:
14635 /*
14636 * nce_timer has detected that this ncec
14637 * is unreachable and initiated deleting
14638 * this ncec.
14639 * This is a harmless race where we found the
14640 * ncec before it was deleted and have
14641 * just sent out a packet using this
14642 * unreachable ncec.
14643 */
14644 mutex_exit(&ncec->ncec_lock);
14645 break;
14646 default:
14647 ASSERT(0);
14648 mutex_exit(&ncec->ncec_lock);
14649 }
14650 }
14651 return (0);
14652
14653 case ND_INCOMPLETE:
14654 /*
14655 * the state could have changed since we didn't hold the lock.
14656 * Re-verify state under lock.
14657 */
14658 is_probe = ipmp_packet_is_probe(mp, nce->nce_ill);
14659 mutex_enter(&ncec->ncec_lock);
14660 if (NCE_ISREACHABLE(ncec)) {
14661 mutex_exit(&ncec->ncec_lock);
14662 goto sendit;
14663 }
14664 /* queue the packet */
14665 nce_queue_mp(ncec, mp, is_probe);
14666 mutex_exit(&ncec->ncec_lock);
14667 DTRACE_PROBE2(ip__xmit__incomplete,
14668 (ncec_t *), ncec, (mblk_t *), mp);
14669 return (0);
14670
14671 case ND_INITIAL:
14672 /*
14673 * State could have changed since we didn't hold the lock, so
14674 * re-verify state.
14675 */
14676 is_probe = ipmp_packet_is_probe(mp, nce->nce_ill);
14677 mutex_enter(&ncec->ncec_lock);
14678 if (NCE_ISREACHABLE(ncec)) {
14679 mutex_exit(&ncec->ncec_lock);
14680 goto sendit;
14681 }
14682 nce_queue_mp(ncec, mp, is_probe);
14683 if (ncec->ncec_state == ND_INITIAL) {
14684 ncec->ncec_state = ND_INCOMPLETE;
14685 mutex_exit(&ncec->ncec_lock);
14686 /*
14687 * figure out the source we want to use
14688 * and resolve it.
14689 */
14690 ip_ndp_resolve(ncec);
14691 } else {
14692 mutex_exit(&ncec->ncec_lock);
14693 }
14694 return (0);
14695
14696 case ND_UNREACHABLE:
14697 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14698 ip_drop_output("ipIfStatsOutDiscards - ND_UNREACHABLE",
14699 mp, ill);
14700 freemsg(mp);
14701 return (0);
14702
14703 default:
14704 ASSERT(0);
14705 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14706 ip_drop_output("ipIfStatsOutDiscards - ND_other",
14707 mp, ill);
14708 freemsg(mp);
14709 return (ENETUNREACH);
14710 }
14711 }
14712
14713 /*
14714 * Return B_TRUE if the buffers differ in length or content.
14715 * This is used for comparing extension header buffers.
14716 * Note that an extension header would be declared different
14717 * even if all that changed was the next header value in that header i.e.
14718 * what really changed is the next extension header.
14719 */
14720 boolean_t
14721 ip_cmpbuf(const void *abuf, uint_t alen, boolean_t b_valid, const void *bbuf,
14722 uint_t blen)
14723 {
14724 if (!b_valid)
14725 blen = 0;
14726
14727 if (alen != blen)
14728 return (B_TRUE);
14729 if (alen == 0)
14730 return (B_FALSE); /* Both zero length */
14731 return (bcmp(abuf, bbuf, alen));
14732 }
14733
14734 /*
14735 * Preallocate memory for ip_savebuf(). Returns B_TRUE if ok.
14736 * Return B_FALSE if memory allocation fails - don't change any state!
14737 */
14738 boolean_t
14739 ip_allocbuf(void **dstp, uint_t *dstlenp, boolean_t src_valid,
14740 const void *src, uint_t srclen)
14741 {
14742 void *dst;
14743
14744 if (!src_valid)
14745 srclen = 0;
14746
14747 ASSERT(*dstlenp == 0);
14748 if (src != NULL && srclen != 0) {
14749 dst = mi_alloc(srclen, BPRI_MED);
14750 if (dst == NULL)
14751 return (B_FALSE);
14752 } else {
14753 dst = NULL;
14754 }
14755 if (*dstp != NULL)
14756 mi_free(*dstp);
14757 *dstp = dst;
14758 *dstlenp = dst == NULL ? 0 : srclen;
14759 return (B_TRUE);
14760 }
14761
14762 /*
14763 * Replace what is in *dst, *dstlen with the source.
14764 * Assumes ip_allocbuf has already been called.
14765 */
14766 void
14767 ip_savebuf(void **dstp, uint_t *dstlenp, boolean_t src_valid,
14768 const void *src, uint_t srclen)
14769 {
14770 if (!src_valid)
14771 srclen = 0;
14772
14773 ASSERT(*dstlenp == srclen);
14774 if (src != NULL && srclen != 0)
14775 bcopy(src, *dstp, srclen);
14776 }
14777
14778 /*
14779 * Free the storage pointed to by the members of an ip_pkt_t.
14780 */
14781 void
14782 ip_pkt_free(ip_pkt_t *ipp)
14783 {
14784 uint_t fields = ipp->ipp_fields;
14785
14786 if (fields & IPPF_HOPOPTS) {
14787 kmem_free(ipp->ipp_hopopts, ipp->ipp_hopoptslen);
14788 ipp->ipp_hopopts = NULL;
14789 ipp->ipp_hopoptslen = 0;
14790 }
14791 if (fields & IPPF_RTHDRDSTOPTS) {
14792 kmem_free(ipp->ipp_rthdrdstopts, ipp->ipp_rthdrdstoptslen);
14793 ipp->ipp_rthdrdstopts = NULL;
14794 ipp->ipp_rthdrdstoptslen = 0;
14795 }
14796 if (fields & IPPF_DSTOPTS) {
14797 kmem_free(ipp->ipp_dstopts, ipp->ipp_dstoptslen);
14798 ipp->ipp_dstopts = NULL;
14799 ipp->ipp_dstoptslen = 0;
14800 }
14801 if (fields & IPPF_RTHDR) {
14802 kmem_free(ipp->ipp_rthdr, ipp->ipp_rthdrlen);
14803 ipp->ipp_rthdr = NULL;
14804 ipp->ipp_rthdrlen = 0;
14805 }
14806 if (fields & IPPF_IPV4_OPTIONS) {
14807 kmem_free(ipp->ipp_ipv4_options, ipp->ipp_ipv4_options_len);
14808 ipp->ipp_ipv4_options = NULL;
14809 ipp->ipp_ipv4_options_len = 0;
14810 }
14811 if (fields & IPPF_LABEL_V4) {
14812 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
14813 ipp->ipp_label_v4 = NULL;
14814 ipp->ipp_label_len_v4 = 0;
14815 }
14816 if (fields & IPPF_LABEL_V6) {
14817 kmem_free(ipp->ipp_label_v6, ipp->ipp_label_len_v6);
14818 ipp->ipp_label_v6 = NULL;
14819 ipp->ipp_label_len_v6 = 0;
14820 }
14821 ipp->ipp_fields &= ~(IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14822 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6);
14823 }
14824
14825 /*
14826 * Copy from src to dst and allocate as needed.
14827 * Returns zero or ENOMEM.
14828 *
14829 * The caller must initialize dst to zero.
14830 */
14831 int
14832 ip_pkt_copy(ip_pkt_t *src, ip_pkt_t *dst, int kmflag)
14833 {
14834 uint_t fields = src->ipp_fields;
14835
14836 /* Start with fields that don't require memory allocation */
14837 dst->ipp_fields = fields &
14838 ~(IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14839 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6);
14840
14841 dst->ipp_addr = src->ipp_addr;
14842 dst->ipp_unicast_hops = src->ipp_unicast_hops;
14843 dst->ipp_hoplimit = src->ipp_hoplimit;
14844 dst->ipp_tclass = src->ipp_tclass;
14845 dst->ipp_type_of_service = src->ipp_type_of_service;
14846
14847 if (!(fields & (IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14848 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6)))
14849 return (0);
14850
14851 if (fields & IPPF_HOPOPTS) {
14852 dst->ipp_hopopts = kmem_alloc(src->ipp_hopoptslen, kmflag);
14853 if (dst->ipp_hopopts == NULL) {
14854 ip_pkt_free(dst);
14855 return (ENOMEM);
14856 }
14857 dst->ipp_fields |= IPPF_HOPOPTS;
14858 bcopy(src->ipp_hopopts, dst->ipp_hopopts,
14859 src->ipp_hopoptslen);
14860 dst->ipp_hopoptslen = src->ipp_hopoptslen;
14861 }
14862 if (fields & IPPF_RTHDRDSTOPTS) {
14863 dst->ipp_rthdrdstopts = kmem_alloc(src->ipp_rthdrdstoptslen,
14864 kmflag);
14865 if (dst->ipp_rthdrdstopts == NULL) {
14866 ip_pkt_free(dst);
14867 return (ENOMEM);
14868 }
14869 dst->ipp_fields |= IPPF_RTHDRDSTOPTS;
14870 bcopy(src->ipp_rthdrdstopts, dst->ipp_rthdrdstopts,
14871 src->ipp_rthdrdstoptslen);
14872 dst->ipp_rthdrdstoptslen = src->ipp_rthdrdstoptslen;
14873 }
14874 if (fields & IPPF_DSTOPTS) {
14875 dst->ipp_dstopts = kmem_alloc(src->ipp_dstoptslen, kmflag);
14876 if (dst->ipp_dstopts == NULL) {
14877 ip_pkt_free(dst);
14878 return (ENOMEM);
14879 }
14880 dst->ipp_fields |= IPPF_DSTOPTS;
14881 bcopy(src->ipp_dstopts, dst->ipp_dstopts,
14882 src->ipp_dstoptslen);
14883 dst->ipp_dstoptslen = src->ipp_dstoptslen;
14884 }
14885 if (fields & IPPF_RTHDR) {
14886 dst->ipp_rthdr = kmem_alloc(src->ipp_rthdrlen, kmflag);
14887 if (dst->ipp_rthdr == NULL) {
14888 ip_pkt_free(dst);
14889 return (ENOMEM);
14890 }
14891 dst->ipp_fields |= IPPF_RTHDR;
14892 bcopy(src->ipp_rthdr, dst->ipp_rthdr,
14893 src->ipp_rthdrlen);
14894 dst->ipp_rthdrlen = src->ipp_rthdrlen;
14895 }
14896 if (fields & IPPF_IPV4_OPTIONS) {
14897 dst->ipp_ipv4_options = kmem_alloc(src->ipp_ipv4_options_len,
14898 kmflag);
14899 if (dst->ipp_ipv4_options == NULL) {
14900 ip_pkt_free(dst);
14901 return (ENOMEM);
14902 }
14903 dst->ipp_fields |= IPPF_IPV4_OPTIONS;
14904 bcopy(src->ipp_ipv4_options, dst->ipp_ipv4_options,
14905 src->ipp_ipv4_options_len);
14906 dst->ipp_ipv4_options_len = src->ipp_ipv4_options_len;
14907 }
14908 if (fields & IPPF_LABEL_V4) {
14909 dst->ipp_label_v4 = kmem_alloc(src->ipp_label_len_v4, kmflag);
14910 if (dst->ipp_label_v4 == NULL) {
14911 ip_pkt_free(dst);
14912 return (ENOMEM);
14913 }
14914 dst->ipp_fields |= IPPF_LABEL_V4;
14915 bcopy(src->ipp_label_v4, dst->ipp_label_v4,
14916 src->ipp_label_len_v4);
14917 dst->ipp_label_len_v4 = src->ipp_label_len_v4;
14918 }
14919 if (fields & IPPF_LABEL_V6) {
14920 dst->ipp_label_v6 = kmem_alloc(src->ipp_label_len_v6, kmflag);
14921 if (dst->ipp_label_v6 == NULL) {
14922 ip_pkt_free(dst);
14923 return (ENOMEM);
14924 }
14925 dst->ipp_fields |= IPPF_LABEL_V6;
14926 bcopy(src->ipp_label_v6, dst->ipp_label_v6,
14927 src->ipp_label_len_v6);
14928 dst->ipp_label_len_v6 = src->ipp_label_len_v6;
14929 }
14930 if (fields & IPPF_FRAGHDR) {
14931 dst->ipp_fraghdr = kmem_alloc(src->ipp_fraghdrlen, kmflag);
14932 if (dst->ipp_fraghdr == NULL) {
14933 ip_pkt_free(dst);
14934 return (ENOMEM);
14935 }
14936 dst->ipp_fields |= IPPF_FRAGHDR;
14937 bcopy(src->ipp_fraghdr, dst->ipp_fraghdr,
14938 src->ipp_fraghdrlen);
14939 dst->ipp_fraghdrlen = src->ipp_fraghdrlen;
14940 }
14941 return (0);
14942 }
14943
14944 /*
14945 * Returns INADDR_ANY if no source route
14946 */
14947 ipaddr_t
14948 ip_pkt_source_route_v4(const ip_pkt_t *ipp)
14949 {
14950 ipaddr_t nexthop = INADDR_ANY;
14951 ipoptp_t opts;
14952 uchar_t *opt;
14953 uint8_t optval;
14954 uint8_t optlen;
14955 uint32_t totallen;
14956
14957 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
14958 return (INADDR_ANY);
14959
14960 totallen = ipp->ipp_ipv4_options_len;
14961 if (totallen & 0x3)
14962 return (INADDR_ANY);
14963
14964 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
14965 optval != IPOPT_EOL;
14966 optval = ipoptp_next(&opts)) {
14967 opt = opts.ipoptp_cur;
14968 switch (optval) {
14969 uint8_t off;
14970 case IPOPT_SSRR:
14971 case IPOPT_LSRR:
14972 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
14973 break;
14974 }
14975 optlen = opts.ipoptp_len;
14976 off = opt[IPOPT_OFFSET];
14977 off--;
14978 if (optlen < IP_ADDR_LEN ||
14979 off > optlen - IP_ADDR_LEN) {
14980 /* End of source route */
14981 break;
14982 }
14983 bcopy((char *)opt + off, &nexthop, IP_ADDR_LEN);
14984 if (nexthop == htonl(INADDR_LOOPBACK)) {
14985 /* Ignore */
14986 nexthop = INADDR_ANY;
14987 break;
14988 }
14989 break;
14990 }
14991 }
14992 return (nexthop);
14993 }
14994
14995 /*
14996 * Reverse a source route.
14997 */
14998 void
14999 ip_pkt_source_route_reverse_v4(ip_pkt_t *ipp)
15000 {
15001 ipaddr_t tmp;
15002 ipoptp_t opts;
15003 uchar_t *opt;
15004 uint8_t optval;
15005 uint32_t totallen;
15006
15007 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
15008 return;
15009
15010 totallen = ipp->ipp_ipv4_options_len;
15011 if (totallen & 0x3)
15012 return;
15013
15014 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
15015 optval != IPOPT_EOL;
15016 optval = ipoptp_next(&opts)) {
15017 uint8_t off1, off2;
15018
15019 opt = opts.ipoptp_cur;
15020 switch (optval) {
15021 case IPOPT_SSRR:
15022 case IPOPT_LSRR:
15023 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
15024 break;
15025 }
15026 off1 = IPOPT_MINOFF_SR - 1;
15027 off2 = opt[IPOPT_OFFSET] - IP_ADDR_LEN - 1;
15028 while (off2 > off1) {
15029 bcopy(opt + off2, &tmp, IP_ADDR_LEN);
15030 bcopy(opt + off1, opt + off2, IP_ADDR_LEN);
15031 bcopy(&tmp, opt + off2, IP_ADDR_LEN);
15032 off2 -= IP_ADDR_LEN;
15033 off1 += IP_ADDR_LEN;
15034 }
15035 opt[IPOPT_OFFSET] = IPOPT_MINOFF_SR;
15036 break;
15037 }
15038 }
15039 }
15040
15041 /*
15042 * Returns NULL if no routing header
15043 */
15044 in6_addr_t *
15045 ip_pkt_source_route_v6(const ip_pkt_t *ipp)
15046 {
15047 in6_addr_t *nexthop = NULL;
15048 ip6_rthdr0_t *rthdr;
15049
15050 if (!(ipp->ipp_fields & IPPF_RTHDR))
15051 return (NULL);
15052
15053 rthdr = (ip6_rthdr0_t *)ipp->ipp_rthdr;
15054 if (rthdr->ip6r0_segleft == 0)
15055 return (NULL);
15056
15057 nexthop = (in6_addr_t *)((char *)rthdr + sizeof (*rthdr));
15058 return (nexthop);
15059 }
15060
15061 zoneid_t
15062 ip_get_zoneid_v4(ipaddr_t addr, mblk_t *mp, ip_recv_attr_t *ira,
15063 zoneid_t lookup_zoneid)
15064 {
15065 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
15066 ire_t *ire;
15067 int ire_flags = MATCH_IRE_TYPE;
15068 zoneid_t zoneid = ALL_ZONES;
15069
15070 if (is_system_labeled() && !tsol_can_accept_raw(mp, ira, B_FALSE))
15071 return (ALL_ZONES);
15072
15073 if (lookup_zoneid != ALL_ZONES)
15074 ire_flags |= MATCH_IRE_ZONEONLY;
15075 ire = ire_ftable_lookup_v4(addr, NULL, NULL, IRE_LOCAL | IRE_LOOPBACK,
15076 NULL, lookup_zoneid, NULL, ire_flags, 0, ipst, NULL);
15077 if (ire != NULL) {
15078 zoneid = IP_REAL_ZONEID(ire->ire_zoneid, ipst);
15079 ire_refrele(ire);
15080 }
15081 return (zoneid);
15082 }
15083
15084 zoneid_t
15085 ip_get_zoneid_v6(in6_addr_t *addr, mblk_t *mp, const ill_t *ill,
15086 ip_recv_attr_t *ira, zoneid_t lookup_zoneid)
15087 {
15088 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
15089 ire_t *ire;
15090 int ire_flags = MATCH_IRE_TYPE;
15091 zoneid_t zoneid = ALL_ZONES;
15092
15093 if (is_system_labeled() && !tsol_can_accept_raw(mp, ira, B_FALSE))
15094 return (ALL_ZONES);
15095
15096 if (IN6_IS_ADDR_LINKLOCAL(addr))
15097 ire_flags |= MATCH_IRE_ILL;
15098
15099 if (lookup_zoneid != ALL_ZONES)
15100 ire_flags |= MATCH_IRE_ZONEONLY;
15101 ire = ire_ftable_lookup_v6(addr, NULL, NULL, IRE_LOCAL | IRE_LOOPBACK,
15102 ill, lookup_zoneid, NULL, ire_flags, 0, ipst, NULL);
15103 if (ire != NULL) {
15104 zoneid = IP_REAL_ZONEID(ire->ire_zoneid, ipst);
15105 ire_refrele(ire);
15106 }
15107 return (zoneid);
15108 }
15109
15110 /*
15111 * IP obserability hook support functions.
15112 */
15113 static void
15114 ipobs_init(ip_stack_t *ipst)
15115 {
15116 netid_t id;
15117
15118 id = net_getnetidbynetstackid(ipst->ips_netstack->netstack_stackid);
15119
15120 ipst->ips_ip4_observe_pr = net_protocol_lookup(id, NHF_INET);
15121 VERIFY(ipst->ips_ip4_observe_pr != NULL);
15122
15123 ipst->ips_ip6_observe_pr = net_protocol_lookup(id, NHF_INET6);
15124 VERIFY(ipst->ips_ip6_observe_pr != NULL);
15125 }
15126
15127 static void
15128 ipobs_fini(ip_stack_t *ipst)
15129 {
15130
15131 VERIFY(net_protocol_release(ipst->ips_ip4_observe_pr) == 0);
15132 VERIFY(net_protocol_release(ipst->ips_ip6_observe_pr) == 0);
15133 }
15134
15135 /*
15136 * hook_pkt_observe_t is composed in network byte order so that the
15137 * entire mblk_t chain handed into hook_run can be used as-is.
15138 * The caveat is that use of the fields, such as the zone fields,
15139 * requires conversion into host byte order first.
15140 */
15141 void
15142 ipobs_hook(mblk_t *mp, int htype, zoneid_t zsrc, zoneid_t zdst,
15143 const ill_t *ill, ip_stack_t *ipst)
15144 {
15145 hook_pkt_observe_t *hdr;
15146 uint64_t grifindex;
15147 mblk_t *imp;
15148
15149 imp = allocb(sizeof (*hdr), BPRI_HI);
15150 if (imp == NULL)
15151 return;
15152
15153 hdr = (hook_pkt_observe_t *)imp->b_rptr;
15154 /*
15155 * b_wptr is set to make the apparent size of the data in the mblk_t
15156 * to exclude the pointers at the end of hook_pkt_observer_t.
15157 */
15158 imp->b_wptr = imp->b_rptr + sizeof (dl_ipnetinfo_t);
15159 imp->b_cont = mp;
15160
15161 ASSERT(DB_TYPE(mp) == M_DATA);
15162
15163 if (IS_UNDER_IPMP(ill))
15164 grifindex = ipmp_ill_get_ipmp_ifindex(ill);
15165 else
15166 grifindex = 0;
15167
15168 hdr->hpo_version = 1;
15169 hdr->hpo_htype = htons(htype);
15170 hdr->hpo_pktlen = htonl((ulong_t)msgdsize(mp));
15171 hdr->hpo_ifindex = htonl(ill->ill_phyint->phyint_ifindex);
15172 hdr->hpo_grifindex = htonl(grifindex);
15173 hdr->hpo_zsrc = htonl(zsrc);
15174 hdr->hpo_zdst = htonl(zdst);
15175 hdr->hpo_pkt = imp;
15176 hdr->hpo_ctx = ipst->ips_netstack;
15177
15178 if (ill->ill_isv6) {
15179 hdr->hpo_family = AF_INET6;
15180 (void) hook_run(ipst->ips_ipv6_net_data->netd_hooks,
15181 ipst->ips_ipv6observing, (hook_data_t)hdr);
15182 } else {
15183 hdr->hpo_family = AF_INET;
15184 (void) hook_run(ipst->ips_ipv4_net_data->netd_hooks,
15185 ipst->ips_ipv4observing, (hook_data_t)hdr);
15186 }
15187
15188 imp->b_cont = NULL;
15189 freemsg(imp);
15190 }
15191
15192 /*
15193 * Utility routine that checks if `v4srcp' is a valid address on underlying
15194 * interface `ill'. If `ipifp' is non-NULL, it's set to a held ipif
15195 * associated with `v4srcp' on success. NOTE: if this is not called from
15196 * inside the IPSQ (ill_g_lock is not held), `ill' may be removed from the
15197 * group during or after this lookup.
15198 */
15199 boolean_t
15200 ipif_lookup_testaddr_v4(ill_t *ill, const in_addr_t *v4srcp, ipif_t **ipifp)
15201 {
15202 ipif_t *ipif;
15203
15204 ipif = ipif_lookup_addr_exact(*v4srcp, ill, ill->ill_ipst);
15205 if (ipif != NULL) {
15206 if (ipifp != NULL)
15207 *ipifp = ipif;
15208 else
15209 ipif_refrele(ipif);
15210 return (B_TRUE);
15211 }
15212
15213 ip1dbg(("ipif_lookup_testaddr_v4: cannot find ipif for src %x\n",
15214 *v4srcp));
15215 return (B_FALSE);
15216 }
15217
15218 /*
15219 * Transport protocol call back function for CPU state change.
15220 */
15221 /* ARGSUSED */
15222 static int
15223 ip_tp_cpu_update(cpu_setup_t what, int id, void *arg)
15224 {
15225 processorid_t cpu_seqid;
15226 netstack_handle_t nh;
15227 netstack_t *ns;
15228
15229 ASSERT(MUTEX_HELD(&cpu_lock));
15230
15231 switch (what) {
15232 case CPU_CONFIG:
15233 case CPU_ON:
15234 case CPU_INIT:
15235 case CPU_CPUPART_IN:
15236 cpu_seqid = cpu[id]->cpu_seqid;
15237 netstack_next_init(&nh);
15238 while ((ns = netstack_next(&nh)) != NULL) {
15239 tcp_stack_cpu_add(ns->netstack_tcp, cpu_seqid);
15240 sctp_stack_cpu_add(ns->netstack_sctp, cpu_seqid);
15241 udp_stack_cpu_add(ns->netstack_udp, cpu_seqid);
15242 netstack_rele(ns);
15243 }
15244 netstack_next_fini(&nh);
15245 break;
15246 case CPU_UNCONFIG:
15247 case CPU_OFF:
15248 case CPU_CPUPART_OUT:
15249 /*
15250 * Nothing to do. We don't remove the per CPU stats from
15251 * the IP stack even when the CPU goes offline.
15252 */
15253 break;
15254 default:
15255 break;
15256 }
15257 return (0);
15258 }