1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21
22 /*
23 * Copyright (c) 1991, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 1990 Mentat Inc.
25 * Copyright (c) 2012 Joyent, Inc. All rights reserved.
26 * Copyright (c) 2017 OmniTI Computer Consulting, Inc. All rights reserved.
27 */
28
29 #include <sys/types.h>
30 #include <sys/stream.h>
31 #include <sys/dlpi.h>
32 #include <sys/stropts.h>
33 #include <sys/sysmacros.h>
34 #include <sys/strsubr.h>
35 #include <sys/strlog.h>
36 #include <sys/strsun.h>
37 #include <sys/zone.h>
38 #define _SUN_TPI_VERSION 2
39 #include <sys/tihdr.h>
40 #include <sys/xti_inet.h>
41 #include <sys/ddi.h>
42 #include <sys/suntpi.h>
43 #include <sys/cmn_err.h>
44 #include <sys/debug.h>
45 #include <sys/kobj.h>
46 #include <sys/modctl.h>
47 #include <sys/atomic.h>
48 #include <sys/policy.h>
49 #include <sys/priv.h>
50 #include <sys/taskq.h>
51
52 #include <sys/systm.h>
53 #include <sys/param.h>
54 #include <sys/kmem.h>
55 #include <sys/sdt.h>
56 #include <sys/socket.h>
57 #include <sys/vtrace.h>
58 #include <sys/isa_defs.h>
59 #include <sys/mac.h>
60 #include <net/if.h>
61 #include <net/if_arp.h>
62 #include <net/route.h>
63 #include <sys/sockio.h>
64 #include <netinet/in.h>
65 #include <net/if_dl.h>
66
67 #include <inet/common.h>
68 #include <inet/mi.h>
69 #include <inet/mib2.h>
70 #include <inet/nd.h>
71 #include <inet/arp.h>
72 #include <inet/snmpcom.h>
73 #include <inet/optcom.h>
74 #include <inet/kstatcom.h>
75
76 #include <netinet/igmp_var.h>
77 #include <netinet/ip6.h>
78 #include <netinet/icmp6.h>
79 #include <netinet/sctp.h>
80
81 #include <inet/ip.h>
82 #include <inet/ip_impl.h>
83 #include <inet/ip6.h>
84 #include <inet/ip6_asp.h>
85 #include <inet/tcp.h>
86 #include <inet/tcp_impl.h>
87 #include <inet/ip_multi.h>
88 #include <inet/ip_if.h>
89 #include <inet/ip_ire.h>
90 #include <inet/ip_ftable.h>
91 #include <inet/ip_rts.h>
92 #include <inet/ip_ndp.h>
93 #include <inet/ip_listutils.h>
94 #include <netinet/igmp.h>
95 #include <netinet/ip_mroute.h>
96 #include <inet/ipp_common.h>
97
98 #include <net/pfkeyv2.h>
99 #include <inet/sadb.h>
100 #include <inet/ipsec_impl.h>
101 #include <inet/iptun/iptun_impl.h>
102 #include <inet/ipdrop.h>
103 #include <inet/ip_netinfo.h>
104 #include <inet/ilb_ip.h>
105
106 #include <sys/ethernet.h>
107 #include <net/if_types.h>
108 #include <sys/cpuvar.h>
109
110 #include <ipp/ipp.h>
111 #include <ipp/ipp_impl.h>
112 #include <ipp/ipgpc/ipgpc.h>
113
114 #include <sys/pattr.h>
115 #include <inet/ipclassifier.h>
116 #include <inet/sctp_ip.h>
117 #include <inet/sctp/sctp_impl.h>
118 #include <inet/udp_impl.h>
119 #include <inet/rawip_impl.h>
120 #include <inet/rts_impl.h>
121
122 #include <sys/tsol/label.h>
123 #include <sys/tsol/tnet.h>
124
125 #include <sys/squeue_impl.h>
126 #include <inet/ip_arp.h>
127
128 #include <sys/clock_impl.h> /* For LBOLT_FASTPATH{,64} */
129
130 /*
131 * Values for squeue switch:
132 * IP_SQUEUE_ENTER_NODRAIN: SQ_NODRAIN
133 * IP_SQUEUE_ENTER: SQ_PROCESS
134 * IP_SQUEUE_FILL: SQ_FILL
135 */
136 int ip_squeue_enter = IP_SQUEUE_ENTER; /* Setable in /etc/system */
137
138 int ip_squeue_flag;
139
140 /*
141 * Setable in /etc/system
142 */
143 int ip_poll_normal_ms = 100;
144 int ip_poll_normal_ticks = 0;
145 int ip_modclose_ackwait_ms = 3000;
146
147 /*
148 * It would be nice to have these present only in DEBUG systems, but the
149 * current design of the global symbol checking logic requires them to be
150 * unconditionally present.
151 */
152 uint_t ip_thread_data; /* TSD key for debug support */
153 krwlock_t ip_thread_rwlock;
154 list_t ip_thread_list;
155
156 /*
157 * Structure to represent a linked list of msgblks. Used by ip_snmp_ functions.
158 */
159
160 struct listptr_s {
161 mblk_t *lp_head; /* pointer to the head of the list */
162 mblk_t *lp_tail; /* pointer to the tail of the list */
163 };
164
165 typedef struct listptr_s listptr_t;
166
167 /*
168 * This is used by ip_snmp_get_mib2_ip_route_media and
169 * ip_snmp_get_mib2_ip6_route_media to carry the lists of return data.
170 */
171 typedef struct iproutedata_s {
172 uint_t ird_idx;
173 uint_t ird_flags; /* see below */
174 listptr_t ird_route; /* ipRouteEntryTable */
175 listptr_t ird_netmedia; /* ipNetToMediaEntryTable */
176 listptr_t ird_attrs; /* ipRouteAttributeTable */
177 } iproutedata_t;
178
179 /* Include ire_testhidden and IRE_IF_CLONE routes */
180 #define IRD_REPORT_ALL 0x01
181
182 /*
183 * Cluster specific hooks. These should be NULL when booted as a non-cluster
184 */
185
186 /*
187 * Hook functions to enable cluster networking
188 * On non-clustered systems these vectors must always be NULL.
189 *
190 * Hook function to Check ip specified ip address is a shared ip address
191 * in the cluster
192 *
193 */
194 int (*cl_inet_isclusterwide)(netstackid_t stack_id, uint8_t protocol,
195 sa_family_t addr_family, uint8_t *laddrp, void *args) = NULL;
196
197 /*
198 * Hook function to generate cluster wide ip fragment identifier
199 */
200 uint32_t (*cl_inet_ipident)(netstackid_t stack_id, uint8_t protocol,
201 sa_family_t addr_family, uint8_t *laddrp, uint8_t *faddrp,
202 void *args) = NULL;
203
204 /*
205 * Hook function to generate cluster wide SPI.
206 */
207 void (*cl_inet_getspi)(netstackid_t, uint8_t, uint8_t *, size_t,
208 void *) = NULL;
209
210 /*
211 * Hook function to verify if the SPI is already utlized.
212 */
213
214 int (*cl_inet_checkspi)(netstackid_t, uint8_t, uint32_t, void *) = NULL;
215
216 /*
217 * Hook function to delete the SPI from the cluster wide repository.
218 */
219
220 void (*cl_inet_deletespi)(netstackid_t, uint8_t, uint32_t, void *) = NULL;
221
222 /*
223 * Hook function to inform the cluster when packet received on an IDLE SA
224 */
225
226 void (*cl_inet_idlesa)(netstackid_t, uint8_t, uint32_t, sa_family_t,
227 in6_addr_t, in6_addr_t, void *) = NULL;
228
229 /*
230 * Synchronization notes:
231 *
232 * IP is a fully D_MP STREAMS module/driver. Thus it does not depend on any
233 * MT level protection given by STREAMS. IP uses a combination of its own
234 * internal serialization mechanism and standard Solaris locking techniques.
235 * The internal serialization is per phyint. This is used to serialize
236 * plumbing operations, IPMP operations, most set ioctls, etc.
237 *
238 * Plumbing is a long sequence of operations involving message
239 * exchanges between IP, ARP and device drivers. Many set ioctls are typically
240 * involved in plumbing operations. A natural model is to serialize these
241 * ioctls one per ill. For example plumbing of hme0 and qfe0 can go on in
242 * parallel without any interference. But various set ioctls on hme0 are best
243 * serialized, along with IPMP operations and processing of DLPI control
244 * messages received from drivers on a per phyint basis. This serialization is
245 * provided by the ipsq_t and primitives operating on this. Details can
246 * be found in ip_if.c above the core primitives operating on ipsq_t.
247 *
248 * Lookups of an ipif or ill by a thread return a refheld ipif / ill.
249 * Simiarly lookup of an ire by a thread also returns a refheld ire.
250 * In addition ipif's and ill's referenced by the ire are also indirectly
251 * refheld. Thus no ipif or ill can vanish as long as an ipif is refheld
252 * directly or indirectly. For example an SIOCSLIFADDR ioctl that changes the
253 * address of an ipif has to go through the ipsq_t. This ensures that only
254 * one such exclusive operation proceeds at any time on the ipif. It then
255 * waits for all refcnts
256 * associated with this ipif to come down to zero. The address is changed
257 * only after the ipif has been quiesced. Then the ipif is brought up again.
258 * More details are described above the comment in ip_sioctl_flags.
259 *
260 * Packet processing is based mostly on IREs and are fully multi-threaded
261 * using standard Solaris MT techniques.
262 *
263 * There are explicit locks in IP to handle:
264 * - The ip_g_head list maintained by mi_open_link() and friends.
265 *
266 * - The reassembly data structures (one lock per hash bucket)
267 *
268 * - conn_lock is meant to protect conn_t fields. The fields actually
269 * protected by conn_lock are documented in the conn_t definition.
270 *
271 * - ire_lock to protect some of the fields of the ire, IRE tables
272 * (one lock per hash bucket). Refer to ip_ire.c for details.
273 *
274 * - ndp_g_lock and ncec_lock for protecting NCEs.
275 *
276 * - ill_lock protects fields of the ill and ipif. Details in ip.h
277 *
278 * - ill_g_lock: This is a global reader/writer lock. Protects the following
279 * * The AVL tree based global multi list of all ills.
280 * * The linked list of all ipifs of an ill
281 * * The <ipsq-xop> mapping
282 * * <ill-phyint> association
283 * Insertion/deletion of an ill in the system, insertion/deletion of an ipif
284 * into an ill, changing the <ipsq-xop> mapping of an ill, changing the
285 * <ill-phyint> assoc of an ill will all have to hold the ill_g_lock as
286 * writer for the actual duration of the insertion/deletion/change.
287 *
288 * - ill_lock: This is a per ill mutex.
289 * It protects some members of the ill_t struct; see ip.h for details.
290 * It also protects the <ill-phyint> assoc.
291 * It also protects the list of ipifs hanging off the ill.
292 *
293 * - ipsq_lock: This is a per ipsq_t mutex lock.
294 * This protects some members of the ipsq_t struct; see ip.h for details.
295 * It also protects the <ipsq-ipxop> mapping
296 *
297 * - ipx_lock: This is a per ipxop_t mutex lock.
298 * This protects some members of the ipxop_t struct; see ip.h for details.
299 *
300 * - phyint_lock: This is a per phyint mutex lock. Protects just the
301 * phyint_flags
302 *
303 * - ip_addr_avail_lock: This is used to ensure the uniqueness of IP addresses.
304 * This lock is held in ipif_up_done and the ipif is marked IPIF_UP and the
305 * uniqueness check also done atomically.
306 *
307 * - ill_g_usesrc_lock: This readers/writer lock protects the usesrc
308 * group list linked by ill_usesrc_grp_next. It also protects the
309 * ill_usesrc_ifindex field. It is taken as a writer when a member of the
310 * group is being added or deleted. This lock is taken as a reader when
311 * walking the list/group(eg: to get the number of members in a usesrc group).
312 * Note, it is only necessary to take this lock if the ill_usesrc_grp_next
313 * field is changing state i.e from NULL to non-NULL or vice-versa. For
314 * example, it is not necessary to take this lock in the initial portion
315 * of ip_sioctl_slifusesrc or at all in ip_sioctl_flags since these
316 * operations are executed exclusively and that ensures that the "usesrc
317 * group state" cannot change. The "usesrc group state" change can happen
318 * only in the latter part of ip_sioctl_slifusesrc and in ill_delete.
319 *
320 * Changing <ill-phyint>, <ipsq-xop> assocications:
321 *
322 * To change the <ill-phyint> association, the ill_g_lock must be held
323 * as writer, and the ill_locks of both the v4 and v6 instance of the ill
324 * must be held.
325 *
326 * To change the <ipsq-xop> association, the ill_g_lock must be held as
327 * writer, the ipsq_lock must be held, and one must be writer on the ipsq.
328 * This is only done when ills are added or removed from IPMP groups.
329 *
330 * To add or delete an ipif from the list of ipifs hanging off the ill,
331 * ill_g_lock (writer) and ill_lock must be held and the thread must be
332 * a writer on the associated ipsq.
333 *
334 * To add or delete an ill to the system, the ill_g_lock must be held as
335 * writer and the thread must be a writer on the associated ipsq.
336 *
337 * To add or delete an ilm to an ill, the ill_lock must be held and the thread
338 * must be a writer on the associated ipsq.
339 *
340 * Lock hierarchy
341 *
342 * Some lock hierarchy scenarios are listed below.
343 *
344 * ill_g_lock -> conn_lock -> ill_lock -> ipsq_lock -> ipx_lock
345 * ill_g_lock -> ill_lock(s) -> phyint_lock
346 * ill_g_lock -> ndp_g_lock -> ill_lock -> ncec_lock
347 * ill_g_lock -> ip_addr_avail_lock
348 * conn_lock -> irb_lock -> ill_lock -> ire_lock
349 * ill_g_lock -> ip_g_nd_lock
350 * ill_g_lock -> ips_ipmp_lock -> ill_lock -> nce_lock
351 * ill_g_lock -> ndp_g_lock -> ill_lock -> ncec_lock -> nce_lock
352 * arl_lock -> ill_lock
353 * ips_ire_dep_lock -> irb_lock
354 *
355 * When more than 1 ill lock is needed to be held, all ill lock addresses
356 * are sorted on address and locked starting from highest addressed lock
357 * downward.
358 *
359 * Multicast scenarios
360 * ips_ill_g_lock -> ill_mcast_lock
361 * conn_ilg_lock -> ips_ill_g_lock -> ill_lock
362 * ill_mcast_serializer -> ill_mcast_lock -> ips_ipmp_lock -> ill_lock
363 * ill_mcast_serializer -> ill_mcast_lock -> connf_lock -> conn_lock
364 * ill_mcast_serializer -> ill_mcast_lock -> conn_ilg_lock
365 * ill_mcast_serializer -> ill_mcast_lock -> ips_igmp_timer_lock
366 *
367 * IPsec scenarios
368 *
369 * ipsa_lock -> ill_g_lock -> ill_lock
370 * ill_g_usesrc_lock -> ill_g_lock -> ill_lock
371 *
372 * Trusted Solaris scenarios
373 *
374 * igsa_lock -> gcgrp_rwlock -> gcgrp_lock
375 * igsa_lock -> gcdb_lock
376 * gcgrp_rwlock -> ire_lock
377 * gcgrp_rwlock -> gcdb_lock
378 *
379 * squeue(sq_lock), flow related (ft_lock, fe_lock) locking
380 *
381 * cpu_lock --> ill_lock --> sqset_lock --> sq_lock
382 * sq_lock -> conn_lock -> QLOCK(q)
383 * ill_lock -> ft_lock -> fe_lock
384 *
385 * Routing/forwarding table locking notes:
386 *
387 * Lock acquisition order: Radix tree lock, irb_lock.
388 * Requirements:
389 * i. Walker must not hold any locks during the walker callback.
390 * ii Walker must not see a truncated tree during the walk because of any node
391 * deletion.
392 * iii Existing code assumes ire_bucket is valid if it is non-null and is used
393 * in many places in the code to walk the irb list. Thus even if all the
394 * ires in a bucket have been deleted, we still can't free the radix node
395 * until the ires have actually been inactive'd (freed).
396 *
397 * Tree traversal - Need to hold the global tree lock in read mode.
398 * Before dropping the global tree lock, need to either increment the ire_refcnt
399 * to ensure that the radix node can't be deleted.
400 *
401 * Tree add - Need to hold the global tree lock in write mode to add a
402 * radix node. To prevent the node from being deleted, increment the
403 * irb_refcnt, after the node is added to the tree. The ire itself is
404 * added later while holding the irb_lock, but not the tree lock.
405 *
406 * Tree delete - Need to hold the global tree lock and irb_lock in write mode.
407 * All associated ires must be inactive (i.e. freed), and irb_refcnt
408 * must be zero.
409 *
410 * Walker - Increment irb_refcnt before calling the walker callback. Hold the
411 * global tree lock (read mode) for traversal.
412 *
413 * IRE dependencies - In some cases we hold ips_ire_dep_lock across ire_refrele
414 * hence we will acquire irb_lock while holding ips_ire_dep_lock.
415 *
416 * IPsec notes :
417 *
418 * IP interacts with the IPsec code (AH/ESP) by storing IPsec attributes
419 * in the ip_xmit_attr_t ip_recv_attr_t. For outbound datagrams, the
420 * ip_xmit_attr_t has the
421 * information used by the IPsec code for applying the right level of
422 * protection. The information initialized by IP in the ip_xmit_attr_t
423 * is determined by the per-socket policy or global policy in the system.
424 * For inbound datagrams, the ip_recv_attr_t
425 * starts out with nothing in it. It gets filled
426 * with the right information if it goes through the AH/ESP code, which
427 * happens if the incoming packet is secure. The information initialized
428 * by AH/ESP, is later used by IP (during fanouts to ULP) to see whether
429 * the policy requirements needed by per-socket policy or global policy
430 * is met or not.
431 *
432 * For fully connected sockets i.e dst, src [addr, port] is known,
433 * conn_policy_cached is set indicating that policy has been cached.
434 * conn_in_enforce_policy may or may not be set depending on whether
435 * there is a global policy match or per-socket policy match.
436 * Policy inheriting happpens in ip_policy_set once the destination is known.
437 * Once the right policy is set on the conn_t, policy cannot change for
438 * this socket. This makes life simpler for TCP (UDP ?) where
439 * re-transmissions go out with the same policy. For symmetry, policy
440 * is cached for fully connected UDP sockets also. Thus if policy is cached,
441 * it also implies that policy is latched i.e policy cannot change
442 * on these sockets. As we have the right policy on the conn, we don't
443 * have to lookup global policy for every outbound and inbound datagram
444 * and thus serving as an optimization. Note that a global policy change
445 * does not affect fully connected sockets if they have policy. If fully
446 * connected sockets did not have any policy associated with it, global
447 * policy change may affect them.
448 *
449 * IP Flow control notes:
450 * ---------------------
451 * Non-TCP streams are flow controlled by IP. The way this is accomplished
452 * differs when ILL_CAPAB_DLD_DIRECT is enabled for that IP instance. When
453 * ILL_DIRECT_CAPABLE(ill) is TRUE, IP can do direct function calls into
454 * GLDv3. Otherwise packets are sent down to lower layers using STREAMS
455 * functions.
456 *
457 * Per Tx ring udp flow control:
458 * This is applicable only when ILL_CAPAB_DLD_DIRECT capability is set in
459 * the ill (i.e. ILL_DIRECT_CAPABLE(ill) is true).
460 *
461 * The underlying link can expose multiple Tx rings to the GLDv3 mac layer.
462 * To achieve best performance, outgoing traffic need to be fanned out among
463 * these Tx ring. mac_tx() is called (via str_mdata_fastpath_put()) to send
464 * traffic out of the NIC and it takes a fanout hint. UDP connections pass
465 * the address of connp as fanout hint to mac_tx(). Under flow controlled
466 * condition, mac_tx() returns a non-NULL cookie (ip_mac_tx_cookie_t). This
467 * cookie points to a specific Tx ring that is blocked. The cookie is used to
468 * hash into an idl_tx_list[] entry in idl_tx_list[] array. Each idl_tx_list_t
469 * point to drain_lists (idl_t's). These drain list will store the blocked UDP
470 * connp's. The drain list is not a single list but a configurable number of
471 * lists.
472 *
473 * The diagram below shows idl_tx_list_t's and their drain_lists. ip_stack_t
474 * has an array of idl_tx_list_t. The size of the array is TX_FANOUT_SIZE
475 * which is equal to 128. This array in turn contains a pointer to idl_t[],
476 * the ip drain list. The idl_t[] array size is MIN(max_ncpus, 8). The drain
477 * list will point to the list of connp's that are flow controlled.
478 *
479 * --------------- ------- ------- -------
480 * |->|drain_list[0]|-->|connp|-->|connp|-->|connp|-->
481 * | --------------- ------- ------- -------
482 * | --------------- ------- ------- -------
483 * |->|drain_list[1]|-->|connp|-->|connp|-->|connp|-->
484 * ---------------- | --------------- ------- ------- -------
485 * |idl_tx_list[0]|->| --------------- ------- ------- -------
486 * ---------------- |->|drain_list[2]|-->|connp|-->|connp|-->|connp|-->
487 * | --------------- ------- ------- -------
488 * . . . . .
489 * | --------------- ------- ------- -------
490 * |->|drain_list[n]|-->|connp|-->|connp|-->|connp|-->
491 * --------------- ------- ------- -------
492 * --------------- ------- ------- -------
493 * |->|drain_list[0]|-->|connp|-->|connp|-->|connp|-->
494 * | --------------- ------- ------- -------
495 * | --------------- ------- ------- -------
496 * ---------------- |->|drain_list[1]|-->|connp|-->|connp|-->|connp|-->
497 * |idl_tx_list[1]|->| --------------- ------- ------- -------
498 * ---------------- | . . . .
499 * | --------------- ------- ------- -------
500 * |->|drain_list[n]|-->|connp|-->|connp|-->|connp|-->
501 * --------------- ------- ------- -------
502 * .....
503 * ----------------
504 * |idl_tx_list[n]|-> ...
505 * ----------------
506 *
507 * When mac_tx() returns a cookie, the cookie is hashed into an index into
508 * ips_idl_tx_list[], and conn_drain_insert() is called with the idl_tx_list
509 * to insert the conn onto. conn_drain_insert() asserts flow control for the
510 * sockets via su_txq_full() (non-STREAMS) or QFULL on conn_wq (STREAMS).
511 * Further, conn_blocked is set to indicate that the conn is blocked.
512 *
513 * GLDv3 calls ill_flow_enable() when flow control is relieved. The cookie
514 * passed in the call to ill_flow_enable() identifies the blocked Tx ring and
515 * is again hashed to locate the appropriate idl_tx_list, which is then
516 * drained via conn_walk_drain(). conn_walk_drain() goes through each conn in
517 * the drain list and calls conn_drain_remove() to clear flow control (via
518 * calling su_txq_full() or clearing QFULL), and remove the conn from the
519 * drain list.
520 *
521 * Note that the drain list is not a single list but a (configurable) array of
522 * lists (8 elements by default). Synchronization between drain insertion and
523 * flow control wakeup is handled by using idl_txl->txl_lock, and only
524 * conn_drain_insert() and conn_drain_remove() manipulate the drain list.
525 *
526 * Flow control via STREAMS is used when ILL_DIRECT_CAPABLE() returns FALSE.
527 * On the send side, if the packet cannot be sent down to the driver by IP
528 * (canput() fails), ip_xmit() drops the packet and returns EWOULDBLOCK to the
529 * caller, who may then invoke ixa_check_drain_insert() to insert the conn on
530 * the 0'th drain list. When ip_wsrv() runs on the ill_wq because flow
531 * control has been relieved, the blocked conns in the 0'th drain list are
532 * drained as in the non-STREAMS case.
533 *
534 * In both the STREAMS and non-STREAMS cases, the sockfs upcall to set QFULL
535 * is done when the conn is inserted into the drain list (conn_drain_insert())
536 * and cleared when the conn is removed from the it (conn_drain_remove()).
537 *
538 * IPQOS notes:
539 *
540 * IPQoS Policies are applied to packets using IPPF (IP Policy framework)
541 * and IPQoS modules. IPPF includes hooks in IP at different control points
542 * (callout positions) which direct packets to IPQoS modules for policy
543 * processing. Policies, if present, are global.
544 *
545 * The callout positions are located in the following paths:
546 * o local_in (packets destined for this host)
547 * o local_out (packets orginating from this host )
548 * o fwd_in (packets forwarded by this m/c - inbound)
549 * o fwd_out (packets forwarded by this m/c - outbound)
550 * Hooks at these callout points can be enabled/disabled using the ndd variable
551 * ip_policy_mask (a bit mask with the 4 LSB indicating the callout positions).
552 * By default all the callout positions are enabled.
553 *
554 * Outbound (local_out)
555 * Hooks are placed in ire_send_wire_v4 and ire_send_wire_v6.
556 *
557 * Inbound (local_in)
558 * Hooks are placed in ip_fanout_v4 and ip_fanout_v6.
559 *
560 * Forwarding (in and out)
561 * Hooks are placed in ire_recv_forward_v4/v6.
562 *
563 * IP Policy Framework processing (IPPF processing)
564 * Policy processing for a packet is initiated by ip_process, which ascertains
565 * that the classifier (ipgpc) is loaded and configured, failing which the
566 * packet resumes normal processing in IP. If the clasifier is present, the
567 * packet is acted upon by one or more IPQoS modules (action instances), per
568 * filters configured in ipgpc and resumes normal IP processing thereafter.
569 * An action instance can drop a packet in course of its processing.
570 *
571 * Zones notes:
572 *
573 * The partitioning rules for networking are as follows:
574 * 1) Packets coming from a zone must have a source address belonging to that
575 * zone.
576 * 2) Packets coming from a zone can only be sent on a physical interface on
577 * which the zone has an IP address.
578 * 3) Between two zones on the same machine, packet delivery is only allowed if
579 * there's a matching route for the destination and zone in the forwarding
580 * table.
581 * 4) The TCP and UDP port spaces are per-zone; that is, two processes in
582 * different zones can bind to the same port with the wildcard address
583 * (INADDR_ANY).
584 *
585 * The granularity of interface partitioning is at the logical interface level.
586 * Therefore, every zone has its own IP addresses, and incoming packets can be
587 * attributed to a zone unambiguously. A logical interface is placed into a zone
588 * using the SIOCSLIFZONE ioctl; this sets the ipif_zoneid field in the ipif_t
589 * structure. Rule (1) is implemented by modifying the source address selection
590 * algorithm so that the list of eligible addresses is filtered based on the
591 * sending process zone.
592 *
593 * The Internet Routing Entries (IREs) are either exclusive to a zone or shared
594 * across all zones, depending on their type. Here is the break-up:
595 *
596 * IRE type Shared/exclusive
597 * -------- ----------------
598 * IRE_BROADCAST Exclusive
599 * IRE_DEFAULT (default routes) Shared (*)
600 * IRE_LOCAL Exclusive (x)
601 * IRE_LOOPBACK Exclusive
602 * IRE_PREFIX (net routes) Shared (*)
603 * IRE_IF_NORESOLVER (interface routes) Exclusive
604 * IRE_IF_RESOLVER (interface routes) Exclusive
605 * IRE_IF_CLONE (interface routes) Exclusive
606 * IRE_HOST (host routes) Shared (*)
607 *
608 * (*) A zone can only use a default or off-subnet route if the gateway is
609 * directly reachable from the zone, that is, if the gateway's address matches
610 * one of the zone's logical interfaces.
611 *
612 * (x) IRE_LOCAL are handled a bit differently.
613 * When ip_restrict_interzone_loopback is set (the default),
614 * ire_route_recursive restricts loopback using an IRE_LOCAL
615 * between zone to the case when L2 would have conceptually looped the packet
616 * back, i.e. the loopback which is required since neither Ethernet drivers
617 * nor Ethernet hardware loops them back. This is the case when the normal
618 * routes (ignoring IREs with different zoneids) would send out the packet on
619 * the same ill as the ill with which is IRE_LOCAL is associated.
620 *
621 * Multiple zones can share a common broadcast address; typically all zones
622 * share the 255.255.255.255 address. Incoming as well as locally originated
623 * broadcast packets must be dispatched to all the zones on the broadcast
624 * network. For directed broadcasts (e.g. 10.16.72.255) this is not trivial
625 * since some zones may not be on the 10.16.72/24 network. To handle this, each
626 * zone has its own set of IRE_BROADCAST entries; then, broadcast packets are
627 * sent to every zone that has an IRE_BROADCAST entry for the destination
628 * address on the input ill, see ip_input_broadcast().
629 *
630 * Applications in different zones can join the same multicast group address.
631 * The same logic applies for multicast as for broadcast. ip_input_multicast
632 * dispatches packets to all zones that have members on the physical interface.
633 */
634
635 /*
636 * Squeue Fanout flags:
637 * 0: No fanout.
638 * 1: Fanout across all squeues
639 */
640 boolean_t ip_squeue_fanout = 0;
641
642 /*
643 * Maximum dups allowed per packet.
644 */
645 uint_t ip_max_frag_dups = 10;
646
647 static int ip_open(queue_t *q, dev_t *devp, int flag, int sflag,
648 cred_t *credp, boolean_t isv6);
649 static mblk_t *ip_xmit_attach_llhdr(mblk_t *, nce_t *);
650
651 static boolean_t icmp_inbound_verify_v4(mblk_t *, icmph_t *, ip_recv_attr_t *);
652 static void icmp_inbound_too_big_v4(icmph_t *, ip_recv_attr_t *);
653 static void icmp_inbound_error_fanout_v4(mblk_t *, icmph_t *,
654 ip_recv_attr_t *);
655 static void icmp_options_update(ipha_t *);
656 static void icmp_param_problem(mblk_t *, uint8_t, ip_recv_attr_t *);
657 static void icmp_pkt(mblk_t *, void *, size_t, ip_recv_attr_t *);
658 static mblk_t *icmp_pkt_err_ok(mblk_t *, ip_recv_attr_t *);
659 static void icmp_redirect_v4(mblk_t *mp, ipha_t *, icmph_t *,
660 ip_recv_attr_t *);
661 static void icmp_send_redirect(mblk_t *, ipaddr_t, ip_recv_attr_t *);
662 static void icmp_send_reply_v4(mblk_t *, ipha_t *, icmph_t *,
663 ip_recv_attr_t *);
664
665 mblk_t *ip_dlpi_alloc(size_t, t_uscalar_t);
666 char *ip_dot_addr(ipaddr_t, char *);
667 mblk_t *ip_carve_mp(mblk_t **, ssize_t);
668 int ip_close(queue_t *, int);
669 static char *ip_dot_saddr(uchar_t *, char *);
670 static void ip_lrput(queue_t *, mblk_t *);
671 ipaddr_t ip_net_mask(ipaddr_t);
672 char *ip_nv_lookup(nv_t *, int);
673 void ip_rput(queue_t *, mblk_t *);
674 static void ip_rput_dlpi_writer(ipsq_t *dummy_sq, queue_t *q, mblk_t *mp,
675 void *dummy_arg);
676 int ip_snmp_get(queue_t *, mblk_t *, int, boolean_t);
677 static mblk_t *ip_snmp_get_mib2_ip(queue_t *, mblk_t *,
678 mib2_ipIfStatsEntry_t *, ip_stack_t *, boolean_t);
679 static mblk_t *ip_snmp_get_mib2_ip_traffic_stats(queue_t *, mblk_t *,
680 ip_stack_t *, boolean_t);
681 static mblk_t *ip_snmp_get_mib2_ip6(queue_t *, mblk_t *, ip_stack_t *,
682 boolean_t);
683 static mblk_t *ip_snmp_get_mib2_icmp(queue_t *, mblk_t *, ip_stack_t *ipst);
684 static mblk_t *ip_snmp_get_mib2_icmp6(queue_t *, mblk_t *, ip_stack_t *ipst);
685 static mblk_t *ip_snmp_get_mib2_igmp(queue_t *, mblk_t *, ip_stack_t *ipst);
686 static mblk_t *ip_snmp_get_mib2_multi(queue_t *, mblk_t *, ip_stack_t *ipst);
687 static mblk_t *ip_snmp_get_mib2_ip_addr(queue_t *, mblk_t *,
688 ip_stack_t *ipst, boolean_t);
689 static mblk_t *ip_snmp_get_mib2_ip6_addr(queue_t *, mblk_t *,
690 ip_stack_t *ipst, boolean_t);
691 static mblk_t *ip_snmp_get_mib2_ip_group_src(queue_t *, mblk_t *,
692 ip_stack_t *ipst);
693 static mblk_t *ip_snmp_get_mib2_ip6_group_src(queue_t *, mblk_t *,
694 ip_stack_t *ipst);
695 static mblk_t *ip_snmp_get_mib2_ip_group_mem(queue_t *, mblk_t *,
696 ip_stack_t *ipst);
697 static mblk_t *ip_snmp_get_mib2_ip6_group_mem(queue_t *, mblk_t *,
698 ip_stack_t *ipst);
699 static mblk_t *ip_snmp_get_mib2_virt_multi(queue_t *, mblk_t *,
700 ip_stack_t *ipst);
701 static mblk_t *ip_snmp_get_mib2_multi_rtable(queue_t *, mblk_t *,
702 ip_stack_t *ipst);
703 static mblk_t *ip_snmp_get_mib2_ip_route_media(queue_t *, mblk_t *, int,
704 ip_stack_t *ipst);
705 static mblk_t *ip_snmp_get_mib2_ip6_route_media(queue_t *, mblk_t *, int,
706 ip_stack_t *ipst);
707 static void ip_snmp_get2_v4(ire_t *, iproutedata_t *);
708 static void ip_snmp_get2_v6_route(ire_t *, iproutedata_t *);
709 static int ip_snmp_get2_v4_media(ncec_t *, iproutedata_t *);
710 static int ip_snmp_get2_v6_media(ncec_t *, iproutedata_t *);
711 int ip_snmp_set(queue_t *, int, int, uchar_t *, int);
712
713 static mblk_t *ip_fragment_copyhdr(uchar_t *, int, int, ip_stack_t *,
714 mblk_t *);
715
716 static void conn_drain_init(ip_stack_t *);
717 static void conn_drain_fini(ip_stack_t *);
718 static void conn_drain(conn_t *connp, boolean_t closing);
719
720 static void conn_walk_drain(ip_stack_t *, idl_tx_list_t *);
721 static void conn_walk_sctp(pfv_t, void *, zoneid_t, netstack_t *);
722
723 static void *ip_stack_init(netstackid_t stackid, netstack_t *ns);
724 static void ip_stack_shutdown(netstackid_t stackid, void *arg);
725 static void ip_stack_fini(netstackid_t stackid, void *arg);
726
727 static int ip_multirt_apply_membership(int (*fn)(conn_t *, boolean_t,
728 const in6_addr_t *, ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *),
729 ire_t *, conn_t *, boolean_t, const in6_addr_t *, mcast_record_t,
730 const in6_addr_t *);
731
732 static int ip_squeue_switch(int);
733
734 static void *ip_kstat_init(netstackid_t, ip_stack_t *);
735 static void ip_kstat_fini(netstackid_t, kstat_t *);
736 static int ip_kstat_update(kstat_t *kp, int rw);
737 static void *icmp_kstat_init(netstackid_t);
738 static void icmp_kstat_fini(netstackid_t, kstat_t *);
739 static int icmp_kstat_update(kstat_t *kp, int rw);
740 static void *ip_kstat2_init(netstackid_t, ip_stat_t *);
741 static void ip_kstat2_fini(netstackid_t, kstat_t *);
742
743 static void ipobs_init(ip_stack_t *);
744 static void ipobs_fini(ip_stack_t *);
745
746 static int ip_tp_cpu_update(cpu_setup_t, int, void *);
747
748 ipaddr_t ip_g_all_ones = IP_HOST_MASK;
749
750 static long ip_rput_pullups;
751 int dohwcksum = 1; /* use h/w cksum if supported by the hardware */
752
753 vmem_t *ip_minor_arena_sa; /* for minor nos. from INET_MIN_DEV+2 thru 2^^18-1 */
754 vmem_t *ip_minor_arena_la; /* for minor nos. from 2^^18 thru 2^^32-1 */
755
756 int ip_debug;
757
758 /*
759 * Multirouting/CGTP stuff
760 */
761 int ip_cgtp_filter_rev = CGTP_FILTER_REV; /* CGTP hooks version */
762
763 /*
764 * IP tunables related declarations. Definitions are in ip_tunables.c
765 */
766 extern mod_prop_info_t ip_propinfo_tbl[];
767 extern int ip_propinfo_count;
768
769 /*
770 * Table of IP ioctls encoding the various properties of the ioctl and
771 * indexed based on the last byte of the ioctl command. Occasionally there
772 * is a clash, and there is more than 1 ioctl with the same last byte.
773 * In such a case 1 ioctl is encoded in the ndx table and the remaining
774 * ioctls are encoded in the misc table. An entry in the ndx table is
775 * retrieved by indexing on the last byte of the ioctl command and comparing
776 * the ioctl command with the value in the ndx table. In the event of a
777 * mismatch the misc table is then searched sequentially for the desired
778 * ioctl command.
779 *
780 * Entry: <command> <copyin_size> <flags> <cmd_type> <function> <restart_func>
781 */
782 ip_ioctl_cmd_t ip_ndx_ioctl_table[] = {
783 /* 000 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
784 /* 001 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
785 /* 002 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
786 /* 003 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
787 /* 004 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
788 /* 005 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
789 /* 006 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
790 /* 007 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
791 /* 008 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
792 /* 009 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
793
794 /* 010 */ { SIOCADDRT, sizeof (struct rtentry), IPI_PRIV,
795 MISC_CMD, ip_siocaddrt, NULL },
796 /* 011 */ { SIOCDELRT, sizeof (struct rtentry), IPI_PRIV,
797 MISC_CMD, ip_siocdelrt, NULL },
798
799 /* 012 */ { SIOCSIFADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
800 IF_CMD, ip_sioctl_addr, ip_sioctl_addr_restart },
801 /* 013 */ { SIOCGIFADDR, sizeof (struct ifreq), IPI_GET_CMD,
802 IF_CMD, ip_sioctl_get_addr, NULL },
803
804 /* 014 */ { SIOCSIFDSTADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
805 IF_CMD, ip_sioctl_dstaddr, ip_sioctl_dstaddr_restart },
806 /* 015 */ { SIOCGIFDSTADDR, sizeof (struct ifreq),
807 IPI_GET_CMD, IF_CMD, ip_sioctl_get_dstaddr, NULL },
808
809 /* 016 */ { SIOCSIFFLAGS, sizeof (struct ifreq),
810 IPI_PRIV | IPI_WR,
811 IF_CMD, ip_sioctl_flags, ip_sioctl_flags_restart },
812 /* 017 */ { SIOCGIFFLAGS, sizeof (struct ifreq),
813 IPI_MODOK | IPI_GET_CMD,
814 IF_CMD, ip_sioctl_get_flags, NULL },
815
816 /* 018 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
817 /* 019 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
818
819 /* copyin size cannot be coded for SIOCGIFCONF */
820 /* 020 */ { O_SIOCGIFCONF, 0, IPI_GET_CMD,
821 MISC_CMD, ip_sioctl_get_ifconf, NULL },
822
823 /* 021 */ { SIOCSIFMTU, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
824 IF_CMD, ip_sioctl_mtu, NULL },
825 /* 022 */ { SIOCGIFMTU, sizeof (struct ifreq), IPI_GET_CMD,
826 IF_CMD, ip_sioctl_get_mtu, NULL },
827 /* 023 */ { SIOCGIFBRDADDR, sizeof (struct ifreq),
828 IPI_GET_CMD, IF_CMD, ip_sioctl_get_brdaddr, NULL },
829 /* 024 */ { SIOCSIFBRDADDR, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
830 IF_CMD, ip_sioctl_brdaddr, NULL },
831 /* 025 */ { SIOCGIFNETMASK, sizeof (struct ifreq),
832 IPI_GET_CMD, IF_CMD, ip_sioctl_get_netmask, NULL },
833 /* 026 */ { SIOCSIFNETMASK, sizeof (struct ifreq), IPI_PRIV | IPI_WR,
834 IF_CMD, ip_sioctl_netmask, ip_sioctl_netmask_restart },
835 /* 027 */ { SIOCGIFMETRIC, sizeof (struct ifreq),
836 IPI_GET_CMD, IF_CMD, ip_sioctl_get_metric, NULL },
837 /* 028 */ { SIOCSIFMETRIC, sizeof (struct ifreq), IPI_PRIV,
838 IF_CMD, ip_sioctl_metric, NULL },
839 /* 029 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
840
841 /* See 166-168 below for extended SIOC*XARP ioctls */
842 /* 030 */ { SIOCSARP, sizeof (struct arpreq), IPI_PRIV | IPI_WR,
843 ARP_CMD, ip_sioctl_arp, NULL },
844 /* 031 */ { SIOCGARP, sizeof (struct arpreq), IPI_GET_CMD,
845 ARP_CMD, ip_sioctl_arp, NULL },
846 /* 032 */ { SIOCDARP, sizeof (struct arpreq), IPI_PRIV | IPI_WR,
847 ARP_CMD, ip_sioctl_arp, NULL },
848
849 /* 033 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
850 /* 034 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
851 /* 035 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
852 /* 036 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
853 /* 037 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
854 /* 038 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
855 /* 039 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
856 /* 040 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
857 /* 041 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
858 /* 042 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
859 /* 043 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
860 /* 044 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
861 /* 045 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
862 /* 046 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
863 /* 047 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
864 /* 048 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
865 /* 049 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
866 /* 050 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
867 /* 051 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
868 /* 052 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
869 /* 053 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
870
871 /* 054 */ { IF_UNITSEL, sizeof (int), IPI_PRIV | IPI_WR | IPI_MODOK,
872 MISC_CMD, if_unitsel, if_unitsel_restart },
873
874 /* 055 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
875 /* 056 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
876 /* 057 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
877 /* 058 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
878 /* 059 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
879 /* 060 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
880 /* 061 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
881 /* 062 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
882 /* 063 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
883 /* 064 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
884 /* 065 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
885 /* 066 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
886 /* 067 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
887 /* 068 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
888 /* 069 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
889 /* 070 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
890 /* 071 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
891 /* 072 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
892
893 /* 073 */ { SIOCSIFNAME, sizeof (struct ifreq),
894 IPI_PRIV | IPI_WR | IPI_MODOK,
895 IF_CMD, ip_sioctl_sifname, NULL },
896
897 /* 074 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
898 /* 075 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
899 /* 076 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
900 /* 077 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
901 /* 078 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
902 /* 079 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
903 /* 080 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
904 /* 081 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
905 /* 082 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
906 /* 083 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
907 /* 084 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
908 /* 085 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
909 /* 086 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
910
911 /* 087 */ { SIOCGIFNUM, sizeof (int), IPI_GET_CMD,
912 MISC_CMD, ip_sioctl_get_ifnum, NULL },
913 /* 088 */ { SIOCGIFMUXID, sizeof (struct ifreq), IPI_GET_CMD,
914 IF_CMD, ip_sioctl_get_muxid, NULL },
915 /* 089 */ { SIOCSIFMUXID, sizeof (struct ifreq),
916 IPI_PRIV | IPI_WR, IF_CMD, ip_sioctl_muxid, NULL },
917
918 /* Both if and lif variants share same func */
919 /* 090 */ { SIOCGIFINDEX, sizeof (struct ifreq), IPI_GET_CMD,
920 IF_CMD, ip_sioctl_get_lifindex, NULL },
921 /* Both if and lif variants share same func */
922 /* 091 */ { SIOCSIFINDEX, sizeof (struct ifreq),
923 IPI_PRIV | IPI_WR, IF_CMD, ip_sioctl_slifindex, NULL },
924
925 /* copyin size cannot be coded for SIOCGIFCONF */
926 /* 092 */ { SIOCGIFCONF, 0, IPI_GET_CMD,
927 MISC_CMD, ip_sioctl_get_ifconf, NULL },
928 /* 093 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
929 /* 094 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
930 /* 095 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
931 /* 096 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
932 /* 097 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
933 /* 098 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
934 /* 099 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
935 /* 100 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
936 /* 101 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
937 /* 102 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
938 /* 103 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
939 /* 104 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
940 /* 105 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
941 /* 106 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
942 /* 107 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
943 /* 108 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
944 /* 109 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
945
946 /* 110 */ { SIOCLIFREMOVEIF, sizeof (struct lifreq),
947 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_removeif,
948 ip_sioctl_removeif_restart },
949 /* 111 */ { SIOCLIFADDIF, sizeof (struct lifreq),
950 IPI_GET_CMD | IPI_PRIV | IPI_WR,
951 LIF_CMD, ip_sioctl_addif, NULL },
952 #define SIOCLIFADDR_NDX 112
953 /* 112 */ { SIOCSLIFADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
954 LIF_CMD, ip_sioctl_addr, ip_sioctl_addr_restart },
955 /* 113 */ { SIOCGLIFADDR, sizeof (struct lifreq),
956 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_addr, NULL },
957 /* 114 */ { SIOCSLIFDSTADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
958 LIF_CMD, ip_sioctl_dstaddr, ip_sioctl_dstaddr_restart },
959 /* 115 */ { SIOCGLIFDSTADDR, sizeof (struct lifreq),
960 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_dstaddr, NULL },
961 /* 116 */ { SIOCSLIFFLAGS, sizeof (struct lifreq),
962 IPI_PRIV | IPI_WR,
963 LIF_CMD, ip_sioctl_flags, ip_sioctl_flags_restart },
964 /* 117 */ { SIOCGLIFFLAGS, sizeof (struct lifreq),
965 IPI_GET_CMD | IPI_MODOK,
966 LIF_CMD, ip_sioctl_get_flags, NULL },
967
968 /* 118 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
969 /* 119 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
970
971 /* 120 */ { O_SIOCGLIFCONF, 0, IPI_GET_CMD, MISC_CMD,
972 ip_sioctl_get_lifconf, NULL },
973 /* 121 */ { SIOCSLIFMTU, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
974 LIF_CMD, ip_sioctl_mtu, NULL },
975 /* 122 */ { SIOCGLIFMTU, sizeof (struct lifreq), IPI_GET_CMD,
976 LIF_CMD, ip_sioctl_get_mtu, NULL },
977 /* 123 */ { SIOCGLIFBRDADDR, sizeof (struct lifreq),
978 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_brdaddr, NULL },
979 /* 124 */ { SIOCSLIFBRDADDR, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
980 LIF_CMD, ip_sioctl_brdaddr, NULL },
981 /* 125 */ { SIOCGLIFNETMASK, sizeof (struct lifreq),
982 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_netmask, NULL },
983 /* 126 */ { SIOCSLIFNETMASK, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
984 LIF_CMD, ip_sioctl_netmask, ip_sioctl_netmask_restart },
985 /* 127 */ { SIOCGLIFMETRIC, sizeof (struct lifreq),
986 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_metric, NULL },
987 /* 128 */ { SIOCSLIFMETRIC, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
988 LIF_CMD, ip_sioctl_metric, NULL },
989 /* 129 */ { SIOCSLIFNAME, sizeof (struct lifreq),
990 IPI_PRIV | IPI_WR | IPI_MODOK,
991 LIF_CMD, ip_sioctl_slifname,
992 ip_sioctl_slifname_restart },
993
994 /* 130 */ { SIOCGLIFNUM, sizeof (struct lifnum), IPI_GET_CMD,
995 MISC_CMD, ip_sioctl_get_lifnum, NULL },
996 /* 131 */ { SIOCGLIFMUXID, sizeof (struct lifreq),
997 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_muxid, NULL },
998 /* 132 */ { SIOCSLIFMUXID, sizeof (struct lifreq),
999 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_muxid, NULL },
1000 /* 133 */ { SIOCGLIFINDEX, sizeof (struct lifreq),
1001 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lifindex, 0 },
1002 /* 134 */ { SIOCSLIFINDEX, sizeof (struct lifreq),
1003 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_slifindex, 0 },
1004 /* 135 */ { SIOCSLIFTOKEN, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1005 LIF_CMD, ip_sioctl_token, NULL },
1006 /* 136 */ { SIOCGLIFTOKEN, sizeof (struct lifreq),
1007 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_token, NULL },
1008 /* 137 */ { SIOCSLIFSUBNET, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1009 LIF_CMD, ip_sioctl_subnet, ip_sioctl_subnet_restart },
1010 /* 138 */ { SIOCGLIFSUBNET, sizeof (struct lifreq),
1011 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_subnet, NULL },
1012 /* 139 */ { SIOCSLIFLNKINFO, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1013 LIF_CMD, ip_sioctl_lnkinfo, NULL },
1014
1015 /* 140 */ { SIOCGLIFLNKINFO, sizeof (struct lifreq),
1016 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lnkinfo, NULL },
1017 /* 141 */ { SIOCLIFDELND, sizeof (struct lifreq), IPI_PRIV,
1018 LIF_CMD, ip_siocdelndp_v6, NULL },
1019 /* 142 */ { SIOCLIFGETND, sizeof (struct lifreq), IPI_GET_CMD,
1020 LIF_CMD, ip_siocqueryndp_v6, NULL },
1021 /* 143 */ { SIOCLIFSETND, sizeof (struct lifreq), IPI_PRIV,
1022 LIF_CMD, ip_siocsetndp_v6, NULL },
1023 /* 144 */ { SIOCTMYADDR, sizeof (struct sioc_addrreq), IPI_GET_CMD,
1024 MISC_CMD, ip_sioctl_tmyaddr, NULL },
1025 /* 145 */ { SIOCTONLINK, sizeof (struct sioc_addrreq), IPI_GET_CMD,
1026 MISC_CMD, ip_sioctl_tonlink, NULL },
1027 /* 146 */ { SIOCTMYSITE, sizeof (struct sioc_addrreq), 0,
1028 MISC_CMD, ip_sioctl_tmysite, NULL },
1029 /* 147 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1030 /* 148 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1031
1032 /* Old *IPSECONFIG ioctls are now deprecated, now see spdsock.c */
1033 /* 149 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1034 /* 150 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1035 /* 151 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1036 /* 152 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1037
1038 /* 153 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1039
1040 /* 154 */ { SIOCGLIFBINDING, sizeof (struct lifreq), IPI_GET_CMD,
1041 LIF_CMD, ip_sioctl_get_binding, NULL },
1042 /* 155 */ { SIOCSLIFGROUPNAME, sizeof (struct lifreq),
1043 IPI_PRIV | IPI_WR,
1044 LIF_CMD, ip_sioctl_groupname, ip_sioctl_groupname },
1045 /* 156 */ { SIOCGLIFGROUPNAME, sizeof (struct lifreq),
1046 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_groupname, NULL },
1047 /* 157 */ { SIOCGLIFGROUPINFO, sizeof (lifgroupinfo_t),
1048 IPI_GET_CMD, MISC_CMD, ip_sioctl_groupinfo, NULL },
1049
1050 /* Leave 158-160 unused; used to be SIOC*IFARP ioctls */
1051 /* 158 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1052 /* 159 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1053 /* 160 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1054
1055 /* 161 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1056
1057 /* These are handled in ip_sioctl_copyin_setup itself */
1058 /* 162 */ { SIOCGIP6ADDRPOLICY, 0, IPI_NULL_BCONT,
1059 MISC_CMD, NULL, NULL },
1060 /* 163 */ { SIOCSIP6ADDRPOLICY, 0, IPI_PRIV | IPI_NULL_BCONT,
1061 MISC_CMD, NULL, NULL },
1062 /* 164 */ { SIOCGDSTINFO, 0, IPI_GET_CMD, MISC_CMD, NULL, NULL },
1063
1064 /* 165 */ { SIOCGLIFCONF, 0, IPI_GET_CMD, MISC_CMD,
1065 ip_sioctl_get_lifconf, NULL },
1066
1067 /* 166 */ { SIOCSXARP, sizeof (struct xarpreq), IPI_PRIV | IPI_WR,
1068 XARP_CMD, ip_sioctl_arp, NULL },
1069 /* 167 */ { SIOCGXARP, sizeof (struct xarpreq), IPI_GET_CMD,
1070 XARP_CMD, ip_sioctl_arp, NULL },
1071 /* 168 */ { SIOCDXARP, sizeof (struct xarpreq), IPI_PRIV | IPI_WR,
1072 XARP_CMD, ip_sioctl_arp, NULL },
1073
1074 /* SIOCPOPSOCKFS is not handled by IP */
1075 /* 169 */ { IPI_DONTCARE /* SIOCPOPSOCKFS */, 0, 0, 0, NULL, NULL },
1076
1077 /* 170 */ { SIOCGLIFZONE, sizeof (struct lifreq),
1078 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_lifzone, NULL },
1079 /* 171 */ { SIOCSLIFZONE, sizeof (struct lifreq),
1080 IPI_PRIV | IPI_WR, LIF_CMD, ip_sioctl_slifzone,
1081 ip_sioctl_slifzone_restart },
1082 /* 172-174 are SCTP ioctls and not handled by IP */
1083 /* 172 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1084 /* 173 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1085 /* 174 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1086 /* 175 */ { SIOCGLIFUSESRC, sizeof (struct lifreq),
1087 IPI_GET_CMD, LIF_CMD,
1088 ip_sioctl_get_lifusesrc, 0 },
1089 /* 176 */ { SIOCSLIFUSESRC, sizeof (struct lifreq),
1090 IPI_PRIV | IPI_WR,
1091 LIF_CMD, ip_sioctl_slifusesrc,
1092 NULL },
1093 /* 177 */ { SIOCGLIFSRCOF, 0, IPI_GET_CMD, MISC_CMD,
1094 ip_sioctl_get_lifsrcof, NULL },
1095 /* 178 */ { SIOCGMSFILTER, sizeof (struct group_filter), IPI_GET_CMD,
1096 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1097 /* 179 */ { SIOCSMSFILTER, sizeof (struct group_filter), 0,
1098 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1099 /* 180 */ { SIOCGIPMSFILTER, sizeof (struct ip_msfilter), IPI_GET_CMD,
1100 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1101 /* 181 */ { SIOCSIPMSFILTER, sizeof (struct ip_msfilter), 0,
1102 MSFILT_CMD, ip_sioctl_msfilter, NULL },
1103 /* 182 */ { IPI_DONTCARE, 0, 0, 0, NULL, NULL },
1104 /* SIOCSENABLESDP is handled by SDP */
1105 /* 183 */ { IPI_DONTCARE /* SIOCSENABLESDP */, 0, 0, 0, NULL, NULL },
1106 /* 184 */ { IPI_DONTCARE /* SIOCSQPTR */, 0, 0, 0, NULL, NULL },
1107 /* 185 */ { SIOCGIFHWADDR, sizeof (struct ifreq), IPI_GET_CMD,
1108 IF_CMD, ip_sioctl_get_ifhwaddr, NULL },
1109 /* 186 */ { IPI_DONTCARE /* SIOCGSTAMP */, 0, 0, 0, NULL, NULL },
1110 /* 187 */ { SIOCILB, 0, IPI_PRIV | IPI_GET_CMD, MISC_CMD,
1111 ip_sioctl_ilb_cmd, NULL },
1112 /* 188 */ { SIOCGETPROP, 0, IPI_GET_CMD, 0, NULL, NULL },
1113 /* 189 */ { SIOCSETPROP, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL},
1114 /* 190 */ { SIOCGLIFDADSTATE, sizeof (struct lifreq),
1115 IPI_GET_CMD, LIF_CMD, ip_sioctl_get_dadstate, NULL },
1116 /* 191 */ { SIOCSLIFPREFIX, sizeof (struct lifreq), IPI_PRIV | IPI_WR,
1117 LIF_CMD, ip_sioctl_prefix, ip_sioctl_prefix_restart },
1118 /* 192 */ { SIOCGLIFHWADDR, sizeof (struct lifreq), IPI_GET_CMD,
1119 LIF_CMD, ip_sioctl_get_lifhwaddr, NULL }
1120 };
1121
1122 int ip_ndx_ioctl_count = sizeof (ip_ndx_ioctl_table) / sizeof (ip_ioctl_cmd_t);
1123
1124 ip_ioctl_cmd_t ip_misc_ioctl_table[] = {
1125 { I_LINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1126 { I_UNLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1127 { I_PLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1128 { I_PUNLINK, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1129 { ND_GET, 0, 0, 0, NULL, NULL },
1130 { ND_SET, 0, IPI_PRIV | IPI_WR, 0, NULL, NULL },
1131 { IP_IOCTL, 0, 0, 0, NULL, NULL },
1132 { SIOCGETVIFCNT, sizeof (struct sioc_vif_req), IPI_GET_CMD,
1133 MISC_CMD, mrt_ioctl},
1134 { SIOCGETSGCNT, sizeof (struct sioc_sg_req), IPI_GET_CMD,
1135 MISC_CMD, mrt_ioctl},
1136 { SIOCGETLSGCNT, sizeof (struct sioc_lsg_req), IPI_GET_CMD,
1137 MISC_CMD, mrt_ioctl}
1138 };
1139
1140 int ip_misc_ioctl_count =
1141 sizeof (ip_misc_ioctl_table) / sizeof (ip_ioctl_cmd_t);
1142
1143 int conn_drain_nthreads; /* Number of drainers reqd. */
1144 /* Settable in /etc/system */
1145 /* Defined in ip_ire.c */
1146 extern uint32_t ip_ire_max_bucket_cnt, ip6_ire_max_bucket_cnt;
1147 extern uint32_t ip_ire_min_bucket_cnt, ip6_ire_min_bucket_cnt;
1148 extern uint32_t ip_ire_mem_ratio, ip_ire_cpu_ratio;
1149
1150 static nv_t ire_nv_arr[] = {
1151 { IRE_BROADCAST, "BROADCAST" },
1152 { IRE_LOCAL, "LOCAL" },
1153 { IRE_LOOPBACK, "LOOPBACK" },
1154 { IRE_DEFAULT, "DEFAULT" },
1155 { IRE_PREFIX, "PREFIX" },
1156 { IRE_IF_NORESOLVER, "IF_NORESOL" },
1157 { IRE_IF_RESOLVER, "IF_RESOLV" },
1158 { IRE_IF_CLONE, "IF_CLONE" },
1159 { IRE_HOST, "HOST" },
1160 { IRE_MULTICAST, "MULTICAST" },
1161 { IRE_NOROUTE, "NOROUTE" },
1162 { 0 }
1163 };
1164
1165 nv_t *ire_nv_tbl = ire_nv_arr;
1166
1167 /* Simple ICMP IP Header Template */
1168 static ipha_t icmp_ipha = {
1169 IP_SIMPLE_HDR_VERSION, 0, 0, 0, 0, 0, IPPROTO_ICMP
1170 };
1171
1172 struct module_info ip_mod_info = {
1173 IP_MOD_ID, IP_MOD_NAME, IP_MOD_MINPSZ, IP_MOD_MAXPSZ, IP_MOD_HIWAT,
1174 IP_MOD_LOWAT
1175 };
1176
1177 /*
1178 * Duplicate static symbols within a module confuses mdb; so we avoid the
1179 * problem by making the symbols here distinct from those in udp.c.
1180 */
1181
1182 /*
1183 * Entry points for IP as a device and as a module.
1184 * We have separate open functions for the /dev/ip and /dev/ip6 devices.
1185 */
1186 static struct qinit iprinitv4 = {
1187 (pfi_t)ip_rput, NULL, ip_openv4, ip_close, NULL,
1188 &ip_mod_info
1189 };
1190
1191 struct qinit iprinitv6 = {
1192 (pfi_t)ip_rput_v6, NULL, ip_openv6, ip_close, NULL,
1193 &ip_mod_info
1194 };
1195
1196 static struct qinit ipwinit = {
1197 (pfi_t)ip_wput_nondata, (pfi_t)ip_wsrv, NULL, NULL, NULL,
1198 &ip_mod_info
1199 };
1200
1201 static struct qinit iplrinit = {
1202 (pfi_t)ip_lrput, NULL, ip_openv4, ip_close, NULL,
1203 &ip_mod_info
1204 };
1205
1206 static struct qinit iplwinit = {
1207 (pfi_t)ip_lwput, NULL, NULL, NULL, NULL,
1208 &ip_mod_info
1209 };
1210
1211 /* For AF_INET aka /dev/ip */
1212 struct streamtab ipinfov4 = {
1213 &iprinitv4, &ipwinit, &iplrinit, &iplwinit
1214 };
1215
1216 /* For AF_INET6 aka /dev/ip6 */
1217 struct streamtab ipinfov6 = {
1218 &iprinitv6, &ipwinit, &iplrinit, &iplwinit
1219 };
1220
1221 #ifdef DEBUG
1222 boolean_t skip_sctp_cksum = B_FALSE;
1223 #endif
1224
1225 /*
1226 * Generate an ICMP fragmentation needed message.
1227 * When called from ip_output side a minimal ip_recv_attr_t needs to be
1228 * constructed by the caller.
1229 */
1230 void
1231 icmp_frag_needed(mblk_t *mp, int mtu, ip_recv_attr_t *ira)
1232 {
1233 icmph_t icmph;
1234 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
1235
1236 mp = icmp_pkt_err_ok(mp, ira);
1237 if (mp == NULL)
1238 return;
1239
1240 bzero(&icmph, sizeof (icmph_t));
1241 icmph.icmph_type = ICMP_DEST_UNREACHABLE;
1242 icmph.icmph_code = ICMP_FRAGMENTATION_NEEDED;
1243 icmph.icmph_du_mtu = htons((uint16_t)mtu);
1244 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutFragNeeded);
1245 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDestUnreachs);
1246
1247 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
1248 }
1249
1250 /*
1251 * icmp_inbound_v4 deals with ICMP messages that are handled by IP.
1252 * If the ICMP message is consumed by IP, i.e., it should not be delivered
1253 * to any IPPROTO_ICMP raw sockets, then it returns NULL.
1254 * Likewise, if the ICMP error is misformed (too short, etc), then it
1255 * returns NULL. The caller uses this to determine whether or not to send
1256 * to raw sockets.
1257 *
1258 * All error messages are passed to the matching transport stream.
1259 *
1260 * The following cases are handled by icmp_inbound:
1261 * 1) It needs to send a reply back and possibly delivering it
1262 * to the "interested" upper clients.
1263 * 2) Return the mblk so that the caller can pass it to the RAW socket clients.
1264 * 3) It needs to change some values in IP only.
1265 * 4) It needs to change some values in IP and upper layers e.g TCP
1266 * by delivering an error to the upper layers.
1267 *
1268 * We handle the above three cases in the context of IPsec in the
1269 * following way :
1270 *
1271 * 1) Send the reply back in the same way as the request came in.
1272 * If it came in encrypted, it goes out encrypted. If it came in
1273 * clear, it goes out in clear. Thus, this will prevent chosen
1274 * plain text attack.
1275 * 2) The client may or may not expect things to come in secure.
1276 * If it comes in secure, the policy constraints are checked
1277 * before delivering it to the upper layers. If it comes in
1278 * clear, ipsec_inbound_accept_clear will decide whether to
1279 * accept this in clear or not. In both the cases, if the returned
1280 * message (IP header + 8 bytes) that caused the icmp message has
1281 * AH/ESP headers, it is sent up to AH/ESP for validation before
1282 * sending up. If there are only 8 bytes of returned message, then
1283 * upper client will not be notified.
1284 * 3) Check with global policy to see whether it matches the constaints.
1285 * But this will be done only if icmp_accept_messages_in_clear is
1286 * zero.
1287 * 4) If we need to change both in IP and ULP, then the decision taken
1288 * while affecting the values in IP and while delivering up to TCP
1289 * should be the same.
1290 *
1291 * There are two cases.
1292 *
1293 * a) If we reject data at the IP layer (ipsec_check_global_policy()
1294 * failed), we will not deliver it to the ULP, even though they
1295 * are *willing* to accept in *clear*. This is fine as our global
1296 * disposition to icmp messages asks us reject the datagram.
1297 *
1298 * b) If we accept data at the IP layer (ipsec_check_global_policy()
1299 * succeeded or icmp_accept_messages_in_clear is 1), and not able
1300 * to deliver it to ULP (policy failed), it can lead to
1301 * consistency problems. The cases known at this time are
1302 * ICMP_DESTINATION_UNREACHABLE messages with following code
1303 * values :
1304 *
1305 * - ICMP_FRAGMENTATION_NEEDED : IP adapts to the new value
1306 * and Upper layer rejects. Then the communication will
1307 * come to a stop. This is solved by making similar decisions
1308 * at both levels. Currently, when we are unable to deliver
1309 * to the Upper Layer (due to policy failures) while IP has
1310 * adjusted dce_pmtu, the next outbound datagram would
1311 * generate a local ICMP_FRAGMENTATION_NEEDED message - which
1312 * will be with the right level of protection. Thus the right
1313 * value will be communicated even if we are not able to
1314 * communicate when we get from the wire initially. But this
1315 * assumes there would be at least one outbound datagram after
1316 * IP has adjusted its dce_pmtu value. To make things
1317 * simpler, we accept in clear after the validation of
1318 * AH/ESP headers.
1319 *
1320 * - Other ICMP ERRORS : We may not be able to deliver it to the
1321 * upper layer depending on the level of protection the upper
1322 * layer expects and the disposition in ipsec_inbound_accept_clear().
1323 * ipsec_inbound_accept_clear() decides whether a given ICMP error
1324 * should be accepted in clear when the Upper layer expects secure.
1325 * Thus the communication may get aborted by some bad ICMP
1326 * packets.
1327 */
1328 mblk_t *
1329 icmp_inbound_v4(mblk_t *mp, ip_recv_attr_t *ira)
1330 {
1331 icmph_t *icmph;
1332 ipha_t *ipha; /* Outer header */
1333 int ip_hdr_length; /* Outer header length */
1334 boolean_t interested;
1335 ipif_t *ipif;
1336 uint32_t ts;
1337 uint32_t *tsp;
1338 timestruc_t now;
1339 ill_t *ill = ira->ira_ill;
1340 ip_stack_t *ipst = ill->ill_ipst;
1341 zoneid_t zoneid = ira->ira_zoneid;
1342 int len_needed;
1343 mblk_t *mp_ret = NULL;
1344
1345 ipha = (ipha_t *)mp->b_rptr;
1346
1347 BUMP_MIB(&ipst->ips_icmp_mib, icmpInMsgs);
1348
1349 ip_hdr_length = ira->ira_ip_hdr_length;
1350 if ((mp->b_wptr - mp->b_rptr) < (ip_hdr_length + ICMPH_SIZE)) {
1351 if (ira->ira_pktlen < (ip_hdr_length + ICMPH_SIZE)) {
1352 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
1353 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
1354 freemsg(mp);
1355 return (NULL);
1356 }
1357 /* Last chance to get real. */
1358 ipha = ip_pullup(mp, ip_hdr_length + ICMPH_SIZE, ira);
1359 if (ipha == NULL) {
1360 BUMP_MIB(&ipst->ips_icmp_mib, icmpInErrors);
1361 freemsg(mp);
1362 return (NULL);
1363 }
1364 }
1365
1366 /* The IP header will always be a multiple of four bytes */
1367 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1368 ip2dbg(("icmp_inbound_v4: type %d code %d\n", icmph->icmph_type,
1369 icmph->icmph_code));
1370
1371 /*
1372 * We will set "interested" to "true" if we should pass a copy to
1373 * the transport or if we handle the packet locally.
1374 */
1375 interested = B_FALSE;
1376 switch (icmph->icmph_type) {
1377 case ICMP_ECHO_REPLY:
1378 BUMP_MIB(&ipst->ips_icmp_mib, icmpInEchoReps);
1379 break;
1380 case ICMP_DEST_UNREACHABLE:
1381 if (icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED)
1382 BUMP_MIB(&ipst->ips_icmp_mib, icmpInFragNeeded);
1383 interested = B_TRUE; /* Pass up to transport */
1384 BUMP_MIB(&ipst->ips_icmp_mib, icmpInDestUnreachs);
1385 break;
1386 case ICMP_SOURCE_QUENCH:
1387 interested = B_TRUE; /* Pass up to transport */
1388 BUMP_MIB(&ipst->ips_icmp_mib, icmpInSrcQuenchs);
1389 break;
1390 case ICMP_REDIRECT:
1391 if (!ipst->ips_ip_ignore_redirect)
1392 interested = B_TRUE;
1393 BUMP_MIB(&ipst->ips_icmp_mib, icmpInRedirects);
1394 break;
1395 case ICMP_ECHO_REQUEST:
1396 /*
1397 * Whether to respond to echo requests that come in as IP
1398 * broadcasts or as IP multicast is subject to debate
1399 * (what isn't?). We aim to please, you pick it.
1400 * Default is do it.
1401 */
1402 if (ira->ira_flags & IRAF_MULTICAST) {
1403 /* multicast: respond based on tunable */
1404 interested = ipst->ips_ip_g_resp_to_echo_mcast;
1405 } else if (ira->ira_flags & IRAF_BROADCAST) {
1406 /* broadcast: respond based on tunable */
1407 interested = ipst->ips_ip_g_resp_to_echo_bcast;
1408 } else {
1409 /* unicast: always respond */
1410 interested = B_TRUE;
1411 }
1412 BUMP_MIB(&ipst->ips_icmp_mib, icmpInEchos);
1413 if (!interested) {
1414 /* We never pass these to RAW sockets */
1415 freemsg(mp);
1416 return (NULL);
1417 }
1418
1419 /* Check db_ref to make sure we can modify the packet. */
1420 if (mp->b_datap->db_ref > 1) {
1421 mblk_t *mp1;
1422
1423 mp1 = copymsg(mp);
1424 freemsg(mp);
1425 if (!mp1) {
1426 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1427 return (NULL);
1428 }
1429 mp = mp1;
1430 ipha = (ipha_t *)mp->b_rptr;
1431 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1432 }
1433 icmph->icmph_type = ICMP_ECHO_REPLY;
1434 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutEchoReps);
1435 icmp_send_reply_v4(mp, ipha, icmph, ira);
1436 return (NULL);
1437
1438 case ICMP_ROUTER_ADVERTISEMENT:
1439 case ICMP_ROUTER_SOLICITATION:
1440 break;
1441 case ICMP_TIME_EXCEEDED:
1442 interested = B_TRUE; /* Pass up to transport */
1443 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimeExcds);
1444 break;
1445 case ICMP_PARAM_PROBLEM:
1446 interested = B_TRUE; /* Pass up to transport */
1447 BUMP_MIB(&ipst->ips_icmp_mib, icmpInParmProbs);
1448 break;
1449 case ICMP_TIME_STAMP_REQUEST:
1450 /* Response to Time Stamp Requests is local policy. */
1451 if (ipst->ips_ip_g_resp_to_timestamp) {
1452 if (ira->ira_flags & IRAF_MULTIBROADCAST)
1453 interested =
1454 ipst->ips_ip_g_resp_to_timestamp_bcast;
1455 else
1456 interested = B_TRUE;
1457 }
1458 if (!interested) {
1459 /* We never pass these to RAW sockets */
1460 freemsg(mp);
1461 return (NULL);
1462 }
1463
1464 /* Make sure we have enough of the packet */
1465 len_needed = ip_hdr_length + ICMPH_SIZE +
1466 3 * sizeof (uint32_t);
1467
1468 if (mp->b_wptr - mp->b_rptr < len_needed) {
1469 ipha = ip_pullup(mp, len_needed, ira);
1470 if (ipha == NULL) {
1471 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1472 ip_drop_input("ipIfStatsInDiscards - ip_pullup",
1473 mp, ill);
1474 freemsg(mp);
1475 return (NULL);
1476 }
1477 /* Refresh following the pullup. */
1478 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1479 }
1480 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimestamps);
1481 /* Check db_ref to make sure we can modify the packet. */
1482 if (mp->b_datap->db_ref > 1) {
1483 mblk_t *mp1;
1484
1485 mp1 = copymsg(mp);
1486 freemsg(mp);
1487 if (!mp1) {
1488 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1489 return (NULL);
1490 }
1491 mp = mp1;
1492 ipha = (ipha_t *)mp->b_rptr;
1493 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1494 }
1495 icmph->icmph_type = ICMP_TIME_STAMP_REPLY;
1496 tsp = (uint32_t *)&icmph[1];
1497 tsp++; /* Skip past 'originate time' */
1498 /* Compute # of milliseconds since midnight */
1499 gethrestime(&now);
1500 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
1501 NSEC2MSEC(now.tv_nsec);
1502 *tsp++ = htonl(ts); /* Lay in 'receive time' */
1503 *tsp++ = htonl(ts); /* Lay in 'send time' */
1504 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutTimestampReps);
1505 icmp_send_reply_v4(mp, ipha, icmph, ira);
1506 return (NULL);
1507
1508 case ICMP_TIME_STAMP_REPLY:
1509 BUMP_MIB(&ipst->ips_icmp_mib, icmpInTimestampReps);
1510 break;
1511 case ICMP_INFO_REQUEST:
1512 /* Per RFC 1122 3.2.2.7, ignore this. */
1513 case ICMP_INFO_REPLY:
1514 break;
1515 case ICMP_ADDRESS_MASK_REQUEST:
1516 if (ira->ira_flags & IRAF_MULTIBROADCAST) {
1517 interested =
1518 ipst->ips_ip_respond_to_address_mask_broadcast;
1519 } else {
1520 interested = B_TRUE;
1521 }
1522 if (!interested) {
1523 /* We never pass these to RAW sockets */
1524 freemsg(mp);
1525 return (NULL);
1526 }
1527 len_needed = ip_hdr_length + ICMPH_SIZE + IP_ADDR_LEN;
1528 if (mp->b_wptr - mp->b_rptr < len_needed) {
1529 ipha = ip_pullup(mp, len_needed, ira);
1530 if (ipha == NULL) {
1531 BUMP_MIB(ill->ill_ip_mib,
1532 ipIfStatsInTruncatedPkts);
1533 ip_drop_input("ipIfStatsInTruncatedPkts", mp,
1534 ill);
1535 freemsg(mp);
1536 return (NULL);
1537 }
1538 /* Refresh following the pullup. */
1539 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1540 }
1541 BUMP_MIB(&ipst->ips_icmp_mib, icmpInAddrMasks);
1542 /* Check db_ref to make sure we can modify the packet. */
1543 if (mp->b_datap->db_ref > 1) {
1544 mblk_t *mp1;
1545
1546 mp1 = copymsg(mp);
1547 freemsg(mp);
1548 if (!mp1) {
1549 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
1550 return (NULL);
1551 }
1552 mp = mp1;
1553 ipha = (ipha_t *)mp->b_rptr;
1554 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1555 }
1556 /*
1557 * Need the ipif with the mask be the same as the source
1558 * address of the mask reply. For unicast we have a specific
1559 * ipif. For multicast/broadcast we only handle onlink
1560 * senders, and use the source address to pick an ipif.
1561 */
1562 ipif = ipif_lookup_addr(ipha->ipha_dst, ill, zoneid, ipst);
1563 if (ipif == NULL) {
1564 /* Broadcast or multicast */
1565 ipif = ipif_lookup_remote(ill, ipha->ipha_src, zoneid);
1566 if (ipif == NULL) {
1567 freemsg(mp);
1568 return (NULL);
1569 }
1570 }
1571 icmph->icmph_type = ICMP_ADDRESS_MASK_REPLY;
1572 bcopy(&ipif->ipif_net_mask, &icmph[1], IP_ADDR_LEN);
1573 ipif_refrele(ipif);
1574 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutAddrMaskReps);
1575 icmp_send_reply_v4(mp, ipha, icmph, ira);
1576 return (NULL);
1577
1578 case ICMP_ADDRESS_MASK_REPLY:
1579 BUMP_MIB(&ipst->ips_icmp_mib, icmpInAddrMaskReps);
1580 break;
1581 default:
1582 interested = B_TRUE; /* Pass up to transport */
1583 BUMP_MIB(&ipst->ips_icmp_mib, icmpInUnknowns);
1584 break;
1585 }
1586 /*
1587 * See if there is an ICMP client to avoid an extra copymsg/freemsg
1588 * if there isn't one.
1589 */
1590 if (ipst->ips_ipcl_proto_fanout_v4[IPPROTO_ICMP].connf_head != NULL) {
1591 /* If there is an ICMP client and we want one too, copy it. */
1592
1593 if (!interested) {
1594 /* Caller will deliver to RAW sockets */
1595 return (mp);
1596 }
1597 mp_ret = copymsg(mp);
1598 if (mp_ret == NULL) {
1599 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1600 ip_drop_input("ipIfStatsInDiscards - copymsg", mp, ill);
1601 }
1602 } else if (!interested) {
1603 /* Neither we nor raw sockets are interested. Drop packet now */
1604 freemsg(mp);
1605 return (NULL);
1606 }
1607
1608 /*
1609 * ICMP error or redirect packet. Make sure we have enough of
1610 * the header and that db_ref == 1 since we might end up modifying
1611 * the packet.
1612 */
1613 if (mp->b_cont != NULL) {
1614 if (ip_pullup(mp, -1, ira) == NULL) {
1615 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1616 ip_drop_input("ipIfStatsInDiscards - ip_pullup",
1617 mp, ill);
1618 freemsg(mp);
1619 return (mp_ret);
1620 }
1621 }
1622
1623 if (mp->b_datap->db_ref > 1) {
1624 mblk_t *mp1;
1625
1626 mp1 = copymsg(mp);
1627 if (mp1 == NULL) {
1628 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1629 ip_drop_input("ipIfStatsInDiscards - copymsg", mp, ill);
1630 freemsg(mp);
1631 return (mp_ret);
1632 }
1633 freemsg(mp);
1634 mp = mp1;
1635 }
1636
1637 /*
1638 * In case mp has changed, verify the message before any further
1639 * processes.
1640 */
1641 ipha = (ipha_t *)mp->b_rptr;
1642 icmph = (icmph_t *)&mp->b_rptr[ip_hdr_length];
1643 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
1644 freemsg(mp);
1645 return (mp_ret);
1646 }
1647
1648 switch (icmph->icmph_type) {
1649 case ICMP_REDIRECT:
1650 icmp_redirect_v4(mp, ipha, icmph, ira);
1651 break;
1652 case ICMP_DEST_UNREACHABLE:
1653 if (icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED) {
1654 /* Update DCE and adjust MTU is icmp header if needed */
1655 icmp_inbound_too_big_v4(icmph, ira);
1656 }
1657 /* FALLTHRU */
1658 default:
1659 icmp_inbound_error_fanout_v4(mp, icmph, ira);
1660 break;
1661 }
1662 return (mp_ret);
1663 }
1664
1665 /*
1666 * Send an ICMP echo, timestamp or address mask reply.
1667 * The caller has already updated the payload part of the packet.
1668 * We handle the ICMP checksum, IP source address selection and feed
1669 * the packet into ip_output_simple.
1670 */
1671 static void
1672 icmp_send_reply_v4(mblk_t *mp, ipha_t *ipha, icmph_t *icmph,
1673 ip_recv_attr_t *ira)
1674 {
1675 uint_t ip_hdr_length = ira->ira_ip_hdr_length;
1676 ill_t *ill = ira->ira_ill;
1677 ip_stack_t *ipst = ill->ill_ipst;
1678 ip_xmit_attr_t ixas;
1679
1680 /* Send out an ICMP packet */
1681 icmph->icmph_checksum = 0;
1682 icmph->icmph_checksum = IP_CSUM(mp, ip_hdr_length, 0);
1683 /* Reset time to live. */
1684 ipha->ipha_ttl = ipst->ips_ip_def_ttl;
1685 {
1686 /* Swap source and destination addresses */
1687 ipaddr_t tmp;
1688
1689 tmp = ipha->ipha_src;
1690 ipha->ipha_src = ipha->ipha_dst;
1691 ipha->ipha_dst = tmp;
1692 }
1693 ipha->ipha_ident = 0;
1694 if (!IS_SIMPLE_IPH(ipha))
1695 icmp_options_update(ipha);
1696
1697 bzero(&ixas, sizeof (ixas));
1698 ixas.ixa_flags = IXAF_BASIC_SIMPLE_V4;
1699 ixas.ixa_zoneid = ira->ira_zoneid;
1700 ixas.ixa_cred = kcred;
1701 ixas.ixa_cpid = NOPID;
1702 ixas.ixa_tsl = ira->ira_tsl; /* Behave as a multi-level responder */
1703 ixas.ixa_ifindex = 0;
1704 ixas.ixa_ipst = ipst;
1705 ixas.ixa_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
1706
1707 if (!(ira->ira_flags & IRAF_IPSEC_SECURE)) {
1708 /*
1709 * This packet should go out the same way as it
1710 * came in i.e in clear, independent of the IPsec policy
1711 * for transmitting packets.
1712 */
1713 ixas.ixa_flags |= IXAF_NO_IPSEC;
1714 } else {
1715 if (!ipsec_in_to_out(ira, &ixas, mp, ipha, NULL)) {
1716 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1717 /* Note: mp already consumed and ip_drop_packet done */
1718 return;
1719 }
1720 }
1721 if (ira->ira_flags & IRAF_MULTIBROADCAST) {
1722 /*
1723 * Not one or our addresses (IRE_LOCALs), thus we let
1724 * ip_output_simple pick the source.
1725 */
1726 ipha->ipha_src = INADDR_ANY;
1727 ixas.ixa_flags |= IXAF_SET_SOURCE;
1728 }
1729 /* Should we send with DF and use dce_pmtu? */
1730 if (ipst->ips_ipv4_icmp_return_pmtu) {
1731 ixas.ixa_flags |= IXAF_PMTU_DISCOVERY;
1732 ipha->ipha_fragment_offset_and_flags |= IPH_DF_HTONS;
1733 }
1734
1735 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutMsgs);
1736
1737 (void) ip_output_simple(mp, &ixas);
1738 ixa_cleanup(&ixas);
1739 }
1740
1741 /*
1742 * Verify the ICMP messages for either for ICMP error or redirect packet.
1743 * The caller should have fully pulled up the message. If it's a redirect
1744 * packet, only basic checks on IP header will be done; otherwise, verify
1745 * the packet by looking at the included ULP header.
1746 *
1747 * Called before icmp_inbound_error_fanout_v4 is called.
1748 */
1749 static boolean_t
1750 icmp_inbound_verify_v4(mblk_t *mp, icmph_t *icmph, ip_recv_attr_t *ira)
1751 {
1752 ill_t *ill = ira->ira_ill;
1753 int hdr_length;
1754 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
1755 conn_t *connp;
1756 ipha_t *ipha; /* Inner IP header */
1757
1758 ipha = (ipha_t *)&icmph[1];
1759 if ((uchar_t *)ipha + IP_SIMPLE_HDR_LENGTH > mp->b_wptr)
1760 goto truncated;
1761
1762 hdr_length = IPH_HDR_LENGTH(ipha);
1763
1764 if ((IPH_HDR_VERSION(ipha) != IPV4_VERSION))
1765 goto discard_pkt;
1766
1767 if (hdr_length < sizeof (ipha_t))
1768 goto truncated;
1769
1770 if ((uchar_t *)ipha + hdr_length > mp->b_wptr)
1771 goto truncated;
1772
1773 /*
1774 * Stop here for ICMP_REDIRECT.
1775 */
1776 if (icmph->icmph_type == ICMP_REDIRECT)
1777 return (B_TRUE);
1778
1779 /*
1780 * ICMP errors only.
1781 */
1782 switch (ipha->ipha_protocol) {
1783 case IPPROTO_UDP:
1784 /*
1785 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1786 * transport header.
1787 */
1788 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1789 mp->b_wptr)
1790 goto truncated;
1791 break;
1792 case IPPROTO_TCP: {
1793 tcpha_t *tcpha;
1794
1795 /*
1796 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1797 * transport header.
1798 */
1799 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1800 mp->b_wptr)
1801 goto truncated;
1802
1803 tcpha = (tcpha_t *)((uchar_t *)ipha + hdr_length);
1804 connp = ipcl_tcp_lookup_reversed_ipv4(ipha, tcpha, TCPS_LISTEN,
1805 ipst);
1806 if (connp == NULL)
1807 goto discard_pkt;
1808
1809 if ((connp->conn_verifyicmp != NULL) &&
1810 !connp->conn_verifyicmp(connp, tcpha, icmph, NULL, ira)) {
1811 CONN_DEC_REF(connp);
1812 goto discard_pkt;
1813 }
1814 CONN_DEC_REF(connp);
1815 break;
1816 }
1817 case IPPROTO_SCTP:
1818 /*
1819 * Verify we have at least ICMP_MIN_TP_HDR_LEN bytes of
1820 * transport header.
1821 */
1822 if ((uchar_t *)ipha + hdr_length + ICMP_MIN_TP_HDR_LEN >
1823 mp->b_wptr)
1824 goto truncated;
1825 break;
1826 case IPPROTO_ESP:
1827 case IPPROTO_AH:
1828 break;
1829 case IPPROTO_ENCAP:
1830 if ((uchar_t *)ipha + hdr_length + sizeof (ipha_t) >
1831 mp->b_wptr)
1832 goto truncated;
1833 break;
1834 default:
1835 break;
1836 }
1837
1838 return (B_TRUE);
1839
1840 discard_pkt:
1841 /* Bogus ICMP error. */
1842 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
1843 return (B_FALSE);
1844
1845 truncated:
1846 /* We pulled up everthing already. Must be truncated */
1847 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
1848 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
1849 return (B_FALSE);
1850 }
1851
1852 /* Table from RFC 1191 */
1853 static int icmp_frag_size_table[] =
1854 { 32000, 17914, 8166, 4352, 2002, 1496, 1006, 508, 296, 68 };
1855
1856 /*
1857 * Process received ICMP Packet too big.
1858 * Just handles the DCE create/update, including using the above table of
1859 * PMTU guesses. The caller is responsible for validating the packet before
1860 * passing it in and also to fanout the ICMP error to any matching transport
1861 * conns. Assumes the message has been fully pulled up and verified.
1862 *
1863 * Before getting here, the caller has called icmp_inbound_verify_v4()
1864 * that should have verified with ULP to prevent undoing the changes we're
1865 * going to make to DCE. For example, TCP might have verified that the packet
1866 * which generated error is in the send window.
1867 *
1868 * In some cases modified this MTU in the ICMP header packet; the caller
1869 * should pass to the matching ULP after this returns.
1870 */
1871 static void
1872 icmp_inbound_too_big_v4(icmph_t *icmph, ip_recv_attr_t *ira)
1873 {
1874 dce_t *dce;
1875 int old_mtu;
1876 int mtu, orig_mtu;
1877 ipaddr_t dst;
1878 boolean_t disable_pmtud;
1879 ill_t *ill = ira->ira_ill;
1880 ip_stack_t *ipst = ill->ill_ipst;
1881 uint_t hdr_length;
1882 ipha_t *ipha;
1883
1884 /* Caller already pulled up everything. */
1885 ipha = (ipha_t *)&icmph[1];
1886 ASSERT(icmph->icmph_type == ICMP_DEST_UNREACHABLE &&
1887 icmph->icmph_code == ICMP_FRAGMENTATION_NEEDED);
1888 ASSERT(ill != NULL);
1889
1890 hdr_length = IPH_HDR_LENGTH(ipha);
1891
1892 /*
1893 * We handle path MTU for source routed packets since the DCE
1894 * is looked up using the final destination.
1895 */
1896 dst = ip_get_dst(ipha);
1897
1898 dce = dce_lookup_and_add_v4(dst, ipst);
1899 if (dce == NULL) {
1900 /* Couldn't add a unique one - ENOMEM */
1901 ip1dbg(("icmp_inbound_too_big_v4: no dce for 0x%x\n",
1902 ntohl(dst)));
1903 return;
1904 }
1905
1906 /* Check for MTU discovery advice as described in RFC 1191 */
1907 mtu = ntohs(icmph->icmph_du_mtu);
1908 orig_mtu = mtu;
1909 disable_pmtud = B_FALSE;
1910
1911 mutex_enter(&dce->dce_lock);
1912 if (dce->dce_flags & DCEF_PMTU)
1913 old_mtu = dce->dce_pmtu;
1914 else
1915 old_mtu = ill->ill_mtu;
1916
1917 if (icmph->icmph_du_zero != 0 || mtu < ipst->ips_ip_pmtu_min) {
1918 uint32_t length;
1919 int i;
1920
1921 /*
1922 * Use the table from RFC 1191 to figure out
1923 * the next "plateau" based on the length in
1924 * the original IP packet.
1925 */
1926 length = ntohs(ipha->ipha_length);
1927 DTRACE_PROBE2(ip4__pmtu__guess, dce_t *, dce,
1928 uint32_t, length);
1929 if (old_mtu <= length &&
1930 old_mtu >= length - hdr_length) {
1931 /*
1932 * Handle broken BSD 4.2 systems that
1933 * return the wrong ipha_length in ICMP
1934 * errors.
1935 */
1936 ip1dbg(("Wrong mtu: sent %d, dce %d\n",
1937 length, old_mtu));
1938 length -= hdr_length;
1939 }
1940 for (i = 0; i < A_CNT(icmp_frag_size_table); i++) {
1941 if (length > icmp_frag_size_table[i])
1942 break;
1943 }
1944 if (i == A_CNT(icmp_frag_size_table)) {
1945 /* Smaller than IP_MIN_MTU! */
1946 ip1dbg(("Too big for packet size %d\n",
1947 length));
1948 disable_pmtud = B_TRUE;
1949 mtu = ipst->ips_ip_pmtu_min;
1950 } else {
1951 mtu = icmp_frag_size_table[i];
1952 ip1dbg(("Calculated mtu %d, packet size %d, "
1953 "before %d\n", mtu, length, old_mtu));
1954 if (mtu < ipst->ips_ip_pmtu_min) {
1955 mtu = ipst->ips_ip_pmtu_min;
1956 disable_pmtud = B_TRUE;
1957 }
1958 }
1959 }
1960 if (disable_pmtud)
1961 dce->dce_flags |= DCEF_TOO_SMALL_PMTU;
1962 else
1963 dce->dce_flags &= ~DCEF_TOO_SMALL_PMTU;
1964
1965 dce->dce_pmtu = MIN(old_mtu, mtu);
1966 /* Prepare to send the new max frag size for the ULP. */
1967 icmph->icmph_du_zero = 0;
1968 icmph->icmph_du_mtu = htons((uint16_t)dce->dce_pmtu);
1969 DTRACE_PROBE4(ip4__pmtu__change, icmph_t *, icmph, dce_t *,
1970 dce, int, orig_mtu, int, mtu);
1971
1972 /* We now have a PMTU for sure */
1973 dce->dce_flags |= DCEF_PMTU;
1974 dce->dce_last_change_time = TICK_TO_SEC(ddi_get_lbolt64());
1975 mutex_exit(&dce->dce_lock);
1976 /*
1977 * After dropping the lock the new value is visible to everyone.
1978 * Then we bump the generation number so any cached values reinspect
1979 * the dce_t.
1980 */
1981 dce_increment_generation(dce);
1982 dce_refrele(dce);
1983 }
1984
1985 /*
1986 * If the packet in error is Self-Encapsulated, icmp_inbound_error_fanout_v4
1987 * calls this function.
1988 */
1989 static mblk_t *
1990 icmp_inbound_self_encap_error_v4(mblk_t *mp, ipha_t *ipha, ipha_t *in_ipha)
1991 {
1992 int length;
1993
1994 ASSERT(mp->b_datap->db_type == M_DATA);
1995
1996 /* icmp_inbound_v4 has already pulled up the whole error packet */
1997 ASSERT(mp->b_cont == NULL);
1998
1999 /*
2000 * The length that we want to overlay is the inner header
2001 * and what follows it.
2002 */
2003 length = msgdsize(mp) - ((uchar_t *)in_ipha - mp->b_rptr);
2004
2005 /*
2006 * Overlay the inner header and whatever follows it over the
2007 * outer header.
2008 */
2009 bcopy((uchar_t *)in_ipha, (uchar_t *)ipha, length);
2010
2011 /* Adjust for what we removed */
2012 mp->b_wptr -= (uchar_t *)in_ipha - (uchar_t *)ipha;
2013 return (mp);
2014 }
2015
2016 /*
2017 * Try to pass the ICMP message upstream in case the ULP cares.
2018 *
2019 * If the packet that caused the ICMP error is secure, we send
2020 * it to AH/ESP to make sure that the attached packet has a
2021 * valid association. ipha in the code below points to the
2022 * IP header of the packet that caused the error.
2023 *
2024 * For IPsec cases, we let the next-layer-up (which has access to
2025 * cached policy on the conn_t, or can query the SPD directly)
2026 * subtract out any IPsec overhead if they must. We therefore make no
2027 * adjustments here for IPsec overhead.
2028 *
2029 * IFN could have been generated locally or by some router.
2030 *
2031 * LOCAL : ire_send_wire (before calling ipsec_out_process) can call
2032 * icmp_frag_needed/icmp_pkt2big_v6 to generated a local IFN.
2033 * This happens because IP adjusted its value of MTU on an
2034 * earlier IFN message and could not tell the upper layer,
2035 * the new adjusted value of MTU e.g. Packet was encrypted
2036 * or there was not enough information to fanout to upper
2037 * layers. Thus on the next outbound datagram, ire_send_wire
2038 * generates the IFN, where IPsec processing has *not* been
2039 * done.
2040 *
2041 * Note that we retain ixa_fragsize across IPsec thus once
2042 * we have picking ixa_fragsize and entered ipsec_out_process we do
2043 * no change the fragsize even if the path MTU changes before
2044 * we reach ip_output_post_ipsec.
2045 *
2046 * In the local case, IRAF_LOOPBACK will be set indicating
2047 * that IFN was generated locally.
2048 *
2049 * ROUTER : IFN could be secure or non-secure.
2050 *
2051 * * SECURE : We use the IPSEC_IN to fanout to AH/ESP if the
2052 * packet in error has AH/ESP headers to validate the AH/ESP
2053 * headers. AH/ESP will verify whether there is a valid SA or
2054 * not and send it back. We will fanout again if we have more
2055 * data in the packet.
2056 *
2057 * If the packet in error does not have AH/ESP, we handle it
2058 * like any other case.
2059 *
2060 * * NON_SECURE : If the packet in error has AH/ESP headers, we send it
2061 * up to AH/ESP for validation. AH/ESP will verify whether there is a
2062 * valid SA or not and send it back. We will fanout again if
2063 * we have more data in the packet.
2064 *
2065 * If the packet in error does not have AH/ESP, we handle it
2066 * like any other case.
2067 *
2068 * The caller must have called icmp_inbound_verify_v4.
2069 */
2070 static void
2071 icmp_inbound_error_fanout_v4(mblk_t *mp, icmph_t *icmph, ip_recv_attr_t *ira)
2072 {
2073 uint16_t *up; /* Pointer to ports in ULP header */
2074 uint32_t ports; /* reversed ports for fanout */
2075 ipha_t ripha; /* With reversed addresses */
2076 ipha_t *ipha; /* Inner IP header */
2077 uint_t hdr_length; /* Inner IP header length */
2078 tcpha_t *tcpha;
2079 conn_t *connp;
2080 ill_t *ill = ira->ira_ill;
2081 ip_stack_t *ipst = ill->ill_ipst;
2082 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
2083 ill_t *rill = ira->ira_rill;
2084
2085 /* Caller already pulled up everything. */
2086 ipha = (ipha_t *)&icmph[1];
2087 ASSERT((uchar_t *)&ipha[1] <= mp->b_wptr);
2088 ASSERT(mp->b_cont == NULL);
2089
2090 hdr_length = IPH_HDR_LENGTH(ipha);
2091 ira->ira_protocol = ipha->ipha_protocol;
2092
2093 /*
2094 * We need a separate IP header with the source and destination
2095 * addresses reversed to do fanout/classification because the ipha in
2096 * the ICMP error is in the form we sent it out.
2097 */
2098 ripha.ipha_src = ipha->ipha_dst;
2099 ripha.ipha_dst = ipha->ipha_src;
2100 ripha.ipha_protocol = ipha->ipha_protocol;
2101 ripha.ipha_version_and_hdr_length = ipha->ipha_version_and_hdr_length;
2102
2103 ip2dbg(("icmp_inbound_error_v4: proto %d %x to %x: %d/%d\n",
2104 ripha.ipha_protocol, ntohl(ipha->ipha_src),
2105 ntohl(ipha->ipha_dst),
2106 icmph->icmph_type, icmph->icmph_code));
2107
2108 switch (ipha->ipha_protocol) {
2109 case IPPROTO_UDP:
2110 up = (uint16_t *)((uchar_t *)ipha + hdr_length);
2111
2112 /* Attempt to find a client stream based on port. */
2113 ip2dbg(("icmp_inbound_error_v4: UDP ports %d to %d\n",
2114 ntohs(up[0]), ntohs(up[1])));
2115
2116 /* Note that we send error to all matches. */
2117 ira->ira_flags |= IRAF_ICMP_ERROR;
2118 ip_fanout_udp_multi_v4(mp, &ripha, up[0], up[1], ira);
2119 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2120 return;
2121
2122 case IPPROTO_TCP:
2123 /*
2124 * Find a TCP client stream for this packet.
2125 * Note that we do a reverse lookup since the header is
2126 * in the form we sent it out.
2127 */
2128 tcpha = (tcpha_t *)((uchar_t *)ipha + hdr_length);
2129 connp = ipcl_tcp_lookup_reversed_ipv4(ipha, tcpha, TCPS_LISTEN,
2130 ipst);
2131 if (connp == NULL)
2132 goto discard_pkt;
2133
2134 if (CONN_INBOUND_POLICY_PRESENT(connp, ipss) ||
2135 (ira->ira_flags & IRAF_IPSEC_SECURE)) {
2136 mp = ipsec_check_inbound_policy(mp, connp,
2137 ipha, NULL, ira);
2138 if (mp == NULL) {
2139 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2140 /* Note that mp is NULL */
2141 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2142 CONN_DEC_REF(connp);
2143 return;
2144 }
2145 }
2146
2147 ira->ira_flags |= IRAF_ICMP_ERROR;
2148 ira->ira_ill = ira->ira_rill = NULL;
2149 if (IPCL_IS_TCP(connp)) {
2150 SQUEUE_ENTER_ONE(connp->conn_sqp, mp,
2151 connp->conn_recvicmp, connp, ira, SQ_FILL,
2152 SQTAG_TCP_INPUT_ICMP_ERR);
2153 } else {
2154 /* Not TCP; must be SOCK_RAW, IPPROTO_TCP */
2155 (connp->conn_recv)(connp, mp, NULL, ira);
2156 CONN_DEC_REF(connp);
2157 }
2158 ira->ira_ill = ill;
2159 ira->ira_rill = rill;
2160 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2161 return;
2162
2163 case IPPROTO_SCTP:
2164 up = (uint16_t *)((uchar_t *)ipha + hdr_length);
2165 /* Find a SCTP client stream for this packet. */
2166 ((uint16_t *)&ports)[0] = up[1];
2167 ((uint16_t *)&ports)[1] = up[0];
2168
2169 ira->ira_flags |= IRAF_ICMP_ERROR;
2170 ip_fanout_sctp(mp, &ripha, NULL, ports, ira);
2171 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2172 return;
2173
2174 case IPPROTO_ESP:
2175 case IPPROTO_AH:
2176 if (!ipsec_loaded(ipss)) {
2177 ip_proto_not_sup(mp, ira);
2178 return;
2179 }
2180
2181 if (ipha->ipha_protocol == IPPROTO_ESP)
2182 mp = ipsecesp_icmp_error(mp, ira);
2183 else
2184 mp = ipsecah_icmp_error(mp, ira);
2185 if (mp == NULL)
2186 return;
2187
2188 /* Just in case ipsec didn't preserve the NULL b_cont */
2189 if (mp->b_cont != NULL) {
2190 if (!pullupmsg(mp, -1))
2191 goto discard_pkt;
2192 }
2193
2194 /*
2195 * Note that ira_pktlen and ira_ip_hdr_length are no longer
2196 * correct, but we don't use them any more here.
2197 *
2198 * If succesful, the mp has been modified to not include
2199 * the ESP/AH header so we can fanout to the ULP's icmp
2200 * error handler.
2201 */
2202 if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH)
2203 goto truncated;
2204
2205 /* Verify the modified message before any further processes. */
2206 ipha = (ipha_t *)mp->b_rptr;
2207 hdr_length = IPH_HDR_LENGTH(ipha);
2208 icmph = (icmph_t *)&mp->b_rptr[hdr_length];
2209 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
2210 freemsg(mp);
2211 return;
2212 }
2213
2214 icmp_inbound_error_fanout_v4(mp, icmph, ira);
2215 return;
2216
2217 case IPPROTO_ENCAP: {
2218 /* Look for self-encapsulated packets that caused an error */
2219 ipha_t *in_ipha;
2220
2221 /*
2222 * Caller has verified that length has to be
2223 * at least the size of IP header.
2224 */
2225 ASSERT(hdr_length >= sizeof (ipha_t));
2226 /*
2227 * Check the sanity of the inner IP header like
2228 * we did for the outer header.
2229 */
2230 in_ipha = (ipha_t *)((uchar_t *)ipha + hdr_length);
2231 if ((IPH_HDR_VERSION(in_ipha) != IPV4_VERSION)) {
2232 goto discard_pkt;
2233 }
2234 if (IPH_HDR_LENGTH(in_ipha) < sizeof (ipha_t)) {
2235 goto discard_pkt;
2236 }
2237 /* Check for Self-encapsulated tunnels */
2238 if (in_ipha->ipha_src == ipha->ipha_src &&
2239 in_ipha->ipha_dst == ipha->ipha_dst) {
2240
2241 mp = icmp_inbound_self_encap_error_v4(mp, ipha,
2242 in_ipha);
2243 if (mp == NULL)
2244 goto discard_pkt;
2245
2246 /*
2247 * Just in case self_encap didn't preserve the NULL
2248 * b_cont
2249 */
2250 if (mp->b_cont != NULL) {
2251 if (!pullupmsg(mp, -1))
2252 goto discard_pkt;
2253 }
2254 /*
2255 * Note that ira_pktlen and ira_ip_hdr_length are no
2256 * longer correct, but we don't use them any more here.
2257 */
2258 if (mp->b_wptr - mp->b_rptr < IP_SIMPLE_HDR_LENGTH)
2259 goto truncated;
2260
2261 /*
2262 * Verify the modified message before any further
2263 * processes.
2264 */
2265 ipha = (ipha_t *)mp->b_rptr;
2266 hdr_length = IPH_HDR_LENGTH(ipha);
2267 icmph = (icmph_t *)&mp->b_rptr[hdr_length];
2268 if (!icmp_inbound_verify_v4(mp, icmph, ira)) {
2269 freemsg(mp);
2270 return;
2271 }
2272
2273 /*
2274 * The packet in error is self-encapsualted.
2275 * And we are finding it further encapsulated
2276 * which we could not have possibly generated.
2277 */
2278 if (ipha->ipha_protocol == IPPROTO_ENCAP) {
2279 goto discard_pkt;
2280 }
2281 icmp_inbound_error_fanout_v4(mp, icmph, ira);
2282 return;
2283 }
2284 /* No self-encapsulated */
2285 /* FALLTHRU */
2286 }
2287 case IPPROTO_IPV6:
2288 if ((connp = ipcl_iptun_classify_v4(&ripha.ipha_src,
2289 &ripha.ipha_dst, ipst)) != NULL) {
2290 ira->ira_flags |= IRAF_ICMP_ERROR;
2291 connp->conn_recvicmp(connp, mp, NULL, ira);
2292 CONN_DEC_REF(connp);
2293 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2294 return;
2295 }
2296 /*
2297 * No IP tunnel is interested, fallthrough and see
2298 * if a raw socket will want it.
2299 */
2300 /* FALLTHRU */
2301 default:
2302 ira->ira_flags |= IRAF_ICMP_ERROR;
2303 ip_fanout_proto_v4(mp, &ripha, ira);
2304 ira->ira_flags &= ~IRAF_ICMP_ERROR;
2305 return;
2306 }
2307 /* NOTREACHED */
2308 discard_pkt:
2309 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
2310 ip1dbg(("icmp_inbound_error_fanout_v4: drop pkt\n"));
2311 ip_drop_input("ipIfStatsInDiscards", mp, ill);
2312 freemsg(mp);
2313 return;
2314
2315 truncated:
2316 /* We pulled up everthing already. Must be truncated */
2317 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
2318 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
2319 freemsg(mp);
2320 }
2321
2322 /*
2323 * Common IP options parser.
2324 *
2325 * Setup routine: fill in *optp with options-parsing state, then
2326 * tail-call ipoptp_next to return the first option.
2327 */
2328 uint8_t
2329 ipoptp_first(ipoptp_t *optp, ipha_t *ipha)
2330 {
2331 uint32_t totallen; /* total length of all options */
2332
2333 totallen = ipha->ipha_version_and_hdr_length -
2334 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
2335 totallen <<= 2;
2336 optp->ipoptp_next = (uint8_t *)(&ipha[1]);
2337 optp->ipoptp_end = optp->ipoptp_next + totallen;
2338 optp->ipoptp_flags = 0;
2339 return (ipoptp_next(optp));
2340 }
2341
2342 /* Like above but without an ipha_t */
2343 uint8_t
2344 ipoptp_first2(ipoptp_t *optp, uint32_t totallen, uint8_t *opt)
2345 {
2346 optp->ipoptp_next = opt;
2347 optp->ipoptp_end = optp->ipoptp_next + totallen;
2348 optp->ipoptp_flags = 0;
2349 return (ipoptp_next(optp));
2350 }
2351
2352 /*
2353 * Common IP options parser: extract next option.
2354 */
2355 uint8_t
2356 ipoptp_next(ipoptp_t *optp)
2357 {
2358 uint8_t *end = optp->ipoptp_end;
2359 uint8_t *cur = optp->ipoptp_next;
2360 uint8_t opt, len, pointer;
2361
2362 /*
2363 * If cur > end already, then the ipoptp_end or ipoptp_next pointer
2364 * has been corrupted.
2365 */
2366 ASSERT(cur <= end);
2367
2368 if (cur == end)
2369 return (IPOPT_EOL);
2370
2371 opt = cur[IPOPT_OPTVAL];
2372
2373 /*
2374 * Skip any NOP options.
2375 */
2376 while (opt == IPOPT_NOP) {
2377 cur++;
2378 if (cur == end)
2379 return (IPOPT_EOL);
2380 opt = cur[IPOPT_OPTVAL];
2381 }
2382
2383 if (opt == IPOPT_EOL)
2384 return (IPOPT_EOL);
2385
2386 /*
2387 * Option requiring a length.
2388 */
2389 if ((cur + 1) >= end) {
2390 optp->ipoptp_flags |= IPOPTP_ERROR;
2391 return (IPOPT_EOL);
2392 }
2393 len = cur[IPOPT_OLEN];
2394 if (len < 2) {
2395 optp->ipoptp_flags |= IPOPTP_ERROR;
2396 return (IPOPT_EOL);
2397 }
2398 optp->ipoptp_cur = cur;
2399 optp->ipoptp_len = len;
2400 optp->ipoptp_next = cur + len;
2401 if (cur + len > end) {
2402 optp->ipoptp_flags |= IPOPTP_ERROR;
2403 return (IPOPT_EOL);
2404 }
2405
2406 /*
2407 * For the options which require a pointer field, make sure
2408 * its there, and make sure it points to either something
2409 * inside this option, or the end of the option.
2410 */
2411 switch (opt) {
2412 case IPOPT_RR:
2413 case IPOPT_TS:
2414 case IPOPT_LSRR:
2415 case IPOPT_SSRR:
2416 if (len <= IPOPT_OFFSET) {
2417 optp->ipoptp_flags |= IPOPTP_ERROR;
2418 return (opt);
2419 }
2420 pointer = cur[IPOPT_OFFSET];
2421 if (pointer - 1 > len) {
2422 optp->ipoptp_flags |= IPOPTP_ERROR;
2423 return (opt);
2424 }
2425 break;
2426 }
2427
2428 /*
2429 * Sanity check the pointer field based on the type of the
2430 * option.
2431 */
2432 switch (opt) {
2433 case IPOPT_RR:
2434 case IPOPT_SSRR:
2435 case IPOPT_LSRR:
2436 if (pointer < IPOPT_MINOFF_SR)
2437 optp->ipoptp_flags |= IPOPTP_ERROR;
2438 break;
2439 case IPOPT_TS:
2440 if (pointer < IPOPT_MINOFF_IT)
2441 optp->ipoptp_flags |= IPOPTP_ERROR;
2442 /*
2443 * Note that the Internet Timestamp option also
2444 * contains two four bit fields (the Overflow field,
2445 * and the Flag field), which follow the pointer
2446 * field. We don't need to check that these fields
2447 * fall within the length of the option because this
2448 * was implicitely done above. We've checked that the
2449 * pointer value is at least IPOPT_MINOFF_IT, and that
2450 * it falls within the option. Since IPOPT_MINOFF_IT >
2451 * IPOPT_POS_OV_FLG, we don't need the explicit check.
2452 */
2453 ASSERT(len > IPOPT_POS_OV_FLG);
2454 break;
2455 }
2456
2457 return (opt);
2458 }
2459
2460 /*
2461 * Use the outgoing IP header to create an IP_OPTIONS option the way
2462 * it was passed down from the application.
2463 *
2464 * This is compatible with BSD in that it returns
2465 * the reverse source route with the final destination
2466 * as the last entry. The first 4 bytes of the option
2467 * will contain the final destination.
2468 */
2469 int
2470 ip_opt_get_user(conn_t *connp, uchar_t *buf)
2471 {
2472 ipoptp_t opts;
2473 uchar_t *opt;
2474 uint8_t optval;
2475 uint8_t optlen;
2476 uint32_t len = 0;
2477 uchar_t *buf1 = buf;
2478 uint32_t totallen;
2479 ipaddr_t dst;
2480 ip_pkt_t *ipp = &connp->conn_xmit_ipp;
2481
2482 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
2483 return (0);
2484
2485 totallen = ipp->ipp_ipv4_options_len;
2486 if (totallen & 0x3)
2487 return (0);
2488
2489 buf += IP_ADDR_LEN; /* Leave room for final destination */
2490 len += IP_ADDR_LEN;
2491 bzero(buf1, IP_ADDR_LEN);
2492
2493 dst = connp->conn_faddr_v4;
2494
2495 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
2496 optval != IPOPT_EOL;
2497 optval = ipoptp_next(&opts)) {
2498 int off;
2499
2500 opt = opts.ipoptp_cur;
2501 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
2502 break;
2503 }
2504 optlen = opts.ipoptp_len;
2505
2506 switch (optval) {
2507 case IPOPT_SSRR:
2508 case IPOPT_LSRR:
2509
2510 /*
2511 * Insert destination as the first entry in the source
2512 * route and move down the entries on step.
2513 * The last entry gets placed at buf1.
2514 */
2515 buf[IPOPT_OPTVAL] = optval;
2516 buf[IPOPT_OLEN] = optlen;
2517 buf[IPOPT_OFFSET] = optlen;
2518
2519 off = optlen - IP_ADDR_LEN;
2520 if (off < 0) {
2521 /* No entries in source route */
2522 break;
2523 }
2524 /* Last entry in source route if not already set */
2525 if (dst == INADDR_ANY)
2526 bcopy(opt + off, buf1, IP_ADDR_LEN);
2527 off -= IP_ADDR_LEN;
2528
2529 while (off > 0) {
2530 bcopy(opt + off,
2531 buf + off + IP_ADDR_LEN,
2532 IP_ADDR_LEN);
2533 off -= IP_ADDR_LEN;
2534 }
2535 /* ipha_dst into first slot */
2536 bcopy(&dst, buf + off + IP_ADDR_LEN,
2537 IP_ADDR_LEN);
2538 buf += optlen;
2539 len += optlen;
2540 break;
2541
2542 default:
2543 bcopy(opt, buf, optlen);
2544 buf += optlen;
2545 len += optlen;
2546 break;
2547 }
2548 }
2549 done:
2550 /* Pad the resulting options */
2551 while (len & 0x3) {
2552 *buf++ = IPOPT_EOL;
2553 len++;
2554 }
2555 return (len);
2556 }
2557
2558 /*
2559 * Update any record route or timestamp options to include this host.
2560 * Reverse any source route option.
2561 * This routine assumes that the options are well formed i.e. that they
2562 * have already been checked.
2563 */
2564 static void
2565 icmp_options_update(ipha_t *ipha)
2566 {
2567 ipoptp_t opts;
2568 uchar_t *opt;
2569 uint8_t optval;
2570 ipaddr_t src; /* Our local address */
2571 ipaddr_t dst;
2572
2573 ip2dbg(("icmp_options_update\n"));
2574 src = ipha->ipha_src;
2575 dst = ipha->ipha_dst;
2576
2577 for (optval = ipoptp_first(&opts, ipha);
2578 optval != IPOPT_EOL;
2579 optval = ipoptp_next(&opts)) {
2580 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
2581 opt = opts.ipoptp_cur;
2582 ip2dbg(("icmp_options_update: opt %d, len %d\n",
2583 optval, opts.ipoptp_len));
2584 switch (optval) {
2585 int off1, off2;
2586 case IPOPT_SSRR:
2587 case IPOPT_LSRR:
2588 /*
2589 * Reverse the source route. The first entry
2590 * should be the next to last one in the current
2591 * source route (the last entry is our address).
2592 * The last entry should be the final destination.
2593 */
2594 off1 = IPOPT_MINOFF_SR - 1;
2595 off2 = opt[IPOPT_OFFSET] - IP_ADDR_LEN - 1;
2596 if (off2 < 0) {
2597 /* No entries in source route */
2598 ip1dbg((
2599 "icmp_options_update: bad src route\n"));
2600 break;
2601 }
2602 bcopy((char *)opt + off2, &dst, IP_ADDR_LEN);
2603 bcopy(&ipha->ipha_dst, (char *)opt + off2, IP_ADDR_LEN);
2604 bcopy(&dst, &ipha->ipha_dst, IP_ADDR_LEN);
2605 off2 -= IP_ADDR_LEN;
2606
2607 while (off1 < off2) {
2608 bcopy((char *)opt + off1, &src, IP_ADDR_LEN);
2609 bcopy((char *)opt + off2, (char *)opt + off1,
2610 IP_ADDR_LEN);
2611 bcopy(&src, (char *)opt + off2, IP_ADDR_LEN);
2612 off1 += IP_ADDR_LEN;
2613 off2 -= IP_ADDR_LEN;
2614 }
2615 opt[IPOPT_OFFSET] = IPOPT_MINOFF_SR;
2616 break;
2617 }
2618 }
2619 }
2620
2621 /*
2622 * Process received ICMP Redirect messages.
2623 * Assumes the caller has verified that the headers are in the pulled up mblk.
2624 * Consumes mp.
2625 */
2626 static void
2627 icmp_redirect_v4(mblk_t *mp, ipha_t *ipha, icmph_t *icmph, ip_recv_attr_t *ira)
2628 {
2629 ire_t *ire, *nire;
2630 ire_t *prev_ire;
2631 ipaddr_t src, dst, gateway;
2632 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2633 ipha_t *inner_ipha; /* Inner IP header */
2634
2635 /* Caller already pulled up everything. */
2636 inner_ipha = (ipha_t *)&icmph[1];
2637 src = ipha->ipha_src;
2638 dst = inner_ipha->ipha_dst;
2639 gateway = icmph->icmph_rd_gateway;
2640 /* Make sure the new gateway is reachable somehow. */
2641 ire = ire_ftable_lookup_v4(gateway, 0, 0, IRE_ONLINK, NULL,
2642 ALL_ZONES, NULL, MATCH_IRE_TYPE, 0, ipst, NULL);
2643 /*
2644 * Make sure we had a route for the dest in question and that
2645 * that route was pointing to the old gateway (the source of the
2646 * redirect packet.)
2647 * We do longest match and then compare ire_gateway_addr below.
2648 */
2649 prev_ire = ire_ftable_lookup_v4(dst, 0, 0, 0, NULL, ALL_ZONES,
2650 NULL, MATCH_IRE_DSTONLY, 0, ipst, NULL);
2651 /*
2652 * Check that
2653 * the redirect was not from ourselves
2654 * the new gateway and the old gateway are directly reachable
2655 */
2656 if (prev_ire == NULL || ire == NULL ||
2657 (prev_ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK)) ||
2658 (prev_ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) ||
2659 !(ire->ire_type & IRE_IF_ALL) ||
2660 prev_ire->ire_gateway_addr != src) {
2661 BUMP_MIB(&ipst->ips_icmp_mib, icmpInBadRedirects);
2662 ip_drop_input("icmpInBadRedirects - ire", mp, ira->ira_ill);
2663 freemsg(mp);
2664 if (ire != NULL)
2665 ire_refrele(ire);
2666 if (prev_ire != NULL)
2667 ire_refrele(prev_ire);
2668 return;
2669 }
2670
2671 ire_refrele(prev_ire);
2672 ire_refrele(ire);
2673
2674 /*
2675 * TODO: more precise handling for cases 0, 2, 3, the latter two
2676 * require TOS routing
2677 */
2678 switch (icmph->icmph_code) {
2679 case 0:
2680 case 1:
2681 /* TODO: TOS specificity for cases 2 and 3 */
2682 case 2:
2683 case 3:
2684 break;
2685 default:
2686 BUMP_MIB(&ipst->ips_icmp_mib, icmpInBadRedirects);
2687 ip_drop_input("icmpInBadRedirects - code", mp, ira->ira_ill);
2688 freemsg(mp);
2689 return;
2690 }
2691 /*
2692 * Create a Route Association. This will allow us to remember that
2693 * someone we believe told us to use the particular gateway.
2694 */
2695 ire = ire_create(
2696 (uchar_t *)&dst, /* dest addr */
2697 (uchar_t *)&ip_g_all_ones, /* mask */
2698 (uchar_t *)&gateway, /* gateway addr */
2699 IRE_HOST,
2700 NULL, /* ill */
2701 ALL_ZONES,
2702 (RTF_DYNAMIC | RTF_GATEWAY | RTF_HOST),
2703 NULL, /* tsol_gc_t */
2704 ipst);
2705
2706 if (ire == NULL) {
2707 freemsg(mp);
2708 return;
2709 }
2710 nire = ire_add(ire);
2711 /* Check if it was a duplicate entry */
2712 if (nire != NULL && nire != ire) {
2713 ASSERT(nire->ire_identical_ref > 1);
2714 ire_delete(nire);
2715 ire_refrele(nire);
2716 nire = NULL;
2717 }
2718 ire = nire;
2719 if (ire != NULL) {
2720 ire_refrele(ire); /* Held in ire_add */
2721
2722 /* tell routing sockets that we received a redirect */
2723 ip_rts_change(RTM_REDIRECT, dst, gateway, IP_HOST_MASK, 0, src,
2724 (RTF_DYNAMIC | RTF_GATEWAY | RTF_HOST), 0,
2725 (RTA_DST | RTA_GATEWAY | RTA_NETMASK | RTA_AUTHOR), ipst);
2726 }
2727
2728 /*
2729 * Delete any existing IRE_HOST type redirect ires for this destination.
2730 * This together with the added IRE has the effect of
2731 * modifying an existing redirect.
2732 */
2733 prev_ire = ire_ftable_lookup_v4(dst, 0, src, IRE_HOST, NULL,
2734 ALL_ZONES, NULL, (MATCH_IRE_GW | MATCH_IRE_TYPE), 0, ipst, NULL);
2735 if (prev_ire != NULL) {
2736 if (prev_ire ->ire_flags & RTF_DYNAMIC)
2737 ire_delete(prev_ire);
2738 ire_refrele(prev_ire);
2739 }
2740
2741 freemsg(mp);
2742 }
2743
2744 /*
2745 * Generate an ICMP parameter problem message.
2746 * When called from ip_output side a minimal ip_recv_attr_t needs to be
2747 * constructed by the caller.
2748 */
2749 static void
2750 icmp_param_problem(mblk_t *mp, uint8_t ptr, ip_recv_attr_t *ira)
2751 {
2752 icmph_t icmph;
2753 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2754
2755 mp = icmp_pkt_err_ok(mp, ira);
2756 if (mp == NULL)
2757 return;
2758
2759 bzero(&icmph, sizeof (icmph_t));
2760 icmph.icmph_type = ICMP_PARAM_PROBLEM;
2761 icmph.icmph_pp_ptr = ptr;
2762 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutParmProbs);
2763 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
2764 }
2765
2766 /*
2767 * Build and ship an IPv4 ICMP message using the packet data in mp, and
2768 * the ICMP header pointed to by "stuff". (May be called as writer.)
2769 * Note: assumes that icmp_pkt_err_ok has been called to verify that
2770 * an icmp error packet can be sent.
2771 * Assigns an appropriate source address to the packet. If ipha_dst is
2772 * one of our addresses use it for source. Otherwise let ip_output_simple
2773 * pick the source address.
2774 */
2775 static void
2776 icmp_pkt(mblk_t *mp, void *stuff, size_t len, ip_recv_attr_t *ira)
2777 {
2778 ipaddr_t dst;
2779 icmph_t *icmph;
2780 ipha_t *ipha;
2781 uint_t len_needed;
2782 size_t msg_len;
2783 mblk_t *mp1;
2784 ipaddr_t src;
2785 ire_t *ire;
2786 ip_xmit_attr_t ixas;
2787 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2788
2789 ipha = (ipha_t *)mp->b_rptr;
2790
2791 bzero(&ixas, sizeof (ixas));
2792 ixas.ixa_flags = IXAF_BASIC_SIMPLE_V4;
2793 ixas.ixa_zoneid = ira->ira_zoneid;
2794 ixas.ixa_ifindex = 0;
2795 ixas.ixa_ipst = ipst;
2796 ixas.ixa_cred = kcred;
2797 ixas.ixa_cpid = NOPID;
2798 ixas.ixa_tsl = ira->ira_tsl; /* Behave as a multi-level responder */
2799 ixas.ixa_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
2800
2801 if (ira->ira_flags & IRAF_IPSEC_SECURE) {
2802 /*
2803 * Apply IPsec based on how IPsec was applied to
2804 * the packet that had the error.
2805 *
2806 * If it was an outbound packet that caused the ICMP
2807 * error, then the caller will have setup the IRA
2808 * appropriately.
2809 */
2810 if (!ipsec_in_to_out(ira, &ixas, mp, ipha, NULL)) {
2811 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
2812 /* Note: mp already consumed and ip_drop_packet done */
2813 return;
2814 }
2815 } else {
2816 /*
2817 * This is in clear. The icmp message we are building
2818 * here should go out in clear, independent of our policy.
2819 */
2820 ixas.ixa_flags |= IXAF_NO_IPSEC;
2821 }
2822
2823 /* Remember our eventual destination */
2824 dst = ipha->ipha_src;
2825
2826 /*
2827 * If the packet was for one of our unicast addresses, make
2828 * sure we respond with that as the source. Otherwise
2829 * have ip_output_simple pick the source address.
2830 */
2831 ire = ire_ftable_lookup_v4(ipha->ipha_dst, 0, 0,
2832 (IRE_LOCAL|IRE_LOOPBACK), NULL, ira->ira_zoneid, NULL,
2833 MATCH_IRE_TYPE|MATCH_IRE_ZONEONLY, 0, ipst, NULL);
2834 if (ire != NULL) {
2835 ire_refrele(ire);
2836 src = ipha->ipha_dst;
2837 } else {
2838 src = INADDR_ANY;
2839 ixas.ixa_flags |= IXAF_SET_SOURCE;
2840 }
2841
2842 /*
2843 * Check if we can send back more then 8 bytes in addition to
2844 * the IP header. We try to send 64 bytes of data and the internal
2845 * header in the special cases of ipv4 encapsulated ipv4 or ipv6.
2846 */
2847 len_needed = IPH_HDR_LENGTH(ipha);
2848 if (ipha->ipha_protocol == IPPROTO_ENCAP ||
2849 ipha->ipha_protocol == IPPROTO_IPV6) {
2850 if (!pullupmsg(mp, -1)) {
2851 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsOutDiscards);
2852 ip_drop_output("ipIfStatsOutDiscards", mp, NULL);
2853 freemsg(mp);
2854 return;
2855 }
2856 ipha = (ipha_t *)mp->b_rptr;
2857
2858 if (ipha->ipha_protocol == IPPROTO_ENCAP) {
2859 len_needed += IPH_HDR_LENGTH(((uchar_t *)ipha +
2860 len_needed));
2861 } else {
2862 ip6_t *ip6h = (ip6_t *)((uchar_t *)ipha + len_needed);
2863
2864 ASSERT(ipha->ipha_protocol == IPPROTO_IPV6);
2865 len_needed += ip_hdr_length_v6(mp, ip6h);
2866 }
2867 }
2868 len_needed += ipst->ips_ip_icmp_return;
2869 msg_len = msgdsize(mp);
2870 if (msg_len > len_needed) {
2871 (void) adjmsg(mp, len_needed - msg_len);
2872 msg_len = len_needed;
2873 }
2874 mp1 = allocb(sizeof (icmp_ipha) + len, BPRI_MED);
2875 if (mp1 == NULL) {
2876 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutErrors);
2877 freemsg(mp);
2878 return;
2879 }
2880 mp1->b_cont = mp;
2881 mp = mp1;
2882
2883 /*
2884 * Set IXAF_TRUSTED_ICMP so we can let the ICMP messages this
2885 * node generates be accepted in peace by all on-host destinations.
2886 * If we do NOT assume that all on-host destinations trust
2887 * self-generated ICMP messages, then rework here, ip6.c, and spd.c.
2888 * (Look for IXAF_TRUSTED_ICMP).
2889 */
2890 ixas.ixa_flags |= IXAF_TRUSTED_ICMP;
2891
2892 ipha = (ipha_t *)mp->b_rptr;
2893 mp1->b_wptr = (uchar_t *)ipha + (sizeof (icmp_ipha) + len);
2894 *ipha = icmp_ipha;
2895 ipha->ipha_src = src;
2896 ipha->ipha_dst = dst;
2897 ipha->ipha_ttl = ipst->ips_ip_def_ttl;
2898 msg_len += sizeof (icmp_ipha) + len;
2899 if (msg_len > IP_MAXPACKET) {
2900 (void) adjmsg(mp, IP_MAXPACKET - msg_len);
2901 msg_len = IP_MAXPACKET;
2902 }
2903 ipha->ipha_length = htons((uint16_t)msg_len);
2904 icmph = (icmph_t *)&ipha[1];
2905 bcopy(stuff, icmph, len);
2906 icmph->icmph_checksum = 0;
2907 icmph->icmph_checksum = IP_CSUM(mp, (int32_t)sizeof (ipha_t), 0);
2908 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutMsgs);
2909
2910 (void) ip_output_simple(mp, &ixas);
2911 ixa_cleanup(&ixas);
2912 }
2913
2914 /*
2915 * Determine if an ICMP error packet can be sent given the rate limit.
2916 * The limit consists of an average frequency (icmp_pkt_err_interval measured
2917 * in milliseconds) and a burst size. Burst size number of packets can
2918 * be sent arbitrarely closely spaced.
2919 * The state is tracked using two variables to implement an approximate
2920 * token bucket filter:
2921 * icmp_pkt_err_last - lbolt value when the last burst started
2922 * icmp_pkt_err_sent - number of packets sent in current burst
2923 */
2924 boolean_t
2925 icmp_err_rate_limit(ip_stack_t *ipst)
2926 {
2927 clock_t now = TICK_TO_MSEC(ddi_get_lbolt());
2928 uint_t refilled; /* Number of packets refilled in tbf since last */
2929 /* Guard against changes by loading into local variable */
2930 uint_t err_interval = ipst->ips_ip_icmp_err_interval;
2931
2932 if (err_interval == 0)
2933 return (B_FALSE);
2934
2935 if (ipst->ips_icmp_pkt_err_last > now) {
2936 /* 100HZ lbolt in ms for 32bit arch wraps every 49.7 days */
2937 ipst->ips_icmp_pkt_err_last = 0;
2938 ipst->ips_icmp_pkt_err_sent = 0;
2939 }
2940 /*
2941 * If we are in a burst update the token bucket filter.
2942 * Update the "last" time to be close to "now" but make sure
2943 * we don't loose precision.
2944 */
2945 if (ipst->ips_icmp_pkt_err_sent != 0) {
2946 refilled = (now - ipst->ips_icmp_pkt_err_last)/err_interval;
2947 if (refilled > ipst->ips_icmp_pkt_err_sent) {
2948 ipst->ips_icmp_pkt_err_sent = 0;
2949 } else {
2950 ipst->ips_icmp_pkt_err_sent -= refilled;
2951 ipst->ips_icmp_pkt_err_last += refilled * err_interval;
2952 }
2953 }
2954 if (ipst->ips_icmp_pkt_err_sent == 0) {
2955 /* Start of new burst */
2956 ipst->ips_icmp_pkt_err_last = now;
2957 }
2958 if (ipst->ips_icmp_pkt_err_sent < ipst->ips_ip_icmp_err_burst) {
2959 ipst->ips_icmp_pkt_err_sent++;
2960 ip1dbg(("icmp_err_rate_limit: %d sent in burst\n",
2961 ipst->ips_icmp_pkt_err_sent));
2962 return (B_FALSE);
2963 }
2964 ip1dbg(("icmp_err_rate_limit: dropped\n"));
2965 return (B_TRUE);
2966 }
2967
2968 /*
2969 * Check if it is ok to send an IPv4 ICMP error packet in
2970 * response to the IPv4 packet in mp.
2971 * Free the message and return null if no
2972 * ICMP error packet should be sent.
2973 */
2974 static mblk_t *
2975 icmp_pkt_err_ok(mblk_t *mp, ip_recv_attr_t *ira)
2976 {
2977 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
2978 icmph_t *icmph;
2979 ipha_t *ipha;
2980 uint_t len_needed;
2981
2982 if (!mp)
2983 return (NULL);
2984 ipha = (ipha_t *)mp->b_rptr;
2985 if (ip_csum_hdr(ipha)) {
2986 BUMP_MIB(&ipst->ips_ip_mib, ipIfStatsInCksumErrs);
2987 ip_drop_input("ipIfStatsInCksumErrs", mp, NULL);
2988 freemsg(mp);
2989 return (NULL);
2990 }
2991 if (ip_type_v4(ipha->ipha_dst, ipst) == IRE_BROADCAST ||
2992 ip_type_v4(ipha->ipha_src, ipst) == IRE_BROADCAST ||
2993 CLASSD(ipha->ipha_dst) ||
2994 CLASSD(ipha->ipha_src) ||
2995 (ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_OFFSET)) {
2996 /* Note: only errors to the fragment with offset 0 */
2997 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
2998 freemsg(mp);
2999 return (NULL);
3000 }
3001 if (ipha->ipha_protocol == IPPROTO_ICMP) {
3002 /*
3003 * Check the ICMP type. RFC 1122 sez: don't send ICMP
3004 * errors in response to any ICMP errors.
3005 */
3006 len_needed = IPH_HDR_LENGTH(ipha) + ICMPH_SIZE;
3007 if (mp->b_wptr - mp->b_rptr < len_needed) {
3008 if (!pullupmsg(mp, len_needed)) {
3009 BUMP_MIB(&ipst->ips_icmp_mib, icmpInErrors);
3010 freemsg(mp);
3011 return (NULL);
3012 }
3013 ipha = (ipha_t *)mp->b_rptr;
3014 }
3015 icmph = (icmph_t *)
3016 (&((char *)ipha)[IPH_HDR_LENGTH(ipha)]);
3017 switch (icmph->icmph_type) {
3018 case ICMP_DEST_UNREACHABLE:
3019 case ICMP_SOURCE_QUENCH:
3020 case ICMP_TIME_EXCEEDED:
3021 case ICMP_PARAM_PROBLEM:
3022 case ICMP_REDIRECT:
3023 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
3024 freemsg(mp);
3025 return (NULL);
3026 default:
3027 break;
3028 }
3029 }
3030 /*
3031 * If this is a labeled system, then check to see if we're allowed to
3032 * send a response to this particular sender. If not, then just drop.
3033 */
3034 if (is_system_labeled() && !tsol_can_reply_error(mp, ira)) {
3035 ip2dbg(("icmp_pkt_err_ok: can't respond to packet\n"));
3036 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDrops);
3037 freemsg(mp);
3038 return (NULL);
3039 }
3040 if (icmp_err_rate_limit(ipst)) {
3041 /*
3042 * Only send ICMP error packets every so often.
3043 * This should be done on a per port/source basis,
3044 * but for now this will suffice.
3045 */
3046 freemsg(mp);
3047 return (NULL);
3048 }
3049 return (mp);
3050 }
3051
3052 /*
3053 * Called when a packet was sent out the same link that it arrived on.
3054 * Check if it is ok to send a redirect and then send it.
3055 */
3056 void
3057 ip_send_potential_redirect_v4(mblk_t *mp, ipha_t *ipha, ire_t *ire,
3058 ip_recv_attr_t *ira)
3059 {
3060 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3061 ipaddr_t src, nhop;
3062 mblk_t *mp1;
3063 ire_t *nhop_ire;
3064
3065 /*
3066 * Check the source address to see if it originated
3067 * on the same logical subnet it is going back out on.
3068 * If so, we should be able to send it a redirect.
3069 * Avoid sending a redirect if the destination
3070 * is directly connected (i.e., we matched an IRE_ONLINK),
3071 * or if the packet was source routed out this interface.
3072 *
3073 * We avoid sending a redirect if the
3074 * destination is directly connected
3075 * because it is possible that multiple
3076 * IP subnets may have been configured on
3077 * the link, and the source may not
3078 * be on the same subnet as ip destination,
3079 * even though they are on the same
3080 * physical link.
3081 */
3082 if ((ire->ire_type & IRE_ONLINK) ||
3083 ip_source_routed(ipha, ipst))
3084 return;
3085
3086 nhop_ire = ire_nexthop(ire);
3087 if (nhop_ire == NULL)
3088 return;
3089
3090 nhop = nhop_ire->ire_addr;
3091
3092 if (nhop_ire->ire_type & IRE_IF_CLONE) {
3093 ire_t *ire2;
3094
3095 /* Follow ire_dep_parent to find non-clone IRE_INTERFACE */
3096 mutex_enter(&nhop_ire->ire_lock);
3097 ire2 = nhop_ire->ire_dep_parent;
3098 if (ire2 != NULL)
3099 ire_refhold(ire2);
3100 mutex_exit(&nhop_ire->ire_lock);
3101 ire_refrele(nhop_ire);
3102 nhop_ire = ire2;
3103 }
3104 if (nhop_ire == NULL)
3105 return;
3106
3107 ASSERT(!(nhop_ire->ire_type & IRE_IF_CLONE));
3108
3109 src = ipha->ipha_src;
3110
3111 /*
3112 * We look at the interface ire for the nexthop,
3113 * to see if ipha_src is in the same subnet
3114 * as the nexthop.
3115 */
3116 if ((src & nhop_ire->ire_mask) == (nhop & nhop_ire->ire_mask)) {
3117 /*
3118 * The source is directly connected.
3119 */
3120 mp1 = copymsg(mp);
3121 if (mp1 != NULL) {
3122 icmp_send_redirect(mp1, nhop, ira);
3123 }
3124 }
3125 ire_refrele(nhop_ire);
3126 }
3127
3128 /*
3129 * Generate an ICMP redirect message.
3130 */
3131 static void
3132 icmp_send_redirect(mblk_t *mp, ipaddr_t gateway, ip_recv_attr_t *ira)
3133 {
3134 icmph_t icmph;
3135 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3136
3137 mp = icmp_pkt_err_ok(mp, ira);
3138 if (mp == NULL)
3139 return;
3140
3141 bzero(&icmph, sizeof (icmph_t));
3142 icmph.icmph_type = ICMP_REDIRECT;
3143 icmph.icmph_code = 1;
3144 icmph.icmph_rd_gateway = gateway;
3145 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutRedirects);
3146 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3147 }
3148
3149 /*
3150 * Generate an ICMP time exceeded message.
3151 */
3152 void
3153 icmp_time_exceeded(mblk_t *mp, uint8_t code, ip_recv_attr_t *ira)
3154 {
3155 icmph_t icmph;
3156 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3157
3158 mp = icmp_pkt_err_ok(mp, ira);
3159 if (mp == NULL)
3160 return;
3161
3162 bzero(&icmph, sizeof (icmph_t));
3163 icmph.icmph_type = ICMP_TIME_EXCEEDED;
3164 icmph.icmph_code = code;
3165 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutTimeExcds);
3166 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3167 }
3168
3169 /*
3170 * Generate an ICMP unreachable message.
3171 * When called from ip_output side a minimal ip_recv_attr_t needs to be
3172 * constructed by the caller.
3173 */
3174 void
3175 icmp_unreachable(mblk_t *mp, uint8_t code, ip_recv_attr_t *ira)
3176 {
3177 icmph_t icmph;
3178 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
3179
3180 mp = icmp_pkt_err_ok(mp, ira);
3181 if (mp == NULL)
3182 return;
3183
3184 bzero(&icmph, sizeof (icmph_t));
3185 icmph.icmph_type = ICMP_DEST_UNREACHABLE;
3186 icmph.icmph_code = code;
3187 BUMP_MIB(&ipst->ips_icmp_mib, icmpOutDestUnreachs);
3188 icmp_pkt(mp, &icmph, sizeof (icmph_t), ira);
3189 }
3190
3191 /*
3192 * Latch in the IPsec state for a stream based the policy in the listener
3193 * and the actions in the ip_recv_attr_t.
3194 * Called directly from TCP and SCTP.
3195 */
3196 boolean_t
3197 ip_ipsec_policy_inherit(conn_t *connp, conn_t *lconnp, ip_recv_attr_t *ira)
3198 {
3199 ASSERT(lconnp->conn_policy != NULL);
3200 ASSERT(connp->conn_policy == NULL);
3201
3202 IPPH_REFHOLD(lconnp->conn_policy);
3203 connp->conn_policy = lconnp->conn_policy;
3204
3205 if (ira->ira_ipsec_action != NULL) {
3206 if (connp->conn_latch == NULL) {
3207 connp->conn_latch = iplatch_create();
3208 if (connp->conn_latch == NULL)
3209 return (B_FALSE);
3210 }
3211 ipsec_latch_inbound(connp, ira);
3212 }
3213 return (B_TRUE);
3214 }
3215
3216 /*
3217 * Verify whether or not the IP address is a valid local address.
3218 * Could be a unicast, including one for a down interface.
3219 * If allow_mcbc then a multicast or broadcast address is also
3220 * acceptable.
3221 *
3222 * In the case of a broadcast/multicast address, however, the
3223 * upper protocol is expected to reset the src address
3224 * to zero when we return IPVL_MCAST/IPVL_BCAST so that
3225 * no packets are emitted with broadcast/multicast address as
3226 * source address (that violates hosts requirements RFC 1122)
3227 * The addresses valid for bind are:
3228 * (1) - INADDR_ANY (0)
3229 * (2) - IP address of an UP interface
3230 * (3) - IP address of a DOWN interface
3231 * (4) - valid local IP broadcast addresses. In this case
3232 * the conn will only receive packets destined to
3233 * the specified broadcast address.
3234 * (5) - a multicast address. In this case
3235 * the conn will only receive packets destined to
3236 * the specified multicast address. Note: the
3237 * application still has to issue an
3238 * IP_ADD_MEMBERSHIP socket option.
3239 *
3240 * In all the above cases, the bound address must be valid in the current zone.
3241 * When the address is loopback, multicast or broadcast, there might be many
3242 * matching IREs so bind has to look up based on the zone.
3243 */
3244 ip_laddr_t
3245 ip_laddr_verify_v4(ipaddr_t src_addr, zoneid_t zoneid,
3246 ip_stack_t *ipst, boolean_t allow_mcbc)
3247 {
3248 ire_t *src_ire;
3249
3250 ASSERT(src_addr != INADDR_ANY);
3251
3252 src_ire = ire_ftable_lookup_v4(src_addr, 0, 0, 0,
3253 NULL, zoneid, NULL, MATCH_IRE_ZONEONLY, 0, ipst, NULL);
3254
3255 /*
3256 * If an address other than in6addr_any is requested,
3257 * we verify that it is a valid address for bind
3258 * Note: Following code is in if-else-if form for
3259 * readability compared to a condition check.
3260 */
3261 if (src_ire != NULL && (src_ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK))) {
3262 /*
3263 * (2) Bind to address of local UP interface
3264 */
3265 ire_refrele(src_ire);
3266 return (IPVL_UNICAST_UP);
3267 } else if (src_ire != NULL && src_ire->ire_type & IRE_BROADCAST) {
3268 /*
3269 * (4) Bind to broadcast address
3270 */
3271 ire_refrele(src_ire);
3272 if (allow_mcbc)
3273 return (IPVL_BCAST);
3274 else
3275 return (IPVL_BAD);
3276 } else if (CLASSD(src_addr)) {
3277 /* (5) bind to multicast address. */
3278 if (src_ire != NULL)
3279 ire_refrele(src_ire);
3280
3281 if (allow_mcbc)
3282 return (IPVL_MCAST);
3283 else
3284 return (IPVL_BAD);
3285 } else {
3286 ipif_t *ipif;
3287
3288 /*
3289 * (3) Bind to address of local DOWN interface?
3290 * (ipif_lookup_addr() looks up all interfaces
3291 * but we do not get here for UP interfaces
3292 * - case (2) above)
3293 */
3294 if (src_ire != NULL)
3295 ire_refrele(src_ire);
3296
3297 ipif = ipif_lookup_addr(src_addr, NULL, zoneid, ipst);
3298 if (ipif == NULL)
3299 return (IPVL_BAD);
3300
3301 /* Not a useful source? */
3302 if (ipif->ipif_flags & (IPIF_NOLOCAL | IPIF_ANYCAST)) {
3303 ipif_refrele(ipif);
3304 return (IPVL_BAD);
3305 }
3306 ipif_refrele(ipif);
3307 return (IPVL_UNICAST_DOWN);
3308 }
3309 }
3310
3311 /*
3312 * Insert in the bind fanout for IPv4 and IPv6.
3313 * The caller should already have used ip_laddr_verify_v*() before calling
3314 * this.
3315 */
3316 int
3317 ip_laddr_fanout_insert(conn_t *connp)
3318 {
3319 int error;
3320
3321 /*
3322 * Allow setting new policies. For example, disconnects result
3323 * in us being called. As we would have set conn_policy_cached
3324 * to B_TRUE before, we should set it to B_FALSE, so that policy
3325 * can change after the disconnect.
3326 */
3327 connp->conn_policy_cached = B_FALSE;
3328
3329 error = ipcl_bind_insert(connp);
3330 if (error != 0) {
3331 if (connp->conn_anon_port) {
3332 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
3333 connp->conn_mlp_type, connp->conn_proto,
3334 ntohs(connp->conn_lport), B_FALSE);
3335 }
3336 connp->conn_mlp_type = mlptSingle;
3337 }
3338 return (error);
3339 }
3340
3341 /*
3342 * Verify that both the source and destination addresses are valid. If
3343 * IPDF_VERIFY_DST is not set, then the destination address may be unreachable,
3344 * i.e. have no route to it. Protocols like TCP want to verify destination
3345 * reachability, while tunnels do not.
3346 *
3347 * Determine the route, the interface, and (optionally) the source address
3348 * to use to reach a given destination.
3349 * Note that we allow connect to broadcast and multicast addresses when
3350 * IPDF_ALLOW_MCBC is set.
3351 * first_hop and dst_addr are normally the same, but if source routing
3352 * they will differ; in that case the first_hop is what we'll use for the
3353 * routing lookup but the dce and label checks will be done on dst_addr,
3354 *
3355 * If uinfo is set, then we fill in the best available information
3356 * we have for the destination. This is based on (in priority order) any
3357 * metrics and path MTU stored in a dce_t, route metrics, and finally the
3358 * ill_mtu/ill_mc_mtu.
3359 *
3360 * Tsol note: If we have a source route then dst_addr != firsthop. But we
3361 * always do the label check on dst_addr.
3362 */
3363 int
3364 ip_set_destination_v4(ipaddr_t *src_addrp, ipaddr_t dst_addr, ipaddr_t firsthop,
3365 ip_xmit_attr_t *ixa, iulp_t *uinfo, uint32_t flags, uint_t mac_mode)
3366 {
3367 ire_t *ire = NULL;
3368 int error = 0;
3369 ipaddr_t setsrc; /* RTF_SETSRC */
3370 zoneid_t zoneid = ixa->ixa_zoneid; /* Honors SO_ALLZONES */
3371 ip_stack_t *ipst = ixa->ixa_ipst;
3372 dce_t *dce;
3373 uint_t pmtu;
3374 uint_t generation;
3375 nce_t *nce;
3376 ill_t *ill = NULL;
3377 boolean_t multirt = B_FALSE;
3378
3379 ASSERT(ixa->ixa_flags & IXAF_IS_IPV4);
3380
3381 /*
3382 * We never send to zero; the ULPs map it to the loopback address.
3383 * We can't allow it since we use zero to mean unitialized in some
3384 * places.
3385 */
3386 ASSERT(dst_addr != INADDR_ANY);
3387
3388 if (is_system_labeled()) {
3389 ts_label_t *tsl = NULL;
3390
3391 error = tsol_check_dest(ixa->ixa_tsl, &dst_addr, IPV4_VERSION,
3392 mac_mode, (flags & IPDF_ZONE_IS_GLOBAL) != 0, &tsl);
3393 if (error != 0)
3394 return (error);
3395 if (tsl != NULL) {
3396 /* Update the label */
3397 ip_xmit_attr_replace_tsl(ixa, tsl);
3398 }
3399 }
3400
3401 setsrc = INADDR_ANY;
3402 /*
3403 * Select a route; For IPMP interfaces, we would only select
3404 * a "hidden" route (i.e., going through a specific under_ill)
3405 * if ixa_ifindex has been specified.
3406 */
3407 ire = ip_select_route_v4(firsthop, *src_addrp, ixa,
3408 &generation, &setsrc, &error, &multirt);
3409 ASSERT(ire != NULL); /* IRE_NOROUTE if none found */
3410 if (error != 0)
3411 goto bad_addr;
3412
3413 /*
3414 * ire can't be a broadcast or multicast unless IPDF_ALLOW_MCBC is set.
3415 * If IPDF_VERIFY_DST is set, the destination must be reachable;
3416 * Otherwise the destination needn't be reachable.
3417 *
3418 * If we match on a reject or black hole, then we've got a
3419 * local failure. May as well fail out the connect() attempt,
3420 * since it's never going to succeed.
3421 */
3422 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
3423 /*
3424 * If we're verifying destination reachability, we always want
3425 * to complain here.
3426 *
3427 * If we're not verifying destination reachability but the
3428 * destination has a route, we still want to fail on the
3429 * temporary address and broadcast address tests.
3430 *
3431 * In both cases do we let the code continue so some reasonable
3432 * information is returned to the caller. That enables the
3433 * caller to use (and even cache) the IRE. conn_ip_ouput will
3434 * use the generation mismatch path to check for the unreachable
3435 * case thereby avoiding any specific check in the main path.
3436 */
3437 ASSERT(generation == IRE_GENERATION_VERIFY);
3438 if (flags & IPDF_VERIFY_DST) {
3439 /*
3440 * Set errno but continue to set up ixa_ire to be
3441 * the RTF_REJECT|RTF_BLACKHOLE IRE.
3442 * That allows callers to use ip_output to get an
3443 * ICMP error back.
3444 */
3445 if (!(ire->ire_type & IRE_HOST))
3446 error = ENETUNREACH;
3447 else
3448 error = EHOSTUNREACH;
3449 }
3450 }
3451
3452 if ((ire->ire_type & (IRE_BROADCAST|IRE_MULTICAST)) &&
3453 !(flags & IPDF_ALLOW_MCBC)) {
3454 ire_refrele(ire);
3455 ire = ire_reject(ipst, B_FALSE);
3456 generation = IRE_GENERATION_VERIFY;
3457 error = ENETUNREACH;
3458 }
3459
3460 /* Cache things */
3461 if (ixa->ixa_ire != NULL)
3462 ire_refrele_notr(ixa->ixa_ire);
3463 #ifdef DEBUG
3464 ire_refhold_notr(ire);
3465 ire_refrele(ire);
3466 #endif
3467 ixa->ixa_ire = ire;
3468 ixa->ixa_ire_generation = generation;
3469
3470 /*
3471 * Ensure that ixa_dce is always set any time that ixa_ire is set,
3472 * since some callers will send a packet to conn_ip_output() even if
3473 * there's an error.
3474 */
3475 if (flags & IPDF_UNIQUE_DCE) {
3476 /* Fallback to the default dce if allocation fails */
3477 dce = dce_lookup_and_add_v4(dst_addr, ipst);
3478 if (dce != NULL)
3479 generation = dce->dce_generation;
3480 else
3481 dce = dce_lookup_v4(dst_addr, ipst, &generation);
3482 } else {
3483 dce = dce_lookup_v4(dst_addr, ipst, &generation);
3484 }
3485 ASSERT(dce != NULL);
3486 if (ixa->ixa_dce != NULL)
3487 dce_refrele_notr(ixa->ixa_dce);
3488 #ifdef DEBUG
3489 dce_refhold_notr(dce);
3490 dce_refrele(dce);
3491 #endif
3492 ixa->ixa_dce = dce;
3493 ixa->ixa_dce_generation = generation;
3494
3495 /*
3496 * For multicast with multirt we have a flag passed back from
3497 * ire_lookup_multi_ill_v4 since we don't have an IRE for each
3498 * possible multicast address.
3499 * We also need a flag for multicast since we can't check
3500 * whether RTF_MULTIRT is set in ixa_ire for multicast.
3501 */
3502 if (multirt) {
3503 ixa->ixa_postfragfn = ip_postfrag_multirt_v4;
3504 ixa->ixa_flags |= IXAF_MULTIRT_MULTICAST;
3505 } else {
3506 ixa->ixa_postfragfn = ire->ire_postfragfn;
3507 ixa->ixa_flags &= ~IXAF_MULTIRT_MULTICAST;
3508 }
3509 if (!(ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))) {
3510 /* Get an nce to cache. */
3511 nce = ire_to_nce(ire, firsthop, NULL);
3512 if (nce == NULL) {
3513 /* Allocation failure? */
3514 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3515 } else {
3516 if (ixa->ixa_nce != NULL)
3517 nce_refrele(ixa->ixa_nce);
3518 ixa->ixa_nce = nce;
3519 }
3520 }
3521
3522 /*
3523 * If the source address is a loopback address, the
3524 * destination had best be local or multicast.
3525 * If we are sending to an IRE_LOCAL using a loopback source then
3526 * it had better be the same zoneid.
3527 */
3528 if (*src_addrp == htonl(INADDR_LOOPBACK)) {
3529 if ((ire->ire_type & IRE_LOCAL) && ire->ire_zoneid != zoneid) {
3530 ire = NULL; /* Stored in ixa_ire */
3531 error = EADDRNOTAVAIL;
3532 goto bad_addr;
3533 }
3534 if (!(ire->ire_type & (IRE_LOOPBACK|IRE_LOCAL|IRE_MULTICAST))) {
3535 ire = NULL; /* Stored in ixa_ire */
3536 error = EADDRNOTAVAIL;
3537 goto bad_addr;
3538 }
3539 }
3540 if (ire->ire_type & IRE_BROADCAST) {
3541 /*
3542 * If the ULP didn't have a specified source, then we
3543 * make sure we reselect the source when sending
3544 * broadcasts out different interfaces.
3545 */
3546 if (flags & IPDF_SELECT_SRC)
3547 ixa->ixa_flags |= IXAF_SET_SOURCE;
3548 else
3549 ixa->ixa_flags &= ~IXAF_SET_SOURCE;
3550 }
3551
3552 /*
3553 * Does the caller want us to pick a source address?
3554 */
3555 if (flags & IPDF_SELECT_SRC) {
3556 ipaddr_t src_addr;
3557
3558 /*
3559 * We use use ire_nexthop_ill to avoid the under ipmp
3560 * interface for source address selection. Note that for ipmp
3561 * probe packets, ixa_ifindex would have been specified, and
3562 * the ip_select_route() invocation would have picked an ire
3563 * will ire_ill pointing at an under interface.
3564 */
3565 ill = ire_nexthop_ill(ire);
3566
3567 /* If unreachable we have no ill but need some source */
3568 if (ill == NULL) {
3569 src_addr = htonl(INADDR_LOOPBACK);
3570 /* Make sure we look for a better source address */
3571 generation = SRC_GENERATION_VERIFY;
3572 } else {
3573 error = ip_select_source_v4(ill, setsrc, dst_addr,
3574 ixa->ixa_multicast_ifaddr, zoneid,
3575 ipst, &src_addr, &generation, NULL);
3576 if (error != 0) {
3577 ire = NULL; /* Stored in ixa_ire */
3578 goto bad_addr;
3579 }
3580 }
3581
3582 /*
3583 * We allow the source address to to down.
3584 * However, we check that we don't use the loopback address
3585 * as a source when sending out on the wire.
3586 */
3587 if ((src_addr == htonl(INADDR_LOOPBACK)) &&
3588 !(ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK|IRE_MULTICAST)) &&
3589 !(ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))) {
3590 ire = NULL; /* Stored in ixa_ire */
3591 error = EADDRNOTAVAIL;
3592 goto bad_addr;
3593 }
3594
3595 *src_addrp = src_addr;
3596 ixa->ixa_src_generation = generation;
3597 }
3598
3599 /*
3600 * Make sure we don't leave an unreachable ixa_nce in place
3601 * since ip_select_route is used when we unplumb i.e., remove
3602 * references on ixa_ire, ixa_nce, and ixa_dce.
3603 */
3604 nce = ixa->ixa_nce;
3605 if (nce != NULL && nce->nce_is_condemned) {
3606 nce_refrele(nce);
3607 ixa->ixa_nce = NULL;
3608 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3609 }
3610
3611 /*
3612 * The caller has set IXAF_PMTU_DISCOVERY if path MTU is desired.
3613 * However, we can't do it for IPv4 multicast or broadcast.
3614 */
3615 if (ire->ire_type & (IRE_BROADCAST|IRE_MULTICAST))
3616 ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
3617
3618 /*
3619 * Set initial value for fragmentation limit. Either conn_ip_output
3620 * or ULP might updates it when there are routing changes.
3621 * Handles a NULL ixa_ire->ire_ill or a NULL ixa_nce for RTF_REJECT.
3622 */
3623 pmtu = ip_get_pmtu(ixa);
3624 ixa->ixa_fragsize = pmtu;
3625 /* Make sure ixa_fragsize and ixa_pmtu remain identical */
3626 if (ixa->ixa_flags & IXAF_VERIFY_PMTU)
3627 ixa->ixa_pmtu = pmtu;
3628
3629 /*
3630 * Extract information useful for some transports.
3631 * First we look for DCE metrics. Then we take what we have in
3632 * the metrics in the route, where the offlink is used if we have
3633 * one.
3634 */
3635 if (uinfo != NULL) {
3636 bzero(uinfo, sizeof (*uinfo));
3637
3638 if (dce->dce_flags & DCEF_UINFO)
3639 *uinfo = dce->dce_uinfo;
3640
3641 rts_merge_metrics(uinfo, &ire->ire_metrics);
3642
3643 /* Allow ire_metrics to decrease the path MTU from above */
3644 if (uinfo->iulp_mtu == 0 || uinfo->iulp_mtu > pmtu)
3645 uinfo->iulp_mtu = pmtu;
3646
3647 uinfo->iulp_localnet = (ire->ire_type & IRE_ONLINK) != 0;
3648 uinfo->iulp_loopback = (ire->ire_type & IRE_LOOPBACK) != 0;
3649 uinfo->iulp_local = (ire->ire_type & IRE_LOCAL) != 0;
3650 }
3651
3652 if (ill != NULL)
3653 ill_refrele(ill);
3654
3655 return (error);
3656
3657 bad_addr:
3658 if (ire != NULL)
3659 ire_refrele(ire);
3660
3661 if (ill != NULL)
3662 ill_refrele(ill);
3663
3664 /*
3665 * Make sure we don't leave an unreachable ixa_nce in place
3666 * since ip_select_route is used when we unplumb i.e., remove
3667 * references on ixa_ire, ixa_nce, and ixa_dce.
3668 */
3669 nce = ixa->ixa_nce;
3670 if (nce != NULL && nce->nce_is_condemned) {
3671 nce_refrele(nce);
3672 ixa->ixa_nce = NULL;
3673 ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
3674 }
3675
3676 return (error);
3677 }
3678
3679
3680 /*
3681 * Get the base MTU for the case when path MTU discovery is not used.
3682 * Takes the MTU of the IRE into account.
3683 */
3684 uint_t
3685 ip_get_base_mtu(ill_t *ill, ire_t *ire)
3686 {
3687 uint_t mtu;
3688 uint_t iremtu = ire->ire_metrics.iulp_mtu;
3689
3690 if (ire->ire_type & (IRE_MULTICAST|IRE_BROADCAST))
3691 mtu = ill->ill_mc_mtu;
3692 else
3693 mtu = ill->ill_mtu;
3694
3695 if (iremtu != 0 && iremtu < mtu)
3696 mtu = iremtu;
3697
3698 return (mtu);
3699 }
3700
3701 /*
3702 * Get the PMTU for the attributes. Handles both IPv4 and IPv6.
3703 * Assumes that ixa_ire, dce, and nce have already been set up.
3704 *
3705 * The caller has set IXAF_PMTU_DISCOVERY if path MTU discovery is desired.
3706 * We avoid path MTU discovery if it is disabled with ndd.
3707 * Furtermore, if the path MTU is too small, then we don't set DF for IPv4.
3708 *
3709 * NOTE: We also used to turn it off for source routed packets. That
3710 * is no longer required since the dce is per final destination.
3711 */
3712 uint_t
3713 ip_get_pmtu(ip_xmit_attr_t *ixa)
3714 {
3715 ip_stack_t *ipst = ixa->ixa_ipst;
3716 dce_t *dce;
3717 nce_t *nce;
3718 ire_t *ire;
3719 uint_t pmtu;
3720
3721 ire = ixa->ixa_ire;
3722 dce = ixa->ixa_dce;
3723 nce = ixa->ixa_nce;
3724
3725 /*
3726 * If path MTU discovery has been turned off by ndd, then we ignore
3727 * any dce_pmtu and for IPv4 we will not set DF.
3728 */
3729 if (!ipst->ips_ip_path_mtu_discovery)
3730 ixa->ixa_flags &= ~IXAF_PMTU_DISCOVERY;
3731
3732 pmtu = IP_MAXPACKET;
3733 /*
3734 * Decide whether whether IPv4 sets DF
3735 * For IPv6 "no DF" means to use the 1280 mtu
3736 */
3737 if (ixa->ixa_flags & IXAF_PMTU_DISCOVERY) {
3738 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3739 } else {
3740 ixa->ixa_flags &= ~IXAF_PMTU_IPV4_DF;
3741 if (!(ixa->ixa_flags & IXAF_IS_IPV4))
3742 pmtu = IPV6_MIN_MTU;
3743 }
3744
3745 /* Check if the PMTU is to old before we use it */
3746 if ((dce->dce_flags & DCEF_PMTU) &&
3747 TICK_TO_SEC(ddi_get_lbolt64()) - dce->dce_last_change_time >
3748 ipst->ips_ip_pathmtu_interval) {
3749 /*
3750 * Older than 20 minutes. Drop the path MTU information.
3751 */
3752 mutex_enter(&dce->dce_lock);
3753 dce->dce_flags &= ~(DCEF_PMTU|DCEF_TOO_SMALL_PMTU);
3754 dce->dce_last_change_time = TICK_TO_SEC(ddi_get_lbolt64());
3755 mutex_exit(&dce->dce_lock);
3756 dce_increment_generation(dce);
3757 }
3758
3759 /* The metrics on the route can lower the path MTU */
3760 if (ire->ire_metrics.iulp_mtu != 0 &&
3761 ire->ire_metrics.iulp_mtu < pmtu)
3762 pmtu = ire->ire_metrics.iulp_mtu;
3763
3764 /*
3765 * If the path MTU is smaller than some minimum, we still use dce_pmtu
3766 * above (would be 576 for IPv4 and 1280 for IPv6), but we clear
3767 * IXAF_PMTU_IPV4_DF so that we avoid setting DF for IPv4.
3768 */
3769 if (ixa->ixa_flags & IXAF_PMTU_DISCOVERY) {
3770 if (dce->dce_flags & DCEF_PMTU) {
3771 if (dce->dce_pmtu < pmtu)
3772 pmtu = dce->dce_pmtu;
3773
3774 if (dce->dce_flags & DCEF_TOO_SMALL_PMTU) {
3775 ixa->ixa_flags |= IXAF_PMTU_TOO_SMALL;
3776 ixa->ixa_flags &= ~IXAF_PMTU_IPV4_DF;
3777 } else {
3778 ixa->ixa_flags &= ~IXAF_PMTU_TOO_SMALL;
3779 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3780 }
3781 } else {
3782 ixa->ixa_flags &= ~IXAF_PMTU_TOO_SMALL;
3783 ixa->ixa_flags |= IXAF_PMTU_IPV4_DF;
3784 }
3785 }
3786
3787 /*
3788 * If we have an IRE_LOCAL we use the loopback mtu instead of
3789 * the ill for going out the wire i.e., IRE_LOCAL gets the same
3790 * mtu as IRE_LOOPBACK.
3791 */
3792 if (ire->ire_type & (IRE_LOCAL|IRE_LOOPBACK)) {
3793 uint_t loopback_mtu;
3794
3795 loopback_mtu = (ire->ire_ipversion == IPV6_VERSION) ?
3796 ip_loopback_mtu_v6plus : ip_loopback_mtuplus;
3797
3798 if (loopback_mtu < pmtu)
3799 pmtu = loopback_mtu;
3800 } else if (nce != NULL) {
3801 /*
3802 * Make sure we don't exceed the interface MTU.
3803 * In the case of RTF_REJECT or RTF_BLACKHOLE we might not have
3804 * an ill. We'd use the above IP_MAXPACKET in that case just
3805 * to tell the transport something larger than zero.
3806 */
3807 if (ire->ire_type & (IRE_MULTICAST|IRE_BROADCAST)) {
3808 if (nce->nce_common->ncec_ill->ill_mc_mtu < pmtu)
3809 pmtu = nce->nce_common->ncec_ill->ill_mc_mtu;
3810 if (nce->nce_common->ncec_ill != nce->nce_ill &&
3811 nce->nce_ill->ill_mc_mtu < pmtu) {
3812 /*
3813 * for interfaces in an IPMP group, the mtu of
3814 * the nce_ill (under_ill) could be different
3815 * from the mtu of the ncec_ill, so we take the
3816 * min of the two.
3817 */
3818 pmtu = nce->nce_ill->ill_mc_mtu;
3819 }
3820 } else {
3821 if (nce->nce_common->ncec_ill->ill_mtu < pmtu)
3822 pmtu = nce->nce_common->ncec_ill->ill_mtu;
3823 if (nce->nce_common->ncec_ill != nce->nce_ill &&
3824 nce->nce_ill->ill_mtu < pmtu) {
3825 /*
3826 * for interfaces in an IPMP group, the mtu of
3827 * the nce_ill (under_ill) could be different
3828 * from the mtu of the ncec_ill, so we take the
3829 * min of the two.
3830 */
3831 pmtu = nce->nce_ill->ill_mtu;
3832 }
3833 }
3834 }
3835
3836 /*
3837 * Handle the IPV6_USE_MIN_MTU socket option or ancillary data.
3838 * Only applies to IPv6.
3839 */
3840 if (!(ixa->ixa_flags & IXAF_IS_IPV4)) {
3841 if (ixa->ixa_flags & IXAF_USE_MIN_MTU) {
3842 switch (ixa->ixa_use_min_mtu) {
3843 case IPV6_USE_MIN_MTU_MULTICAST:
3844 if (ire->ire_type & IRE_MULTICAST)
3845 pmtu = IPV6_MIN_MTU;
3846 break;
3847 case IPV6_USE_MIN_MTU_ALWAYS:
3848 pmtu = IPV6_MIN_MTU;
3849 break;
3850 case IPV6_USE_MIN_MTU_NEVER:
3851 break;
3852 }
3853 } else {
3854 /* Default is IPV6_USE_MIN_MTU_MULTICAST */
3855 if (ire->ire_type & IRE_MULTICAST)
3856 pmtu = IPV6_MIN_MTU;
3857 }
3858 }
3859
3860 /*
3861 * For multirouted IPv6 packets, the IP layer will insert a 8-byte
3862 * fragment header in every packet. We compensate for those cases by
3863 * returning a smaller path MTU to the ULP.
3864 *
3865 * In the case of CGTP then ip_output will add a fragment header.
3866 * Make sure there is room for it by telling a smaller number
3867 * to the transport.
3868 *
3869 * When IXAF_IPV6_ADDR_FRAGHDR we subtract the frag hdr here
3870 * so the ULPs consistently see a iulp_pmtu and ip_get_pmtu()
3871 * which is the size of the packets it can send.
3872 */
3873 if (!(ixa->ixa_flags & IXAF_IS_IPV4)) {
3874 if ((ire->ire_flags & RTF_MULTIRT) ||
3875 (ixa->ixa_flags & IXAF_MULTIRT_MULTICAST)) {
3876 pmtu -= sizeof (ip6_frag_t);
3877 ixa->ixa_flags |= IXAF_IPV6_ADD_FRAGHDR;
3878 }
3879 }
3880
3881 return (pmtu);
3882 }
3883
3884 /*
3885 * Carve "len" bytes out of an mblk chain, consuming any we empty, and duping
3886 * the final piece where we don't. Return a pointer to the first mblk in the
3887 * result, and update the pointer to the next mblk to chew on. If anything
3888 * goes wrong (i.e., dupb fails), we waste everything in sight and return a
3889 * NULL pointer.
3890 */
3891 mblk_t *
3892 ip_carve_mp(mblk_t **mpp, ssize_t len)
3893 {
3894 mblk_t *mp0;
3895 mblk_t *mp1;
3896 mblk_t *mp2;
3897
3898 if (!len || !mpp || !(mp0 = *mpp))
3899 return (NULL);
3900 /* If we aren't going to consume the first mblk, we need a dup. */
3901 if (mp0->b_wptr - mp0->b_rptr > len) {
3902 mp1 = dupb(mp0);
3903 if (mp1) {
3904 /* Partition the data between the two mblks. */
3905 mp1->b_wptr = mp1->b_rptr + len;
3906 mp0->b_rptr = mp1->b_wptr;
3907 /*
3908 * after adjustments if mblk not consumed is now
3909 * unaligned, try to align it. If this fails free
3910 * all messages and let upper layer recover.
3911 */
3912 if (!OK_32PTR(mp0->b_rptr)) {
3913 if (!pullupmsg(mp0, -1)) {
3914 freemsg(mp0);
3915 freemsg(mp1);
3916 *mpp = NULL;
3917 return (NULL);
3918 }
3919 }
3920 }
3921 return (mp1);
3922 }
3923 /* Eat through as many mblks as we need to get len bytes. */
3924 len -= mp0->b_wptr - mp0->b_rptr;
3925 for (mp2 = mp1 = mp0; (mp2 = mp2->b_cont) != 0 && len; mp1 = mp2) {
3926 if (mp2->b_wptr - mp2->b_rptr > len) {
3927 /*
3928 * We won't consume the entire last mblk. Like
3929 * above, dup and partition it.
3930 */
3931 mp1->b_cont = dupb(mp2);
3932 mp1 = mp1->b_cont;
3933 if (!mp1) {
3934 /*
3935 * Trouble. Rather than go to a lot of
3936 * trouble to clean up, we free the messages.
3937 * This won't be any worse than losing it on
3938 * the wire.
3939 */
3940 freemsg(mp0);
3941 freemsg(mp2);
3942 *mpp = NULL;
3943 return (NULL);
3944 }
3945 mp1->b_wptr = mp1->b_rptr + len;
3946 mp2->b_rptr = mp1->b_wptr;
3947 /*
3948 * after adjustments if mblk not consumed is now
3949 * unaligned, try to align it. If this fails free
3950 * all messages and let upper layer recover.
3951 */
3952 if (!OK_32PTR(mp2->b_rptr)) {
3953 if (!pullupmsg(mp2, -1)) {
3954 freemsg(mp0);
3955 freemsg(mp2);
3956 *mpp = NULL;
3957 return (NULL);
3958 }
3959 }
3960 *mpp = mp2;
3961 return (mp0);
3962 }
3963 /* Decrement len by the amount we just got. */
3964 len -= mp2->b_wptr - mp2->b_rptr;
3965 }
3966 /*
3967 * len should be reduced to zero now. If not our caller has
3968 * screwed up.
3969 */
3970 if (len) {
3971 /* Shouldn't happen! */
3972 freemsg(mp0);
3973 *mpp = NULL;
3974 return (NULL);
3975 }
3976 /*
3977 * We consumed up to exactly the end of an mblk. Detach the part
3978 * we are returning from the rest of the chain.
3979 */
3980 mp1->b_cont = NULL;
3981 *mpp = mp2;
3982 return (mp0);
3983 }
3984
3985 /* The ill stream is being unplumbed. Called from ip_close */
3986 int
3987 ip_modclose(ill_t *ill)
3988 {
3989 boolean_t success;
3990 ipsq_t *ipsq;
3991 ipif_t *ipif;
3992 queue_t *q = ill->ill_rq;
3993 ip_stack_t *ipst = ill->ill_ipst;
3994 int i;
3995 arl_ill_common_t *ai = ill->ill_common;
3996
3997 /*
3998 * The punlink prior to this may have initiated a capability
3999 * negotiation. But ipsq_enter will block until that finishes or
4000 * times out.
4001 */
4002 success = ipsq_enter(ill, B_FALSE, NEW_OP);
4003
4004 /*
4005 * Open/close/push/pop is guaranteed to be single threaded
4006 * per stream by STREAMS. FS guarantees that all references
4007 * from top are gone before close is called. So there can't
4008 * be another close thread that has set CONDEMNED on this ill.
4009 * and cause ipsq_enter to return failure.
4010 */
4011 ASSERT(success);
4012 ipsq = ill->ill_phyint->phyint_ipsq;
4013
4014 /*
4015 * Mark it condemned. No new reference will be made to this ill.
4016 * Lookup functions will return an error. Threads that try to
4017 * increment the refcnt must check for ILL_CAN_LOOKUP. This ensures
4018 * that the refcnt will drop down to zero.
4019 */
4020 mutex_enter(&ill->ill_lock);
4021 ill->ill_state_flags |= ILL_CONDEMNED;
4022 for (ipif = ill->ill_ipif; ipif != NULL;
4023 ipif = ipif->ipif_next) {
4024 ipif->ipif_state_flags |= IPIF_CONDEMNED;
4025 }
4026 /*
4027 * Wake up anybody waiting to enter the ipsq. ipsq_enter
4028 * returns error if ILL_CONDEMNED is set
4029 */
4030 cv_broadcast(&ill->ill_cv);
4031 mutex_exit(&ill->ill_lock);
4032
4033 /*
4034 * Send all the deferred DLPI messages downstream which came in
4035 * during the small window right before ipsq_enter(). We do this
4036 * without waiting for the ACKs because all the ACKs for M_PROTO
4037 * messages are ignored in ip_rput() when ILL_CONDEMNED is set.
4038 */
4039 ill_dlpi_send_deferred(ill);
4040
4041 /*
4042 * Shut down fragmentation reassembly.
4043 * ill_frag_timer won't start a timer again.
4044 * Now cancel any existing timer
4045 */
4046 (void) untimeout(ill->ill_frag_timer_id);
4047 (void) ill_frag_timeout(ill, 0);
4048
4049 /*
4050 * Call ill_delete to bring down the ipifs, ilms and ill on
4051 * this ill. Then wait for the refcnts to drop to zero.
4052 * ill_is_freeable checks whether the ill is really quiescent.
4053 * Then make sure that threads that are waiting to enter the
4054 * ipsq have seen the error returned by ipsq_enter and have
4055 * gone away. Then we call ill_delete_tail which does the
4056 * DL_UNBIND_REQ with the driver and then qprocsoff.
4057 */
4058 ill_delete(ill);
4059 mutex_enter(&ill->ill_lock);
4060 while (!ill_is_freeable(ill))
4061 cv_wait(&ill->ill_cv, &ill->ill_lock);
4062
4063 while (ill->ill_waiters)
4064 cv_wait(&ill->ill_cv, &ill->ill_lock);
4065
4066 mutex_exit(&ill->ill_lock);
4067
4068 /*
4069 * ill_delete_tail drops reference on ill_ipst, but we need to keep
4070 * it held until the end of the function since the cleanup
4071 * below needs to be able to use the ip_stack_t.
4072 */
4073 netstack_hold(ipst->ips_netstack);
4074
4075 /* qprocsoff is done via ill_delete_tail */
4076 ill_delete_tail(ill);
4077 /*
4078 * synchronously wait for arp stream to unbind. After this, we
4079 * cannot get any data packets up from the driver.
4080 */
4081 arp_unbind_complete(ill);
4082 ASSERT(ill->ill_ipst == NULL);
4083
4084 /*
4085 * Walk through all conns and qenable those that have queued data.
4086 * Close synchronization needs this to
4087 * be done to ensure that all upper layers blocked
4088 * due to flow control to the closing device
4089 * get unblocked.
4090 */
4091 ip1dbg(("ip_wsrv: walking\n"));
4092 for (i = 0; i < TX_FANOUT_SIZE; i++) {
4093 conn_walk_drain(ipst, &ipst->ips_idl_tx_list[i]);
4094 }
4095
4096 /*
4097 * ai can be null if this is an IPv6 ill, or if the IPv4
4098 * stream is being torn down before ARP was plumbed (e.g.,
4099 * /sbin/ifconfig plumbing a stream twice, and encountering
4100 * an error
4101 */
4102 if (ai != NULL) {
4103 ASSERT(!ill->ill_isv6);
4104 mutex_enter(&ai->ai_lock);
4105 ai->ai_ill = NULL;
4106 if (ai->ai_arl == NULL) {
4107 mutex_destroy(&ai->ai_lock);
4108 kmem_free(ai, sizeof (*ai));
4109 } else {
4110 cv_signal(&ai->ai_ill_unplumb_done);
4111 mutex_exit(&ai->ai_lock);
4112 }
4113 }
4114
4115 mutex_enter(&ipst->ips_ip_mi_lock);
4116 mi_close_unlink(&ipst->ips_ip_g_head, (IDP)ill);
4117 mutex_exit(&ipst->ips_ip_mi_lock);
4118
4119 /*
4120 * credp could be null if the open didn't succeed and ip_modopen
4121 * itself calls ip_close.
4122 */
4123 if (ill->ill_credp != NULL)
4124 crfree(ill->ill_credp);
4125
4126 mutex_destroy(&ill->ill_saved_ire_lock);
4127 mutex_destroy(&ill->ill_lock);
4128 rw_destroy(&ill->ill_mcast_lock);
4129 mutex_destroy(&ill->ill_mcast_serializer);
4130 list_destroy(&ill->ill_nce);
4131
4132 /*
4133 * Now we are done with the module close pieces that
4134 * need the netstack_t.
4135 */
4136 netstack_rele(ipst->ips_netstack);
4137
4138 mi_close_free((IDP)ill);
4139 q->q_ptr = WR(q)->q_ptr = NULL;
4140
4141 ipsq_exit(ipsq);
4142
4143 return (0);
4144 }
4145
4146 /*
4147 * This is called as part of close() for IP, UDP, ICMP, and RTS
4148 * in order to quiesce the conn.
4149 */
4150 void
4151 ip_quiesce_conn(conn_t *connp)
4152 {
4153 boolean_t drain_cleanup_reqd = B_FALSE;
4154 boolean_t conn_ioctl_cleanup_reqd = B_FALSE;
4155 boolean_t ilg_cleanup_reqd = B_FALSE;
4156 ip_stack_t *ipst;
4157
4158 ASSERT(!IPCL_IS_TCP(connp));
4159 ipst = connp->conn_netstack->netstack_ip;
4160
4161 /*
4162 * Mark the conn as closing, and this conn must not be
4163 * inserted in future into any list. Eg. conn_drain_insert(),
4164 * won't insert this conn into the conn_drain_list.
4165 *
4166 * conn_idl, and conn_ilg cannot get set henceforth.
4167 */
4168 mutex_enter(&connp->conn_lock);
4169 ASSERT(!(connp->conn_state_flags & CONN_QUIESCED));
4170 connp->conn_state_flags |= CONN_CLOSING;
4171 if (connp->conn_idl != NULL)
4172 drain_cleanup_reqd = B_TRUE;
4173 if (connp->conn_oper_pending_ill != NULL)
4174 conn_ioctl_cleanup_reqd = B_TRUE;
4175 if (connp->conn_dhcpinit_ill != NULL) {
4176 ASSERT(connp->conn_dhcpinit_ill->ill_dhcpinit != 0);
4177 atomic_dec_32(&connp->conn_dhcpinit_ill->ill_dhcpinit);
4178 ill_set_inputfn(connp->conn_dhcpinit_ill);
4179 connp->conn_dhcpinit_ill = NULL;
4180 }
4181 if (connp->conn_ilg != NULL)
4182 ilg_cleanup_reqd = B_TRUE;
4183 mutex_exit(&connp->conn_lock);
4184
4185 if (conn_ioctl_cleanup_reqd)
4186 conn_ioctl_cleanup(connp);
4187
4188 if (is_system_labeled() && connp->conn_anon_port) {
4189 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
4190 connp->conn_mlp_type, connp->conn_proto,
4191 ntohs(connp->conn_lport), B_FALSE);
4192 connp->conn_anon_port = 0;
4193 }
4194 connp->conn_mlp_type = mlptSingle;
4195
4196 /*
4197 * Remove this conn from any fanout list it is on.
4198 * and then wait for any threads currently operating
4199 * on this endpoint to finish
4200 */
4201 ipcl_hash_remove(connp);
4202
4203 /*
4204 * Remove this conn from the drain list, and do any other cleanup that
4205 * may be required. (TCP conns are never flow controlled, and
4206 * conn_idl will be NULL.)
4207 */
4208 if (drain_cleanup_reqd && connp->conn_idl != NULL) {
4209 idl_t *idl = connp->conn_idl;
4210
4211 mutex_enter(&idl->idl_lock);
4212 conn_drain(connp, B_TRUE);
4213 mutex_exit(&idl->idl_lock);
4214 }
4215
4216 if (connp == ipst->ips_ip_g_mrouter)
4217 (void) ip_mrouter_done(ipst);
4218
4219 if (ilg_cleanup_reqd)
4220 ilg_delete_all(connp);
4221
4222 /*
4223 * Now conn refcnt can increase only thru CONN_INC_REF_LOCKED.
4224 * callers from write side can't be there now because close
4225 * is in progress. The only other caller is ipcl_walk
4226 * which checks for the condemned flag.
4227 */
4228 mutex_enter(&connp->conn_lock);
4229 connp->conn_state_flags |= CONN_CONDEMNED;
4230 while (connp->conn_ref != 1)
4231 cv_wait(&connp->conn_cv, &connp->conn_lock);
4232 connp->conn_state_flags |= CONN_QUIESCED;
4233 mutex_exit(&connp->conn_lock);
4234 }
4235
4236 /* ARGSUSED */
4237 int
4238 ip_close(queue_t *q, int flags)
4239 {
4240 conn_t *connp;
4241
4242 /*
4243 * Call the appropriate delete routine depending on whether this is
4244 * a module or device.
4245 */
4246 if (WR(q)->q_next != NULL) {
4247 /* This is a module close */
4248 return (ip_modclose((ill_t *)q->q_ptr));
4249 }
4250
4251 connp = q->q_ptr;
4252 ip_quiesce_conn(connp);
4253
4254 qprocsoff(q);
4255
4256 /*
4257 * Now we are truly single threaded on this stream, and can
4258 * delete the things hanging off the connp, and finally the connp.
4259 * We removed this connp from the fanout list, it cannot be
4260 * accessed thru the fanouts, and we already waited for the
4261 * conn_ref to drop to 0. We are already in close, so
4262 * there cannot be any other thread from the top. qprocsoff
4263 * has completed, and service has completed or won't run in
4264 * future.
4265 */
4266 ASSERT(connp->conn_ref == 1);
4267
4268 inet_minor_free(connp->conn_minor_arena, connp->conn_dev);
4269
4270 connp->conn_ref--;
4271 ipcl_conn_destroy(connp);
4272
4273 q->q_ptr = WR(q)->q_ptr = NULL;
4274 return (0);
4275 }
4276
4277 /*
4278 * Wapper around putnext() so that ip_rts_request can merely use
4279 * conn_recv.
4280 */
4281 /*ARGSUSED2*/
4282 static void
4283 ip_conn_input(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira)
4284 {
4285 conn_t *connp = (conn_t *)arg1;
4286
4287 putnext(connp->conn_rq, mp);
4288 }
4289
4290 /* Dummy in case ICMP error delivery is attempted to a /dev/ip instance */
4291 /* ARGSUSED */
4292 static void
4293 ip_conn_input_icmp(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira)
4294 {
4295 freemsg(mp);
4296 }
4297
4298 /*
4299 * Called when the module is about to be unloaded
4300 */
4301 void
4302 ip_ddi_destroy(void)
4303 {
4304 /* This needs to be called before destroying any transports. */
4305 mutex_enter(&cpu_lock);
4306 unregister_cpu_setup_func(ip_tp_cpu_update, NULL);
4307 mutex_exit(&cpu_lock);
4308
4309 tnet_fini();
4310
4311 icmp_ddi_g_destroy();
4312 rts_ddi_g_destroy();
4313 udp_ddi_g_destroy();
4314 sctp_ddi_g_destroy();
4315 tcp_ddi_g_destroy();
4316 ilb_ddi_g_destroy();
4317 dce_g_destroy();
4318 ipsec_policy_g_destroy();
4319 ipcl_g_destroy();
4320 ip_net_g_destroy();
4321 ip_ire_g_fini();
4322 inet_minor_destroy(ip_minor_arena_sa);
4323 #if defined(_LP64)
4324 inet_minor_destroy(ip_minor_arena_la);
4325 #endif
4326
4327 #ifdef DEBUG
4328 list_destroy(&ip_thread_list);
4329 rw_destroy(&ip_thread_rwlock);
4330 tsd_destroy(&ip_thread_data);
4331 #endif
4332
4333 netstack_unregister(NS_IP);
4334 }
4335
4336 /*
4337 * First step in cleanup.
4338 */
4339 /* ARGSUSED */
4340 static void
4341 ip_stack_shutdown(netstackid_t stackid, void *arg)
4342 {
4343 ip_stack_t *ipst = (ip_stack_t *)arg;
4344 kt_did_t ktid;
4345
4346 #ifdef NS_DEBUG
4347 printf("ip_stack_shutdown(%p, stack %d)\n", (void *)ipst, stackid);
4348 #endif
4349
4350 /*
4351 * Perform cleanup for special interfaces (loopback and IPMP).
4352 */
4353 ip_interface_cleanup(ipst);
4354
4355 /*
4356 * The *_hook_shutdown()s start the process of notifying any
4357 * consumers that things are going away.... nothing is destroyed.
4358 */
4359 ipv4_hook_shutdown(ipst);
4360 ipv6_hook_shutdown(ipst);
4361 arp_hook_shutdown(ipst);
4362
4363 mutex_enter(&ipst->ips_capab_taskq_lock);
4364 ktid = ipst->ips_capab_taskq_thread->t_did;
4365 ipst->ips_capab_taskq_quit = B_TRUE;
4366 cv_signal(&ipst->ips_capab_taskq_cv);
4367 mutex_exit(&ipst->ips_capab_taskq_lock);
4368
4369 /*
4370 * In rare occurrences, particularly on virtual hardware where CPUs can
4371 * be de-scheduled, the thread that we just signaled will not run until
4372 * after we have gotten through parts of ip_stack_fini. If that happens
4373 * then we'll try to grab the ips_capab_taskq_lock as part of returning
4374 * from cv_wait which no longer exists.
4375 */
4376 thread_join(ktid);
4377 }
4378
4379 /*
4380 * Free the IP stack instance.
4381 */
4382 static void
4383 ip_stack_fini(netstackid_t stackid, void *arg)
4384 {
4385 ip_stack_t *ipst = (ip_stack_t *)arg;
4386 int ret;
4387
4388 #ifdef NS_DEBUG
4389 printf("ip_stack_fini(%p, stack %d)\n", (void *)ipst, stackid);
4390 #endif
4391 /*
4392 * At this point, all of the notifications that the events and
4393 * protocols are going away have been run, meaning that we can
4394 * now set about starting to clean things up.
4395 */
4396 ipobs_fini(ipst);
4397 ipv4_hook_destroy(ipst);
4398 ipv6_hook_destroy(ipst);
4399 arp_hook_destroy(ipst);
4400 ip_net_destroy(ipst);
4401
4402 ipmp_destroy(ipst);
4403
4404 ip_kstat_fini(stackid, ipst->ips_ip_mibkp);
4405 ipst->ips_ip_mibkp = NULL;
4406 icmp_kstat_fini(stackid, ipst->ips_icmp_mibkp);
4407 ipst->ips_icmp_mibkp = NULL;
4408 ip_kstat2_fini(stackid, ipst->ips_ip_kstat);
4409 ipst->ips_ip_kstat = NULL;
4410 bzero(&ipst->ips_ip_statistics, sizeof (ipst->ips_ip_statistics));
4411 ip6_kstat_fini(stackid, ipst->ips_ip6_kstat);
4412 ipst->ips_ip6_kstat = NULL;
4413 bzero(&ipst->ips_ip6_statistics, sizeof (ipst->ips_ip6_statistics));
4414
4415 kmem_free(ipst->ips_propinfo_tbl,
4416 ip_propinfo_count * sizeof (mod_prop_info_t));
4417 ipst->ips_propinfo_tbl = NULL;
4418
4419 dce_stack_destroy(ipst);
4420 ip_mrouter_stack_destroy(ipst);
4421
4422 /*
4423 * Quiesce all of our timers. Note we set the quiesce flags before we
4424 * call untimeout. The slowtimers may actually kick off another instance
4425 * of the non-slow timers.
4426 */
4427 mutex_enter(&ipst->ips_igmp_timer_lock);
4428 ipst->ips_igmp_timer_quiesce = B_TRUE;
4429 mutex_exit(&ipst->ips_igmp_timer_lock);
4430
4431 mutex_enter(&ipst->ips_mld_timer_lock);
4432 ipst->ips_mld_timer_quiesce = B_TRUE;
4433 mutex_exit(&ipst->ips_mld_timer_lock);
4434
4435 mutex_enter(&ipst->ips_igmp_slowtimeout_lock);
4436 ipst->ips_igmp_slowtimeout_quiesce = B_TRUE;
4437 mutex_exit(&ipst->ips_igmp_slowtimeout_lock);
4438
4439 mutex_enter(&ipst->ips_mld_slowtimeout_lock);
4440 ipst->ips_mld_slowtimeout_quiesce = B_TRUE;
4441 mutex_exit(&ipst->ips_mld_slowtimeout_lock);
4442
4443 ret = untimeout(ipst->ips_igmp_timeout_id);
4444 if (ret == -1) {
4445 ASSERT(ipst->ips_igmp_timeout_id == 0);
4446 } else {
4447 ASSERT(ipst->ips_igmp_timeout_id != 0);
4448 ipst->ips_igmp_timeout_id = 0;
4449 }
4450 ret = untimeout(ipst->ips_igmp_slowtimeout_id);
4451 if (ret == -1) {
4452 ASSERT(ipst->ips_igmp_slowtimeout_id == 0);
4453 } else {
4454 ASSERT(ipst->ips_igmp_slowtimeout_id != 0);
4455 ipst->ips_igmp_slowtimeout_id = 0;
4456 }
4457 ret = untimeout(ipst->ips_mld_timeout_id);
4458 if (ret == -1) {
4459 ASSERT(ipst->ips_mld_timeout_id == 0);
4460 } else {
4461 ASSERT(ipst->ips_mld_timeout_id != 0);
4462 ipst->ips_mld_timeout_id = 0;
4463 }
4464 ret = untimeout(ipst->ips_mld_slowtimeout_id);
4465 if (ret == -1) {
4466 ASSERT(ipst->ips_mld_slowtimeout_id == 0);
4467 } else {
4468 ASSERT(ipst->ips_mld_slowtimeout_id != 0);
4469 ipst->ips_mld_slowtimeout_id = 0;
4470 }
4471
4472 ip_ire_fini(ipst);
4473 ip6_asp_free(ipst);
4474 conn_drain_fini(ipst);
4475 ipcl_destroy(ipst);
4476
4477 mutex_destroy(&ipst->ips_ndp4->ndp_g_lock);
4478 mutex_destroy(&ipst->ips_ndp6->ndp_g_lock);
4479 kmem_free(ipst->ips_ndp4, sizeof (ndp_g_t));
4480 ipst->ips_ndp4 = NULL;
4481 kmem_free(ipst->ips_ndp6, sizeof (ndp_g_t));
4482 ipst->ips_ndp6 = NULL;
4483
4484 if (ipst->ips_loopback_ksp != NULL) {
4485 kstat_delete_netstack(ipst->ips_loopback_ksp, stackid);
4486 ipst->ips_loopback_ksp = NULL;
4487 }
4488
4489 mutex_destroy(&ipst->ips_capab_taskq_lock);
4490 cv_destroy(&ipst->ips_capab_taskq_cv);
4491
4492 rw_destroy(&ipst->ips_srcid_lock);
4493
4494 mutex_destroy(&ipst->ips_ip_mi_lock);
4495 rw_destroy(&ipst->ips_ill_g_usesrc_lock);
4496
4497 mutex_destroy(&ipst->ips_igmp_timer_lock);
4498 mutex_destroy(&ipst->ips_mld_timer_lock);
4499 mutex_destroy(&ipst->ips_igmp_slowtimeout_lock);
4500 mutex_destroy(&ipst->ips_mld_slowtimeout_lock);
4501 mutex_destroy(&ipst->ips_ip_addr_avail_lock);
4502 rw_destroy(&ipst->ips_ill_g_lock);
4503
4504 kmem_free(ipst->ips_phyint_g_list, sizeof (phyint_list_t));
4505 ipst->ips_phyint_g_list = NULL;
4506 kmem_free(ipst->ips_ill_g_heads, sizeof (ill_g_head_t) * MAX_G_HEADS);
4507 ipst->ips_ill_g_heads = NULL;
4508
4509 ldi_ident_release(ipst->ips_ldi_ident);
4510 kmem_free(ipst, sizeof (*ipst));
4511 }
4512
4513 /*
4514 * This function is called from the TSD destructor, and is used to debug
4515 * reference count issues in IP. See block comment in <inet/ip_if.h> for
4516 * details.
4517 */
4518 static void
4519 ip_thread_exit(void *phash)
4520 {
4521 th_hash_t *thh = phash;
4522
4523 rw_enter(&ip_thread_rwlock, RW_WRITER);
4524 list_remove(&ip_thread_list, thh);
4525 rw_exit(&ip_thread_rwlock);
4526 mod_hash_destroy_hash(thh->thh_hash);
4527 kmem_free(thh, sizeof (*thh));
4528 }
4529
4530 /*
4531 * Called when the IP kernel module is loaded into the kernel
4532 */
4533 void
4534 ip_ddi_init(void)
4535 {
4536 ip_squeue_flag = ip_squeue_switch(ip_squeue_enter);
4537
4538 /*
4539 * For IP and TCP the minor numbers should start from 2 since we have 4
4540 * initial devices: ip, ip6, tcp, tcp6.
4541 */
4542 /*
4543 * If this is a 64-bit kernel, then create two separate arenas -
4544 * one for TLIs in the range of INET_MIN_DEV+2 through 2^^18-1, and the
4545 * other for socket apps in the range 2^^18 through 2^^32-1.
4546 */
4547 ip_minor_arena_la = NULL;
4548 ip_minor_arena_sa = NULL;
4549 #if defined(_LP64)
4550 if ((ip_minor_arena_sa = inet_minor_create("ip_minor_arena_sa",
4551 INET_MIN_DEV + 2, MAXMIN32, KM_SLEEP)) == NULL) {
4552 cmn_err(CE_PANIC,
4553 "ip_ddi_init: ip_minor_arena_sa creation failed\n");
4554 }
4555 if ((ip_minor_arena_la = inet_minor_create("ip_minor_arena_la",
4556 MAXMIN32 + 1, MAXMIN64, KM_SLEEP)) == NULL) {
4557 cmn_err(CE_PANIC,
4558 "ip_ddi_init: ip_minor_arena_la creation failed\n");
4559 }
4560 #else
4561 if ((ip_minor_arena_sa = inet_minor_create("ip_minor_arena_sa",
4562 INET_MIN_DEV + 2, MAXMIN, KM_SLEEP)) == NULL) {
4563 cmn_err(CE_PANIC,
4564 "ip_ddi_init: ip_minor_arena_sa creation failed\n");
4565 }
4566 #endif
4567 ip_poll_normal_ticks = MSEC_TO_TICK_ROUNDUP(ip_poll_normal_ms);
4568
4569 ipcl_g_init();
4570 ip_ire_g_init();
4571 ip_net_g_init();
4572
4573 #ifdef DEBUG
4574 tsd_create(&ip_thread_data, ip_thread_exit);
4575 rw_init(&ip_thread_rwlock, NULL, RW_DEFAULT, NULL);
4576 list_create(&ip_thread_list, sizeof (th_hash_t),
4577 offsetof(th_hash_t, thh_link));
4578 #endif
4579 ipsec_policy_g_init();
4580 tcp_ddi_g_init();
4581 sctp_ddi_g_init();
4582 dce_g_init();
4583
4584 /*
4585 * We want to be informed each time a stack is created or
4586 * destroyed in the kernel, so we can maintain the
4587 * set of udp_stack_t's.
4588 */
4589 netstack_register(NS_IP, ip_stack_init, ip_stack_shutdown,
4590 ip_stack_fini);
4591
4592 tnet_init();
4593
4594 udp_ddi_g_init();
4595 rts_ddi_g_init();
4596 icmp_ddi_g_init();
4597 ilb_ddi_g_init();
4598
4599 /* This needs to be called after all transports are initialized. */
4600 mutex_enter(&cpu_lock);
4601 register_cpu_setup_func(ip_tp_cpu_update, NULL);
4602 mutex_exit(&cpu_lock);
4603 }
4604
4605 /*
4606 * Initialize the IP stack instance.
4607 */
4608 static void *
4609 ip_stack_init(netstackid_t stackid, netstack_t *ns)
4610 {
4611 ip_stack_t *ipst;
4612 size_t arrsz;
4613 major_t major;
4614
4615 #ifdef NS_DEBUG
4616 printf("ip_stack_init(stack %d)\n", stackid);
4617 #endif
4618
4619 ipst = (ip_stack_t *)kmem_zalloc(sizeof (*ipst), KM_SLEEP);
4620 ipst->ips_netstack = ns;
4621
4622 ipst->ips_ill_g_heads = kmem_zalloc(sizeof (ill_g_head_t) * MAX_G_HEADS,
4623 KM_SLEEP);
4624 ipst->ips_phyint_g_list = kmem_zalloc(sizeof (phyint_list_t),
4625 KM_SLEEP);
4626 ipst->ips_ndp4 = kmem_zalloc(sizeof (ndp_g_t), KM_SLEEP);
4627 ipst->ips_ndp6 = kmem_zalloc(sizeof (ndp_g_t), KM_SLEEP);
4628 mutex_init(&ipst->ips_ndp4->ndp_g_lock, NULL, MUTEX_DEFAULT, NULL);
4629 mutex_init(&ipst->ips_ndp6->ndp_g_lock, NULL, MUTEX_DEFAULT, NULL);
4630
4631 mutex_init(&ipst->ips_igmp_timer_lock, NULL, MUTEX_DEFAULT, NULL);
4632 ipst->ips_igmp_deferred_next = INFINITY;
4633 mutex_init(&ipst->ips_mld_timer_lock, NULL, MUTEX_DEFAULT, NULL);
4634 ipst->ips_mld_deferred_next = INFINITY;
4635 mutex_init(&ipst->ips_igmp_slowtimeout_lock, NULL, MUTEX_DEFAULT, NULL);
4636 mutex_init(&ipst->ips_mld_slowtimeout_lock, NULL, MUTEX_DEFAULT, NULL);
4637 mutex_init(&ipst->ips_ip_mi_lock, NULL, MUTEX_DEFAULT, NULL);
4638 mutex_init(&ipst->ips_ip_addr_avail_lock, NULL, MUTEX_DEFAULT, NULL);
4639 rw_init(&ipst->ips_ill_g_lock, NULL, RW_DEFAULT, NULL);
4640 rw_init(&ipst->ips_ill_g_usesrc_lock, NULL, RW_DEFAULT, NULL);
4641
4642 ipcl_init(ipst);
4643 ip_ire_init(ipst);
4644 ip6_asp_init(ipst);
4645 ipif_init(ipst);
4646 conn_drain_init(ipst);
4647 ip_mrouter_stack_init(ipst);
4648 dce_stack_init(ipst);
4649
4650 ipst->ips_ip_multirt_log_interval = 1000;
4651
4652 ipst->ips_ill_index = 1;
4653
4654 ipst->ips_saved_ip_forwarding = -1;
4655 ipst->ips_reg_vif_num = ALL_VIFS; /* Index to Register vif */
4656
4657 arrsz = ip_propinfo_count * sizeof (mod_prop_info_t);
4658 ipst->ips_propinfo_tbl = (mod_prop_info_t *)kmem_alloc(arrsz, KM_SLEEP);
4659 bcopy(ip_propinfo_tbl, ipst->ips_propinfo_tbl, arrsz);
4660
4661 ipst->ips_ip_mibkp = ip_kstat_init(stackid, ipst);
4662 ipst->ips_icmp_mibkp = icmp_kstat_init(stackid);
4663 ipst->ips_ip_kstat = ip_kstat2_init(stackid, &ipst->ips_ip_statistics);
4664 ipst->ips_ip6_kstat =
4665 ip6_kstat_init(stackid, &ipst->ips_ip6_statistics);
4666
4667 ipst->ips_ip_src_id = 1;
4668 rw_init(&ipst->ips_srcid_lock, NULL, RW_DEFAULT, NULL);
4669
4670 ipst->ips_src_generation = SRC_GENERATION_INITIAL;
4671
4672 ip_net_init(ipst, ns);
4673 ipv4_hook_init(ipst);
4674 ipv6_hook_init(ipst);
4675 arp_hook_init(ipst);
4676 ipmp_init(ipst);
4677 ipobs_init(ipst);
4678
4679 /*
4680 * Create the taskq dispatcher thread and initialize related stuff.
4681 */
4682 mutex_init(&ipst->ips_capab_taskq_lock, NULL, MUTEX_DEFAULT, NULL);
4683 cv_init(&ipst->ips_capab_taskq_cv, NULL, CV_DEFAULT, NULL);
4684 ipst->ips_capab_taskq_thread = thread_create(NULL, 0,
4685 ill_taskq_dispatch, ipst, 0, &p0, TS_RUN, minclsyspri);
4686
4687 major = mod_name_to_major(INET_NAME);
4688 (void) ldi_ident_from_major(major, &ipst->ips_ldi_ident);
4689 return (ipst);
4690 }
4691
4692 /*
4693 * Allocate and initialize a DLPI template of the specified length. (May be
4694 * called as writer.)
4695 */
4696 mblk_t *
4697 ip_dlpi_alloc(size_t len, t_uscalar_t prim)
4698 {
4699 mblk_t *mp;
4700
4701 mp = allocb(len, BPRI_MED);
4702 if (!mp)
4703 return (NULL);
4704
4705 /*
4706 * DLPIv2 says that DL_INFO_REQ and DL_TOKEN_REQ (the latter
4707 * of which we don't seem to use) are sent with M_PCPROTO, and
4708 * that other DLPI are M_PROTO.
4709 */
4710 if (prim == DL_INFO_REQ) {
4711 mp->b_datap->db_type = M_PCPROTO;
4712 } else {
4713 mp->b_datap->db_type = M_PROTO;
4714 }
4715
4716 mp->b_wptr = mp->b_rptr + len;
4717 bzero(mp->b_rptr, len);
4718 ((dl_unitdata_req_t *)mp->b_rptr)->dl_primitive = prim;
4719 return (mp);
4720 }
4721
4722 /*
4723 * Allocate and initialize a DLPI notification. (May be called as writer.)
4724 */
4725 mblk_t *
4726 ip_dlnotify_alloc(uint_t notification, uint_t data)
4727 {
4728 dl_notify_ind_t *notifyp;
4729 mblk_t *mp;
4730
4731 if ((mp = ip_dlpi_alloc(DL_NOTIFY_IND_SIZE, DL_NOTIFY_IND)) == NULL)
4732 return (NULL);
4733
4734 notifyp = (dl_notify_ind_t *)mp->b_rptr;
4735 notifyp->dl_notification = notification;
4736 notifyp->dl_data = data;
4737 return (mp);
4738 }
4739
4740 mblk_t *
4741 ip_dlnotify_alloc2(uint_t notification, uint_t data1, uint_t data2)
4742 {
4743 dl_notify_ind_t *notifyp;
4744 mblk_t *mp;
4745
4746 if ((mp = ip_dlpi_alloc(DL_NOTIFY_IND_SIZE, DL_NOTIFY_IND)) == NULL)
4747 return (NULL);
4748
4749 notifyp = (dl_notify_ind_t *)mp->b_rptr;
4750 notifyp->dl_notification = notification;
4751 notifyp->dl_data1 = data1;
4752 notifyp->dl_data2 = data2;
4753 return (mp);
4754 }
4755
4756 /*
4757 * Debug formatting routine. Returns a character string representation of the
4758 * addr in buf, of the form xxx.xxx.xxx.xxx. This routine takes the address
4759 * in the form of a ipaddr_t and calls ip_dot_saddr with a pointer.
4760 *
4761 * Once the ndd table-printing interfaces are removed, this can be changed to
4762 * standard dotted-decimal form.
4763 */
4764 char *
4765 ip_dot_addr(ipaddr_t addr, char *buf)
4766 {
4767 uint8_t *ap = (uint8_t *)&addr;
4768
4769 (void) mi_sprintf(buf, "%03d.%03d.%03d.%03d",
4770 ap[0] & 0xFF, ap[1] & 0xFF, ap[2] & 0xFF, ap[3] & 0xFF);
4771 return (buf);
4772 }
4773
4774 /*
4775 * Write the given MAC address as a printable string in the usual colon-
4776 * separated format.
4777 */
4778 const char *
4779 mac_colon_addr(const uint8_t *addr, size_t alen, char *buf, size_t buflen)
4780 {
4781 char *bp;
4782
4783 if (alen == 0 || buflen < 4)
4784 return ("?");
4785 bp = buf;
4786 for (;;) {
4787 /*
4788 * If there are more MAC address bytes available, but we won't
4789 * have any room to print them, then add "..." to the string
4790 * instead. See below for the 'magic number' explanation.
4791 */
4792 if ((alen == 2 && buflen < 6) || (alen > 2 && buflen < 7)) {
4793 (void) strcpy(bp, "...");
4794 break;
4795 }
4796 (void) sprintf(bp, "%02x", *addr++);
4797 bp += 2;
4798 if (--alen == 0)
4799 break;
4800 *bp++ = ':';
4801 buflen -= 3;
4802 /*
4803 * At this point, based on the first 'if' statement above,
4804 * either alen == 1 and buflen >= 3, or alen > 1 and
4805 * buflen >= 4. The first case leaves room for the final "xx"
4806 * number and trailing NUL byte. The second leaves room for at
4807 * least "...". Thus the apparently 'magic' numbers chosen for
4808 * that statement.
4809 */
4810 }
4811 return (buf);
4812 }
4813
4814 /*
4815 * Called when it is conceptually a ULP that would sent the packet
4816 * e.g., port unreachable and protocol unreachable. Check that the packet
4817 * would have passed the IPsec global policy before sending the error.
4818 *
4819 * Send an ICMP error after patching up the packet appropriately.
4820 * Uses ip_drop_input and bumps the appropriate MIB.
4821 */
4822 void
4823 ip_fanout_send_icmp_v4(mblk_t *mp, uint_t icmp_type, uint_t icmp_code,
4824 ip_recv_attr_t *ira)
4825 {
4826 ipha_t *ipha;
4827 boolean_t secure;
4828 ill_t *ill = ira->ira_ill;
4829 ip_stack_t *ipst = ill->ill_ipst;
4830 netstack_t *ns = ipst->ips_netstack;
4831 ipsec_stack_t *ipss = ns->netstack_ipsec;
4832
4833 secure = ira->ira_flags & IRAF_IPSEC_SECURE;
4834
4835 /*
4836 * We are generating an icmp error for some inbound packet.
4837 * Called from all ip_fanout_(udp, tcp, proto) functions.
4838 * Before we generate an error, check with global policy
4839 * to see whether this is allowed to enter the system. As
4840 * there is no "conn", we are checking with global policy.
4841 */
4842 ipha = (ipha_t *)mp->b_rptr;
4843 if (secure || ipss->ipsec_inbound_v4_policy_present) {
4844 mp = ipsec_check_global_policy(mp, NULL, ipha, NULL, ira, ns);
4845 if (mp == NULL)
4846 return;
4847 }
4848
4849 /* We never send errors for protocols that we do implement */
4850 if (ira->ira_protocol == IPPROTO_ICMP ||
4851 ira->ira_protocol == IPPROTO_IGMP) {
4852 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
4853 ip_drop_input("ip_fanout_send_icmp_v4", mp, ill);
4854 freemsg(mp);
4855 return;
4856 }
4857 /*
4858 * Have to correct checksum since
4859 * the packet might have been
4860 * fragmented and the reassembly code in ip_rput
4861 * does not restore the IP checksum.
4862 */
4863 ipha->ipha_hdr_checksum = 0;
4864 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
4865
4866 switch (icmp_type) {
4867 case ICMP_DEST_UNREACHABLE:
4868 switch (icmp_code) {
4869 case ICMP_PROTOCOL_UNREACHABLE:
4870 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInUnknownProtos);
4871 ip_drop_input("ipIfStatsInUnknownProtos", mp, ill);
4872 break;
4873 case ICMP_PORT_UNREACHABLE:
4874 BUMP_MIB(ill->ill_ip_mib, udpIfStatsNoPorts);
4875 ip_drop_input("ipIfStatsNoPorts", mp, ill);
4876 break;
4877 }
4878
4879 icmp_unreachable(mp, icmp_code, ira);
4880 break;
4881 default:
4882 #ifdef DEBUG
4883 panic("ip_fanout_send_icmp_v4: wrong type");
4884 /*NOTREACHED*/
4885 #else
4886 freemsg(mp);
4887 break;
4888 #endif
4889 }
4890 }
4891
4892 /*
4893 * Used to send an ICMP error message when a packet is received for
4894 * a protocol that is not supported. The mblk passed as argument
4895 * is consumed by this function.
4896 */
4897 void
4898 ip_proto_not_sup(mblk_t *mp, ip_recv_attr_t *ira)
4899 {
4900 ipha_t *ipha;
4901
4902 ipha = (ipha_t *)mp->b_rptr;
4903 if (ira->ira_flags & IRAF_IS_IPV4) {
4904 ASSERT(IPH_HDR_VERSION(ipha) == IP_VERSION);
4905 ip_fanout_send_icmp_v4(mp, ICMP_DEST_UNREACHABLE,
4906 ICMP_PROTOCOL_UNREACHABLE, ira);
4907 } else {
4908 ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION);
4909 ip_fanout_send_icmp_v6(mp, ICMP6_PARAM_PROB,
4910 ICMP6_PARAMPROB_NEXTHEADER, ira);
4911 }
4912 }
4913
4914 /*
4915 * Deliver a rawip packet to the given conn, possibly applying ipsec policy.
4916 * Handles IPv4 and IPv6.
4917 * We are responsible for disposing of mp, such as by freemsg() or putnext()
4918 * Caller is responsible for dropping references to the conn.
4919 */
4920 void
4921 ip_fanout_proto_conn(conn_t *connp, mblk_t *mp, ipha_t *ipha, ip6_t *ip6h,
4922 ip_recv_attr_t *ira)
4923 {
4924 ill_t *ill = ira->ira_ill;
4925 ip_stack_t *ipst = ill->ill_ipst;
4926 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
4927 boolean_t secure;
4928 uint_t protocol = ira->ira_protocol;
4929 iaflags_t iraflags = ira->ira_flags;
4930 queue_t *rq;
4931
4932 secure = iraflags & IRAF_IPSEC_SECURE;
4933
4934 rq = connp->conn_rq;
4935 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld : !canputnext(rq)) {
4936 switch (protocol) {
4937 case IPPROTO_ICMPV6:
4938 BUMP_MIB(ill->ill_icmp6_mib, ipv6IfIcmpInOverflows);
4939 break;
4940 case IPPROTO_ICMP:
4941 BUMP_MIB(&ipst->ips_icmp_mib, icmpInOverflows);
4942 break;
4943 default:
4944 BUMP_MIB(ill->ill_ip_mib, rawipIfStatsInOverflows);
4945 break;
4946 }
4947 freemsg(mp);
4948 return;
4949 }
4950
4951 ASSERT(!(IPCL_IS_IPTUN(connp)));
4952
4953 if (((iraflags & IRAF_IS_IPV4) ?
4954 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
4955 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
4956 secure) {
4957 mp = ipsec_check_inbound_policy(mp, connp, ipha,
4958 ip6h, ira);
4959 if (mp == NULL) {
4960 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
4961 /* Note that mp is NULL */
4962 ip_drop_input("ipIfStatsInDiscards", mp, ill);
4963 return;
4964 }
4965 }
4966
4967 if (iraflags & IRAF_ICMP_ERROR) {
4968 (connp->conn_recvicmp)(connp, mp, NULL, ira);
4969 } else {
4970 ill_t *rill = ira->ira_rill;
4971
4972 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
4973 ira->ira_ill = ira->ira_rill = NULL;
4974 /* Send it upstream */
4975 (connp->conn_recv)(connp, mp, NULL, ira);
4976 ira->ira_ill = ill;
4977 ira->ira_rill = rill;
4978 }
4979 }
4980
4981 /*
4982 * Handle protocols with which IP is less intimate. There
4983 * can be more than one stream bound to a particular
4984 * protocol. When this is the case, normally each one gets a copy
4985 * of any incoming packets.
4986 *
4987 * IPsec NOTE :
4988 *
4989 * Don't allow a secure packet going up a non-secure connection.
4990 * We don't allow this because
4991 *
4992 * 1) Reply might go out in clear which will be dropped at
4993 * the sending side.
4994 * 2) If the reply goes out in clear it will give the
4995 * adversary enough information for getting the key in
4996 * most of the cases.
4997 *
4998 * Moreover getting a secure packet when we expect clear
4999 * implies that SA's were added without checking for
5000 * policy on both ends. This should not happen once ISAKMP
5001 * is used to negotiate SAs as SAs will be added only after
5002 * verifying the policy.
5003 *
5004 * Zones notes:
5005 * Earlier in ip_input on a system with multiple shared-IP zones we
5006 * duplicate the multicast and broadcast packets and send them up
5007 * with each explicit zoneid that exists on that ill.
5008 * This means that here we can match the zoneid with SO_ALLZONES being special.
5009 */
5010 void
5011 ip_fanout_proto_v4(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
5012 {
5013 mblk_t *mp1;
5014 ipaddr_t laddr;
5015 conn_t *connp, *first_connp, *next_connp;
5016 connf_t *connfp;
5017 ill_t *ill = ira->ira_ill;
5018 ip_stack_t *ipst = ill->ill_ipst;
5019
5020 laddr = ipha->ipha_dst;
5021
5022 connfp = &ipst->ips_ipcl_proto_fanout_v4[ira->ira_protocol];
5023 mutex_enter(&connfp->connf_lock);
5024 connp = connfp->connf_head;
5025 for (connp = connfp->connf_head; connp != NULL;
5026 connp = connp->conn_next) {
5027 /* Note: IPCL_PROTO_MATCH includes conn_wantpacket */
5028 if (IPCL_PROTO_MATCH(connp, ira, ipha) &&
5029 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5030 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp))) {
5031 break;
5032 }
5033 }
5034
5035 if (connp == NULL) {
5036 /*
5037 * No one bound to these addresses. Is
5038 * there a client that wants all
5039 * unclaimed datagrams?
5040 */
5041 mutex_exit(&connfp->connf_lock);
5042 ip_fanout_send_icmp_v4(mp, ICMP_DEST_UNREACHABLE,
5043 ICMP_PROTOCOL_UNREACHABLE, ira);
5044 return;
5045 }
5046
5047 ASSERT(IPCL_IS_NONSTR(connp) || connp->conn_rq != NULL);
5048
5049 CONN_INC_REF(connp);
5050 first_connp = connp;
5051 connp = connp->conn_next;
5052
5053 for (;;) {
5054 while (connp != NULL) {
5055 /* Note: IPCL_PROTO_MATCH includes conn_wantpacket */
5056 if (IPCL_PROTO_MATCH(connp, ira, ipha) &&
5057 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5058 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5059 ira, connp)))
5060 break;
5061 connp = connp->conn_next;
5062 }
5063
5064 if (connp == NULL) {
5065 /* No more interested clients */
5066 connp = first_connp;
5067 break;
5068 }
5069 if (((mp1 = dupmsg(mp)) == NULL) &&
5070 ((mp1 = copymsg(mp)) == NULL)) {
5071 /* Memory allocation failed */
5072 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5073 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5074 connp = first_connp;
5075 break;
5076 }
5077
5078 CONN_INC_REF(connp);
5079 mutex_exit(&connfp->connf_lock);
5080
5081 ip_fanout_proto_conn(connp, mp1, (ipha_t *)mp1->b_rptr, NULL,
5082 ira);
5083
5084 mutex_enter(&connfp->connf_lock);
5085 /* Follow the next pointer before releasing the conn. */
5086 next_connp = connp->conn_next;
5087 CONN_DEC_REF(connp);
5088 connp = next_connp;
5089 }
5090
5091 /* Last one. Send it upstream. */
5092 mutex_exit(&connfp->connf_lock);
5093
5094 ip_fanout_proto_conn(connp, mp, ipha, NULL, ira);
5095
5096 CONN_DEC_REF(connp);
5097 }
5098
5099 /*
5100 * If we have a IPsec NAT-Traversal packet, strip the zero-SPI or
5101 * pass it along to ESP if the SPI is non-zero. Returns the mblk if the mblk
5102 * is not consumed.
5103 *
5104 * One of three things can happen, all of which affect the passed-in mblk:
5105 *
5106 * 1.) The packet is stock UDP and gets its zero-SPI stripped. Return mblk..
5107 *
5108 * 2.) The packet is ESP-in-UDP, gets transformed into an equivalent
5109 * ESP packet, and is passed along to ESP for consumption. Return NULL.
5110 *
5111 * 3.) The packet is an ESP-in-UDP Keepalive. Drop it and return NULL.
5112 */
5113 mblk_t *
5114 zero_spi_check(mblk_t *mp, ip_recv_attr_t *ira)
5115 {
5116 int shift, plen, iph_len;
5117 ipha_t *ipha;
5118 udpha_t *udpha;
5119 uint32_t *spi;
5120 uint32_t esp_ports;
5121 uint8_t *orptr;
5122 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
5123 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
5124
5125 ipha = (ipha_t *)mp->b_rptr;
5126 iph_len = ira->ira_ip_hdr_length;
5127 plen = ira->ira_pktlen;
5128
5129 if (plen - iph_len - sizeof (udpha_t) < sizeof (uint32_t)) {
5130 /*
5131 * Most likely a keepalive for the benefit of an intervening
5132 * NAT. These aren't for us, per se, so drop it.
5133 *
5134 * RFC 3947/8 doesn't say for sure what to do for 2-3
5135 * byte packets (keepalives are 1-byte), but we'll drop them
5136 * also.
5137 */
5138 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
5139 DROPPER(ipss, ipds_esp_nat_t_ka), &ipss->ipsec_dropper);
5140 return (NULL);
5141 }
5142
5143 if (MBLKL(mp) < iph_len + sizeof (udpha_t) + sizeof (*spi)) {
5144 /* might as well pull it all up - it might be ESP. */
5145 if (!pullupmsg(mp, -1)) {
5146 ip_drop_packet(mp, B_TRUE, ira->ira_ill,
5147 DROPPER(ipss, ipds_esp_nomem),
5148 &ipss->ipsec_dropper);
5149 return (NULL);
5150 }
5151
5152 ipha = (ipha_t *)mp->b_rptr;
5153 }
5154 spi = (uint32_t *)(mp->b_rptr + iph_len + sizeof (udpha_t));
5155 if (*spi == 0) {
5156 /* UDP packet - remove 0-spi. */
5157 shift = sizeof (uint32_t);
5158 } else {
5159 /* ESP-in-UDP packet - reduce to ESP. */
5160 ipha->ipha_protocol = IPPROTO_ESP;
5161 shift = sizeof (udpha_t);
5162 }
5163
5164 /* Fix IP header */
5165 ira->ira_pktlen = (plen - shift);
5166 ipha->ipha_length = htons(ira->ira_pktlen);
5167 ipha->ipha_hdr_checksum = 0;
5168
5169 orptr = mp->b_rptr;
5170 mp->b_rptr += shift;
5171
5172 udpha = (udpha_t *)(orptr + iph_len);
5173 if (*spi == 0) {
5174 ASSERT((uint8_t *)ipha == orptr);
5175 udpha->uha_length = htons(plen - shift - iph_len);
5176 iph_len += sizeof (udpha_t); /* For the call to ovbcopy(). */
5177 esp_ports = 0;
5178 } else {
5179 esp_ports = *((uint32_t *)udpha);
5180 ASSERT(esp_ports != 0);
5181 }
5182 ovbcopy(orptr, orptr + shift, iph_len);
5183 if (esp_ports != 0) /* Punt up for ESP processing. */ {
5184 ipha = (ipha_t *)(orptr + shift);
5185
5186 ira->ira_flags |= IRAF_ESP_UDP_PORTS;
5187 ira->ira_esp_udp_ports = esp_ports;
5188 ip_fanout_v4(mp, ipha, ira);
5189 return (NULL);
5190 }
5191 return (mp);
5192 }
5193
5194 /*
5195 * Deliver a udp packet to the given conn, possibly applying ipsec policy.
5196 * Handles IPv4 and IPv6.
5197 * We are responsible for disposing of mp, such as by freemsg() or putnext()
5198 * Caller is responsible for dropping references to the conn.
5199 */
5200 void
5201 ip_fanout_udp_conn(conn_t *connp, mblk_t *mp, ipha_t *ipha, ip6_t *ip6h,
5202 ip_recv_attr_t *ira)
5203 {
5204 ill_t *ill = ira->ira_ill;
5205 ip_stack_t *ipst = ill->ill_ipst;
5206 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
5207 boolean_t secure;
5208 iaflags_t iraflags = ira->ira_flags;
5209
5210 secure = iraflags & IRAF_IPSEC_SECURE;
5211
5212 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld :
5213 !canputnext(connp->conn_rq)) {
5214 BUMP_MIB(ill->ill_ip_mib, udpIfStatsInOverflows);
5215 freemsg(mp);
5216 return;
5217 }
5218
5219 if (((iraflags & IRAF_IS_IPV4) ?
5220 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
5221 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
5222 secure) {
5223 mp = ipsec_check_inbound_policy(mp, connp, ipha,
5224 ip6h, ira);
5225 if (mp == NULL) {
5226 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5227 /* Note that mp is NULL */
5228 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5229 return;
5230 }
5231 }
5232
5233 /*
5234 * Since this code is not used for UDP unicast we don't need a NAT_T
5235 * check. Only ip_fanout_v4 has that check.
5236 */
5237 if (ira->ira_flags & IRAF_ICMP_ERROR) {
5238 (connp->conn_recvicmp)(connp, mp, NULL, ira);
5239 } else {
5240 ill_t *rill = ira->ira_rill;
5241
5242 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
5243 ira->ira_ill = ira->ira_rill = NULL;
5244 /* Send it upstream */
5245 (connp->conn_recv)(connp, mp, NULL, ira);
5246 ira->ira_ill = ill;
5247 ira->ira_rill = rill;
5248 }
5249 }
5250
5251 /*
5252 * Fanout for UDP packets that are multicast or broadcast, and ICMP errors.
5253 * (Unicast fanout is handled in ip_input_v4.)
5254 *
5255 * If SO_REUSEADDR is set all multicast and broadcast packets
5256 * will be delivered to all conns bound to the same port.
5257 *
5258 * If there is at least one matching AF_INET receiver, then we will
5259 * ignore any AF_INET6 receivers.
5260 * In the special case where an AF_INET socket binds to 0.0.0.0/<port> and an
5261 * AF_INET6 socket binds to ::/<port>, only the AF_INET socket receives the IPv4
5262 * packets.
5263 *
5264 * Zones notes:
5265 * Earlier in ip_input on a system with multiple shared-IP zones we
5266 * duplicate the multicast and broadcast packets and send them up
5267 * with each explicit zoneid that exists on that ill.
5268 * This means that here we can match the zoneid with SO_ALLZONES being special.
5269 */
5270 void
5271 ip_fanout_udp_multi_v4(mblk_t *mp, ipha_t *ipha, uint16_t lport, uint16_t fport,
5272 ip_recv_attr_t *ira)
5273 {
5274 ipaddr_t laddr;
5275 in6_addr_t v6faddr;
5276 conn_t *connp;
5277 connf_t *connfp;
5278 ipaddr_t faddr;
5279 ill_t *ill = ira->ira_ill;
5280 ip_stack_t *ipst = ill->ill_ipst;
5281
5282 ASSERT(ira->ira_flags & (IRAF_MULTIBROADCAST|IRAF_ICMP_ERROR));
5283
5284 laddr = ipha->ipha_dst;
5285 faddr = ipha->ipha_src;
5286
5287 connfp = &ipst->ips_ipcl_udp_fanout[IPCL_UDP_HASH(lport, ipst)];
5288 mutex_enter(&connfp->connf_lock);
5289 connp = connfp->connf_head;
5290
5291 /*
5292 * If SO_REUSEADDR has been set on the first we send the
5293 * packet to all clients that have joined the group and
5294 * match the port.
5295 */
5296 while (connp != NULL) {
5297 if ((IPCL_UDP_MATCH(connp, lport, laddr, fport, faddr)) &&
5298 conn_wantpacket(connp, ira, ipha) &&
5299 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5300 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp)))
5301 break;
5302 connp = connp->conn_next;
5303 }
5304
5305 if (connp == NULL)
5306 goto notfound;
5307
5308 CONN_INC_REF(connp);
5309
5310 if (connp->conn_reuseaddr) {
5311 conn_t *first_connp = connp;
5312 conn_t *next_connp;
5313 mblk_t *mp1;
5314
5315 connp = connp->conn_next;
5316 for (;;) {
5317 while (connp != NULL) {
5318 if (IPCL_UDP_MATCH(connp, lport, laddr,
5319 fport, faddr) &&
5320 conn_wantpacket(connp, ira, ipha) &&
5321 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5322 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5323 ira, connp)))
5324 break;
5325 connp = connp->conn_next;
5326 }
5327 if (connp == NULL) {
5328 /* No more interested clients */
5329 connp = first_connp;
5330 break;
5331 }
5332 if (((mp1 = dupmsg(mp)) == NULL) &&
5333 ((mp1 = copymsg(mp)) == NULL)) {
5334 /* Memory allocation failed */
5335 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5336 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5337 connp = first_connp;
5338 break;
5339 }
5340 CONN_INC_REF(connp);
5341 mutex_exit(&connfp->connf_lock);
5342
5343 IP_STAT(ipst, ip_udp_fanmb);
5344 ip_fanout_udp_conn(connp, mp1, (ipha_t *)mp1->b_rptr,
5345 NULL, ira);
5346 mutex_enter(&connfp->connf_lock);
5347 /* Follow the next pointer before releasing the conn */
5348 next_connp = connp->conn_next;
5349 CONN_DEC_REF(connp);
5350 connp = next_connp;
5351 }
5352 }
5353
5354 /* Last one. Send it upstream. */
5355 mutex_exit(&connfp->connf_lock);
5356 IP_STAT(ipst, ip_udp_fanmb);
5357 ip_fanout_udp_conn(connp, mp, ipha, NULL, ira);
5358 CONN_DEC_REF(connp);
5359 return;
5360
5361 notfound:
5362 mutex_exit(&connfp->connf_lock);
5363 /*
5364 * IPv6 endpoints bound to multicast IPv4-mapped addresses
5365 * have already been matched above, since they live in the IPv4
5366 * fanout tables. This implies we only need to
5367 * check for IPv6 in6addr_any endpoints here.
5368 * Thus we compare using ipv6_all_zeros instead of the destination
5369 * address, except for the multicast group membership lookup which
5370 * uses the IPv4 destination.
5371 */
5372 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_src, &v6faddr);
5373 connfp = &ipst->ips_ipcl_udp_fanout[IPCL_UDP_HASH(lport, ipst)];
5374 mutex_enter(&connfp->connf_lock);
5375 connp = connfp->connf_head;
5376 /*
5377 * IPv4 multicast packet being delivered to an AF_INET6
5378 * in6addr_any endpoint.
5379 * Need to check conn_wantpacket(). Note that we use conn_wantpacket()
5380 * and not conn_wantpacket_v6() since any multicast membership is
5381 * for an IPv4-mapped multicast address.
5382 */
5383 while (connp != NULL) {
5384 if (IPCL_UDP_MATCH_V6(connp, lport, ipv6_all_zeros,
5385 fport, v6faddr) &&
5386 conn_wantpacket(connp, ira, ipha) &&
5387 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5388 tsol_receive_local(mp, &laddr, IPV4_VERSION, ira, connp)))
5389 break;
5390 connp = connp->conn_next;
5391 }
5392
5393 if (connp == NULL) {
5394 /*
5395 * No one bound to this port. Is
5396 * there a client that wants all
5397 * unclaimed datagrams?
5398 */
5399 mutex_exit(&connfp->connf_lock);
5400
5401 if (ipst->ips_ipcl_proto_fanout_v4[IPPROTO_UDP].connf_head !=
5402 NULL) {
5403 ASSERT(ira->ira_protocol == IPPROTO_UDP);
5404 ip_fanout_proto_v4(mp, ipha, ira);
5405 } else {
5406 /*
5407 * We used to attempt to send an icmp error here, but
5408 * since this is known to be a multicast packet
5409 * and we don't send icmp errors in response to
5410 * multicast, just drop the packet and give up sooner.
5411 */
5412 BUMP_MIB(ill->ill_ip_mib, udpIfStatsNoPorts);
5413 freemsg(mp);
5414 }
5415 return;
5416 }
5417 ASSERT(IPCL_IS_NONSTR(connp) || connp->conn_rq != NULL);
5418
5419 /*
5420 * If SO_REUSEADDR has been set on the first we send the
5421 * packet to all clients that have joined the group and
5422 * match the port.
5423 */
5424 if (connp->conn_reuseaddr) {
5425 conn_t *first_connp = connp;
5426 conn_t *next_connp;
5427 mblk_t *mp1;
5428
5429 CONN_INC_REF(connp);
5430 connp = connp->conn_next;
5431 for (;;) {
5432 while (connp != NULL) {
5433 if (IPCL_UDP_MATCH_V6(connp, lport,
5434 ipv6_all_zeros, fport, v6faddr) &&
5435 conn_wantpacket(connp, ira, ipha) &&
5436 (!(ira->ira_flags & IRAF_SYSTEM_LABELED) ||
5437 tsol_receive_local(mp, &laddr, IPV4_VERSION,
5438 ira, connp)))
5439 break;
5440 connp = connp->conn_next;
5441 }
5442 if (connp == NULL) {
5443 /* No more interested clients */
5444 connp = first_connp;
5445 break;
5446 }
5447 if (((mp1 = dupmsg(mp)) == NULL) &&
5448 ((mp1 = copymsg(mp)) == NULL)) {
5449 /* Memory allocation failed */
5450 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
5451 ip_drop_input("ipIfStatsInDiscards", mp, ill);
5452 connp = first_connp;
5453 break;
5454 }
5455 CONN_INC_REF(connp);
5456 mutex_exit(&connfp->connf_lock);
5457
5458 IP_STAT(ipst, ip_udp_fanmb);
5459 ip_fanout_udp_conn(connp, mp1, (ipha_t *)mp1->b_rptr,
5460 NULL, ira);
5461 mutex_enter(&connfp->connf_lock);
5462 /* Follow the next pointer before releasing the conn */
5463 next_connp = connp->conn_next;
5464 CONN_DEC_REF(connp);
5465 connp = next_connp;
5466 }
5467 }
5468
5469 /* Last one. Send it upstream. */
5470 mutex_exit(&connfp->connf_lock);
5471 IP_STAT(ipst, ip_udp_fanmb);
5472 ip_fanout_udp_conn(connp, mp, ipha, NULL, ira);
5473 CONN_DEC_REF(connp);
5474 }
5475
5476 /*
5477 * Split an incoming packet's IPv4 options into the label and the other options.
5478 * If 'allocate' is set it does memory allocation for the ip_pkt_t, including
5479 * clearing out any leftover label or options.
5480 * Otherwise it just makes ipp point into the packet.
5481 *
5482 * Returns zero if ok; ENOMEM if the buffer couldn't be allocated.
5483 */
5484 int
5485 ip_find_hdr_v4(ipha_t *ipha, ip_pkt_t *ipp, boolean_t allocate)
5486 {
5487 uchar_t *opt;
5488 uint32_t totallen;
5489 uint32_t optval;
5490 uint32_t optlen;
5491
5492 ipp->ipp_fields |= IPPF_HOPLIMIT | IPPF_TCLASS | IPPF_ADDR;
5493 ipp->ipp_hoplimit = ipha->ipha_ttl;
5494 ipp->ipp_type_of_service = ipha->ipha_type_of_service;
5495 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &ipp->ipp_addr);
5496
5497 /*
5498 * Get length (in 4 byte octets) of IP header options.
5499 */
5500 totallen = ipha->ipha_version_and_hdr_length -
5501 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
5502
5503 if (totallen == 0) {
5504 if (!allocate)
5505 return (0);
5506
5507 /* Clear out anything from a previous packet */
5508 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
5509 kmem_free(ipp->ipp_ipv4_options,
5510 ipp->ipp_ipv4_options_len);
5511 ipp->ipp_ipv4_options = NULL;
5512 ipp->ipp_ipv4_options_len = 0;
5513 ipp->ipp_fields &= ~IPPF_IPV4_OPTIONS;
5514 }
5515 if (ipp->ipp_fields & IPPF_LABEL_V4) {
5516 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
5517 ipp->ipp_label_v4 = NULL;
5518 ipp->ipp_label_len_v4 = 0;
5519 ipp->ipp_fields &= ~IPPF_LABEL_V4;
5520 }
5521 return (0);
5522 }
5523
5524 totallen <<= 2;
5525 opt = (uchar_t *)&ipha[1];
5526 if (!is_system_labeled()) {
5527
5528 copyall:
5529 if (!allocate) {
5530 if (totallen != 0) {
5531 ipp->ipp_ipv4_options = opt;
5532 ipp->ipp_ipv4_options_len = totallen;
5533 ipp->ipp_fields |= IPPF_IPV4_OPTIONS;
5534 }
5535 return (0);
5536 }
5537 /* Just copy all of options */
5538 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
5539 if (totallen == ipp->ipp_ipv4_options_len) {
5540 bcopy(opt, ipp->ipp_ipv4_options, totallen);
5541 return (0);
5542 }
5543 kmem_free(ipp->ipp_ipv4_options,
5544 ipp->ipp_ipv4_options_len);
5545 ipp->ipp_ipv4_options = NULL;
5546 ipp->ipp_ipv4_options_len = 0;
5547 ipp->ipp_fields &= ~IPPF_IPV4_OPTIONS;
5548 }
5549 if (totallen == 0)
5550 return (0);
5551
5552 ipp->ipp_ipv4_options = kmem_alloc(totallen, KM_NOSLEEP);
5553 if (ipp->ipp_ipv4_options == NULL)
5554 return (ENOMEM);
5555 ipp->ipp_ipv4_options_len = totallen;
5556 ipp->ipp_fields |= IPPF_IPV4_OPTIONS;
5557 bcopy(opt, ipp->ipp_ipv4_options, totallen);
5558 return (0);
5559 }
5560
5561 if (allocate && (ipp->ipp_fields & IPPF_LABEL_V4)) {
5562 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
5563 ipp->ipp_label_v4 = NULL;
5564 ipp->ipp_label_len_v4 = 0;
5565 ipp->ipp_fields &= ~IPPF_LABEL_V4;
5566 }
5567
5568 /*
5569 * Search for CIPSO option.
5570 * We assume CIPSO is first in options if it is present.
5571 * If it isn't, then ipp_opt_ipv4_options will not include the options
5572 * prior to the CIPSO option.
5573 */
5574 while (totallen != 0) {
5575 switch (optval = opt[IPOPT_OPTVAL]) {
5576 case IPOPT_EOL:
5577 return (0);
5578 case IPOPT_NOP:
5579 optlen = 1;
5580 break;
5581 default:
5582 if (totallen <= IPOPT_OLEN)
5583 return (EINVAL);
5584 optlen = opt[IPOPT_OLEN];
5585 if (optlen < 2)
5586 return (EINVAL);
5587 }
5588 if (optlen > totallen)
5589 return (EINVAL);
5590
5591 switch (optval) {
5592 case IPOPT_COMSEC:
5593 if (!allocate) {
5594 ipp->ipp_label_v4 = opt;
5595 ipp->ipp_label_len_v4 = optlen;
5596 ipp->ipp_fields |= IPPF_LABEL_V4;
5597 } else {
5598 ipp->ipp_label_v4 = kmem_alloc(optlen,
5599 KM_NOSLEEP);
5600 if (ipp->ipp_label_v4 == NULL)
5601 return (ENOMEM);
5602 ipp->ipp_label_len_v4 = optlen;
5603 ipp->ipp_fields |= IPPF_LABEL_V4;
5604 bcopy(opt, ipp->ipp_label_v4, optlen);
5605 }
5606 totallen -= optlen;
5607 opt += optlen;
5608
5609 /* Skip padding bytes until we get to a multiple of 4 */
5610 while ((totallen & 3) != 0 && opt[0] == IPOPT_NOP) {
5611 totallen--;
5612 opt++;
5613 }
5614 /* Remaining as ipp_ipv4_options */
5615 goto copyall;
5616 }
5617 totallen -= optlen;
5618 opt += optlen;
5619 }
5620 /* No CIPSO found; return everything as ipp_ipv4_options */
5621 totallen = ipha->ipha_version_and_hdr_length -
5622 (uint8_t)((IP_VERSION << 4) + IP_SIMPLE_HDR_LENGTH_IN_WORDS);
5623 totallen <<= 2;
5624 opt = (uchar_t *)&ipha[1];
5625 goto copyall;
5626 }
5627
5628 /*
5629 * Efficient versions of lookup for an IRE when we only
5630 * match the address.
5631 * For RTF_REJECT or BLACKHOLE we return IRE_NOROUTE.
5632 * Does not handle multicast addresses.
5633 */
5634 uint_t
5635 ip_type_v4(ipaddr_t addr, ip_stack_t *ipst)
5636 {
5637 ire_t *ire;
5638 uint_t result;
5639
5640 ire = ire_ftable_lookup_simple_v4(addr, 0, ipst, NULL);
5641 ASSERT(ire != NULL);
5642 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))
5643 result = IRE_NOROUTE;
5644 else
5645 result = ire->ire_type;
5646 ire_refrele(ire);
5647 return (result);
5648 }
5649
5650 /*
5651 * Efficient versions of lookup for an IRE when we only
5652 * match the address.
5653 * For RTF_REJECT or BLACKHOLE we return IRE_NOROUTE.
5654 * Does not handle multicast addresses.
5655 */
5656 uint_t
5657 ip_type_v6(const in6_addr_t *addr, ip_stack_t *ipst)
5658 {
5659 ire_t *ire;
5660 uint_t result;
5661
5662 ire = ire_ftable_lookup_simple_v6(addr, 0, ipst, NULL);
5663 ASSERT(ire != NULL);
5664 if (ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE))
5665 result = IRE_NOROUTE;
5666 else
5667 result = ire->ire_type;
5668 ire_refrele(ire);
5669 return (result);
5670 }
5671
5672 /*
5673 * Nobody should be sending
5674 * packets up this stream
5675 */
5676 static void
5677 ip_lrput(queue_t *q, mblk_t *mp)
5678 {
5679 switch (mp->b_datap->db_type) {
5680 case M_FLUSH:
5681 /* Turn around */
5682 if (*mp->b_rptr & FLUSHW) {
5683 *mp->b_rptr &= ~FLUSHR;
5684 qreply(q, mp);
5685 return;
5686 }
5687 break;
5688 }
5689 freemsg(mp);
5690 }
5691
5692 /* Nobody should be sending packets down this stream */
5693 /* ARGSUSED */
5694 void
5695 ip_lwput(queue_t *q, mblk_t *mp)
5696 {
5697 freemsg(mp);
5698 }
5699
5700 /*
5701 * Move the first hop in any source route to ipha_dst and remove that part of
5702 * the source route. Called by other protocols. Errors in option formatting
5703 * are ignored - will be handled by ip_output_options. Return the final
5704 * destination (either ipha_dst or the last entry in a source route.)
5705 */
5706 ipaddr_t
5707 ip_massage_options(ipha_t *ipha, netstack_t *ns)
5708 {
5709 ipoptp_t opts;
5710 uchar_t *opt;
5711 uint8_t optval;
5712 uint8_t optlen;
5713 ipaddr_t dst;
5714 int i;
5715 ip_stack_t *ipst = ns->netstack_ip;
5716
5717 ip2dbg(("ip_massage_options\n"));
5718 dst = ipha->ipha_dst;
5719 for (optval = ipoptp_first(&opts, ipha);
5720 optval != IPOPT_EOL;
5721 optval = ipoptp_next(&opts)) {
5722 opt = opts.ipoptp_cur;
5723 switch (optval) {
5724 uint8_t off;
5725 case IPOPT_SSRR:
5726 case IPOPT_LSRR:
5727 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
5728 ip1dbg(("ip_massage_options: bad src route\n"));
5729 break;
5730 }
5731 optlen = opts.ipoptp_len;
5732 off = opt[IPOPT_OFFSET];
5733 off--;
5734 redo_srr:
5735 if (optlen < IP_ADDR_LEN ||
5736 off > optlen - IP_ADDR_LEN) {
5737 /* End of source route */
5738 ip1dbg(("ip_massage_options: end of SR\n"));
5739 break;
5740 }
5741 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
5742 ip1dbg(("ip_massage_options: next hop 0x%x\n",
5743 ntohl(dst)));
5744 /*
5745 * Check if our address is present more than
5746 * once as consecutive hops in source route.
5747 * XXX verify per-interface ip_forwarding
5748 * for source route?
5749 */
5750 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
5751 off += IP_ADDR_LEN;
5752 goto redo_srr;
5753 }
5754 if (dst == htonl(INADDR_LOOPBACK)) {
5755 ip1dbg(("ip_massage_options: loopback addr in "
5756 "source route!\n"));
5757 break;
5758 }
5759 /*
5760 * Update ipha_dst to be the first hop and remove the
5761 * first hop from the source route (by overwriting
5762 * part of the option with NOP options).
5763 */
5764 ipha->ipha_dst = dst;
5765 /* Put the last entry in dst */
5766 off = ((optlen - IP_ADDR_LEN - 3) & ~(IP_ADDR_LEN-1)) +
5767 3;
5768 bcopy(&opt[off], &dst, IP_ADDR_LEN);
5769
5770 ip1dbg(("ip_massage_options: last hop 0x%x\n",
5771 ntohl(dst)));
5772 /* Move down and overwrite */
5773 opt[IP_ADDR_LEN] = opt[0];
5774 opt[IP_ADDR_LEN+1] = opt[IPOPT_OLEN] - IP_ADDR_LEN;
5775 opt[IP_ADDR_LEN+2] = opt[IPOPT_OFFSET];
5776 for (i = 0; i < IP_ADDR_LEN; i++)
5777 opt[i] = IPOPT_NOP;
5778 break;
5779 }
5780 }
5781 return (dst);
5782 }
5783
5784 /*
5785 * Return the network mask
5786 * associated with the specified address.
5787 */
5788 ipaddr_t
5789 ip_net_mask(ipaddr_t addr)
5790 {
5791 uchar_t *up = (uchar_t *)&addr;
5792 ipaddr_t mask = 0;
5793 uchar_t *maskp = (uchar_t *)&mask;
5794
5795 #if defined(__i386) || defined(__amd64)
5796 #define TOTALLY_BRAIN_DAMAGED_C_COMPILER
5797 #endif
5798 #ifdef TOTALLY_BRAIN_DAMAGED_C_COMPILER
5799 maskp[0] = maskp[1] = maskp[2] = maskp[3] = 0;
5800 #endif
5801 if (CLASSD(addr)) {
5802 maskp[0] = 0xF0;
5803 return (mask);
5804 }
5805
5806 /* We assume Class E default netmask to be 32 */
5807 if (CLASSE(addr))
5808 return (0xffffffffU);
5809
5810 if (addr == 0)
5811 return (0);
5812 maskp[0] = 0xFF;
5813 if ((up[0] & 0x80) == 0)
5814 return (mask);
5815
5816 maskp[1] = 0xFF;
5817 if ((up[0] & 0xC0) == 0x80)
5818 return (mask);
5819
5820 maskp[2] = 0xFF;
5821 if ((up[0] & 0xE0) == 0xC0)
5822 return (mask);
5823
5824 /* Otherwise return no mask */
5825 return ((ipaddr_t)0);
5826 }
5827
5828 /* Name/Value Table Lookup Routine */
5829 char *
5830 ip_nv_lookup(nv_t *nv, int value)
5831 {
5832 if (!nv)
5833 return (NULL);
5834 for (; nv->nv_name; nv++) {
5835 if (nv->nv_value == value)
5836 return (nv->nv_name);
5837 }
5838 return ("unknown");
5839 }
5840
5841 static int
5842 ip_wait_for_info_ack(ill_t *ill)
5843 {
5844 int err;
5845
5846 mutex_enter(&ill->ill_lock);
5847 while (ill->ill_state_flags & ILL_LL_SUBNET_PENDING) {
5848 /*
5849 * Return value of 0 indicates a pending signal.
5850 */
5851 err = cv_wait_sig(&ill->ill_cv, &ill->ill_lock);
5852 if (err == 0) {
5853 mutex_exit(&ill->ill_lock);
5854 return (EINTR);
5855 }
5856 }
5857 mutex_exit(&ill->ill_lock);
5858 /*
5859 * ip_rput_other could have set an error in ill_error on
5860 * receipt of M_ERROR.
5861 */
5862 return (ill->ill_error);
5863 }
5864
5865 /*
5866 * This is a module open, i.e. this is a control stream for access
5867 * to a DLPI device. We allocate an ill_t as the instance data in
5868 * this case.
5869 */
5870 static int
5871 ip_modopen(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5872 {
5873 ill_t *ill;
5874 int err;
5875 zoneid_t zoneid;
5876 netstack_t *ns;
5877 ip_stack_t *ipst;
5878
5879 /*
5880 * Prevent unprivileged processes from pushing IP so that
5881 * they can't send raw IP.
5882 */
5883 if (secpolicy_net_rawaccess(credp) != 0)
5884 return (EPERM);
5885
5886 ns = netstack_find_by_cred(credp);
5887 ASSERT(ns != NULL);
5888 ipst = ns->netstack_ip;
5889 ASSERT(ipst != NULL);
5890
5891 /*
5892 * For exclusive stacks we set the zoneid to zero
5893 * to make IP operate as if in the global zone.
5894 */
5895 if (ipst->ips_netstack->netstack_stackid != GLOBAL_NETSTACKID)
5896 zoneid = GLOBAL_ZONEID;
5897 else
5898 zoneid = crgetzoneid(credp);
5899
5900 ill = (ill_t *)mi_open_alloc_sleep(sizeof (ill_t));
5901 q->q_ptr = WR(q)->q_ptr = ill;
5902 ill->ill_ipst = ipst;
5903 ill->ill_zoneid = zoneid;
5904
5905 /*
5906 * ill_init initializes the ill fields and then sends down
5907 * down a DL_INFO_REQ after calling qprocson.
5908 */
5909 err = ill_init(q, ill);
5910
5911 if (err != 0) {
5912 mi_free(ill);
5913 netstack_rele(ipst->ips_netstack);
5914 q->q_ptr = NULL;
5915 WR(q)->q_ptr = NULL;
5916 return (err);
5917 }
5918
5919 /*
5920 * Wait for the DL_INFO_ACK if a DL_INFO_REQ was sent.
5921 *
5922 * ill_init initializes the ipsq marking this thread as
5923 * writer
5924 */
5925 ipsq_exit(ill->ill_phyint->phyint_ipsq);
5926 err = ip_wait_for_info_ack(ill);
5927 if (err == 0)
5928 ill->ill_credp = credp;
5929 else
5930 goto fail;
5931
5932 crhold(credp);
5933
5934 mutex_enter(&ipst->ips_ip_mi_lock);
5935 err = mi_open_link(&ipst->ips_ip_g_head, (IDP)q->q_ptr, devp, flag,
5936 sflag, credp);
5937 mutex_exit(&ipst->ips_ip_mi_lock);
5938 fail:
5939 if (err) {
5940 (void) ip_close(q, 0);
5941 return (err);
5942 }
5943 return (0);
5944 }
5945
5946 /* For /dev/ip aka AF_INET open */
5947 int
5948 ip_openv4(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5949 {
5950 return (ip_open(q, devp, flag, sflag, credp, B_FALSE));
5951 }
5952
5953 /* For /dev/ip6 aka AF_INET6 open */
5954 int
5955 ip_openv6(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
5956 {
5957 return (ip_open(q, devp, flag, sflag, credp, B_TRUE));
5958 }
5959
5960 /* IP open routine. */
5961 int
5962 ip_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp,
5963 boolean_t isv6)
5964 {
5965 conn_t *connp;
5966 major_t maj;
5967 zoneid_t zoneid;
5968 netstack_t *ns;
5969 ip_stack_t *ipst;
5970
5971 /* Allow reopen. */
5972 if (q->q_ptr != NULL)
5973 return (0);
5974
5975 if (sflag & MODOPEN) {
5976 /* This is a module open */
5977 return (ip_modopen(q, devp, flag, sflag, credp));
5978 }
5979
5980 if ((flag & ~(FKLYR)) == IP_HELPER_STR) {
5981 /*
5982 * Non streams based socket looking for a stream
5983 * to access IP
5984 */
5985 return (ip_helper_stream_setup(q, devp, flag, sflag,
5986 credp, isv6));
5987 }
5988
5989 ns = netstack_find_by_cred(credp);
5990 ASSERT(ns != NULL);
5991 ipst = ns->netstack_ip;
5992 ASSERT(ipst != NULL);
5993
5994 /*
5995 * For exclusive stacks we set the zoneid to zero
5996 * to make IP operate as if in the global zone.
5997 */
5998 if (ipst->ips_netstack->netstack_stackid != GLOBAL_NETSTACKID)
5999 zoneid = GLOBAL_ZONEID;
6000 else
6001 zoneid = crgetzoneid(credp);
6002
6003 /*
6004 * We are opening as a device. This is an IP client stream, and we
6005 * allocate an conn_t as the instance data.
6006 */
6007 connp = ipcl_conn_create(IPCL_IPCCONN, KM_SLEEP, ipst->ips_netstack);
6008
6009 /*
6010 * ipcl_conn_create did a netstack_hold. Undo the hold that was
6011 * done by netstack_find_by_cred()
6012 */
6013 netstack_rele(ipst->ips_netstack);
6014
6015 connp->conn_ixa->ixa_flags |= IXAF_MULTICAST_LOOP | IXAF_SET_ULP_CKSUM;
6016 /* conn_allzones can not be set this early, hence no IPCL_ZONEID */
6017 connp->conn_ixa->ixa_zoneid = zoneid;
6018 connp->conn_zoneid = zoneid;
6019
6020 connp->conn_rq = q;
6021 q->q_ptr = WR(q)->q_ptr = connp;
6022
6023 /* Minor tells us which /dev entry was opened */
6024 if (isv6) {
6025 connp->conn_family = AF_INET6;
6026 connp->conn_ipversion = IPV6_VERSION;
6027 connp->conn_ixa->ixa_flags &= ~IXAF_IS_IPV4;
6028 connp->conn_ixa->ixa_src_preferences = IPV6_PREFER_SRC_DEFAULT;
6029 } else {
6030 connp->conn_family = AF_INET;
6031 connp->conn_ipversion = IPV4_VERSION;
6032 connp->conn_ixa->ixa_flags |= IXAF_IS_IPV4;
6033 }
6034
6035 if ((ip_minor_arena_la != NULL) && (flag & SO_SOCKSTR) &&
6036 ((connp->conn_dev = inet_minor_alloc(ip_minor_arena_la)) != 0)) {
6037 connp->conn_minor_arena = ip_minor_arena_la;
6038 } else {
6039 /*
6040 * Either minor numbers in the large arena were exhausted
6041 * or a non socket application is doing the open.
6042 * Try to allocate from the small arena.
6043 */
6044 if ((connp->conn_dev =
6045 inet_minor_alloc(ip_minor_arena_sa)) == 0) {
6046 /* CONN_DEC_REF takes care of netstack_rele() */
6047 q->q_ptr = WR(q)->q_ptr = NULL;
6048 CONN_DEC_REF(connp);
6049 return (EBUSY);
6050 }
6051 connp->conn_minor_arena = ip_minor_arena_sa;
6052 }
6053
6054 maj = getemajor(*devp);
6055 *devp = makedevice(maj, (minor_t)connp->conn_dev);
6056
6057 /*
6058 * connp->conn_cred is crfree()ed in ipcl_conn_destroy()
6059 */
6060 connp->conn_cred = credp;
6061 connp->conn_cpid = curproc->p_pid;
6062 /* Cache things in ixa without an extra refhold */
6063 ASSERT(!(connp->conn_ixa->ixa_free_flags & IXA_FREE_CRED));
6064 connp->conn_ixa->ixa_cred = connp->conn_cred;
6065 connp->conn_ixa->ixa_cpid = connp->conn_cpid;
6066 if (is_system_labeled())
6067 connp->conn_ixa->ixa_tsl = crgetlabel(connp->conn_cred);
6068
6069 /*
6070 * Handle IP_IOC_RTS_REQUEST and other ioctls which use conn_recv
6071 */
6072 connp->conn_recv = ip_conn_input;
6073 connp->conn_recvicmp = ip_conn_input_icmp;
6074
6075 crhold(connp->conn_cred);
6076
6077 /*
6078 * If the caller has the process-wide flag set, then default to MAC
6079 * exempt mode. This allows read-down to unlabeled hosts.
6080 */
6081 if (getpflags(NET_MAC_AWARE, credp) != 0)
6082 connp->conn_mac_mode = CONN_MAC_AWARE;
6083
6084 connp->conn_zone_is_global = (crgetzoneid(credp) == GLOBAL_ZONEID);
6085
6086 connp->conn_rq = q;
6087 connp->conn_wq = WR(q);
6088
6089 /* Non-zero default values */
6090 connp->conn_ixa->ixa_flags |= IXAF_MULTICAST_LOOP;
6091
6092 /*
6093 * Make the conn globally visible to walkers
6094 */
6095 ASSERT(connp->conn_ref == 1);
6096 mutex_enter(&connp->conn_lock);
6097 connp->conn_state_flags &= ~CONN_INCIPIENT;
6098 mutex_exit(&connp->conn_lock);
6099
6100 qprocson(q);
6101
6102 return (0);
6103 }
6104
6105 /*
6106 * Set IPsec policy from an ipsec_req_t. If the req is not "zero" and valid,
6107 * all of them are copied to the conn_t. If the req is "zero", the policy is
6108 * zeroed out. A "zero" policy has zero ipsr_{ah,req,self_encap}_req
6109 * fields.
6110 * We keep only the latest setting of the policy and thus policy setting
6111 * is not incremental/cumulative.
6112 *
6113 * Requests to set policies with multiple alternative actions will
6114 * go through a different API.
6115 */
6116 int
6117 ipsec_set_req(cred_t *cr, conn_t *connp, ipsec_req_t *req)
6118 {
6119 uint_t ah_req = 0;
6120 uint_t esp_req = 0;
6121 uint_t se_req = 0;
6122 ipsec_act_t *actp = NULL;
6123 uint_t nact;
6124 ipsec_policy_head_t *ph;
6125 boolean_t is_pol_reset, is_pol_inserted = B_FALSE;
6126 int error = 0;
6127 netstack_t *ns = connp->conn_netstack;
6128 ip_stack_t *ipst = ns->netstack_ip;
6129 ipsec_stack_t *ipss = ns->netstack_ipsec;
6130
6131 #define REQ_MASK (IPSEC_PREF_REQUIRED|IPSEC_PREF_NEVER)
6132
6133 /*
6134 * The IP_SEC_OPT option does not allow variable length parameters,
6135 * hence a request cannot be NULL.
6136 */
6137 if (req == NULL)
6138 return (EINVAL);
6139
6140 ah_req = req->ipsr_ah_req;
6141 esp_req = req->ipsr_esp_req;
6142 se_req = req->ipsr_self_encap_req;
6143
6144 /* Don't allow setting self-encap without one or more of AH/ESP. */
6145 if (se_req != 0 && esp_req == 0 && ah_req == 0)
6146 return (EINVAL);
6147
6148 /*
6149 * Are we dealing with a request to reset the policy (i.e.
6150 * zero requests).
6151 */
6152 is_pol_reset = ((ah_req & REQ_MASK) == 0 &&
6153 (esp_req & REQ_MASK) == 0 &&
6154 (se_req & REQ_MASK) == 0);
6155
6156 if (!is_pol_reset) {
6157 /*
6158 * If we couldn't load IPsec, fail with "protocol
6159 * not supported".
6160 * IPsec may not have been loaded for a request with zero
6161 * policies, so we don't fail in this case.
6162 */
6163 mutex_enter(&ipss->ipsec_loader_lock);
6164 if (ipss->ipsec_loader_state != IPSEC_LOADER_SUCCEEDED) {
6165 mutex_exit(&ipss->ipsec_loader_lock);
6166 return (EPROTONOSUPPORT);
6167 }
6168 mutex_exit(&ipss->ipsec_loader_lock);
6169
6170 /*
6171 * Test for valid requests. Invalid algorithms
6172 * need to be tested by IPsec code because new
6173 * algorithms can be added dynamically.
6174 */
6175 if ((ah_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0 ||
6176 (esp_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0 ||
6177 (se_req & ~(REQ_MASK|IPSEC_PREF_UNIQUE)) != 0) {
6178 return (EINVAL);
6179 }
6180
6181 /*
6182 * Only privileged users can issue these
6183 * requests.
6184 */
6185 if (((ah_req & IPSEC_PREF_NEVER) ||
6186 (esp_req & IPSEC_PREF_NEVER) ||
6187 (se_req & IPSEC_PREF_NEVER)) &&
6188 secpolicy_ip_config(cr, B_FALSE) != 0) {
6189 return (EPERM);
6190 }
6191
6192 /*
6193 * The IPSEC_PREF_REQUIRED and IPSEC_PREF_NEVER
6194 * are mutually exclusive.
6195 */
6196 if (((ah_req & REQ_MASK) == REQ_MASK) ||
6197 ((esp_req & REQ_MASK) == REQ_MASK) ||
6198 ((se_req & REQ_MASK) == REQ_MASK)) {
6199 /* Both of them are set */
6200 return (EINVAL);
6201 }
6202 }
6203
6204 ASSERT(MUTEX_HELD(&connp->conn_lock));
6205
6206 /*
6207 * If we have already cached policies in conn_connect(), don't
6208 * let them change now. We cache policies for connections
6209 * whose src,dst [addr, port] is known.
6210 */
6211 if (connp->conn_policy_cached) {
6212 return (EINVAL);
6213 }
6214
6215 /*
6216 * We have a zero policies, reset the connection policy if already
6217 * set. This will cause the connection to inherit the
6218 * global policy, if any.
6219 */
6220 if (is_pol_reset) {
6221 if (connp->conn_policy != NULL) {
6222 IPPH_REFRELE(connp->conn_policy, ipst->ips_netstack);
6223 connp->conn_policy = NULL;
6224 }
6225 connp->conn_in_enforce_policy = B_FALSE;
6226 connp->conn_out_enforce_policy = B_FALSE;
6227 return (0);
6228 }
6229
6230 ph = connp->conn_policy = ipsec_polhead_split(connp->conn_policy,
6231 ipst->ips_netstack);
6232 if (ph == NULL)
6233 goto enomem;
6234
6235 ipsec_actvec_from_req(req, &actp, &nact, ipst->ips_netstack);
6236 if (actp == NULL)
6237 goto enomem;
6238
6239 /*
6240 * Always insert IPv4 policy entries, since they can also apply to
6241 * ipv6 sockets being used in ipv4-compat mode.
6242 */
6243 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V4,
6244 IPSEC_TYPE_INBOUND, ns))
6245 goto enomem;
6246 is_pol_inserted = B_TRUE;
6247 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V4,
6248 IPSEC_TYPE_OUTBOUND, ns))
6249 goto enomem;
6250
6251 /*
6252 * We're looking at a v6 socket, also insert the v6-specific
6253 * entries.
6254 */
6255 if (connp->conn_family == AF_INET6) {
6256 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V6,
6257 IPSEC_TYPE_INBOUND, ns))
6258 goto enomem;
6259 if (!ipsec_polhead_insert(ph, actp, nact, IPSEC_AF_V6,
6260 IPSEC_TYPE_OUTBOUND, ns))
6261 goto enomem;
6262 }
6263
6264 ipsec_actvec_free(actp, nact);
6265
6266 /*
6267 * If the requests need security, set enforce_policy.
6268 * If the requests are IPSEC_PREF_NEVER, one should
6269 * still set conn_out_enforce_policy so that ip_set_destination
6270 * marks the ip_xmit_attr_t appropriatly. This is needed so that
6271 * for connections that we don't cache policy in at connect time,
6272 * if global policy matches in ip_output_attach_policy, we
6273 * don't wrongly inherit global policy. Similarly, we need
6274 * to set conn_in_enforce_policy also so that we don't verify
6275 * policy wrongly.
6276 */
6277 if ((ah_req & REQ_MASK) != 0 ||
6278 (esp_req & REQ_MASK) != 0 ||
6279 (se_req & REQ_MASK) != 0) {
6280 connp->conn_in_enforce_policy = B_TRUE;
6281 connp->conn_out_enforce_policy = B_TRUE;
6282 }
6283
6284 return (error);
6285 #undef REQ_MASK
6286
6287 /*
6288 * Common memory-allocation-failure exit path.
6289 */
6290 enomem:
6291 if (actp != NULL)
6292 ipsec_actvec_free(actp, nact);
6293 if (is_pol_inserted)
6294 ipsec_polhead_flush(ph, ns);
6295 return (ENOMEM);
6296 }
6297
6298 /*
6299 * Set socket options for joining and leaving multicast groups.
6300 * Common to IPv4 and IPv6; inet6 indicates the type of socket.
6301 * The caller has already check that the option name is consistent with
6302 * the address family of the socket.
6303 */
6304 int
6305 ip_opt_set_multicast_group(conn_t *connp, t_scalar_t name,
6306 uchar_t *invalp, boolean_t inet6, boolean_t checkonly)
6307 {
6308 int *i1 = (int *)invalp;
6309 int error = 0;
6310 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
6311 struct ip_mreq *v4_mreqp;
6312 struct ipv6_mreq *v6_mreqp;
6313 struct group_req *greqp;
6314 ire_t *ire;
6315 boolean_t done = B_FALSE;
6316 ipaddr_t ifaddr;
6317 in6_addr_t v6group;
6318 uint_t ifindex;
6319 boolean_t mcast_opt = B_TRUE;
6320 mcast_record_t fmode;
6321 int (*optfn)(conn_t *, boolean_t, const in6_addr_t *,
6322 ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *);
6323
6324 switch (name) {
6325 case IP_ADD_MEMBERSHIP:
6326 case IPV6_JOIN_GROUP:
6327 mcast_opt = B_FALSE;
6328 /* FALLTHRU */
6329 case MCAST_JOIN_GROUP:
6330 fmode = MODE_IS_EXCLUDE;
6331 optfn = ip_opt_add_group;
6332 break;
6333
6334 case IP_DROP_MEMBERSHIP:
6335 case IPV6_LEAVE_GROUP:
6336 mcast_opt = B_FALSE;
6337 /* FALLTHRU */
6338 case MCAST_LEAVE_GROUP:
6339 fmode = MODE_IS_INCLUDE;
6340 optfn = ip_opt_delete_group;
6341 break;
6342 default:
6343 ASSERT(0);
6344 }
6345
6346 if (mcast_opt) {
6347 struct sockaddr_in *sin;
6348 struct sockaddr_in6 *sin6;
6349
6350 greqp = (struct group_req *)i1;
6351 if (greqp->gr_group.ss_family == AF_INET) {
6352 sin = (struct sockaddr_in *)&(greqp->gr_group);
6353 IN6_INADDR_TO_V4MAPPED(&sin->sin_addr, &v6group);
6354 } else {
6355 if (!inet6)
6356 return (EINVAL); /* Not on INET socket */
6357
6358 sin6 = (struct sockaddr_in6 *)&(greqp->gr_group);
6359 v6group = sin6->sin6_addr;
6360 }
6361 ifaddr = INADDR_ANY;
6362 ifindex = greqp->gr_interface;
6363 } else if (inet6) {
6364 v6_mreqp = (struct ipv6_mreq *)i1;
6365 v6group = v6_mreqp->ipv6mr_multiaddr;
6366 ifaddr = INADDR_ANY;
6367 ifindex = v6_mreqp->ipv6mr_interface;
6368 } else {
6369 v4_mreqp = (struct ip_mreq *)i1;
6370 IN6_INADDR_TO_V4MAPPED(&v4_mreqp->imr_multiaddr, &v6group);
6371 ifaddr = (ipaddr_t)v4_mreqp->imr_interface.s_addr;
6372 ifindex = 0;
6373 }
6374
6375 /*
6376 * In the multirouting case, we need to replicate
6377 * the request on all interfaces that will take part
6378 * in replication. We do so because multirouting is
6379 * reflective, thus we will probably receive multi-
6380 * casts on those interfaces.
6381 * The ip_multirt_apply_membership() succeeds if
6382 * the operation succeeds on at least one interface.
6383 */
6384 if (IN6_IS_ADDR_V4MAPPED(&v6group)) {
6385 ipaddr_t group;
6386
6387 IN6_V4MAPPED_TO_IPADDR(&v6group, group);
6388
6389 ire = ire_ftable_lookup_v4(group, IP_HOST_MASK, 0,
6390 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6391 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6392 } else {
6393 ire = ire_ftable_lookup_v6(&v6group, &ipv6_all_ones, 0,
6394 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6395 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6396 }
6397 if (ire != NULL) {
6398 if (ire->ire_flags & RTF_MULTIRT) {
6399 error = ip_multirt_apply_membership(optfn, ire, connp,
6400 checkonly, &v6group, fmode, &ipv6_all_zeros);
6401 done = B_TRUE;
6402 }
6403 ire_refrele(ire);
6404 }
6405
6406 if (!done) {
6407 error = optfn(connp, checkonly, &v6group, ifaddr, ifindex,
6408 fmode, &ipv6_all_zeros);
6409 }
6410 return (error);
6411 }
6412
6413 /*
6414 * Set socket options for joining and leaving multicast groups
6415 * for specific sources.
6416 * Common to IPv4 and IPv6; inet6 indicates the type of socket.
6417 * The caller has already check that the option name is consistent with
6418 * the address family of the socket.
6419 */
6420 int
6421 ip_opt_set_multicast_sources(conn_t *connp, t_scalar_t name,
6422 uchar_t *invalp, boolean_t inet6, boolean_t checkonly)
6423 {
6424 int *i1 = (int *)invalp;
6425 int error = 0;
6426 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
6427 struct ip_mreq_source *imreqp;
6428 struct group_source_req *gsreqp;
6429 in6_addr_t v6group, v6src;
6430 uint32_t ifindex;
6431 ipaddr_t ifaddr;
6432 boolean_t mcast_opt = B_TRUE;
6433 mcast_record_t fmode;
6434 ire_t *ire;
6435 boolean_t done = B_FALSE;
6436 int (*optfn)(conn_t *, boolean_t, const in6_addr_t *,
6437 ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *);
6438
6439 switch (name) {
6440 case IP_BLOCK_SOURCE:
6441 mcast_opt = B_FALSE;
6442 /* FALLTHRU */
6443 case MCAST_BLOCK_SOURCE:
6444 fmode = MODE_IS_EXCLUDE;
6445 optfn = ip_opt_add_group;
6446 break;
6447
6448 case IP_UNBLOCK_SOURCE:
6449 mcast_opt = B_FALSE;
6450 /* FALLTHRU */
6451 case MCAST_UNBLOCK_SOURCE:
6452 fmode = MODE_IS_EXCLUDE;
6453 optfn = ip_opt_delete_group;
6454 break;
6455
6456 case IP_ADD_SOURCE_MEMBERSHIP:
6457 mcast_opt = B_FALSE;
6458 /* FALLTHRU */
6459 case MCAST_JOIN_SOURCE_GROUP:
6460 fmode = MODE_IS_INCLUDE;
6461 optfn = ip_opt_add_group;
6462 break;
6463
6464 case IP_DROP_SOURCE_MEMBERSHIP:
6465 mcast_opt = B_FALSE;
6466 /* FALLTHRU */
6467 case MCAST_LEAVE_SOURCE_GROUP:
6468 fmode = MODE_IS_INCLUDE;
6469 optfn = ip_opt_delete_group;
6470 break;
6471 default:
6472 ASSERT(0);
6473 }
6474
6475 if (mcast_opt) {
6476 gsreqp = (struct group_source_req *)i1;
6477 ifindex = gsreqp->gsr_interface;
6478 if (gsreqp->gsr_group.ss_family == AF_INET) {
6479 struct sockaddr_in *s;
6480 s = (struct sockaddr_in *)&gsreqp->gsr_group;
6481 IN6_INADDR_TO_V4MAPPED(&s->sin_addr, &v6group);
6482 s = (struct sockaddr_in *)&gsreqp->gsr_source;
6483 IN6_INADDR_TO_V4MAPPED(&s->sin_addr, &v6src);
6484 } else {
6485 struct sockaddr_in6 *s6;
6486
6487 if (!inet6)
6488 return (EINVAL); /* Not on INET socket */
6489
6490 s6 = (struct sockaddr_in6 *)&gsreqp->gsr_group;
6491 v6group = s6->sin6_addr;
6492 s6 = (struct sockaddr_in6 *)&gsreqp->gsr_source;
6493 v6src = s6->sin6_addr;
6494 }
6495 ifaddr = INADDR_ANY;
6496 } else {
6497 imreqp = (struct ip_mreq_source *)i1;
6498 IN6_INADDR_TO_V4MAPPED(&imreqp->imr_multiaddr, &v6group);
6499 IN6_INADDR_TO_V4MAPPED(&imreqp->imr_sourceaddr, &v6src);
6500 ifaddr = (ipaddr_t)imreqp->imr_interface.s_addr;
6501 ifindex = 0;
6502 }
6503
6504 /*
6505 * Handle src being mapped INADDR_ANY by changing it to unspecified.
6506 */
6507 if (IN6_IS_ADDR_V4MAPPED_ANY(&v6src))
6508 v6src = ipv6_all_zeros;
6509
6510 /*
6511 * In the multirouting case, we need to replicate
6512 * the request as noted in the mcast cases above.
6513 */
6514 if (IN6_IS_ADDR_V4MAPPED(&v6group)) {
6515 ipaddr_t group;
6516
6517 IN6_V4MAPPED_TO_IPADDR(&v6group, group);
6518
6519 ire = ire_ftable_lookup_v4(group, IP_HOST_MASK, 0,
6520 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6521 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6522 } else {
6523 ire = ire_ftable_lookup_v6(&v6group, &ipv6_all_ones, 0,
6524 IRE_HOST | IRE_INTERFACE, NULL, ALL_ZONES, NULL,
6525 MATCH_IRE_MASK | MATCH_IRE_TYPE, 0, ipst, NULL);
6526 }
6527 if (ire != NULL) {
6528 if (ire->ire_flags & RTF_MULTIRT) {
6529 error = ip_multirt_apply_membership(optfn, ire, connp,
6530 checkonly, &v6group, fmode, &v6src);
6531 done = B_TRUE;
6532 }
6533 ire_refrele(ire);
6534 }
6535 if (!done) {
6536 error = optfn(connp, checkonly, &v6group, ifaddr, ifindex,
6537 fmode, &v6src);
6538 }
6539 return (error);
6540 }
6541
6542 /*
6543 * Given a destination address and a pointer to where to put the information
6544 * this routine fills in the mtuinfo.
6545 * The socket must be connected.
6546 * For sctp conn_faddr is the primary address.
6547 */
6548 int
6549 ip_fill_mtuinfo(conn_t *connp, ip_xmit_attr_t *ixa, struct ip6_mtuinfo *mtuinfo)
6550 {
6551 uint32_t pmtu = IP_MAXPACKET;
6552 uint_t scopeid;
6553
6554 if (IN6_IS_ADDR_UNSPECIFIED(&connp->conn_faddr_v6))
6555 return (-1);
6556
6557 /* In case we never sent or called ip_set_destination_v4/v6 */
6558 if (ixa->ixa_ire != NULL)
6559 pmtu = ip_get_pmtu(ixa);
6560
6561 if (ixa->ixa_flags & IXAF_SCOPEID_SET)
6562 scopeid = ixa->ixa_scopeid;
6563 else
6564 scopeid = 0;
6565
6566 bzero(mtuinfo, sizeof (*mtuinfo));
6567 mtuinfo->ip6m_addr.sin6_family = AF_INET6;
6568 mtuinfo->ip6m_addr.sin6_port = connp->conn_fport;
6569 mtuinfo->ip6m_addr.sin6_addr = connp->conn_faddr_v6;
6570 mtuinfo->ip6m_addr.sin6_scope_id = scopeid;
6571 mtuinfo->ip6m_mtu = pmtu;
6572
6573 return (sizeof (struct ip6_mtuinfo));
6574 }
6575
6576 /*
6577 * When the src multihoming is changed from weak to [strong, preferred]
6578 * ip_ire_rebind_walker is called to walk the list of all ire_t entries
6579 * and identify routes that were created by user-applications in the
6580 * unbound state (i.e., without RTA_IFP), and for which an ire_ill is not
6581 * currently defined. These routes are then 'rebound', i.e., their ire_ill
6582 * is selected by finding an interface route for the gateway.
6583 */
6584 /* ARGSUSED */
6585 void
6586 ip_ire_rebind_walker(ire_t *ire, void *notused)
6587 {
6588 if (!ire->ire_unbound || ire->ire_ill != NULL)
6589 return;
6590 ire_rebind(ire);
6591 ire_delete(ire);
6592 }
6593
6594 /*
6595 * When the src multihoming is changed from [strong, preferred] to weak,
6596 * ip_ire_unbind_walker is called to walk the list of all ire_t entries, and
6597 * set any entries that were created by user-applications in the unbound state
6598 * (i.e., without RTA_IFP) back to having a NULL ire_ill.
6599 */
6600 /* ARGSUSED */
6601 void
6602 ip_ire_unbind_walker(ire_t *ire, void *notused)
6603 {
6604 ire_t *new_ire;
6605
6606 if (!ire->ire_unbound || ire->ire_ill == NULL)
6607 return;
6608 if (ire->ire_ipversion == IPV6_VERSION) {
6609 new_ire = ire_create_v6(&ire->ire_addr_v6, &ire->ire_mask_v6,
6610 &ire->ire_gateway_addr_v6, ire->ire_type, NULL,
6611 ire->ire_zoneid, ire->ire_flags, NULL, ire->ire_ipst);
6612 } else {
6613 new_ire = ire_create((uchar_t *)&ire->ire_addr,
6614 (uchar_t *)&ire->ire_mask,
6615 (uchar_t *)&ire->ire_gateway_addr, ire->ire_type, NULL,
6616 ire->ire_zoneid, ire->ire_flags, NULL, ire->ire_ipst);
6617 }
6618 if (new_ire == NULL)
6619 return;
6620 new_ire->ire_unbound = B_TRUE;
6621 /*
6622 * The bound ire must first be deleted so that we don't return
6623 * the existing one on the attempt to add the unbound new_ire.
6624 */
6625 ire_delete(ire);
6626 new_ire = ire_add(new_ire);
6627 if (new_ire != NULL)
6628 ire_refrele(new_ire);
6629 }
6630
6631 /*
6632 * When the settings of ip*_strict_src_multihoming tunables are changed,
6633 * all cached routes need to be recomputed. This recomputation needs to be
6634 * done when going from weaker to stronger modes so that the cached ire
6635 * for the connection does not violate the current ip*_strict_src_multihoming
6636 * setting. It also needs to be done when going from stronger to weaker modes,
6637 * so that we fall back to matching on the longest-matching-route (as opposed
6638 * to a shorter match that may have been selected in the strong mode
6639 * to satisfy src_multihoming settings).
6640 *
6641 * The cached ixa_ire entires for all conn_t entries are marked as
6642 * "verify" so that they will be recomputed for the next packet.
6643 */
6644 void
6645 conn_ire_revalidate(conn_t *connp, void *arg)
6646 {
6647 boolean_t isv6 = (boolean_t)arg;
6648
6649 if ((isv6 && connp->conn_ipversion != IPV6_VERSION) ||
6650 (!isv6 && connp->conn_ipversion != IPV4_VERSION))
6651 return;
6652 connp->conn_ixa->ixa_ire_generation = IRE_GENERATION_VERIFY;
6653 }
6654
6655 /*
6656 * Handles both IPv4 and IPv6 reassembly - doing the out-of-order cases,
6657 * When an ipf is passed here for the first time, if
6658 * we already have in-order fragments on the queue, we convert from the fast-
6659 * path reassembly scheme to the hard-case scheme. From then on, additional
6660 * fragments are reassembled here. We keep track of the start and end offsets
6661 * of each piece, and the number of holes in the chain. When the hole count
6662 * goes to zero, we are done!
6663 *
6664 * The ipf_count will be updated to account for any mblk(s) added (pointed to
6665 * by mp) or subtracted (freeb()ed dups), upon return the caller must update
6666 * ipfb_count and ill_frag_count by the difference of ipf_count before and
6667 * after the call to ip_reassemble().
6668 */
6669 int
6670 ip_reassemble(mblk_t *mp, ipf_t *ipf, uint_t start, boolean_t more, ill_t *ill,
6671 size_t msg_len)
6672 {
6673 uint_t end;
6674 mblk_t *next_mp;
6675 mblk_t *mp1;
6676 uint_t offset;
6677 boolean_t incr_dups = B_TRUE;
6678 boolean_t offset_zero_seen = B_FALSE;
6679 boolean_t pkt_boundary_checked = B_FALSE;
6680
6681 /* If start == 0 then ipf_nf_hdr_len has to be set. */
6682 ASSERT(start != 0 || ipf->ipf_nf_hdr_len != 0);
6683
6684 /* Add in byte count */
6685 ipf->ipf_count += msg_len;
6686 if (ipf->ipf_end) {
6687 /*
6688 * We were part way through in-order reassembly, but now there
6689 * is a hole. We walk through messages already queued, and
6690 * mark them for hard case reassembly. We know that up till
6691 * now they were in order starting from offset zero.
6692 */
6693 offset = 0;
6694 for (mp1 = ipf->ipf_mp->b_cont; mp1; mp1 = mp1->b_cont) {
6695 IP_REASS_SET_START(mp1, offset);
6696 if (offset == 0) {
6697 ASSERT(ipf->ipf_nf_hdr_len != 0);
6698 offset = -ipf->ipf_nf_hdr_len;
6699 }
6700 offset += mp1->b_wptr - mp1->b_rptr;
6701 IP_REASS_SET_END(mp1, offset);
6702 }
6703 /* One hole at the end. */
6704 ipf->ipf_hole_cnt = 1;
6705 /* Brand it as a hard case, forever. */
6706 ipf->ipf_end = 0;
6707 }
6708 /* Walk through all the new pieces. */
6709 do {
6710 end = start + (mp->b_wptr - mp->b_rptr);
6711 /*
6712 * If start is 0, decrease 'end' only for the first mblk of
6713 * the fragment. Otherwise 'end' can get wrong value in the
6714 * second pass of the loop if first mblk is exactly the
6715 * size of ipf_nf_hdr_len.
6716 */
6717 if (start == 0 && !offset_zero_seen) {
6718 /* First segment */
6719 ASSERT(ipf->ipf_nf_hdr_len != 0);
6720 end -= ipf->ipf_nf_hdr_len;
6721 offset_zero_seen = B_TRUE;
6722 }
6723 next_mp = mp->b_cont;
6724 /*
6725 * We are checking to see if there is any interesing data
6726 * to process. If there isn't and the mblk isn't the
6727 * one which carries the unfragmentable header then we
6728 * drop it. It's possible to have just the unfragmentable
6729 * header come through without any data. That needs to be
6730 * saved.
6731 *
6732 * If the assert at the top of this function holds then the
6733 * term "ipf->ipf_nf_hdr_len != 0" isn't needed. This code
6734 * is infrequently traveled enough that the test is left in
6735 * to protect against future code changes which break that
6736 * invariant.
6737 */
6738 if (start == end && start != 0 && ipf->ipf_nf_hdr_len != 0) {
6739 /* Empty. Blast it. */
6740 IP_REASS_SET_START(mp, 0);
6741 IP_REASS_SET_END(mp, 0);
6742 /*
6743 * If the ipf points to the mblk we are about to free,
6744 * update ipf to point to the next mblk (or NULL
6745 * if none).
6746 */
6747 if (ipf->ipf_mp->b_cont == mp)
6748 ipf->ipf_mp->b_cont = next_mp;
6749 freeb(mp);
6750 continue;
6751 }
6752 mp->b_cont = NULL;
6753 IP_REASS_SET_START(mp, start);
6754 IP_REASS_SET_END(mp, end);
6755 if (!ipf->ipf_tail_mp) {
6756 ipf->ipf_tail_mp = mp;
6757 ipf->ipf_mp->b_cont = mp;
6758 if (start == 0 || !more) {
6759 ipf->ipf_hole_cnt = 1;
6760 /*
6761 * if the first fragment comes in more than one
6762 * mblk, this loop will be executed for each
6763 * mblk. Need to adjust hole count so exiting
6764 * this routine will leave hole count at 1.
6765 */
6766 if (next_mp)
6767 ipf->ipf_hole_cnt++;
6768 } else
6769 ipf->ipf_hole_cnt = 2;
6770 continue;
6771 } else if (ipf->ipf_last_frag_seen && !more &&
6772 !pkt_boundary_checked) {
6773 /*
6774 * We check datagram boundary only if this fragment
6775 * claims to be the last fragment and we have seen a
6776 * last fragment in the past too. We do this only
6777 * once for a given fragment.
6778 *
6779 * start cannot be 0 here as fragments with start=0
6780 * and MF=0 gets handled as a complete packet. These
6781 * fragments should not reach here.
6782 */
6783
6784 if (start + msgdsize(mp) !=
6785 IP_REASS_END(ipf->ipf_tail_mp)) {
6786 /*
6787 * We have two fragments both of which claim
6788 * to be the last fragment but gives conflicting
6789 * information about the whole datagram size.
6790 * Something fishy is going on. Drop the
6791 * fragment and free up the reassembly list.
6792 */
6793 return (IP_REASS_FAILED);
6794 }
6795
6796 /*
6797 * We shouldn't come to this code block again for this
6798 * particular fragment.
6799 */
6800 pkt_boundary_checked = B_TRUE;
6801 }
6802
6803 /* New stuff at or beyond tail? */
6804 offset = IP_REASS_END(ipf->ipf_tail_mp);
6805 if (start >= offset) {
6806 if (ipf->ipf_last_frag_seen) {
6807 /* current fragment is beyond last fragment */
6808 return (IP_REASS_FAILED);
6809 }
6810 /* Link it on end. */
6811 ipf->ipf_tail_mp->b_cont = mp;
6812 ipf->ipf_tail_mp = mp;
6813 if (more) {
6814 if (start != offset)
6815 ipf->ipf_hole_cnt++;
6816 } else if (start == offset && next_mp == NULL)
6817 ipf->ipf_hole_cnt--;
6818 continue;
6819 }
6820 mp1 = ipf->ipf_mp->b_cont;
6821 offset = IP_REASS_START(mp1);
6822 /* New stuff at the front? */
6823 if (start < offset) {
6824 if (start == 0) {
6825 if (end >= offset) {
6826 /* Nailed the hole at the begining. */
6827 ipf->ipf_hole_cnt--;
6828 }
6829 } else if (end < offset) {
6830 /*
6831 * A hole, stuff, and a hole where there used
6832 * to be just a hole.
6833 */
6834 ipf->ipf_hole_cnt++;
6835 }
6836 mp->b_cont = mp1;
6837 /* Check for overlap. */
6838 while (end > offset) {
6839 if (end < IP_REASS_END(mp1)) {
6840 mp->b_wptr -= end - offset;
6841 IP_REASS_SET_END(mp, offset);
6842 BUMP_MIB(ill->ill_ip_mib,
6843 ipIfStatsReasmPartDups);
6844 break;
6845 }
6846 /* Did we cover another hole? */
6847 if ((mp1->b_cont &&
6848 IP_REASS_END(mp1) !=
6849 IP_REASS_START(mp1->b_cont) &&
6850 end >= IP_REASS_START(mp1->b_cont)) ||
6851 (!ipf->ipf_last_frag_seen && !more)) {
6852 ipf->ipf_hole_cnt--;
6853 }
6854 /* Clip out mp1. */
6855 if ((mp->b_cont = mp1->b_cont) == NULL) {
6856 /*
6857 * After clipping out mp1, this guy
6858 * is now hanging off the end.
6859 */
6860 ipf->ipf_tail_mp = mp;
6861 }
6862 IP_REASS_SET_START(mp1, 0);
6863 IP_REASS_SET_END(mp1, 0);
6864 /* Subtract byte count */
6865 ipf->ipf_count -= mp1->b_datap->db_lim -
6866 mp1->b_datap->db_base;
6867 freeb(mp1);
6868 BUMP_MIB(ill->ill_ip_mib,
6869 ipIfStatsReasmPartDups);
6870 mp1 = mp->b_cont;
6871 if (!mp1)
6872 break;
6873 offset = IP_REASS_START(mp1);
6874 }
6875 ipf->ipf_mp->b_cont = mp;
6876 continue;
6877 }
6878 /*
6879 * The new piece starts somewhere between the start of the head
6880 * and before the end of the tail.
6881 */
6882 for (; mp1; mp1 = mp1->b_cont) {
6883 offset = IP_REASS_END(mp1);
6884 if (start < offset) {
6885 if (end <= offset) {
6886 /* Nothing new. */
6887 IP_REASS_SET_START(mp, 0);
6888 IP_REASS_SET_END(mp, 0);
6889 /* Subtract byte count */
6890 ipf->ipf_count -= mp->b_datap->db_lim -
6891 mp->b_datap->db_base;
6892 if (incr_dups) {
6893 ipf->ipf_num_dups++;
6894 incr_dups = B_FALSE;
6895 }
6896 freeb(mp);
6897 BUMP_MIB(ill->ill_ip_mib,
6898 ipIfStatsReasmDuplicates);
6899 break;
6900 }
6901 /*
6902 * Trim redundant stuff off beginning of new
6903 * piece.
6904 */
6905 IP_REASS_SET_START(mp, offset);
6906 mp->b_rptr += offset - start;
6907 BUMP_MIB(ill->ill_ip_mib,
6908 ipIfStatsReasmPartDups);
6909 start = offset;
6910 if (!mp1->b_cont) {
6911 /*
6912 * After trimming, this guy is now
6913 * hanging off the end.
6914 */
6915 mp1->b_cont = mp;
6916 ipf->ipf_tail_mp = mp;
6917 if (!more) {
6918 ipf->ipf_hole_cnt--;
6919 }
6920 break;
6921 }
6922 }
6923 if (start >= IP_REASS_START(mp1->b_cont))
6924 continue;
6925 /* Fill a hole */
6926 if (start > offset)
6927 ipf->ipf_hole_cnt++;
6928 mp->b_cont = mp1->b_cont;
6929 mp1->b_cont = mp;
6930 mp1 = mp->b_cont;
6931 offset = IP_REASS_START(mp1);
6932 if (end >= offset) {
6933 ipf->ipf_hole_cnt--;
6934 /* Check for overlap. */
6935 while (end > offset) {
6936 if (end < IP_REASS_END(mp1)) {
6937 mp->b_wptr -= end - offset;
6938 IP_REASS_SET_END(mp, offset);
6939 /*
6940 * TODO we might bump
6941 * this up twice if there is
6942 * overlap at both ends.
6943 */
6944 BUMP_MIB(ill->ill_ip_mib,
6945 ipIfStatsReasmPartDups);
6946 break;
6947 }
6948 /* Did we cover another hole? */
6949 if ((mp1->b_cont &&
6950 IP_REASS_END(mp1)
6951 != IP_REASS_START(mp1->b_cont) &&
6952 end >=
6953 IP_REASS_START(mp1->b_cont)) ||
6954 (!ipf->ipf_last_frag_seen &&
6955 !more)) {
6956 ipf->ipf_hole_cnt--;
6957 }
6958 /* Clip out mp1. */
6959 if ((mp->b_cont = mp1->b_cont) ==
6960 NULL) {
6961 /*
6962 * After clipping out mp1,
6963 * this guy is now hanging
6964 * off the end.
6965 */
6966 ipf->ipf_tail_mp = mp;
6967 }
6968 IP_REASS_SET_START(mp1, 0);
6969 IP_REASS_SET_END(mp1, 0);
6970 /* Subtract byte count */
6971 ipf->ipf_count -=
6972 mp1->b_datap->db_lim -
6973 mp1->b_datap->db_base;
6974 freeb(mp1);
6975 BUMP_MIB(ill->ill_ip_mib,
6976 ipIfStatsReasmPartDups);
6977 mp1 = mp->b_cont;
6978 if (!mp1)
6979 break;
6980 offset = IP_REASS_START(mp1);
6981 }
6982 }
6983 break;
6984 }
6985 } while (start = end, mp = next_mp);
6986
6987 /* Fragment just processed could be the last one. Remember this fact */
6988 if (!more)
6989 ipf->ipf_last_frag_seen = B_TRUE;
6990
6991 /* Still got holes? */
6992 if (ipf->ipf_hole_cnt)
6993 return (IP_REASS_PARTIAL);
6994 /* Clean up overloaded fields to avoid upstream disasters. */
6995 for (mp1 = ipf->ipf_mp->b_cont; mp1; mp1 = mp1->b_cont) {
6996 IP_REASS_SET_START(mp1, 0);
6997 IP_REASS_SET_END(mp1, 0);
6998 }
6999 return (IP_REASS_COMPLETE);
7000 }
7001
7002 /*
7003 * Fragmentation reassembly. Each ILL has a hash table for
7004 * queuing packets undergoing reassembly for all IPIFs
7005 * associated with the ILL. The hash is based on the packet
7006 * IP ident field. The ILL frag hash table was allocated
7007 * as a timer block at the time the ILL was created. Whenever
7008 * there is anything on the reassembly queue, the timer will
7009 * be running. Returns the reassembled packet if reassembly completes.
7010 */
7011 mblk_t *
7012 ip_input_fragment(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
7013 {
7014 uint32_t frag_offset_flags;
7015 mblk_t *t_mp;
7016 ipaddr_t dst;
7017 uint8_t proto = ipha->ipha_protocol;
7018 uint32_t sum_val;
7019 uint16_t sum_flags;
7020 ipf_t *ipf;
7021 ipf_t **ipfp;
7022 ipfb_t *ipfb;
7023 uint16_t ident;
7024 uint32_t offset;
7025 ipaddr_t src;
7026 uint_t hdr_length;
7027 uint32_t end;
7028 mblk_t *mp1;
7029 mblk_t *tail_mp;
7030 size_t count;
7031 size_t msg_len;
7032 uint8_t ecn_info = 0;
7033 uint32_t packet_size;
7034 boolean_t pruned = B_FALSE;
7035 ill_t *ill = ira->ira_ill;
7036 ip_stack_t *ipst = ill->ill_ipst;
7037
7038 /*
7039 * Drop the fragmented as early as possible, if
7040 * we don't have resource(s) to re-assemble.
7041 */
7042 if (ipst->ips_ip_reass_queue_bytes == 0) {
7043 freemsg(mp);
7044 return (NULL);
7045 }
7046
7047 /* Check for fragmentation offset; return if there's none */
7048 if ((frag_offset_flags = ntohs(ipha->ipha_fragment_offset_and_flags) &
7049 (IPH_MF | IPH_OFFSET)) == 0)
7050 return (mp);
7051
7052 /*
7053 * We utilize hardware computed checksum info only for UDP since
7054 * IP fragmentation is a normal occurrence for the protocol. In
7055 * addition, checksum offload support for IP fragments carrying
7056 * UDP payload is commonly implemented across network adapters.
7057 */
7058 ASSERT(ira->ira_rill != NULL);
7059 if (proto == IPPROTO_UDP && dohwcksum &&
7060 ILL_HCKSUM_CAPABLE(ira->ira_rill) &&
7061 (DB_CKSUMFLAGS(mp) & (HCK_FULLCKSUM | HCK_PARTIALCKSUM))) {
7062 mblk_t *mp1 = mp->b_cont;
7063 int32_t len;
7064
7065 /* Record checksum information from the packet */
7066 sum_val = (uint32_t)DB_CKSUM16(mp);
7067 sum_flags = DB_CKSUMFLAGS(mp);
7068
7069 /* IP payload offset from beginning of mblk */
7070 offset = ((uchar_t *)ipha + IPH_HDR_LENGTH(ipha)) - mp->b_rptr;
7071
7072 if ((sum_flags & HCK_PARTIALCKSUM) &&
7073 (mp1 == NULL || mp1->b_cont == NULL) &&
7074 offset >= DB_CKSUMSTART(mp) &&
7075 ((len = offset - DB_CKSUMSTART(mp)) & 1) == 0) {
7076 uint32_t adj;
7077 /*
7078 * Partial checksum has been calculated by hardware
7079 * and attached to the packet; in addition, any
7080 * prepended extraneous data is even byte aligned.
7081 * If any such data exists, we adjust the checksum;
7082 * this would also handle any postpended data.
7083 */
7084 IP_ADJCKSUM_PARTIAL(mp->b_rptr + DB_CKSUMSTART(mp),
7085 mp, mp1, len, adj);
7086
7087 /* One's complement subtract extraneous checksum */
7088 if (adj >= sum_val)
7089 sum_val = ~(adj - sum_val) & 0xFFFF;
7090 else
7091 sum_val -= adj;
7092 }
7093 } else {
7094 sum_val = 0;
7095 sum_flags = 0;
7096 }
7097
7098 /* Clear hardware checksumming flag */
7099 DB_CKSUMFLAGS(mp) = 0;
7100
7101 ident = ipha->ipha_ident;
7102 offset = (frag_offset_flags << 3) & 0xFFFF;
7103 src = ipha->ipha_src;
7104 dst = ipha->ipha_dst;
7105 hdr_length = IPH_HDR_LENGTH(ipha);
7106 end = ntohs(ipha->ipha_length) - hdr_length;
7107
7108 /* If end == 0 then we have a packet with no data, so just free it */
7109 if (end == 0) {
7110 freemsg(mp);
7111 return (NULL);
7112 }
7113
7114 /* Record the ECN field info. */
7115 ecn_info = (ipha->ipha_type_of_service & 0x3);
7116 if (offset != 0) {
7117 /*
7118 * If this isn't the first piece, strip the header, and
7119 * add the offset to the end value.
7120 */
7121 mp->b_rptr += hdr_length;
7122 end += offset;
7123 }
7124
7125 /* Handle vnic loopback of fragments */
7126 if (mp->b_datap->db_ref > 2)
7127 msg_len = 0;
7128 else
7129 msg_len = MBLKSIZE(mp);
7130
7131 tail_mp = mp;
7132 while (tail_mp->b_cont != NULL) {
7133 tail_mp = tail_mp->b_cont;
7134 if (tail_mp->b_datap->db_ref <= 2)
7135 msg_len += MBLKSIZE(tail_mp);
7136 }
7137
7138 /* If the reassembly list for this ILL will get too big, prune it */
7139 if ((msg_len + sizeof (*ipf) + ill->ill_frag_count) >=
7140 ipst->ips_ip_reass_queue_bytes) {
7141 DTRACE_PROBE3(ip_reass_queue_bytes, uint_t, msg_len,
7142 uint_t, ill->ill_frag_count,
7143 uint_t, ipst->ips_ip_reass_queue_bytes);
7144 ill_frag_prune(ill,
7145 (ipst->ips_ip_reass_queue_bytes < msg_len) ? 0 :
7146 (ipst->ips_ip_reass_queue_bytes - msg_len));
7147 pruned = B_TRUE;
7148 }
7149
7150 ipfb = &ill->ill_frag_hash_tbl[ILL_FRAG_HASH(src, ident)];
7151 mutex_enter(&ipfb->ipfb_lock);
7152
7153 ipfp = &ipfb->ipfb_ipf;
7154 /* Try to find an existing fragment queue for this packet. */
7155 for (;;) {
7156 ipf = ipfp[0];
7157 if (ipf != NULL) {
7158 /*
7159 * It has to match on ident and src/dst address.
7160 */
7161 if (ipf->ipf_ident == ident &&
7162 ipf->ipf_src == src &&
7163 ipf->ipf_dst == dst &&
7164 ipf->ipf_protocol == proto) {
7165 /*
7166 * If we have received too many
7167 * duplicate fragments for this packet
7168 * free it.
7169 */
7170 if (ipf->ipf_num_dups > ip_max_frag_dups) {
7171 ill_frag_free_pkts(ill, ipfb, ipf, 1);
7172 freemsg(mp);
7173 mutex_exit(&ipfb->ipfb_lock);
7174 return (NULL);
7175 }
7176 /* Found it. */
7177 break;
7178 }
7179 ipfp = &ipf->ipf_hash_next;
7180 continue;
7181 }
7182
7183 /*
7184 * If we pruned the list, do we want to store this new
7185 * fragment?. We apply an optimization here based on the
7186 * fact that most fragments will be received in order.
7187 * So if the offset of this incoming fragment is zero,
7188 * it is the first fragment of a new packet. We will
7189 * keep it. Otherwise drop the fragment, as we have
7190 * probably pruned the packet already (since the
7191 * packet cannot be found).
7192 */
7193 if (pruned && offset != 0) {
7194 mutex_exit(&ipfb->ipfb_lock);
7195 freemsg(mp);
7196 return (NULL);
7197 }
7198
7199 if (ipfb->ipfb_frag_pkts >= MAX_FRAG_PKTS(ipst)) {
7200 /*
7201 * Too many fragmented packets in this hash
7202 * bucket. Free the oldest.
7203 */
7204 ill_frag_free_pkts(ill, ipfb, ipfb->ipfb_ipf, 1);
7205 }
7206
7207 /* New guy. Allocate a frag message. */
7208 mp1 = allocb(sizeof (*ipf), BPRI_MED);
7209 if (mp1 == NULL) {
7210 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7211 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7212 freemsg(mp);
7213 reass_done:
7214 mutex_exit(&ipfb->ipfb_lock);
7215 return (NULL);
7216 }
7217
7218 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmReqds);
7219 mp1->b_cont = mp;
7220
7221 /* Initialize the fragment header. */
7222 ipf = (ipf_t *)mp1->b_rptr;
7223 ipf->ipf_mp = mp1;
7224 ipf->ipf_ptphn = ipfp;
7225 ipfp[0] = ipf;
7226 ipf->ipf_hash_next = NULL;
7227 ipf->ipf_ident = ident;
7228 ipf->ipf_protocol = proto;
7229 ipf->ipf_src = src;
7230 ipf->ipf_dst = dst;
7231 ipf->ipf_nf_hdr_len = 0;
7232 /* Record reassembly start time. */
7233 ipf->ipf_timestamp = gethrestime_sec();
7234 /* Record ipf generation and account for frag header */
7235 ipf->ipf_gen = ill->ill_ipf_gen++;
7236 ipf->ipf_count = MBLKSIZE(mp1);
7237 ipf->ipf_last_frag_seen = B_FALSE;
7238 ipf->ipf_ecn = ecn_info;
7239 ipf->ipf_num_dups = 0;
7240 ipfb->ipfb_frag_pkts++;
7241 ipf->ipf_checksum = 0;
7242 ipf->ipf_checksum_flags = 0;
7243
7244 /* Store checksum value in fragment header */
7245 if (sum_flags != 0) {
7246 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7247 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7248 ipf->ipf_checksum = sum_val;
7249 ipf->ipf_checksum_flags = sum_flags;
7250 }
7251
7252 /*
7253 * We handle reassembly two ways. In the easy case,
7254 * where all the fragments show up in order, we do
7255 * minimal bookkeeping, and just clip new pieces on
7256 * the end. If we ever see a hole, then we go off
7257 * to ip_reassemble which has to mark the pieces and
7258 * keep track of the number of holes, etc. Obviously,
7259 * the point of having both mechanisms is so we can
7260 * handle the easy case as efficiently as possible.
7261 */
7262 if (offset == 0) {
7263 /* Easy case, in-order reassembly so far. */
7264 ipf->ipf_count += msg_len;
7265 ipf->ipf_tail_mp = tail_mp;
7266 /*
7267 * Keep track of next expected offset in
7268 * ipf_end.
7269 */
7270 ipf->ipf_end = end;
7271 ipf->ipf_nf_hdr_len = hdr_length;
7272 } else {
7273 /* Hard case, hole at the beginning. */
7274 ipf->ipf_tail_mp = NULL;
7275 /*
7276 * ipf_end == 0 means that we have given up
7277 * on easy reassembly.
7278 */
7279 ipf->ipf_end = 0;
7280
7281 /* Forget checksum offload from now on */
7282 ipf->ipf_checksum_flags = 0;
7283
7284 /*
7285 * ipf_hole_cnt is set by ip_reassemble.
7286 * ipf_count is updated by ip_reassemble.
7287 * No need to check for return value here
7288 * as we don't expect reassembly to complete
7289 * or fail for the first fragment itself.
7290 */
7291 (void) ip_reassemble(mp, ipf,
7292 (frag_offset_flags & IPH_OFFSET) << 3,
7293 (frag_offset_flags & IPH_MF), ill, msg_len);
7294 }
7295 /* Update per ipfb and ill byte counts */
7296 ipfb->ipfb_count += ipf->ipf_count;
7297 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7298 atomic_add_32(&ill->ill_frag_count, ipf->ipf_count);
7299 /* If the frag timer wasn't already going, start it. */
7300 mutex_enter(&ill->ill_lock);
7301 ill_frag_timer_start(ill);
7302 mutex_exit(&ill->ill_lock);
7303 goto reass_done;
7304 }
7305
7306 /*
7307 * If the packet's flag has changed (it could be coming up
7308 * from an interface different than the previous, therefore
7309 * possibly different checksum capability), then forget about
7310 * any stored checksum states. Otherwise add the value to
7311 * the existing one stored in the fragment header.
7312 */
7313 if (sum_flags != 0 && sum_flags == ipf->ipf_checksum_flags) {
7314 sum_val += ipf->ipf_checksum;
7315 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7316 sum_val = (sum_val & 0xFFFF) + (sum_val >> 16);
7317 ipf->ipf_checksum = sum_val;
7318 } else if (ipf->ipf_checksum_flags != 0) {
7319 /* Forget checksum offload from now on */
7320 ipf->ipf_checksum_flags = 0;
7321 }
7322
7323 /*
7324 * We have a new piece of a datagram which is already being
7325 * reassembled. Update the ECN info if all IP fragments
7326 * are ECN capable. If there is one which is not, clear
7327 * all the info. If there is at least one which has CE
7328 * code point, IP needs to report that up to transport.
7329 */
7330 if (ecn_info != IPH_ECN_NECT && ipf->ipf_ecn != IPH_ECN_NECT) {
7331 if (ecn_info == IPH_ECN_CE)
7332 ipf->ipf_ecn = IPH_ECN_CE;
7333 } else {
7334 ipf->ipf_ecn = IPH_ECN_NECT;
7335 }
7336 if (offset && ipf->ipf_end == offset) {
7337 /* The new fragment fits at the end */
7338 ipf->ipf_tail_mp->b_cont = mp;
7339 /* Update the byte count */
7340 ipf->ipf_count += msg_len;
7341 /* Update per ipfb and ill byte counts */
7342 ipfb->ipfb_count += msg_len;
7343 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7344 atomic_add_32(&ill->ill_frag_count, msg_len);
7345 if (frag_offset_flags & IPH_MF) {
7346 /* More to come. */
7347 ipf->ipf_end = end;
7348 ipf->ipf_tail_mp = tail_mp;
7349 goto reass_done;
7350 }
7351 } else {
7352 /* Go do the hard cases. */
7353 int ret;
7354
7355 if (offset == 0)
7356 ipf->ipf_nf_hdr_len = hdr_length;
7357
7358 /* Save current byte count */
7359 count = ipf->ipf_count;
7360 ret = ip_reassemble(mp, ipf,
7361 (frag_offset_flags & IPH_OFFSET) << 3,
7362 (frag_offset_flags & IPH_MF), ill, msg_len);
7363 /* Count of bytes added and subtracted (freeb()ed) */
7364 count = ipf->ipf_count - count;
7365 if (count) {
7366 /* Update per ipfb and ill byte counts */
7367 ipfb->ipfb_count += count;
7368 ASSERT(ipfb->ipfb_count > 0); /* Wraparound */
7369 atomic_add_32(&ill->ill_frag_count, count);
7370 }
7371 if (ret == IP_REASS_PARTIAL) {
7372 goto reass_done;
7373 } else if (ret == IP_REASS_FAILED) {
7374 /* Reassembly failed. Free up all resources */
7375 ill_frag_free_pkts(ill, ipfb, ipf, 1);
7376 for (t_mp = mp; t_mp != NULL; t_mp = t_mp->b_cont) {
7377 IP_REASS_SET_START(t_mp, 0);
7378 IP_REASS_SET_END(t_mp, 0);
7379 }
7380 freemsg(mp);
7381 goto reass_done;
7382 }
7383 /* We will reach here iff 'ret' is IP_REASS_COMPLETE */
7384 }
7385 /*
7386 * We have completed reassembly. Unhook the frag header from
7387 * the reassembly list.
7388 *
7389 * Before we free the frag header, record the ECN info
7390 * to report back to the transport.
7391 */
7392 ecn_info = ipf->ipf_ecn;
7393 BUMP_MIB(ill->ill_ip_mib, ipIfStatsReasmOKs);
7394 ipfp = ipf->ipf_ptphn;
7395
7396 /* We need to supply these to caller */
7397 if ((sum_flags = ipf->ipf_checksum_flags) != 0)
7398 sum_val = ipf->ipf_checksum;
7399 else
7400 sum_val = 0;
7401
7402 mp1 = ipf->ipf_mp;
7403 count = ipf->ipf_count;
7404 ipf = ipf->ipf_hash_next;
7405 if (ipf != NULL)
7406 ipf->ipf_ptphn = ipfp;
7407 ipfp[0] = ipf;
7408 atomic_add_32(&ill->ill_frag_count, -count);
7409 ASSERT(ipfb->ipfb_count >= count);
7410 ipfb->ipfb_count -= count;
7411 ipfb->ipfb_frag_pkts--;
7412 mutex_exit(&ipfb->ipfb_lock);
7413 /* Ditch the frag header. */
7414 mp = mp1->b_cont;
7415
7416 freeb(mp1);
7417
7418 /* Restore original IP length in header. */
7419 packet_size = (uint32_t)msgdsize(mp);
7420 if (packet_size > IP_MAXPACKET) {
7421 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7422 ip_drop_input("Reassembled packet too large", mp, ill);
7423 freemsg(mp);
7424 return (NULL);
7425 }
7426
7427 if (DB_REF(mp) > 1) {
7428 mblk_t *mp2 = copymsg(mp);
7429
7430 if (mp2 == NULL) {
7431 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7432 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7433 freemsg(mp);
7434 return (NULL);
7435 }
7436 freemsg(mp);
7437 mp = mp2;
7438 }
7439 ipha = (ipha_t *)mp->b_rptr;
7440
7441 ipha->ipha_length = htons((uint16_t)packet_size);
7442 /* We're now complete, zip the frag state */
7443 ipha->ipha_fragment_offset_and_flags = 0;
7444 /* Record the ECN info. */
7445 ipha->ipha_type_of_service &= 0xFC;
7446 ipha->ipha_type_of_service |= ecn_info;
7447
7448 /* Update the receive attributes */
7449 ira->ira_pktlen = packet_size;
7450 ira->ira_ip_hdr_length = IPH_HDR_LENGTH(ipha);
7451
7452 /* Reassembly is successful; set checksum information in packet */
7453 DB_CKSUM16(mp) = (uint16_t)sum_val;
7454 DB_CKSUMFLAGS(mp) = sum_flags;
7455 DB_CKSUMSTART(mp) = ira->ira_ip_hdr_length;
7456
7457 return (mp);
7458 }
7459
7460 /*
7461 * Pullup function that should be used for IP input in order to
7462 * ensure we do not loose the L2 source address; we need the l2 source
7463 * address for IP_RECVSLLA and for ndp_input.
7464 *
7465 * We return either NULL or b_rptr.
7466 */
7467 void *
7468 ip_pullup(mblk_t *mp, ssize_t len, ip_recv_attr_t *ira)
7469 {
7470 ill_t *ill = ira->ira_ill;
7471
7472 if (ip_rput_pullups++ == 0) {
7473 (void) mi_strlog(ill->ill_rq, 1, SL_ERROR|SL_TRACE,
7474 "ip_pullup: %s forced us to "
7475 " pullup pkt, hdr len %ld, hdr addr %p",
7476 ill->ill_name, len, (void *)mp->b_rptr);
7477 }
7478 if (!(ira->ira_flags & IRAF_L2SRC_SET))
7479 ip_setl2src(mp, ira, ira->ira_rill);
7480 ASSERT(ira->ira_flags & IRAF_L2SRC_SET);
7481 if (!pullupmsg(mp, len))
7482 return (NULL);
7483 else
7484 return (mp->b_rptr);
7485 }
7486
7487 /*
7488 * Make sure ira_l2src has an address. If we don't have one fill with zeros.
7489 * When called from the ULP ira_rill will be NULL hence the caller has to
7490 * pass in the ill.
7491 */
7492 /* ARGSUSED */
7493 void
7494 ip_setl2src(mblk_t *mp, ip_recv_attr_t *ira, ill_t *ill)
7495 {
7496 const uchar_t *addr;
7497 int alen;
7498
7499 if (ira->ira_flags & IRAF_L2SRC_SET)
7500 return;
7501
7502 ASSERT(ill != NULL);
7503 alen = ill->ill_phys_addr_length;
7504 ASSERT(alen <= sizeof (ira->ira_l2src));
7505 if (ira->ira_mhip != NULL &&
7506 (addr = ira->ira_mhip->mhi_saddr) != NULL) {
7507 bcopy(addr, ira->ira_l2src, alen);
7508 } else if ((ira->ira_flags & IRAF_L2SRC_LOOPBACK) &&
7509 (addr = ill->ill_phys_addr) != NULL) {
7510 bcopy(addr, ira->ira_l2src, alen);
7511 } else {
7512 bzero(ira->ira_l2src, alen);
7513 }
7514 ira->ira_flags |= IRAF_L2SRC_SET;
7515 }
7516
7517 /*
7518 * check ip header length and align it.
7519 */
7520 mblk_t *
7521 ip_check_and_align_header(mblk_t *mp, uint_t min_size, ip_recv_attr_t *ira)
7522 {
7523 ill_t *ill = ira->ira_ill;
7524 ssize_t len;
7525
7526 len = MBLKL(mp);
7527
7528 if (!OK_32PTR(mp->b_rptr))
7529 IP_STAT(ill->ill_ipst, ip_notaligned);
7530 else
7531 IP_STAT(ill->ill_ipst, ip_recv_pullup);
7532
7533 /* Guard against bogus device drivers */
7534 if (len < 0) {
7535 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7536 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7537 freemsg(mp);
7538 return (NULL);
7539 }
7540
7541 if (len == 0) {
7542 /* GLD sometimes sends up mblk with b_rptr == b_wptr! */
7543 mblk_t *mp1 = mp->b_cont;
7544
7545 if (!(ira->ira_flags & IRAF_L2SRC_SET))
7546 ip_setl2src(mp, ira, ira->ira_rill);
7547 ASSERT(ira->ira_flags & IRAF_L2SRC_SET);
7548
7549 freeb(mp);
7550 mp = mp1;
7551 if (mp == NULL)
7552 return (NULL);
7553
7554 if (OK_32PTR(mp->b_rptr) && MBLKL(mp) >= min_size)
7555 return (mp);
7556 }
7557 if (ip_pullup(mp, min_size, ira) == NULL) {
7558 if (msgdsize(mp) < min_size) {
7559 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7560 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7561 } else {
7562 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7563 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7564 }
7565 freemsg(mp);
7566 return (NULL);
7567 }
7568 return (mp);
7569 }
7570
7571 /*
7572 * Common code for IPv4 and IPv6 to check and pullup multi-mblks
7573 */
7574 mblk_t *
7575 ip_check_length(mblk_t *mp, uchar_t *rptr, ssize_t len, uint_t pkt_len,
7576 uint_t min_size, ip_recv_attr_t *ira)
7577 {
7578 ill_t *ill = ira->ira_ill;
7579
7580 /*
7581 * Make sure we have data length consistent
7582 * with the IP header.
7583 */
7584 if (mp->b_cont == NULL) {
7585 /* pkt_len is based on ipha_len, not the mblk length */
7586 if (pkt_len < min_size) {
7587 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7588 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7589 freemsg(mp);
7590 return (NULL);
7591 }
7592 if (len < 0) {
7593 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
7594 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
7595 freemsg(mp);
7596 return (NULL);
7597 }
7598 /* Drop any pad */
7599 mp->b_wptr = rptr + pkt_len;
7600 } else if ((len += msgdsize(mp->b_cont)) != 0) {
7601 ASSERT(pkt_len >= min_size);
7602 if (pkt_len < min_size) {
7603 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7604 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7605 freemsg(mp);
7606 return (NULL);
7607 }
7608 if (len < 0) {
7609 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInTruncatedPkts);
7610 ip_drop_input("ipIfStatsInTruncatedPkts", mp, ill);
7611 freemsg(mp);
7612 return (NULL);
7613 }
7614 /* Drop any pad */
7615 (void) adjmsg(mp, -len);
7616 /*
7617 * adjmsg may have freed an mblk from the chain, hence
7618 * invalidate any hw checksum here. This will force IP to
7619 * calculate the checksum in sw, but only for this packet.
7620 */
7621 DB_CKSUMFLAGS(mp) = 0;
7622 IP_STAT(ill->ill_ipst, ip_multimblk);
7623 }
7624 return (mp);
7625 }
7626
7627 /*
7628 * Check that the IPv4 opt_len is consistent with the packet and pullup
7629 * the options.
7630 */
7631 mblk_t *
7632 ip_check_optlen(mblk_t *mp, ipha_t *ipha, uint_t opt_len, uint_t pkt_len,
7633 ip_recv_attr_t *ira)
7634 {
7635 ill_t *ill = ira->ira_ill;
7636 ssize_t len;
7637
7638 /* Assume no IPv6 packets arrive over the IPv4 queue */
7639 if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) {
7640 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7641 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInWrongIPVersion);
7642 ip_drop_input("IPvN packet on IPv4 ill", mp, ill);
7643 freemsg(mp);
7644 return (NULL);
7645 }
7646
7647 if (opt_len > (15 - IP_SIMPLE_HDR_LENGTH_IN_WORDS)) {
7648 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7649 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7650 freemsg(mp);
7651 return (NULL);
7652 }
7653 /*
7654 * Recompute complete header length and make sure we
7655 * have access to all of it.
7656 */
7657 len = ((size_t)opt_len + IP_SIMPLE_HDR_LENGTH_IN_WORDS) << 2;
7658 if (len > (mp->b_wptr - mp->b_rptr)) {
7659 if (len > pkt_len) {
7660 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInHdrErrors);
7661 ip_drop_input("ipIfStatsInHdrErrors", mp, ill);
7662 freemsg(mp);
7663 return (NULL);
7664 }
7665 if (ip_pullup(mp, len, ira) == NULL) {
7666 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
7667 ip_drop_input("ipIfStatsInDiscards", mp, ill);
7668 freemsg(mp);
7669 return (NULL);
7670 }
7671 }
7672 return (mp);
7673 }
7674
7675 /*
7676 * Returns a new ire, or the same ire, or NULL.
7677 * If a different IRE is returned, then it is held; the caller
7678 * needs to release it.
7679 * In no case is there any hold/release on the ire argument.
7680 */
7681 ire_t *
7682 ip_check_multihome(void *addr, ire_t *ire, ill_t *ill)
7683 {
7684 ire_t *new_ire;
7685 ill_t *ire_ill;
7686 uint_t ifindex;
7687 ip_stack_t *ipst = ill->ill_ipst;
7688 boolean_t strict_check = B_FALSE;
7689
7690 /*
7691 * IPMP common case: if IRE and ILL are in the same group, there's no
7692 * issue (e.g. packet received on an underlying interface matched an
7693 * IRE_LOCAL on its associated group interface).
7694 */
7695 ASSERT(ire->ire_ill != NULL);
7696 if (IS_IN_SAME_ILLGRP(ill, ire->ire_ill))
7697 return (ire);
7698
7699 /*
7700 * Do another ire lookup here, using the ingress ill, to see if the
7701 * interface is in a usesrc group.
7702 * As long as the ills belong to the same group, we don't consider
7703 * them to be arriving on the wrong interface. Thus, if the switch
7704 * is doing inbound load spreading, we won't drop packets when the
7705 * ip*_strict_dst_multihoming switch is on.
7706 * We also need to check for IPIF_UNNUMBERED point2point interfaces
7707 * where the local address may not be unique. In this case we were
7708 * at the mercy of the initial ire lookup and the IRE_LOCAL it
7709 * actually returned. The new lookup, which is more specific, should
7710 * only find the IRE_LOCAL associated with the ingress ill if one
7711 * exists.
7712 */
7713 if (ire->ire_ipversion == IPV4_VERSION) {
7714 if (ipst->ips_ip_strict_dst_multihoming)
7715 strict_check = B_TRUE;
7716 new_ire = ire_ftable_lookup_v4(*((ipaddr_t *)addr), 0, 0,
7717 IRE_LOCAL, ill, ALL_ZONES, NULL,
7718 (MATCH_IRE_TYPE|MATCH_IRE_ILL), 0, ipst, NULL);
7719 } else {
7720 ASSERT(!IN6_IS_ADDR_MULTICAST((in6_addr_t *)addr));
7721 if (ipst->ips_ipv6_strict_dst_multihoming)
7722 strict_check = B_TRUE;
7723 new_ire = ire_ftable_lookup_v6((in6_addr_t *)addr, NULL, NULL,
7724 IRE_LOCAL, ill, ALL_ZONES, NULL,
7725 (MATCH_IRE_TYPE|MATCH_IRE_ILL), 0, ipst, NULL);
7726 }
7727 /*
7728 * If the same ire that was returned in ip_input() is found then this
7729 * is an indication that usesrc groups are in use. The packet
7730 * arrived on a different ill in the group than the one associated with
7731 * the destination address. If a different ire was found then the same
7732 * IP address must be hosted on multiple ills. This is possible with
7733 * unnumbered point2point interfaces. We switch to use this new ire in
7734 * order to have accurate interface statistics.
7735 */
7736 if (new_ire != NULL) {
7737 /* Note: held in one case but not the other? Caller handles */
7738 if (new_ire != ire)
7739 return (new_ire);
7740 /* Unchanged */
7741 ire_refrele(new_ire);
7742 return (ire);
7743 }
7744
7745 /*
7746 * Chase pointers once and store locally.
7747 */
7748 ASSERT(ire->ire_ill != NULL);
7749 ire_ill = ire->ire_ill;
7750 ifindex = ill->ill_usesrc_ifindex;
7751
7752 /*
7753 * Check if it's a legal address on the 'usesrc' interface.
7754 * For IPMP data addresses the IRE_LOCAL is the upper, hence we
7755 * can just check phyint_ifindex.
7756 */
7757 if (ifindex != 0 && ifindex == ire_ill->ill_phyint->phyint_ifindex) {
7758 return (ire);
7759 }
7760
7761 /*
7762 * If the ip*_strict_dst_multihoming switch is on then we can
7763 * only accept this packet if the interface is marked as routing.
7764 */
7765 if (!(strict_check))
7766 return (ire);
7767
7768 if ((ill->ill_flags & ire->ire_ill->ill_flags & ILLF_ROUTER) != 0) {
7769 return (ire);
7770 }
7771 return (NULL);
7772 }
7773
7774 /*
7775 * This function is used to construct a mac_header_info_s from a
7776 * DL_UNITDATA_IND message.
7777 * The address fields in the mhi structure points into the message,
7778 * thus the caller can't use those fields after freeing the message.
7779 *
7780 * We determine whether the packet received is a non-unicast packet
7781 * and in doing so, determine whether or not it is broadcast vs multicast.
7782 * For it to be a broadcast packet, we must have the appropriate mblk_t
7783 * hanging off the ill_t. If this is either not present or doesn't match
7784 * the destination mac address in the DL_UNITDATA_IND, the packet is deemed
7785 * to be multicast. Thus NICs that have no broadcast address (or no
7786 * capability for one, such as point to point links) cannot return as
7787 * the packet being broadcast.
7788 */
7789 void
7790 ip_dlur_to_mhi(ill_t *ill, mblk_t *mb, struct mac_header_info_s *mhip)
7791 {
7792 dl_unitdata_ind_t *ind = (dl_unitdata_ind_t *)mb->b_rptr;
7793 mblk_t *bmp;
7794 uint_t extra_offset;
7795
7796 bzero(mhip, sizeof (struct mac_header_info_s));
7797
7798 mhip->mhi_dsttype = MAC_ADDRTYPE_UNICAST;
7799
7800 if (ill->ill_sap_length < 0)
7801 extra_offset = 0;
7802 else
7803 extra_offset = ill->ill_sap_length;
7804
7805 mhip->mhi_daddr = (uchar_t *)ind + ind->dl_dest_addr_offset +
7806 extra_offset;
7807 mhip->mhi_saddr = (uchar_t *)ind + ind->dl_src_addr_offset +
7808 extra_offset;
7809
7810 if (!ind->dl_group_address)
7811 return;
7812
7813 /* Multicast or broadcast */
7814 mhip->mhi_dsttype = MAC_ADDRTYPE_MULTICAST;
7815
7816 if (ind->dl_dest_addr_offset > sizeof (*ind) &&
7817 ind->dl_dest_addr_offset + ind->dl_dest_addr_length < MBLKL(mb) &&
7818 (bmp = ill->ill_bcast_mp) != NULL) {
7819 dl_unitdata_req_t *dlur;
7820 uint8_t *bphys_addr;
7821
7822 dlur = (dl_unitdata_req_t *)bmp->b_rptr;
7823 bphys_addr = (uchar_t *)dlur + dlur->dl_dest_addr_offset +
7824 extra_offset;
7825
7826 if (bcmp(mhip->mhi_daddr, bphys_addr,
7827 ind->dl_dest_addr_length) == 0)
7828 mhip->mhi_dsttype = MAC_ADDRTYPE_BROADCAST;
7829 }
7830 }
7831
7832 /*
7833 * This function is used to construct a mac_header_info_s from a
7834 * M_DATA fastpath message from a DLPI driver.
7835 * The address fields in the mhi structure points into the message,
7836 * thus the caller can't use those fields after freeing the message.
7837 *
7838 * We determine whether the packet received is a non-unicast packet
7839 * and in doing so, determine whether or not it is broadcast vs multicast.
7840 * For it to be a broadcast packet, we must have the appropriate mblk_t
7841 * hanging off the ill_t. If this is either not present or doesn't match
7842 * the destination mac address in the DL_UNITDATA_IND, the packet is deemed
7843 * to be multicast. Thus NICs that have no broadcast address (or no
7844 * capability for one, such as point to point links) cannot return as
7845 * the packet being broadcast.
7846 */
7847 void
7848 ip_mdata_to_mhi(ill_t *ill, mblk_t *mp, struct mac_header_info_s *mhip)
7849 {
7850 mblk_t *bmp;
7851 struct ether_header *pether;
7852
7853 bzero(mhip, sizeof (struct mac_header_info_s));
7854
7855 mhip->mhi_dsttype = MAC_ADDRTYPE_UNICAST;
7856
7857 pether = (struct ether_header *)((char *)mp->b_rptr
7858 - sizeof (struct ether_header));
7859
7860 /*
7861 * Make sure the interface is an ethernet type, since we don't
7862 * know the header format for anything but Ethernet. Also make
7863 * sure we are pointing correctly above db_base.
7864 */
7865 if (ill->ill_type != IFT_ETHER)
7866 return;
7867
7868 retry:
7869 if ((uchar_t *)pether < mp->b_datap->db_base)
7870 return;
7871
7872 /* Is there a VLAN tag? */
7873 if (ill->ill_isv6) {
7874 if (pether->ether_type != htons(ETHERTYPE_IPV6)) {
7875 pether = (struct ether_header *)((char *)pether - 4);
7876 goto retry;
7877 }
7878 } else {
7879 if (pether->ether_type != htons(ETHERTYPE_IP)) {
7880 pether = (struct ether_header *)((char *)pether - 4);
7881 goto retry;
7882 }
7883 }
7884 mhip->mhi_daddr = (uchar_t *)&pether->ether_dhost;
7885 mhip->mhi_saddr = (uchar_t *)&pether->ether_shost;
7886
7887 if (!(mhip->mhi_daddr[0] & 0x01))
7888 return;
7889
7890 /* Multicast or broadcast */
7891 mhip->mhi_dsttype = MAC_ADDRTYPE_MULTICAST;
7892
7893 if ((bmp = ill->ill_bcast_mp) != NULL) {
7894 dl_unitdata_req_t *dlur;
7895 uint8_t *bphys_addr;
7896 uint_t addrlen;
7897
7898 dlur = (dl_unitdata_req_t *)bmp->b_rptr;
7899 addrlen = dlur->dl_dest_addr_length;
7900 if (ill->ill_sap_length < 0) {
7901 bphys_addr = (uchar_t *)dlur +
7902 dlur->dl_dest_addr_offset;
7903 addrlen += ill->ill_sap_length;
7904 } else {
7905 bphys_addr = (uchar_t *)dlur +
7906 dlur->dl_dest_addr_offset +
7907 ill->ill_sap_length;
7908 addrlen -= ill->ill_sap_length;
7909 }
7910 if (bcmp(mhip->mhi_daddr, bphys_addr, addrlen) == 0)
7911 mhip->mhi_dsttype = MAC_ADDRTYPE_BROADCAST;
7912 }
7913 }
7914
7915 /*
7916 * Handle anything but M_DATA messages
7917 * We see the DL_UNITDATA_IND which are part
7918 * of the data path, and also the other messages from the driver.
7919 */
7920 void
7921 ip_rput_notdata(ill_t *ill, mblk_t *mp)
7922 {
7923 mblk_t *first_mp;
7924 struct iocblk *iocp;
7925 struct mac_header_info_s mhi;
7926
7927 switch (DB_TYPE(mp)) {
7928 case M_PROTO:
7929 case M_PCPROTO: {
7930 if (((dl_unitdata_ind_t *)mp->b_rptr)->dl_primitive !=
7931 DL_UNITDATA_IND) {
7932 /* Go handle anything other than data elsewhere. */
7933 ip_rput_dlpi(ill, mp);
7934 return;
7935 }
7936
7937 first_mp = mp;
7938 mp = first_mp->b_cont;
7939 first_mp->b_cont = NULL;
7940
7941 if (mp == NULL) {
7942 freeb(first_mp);
7943 return;
7944 }
7945 ip_dlur_to_mhi(ill, first_mp, &mhi);
7946 if (ill->ill_isv6)
7947 ip_input_v6(ill, NULL, mp, &mhi);
7948 else
7949 ip_input(ill, NULL, mp, &mhi);
7950
7951 /* Ditch the DLPI header. */
7952 freeb(first_mp);
7953 return;
7954 }
7955 case M_IOCACK:
7956 iocp = (struct iocblk *)mp->b_rptr;
7957 switch (iocp->ioc_cmd) {
7958 case DL_IOC_HDR_INFO:
7959 ill_fastpath_ack(ill, mp);
7960 return;
7961 default:
7962 putnext(ill->ill_rq, mp);
7963 return;
7964 }
7965 /* FALLTHRU */
7966 case M_ERROR:
7967 case M_HANGUP:
7968 mutex_enter(&ill->ill_lock);
7969 if (ill->ill_state_flags & ILL_CONDEMNED) {
7970 mutex_exit(&ill->ill_lock);
7971 freemsg(mp);
7972 return;
7973 }
7974 ill_refhold_locked(ill);
7975 mutex_exit(&ill->ill_lock);
7976 qwriter_ip(ill, ill->ill_rq, mp, ip_rput_other, CUR_OP,
7977 B_FALSE);
7978 return;
7979 case M_CTL:
7980 putnext(ill->ill_rq, mp);
7981 return;
7982 case M_IOCNAK:
7983 ip1dbg(("got iocnak "));
7984 iocp = (struct iocblk *)mp->b_rptr;
7985 switch (iocp->ioc_cmd) {
7986 case DL_IOC_HDR_INFO:
7987 ip_rput_other(NULL, ill->ill_rq, mp, NULL);
7988 return;
7989 default:
7990 break;
7991 }
7992 /* FALLTHRU */
7993 default:
7994 putnext(ill->ill_rq, mp);
7995 return;
7996 }
7997 }
7998
7999 /* Read side put procedure. Packets coming from the wire arrive here. */
8000 void
8001 ip_rput(queue_t *q, mblk_t *mp)
8002 {
8003 ill_t *ill;
8004 union DL_primitives *dl;
8005
8006 ill = (ill_t *)q->q_ptr;
8007
8008 if (ill->ill_state_flags & (ILL_CONDEMNED | ILL_LL_SUBNET_PENDING)) {
8009 /*
8010 * If things are opening or closing, only accept high-priority
8011 * DLPI messages. (On open ill->ill_ipif has not yet been
8012 * created; on close, things hanging off the ill may have been
8013 * freed already.)
8014 */
8015 dl = (union DL_primitives *)mp->b_rptr;
8016 if (DB_TYPE(mp) != M_PCPROTO ||
8017 dl->dl_primitive == DL_UNITDATA_IND) {
8018 inet_freemsg(mp);
8019 return;
8020 }
8021 }
8022 if (DB_TYPE(mp) == M_DATA) {
8023 struct mac_header_info_s mhi;
8024
8025 ip_mdata_to_mhi(ill, mp, &mhi);
8026 ip_input(ill, NULL, mp, &mhi);
8027 } else {
8028 ip_rput_notdata(ill, mp);
8029 }
8030 }
8031
8032 /*
8033 * Move the information to a copy.
8034 */
8035 mblk_t *
8036 ip_fix_dbref(mblk_t *mp, ip_recv_attr_t *ira)
8037 {
8038 mblk_t *mp1;
8039 ill_t *ill = ira->ira_ill;
8040 ip_stack_t *ipst = ill->ill_ipst;
8041
8042 IP_STAT(ipst, ip_db_ref);
8043
8044 /* Make sure we have ira_l2src before we loose the original mblk */
8045 if (!(ira->ira_flags & IRAF_L2SRC_SET))
8046 ip_setl2src(mp, ira, ira->ira_rill);
8047
8048 mp1 = copymsg(mp);
8049 if (mp1 == NULL) {
8050 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
8051 ip_drop_input("ipIfStatsInDiscards", mp, ill);
8052 freemsg(mp);
8053 return (NULL);
8054 }
8055 /* preserve the hardware checksum flags and data, if present */
8056 if (DB_CKSUMFLAGS(mp) != 0) {
8057 DB_CKSUMFLAGS(mp1) = DB_CKSUMFLAGS(mp);
8058 DB_CKSUMSTART(mp1) = DB_CKSUMSTART(mp);
8059 DB_CKSUMSTUFF(mp1) = DB_CKSUMSTUFF(mp);
8060 DB_CKSUMEND(mp1) = DB_CKSUMEND(mp);
8061 DB_CKSUM16(mp1) = DB_CKSUM16(mp);
8062 }
8063 freemsg(mp);
8064 return (mp1);
8065 }
8066
8067 static void
8068 ip_dlpi_error(ill_t *ill, t_uscalar_t prim, t_uscalar_t dl_err,
8069 t_uscalar_t err)
8070 {
8071 if (dl_err == DL_SYSERR) {
8072 (void) mi_strlog(ill->ill_rq, 1, SL_CONSOLE|SL_ERROR|SL_TRACE,
8073 "%s: %s failed: DL_SYSERR (errno %u)\n",
8074 ill->ill_name, dl_primstr(prim), err);
8075 return;
8076 }
8077
8078 (void) mi_strlog(ill->ill_rq, 1, SL_CONSOLE|SL_ERROR|SL_TRACE,
8079 "%s: %s failed: %s\n", ill->ill_name, dl_primstr(prim),
8080 dl_errstr(dl_err));
8081 }
8082
8083 /*
8084 * ip_rput_dlpi is called by ip_rput to handle all DLPI messages other
8085 * than DL_UNITDATA_IND messages. If we need to process this message
8086 * exclusively, we call qwriter_ip, in which case we also need to call
8087 * ill_refhold before that, since qwriter_ip does an ill_refrele.
8088 */
8089 void
8090 ip_rput_dlpi(ill_t *ill, mblk_t *mp)
8091 {
8092 dl_ok_ack_t *dloa = (dl_ok_ack_t *)mp->b_rptr;
8093 dl_error_ack_t *dlea = (dl_error_ack_t *)dloa;
8094 queue_t *q = ill->ill_rq;
8095 t_uscalar_t prim = dloa->dl_primitive;
8096 t_uscalar_t reqprim = DL_PRIM_INVAL;
8097
8098 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi",
8099 char *, dl_primstr(prim), ill_t *, ill);
8100 ip1dbg(("ip_rput_dlpi"));
8101
8102 /*
8103 * If we received an ACK but didn't send a request for it, then it
8104 * can't be part of any pending operation; discard up-front.
8105 */
8106 switch (prim) {
8107 case DL_ERROR_ACK:
8108 reqprim = dlea->dl_error_primitive;
8109 ip2dbg(("ip_rput_dlpi(%s): DL_ERROR_ACK for %s (0x%x): %s "
8110 "(0x%x), unix %u\n", ill->ill_name, dl_primstr(reqprim),
8111 reqprim, dl_errstr(dlea->dl_errno), dlea->dl_errno,
8112 dlea->dl_unix_errno));
8113 break;
8114 case DL_OK_ACK:
8115 reqprim = dloa->dl_correct_primitive;
8116 break;
8117 case DL_INFO_ACK:
8118 reqprim = DL_INFO_REQ;
8119 break;
8120 case DL_BIND_ACK:
8121 reqprim = DL_BIND_REQ;
8122 break;
8123 case DL_PHYS_ADDR_ACK:
8124 reqprim = DL_PHYS_ADDR_REQ;
8125 break;
8126 case DL_NOTIFY_ACK:
8127 reqprim = DL_NOTIFY_REQ;
8128 break;
8129 case DL_CAPABILITY_ACK:
8130 reqprim = DL_CAPABILITY_REQ;
8131 break;
8132 }
8133
8134 if (prim != DL_NOTIFY_IND) {
8135 if (reqprim == DL_PRIM_INVAL ||
8136 !ill_dlpi_pending(ill, reqprim)) {
8137 /* Not a DLPI message we support or expected */
8138 freemsg(mp);
8139 return;
8140 }
8141 ip1dbg(("ip_rput: received %s for %s\n", dl_primstr(prim),
8142 dl_primstr(reqprim)));
8143 }
8144
8145 switch (reqprim) {
8146 case DL_UNBIND_REQ:
8147 /*
8148 * NOTE: we mark the unbind as complete even if we got a
8149 * DL_ERROR_ACK, since there's not much else we can do.
8150 */
8151 mutex_enter(&ill->ill_lock);
8152 ill->ill_state_flags &= ~ILL_DL_UNBIND_IN_PROGRESS;
8153 cv_signal(&ill->ill_cv);
8154 mutex_exit(&ill->ill_lock);
8155 break;
8156
8157 case DL_ENABMULTI_REQ:
8158 if (prim == DL_OK_ACK) {
8159 if (ill->ill_dlpi_multicast_state == IDS_INPROGRESS)
8160 ill->ill_dlpi_multicast_state = IDS_OK;
8161 }
8162 break;
8163 }
8164
8165 /*
8166 * The message is one we're waiting for (or DL_NOTIFY_IND), but we
8167 * need to become writer to continue to process it. Because an
8168 * exclusive operation doesn't complete until replies to all queued
8169 * DLPI messages have been received, we know we're in the middle of an
8170 * exclusive operation and pass CUR_OP (except for DL_NOTIFY_IND).
8171 *
8172 * As required by qwriter_ip(), we refhold the ill; it will refrele.
8173 * Since this is on the ill stream we unconditionally bump up the
8174 * refcount without doing ILL_CAN_LOOKUP().
8175 */
8176 ill_refhold(ill);
8177 if (prim == DL_NOTIFY_IND)
8178 qwriter_ip(ill, q, mp, ip_rput_dlpi_writer, NEW_OP, B_FALSE);
8179 else
8180 qwriter_ip(ill, q, mp, ip_rput_dlpi_writer, CUR_OP, B_FALSE);
8181 }
8182
8183 /*
8184 * Handling of DLPI messages that require exclusive access to the ipsq.
8185 *
8186 * Need to do ipsq_pending_mp_get on ioctl completion, which could
8187 * happen here. (along with mi_copy_done)
8188 */
8189 /* ARGSUSED */
8190 static void
8191 ip_rput_dlpi_writer(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
8192 {
8193 dl_ok_ack_t *dloa = (dl_ok_ack_t *)mp->b_rptr;
8194 dl_error_ack_t *dlea = (dl_error_ack_t *)dloa;
8195 int err = 0;
8196 ill_t *ill = (ill_t *)q->q_ptr;
8197 ipif_t *ipif = NULL;
8198 mblk_t *mp1 = NULL;
8199 conn_t *connp = NULL;
8200 t_uscalar_t paddrreq;
8201 mblk_t *mp_hw;
8202 boolean_t success;
8203 boolean_t ioctl_aborted = B_FALSE;
8204 boolean_t log = B_TRUE;
8205
8206 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer",
8207 char *, dl_primstr(dloa->dl_primitive), ill_t *, ill);
8208
8209 ip1dbg(("ip_rput_dlpi_writer .."));
8210 ASSERT(ipsq->ipsq_xop == ill->ill_phyint->phyint_ipsq->ipsq_xop);
8211 ASSERT(IAM_WRITER_ILL(ill));
8212
8213 ipif = ipsq->ipsq_xop->ipx_pending_ipif;
8214 /*
8215 * The current ioctl could have been aborted by the user and a new
8216 * ioctl to bring up another ill could have started. We could still
8217 * get a response from the driver later.
8218 */
8219 if (ipif != NULL && ipif->ipif_ill != ill)
8220 ioctl_aborted = B_TRUE;
8221
8222 switch (dloa->dl_primitive) {
8223 case DL_ERROR_ACK:
8224 ip1dbg(("ip_rput_dlpi_writer: got DL_ERROR_ACK for %s\n",
8225 dl_primstr(dlea->dl_error_primitive)));
8226
8227 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer error",
8228 char *, dl_primstr(dlea->dl_error_primitive),
8229 ill_t *, ill);
8230
8231 switch (dlea->dl_error_primitive) {
8232 case DL_DISABMULTI_REQ:
8233 ill_dlpi_done(ill, dlea->dl_error_primitive);
8234 break;
8235 case DL_PROMISCON_REQ:
8236 case DL_PROMISCOFF_REQ:
8237 case DL_UNBIND_REQ:
8238 case DL_ATTACH_REQ:
8239 case DL_INFO_REQ:
8240 ill_dlpi_done(ill, dlea->dl_error_primitive);
8241 break;
8242 case DL_NOTIFY_REQ:
8243 ill_dlpi_done(ill, DL_NOTIFY_REQ);
8244 log = B_FALSE;
8245 break;
8246 case DL_PHYS_ADDR_REQ:
8247 /*
8248 * For IPv6 only, there are two additional
8249 * phys_addr_req's sent to the driver to get the
8250 * IPv6 token and lla. This allows IP to acquire
8251 * the hardware address format for a given interface
8252 * without having built in knowledge of the hardware
8253 * address. ill_phys_addr_pend keeps track of the last
8254 * DL_PAR sent so we know which response we are
8255 * dealing with. ill_dlpi_done will update
8256 * ill_phys_addr_pend when it sends the next req.
8257 * We don't complete the IOCTL until all three DL_PARs
8258 * have been attempted, so set *_len to 0 and break.
8259 */
8260 paddrreq = ill->ill_phys_addr_pend;
8261 ill_dlpi_done(ill, DL_PHYS_ADDR_REQ);
8262 if (paddrreq == DL_IPV6_TOKEN) {
8263 ill->ill_token_length = 0;
8264 log = B_FALSE;
8265 break;
8266 } else if (paddrreq == DL_IPV6_LINK_LAYER_ADDR) {
8267 ill->ill_nd_lla_len = 0;
8268 log = B_FALSE;
8269 break;
8270 }
8271 /*
8272 * Something went wrong with the DL_PHYS_ADDR_REQ.
8273 * We presumably have an IOCTL hanging out waiting
8274 * for completion. Find it and complete the IOCTL
8275 * with the error noted.
8276 * However, ill_dl_phys was called on an ill queue
8277 * (from SIOCSLIFNAME), thus conn_pending_ill is not
8278 * set. But the ioctl is known to be pending on ill_wq.
8279 */
8280 if (!ill->ill_ifname_pending)
8281 break;
8282 ill->ill_ifname_pending = 0;
8283 if (!ioctl_aborted)
8284 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8285 if (mp1 != NULL) {
8286 /*
8287 * This operation (SIOCSLIFNAME) must have
8288 * happened on the ill. Assert there is no conn
8289 */
8290 ASSERT(connp == NULL);
8291 q = ill->ill_wq;
8292 }
8293 break;
8294 case DL_BIND_REQ:
8295 ill_dlpi_done(ill, DL_BIND_REQ);
8296 if (ill->ill_ifname_pending)
8297 break;
8298 mutex_enter(&ill->ill_lock);
8299 ill->ill_state_flags &= ~ILL_DOWN_IN_PROGRESS;
8300 mutex_exit(&ill->ill_lock);
8301 /*
8302 * Something went wrong with the bind. We presumably
8303 * have an IOCTL hanging out waiting for completion.
8304 * Find it, take down the interface that was coming
8305 * up, and complete the IOCTL with the error noted.
8306 */
8307 if (!ioctl_aborted)
8308 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8309 if (mp1 != NULL) {
8310 /*
8311 * This might be a result of a DL_NOTE_REPLUMB
8312 * notification. In that case, connp is NULL.
8313 */
8314 if (connp != NULL)
8315 q = CONNP_TO_WQ(connp);
8316
8317 (void) ipif_down(ipif, NULL, NULL);
8318 /* error is set below the switch */
8319 }
8320 break;
8321 case DL_ENABMULTI_REQ:
8322 ill_dlpi_done(ill, DL_ENABMULTI_REQ);
8323
8324 if (ill->ill_dlpi_multicast_state == IDS_INPROGRESS)
8325 ill->ill_dlpi_multicast_state = IDS_FAILED;
8326 if (ill->ill_dlpi_multicast_state == IDS_FAILED) {
8327
8328 printf("ip: joining multicasts failed (%d)"
8329 " on %s - will use link layer "
8330 "broadcasts for multicast\n",
8331 dlea->dl_errno, ill->ill_name);
8332
8333 /*
8334 * Set up for multi_bcast; We are the
8335 * writer, so ok to access ill->ill_ipif
8336 * without any lock.
8337 */
8338 mutex_enter(&ill->ill_phyint->phyint_lock);
8339 ill->ill_phyint->phyint_flags |=
8340 PHYI_MULTI_BCAST;
8341 mutex_exit(&ill->ill_phyint->phyint_lock);
8342
8343 }
8344 freemsg(mp); /* Don't want to pass this up */
8345 return;
8346 case DL_CAPABILITY_REQ:
8347 ip1dbg(("ip_rput_dlpi_writer: got DL_ERROR_ACK for "
8348 "DL_CAPABILITY REQ\n"));
8349 if (ill->ill_dlpi_capab_state == IDCS_PROBE_SENT)
8350 ill->ill_dlpi_capab_state = IDCS_FAILED;
8351 ill_capability_done(ill);
8352 freemsg(mp);
8353 return;
8354 }
8355 /*
8356 * Note the error for IOCTL completion (mp1 is set when
8357 * ready to complete ioctl). If ill_ifname_pending_err is
8358 * set, an error occured during plumbing (ill_ifname_pending),
8359 * so we want to report that error.
8360 *
8361 * NOTE: there are two addtional DL_PHYS_ADDR_REQ's
8362 * (DL_IPV6_TOKEN and DL_IPV6_LINK_LAYER_ADDR) that are
8363 * expected to get errack'd if the driver doesn't support
8364 * these flags (e.g. ethernet). log will be set to B_FALSE
8365 * if these error conditions are encountered.
8366 */
8367 if (mp1 != NULL) {
8368 if (ill->ill_ifname_pending_err != 0) {
8369 err = ill->ill_ifname_pending_err;
8370 ill->ill_ifname_pending_err = 0;
8371 } else {
8372 err = dlea->dl_unix_errno ?
8373 dlea->dl_unix_errno : ENXIO;
8374 }
8375 /*
8376 * If we're plumbing an interface and an error hasn't already
8377 * been saved, set ill_ifname_pending_err to the error passed
8378 * up. Ignore the error if log is B_FALSE (see comment above).
8379 */
8380 } else if (log && ill->ill_ifname_pending &&
8381 ill->ill_ifname_pending_err == 0) {
8382 ill->ill_ifname_pending_err = dlea->dl_unix_errno ?
8383 dlea->dl_unix_errno : ENXIO;
8384 }
8385
8386 if (log)
8387 ip_dlpi_error(ill, dlea->dl_error_primitive,
8388 dlea->dl_errno, dlea->dl_unix_errno);
8389 break;
8390 case DL_CAPABILITY_ACK:
8391 ill_capability_ack(ill, mp);
8392 /*
8393 * The message has been handed off to ill_capability_ack
8394 * and must not be freed below
8395 */
8396 mp = NULL;
8397 break;
8398
8399 case DL_INFO_ACK:
8400 /* Call a routine to handle this one. */
8401 ill_dlpi_done(ill, DL_INFO_REQ);
8402 ip_ll_subnet_defaults(ill, mp);
8403 ASSERT(!MUTEX_HELD(&ill->ill_phyint->phyint_ipsq->ipsq_lock));
8404 return;
8405 case DL_BIND_ACK:
8406 /*
8407 * We should have an IOCTL waiting on this unless
8408 * sent by ill_dl_phys, in which case just return
8409 */
8410 ill_dlpi_done(ill, DL_BIND_REQ);
8411
8412 if (ill->ill_ifname_pending) {
8413 DTRACE_PROBE2(ip__rput__dlpi__ifname__pending,
8414 ill_t *, ill, mblk_t *, mp);
8415 break;
8416 }
8417 mutex_enter(&ill->ill_lock);
8418 ill->ill_dl_up = 1;
8419 ill->ill_state_flags &= ~ILL_DOWN_IN_PROGRESS;
8420 mutex_exit(&ill->ill_lock);
8421
8422 if (!ioctl_aborted)
8423 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8424 if (mp1 == NULL) {
8425 DTRACE_PROBE1(ip__rput__dlpi__no__mblk, ill_t *, ill);
8426 break;
8427 }
8428 /*
8429 * mp1 was added by ill_dl_up(). if that is a result of
8430 * a DL_NOTE_REPLUMB notification, connp could be NULL.
8431 */
8432 if (connp != NULL)
8433 q = CONNP_TO_WQ(connp);
8434 /*
8435 * We are exclusive. So nothing can change even after
8436 * we get the pending mp.
8437 */
8438 ip1dbg(("ip_rput_dlpi: bind_ack %s\n", ill->ill_name));
8439 DTRACE_PROBE1(ip__rput__dlpi__bind__ack, ill_t *, ill);
8440 ill_nic_event_dispatch(ill, 0, NE_UP, NULL, 0);
8441
8442 /*
8443 * Now bring up the resolver; when that is complete, we'll
8444 * create IREs. Note that we intentionally mirror what
8445 * ipif_up() would have done, because we got here by way of
8446 * ill_dl_up(), which stopped ipif_up()'s processing.
8447 */
8448 if (ill->ill_isv6) {
8449 /*
8450 * v6 interfaces.
8451 * Unlike ARP which has to do another bind
8452 * and attach, once we get here we are
8453 * done with NDP
8454 */
8455 (void) ipif_resolver_up(ipif, Res_act_initial);
8456 if ((err = ipif_ndp_up(ipif, B_TRUE)) == 0)
8457 err = ipif_up_done_v6(ipif);
8458 } else if (ill->ill_net_type == IRE_IF_RESOLVER) {
8459 /*
8460 * ARP and other v4 external resolvers.
8461 * Leave the pending mblk intact so that
8462 * the ioctl completes in ip_rput().
8463 */
8464 if (connp != NULL)
8465 mutex_enter(&connp->conn_lock);
8466 mutex_enter(&ill->ill_lock);
8467 success = ipsq_pending_mp_add(connp, ipif, q, mp1, 0);
8468 mutex_exit(&ill->ill_lock);
8469 if (connp != NULL)
8470 mutex_exit(&connp->conn_lock);
8471 if (success) {
8472 err = ipif_resolver_up(ipif, Res_act_initial);
8473 if (err == EINPROGRESS) {
8474 freemsg(mp);
8475 return;
8476 }
8477 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8478 } else {
8479 /* The conn has started closing */
8480 err = EINTR;
8481 }
8482 } else {
8483 /*
8484 * This one is complete. Reply to pending ioctl.
8485 */
8486 (void) ipif_resolver_up(ipif, Res_act_initial);
8487 err = ipif_up_done(ipif);
8488 }
8489
8490 if ((err == 0) && (ill->ill_up_ipifs)) {
8491 err = ill_up_ipifs(ill, q, mp1);
8492 if (err == EINPROGRESS) {
8493 freemsg(mp);
8494 return;
8495 }
8496 }
8497
8498 /*
8499 * If we have a moved ipif to bring up, and everything has
8500 * succeeded to this point, bring it up on the IPMP ill.
8501 * Otherwise, leave it down -- the admin can try to bring it
8502 * up by hand if need be.
8503 */
8504 if (ill->ill_move_ipif != NULL) {
8505 if (err != 0) {
8506 ill->ill_move_ipif = NULL;
8507 } else {
8508 ipif = ill->ill_move_ipif;
8509 ill->ill_move_ipif = NULL;
8510 err = ipif_up(ipif, q, mp1);
8511 if (err == EINPROGRESS) {
8512 freemsg(mp);
8513 return;
8514 }
8515 }
8516 }
8517 break;
8518
8519 case DL_NOTIFY_IND: {
8520 dl_notify_ind_t *notify = (dl_notify_ind_t *)mp->b_rptr;
8521 uint_t orig_mtu, orig_mc_mtu;
8522
8523 switch (notify->dl_notification) {
8524 case DL_NOTE_PHYS_ADDR:
8525 err = ill_set_phys_addr(ill, mp);
8526 break;
8527
8528 case DL_NOTE_REPLUMB:
8529 /*
8530 * Directly return after calling ill_replumb().
8531 * Note that we should not free mp as it is reused
8532 * in the ill_replumb() function.
8533 */
8534 err = ill_replumb(ill, mp);
8535 return;
8536
8537 case DL_NOTE_FASTPATH_FLUSH:
8538 nce_flush(ill, B_FALSE);
8539 break;
8540
8541 case DL_NOTE_SDU_SIZE:
8542 case DL_NOTE_SDU_SIZE2:
8543 /*
8544 * The dce and fragmentation code can cope with
8545 * this changing while packets are being sent.
8546 * When packets are sent ip_output will discover
8547 * a change.
8548 *
8549 * Change the MTU size of the interface.
8550 */
8551 mutex_enter(&ill->ill_lock);
8552 orig_mtu = ill->ill_mtu;
8553 orig_mc_mtu = ill->ill_mc_mtu;
8554 switch (notify->dl_notification) {
8555 case DL_NOTE_SDU_SIZE:
8556 ill->ill_current_frag =
8557 (uint_t)notify->dl_data;
8558 ill->ill_mc_mtu = (uint_t)notify->dl_data;
8559 break;
8560 case DL_NOTE_SDU_SIZE2:
8561 ill->ill_current_frag =
8562 (uint_t)notify->dl_data1;
8563 ill->ill_mc_mtu = (uint_t)notify->dl_data2;
8564 break;
8565 }
8566 if (ill->ill_current_frag > ill->ill_max_frag)
8567 ill->ill_max_frag = ill->ill_current_frag;
8568
8569 if (!(ill->ill_flags & ILLF_FIXEDMTU)) {
8570 ill->ill_mtu = ill->ill_current_frag;
8571
8572 /*
8573 * If ill_user_mtu was set (via
8574 * SIOCSLIFLNKINFO), clamp ill_mtu at it.
8575 */
8576 if (ill->ill_user_mtu != 0 &&
8577 ill->ill_user_mtu < ill->ill_mtu)
8578 ill->ill_mtu = ill->ill_user_mtu;
8579
8580 if (ill->ill_user_mtu != 0 &&
8581 ill->ill_user_mtu < ill->ill_mc_mtu)
8582 ill->ill_mc_mtu = ill->ill_user_mtu;
8583
8584 if (ill->ill_isv6) {
8585 if (ill->ill_mtu < IPV6_MIN_MTU)
8586 ill->ill_mtu = IPV6_MIN_MTU;
8587 if (ill->ill_mc_mtu < IPV6_MIN_MTU)
8588 ill->ill_mc_mtu = IPV6_MIN_MTU;
8589 } else {
8590 if (ill->ill_mtu < IP_MIN_MTU)
8591 ill->ill_mtu = IP_MIN_MTU;
8592 if (ill->ill_mc_mtu < IP_MIN_MTU)
8593 ill->ill_mc_mtu = IP_MIN_MTU;
8594 }
8595 } else if (ill->ill_mc_mtu > ill->ill_mtu) {
8596 ill->ill_mc_mtu = ill->ill_mtu;
8597 }
8598
8599 mutex_exit(&ill->ill_lock);
8600 /*
8601 * Make sure all dce_generation checks find out
8602 * that ill_mtu/ill_mc_mtu has changed.
8603 */
8604 if (orig_mtu != ill->ill_mtu ||
8605 orig_mc_mtu != ill->ill_mc_mtu) {
8606 dce_increment_all_generations(ill->ill_isv6,
8607 ill->ill_ipst);
8608 }
8609
8610 /*
8611 * Refresh IPMP meta-interface MTU if necessary.
8612 */
8613 if (IS_UNDER_IPMP(ill))
8614 ipmp_illgrp_refresh_mtu(ill->ill_grp);
8615 break;
8616
8617 case DL_NOTE_LINK_UP:
8618 case DL_NOTE_LINK_DOWN: {
8619 /*
8620 * We are writer. ill / phyint / ipsq assocs stable.
8621 * The RUNNING flag reflects the state of the link.
8622 */
8623 phyint_t *phyint = ill->ill_phyint;
8624 uint64_t new_phyint_flags;
8625 boolean_t changed = B_FALSE;
8626 boolean_t went_up;
8627
8628 went_up = notify->dl_notification == DL_NOTE_LINK_UP;
8629 mutex_enter(&phyint->phyint_lock);
8630
8631 new_phyint_flags = went_up ?
8632 phyint->phyint_flags | PHYI_RUNNING :
8633 phyint->phyint_flags & ~PHYI_RUNNING;
8634
8635 if (IS_IPMP(ill)) {
8636 new_phyint_flags = went_up ?
8637 new_phyint_flags & ~PHYI_FAILED :
8638 new_phyint_flags | PHYI_FAILED;
8639 }
8640
8641 if (new_phyint_flags != phyint->phyint_flags) {
8642 phyint->phyint_flags = new_phyint_flags;
8643 changed = B_TRUE;
8644 }
8645 mutex_exit(&phyint->phyint_lock);
8646 /*
8647 * ill_restart_dad handles the DAD restart and routing
8648 * socket notification logic.
8649 */
8650 if (changed) {
8651 ill_restart_dad(phyint->phyint_illv4, went_up);
8652 ill_restart_dad(phyint->phyint_illv6, went_up);
8653 }
8654 break;
8655 }
8656 case DL_NOTE_PROMISC_ON_PHYS: {
8657 phyint_t *phyint = ill->ill_phyint;
8658
8659 mutex_enter(&phyint->phyint_lock);
8660 phyint->phyint_flags |= PHYI_PROMISC;
8661 mutex_exit(&phyint->phyint_lock);
8662 break;
8663 }
8664 case DL_NOTE_PROMISC_OFF_PHYS: {
8665 phyint_t *phyint = ill->ill_phyint;
8666
8667 mutex_enter(&phyint->phyint_lock);
8668 phyint->phyint_flags &= ~PHYI_PROMISC;
8669 mutex_exit(&phyint->phyint_lock);
8670 break;
8671 }
8672 case DL_NOTE_CAPAB_RENEG:
8673 /*
8674 * Something changed on the driver side.
8675 * It wants us to renegotiate the capabilities
8676 * on this ill. One possible cause is the aggregation
8677 * interface under us where a port got added or
8678 * went away.
8679 *
8680 * If the capability negotiation is already done
8681 * or is in progress, reset the capabilities and
8682 * mark the ill's ill_capab_reneg to be B_TRUE,
8683 * so that when the ack comes back, we can start
8684 * the renegotiation process.
8685 *
8686 * Note that if ill_capab_reneg is already B_TRUE
8687 * (ill_dlpi_capab_state is IDS_UNKNOWN in this case),
8688 * the capability resetting request has been sent
8689 * and the renegotiation has not been started yet;
8690 * nothing needs to be done in this case.
8691 */
8692 ipsq_current_start(ipsq, ill->ill_ipif, 0);
8693 ill_capability_reset(ill, B_TRUE);
8694 ipsq_current_finish(ipsq);
8695 break;
8696
8697 case DL_NOTE_ALLOWED_IPS:
8698 ill_set_allowed_ips(ill, mp);
8699 break;
8700 default:
8701 ip0dbg(("ip_rput_dlpi_writer: unknown notification "
8702 "type 0x%x for DL_NOTIFY_IND\n",
8703 notify->dl_notification));
8704 break;
8705 }
8706
8707 /*
8708 * As this is an asynchronous operation, we
8709 * should not call ill_dlpi_done
8710 */
8711 break;
8712 }
8713 case DL_NOTIFY_ACK: {
8714 dl_notify_ack_t *noteack = (dl_notify_ack_t *)mp->b_rptr;
8715
8716 if (noteack->dl_notifications & DL_NOTE_LINK_UP)
8717 ill->ill_note_link = 1;
8718 ill_dlpi_done(ill, DL_NOTIFY_REQ);
8719 break;
8720 }
8721 case DL_PHYS_ADDR_ACK: {
8722 /*
8723 * As part of plumbing the interface via SIOCSLIFNAME,
8724 * ill_dl_phys() will queue a series of DL_PHYS_ADDR_REQs,
8725 * whose answers we receive here. As each answer is received,
8726 * we call ill_dlpi_done() to dispatch the next request as
8727 * we're processing the current one. Once all answers have
8728 * been received, we use ipsq_pending_mp_get() to dequeue the
8729 * outstanding IOCTL and reply to it. (Because ill_dl_phys()
8730 * is invoked from an ill queue, conn_oper_pending_ill is not
8731 * available, but we know the ioctl is pending on ill_wq.)
8732 */
8733 uint_t paddrlen, paddroff;
8734 uint8_t *addr;
8735
8736 paddrreq = ill->ill_phys_addr_pend;
8737 paddrlen = ((dl_phys_addr_ack_t *)mp->b_rptr)->dl_addr_length;
8738 paddroff = ((dl_phys_addr_ack_t *)mp->b_rptr)->dl_addr_offset;
8739 addr = mp->b_rptr + paddroff;
8740
8741 ill_dlpi_done(ill, DL_PHYS_ADDR_REQ);
8742 if (paddrreq == DL_IPV6_TOKEN) {
8743 /*
8744 * bcopy to low-order bits of ill_token
8745 *
8746 * XXX Temporary hack - currently, all known tokens
8747 * are 64 bits, so I'll cheat for the moment.
8748 */
8749 bcopy(addr, &ill->ill_token.s6_addr32[2], paddrlen);
8750 ill->ill_token_length = paddrlen;
8751 break;
8752 } else if (paddrreq == DL_IPV6_LINK_LAYER_ADDR) {
8753 ASSERT(ill->ill_nd_lla_mp == NULL);
8754 ill_set_ndmp(ill, mp, paddroff, paddrlen);
8755 mp = NULL;
8756 break;
8757 } else if (paddrreq == DL_CURR_DEST_ADDR) {
8758 ASSERT(ill->ill_dest_addr_mp == NULL);
8759 ill->ill_dest_addr_mp = mp;
8760 ill->ill_dest_addr = addr;
8761 mp = NULL;
8762 if (ill->ill_isv6) {
8763 ill_setdesttoken(ill);
8764 ipif_setdestlinklocal(ill->ill_ipif);
8765 }
8766 break;
8767 }
8768
8769 ASSERT(paddrreq == DL_CURR_PHYS_ADDR);
8770 ASSERT(ill->ill_phys_addr_mp == NULL);
8771 if (!ill->ill_ifname_pending)
8772 break;
8773 ill->ill_ifname_pending = 0;
8774 if (!ioctl_aborted)
8775 mp1 = ipsq_pending_mp_get(ipsq, &connp);
8776 if (mp1 != NULL) {
8777 ASSERT(connp == NULL);
8778 q = ill->ill_wq;
8779 }
8780 /*
8781 * If any error acks received during the plumbing sequence,
8782 * ill_ifname_pending_err will be set. Break out and send up
8783 * the error to the pending ioctl.
8784 */
8785 if (ill->ill_ifname_pending_err != 0) {
8786 err = ill->ill_ifname_pending_err;
8787 ill->ill_ifname_pending_err = 0;
8788 break;
8789 }
8790
8791 ill->ill_phys_addr_mp = mp;
8792 ill->ill_phys_addr = (paddrlen == 0 ? NULL : addr);
8793 mp = NULL;
8794
8795 /*
8796 * If paddrlen or ill_phys_addr_length is zero, the DLPI
8797 * provider doesn't support physical addresses. We check both
8798 * paddrlen and ill_phys_addr_length because sppp (PPP) does
8799 * not have physical addresses, but historically adversises a
8800 * physical address length of 0 in its DL_INFO_ACK, but 6 in
8801 * its DL_PHYS_ADDR_ACK.
8802 */
8803 if (paddrlen == 0 || ill->ill_phys_addr_length == 0) {
8804 ill->ill_phys_addr = NULL;
8805 } else if (paddrlen != ill->ill_phys_addr_length) {
8806 ip0dbg(("DL_PHYS_ADDR_ACK: got addrlen %d, expected %d",
8807 paddrlen, ill->ill_phys_addr_length));
8808 err = EINVAL;
8809 break;
8810 }
8811
8812 if (ill->ill_nd_lla_mp == NULL) {
8813 if ((mp_hw = copyb(ill->ill_phys_addr_mp)) == NULL) {
8814 err = ENOMEM;
8815 break;
8816 }
8817 ill_set_ndmp(ill, mp_hw, paddroff, paddrlen);
8818 }
8819
8820 if (ill->ill_isv6) {
8821 ill_setdefaulttoken(ill);
8822 ipif_setlinklocal(ill->ill_ipif);
8823 }
8824 break;
8825 }
8826 case DL_OK_ACK:
8827 ip2dbg(("DL_OK_ACK %s (0x%x)\n",
8828 dl_primstr((int)dloa->dl_correct_primitive),
8829 dloa->dl_correct_primitive));
8830 DTRACE_PROBE3(ill__dlpi, char *, "ip_rput_dlpi_writer ok",
8831 char *, dl_primstr(dloa->dl_correct_primitive),
8832 ill_t *, ill);
8833
8834 switch (dloa->dl_correct_primitive) {
8835 case DL_ENABMULTI_REQ:
8836 case DL_DISABMULTI_REQ:
8837 ill_dlpi_done(ill, dloa->dl_correct_primitive);
8838 break;
8839 case DL_PROMISCON_REQ:
8840 case DL_PROMISCOFF_REQ:
8841 case DL_UNBIND_REQ:
8842 case DL_ATTACH_REQ:
8843 ill_dlpi_done(ill, dloa->dl_correct_primitive);
8844 break;
8845 }
8846 break;
8847 default:
8848 break;
8849 }
8850
8851 freemsg(mp);
8852 if (mp1 == NULL)
8853 return;
8854
8855 /*
8856 * The operation must complete without EINPROGRESS since
8857 * ipsq_pending_mp_get() has removed the mblk (mp1). Otherwise,
8858 * the operation will be stuck forever inside the IPSQ.
8859 */
8860 ASSERT(err != EINPROGRESS);
8861
8862 DTRACE_PROBE4(ipif__ioctl, char *, "ip_rput_dlpi_writer finish",
8863 int, ipsq->ipsq_xop->ipx_current_ioctl, ill_t *, ill,
8864 ipif_t *, NULL);
8865
8866 switch (ipsq->ipsq_xop->ipx_current_ioctl) {
8867 case 0:
8868 ipsq_current_finish(ipsq);
8869 break;
8870
8871 case SIOCSLIFNAME:
8872 case IF_UNITSEL: {
8873 ill_t *ill_other = ILL_OTHER(ill);
8874
8875 /*
8876 * If SIOCSLIFNAME or IF_UNITSEL is about to succeed, and the
8877 * ill has a peer which is in an IPMP group, then place ill
8878 * into the same group. One catch: although ifconfig plumbs
8879 * the appropriate IPMP meta-interface prior to plumbing this
8880 * ill, it is possible for multiple ifconfig applications to
8881 * race (or for another application to adjust plumbing), in
8882 * which case the IPMP meta-interface we need will be missing.
8883 * If so, kick the phyint out of the group.
8884 */
8885 if (err == 0 && ill_other != NULL && IS_UNDER_IPMP(ill_other)) {
8886 ipmp_grp_t *grp = ill->ill_phyint->phyint_grp;
8887 ipmp_illgrp_t *illg;
8888
8889 illg = ill->ill_isv6 ? grp->gr_v6 : grp->gr_v4;
8890 if (illg == NULL)
8891 ipmp_phyint_leave_grp(ill->ill_phyint);
8892 else
8893 ipmp_ill_join_illgrp(ill, illg);
8894 }
8895
8896 if (ipsq->ipsq_xop->ipx_current_ioctl == IF_UNITSEL)
8897 ip_ioctl_finish(q, mp1, err, NO_COPYOUT, ipsq);
8898 else
8899 ip_ioctl_finish(q, mp1, err, COPYOUT, ipsq);
8900 break;
8901 }
8902 case SIOCLIFADDIF:
8903 ip_ioctl_finish(q, mp1, err, COPYOUT, ipsq);
8904 break;
8905
8906 default:
8907 ip_ioctl_finish(q, mp1, err, NO_COPYOUT, ipsq);
8908 break;
8909 }
8910 }
8911
8912 /*
8913 * ip_rput_other is called by ip_rput to handle messages modifying the global
8914 * state in IP. If 'ipsq' is non-NULL, caller is writer on it.
8915 */
8916 /* ARGSUSED */
8917 void
8918 ip_rput_other(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
8919 {
8920 ill_t *ill = q->q_ptr;
8921 struct iocblk *iocp;
8922
8923 ip1dbg(("ip_rput_other "));
8924 if (ipsq != NULL) {
8925 ASSERT(IAM_WRITER_IPSQ(ipsq));
8926 ASSERT(ipsq->ipsq_xop ==
8927 ill->ill_phyint->phyint_ipsq->ipsq_xop);
8928 }
8929
8930 switch (mp->b_datap->db_type) {
8931 case M_ERROR:
8932 case M_HANGUP:
8933 /*
8934 * The device has a problem. We force the ILL down. It can
8935 * be brought up again manually using SIOCSIFFLAGS (via
8936 * ifconfig or equivalent).
8937 */
8938 ASSERT(ipsq != NULL);
8939 if (mp->b_rptr < mp->b_wptr)
8940 ill->ill_error = (int)(*mp->b_rptr & 0xFF);
8941 if (ill->ill_error == 0)
8942 ill->ill_error = ENXIO;
8943 if (!ill_down_start(q, mp))
8944 return;
8945 ipif_all_down_tail(ipsq, q, mp, NULL);
8946 break;
8947 case M_IOCNAK: {
8948 iocp = (struct iocblk *)mp->b_rptr;
8949
8950 ASSERT(iocp->ioc_cmd == DL_IOC_HDR_INFO);
8951 /*
8952 * If this was the first attempt, turn off the fastpath
8953 * probing.
8954 */
8955 mutex_enter(&ill->ill_lock);
8956 if (ill->ill_dlpi_fastpath_state == IDS_INPROGRESS) {
8957 ill->ill_dlpi_fastpath_state = IDS_FAILED;
8958 mutex_exit(&ill->ill_lock);
8959 /*
8960 * don't flush the nce_t entries: we use them
8961 * as an index to the ncec itself.
8962 */
8963 ip1dbg(("ip_rput: DLPI fastpath off on interface %s\n",
8964 ill->ill_name));
8965 } else {
8966 mutex_exit(&ill->ill_lock);
8967 }
8968 freemsg(mp);
8969 break;
8970 }
8971 default:
8972 ASSERT(0);
8973 break;
8974 }
8975 }
8976
8977 /*
8978 * Update any source route, record route or timestamp options
8979 * When it fails it has consumed the message and BUMPed the MIB.
8980 */
8981 boolean_t
8982 ip_forward_options(mblk_t *mp, ipha_t *ipha, ill_t *dst_ill,
8983 ip_recv_attr_t *ira)
8984 {
8985 ipoptp_t opts;
8986 uchar_t *opt;
8987 uint8_t optval;
8988 uint8_t optlen;
8989 ipaddr_t dst;
8990 ipaddr_t ifaddr;
8991 uint32_t ts;
8992 timestruc_t now;
8993 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
8994
8995 ip2dbg(("ip_forward_options\n"));
8996 dst = ipha->ipha_dst;
8997 for (optval = ipoptp_first(&opts, ipha);
8998 optval != IPOPT_EOL;
8999 optval = ipoptp_next(&opts)) {
9000 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
9001 opt = opts.ipoptp_cur;
9002 optlen = opts.ipoptp_len;
9003 ip2dbg(("ip_forward_options: opt %d, len %d\n",
9004 optval, opts.ipoptp_len));
9005 switch (optval) {
9006 uint32_t off;
9007 case IPOPT_SSRR:
9008 case IPOPT_LSRR:
9009 /* Check if adminstratively disabled */
9010 if (!ipst->ips_ip_forward_src_routed) {
9011 BUMP_MIB(dst_ill->ill_ip_mib,
9012 ipIfStatsForwProhibits);
9013 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED",
9014 mp, dst_ill);
9015 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED,
9016 ira);
9017 return (B_FALSE);
9018 }
9019 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9020 /*
9021 * Must be partial since ip_input_options
9022 * checked for strict.
9023 */
9024 break;
9025 }
9026 off = opt[IPOPT_OFFSET];
9027 off--;
9028 redo_srr:
9029 if (optlen < IP_ADDR_LEN ||
9030 off > optlen - IP_ADDR_LEN) {
9031 /* End of source route */
9032 ip1dbg((
9033 "ip_forward_options: end of SR\n"));
9034 break;
9035 }
9036 /* Pick a reasonable address on the outbound if */
9037 ASSERT(dst_ill != NULL);
9038 if (ip_select_source_v4(dst_ill, INADDR_ANY, dst,
9039 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9040 NULL) != 0) {
9041 /* No source! Shouldn't happen */
9042 ifaddr = INADDR_ANY;
9043 }
9044 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9045 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9046 ip1dbg(("ip_forward_options: next hop 0x%x\n",
9047 ntohl(dst)));
9048
9049 /*
9050 * Check if our address is present more than
9051 * once as consecutive hops in source route.
9052 */
9053 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
9054 off += IP_ADDR_LEN;
9055 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9056 goto redo_srr;
9057 }
9058 ipha->ipha_dst = dst;
9059 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9060 break;
9061 case IPOPT_RR:
9062 off = opt[IPOPT_OFFSET];
9063 off--;
9064 if (optlen < IP_ADDR_LEN ||
9065 off > optlen - IP_ADDR_LEN) {
9066 /* No more room - ignore */
9067 ip1dbg((
9068 "ip_forward_options: end of RR\n"));
9069 break;
9070 }
9071 /* Pick a reasonable address on the outbound if */
9072 ASSERT(dst_ill != NULL);
9073 if (ip_select_source_v4(dst_ill, INADDR_ANY, dst,
9074 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9075 NULL) != 0) {
9076 /* No source! Shouldn't happen */
9077 ifaddr = INADDR_ANY;
9078 }
9079 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9080 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9081 break;
9082 case IPOPT_TS:
9083 /* Insert timestamp if there is room */
9084 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9085 case IPOPT_TS_TSONLY:
9086 off = IPOPT_TS_TIMELEN;
9087 break;
9088 case IPOPT_TS_PRESPEC:
9089 case IPOPT_TS_PRESPEC_RFC791:
9090 /* Verify that the address matched */
9091 off = opt[IPOPT_OFFSET] - 1;
9092 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9093 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9094 /* Not for us */
9095 break;
9096 }
9097 /* FALLTHRU */
9098 case IPOPT_TS_TSANDADDR:
9099 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9100 break;
9101 default:
9102 /*
9103 * ip_*put_options should have already
9104 * dropped this packet.
9105 */
9106 cmn_err(CE_PANIC, "ip_forward_options: "
9107 "unknown IT - bug in ip_input_options?\n");
9108 return (B_TRUE); /* Keep "lint" happy */
9109 }
9110 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
9111 /* Increase overflow counter */
9112 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
9113 opt[IPOPT_POS_OV_FLG] =
9114 (uint8_t)((opt[IPOPT_POS_OV_FLG] & 0x0F) |
9115 (off << 4));
9116 break;
9117 }
9118 off = opt[IPOPT_OFFSET] - 1;
9119 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9120 case IPOPT_TS_PRESPEC:
9121 case IPOPT_TS_PRESPEC_RFC791:
9122 case IPOPT_TS_TSANDADDR:
9123 /* Pick a reasonable addr on the outbound if */
9124 ASSERT(dst_ill != NULL);
9125 if (ip_select_source_v4(dst_ill, INADDR_ANY,
9126 dst, INADDR_ANY, ALL_ZONES, ipst, &ifaddr,
9127 NULL, NULL) != 0) {
9128 /* No source! Shouldn't happen */
9129 ifaddr = INADDR_ANY;
9130 }
9131 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9132 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9133 /* FALLTHRU */
9134 case IPOPT_TS_TSONLY:
9135 off = opt[IPOPT_OFFSET] - 1;
9136 /* Compute # of milliseconds since midnight */
9137 gethrestime(&now);
9138 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
9139 NSEC2MSEC(now.tv_nsec);
9140 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
9141 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
9142 break;
9143 }
9144 break;
9145 }
9146 }
9147 return (B_TRUE);
9148 }
9149
9150 /*
9151 * Call ill_frag_timeout to do garbage collection. ill_frag_timeout
9152 * returns 'true' if there are still fragments left on the queue, in
9153 * which case we restart the timer.
9154 */
9155 void
9156 ill_frag_timer(void *arg)
9157 {
9158 ill_t *ill = (ill_t *)arg;
9159 boolean_t frag_pending;
9160 ip_stack_t *ipst = ill->ill_ipst;
9161 time_t timeout;
9162
9163 mutex_enter(&ill->ill_lock);
9164 ASSERT(!ill->ill_fragtimer_executing);
9165 if (ill->ill_state_flags & ILL_CONDEMNED) {
9166 ill->ill_frag_timer_id = 0;
9167 mutex_exit(&ill->ill_lock);
9168 return;
9169 }
9170 ill->ill_fragtimer_executing = 1;
9171 mutex_exit(&ill->ill_lock);
9172
9173 timeout = (ill->ill_isv6 ? ipst->ips_ipv6_reassembly_timeout :
9174 ipst->ips_ip_reassembly_timeout);
9175
9176 frag_pending = ill_frag_timeout(ill, timeout);
9177
9178 /*
9179 * Restart the timer, if we have fragments pending or if someone
9180 * wanted us to be scheduled again.
9181 */
9182 mutex_enter(&ill->ill_lock);
9183 ill->ill_fragtimer_executing = 0;
9184 ill->ill_frag_timer_id = 0;
9185 if (frag_pending || ill->ill_fragtimer_needrestart)
9186 ill_frag_timer_start(ill);
9187 mutex_exit(&ill->ill_lock);
9188 }
9189
9190 void
9191 ill_frag_timer_start(ill_t *ill)
9192 {
9193 ip_stack_t *ipst = ill->ill_ipst;
9194 clock_t timeo_ms;
9195
9196 ASSERT(MUTEX_HELD(&ill->ill_lock));
9197
9198 /* If the ill is closing or opening don't proceed */
9199 if (ill->ill_state_flags & ILL_CONDEMNED)
9200 return;
9201
9202 if (ill->ill_fragtimer_executing) {
9203 /*
9204 * ill_frag_timer is currently executing. Just record the
9205 * the fact that we want the timer to be restarted.
9206 * ill_frag_timer will post a timeout before it returns,
9207 * ensuring it will be called again.
9208 */
9209 ill->ill_fragtimer_needrestart = 1;
9210 return;
9211 }
9212
9213 if (ill->ill_frag_timer_id == 0) {
9214 timeo_ms = (ill->ill_isv6 ? ipst->ips_ipv6_reassembly_timeout :
9215 ipst->ips_ip_reassembly_timeout) * SECONDS;
9216
9217 /*
9218 * The timer is neither running nor is the timeout handler
9219 * executing. Post a timeout so that ill_frag_timer will be
9220 * called
9221 */
9222 ill->ill_frag_timer_id = timeout(ill_frag_timer, ill,
9223 MSEC_TO_TICK(timeo_ms >> 1));
9224 ill->ill_fragtimer_needrestart = 0;
9225 }
9226 }
9227
9228 /*
9229 * Update any source route, record route or timestamp options.
9230 * Check that we are at end of strict source route.
9231 * The options have already been checked for sanity in ip_input_options().
9232 */
9233 boolean_t
9234 ip_input_local_options(mblk_t *mp, ipha_t *ipha, ip_recv_attr_t *ira)
9235 {
9236 ipoptp_t opts;
9237 uchar_t *opt;
9238 uint8_t optval;
9239 uint8_t optlen;
9240 ipaddr_t dst;
9241 ipaddr_t ifaddr;
9242 uint32_t ts;
9243 timestruc_t now;
9244 ill_t *ill = ira->ira_ill;
9245 ip_stack_t *ipst = ill->ill_ipst;
9246
9247 ip2dbg(("ip_input_local_options\n"));
9248
9249 for (optval = ipoptp_first(&opts, ipha);
9250 optval != IPOPT_EOL;
9251 optval = ipoptp_next(&opts)) {
9252 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
9253 opt = opts.ipoptp_cur;
9254 optlen = opts.ipoptp_len;
9255 ip2dbg(("ip_input_local_options: opt %d, len %d\n",
9256 optval, optlen));
9257 switch (optval) {
9258 uint32_t off;
9259 case IPOPT_SSRR:
9260 case IPOPT_LSRR:
9261 off = opt[IPOPT_OFFSET];
9262 off--;
9263 if (optlen < IP_ADDR_LEN ||
9264 off > optlen - IP_ADDR_LEN) {
9265 /* End of source route */
9266 ip1dbg(("ip_input_local_options: end of SR\n"));
9267 break;
9268 }
9269 /*
9270 * This will only happen if two consecutive entries
9271 * in the source route contains our address or if
9272 * it is a packet with a loose source route which
9273 * reaches us before consuming the whole source route
9274 */
9275 ip1dbg(("ip_input_local_options: not end of SR\n"));
9276 if (optval == IPOPT_SSRR) {
9277 goto bad_src_route;
9278 }
9279 /*
9280 * Hack: instead of dropping the packet truncate the
9281 * source route to what has been used by filling the
9282 * rest with IPOPT_NOP.
9283 */
9284 opt[IPOPT_OLEN] = (uint8_t)off;
9285 while (off < optlen) {
9286 opt[off++] = IPOPT_NOP;
9287 }
9288 break;
9289 case IPOPT_RR:
9290 off = opt[IPOPT_OFFSET];
9291 off--;
9292 if (optlen < IP_ADDR_LEN ||
9293 off > optlen - IP_ADDR_LEN) {
9294 /* No more room - ignore */
9295 ip1dbg((
9296 "ip_input_local_options: end of RR\n"));
9297 break;
9298 }
9299 /* Pick a reasonable address on the outbound if */
9300 if (ip_select_source_v4(ill, INADDR_ANY, ipha->ipha_dst,
9301 INADDR_ANY, ALL_ZONES, ipst, &ifaddr, NULL,
9302 NULL) != 0) {
9303 /* No source! Shouldn't happen */
9304 ifaddr = INADDR_ANY;
9305 }
9306 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9307 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9308 break;
9309 case IPOPT_TS:
9310 /* Insert timestamp if there is romm */
9311 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9312 case IPOPT_TS_TSONLY:
9313 off = IPOPT_TS_TIMELEN;
9314 break;
9315 case IPOPT_TS_PRESPEC:
9316 case IPOPT_TS_PRESPEC_RFC791:
9317 /* Verify that the address matched */
9318 off = opt[IPOPT_OFFSET] - 1;
9319 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9320 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9321 /* Not for us */
9322 break;
9323 }
9324 /* FALLTHRU */
9325 case IPOPT_TS_TSANDADDR:
9326 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9327 break;
9328 default:
9329 /*
9330 * ip_*put_options should have already
9331 * dropped this packet.
9332 */
9333 cmn_err(CE_PANIC, "ip_input_local_options: "
9334 "unknown IT - bug in ip_input_options?\n");
9335 return (B_TRUE); /* Keep "lint" happy */
9336 }
9337 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
9338 /* Increase overflow counter */
9339 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
9340 opt[IPOPT_POS_OV_FLG] =
9341 (uint8_t)((opt[IPOPT_POS_OV_FLG] & 0x0F) |
9342 (off << 4));
9343 break;
9344 }
9345 off = opt[IPOPT_OFFSET] - 1;
9346 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9347 case IPOPT_TS_PRESPEC:
9348 case IPOPT_TS_PRESPEC_RFC791:
9349 case IPOPT_TS_TSANDADDR:
9350 /* Pick a reasonable addr on the outbound if */
9351 if (ip_select_source_v4(ill, INADDR_ANY,
9352 ipha->ipha_dst, INADDR_ANY, ALL_ZONES, ipst,
9353 &ifaddr, NULL, NULL) != 0) {
9354 /* No source! Shouldn't happen */
9355 ifaddr = INADDR_ANY;
9356 }
9357 bcopy(&ifaddr, (char *)opt + off, IP_ADDR_LEN);
9358 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
9359 /* FALLTHRU */
9360 case IPOPT_TS_TSONLY:
9361 off = opt[IPOPT_OFFSET] - 1;
9362 /* Compute # of milliseconds since midnight */
9363 gethrestime(&now);
9364 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
9365 NSEC2MSEC(now.tv_nsec);
9366 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
9367 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
9368 break;
9369 }
9370 break;
9371 }
9372 }
9373 return (B_TRUE);
9374
9375 bad_src_route:
9376 /* make sure we clear any indication of a hardware checksum */
9377 DB_CKSUMFLAGS(mp) = 0;
9378 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ill);
9379 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
9380 return (B_FALSE);
9381
9382 }
9383
9384 /*
9385 * Process IP options in an inbound packet. Always returns the nexthop.
9386 * Normally this is the passed in nexthop, but if there is an option
9387 * that effects the nexthop (such as a source route) that will be returned.
9388 * Sets *errorp if there is an error, in which case an ICMP error has been sent
9389 * and mp freed.
9390 */
9391 ipaddr_t
9392 ip_input_options(ipha_t *ipha, ipaddr_t dst, mblk_t *mp,
9393 ip_recv_attr_t *ira, int *errorp)
9394 {
9395 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
9396 ipoptp_t opts;
9397 uchar_t *opt;
9398 uint8_t optval;
9399 uint8_t optlen;
9400 intptr_t code = 0;
9401 ire_t *ire;
9402
9403 ip2dbg(("ip_input_options\n"));
9404 *errorp = 0;
9405 for (optval = ipoptp_first(&opts, ipha);
9406 optval != IPOPT_EOL;
9407 optval = ipoptp_next(&opts)) {
9408 opt = opts.ipoptp_cur;
9409 optlen = opts.ipoptp_len;
9410 ip2dbg(("ip_input_options: opt %d, len %d\n",
9411 optval, optlen));
9412 /*
9413 * Note: we need to verify the checksum before we
9414 * modify anything thus this routine only extracts the next
9415 * hop dst from any source route.
9416 */
9417 switch (optval) {
9418 uint32_t off;
9419 case IPOPT_SSRR:
9420 case IPOPT_LSRR:
9421 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
9422 if (optval == IPOPT_SSRR) {
9423 ip1dbg(("ip_input_options: not next"
9424 " strict source route 0x%x\n",
9425 ntohl(dst)));
9426 code = (char *)&ipha->ipha_dst -
9427 (char *)ipha;
9428 goto param_prob; /* RouterReq's */
9429 }
9430 ip2dbg(("ip_input_options: "
9431 "not next source route 0x%x\n",
9432 ntohl(dst)));
9433 break;
9434 }
9435
9436 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9437 ip1dbg((
9438 "ip_input_options: bad option offset\n"));
9439 code = (char *)&opt[IPOPT_OLEN] -
9440 (char *)ipha;
9441 goto param_prob;
9442 }
9443 off = opt[IPOPT_OFFSET];
9444 off--;
9445 redo_srr:
9446 if (optlen < IP_ADDR_LEN ||
9447 off > optlen - IP_ADDR_LEN) {
9448 /* End of source route */
9449 ip1dbg(("ip_input_options: end of SR\n"));
9450 break;
9451 }
9452 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
9453 ip1dbg(("ip_input_options: next hop 0x%x\n",
9454 ntohl(dst)));
9455
9456 /*
9457 * Check if our address is present more than
9458 * once as consecutive hops in source route.
9459 * XXX verify per-interface ip_forwarding
9460 * for source route?
9461 */
9462 if (ip_type_v4(dst, ipst) == IRE_LOCAL) {
9463 off += IP_ADDR_LEN;
9464 goto redo_srr;
9465 }
9466
9467 if (dst == htonl(INADDR_LOOPBACK)) {
9468 ip1dbg(("ip_input_options: loopback addr in "
9469 "source route!\n"));
9470 goto bad_src_route;
9471 }
9472 /*
9473 * For strict: verify that dst is directly
9474 * reachable.
9475 */
9476 if (optval == IPOPT_SSRR) {
9477 ire = ire_ftable_lookup_v4(dst, 0, 0,
9478 IRE_INTERFACE, NULL, ALL_ZONES,
9479 ira->ira_tsl,
9480 MATCH_IRE_TYPE | MATCH_IRE_SECATTR, 0, ipst,
9481 NULL);
9482 if (ire == NULL) {
9483 ip1dbg(("ip_input_options: SSRR not "
9484 "directly reachable: 0x%x\n",
9485 ntohl(dst)));
9486 goto bad_src_route;
9487 }
9488 ire_refrele(ire);
9489 }
9490 /*
9491 * Defer update of the offset and the record route
9492 * until the packet is forwarded.
9493 */
9494 break;
9495 case IPOPT_RR:
9496 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9497 ip1dbg((
9498 "ip_input_options: bad option offset\n"));
9499 code = (char *)&opt[IPOPT_OLEN] -
9500 (char *)ipha;
9501 goto param_prob;
9502 }
9503 break;
9504 case IPOPT_TS:
9505 /*
9506 * Verify that length >= 5 and that there is either
9507 * room for another timestamp or that the overflow
9508 * counter is not maxed out.
9509 */
9510 code = (char *)&opt[IPOPT_OLEN] - (char *)ipha;
9511 if (optlen < IPOPT_MINLEN_IT) {
9512 goto param_prob;
9513 }
9514 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
9515 ip1dbg((
9516 "ip_input_options: bad option offset\n"));
9517 code = (char *)&opt[IPOPT_OFFSET] -
9518 (char *)ipha;
9519 goto param_prob;
9520 }
9521 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
9522 case IPOPT_TS_TSONLY:
9523 off = IPOPT_TS_TIMELEN;
9524 break;
9525 case IPOPT_TS_TSANDADDR:
9526 case IPOPT_TS_PRESPEC:
9527 case IPOPT_TS_PRESPEC_RFC791:
9528 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
9529 break;
9530 default:
9531 code = (char *)&opt[IPOPT_POS_OV_FLG] -
9532 (char *)ipha;
9533 goto param_prob;
9534 }
9535 if (opt[IPOPT_OFFSET] - 1 + off > optlen &&
9536 (opt[IPOPT_POS_OV_FLG] & 0xF0) == 0xF0) {
9537 /*
9538 * No room and the overflow counter is 15
9539 * already.
9540 */
9541 goto param_prob;
9542 }
9543 break;
9544 }
9545 }
9546
9547 if ((opts.ipoptp_flags & IPOPTP_ERROR) == 0) {
9548 return (dst);
9549 }
9550
9551 ip1dbg(("ip_input_options: error processing IP options."));
9552 code = (char *)&opt[IPOPT_OFFSET] - (char *)ipha;
9553
9554 param_prob:
9555 /* make sure we clear any indication of a hardware checksum */
9556 DB_CKSUMFLAGS(mp) = 0;
9557 ip_drop_input("ICMP_PARAM_PROBLEM", mp, ira->ira_ill);
9558 icmp_param_problem(mp, (uint8_t)code, ira);
9559 *errorp = -1;
9560 return (dst);
9561
9562 bad_src_route:
9563 /* make sure we clear any indication of a hardware checksum */
9564 DB_CKSUMFLAGS(mp) = 0;
9565 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ira->ira_ill);
9566 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, ira);
9567 *errorp = -1;
9568 return (dst);
9569 }
9570
9571 /*
9572 * IP & ICMP info in >=14 msg's ...
9573 * - ip fixed part (mib2_ip_t)
9574 * - icmp fixed part (mib2_icmp_t)
9575 * - ipAddrEntryTable (ip 20) all IPv4 ipifs
9576 * - ipRouteEntryTable (ip 21) all IPv4 IREs
9577 * - ipNetToMediaEntryTable (ip 22) all IPv4 Neighbor Cache entries
9578 * - ipRouteAttributeTable (ip 102) labeled routes
9579 * - ip multicast membership (ip_member_t)
9580 * - ip multicast source filtering (ip_grpsrc_t)
9581 * - igmp fixed part (struct igmpstat)
9582 * - multicast routing stats (struct mrtstat)
9583 * - multicast routing vifs (array of struct vifctl)
9584 * - multicast routing routes (array of struct mfcctl)
9585 * - ip6 fixed part (mib2_ipv6IfStatsEntry_t)
9586 * One per ill plus one generic
9587 * - icmp6 fixed part (mib2_ipv6IfIcmpEntry_t)
9588 * One per ill plus one generic
9589 * - ipv6RouteEntry all IPv6 IREs
9590 * - ipv6RouteAttributeTable (ip6 102) labeled routes
9591 * - ipv6NetToMediaEntry all IPv6 Neighbor Cache entries
9592 * - ipv6AddrEntry all IPv6 ipifs
9593 * - ipv6 multicast membership (ipv6_member_t)
9594 * - ipv6 multicast source filtering (ipv6_grpsrc_t)
9595 *
9596 * NOTE: original mpctl is copied for msg's 2..N, since its ctl part is
9597 * already filled in by the caller.
9598 * If legacy_req is true then MIB structures needs to be truncated to their
9599 * legacy sizes before being returned.
9600 * Return value of 0 indicates that no messages were sent and caller
9601 * should free mpctl.
9602 */
9603 int
9604 ip_snmp_get(queue_t *q, mblk_t *mpctl, int level, boolean_t legacy_req)
9605 {
9606 ip_stack_t *ipst;
9607 sctp_stack_t *sctps;
9608
9609 if (q->q_next != NULL) {
9610 ipst = ILLQ_TO_IPST(q);
9611 } else {
9612 ipst = CONNQ_TO_IPST(q);
9613 }
9614 ASSERT(ipst != NULL);
9615 sctps = ipst->ips_netstack->netstack_sctp;
9616
9617 if (mpctl == NULL || mpctl->b_cont == NULL) {
9618 return (0);
9619 }
9620
9621 /*
9622 * For the purposes of the (broken) packet shell use
9623 * of the level we make sure MIB2_TCP/MIB2_UDP can be used
9624 * to make TCP and UDP appear first in the list of mib items.
9625 * TBD: We could expand this and use it in netstat so that
9626 * the kernel doesn't have to produce large tables (connections,
9627 * routes, etc) when netstat only wants the statistics or a particular
9628 * table.
9629 */
9630 if (!(level == MIB2_TCP || level == MIB2_UDP)) {
9631 if ((mpctl = icmp_snmp_get(q, mpctl)) == NULL) {
9632 return (1);
9633 }
9634 }
9635
9636 if (level != MIB2_TCP) {
9637 if ((mpctl = udp_snmp_get(q, mpctl, legacy_req)) == NULL) {
9638 return (1);
9639 }
9640 }
9641
9642 if (level != MIB2_UDP) {
9643 if ((mpctl = tcp_snmp_get(q, mpctl, legacy_req)) == NULL) {
9644 return (1);
9645 }
9646 }
9647
9648 if ((mpctl = ip_snmp_get_mib2_ip_traffic_stats(q, mpctl,
9649 ipst, legacy_req)) == NULL) {
9650 return (1);
9651 }
9652
9653 if ((mpctl = ip_snmp_get_mib2_ip6(q, mpctl, ipst,
9654 legacy_req)) == NULL) {
9655 return (1);
9656 }
9657
9658 if ((mpctl = ip_snmp_get_mib2_icmp(q, mpctl, ipst)) == NULL) {
9659 return (1);
9660 }
9661
9662 if ((mpctl = ip_snmp_get_mib2_icmp6(q, mpctl, ipst)) == NULL) {
9663 return (1);
9664 }
9665
9666 if ((mpctl = ip_snmp_get_mib2_igmp(q, mpctl, ipst)) == NULL) {
9667 return (1);
9668 }
9669
9670 if ((mpctl = ip_snmp_get_mib2_multi(q, mpctl, ipst)) == NULL) {
9671 return (1);
9672 }
9673
9674 if ((mpctl = ip_snmp_get_mib2_ip_addr(q, mpctl, ipst,
9675 legacy_req)) == NULL) {
9676 return (1);
9677 }
9678
9679 if ((mpctl = ip_snmp_get_mib2_ip6_addr(q, mpctl, ipst,
9680 legacy_req)) == NULL) {
9681 return (1);
9682 }
9683
9684 if ((mpctl = ip_snmp_get_mib2_ip_group_mem(q, mpctl, ipst)) == NULL) {
9685 return (1);
9686 }
9687
9688 if ((mpctl = ip_snmp_get_mib2_ip6_group_mem(q, mpctl, ipst)) == NULL) {
9689 return (1);
9690 }
9691
9692 if ((mpctl = ip_snmp_get_mib2_ip_group_src(q, mpctl, ipst)) == NULL) {
9693 return (1);
9694 }
9695
9696 if ((mpctl = ip_snmp_get_mib2_ip6_group_src(q, mpctl, ipst)) == NULL) {
9697 return (1);
9698 }
9699
9700 if ((mpctl = ip_snmp_get_mib2_virt_multi(q, mpctl, ipst)) == NULL) {
9701 return (1);
9702 }
9703
9704 if ((mpctl = ip_snmp_get_mib2_multi_rtable(q, mpctl, ipst)) == NULL) {
9705 return (1);
9706 }
9707
9708 mpctl = ip_snmp_get_mib2_ip_route_media(q, mpctl, level, ipst);
9709 if (mpctl == NULL)
9710 return (1);
9711
9712 mpctl = ip_snmp_get_mib2_ip6_route_media(q, mpctl, level, ipst);
9713 if (mpctl == NULL)
9714 return (1);
9715
9716 if ((mpctl = sctp_snmp_get_mib2(q, mpctl, sctps)) == NULL) {
9717 return (1);
9718 }
9719 if ((mpctl = ip_snmp_get_mib2_ip_dce(q, mpctl, ipst)) == NULL) {
9720 return (1);
9721 }
9722 freemsg(mpctl);
9723 return (1);
9724 }
9725
9726 /* Get global (legacy) IPv4 statistics */
9727 static mblk_t *
9728 ip_snmp_get_mib2_ip(queue_t *q, mblk_t *mpctl, mib2_ipIfStatsEntry_t *ipmib,
9729 ip_stack_t *ipst, boolean_t legacy_req)
9730 {
9731 mib2_ip_t old_ip_mib;
9732 struct opthdr *optp;
9733 mblk_t *mp2ctl;
9734 mib2_ipAddrEntry_t mae;
9735
9736 /*
9737 * make a copy of the original message
9738 */
9739 mp2ctl = copymsg(mpctl);
9740
9741 /* fixed length IP structure... */
9742 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9743 optp->level = MIB2_IP;
9744 optp->name = 0;
9745 SET_MIB(old_ip_mib.ipForwarding,
9746 (WE_ARE_FORWARDING(ipst) ? 1 : 2));
9747 SET_MIB(old_ip_mib.ipDefaultTTL,
9748 (uint32_t)ipst->ips_ip_def_ttl);
9749 SET_MIB(old_ip_mib.ipReasmTimeout,
9750 ipst->ips_ip_reassembly_timeout);
9751 SET_MIB(old_ip_mib.ipAddrEntrySize,
9752 (legacy_req) ? LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t) :
9753 sizeof (mib2_ipAddrEntry_t));
9754 SET_MIB(old_ip_mib.ipRouteEntrySize,
9755 sizeof (mib2_ipRouteEntry_t));
9756 SET_MIB(old_ip_mib.ipNetToMediaEntrySize,
9757 sizeof (mib2_ipNetToMediaEntry_t));
9758 SET_MIB(old_ip_mib.ipMemberEntrySize, sizeof (ip_member_t));
9759 SET_MIB(old_ip_mib.ipGroupSourceEntrySize, sizeof (ip_grpsrc_t));
9760 SET_MIB(old_ip_mib.ipRouteAttributeSize,
9761 sizeof (mib2_ipAttributeEntry_t));
9762 SET_MIB(old_ip_mib.transportMLPSize, sizeof (mib2_transportMLPEntry_t));
9763 SET_MIB(old_ip_mib.ipDestEntrySize, sizeof (dest_cache_entry_t));
9764
9765 /*
9766 * Grab the statistics from the new IP MIB
9767 */
9768 SET_MIB(old_ip_mib.ipInReceives,
9769 (uint32_t)ipmib->ipIfStatsHCInReceives);
9770 SET_MIB(old_ip_mib.ipInHdrErrors, ipmib->ipIfStatsInHdrErrors);
9771 SET_MIB(old_ip_mib.ipInAddrErrors, ipmib->ipIfStatsInAddrErrors);
9772 SET_MIB(old_ip_mib.ipForwDatagrams,
9773 (uint32_t)ipmib->ipIfStatsHCOutForwDatagrams);
9774 SET_MIB(old_ip_mib.ipInUnknownProtos,
9775 ipmib->ipIfStatsInUnknownProtos);
9776 SET_MIB(old_ip_mib.ipInDiscards, ipmib->ipIfStatsInDiscards);
9777 SET_MIB(old_ip_mib.ipInDelivers,
9778 (uint32_t)ipmib->ipIfStatsHCInDelivers);
9779 SET_MIB(old_ip_mib.ipOutRequests,
9780 (uint32_t)ipmib->ipIfStatsHCOutRequests);
9781 SET_MIB(old_ip_mib.ipOutDiscards, ipmib->ipIfStatsOutDiscards);
9782 SET_MIB(old_ip_mib.ipOutNoRoutes, ipmib->ipIfStatsOutNoRoutes);
9783 SET_MIB(old_ip_mib.ipReasmReqds, ipmib->ipIfStatsReasmReqds);
9784 SET_MIB(old_ip_mib.ipReasmOKs, ipmib->ipIfStatsReasmOKs);
9785 SET_MIB(old_ip_mib.ipReasmFails, ipmib->ipIfStatsReasmFails);
9786 SET_MIB(old_ip_mib.ipFragOKs, ipmib->ipIfStatsOutFragOKs);
9787 SET_MIB(old_ip_mib.ipFragFails, ipmib->ipIfStatsOutFragFails);
9788 SET_MIB(old_ip_mib.ipFragCreates, ipmib->ipIfStatsOutFragCreates);
9789
9790 /* ipRoutingDiscards is not being used */
9791 SET_MIB(old_ip_mib.ipRoutingDiscards, 0);
9792 SET_MIB(old_ip_mib.tcpInErrs, ipmib->tcpIfStatsInErrs);
9793 SET_MIB(old_ip_mib.udpNoPorts, ipmib->udpIfStatsNoPorts);
9794 SET_MIB(old_ip_mib.ipInCksumErrs, ipmib->ipIfStatsInCksumErrs);
9795 SET_MIB(old_ip_mib.ipReasmDuplicates,
9796 ipmib->ipIfStatsReasmDuplicates);
9797 SET_MIB(old_ip_mib.ipReasmPartDups, ipmib->ipIfStatsReasmPartDups);
9798 SET_MIB(old_ip_mib.ipForwProhibits, ipmib->ipIfStatsForwProhibits);
9799 SET_MIB(old_ip_mib.udpInCksumErrs, ipmib->udpIfStatsInCksumErrs);
9800 SET_MIB(old_ip_mib.udpInOverflows, ipmib->udpIfStatsInOverflows);
9801 SET_MIB(old_ip_mib.rawipInOverflows,
9802 ipmib->rawipIfStatsInOverflows);
9803
9804 SET_MIB(old_ip_mib.ipsecInSucceeded, ipmib->ipsecIfStatsInSucceeded);
9805 SET_MIB(old_ip_mib.ipsecInFailed, ipmib->ipsecIfStatsInFailed);
9806 SET_MIB(old_ip_mib.ipInIPv6, ipmib->ipIfStatsInWrongIPVersion);
9807 SET_MIB(old_ip_mib.ipOutIPv6, ipmib->ipIfStatsOutWrongIPVersion);
9808 SET_MIB(old_ip_mib.ipOutSwitchIPv6,
9809 ipmib->ipIfStatsOutSwitchIPVersion);
9810
9811 if (!snmp_append_data(mpctl->b_cont, (char *)&old_ip_mib,
9812 (int)sizeof (old_ip_mib))) {
9813 ip1dbg(("ip_snmp_get_mib2_ip: failed to allocate %u bytes\n",
9814 (uint_t)sizeof (old_ip_mib)));
9815 }
9816
9817 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9818 ip3dbg(("ip_snmp_get_mib2_ip: level %d, name %d, len %d\n",
9819 (int)optp->level, (int)optp->name, (int)optp->len));
9820 qreply(q, mpctl);
9821 return (mp2ctl);
9822 }
9823
9824 /* Per interface IPv4 statistics */
9825 static mblk_t *
9826 ip_snmp_get_mib2_ip_traffic_stats(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
9827 boolean_t legacy_req)
9828 {
9829 struct opthdr *optp;
9830 mblk_t *mp2ctl;
9831 ill_t *ill;
9832 ill_walk_context_t ctx;
9833 mblk_t *mp_tail = NULL;
9834 mib2_ipIfStatsEntry_t global_ip_mib;
9835 mib2_ipAddrEntry_t mae;
9836
9837 /*
9838 * Make a copy of the original message
9839 */
9840 mp2ctl = copymsg(mpctl);
9841
9842 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9843 optp->level = MIB2_IP;
9844 optp->name = MIB2_IP_TRAFFIC_STATS;
9845 /* Include "unknown interface" ip_mib */
9846 ipst->ips_ip_mib.ipIfStatsIPVersion = MIB2_INETADDRESSTYPE_ipv4;
9847 ipst->ips_ip_mib.ipIfStatsIfIndex =
9848 MIB2_UNKNOWN_INTERFACE; /* Flag to netstat */
9849 SET_MIB(ipst->ips_ip_mib.ipIfStatsForwarding,
9850 (ipst->ips_ip_forwarding ? 1 : 2));
9851 SET_MIB(ipst->ips_ip_mib.ipIfStatsDefaultTTL,
9852 (uint32_t)ipst->ips_ip_def_ttl);
9853 SET_MIB(ipst->ips_ip_mib.ipIfStatsEntrySize,
9854 sizeof (mib2_ipIfStatsEntry_t));
9855 SET_MIB(ipst->ips_ip_mib.ipIfStatsAddrEntrySize,
9856 sizeof (mib2_ipAddrEntry_t));
9857 SET_MIB(ipst->ips_ip_mib.ipIfStatsRouteEntrySize,
9858 sizeof (mib2_ipRouteEntry_t));
9859 SET_MIB(ipst->ips_ip_mib.ipIfStatsNetToMediaEntrySize,
9860 sizeof (mib2_ipNetToMediaEntry_t));
9861 SET_MIB(ipst->ips_ip_mib.ipIfStatsMemberEntrySize,
9862 sizeof (ip_member_t));
9863 SET_MIB(ipst->ips_ip_mib.ipIfStatsGroupSourceEntrySize,
9864 sizeof (ip_grpsrc_t));
9865
9866 bcopy(&ipst->ips_ip_mib, &global_ip_mib, sizeof (global_ip_mib));
9867
9868 if (legacy_req) {
9869 SET_MIB(global_ip_mib.ipIfStatsAddrEntrySize,
9870 LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t));
9871 }
9872
9873 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
9874 (char *)&global_ip_mib, (int)sizeof (global_ip_mib))) {
9875 ip1dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9876 "failed to allocate %u bytes\n",
9877 (uint_t)sizeof (global_ip_mib)));
9878 }
9879
9880 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
9881 ill = ILL_START_WALK_V4(&ctx, ipst);
9882 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
9883 ill->ill_ip_mib->ipIfStatsIfIndex =
9884 ill->ill_phyint->phyint_ifindex;
9885 SET_MIB(ill->ill_ip_mib->ipIfStatsForwarding,
9886 (ipst->ips_ip_forwarding ? 1 : 2));
9887 SET_MIB(ill->ill_ip_mib->ipIfStatsDefaultTTL,
9888 (uint32_t)ipst->ips_ip_def_ttl);
9889
9890 ip_mib2_add_ip_stats(&global_ip_mib, ill->ill_ip_mib);
9891 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
9892 (char *)ill->ill_ip_mib,
9893 (int)sizeof (*ill->ill_ip_mib))) {
9894 ip1dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9895 "failed to allocate %u bytes\n",
9896 (uint_t)sizeof (*ill->ill_ip_mib)));
9897 }
9898 }
9899 rw_exit(&ipst->ips_ill_g_lock);
9900
9901 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9902 ip3dbg(("ip_snmp_get_mib2_ip_traffic_stats: "
9903 "level %d, name %d, len %d\n",
9904 (int)optp->level, (int)optp->name, (int)optp->len));
9905 qreply(q, mpctl);
9906
9907 if (mp2ctl == NULL)
9908 return (NULL);
9909
9910 return (ip_snmp_get_mib2_ip(q, mp2ctl, &global_ip_mib, ipst,
9911 legacy_req));
9912 }
9913
9914 /* Global IPv4 ICMP statistics */
9915 static mblk_t *
9916 ip_snmp_get_mib2_icmp(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9917 {
9918 struct opthdr *optp;
9919 mblk_t *mp2ctl;
9920
9921 /*
9922 * Make a copy of the original message
9923 */
9924 mp2ctl = copymsg(mpctl);
9925
9926 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9927 optp->level = MIB2_ICMP;
9928 optp->name = 0;
9929 if (!snmp_append_data(mpctl->b_cont, (char *)&ipst->ips_icmp_mib,
9930 (int)sizeof (ipst->ips_icmp_mib))) {
9931 ip1dbg(("ip_snmp_get_mib2_icmp: failed to allocate %u bytes\n",
9932 (uint_t)sizeof (ipst->ips_icmp_mib)));
9933 }
9934 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9935 ip3dbg(("ip_snmp_get_mib2_icmp: level %d, name %d, len %d\n",
9936 (int)optp->level, (int)optp->name, (int)optp->len));
9937 qreply(q, mpctl);
9938 return (mp2ctl);
9939 }
9940
9941 /* Global IPv4 IGMP statistics */
9942 static mblk_t *
9943 ip_snmp_get_mib2_igmp(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9944 {
9945 struct opthdr *optp;
9946 mblk_t *mp2ctl;
9947
9948 /*
9949 * make a copy of the original message
9950 */
9951 mp2ctl = copymsg(mpctl);
9952
9953 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9954 optp->level = EXPER_IGMP;
9955 optp->name = 0;
9956 if (!snmp_append_data(mpctl->b_cont, (char *)&ipst->ips_igmpstat,
9957 (int)sizeof (ipst->ips_igmpstat))) {
9958 ip1dbg(("ip_snmp_get_mib2_igmp: failed to allocate %u bytes\n",
9959 (uint_t)sizeof (ipst->ips_igmpstat)));
9960 }
9961 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9962 ip3dbg(("ip_snmp_get_mib2_igmp: level %d, name %d, len %d\n",
9963 (int)optp->level, (int)optp->name, (int)optp->len));
9964 qreply(q, mpctl);
9965 return (mp2ctl);
9966 }
9967
9968 /* Global IPv4 Multicast Routing statistics */
9969 static mblk_t *
9970 ip_snmp_get_mib2_multi(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
9971 {
9972 struct opthdr *optp;
9973 mblk_t *mp2ctl;
9974
9975 /*
9976 * make a copy of the original message
9977 */
9978 mp2ctl = copymsg(mpctl);
9979
9980 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
9981 optp->level = EXPER_DVMRP;
9982 optp->name = 0;
9983 if (!ip_mroute_stats(mpctl->b_cont, ipst)) {
9984 ip0dbg(("ip_mroute_stats: failed\n"));
9985 }
9986 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
9987 ip3dbg(("ip_snmp_get_mib2_multi: level %d, name %d, len %d\n",
9988 (int)optp->level, (int)optp->name, (int)optp->len));
9989 qreply(q, mpctl);
9990 return (mp2ctl);
9991 }
9992
9993 /* IPv4 address information */
9994 static mblk_t *
9995 ip_snmp_get_mib2_ip_addr(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
9996 boolean_t legacy_req)
9997 {
9998 struct opthdr *optp;
9999 mblk_t *mp2ctl;
10000 mblk_t *mp_tail = NULL;
10001 ill_t *ill;
10002 ipif_t *ipif;
10003 uint_t bitval;
10004 mib2_ipAddrEntry_t mae;
10005 size_t mae_size;
10006 zoneid_t zoneid;
10007 ill_walk_context_t ctx;
10008
10009 /*
10010 * make a copy of the original message
10011 */
10012 mp2ctl = copymsg(mpctl);
10013
10014 mae_size = (legacy_req) ? LEGACY_MIB_SIZE(&mae, mib2_ipAddrEntry_t) :
10015 sizeof (mib2_ipAddrEntry_t);
10016
10017 /* ipAddrEntryTable */
10018
10019 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10020 optp->level = MIB2_IP;
10021 optp->name = MIB2_IP_ADDR;
10022 zoneid = Q_TO_CONN(q)->conn_zoneid;
10023
10024 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10025 ill = ILL_START_WALK_V4(&ctx, ipst);
10026 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10027 for (ipif = ill->ill_ipif; ipif != NULL;
10028 ipif = ipif->ipif_next) {
10029 if (ipif->ipif_zoneid != zoneid &&
10030 ipif->ipif_zoneid != ALL_ZONES)
10031 continue;
10032 /* Sum of count from dead IRE_LO* and our current */
10033 mae.ipAdEntInfo.ae_ibcnt = ipif->ipif_ib_pkt_count;
10034 if (ipif->ipif_ire_local != NULL) {
10035 mae.ipAdEntInfo.ae_ibcnt +=
10036 ipif->ipif_ire_local->ire_ib_pkt_count;
10037 }
10038 mae.ipAdEntInfo.ae_obcnt = 0;
10039 mae.ipAdEntInfo.ae_focnt = 0;
10040
10041 ipif_get_name(ipif, mae.ipAdEntIfIndex.o_bytes,
10042 OCTET_LENGTH);
10043 mae.ipAdEntIfIndex.o_length =
10044 mi_strlen(mae.ipAdEntIfIndex.o_bytes);
10045 mae.ipAdEntAddr = ipif->ipif_lcl_addr;
10046 mae.ipAdEntNetMask = ipif->ipif_net_mask;
10047 mae.ipAdEntInfo.ae_subnet = ipif->ipif_subnet;
10048 mae.ipAdEntInfo.ae_subnet_len =
10049 ip_mask_to_plen(ipif->ipif_net_mask);
10050 mae.ipAdEntInfo.ae_src_addr = ipif->ipif_lcl_addr;
10051 for (bitval = 1;
10052 bitval &&
10053 !(bitval & ipif->ipif_brd_addr);
10054 bitval <<= 1)
10055 noop;
10056 mae.ipAdEntBcastAddr = bitval;
10057 mae.ipAdEntReasmMaxSize = IP_MAXPACKET;
10058 mae.ipAdEntInfo.ae_mtu = ipif->ipif_ill->ill_mtu;
10059 mae.ipAdEntInfo.ae_metric = ipif->ipif_ill->ill_metric;
10060 mae.ipAdEntInfo.ae_broadcast_addr =
10061 ipif->ipif_brd_addr;
10062 mae.ipAdEntInfo.ae_pp_dst_addr =
10063 ipif->ipif_pp_dst_addr;
10064 mae.ipAdEntInfo.ae_flags = ipif->ipif_flags |
10065 ill->ill_flags | ill->ill_phyint->phyint_flags;
10066 mae.ipAdEntRetransmitTime =
10067 ill->ill_reachable_retrans_time;
10068
10069 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10070 (char *)&mae, (int)mae_size)) {
10071 ip1dbg(("ip_snmp_get_mib2_ip_addr: failed to "
10072 "allocate %u bytes\n", (uint_t)mae_size));
10073 }
10074 }
10075 }
10076 rw_exit(&ipst->ips_ill_g_lock);
10077
10078 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10079 ip3dbg(("ip_snmp_get_mib2_ip_addr: level %d, name %d, len %d\n",
10080 (int)optp->level, (int)optp->name, (int)optp->len));
10081 qreply(q, mpctl);
10082 return (mp2ctl);
10083 }
10084
10085 /* IPv6 address information */
10086 static mblk_t *
10087 ip_snmp_get_mib2_ip6_addr(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
10088 boolean_t legacy_req)
10089 {
10090 struct opthdr *optp;
10091 mblk_t *mp2ctl;
10092 mblk_t *mp_tail = NULL;
10093 ill_t *ill;
10094 ipif_t *ipif;
10095 mib2_ipv6AddrEntry_t mae6;
10096 size_t mae6_size;
10097 zoneid_t zoneid;
10098 ill_walk_context_t ctx;
10099
10100 /*
10101 * make a copy of the original message
10102 */
10103 mp2ctl = copymsg(mpctl);
10104
10105 mae6_size = (legacy_req) ?
10106 LEGACY_MIB_SIZE(&mae6, mib2_ipv6AddrEntry_t) :
10107 sizeof (mib2_ipv6AddrEntry_t);
10108
10109 /* ipv6AddrEntryTable */
10110
10111 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10112 optp->level = MIB2_IP6;
10113 optp->name = MIB2_IP6_ADDR;
10114 zoneid = Q_TO_CONN(q)->conn_zoneid;
10115
10116 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10117 ill = ILL_START_WALK_V6(&ctx, ipst);
10118 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10119 for (ipif = ill->ill_ipif; ipif != NULL;
10120 ipif = ipif->ipif_next) {
10121 if (ipif->ipif_zoneid != zoneid &&
10122 ipif->ipif_zoneid != ALL_ZONES)
10123 continue;
10124 /* Sum of count from dead IRE_LO* and our current */
10125 mae6.ipv6AddrInfo.ae_ibcnt = ipif->ipif_ib_pkt_count;
10126 if (ipif->ipif_ire_local != NULL) {
10127 mae6.ipv6AddrInfo.ae_ibcnt +=
10128 ipif->ipif_ire_local->ire_ib_pkt_count;
10129 }
10130 mae6.ipv6AddrInfo.ae_obcnt = 0;
10131 mae6.ipv6AddrInfo.ae_focnt = 0;
10132
10133 ipif_get_name(ipif, mae6.ipv6AddrIfIndex.o_bytes,
10134 OCTET_LENGTH);
10135 mae6.ipv6AddrIfIndex.o_length =
10136 mi_strlen(mae6.ipv6AddrIfIndex.o_bytes);
10137 mae6.ipv6AddrAddress = ipif->ipif_v6lcl_addr;
10138 mae6.ipv6AddrPfxLength =
10139 ip_mask_to_plen_v6(&ipif->ipif_v6net_mask);
10140 mae6.ipv6AddrInfo.ae_subnet = ipif->ipif_v6subnet;
10141 mae6.ipv6AddrInfo.ae_subnet_len =
10142 mae6.ipv6AddrPfxLength;
10143 mae6.ipv6AddrInfo.ae_src_addr = ipif->ipif_v6lcl_addr;
10144
10145 /* Type: stateless(1), stateful(2), unknown(3) */
10146 if (ipif->ipif_flags & IPIF_ADDRCONF)
10147 mae6.ipv6AddrType = 1;
10148 else
10149 mae6.ipv6AddrType = 2;
10150 /* Anycast: true(1), false(2) */
10151 if (ipif->ipif_flags & IPIF_ANYCAST)
10152 mae6.ipv6AddrAnycastFlag = 1;
10153 else
10154 mae6.ipv6AddrAnycastFlag = 2;
10155
10156 /*
10157 * Address status: preferred(1), deprecated(2),
10158 * invalid(3), inaccessible(4), unknown(5)
10159 */
10160 if (ipif->ipif_flags & IPIF_NOLOCAL)
10161 mae6.ipv6AddrStatus = 3;
10162 else if (ipif->ipif_flags & IPIF_DEPRECATED)
10163 mae6.ipv6AddrStatus = 2;
10164 else
10165 mae6.ipv6AddrStatus = 1;
10166 mae6.ipv6AddrInfo.ae_mtu = ipif->ipif_ill->ill_mtu;
10167 mae6.ipv6AddrInfo.ae_metric =
10168 ipif->ipif_ill->ill_metric;
10169 mae6.ipv6AddrInfo.ae_pp_dst_addr =
10170 ipif->ipif_v6pp_dst_addr;
10171 mae6.ipv6AddrInfo.ae_flags = ipif->ipif_flags |
10172 ill->ill_flags | ill->ill_phyint->phyint_flags;
10173 mae6.ipv6AddrReasmMaxSize = IP_MAXPACKET;
10174 mae6.ipv6AddrIdentifier = ill->ill_token;
10175 mae6.ipv6AddrIdentifierLen = ill->ill_token_length;
10176 mae6.ipv6AddrReachableTime = ill->ill_reachable_time;
10177 mae6.ipv6AddrRetransmitTime =
10178 ill->ill_reachable_retrans_time;
10179 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10180 (char *)&mae6, (int)mae6_size)) {
10181 ip1dbg(("ip_snmp_get_mib2_ip6_addr: failed to "
10182 "allocate %u bytes\n",
10183 (uint_t)mae6_size));
10184 }
10185 }
10186 }
10187 rw_exit(&ipst->ips_ill_g_lock);
10188
10189 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10190 ip3dbg(("ip_snmp_get_mib2_ip6_addr: level %d, name %d, len %d\n",
10191 (int)optp->level, (int)optp->name, (int)optp->len));
10192 qreply(q, mpctl);
10193 return (mp2ctl);
10194 }
10195
10196 /* IPv4 multicast group membership. */
10197 static mblk_t *
10198 ip_snmp_get_mib2_ip_group_mem(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10199 {
10200 struct opthdr *optp;
10201 mblk_t *mp2ctl;
10202 ill_t *ill;
10203 ipif_t *ipif;
10204 ilm_t *ilm;
10205 ip_member_t ipm;
10206 mblk_t *mp_tail = NULL;
10207 ill_walk_context_t ctx;
10208 zoneid_t zoneid;
10209
10210 /*
10211 * make a copy of the original message
10212 */
10213 mp2ctl = copymsg(mpctl);
10214 zoneid = Q_TO_CONN(q)->conn_zoneid;
10215
10216 /* ipGroupMember table */
10217 optp = (struct opthdr *)&mpctl->b_rptr[
10218 sizeof (struct T_optmgmt_ack)];
10219 optp->level = MIB2_IP;
10220 optp->name = EXPER_IP_GROUP_MEMBERSHIP;
10221
10222 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10223 ill = ILL_START_WALK_V4(&ctx, ipst);
10224 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10225 /* Make sure the ill isn't going away. */
10226 if (!ill_check_and_refhold(ill))
10227 continue;
10228 rw_exit(&ipst->ips_ill_g_lock);
10229 rw_enter(&ill->ill_mcast_lock, RW_READER);
10230 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10231 if (ilm->ilm_zoneid != zoneid &&
10232 ilm->ilm_zoneid != ALL_ZONES)
10233 continue;
10234
10235 /* Is there an ipif for ilm_ifaddr? */
10236 for (ipif = ill->ill_ipif; ipif != NULL;
10237 ipif = ipif->ipif_next) {
10238 if (!IPIF_IS_CONDEMNED(ipif) &&
10239 ipif->ipif_lcl_addr == ilm->ilm_ifaddr &&
10240 ilm->ilm_ifaddr != INADDR_ANY)
10241 break;
10242 }
10243 if (ipif != NULL) {
10244 ipif_get_name(ipif,
10245 ipm.ipGroupMemberIfIndex.o_bytes,
10246 OCTET_LENGTH);
10247 } else {
10248 ill_get_name(ill,
10249 ipm.ipGroupMemberIfIndex.o_bytes,
10250 OCTET_LENGTH);
10251 }
10252 ipm.ipGroupMemberIfIndex.o_length =
10253 mi_strlen(ipm.ipGroupMemberIfIndex.o_bytes);
10254
10255 ipm.ipGroupMemberAddress = ilm->ilm_addr;
10256 ipm.ipGroupMemberRefCnt = ilm->ilm_refcnt;
10257 ipm.ipGroupMemberFilterMode = ilm->ilm_fmode;
10258 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10259 (char *)&ipm, (int)sizeof (ipm))) {
10260 ip1dbg(("ip_snmp_get_mib2_ip_group: "
10261 "failed to allocate %u bytes\n",
10262 (uint_t)sizeof (ipm)));
10263 }
10264 }
10265 rw_exit(&ill->ill_mcast_lock);
10266 ill_refrele(ill);
10267 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10268 }
10269 rw_exit(&ipst->ips_ill_g_lock);
10270 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10271 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10272 (int)optp->level, (int)optp->name, (int)optp->len));
10273 qreply(q, mpctl);
10274 return (mp2ctl);
10275 }
10276
10277 /* IPv6 multicast group membership. */
10278 static mblk_t *
10279 ip_snmp_get_mib2_ip6_group_mem(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10280 {
10281 struct opthdr *optp;
10282 mblk_t *mp2ctl;
10283 ill_t *ill;
10284 ilm_t *ilm;
10285 ipv6_member_t ipm6;
10286 mblk_t *mp_tail = NULL;
10287 ill_walk_context_t ctx;
10288 zoneid_t zoneid;
10289
10290 /*
10291 * make a copy of the original message
10292 */
10293 mp2ctl = copymsg(mpctl);
10294 zoneid = Q_TO_CONN(q)->conn_zoneid;
10295
10296 /* ip6GroupMember table */
10297 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10298 optp->level = MIB2_IP6;
10299 optp->name = EXPER_IP6_GROUP_MEMBERSHIP;
10300
10301 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10302 ill = ILL_START_WALK_V6(&ctx, ipst);
10303 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10304 /* Make sure the ill isn't going away. */
10305 if (!ill_check_and_refhold(ill))
10306 continue;
10307 rw_exit(&ipst->ips_ill_g_lock);
10308 /*
10309 * Normally we don't have any members on under IPMP interfaces.
10310 * We report them as a debugging aid.
10311 */
10312 rw_enter(&ill->ill_mcast_lock, RW_READER);
10313 ipm6.ipv6GroupMemberIfIndex = ill->ill_phyint->phyint_ifindex;
10314 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10315 if (ilm->ilm_zoneid != zoneid &&
10316 ilm->ilm_zoneid != ALL_ZONES)
10317 continue; /* not this zone */
10318 ipm6.ipv6GroupMemberAddress = ilm->ilm_v6addr;
10319 ipm6.ipv6GroupMemberRefCnt = ilm->ilm_refcnt;
10320 ipm6.ipv6GroupMemberFilterMode = ilm->ilm_fmode;
10321 if (!snmp_append_data2(mpctl->b_cont,
10322 &mp_tail,
10323 (char *)&ipm6, (int)sizeof (ipm6))) {
10324 ip1dbg(("ip_snmp_get_mib2_ip6_group: "
10325 "failed to allocate %u bytes\n",
10326 (uint_t)sizeof (ipm6)));
10327 }
10328 }
10329 rw_exit(&ill->ill_mcast_lock);
10330 ill_refrele(ill);
10331 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10332 }
10333 rw_exit(&ipst->ips_ill_g_lock);
10334
10335 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10336 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10337 (int)optp->level, (int)optp->name, (int)optp->len));
10338 qreply(q, mpctl);
10339 return (mp2ctl);
10340 }
10341
10342 /* IP multicast filtered sources */
10343 static mblk_t *
10344 ip_snmp_get_mib2_ip_group_src(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10345 {
10346 struct opthdr *optp;
10347 mblk_t *mp2ctl;
10348 ill_t *ill;
10349 ipif_t *ipif;
10350 ilm_t *ilm;
10351 ip_grpsrc_t ips;
10352 mblk_t *mp_tail = NULL;
10353 ill_walk_context_t ctx;
10354 zoneid_t zoneid;
10355 int i;
10356 slist_t *sl;
10357
10358 /*
10359 * make a copy of the original message
10360 */
10361 mp2ctl = copymsg(mpctl);
10362 zoneid = Q_TO_CONN(q)->conn_zoneid;
10363
10364 /* ipGroupSource table */
10365 optp = (struct opthdr *)&mpctl->b_rptr[
10366 sizeof (struct T_optmgmt_ack)];
10367 optp->level = MIB2_IP;
10368 optp->name = EXPER_IP_GROUP_SOURCES;
10369
10370 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10371 ill = ILL_START_WALK_V4(&ctx, ipst);
10372 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10373 /* Make sure the ill isn't going away. */
10374 if (!ill_check_and_refhold(ill))
10375 continue;
10376 rw_exit(&ipst->ips_ill_g_lock);
10377 rw_enter(&ill->ill_mcast_lock, RW_READER);
10378 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10379 sl = ilm->ilm_filter;
10380 if (ilm->ilm_zoneid != zoneid &&
10381 ilm->ilm_zoneid != ALL_ZONES)
10382 continue;
10383 if (SLIST_IS_EMPTY(sl))
10384 continue;
10385
10386 /* Is there an ipif for ilm_ifaddr? */
10387 for (ipif = ill->ill_ipif; ipif != NULL;
10388 ipif = ipif->ipif_next) {
10389 if (!IPIF_IS_CONDEMNED(ipif) &&
10390 ipif->ipif_lcl_addr == ilm->ilm_ifaddr &&
10391 ilm->ilm_ifaddr != INADDR_ANY)
10392 break;
10393 }
10394 if (ipif != NULL) {
10395 ipif_get_name(ipif,
10396 ips.ipGroupSourceIfIndex.o_bytes,
10397 OCTET_LENGTH);
10398 } else {
10399 ill_get_name(ill,
10400 ips.ipGroupSourceIfIndex.o_bytes,
10401 OCTET_LENGTH);
10402 }
10403 ips.ipGroupSourceIfIndex.o_length =
10404 mi_strlen(ips.ipGroupSourceIfIndex.o_bytes);
10405
10406 ips.ipGroupSourceGroup = ilm->ilm_addr;
10407 for (i = 0; i < sl->sl_numsrc; i++) {
10408 if (!IN6_IS_ADDR_V4MAPPED(&sl->sl_addr[i]))
10409 continue;
10410 IN6_V4MAPPED_TO_IPADDR(&sl->sl_addr[i],
10411 ips.ipGroupSourceAddress);
10412 if (snmp_append_data2(mpctl->b_cont, &mp_tail,
10413 (char *)&ips, (int)sizeof (ips)) == 0) {
10414 ip1dbg(("ip_snmp_get_mib2_ip_group_src:"
10415 " failed to allocate %u bytes\n",
10416 (uint_t)sizeof (ips)));
10417 }
10418 }
10419 }
10420 rw_exit(&ill->ill_mcast_lock);
10421 ill_refrele(ill);
10422 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10423 }
10424 rw_exit(&ipst->ips_ill_g_lock);
10425 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10426 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10427 (int)optp->level, (int)optp->name, (int)optp->len));
10428 qreply(q, mpctl);
10429 return (mp2ctl);
10430 }
10431
10432 /* IPv6 multicast filtered sources. */
10433 static mblk_t *
10434 ip_snmp_get_mib2_ip6_group_src(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10435 {
10436 struct opthdr *optp;
10437 mblk_t *mp2ctl;
10438 ill_t *ill;
10439 ilm_t *ilm;
10440 ipv6_grpsrc_t ips6;
10441 mblk_t *mp_tail = NULL;
10442 ill_walk_context_t ctx;
10443 zoneid_t zoneid;
10444 int i;
10445 slist_t *sl;
10446
10447 /*
10448 * make a copy of the original message
10449 */
10450 mp2ctl = copymsg(mpctl);
10451 zoneid = Q_TO_CONN(q)->conn_zoneid;
10452
10453 /* ip6GroupMember table */
10454 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10455 optp->level = MIB2_IP6;
10456 optp->name = EXPER_IP6_GROUP_SOURCES;
10457
10458 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10459 ill = ILL_START_WALK_V6(&ctx, ipst);
10460 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10461 /* Make sure the ill isn't going away. */
10462 if (!ill_check_and_refhold(ill))
10463 continue;
10464 rw_exit(&ipst->ips_ill_g_lock);
10465 /*
10466 * Normally we don't have any members on under IPMP interfaces.
10467 * We report them as a debugging aid.
10468 */
10469 rw_enter(&ill->ill_mcast_lock, RW_READER);
10470 ips6.ipv6GroupSourceIfIndex = ill->ill_phyint->phyint_ifindex;
10471 for (ilm = ill->ill_ilm; ilm; ilm = ilm->ilm_next) {
10472 sl = ilm->ilm_filter;
10473 if (ilm->ilm_zoneid != zoneid &&
10474 ilm->ilm_zoneid != ALL_ZONES)
10475 continue;
10476 if (SLIST_IS_EMPTY(sl))
10477 continue;
10478 ips6.ipv6GroupSourceGroup = ilm->ilm_v6addr;
10479 for (i = 0; i < sl->sl_numsrc; i++) {
10480 ips6.ipv6GroupSourceAddress = sl->sl_addr[i];
10481 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10482 (char *)&ips6, (int)sizeof (ips6))) {
10483 ip1dbg(("ip_snmp_get_mib2_ip6_"
10484 "group_src: failed to allocate "
10485 "%u bytes\n",
10486 (uint_t)sizeof (ips6)));
10487 }
10488 }
10489 }
10490 rw_exit(&ill->ill_mcast_lock);
10491 ill_refrele(ill);
10492 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10493 }
10494 rw_exit(&ipst->ips_ill_g_lock);
10495
10496 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10497 ip3dbg(("ip_snmp_get: level %d, name %d, len %d\n",
10498 (int)optp->level, (int)optp->name, (int)optp->len));
10499 qreply(q, mpctl);
10500 return (mp2ctl);
10501 }
10502
10503 /* Multicast routing virtual interface table. */
10504 static mblk_t *
10505 ip_snmp_get_mib2_virt_multi(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10506 {
10507 struct opthdr *optp;
10508 mblk_t *mp2ctl;
10509
10510 /*
10511 * make a copy of the original message
10512 */
10513 mp2ctl = copymsg(mpctl);
10514
10515 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10516 optp->level = EXPER_DVMRP;
10517 optp->name = EXPER_DVMRP_VIF;
10518 if (!ip_mroute_vif(mpctl->b_cont, ipst)) {
10519 ip0dbg(("ip_mroute_vif: failed\n"));
10520 }
10521 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10522 ip3dbg(("ip_snmp_get_mib2_virt_multi: level %d, name %d, len %d\n",
10523 (int)optp->level, (int)optp->name, (int)optp->len));
10524 qreply(q, mpctl);
10525 return (mp2ctl);
10526 }
10527
10528 /* Multicast routing table. */
10529 static mblk_t *
10530 ip_snmp_get_mib2_multi_rtable(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10531 {
10532 struct opthdr *optp;
10533 mblk_t *mp2ctl;
10534
10535 /*
10536 * make a copy of the original message
10537 */
10538 mp2ctl = copymsg(mpctl);
10539
10540 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10541 optp->level = EXPER_DVMRP;
10542 optp->name = EXPER_DVMRP_MRT;
10543 if (!ip_mroute_mrt(mpctl->b_cont, ipst)) {
10544 ip0dbg(("ip_mroute_mrt: failed\n"));
10545 }
10546 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10547 ip3dbg(("ip_snmp_get_mib2_multi_rtable: level %d, name %d, len %d\n",
10548 (int)optp->level, (int)optp->name, (int)optp->len));
10549 qreply(q, mpctl);
10550 return (mp2ctl);
10551 }
10552
10553 /*
10554 * Return ipRouteEntryTable, ipNetToMediaEntryTable, and ipRouteAttributeTable
10555 * in one IRE walk.
10556 */
10557 static mblk_t *
10558 ip_snmp_get_mib2_ip_route_media(queue_t *q, mblk_t *mpctl, int level,
10559 ip_stack_t *ipst)
10560 {
10561 struct opthdr *optp;
10562 mblk_t *mp2ctl; /* Returned */
10563 mblk_t *mp3ctl; /* nettomedia */
10564 mblk_t *mp4ctl; /* routeattrs */
10565 iproutedata_t ird;
10566 zoneid_t zoneid;
10567
10568 /*
10569 * make copies of the original message
10570 * - mp2ctl is returned unchanged to the caller for his use
10571 * - mpctl is sent upstream as ipRouteEntryTable
10572 * - mp3ctl is sent upstream as ipNetToMediaEntryTable
10573 * - mp4ctl is sent upstream as ipRouteAttributeTable
10574 */
10575 mp2ctl = copymsg(mpctl);
10576 mp3ctl = copymsg(mpctl);
10577 mp4ctl = copymsg(mpctl);
10578 if (mp3ctl == NULL || mp4ctl == NULL) {
10579 freemsg(mp4ctl);
10580 freemsg(mp3ctl);
10581 freemsg(mp2ctl);
10582 freemsg(mpctl);
10583 return (NULL);
10584 }
10585
10586 bzero(&ird, sizeof (ird));
10587
10588 ird.ird_route.lp_head = mpctl->b_cont;
10589 ird.ird_netmedia.lp_head = mp3ctl->b_cont;
10590 ird.ird_attrs.lp_head = mp4ctl->b_cont;
10591 /*
10592 * If the level has been set the special EXPER_IP_AND_ALL_IRES value,
10593 * then also include ire_testhidden IREs and IRE_IF_CLONE. This is
10594 * intended a temporary solution until a proper MIB API is provided
10595 * that provides complete filtering/caller-opt-in.
10596 */
10597 if (level == EXPER_IP_AND_ALL_IRES)
10598 ird.ird_flags |= IRD_REPORT_ALL;
10599
10600 zoneid = Q_TO_CONN(q)->conn_zoneid;
10601 ire_walk_v4(ip_snmp_get2_v4, &ird, zoneid, ipst);
10602
10603 /* ipRouteEntryTable in mpctl */
10604 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10605 optp->level = MIB2_IP;
10606 optp->name = MIB2_IP_ROUTE;
10607 optp->len = msgdsize(ird.ird_route.lp_head);
10608 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10609 (int)optp->level, (int)optp->name, (int)optp->len));
10610 qreply(q, mpctl);
10611
10612 /* ipNetToMediaEntryTable in mp3ctl */
10613 ncec_walk(NULL, ip_snmp_get2_v4_media, &ird, ipst);
10614
10615 optp = (struct opthdr *)&mp3ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10616 optp->level = MIB2_IP;
10617 optp->name = MIB2_IP_MEDIA;
10618 optp->len = msgdsize(ird.ird_netmedia.lp_head);
10619 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10620 (int)optp->level, (int)optp->name, (int)optp->len));
10621 qreply(q, mp3ctl);
10622
10623 /* ipRouteAttributeTable in mp4ctl */
10624 optp = (struct opthdr *)&mp4ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10625 optp->level = MIB2_IP;
10626 optp->name = EXPER_IP_RTATTR;
10627 optp->len = msgdsize(ird.ird_attrs.lp_head);
10628 ip3dbg(("ip_snmp_get_mib2_ip_route_media: level %d, name %d, len %d\n",
10629 (int)optp->level, (int)optp->name, (int)optp->len));
10630 if (optp->len == 0)
10631 freemsg(mp4ctl);
10632 else
10633 qreply(q, mp4ctl);
10634
10635 return (mp2ctl);
10636 }
10637
10638 /*
10639 * Return ipv6RouteEntryTable and ipv6RouteAttributeTable in one IRE walk, and
10640 * ipv6NetToMediaEntryTable in an NDP walk.
10641 */
10642 static mblk_t *
10643 ip_snmp_get_mib2_ip6_route_media(queue_t *q, mblk_t *mpctl, int level,
10644 ip_stack_t *ipst)
10645 {
10646 struct opthdr *optp;
10647 mblk_t *mp2ctl; /* Returned */
10648 mblk_t *mp3ctl; /* nettomedia */
10649 mblk_t *mp4ctl; /* routeattrs */
10650 iproutedata_t ird;
10651 zoneid_t zoneid;
10652
10653 /*
10654 * make copies of the original message
10655 * - mp2ctl is returned unchanged to the caller for his use
10656 * - mpctl is sent upstream as ipv6RouteEntryTable
10657 * - mp3ctl is sent upstream as ipv6NetToMediaEntryTable
10658 * - mp4ctl is sent upstream as ipv6RouteAttributeTable
10659 */
10660 mp2ctl = copymsg(mpctl);
10661 mp3ctl = copymsg(mpctl);
10662 mp4ctl = copymsg(mpctl);
10663 if (mp3ctl == NULL || mp4ctl == NULL) {
10664 freemsg(mp4ctl);
10665 freemsg(mp3ctl);
10666 freemsg(mp2ctl);
10667 freemsg(mpctl);
10668 return (NULL);
10669 }
10670
10671 bzero(&ird, sizeof (ird));
10672
10673 ird.ird_route.lp_head = mpctl->b_cont;
10674 ird.ird_netmedia.lp_head = mp3ctl->b_cont;
10675 ird.ird_attrs.lp_head = mp4ctl->b_cont;
10676 /*
10677 * If the level has been set the special EXPER_IP_AND_ALL_IRES value,
10678 * then also include ire_testhidden IREs and IRE_IF_CLONE. This is
10679 * intended a temporary solution until a proper MIB API is provided
10680 * that provides complete filtering/caller-opt-in.
10681 */
10682 if (level == EXPER_IP_AND_ALL_IRES)
10683 ird.ird_flags |= IRD_REPORT_ALL;
10684
10685 zoneid = Q_TO_CONN(q)->conn_zoneid;
10686 ire_walk_v6(ip_snmp_get2_v6_route, &ird, zoneid, ipst);
10687
10688 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10689 optp->level = MIB2_IP6;
10690 optp->name = MIB2_IP6_ROUTE;
10691 optp->len = msgdsize(ird.ird_route.lp_head);
10692 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10693 (int)optp->level, (int)optp->name, (int)optp->len));
10694 qreply(q, mpctl);
10695
10696 /* ipv6NetToMediaEntryTable in mp3ctl */
10697 ncec_walk(NULL, ip_snmp_get2_v6_media, &ird, ipst);
10698
10699 optp = (struct opthdr *)&mp3ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10700 optp->level = MIB2_IP6;
10701 optp->name = MIB2_IP6_MEDIA;
10702 optp->len = msgdsize(ird.ird_netmedia.lp_head);
10703 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10704 (int)optp->level, (int)optp->name, (int)optp->len));
10705 qreply(q, mp3ctl);
10706
10707 /* ipv6RouteAttributeTable in mp4ctl */
10708 optp = (struct opthdr *)&mp4ctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10709 optp->level = MIB2_IP6;
10710 optp->name = EXPER_IP_RTATTR;
10711 optp->len = msgdsize(ird.ird_attrs.lp_head);
10712 ip3dbg(("ip_snmp_get_mib2_ip6_route_media: level %d, name %d, len %d\n",
10713 (int)optp->level, (int)optp->name, (int)optp->len));
10714 if (optp->len == 0)
10715 freemsg(mp4ctl);
10716 else
10717 qreply(q, mp4ctl);
10718
10719 return (mp2ctl);
10720 }
10721
10722 /*
10723 * IPv6 mib: One per ill
10724 */
10725 static mblk_t *
10726 ip_snmp_get_mib2_ip6(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst,
10727 boolean_t legacy_req)
10728 {
10729 struct opthdr *optp;
10730 mblk_t *mp2ctl;
10731 ill_t *ill;
10732 ill_walk_context_t ctx;
10733 mblk_t *mp_tail = NULL;
10734 mib2_ipv6AddrEntry_t mae6;
10735 mib2_ipIfStatsEntry_t *ise;
10736 size_t ise_size, iae_size;
10737
10738 /*
10739 * Make a copy of the original message
10740 */
10741 mp2ctl = copymsg(mpctl);
10742
10743 /* fixed length IPv6 structure ... */
10744
10745 if (legacy_req) {
10746 ise_size = LEGACY_MIB_SIZE(&ipst->ips_ip6_mib,
10747 mib2_ipIfStatsEntry_t);
10748 iae_size = LEGACY_MIB_SIZE(&mae6, mib2_ipv6AddrEntry_t);
10749 } else {
10750 ise_size = sizeof (mib2_ipIfStatsEntry_t);
10751 iae_size = sizeof (mib2_ipv6AddrEntry_t);
10752 }
10753
10754 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10755 optp->level = MIB2_IP6;
10756 optp->name = 0;
10757 /* Include "unknown interface" ip6_mib */
10758 ipst->ips_ip6_mib.ipIfStatsIPVersion = MIB2_INETADDRESSTYPE_ipv6;
10759 ipst->ips_ip6_mib.ipIfStatsIfIndex =
10760 MIB2_UNKNOWN_INTERFACE; /* Flag to netstat */
10761 SET_MIB(ipst->ips_ip6_mib.ipIfStatsForwarding,
10762 ipst->ips_ipv6_forwarding ? 1 : 2);
10763 SET_MIB(ipst->ips_ip6_mib.ipIfStatsDefaultHopLimit,
10764 ipst->ips_ipv6_def_hops);
10765 SET_MIB(ipst->ips_ip6_mib.ipIfStatsEntrySize,
10766 sizeof (mib2_ipIfStatsEntry_t));
10767 SET_MIB(ipst->ips_ip6_mib.ipIfStatsAddrEntrySize,
10768 sizeof (mib2_ipv6AddrEntry_t));
10769 SET_MIB(ipst->ips_ip6_mib.ipIfStatsRouteEntrySize,
10770 sizeof (mib2_ipv6RouteEntry_t));
10771 SET_MIB(ipst->ips_ip6_mib.ipIfStatsNetToMediaEntrySize,
10772 sizeof (mib2_ipv6NetToMediaEntry_t));
10773 SET_MIB(ipst->ips_ip6_mib.ipIfStatsMemberEntrySize,
10774 sizeof (ipv6_member_t));
10775 SET_MIB(ipst->ips_ip6_mib.ipIfStatsGroupSourceEntrySize,
10776 sizeof (ipv6_grpsrc_t));
10777
10778 /*
10779 * Synchronize 64- and 32-bit counters
10780 */
10781 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInReceives,
10782 ipIfStatsHCInReceives);
10783 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInDelivers,
10784 ipIfStatsHCInDelivers);
10785 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutRequests,
10786 ipIfStatsHCOutRequests);
10787 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutForwDatagrams,
10788 ipIfStatsHCOutForwDatagrams);
10789 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsOutMcastPkts,
10790 ipIfStatsHCOutMcastPkts);
10791 SYNC32_MIB(&ipst->ips_ip6_mib, ipIfStatsInMcastPkts,
10792 ipIfStatsHCInMcastPkts);
10793
10794 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10795 (char *)&ipst->ips_ip6_mib, (int)ise_size)) {
10796 ip1dbg(("ip_snmp_get_mib2_ip6: failed to allocate %u bytes\n",
10797 (uint_t)ise_size));
10798 } else if (legacy_req) {
10799 /* Adjust the EntrySize fields for legacy requests. */
10800 ise =
10801 (mib2_ipIfStatsEntry_t *)(mp_tail->b_wptr - (int)ise_size);
10802 SET_MIB(ise->ipIfStatsEntrySize, ise_size);
10803 SET_MIB(ise->ipIfStatsAddrEntrySize, iae_size);
10804 }
10805
10806 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10807 ill = ILL_START_WALK_V6(&ctx, ipst);
10808 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10809 ill->ill_ip_mib->ipIfStatsIfIndex =
10810 ill->ill_phyint->phyint_ifindex;
10811 SET_MIB(ill->ill_ip_mib->ipIfStatsForwarding,
10812 ipst->ips_ipv6_forwarding ? 1 : 2);
10813 SET_MIB(ill->ill_ip_mib->ipIfStatsDefaultHopLimit,
10814 ill->ill_max_hops);
10815
10816 /*
10817 * Synchronize 64- and 32-bit counters
10818 */
10819 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInReceives,
10820 ipIfStatsHCInReceives);
10821 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInDelivers,
10822 ipIfStatsHCInDelivers);
10823 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutRequests,
10824 ipIfStatsHCOutRequests);
10825 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutForwDatagrams,
10826 ipIfStatsHCOutForwDatagrams);
10827 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsOutMcastPkts,
10828 ipIfStatsHCOutMcastPkts);
10829 SYNC32_MIB(ill->ill_ip_mib, ipIfStatsInMcastPkts,
10830 ipIfStatsHCInMcastPkts);
10831
10832 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10833 (char *)ill->ill_ip_mib, (int)ise_size)) {
10834 ip1dbg(("ip_snmp_get_mib2_ip6: failed to allocate "
10835 "%u bytes\n", (uint_t)ise_size));
10836 } else if (legacy_req) {
10837 /* Adjust the EntrySize fields for legacy requests. */
10838 ise = (mib2_ipIfStatsEntry_t *)(mp_tail->b_wptr -
10839 (int)ise_size);
10840 SET_MIB(ise->ipIfStatsEntrySize, ise_size);
10841 SET_MIB(ise->ipIfStatsAddrEntrySize, iae_size);
10842 }
10843 }
10844 rw_exit(&ipst->ips_ill_g_lock);
10845
10846 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10847 ip3dbg(("ip_snmp_get_mib2_ip6: level %d, name %d, len %d\n",
10848 (int)optp->level, (int)optp->name, (int)optp->len));
10849 qreply(q, mpctl);
10850 return (mp2ctl);
10851 }
10852
10853 /*
10854 * ICMPv6 mib: One per ill
10855 */
10856 static mblk_t *
10857 ip_snmp_get_mib2_icmp6(queue_t *q, mblk_t *mpctl, ip_stack_t *ipst)
10858 {
10859 struct opthdr *optp;
10860 mblk_t *mp2ctl;
10861 ill_t *ill;
10862 ill_walk_context_t ctx;
10863 mblk_t *mp_tail = NULL;
10864 /*
10865 * Make a copy of the original message
10866 */
10867 mp2ctl = copymsg(mpctl);
10868
10869 /* fixed length ICMPv6 structure ... */
10870
10871 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
10872 optp->level = MIB2_ICMP6;
10873 optp->name = 0;
10874 /* Include "unknown interface" icmp6_mib */
10875 ipst->ips_icmp6_mib.ipv6IfIcmpIfIndex =
10876 MIB2_UNKNOWN_INTERFACE; /* netstat flag */
10877 ipst->ips_icmp6_mib.ipv6IfIcmpEntrySize =
10878 sizeof (mib2_ipv6IfIcmpEntry_t);
10879 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10880 (char *)&ipst->ips_icmp6_mib,
10881 (int)sizeof (ipst->ips_icmp6_mib))) {
10882 ip1dbg(("ip_snmp_get_mib2_icmp6: failed to allocate %u bytes\n",
10883 (uint_t)sizeof (ipst->ips_icmp6_mib)));
10884 }
10885
10886 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
10887 ill = ILL_START_WALK_V6(&ctx, ipst);
10888 for (; ill != NULL; ill = ill_next(&ctx, ill)) {
10889 ill->ill_icmp6_mib->ipv6IfIcmpIfIndex =
10890 ill->ill_phyint->phyint_ifindex;
10891 if (!snmp_append_data2(mpctl->b_cont, &mp_tail,
10892 (char *)ill->ill_icmp6_mib,
10893 (int)sizeof (*ill->ill_icmp6_mib))) {
10894 ip1dbg(("ip_snmp_get_mib2_icmp6: failed to allocate "
10895 "%u bytes\n",
10896 (uint_t)sizeof (*ill->ill_icmp6_mib)));
10897 }
10898 }
10899 rw_exit(&ipst->ips_ill_g_lock);
10900
10901 optp->len = (t_uscalar_t)msgdsize(mpctl->b_cont);
10902 ip3dbg(("ip_snmp_get_mib2_icmp6: level %d, name %d, len %d\n",
10903 (int)optp->level, (int)optp->name, (int)optp->len));
10904 qreply(q, mpctl);
10905 return (mp2ctl);
10906 }
10907
10908 /*
10909 * ire_walk routine to create both ipRouteEntryTable and
10910 * ipRouteAttributeTable in one IRE walk
10911 */
10912 static void
10913 ip_snmp_get2_v4(ire_t *ire, iproutedata_t *ird)
10914 {
10915 ill_t *ill;
10916 mib2_ipRouteEntry_t *re;
10917 mib2_ipAttributeEntry_t iaes;
10918 tsol_ire_gw_secattr_t *attrp;
10919 tsol_gc_t *gc = NULL;
10920 tsol_gcgrp_t *gcgrp = NULL;
10921 ip_stack_t *ipst = ire->ire_ipst;
10922
10923 ASSERT(ire->ire_ipversion == IPV4_VERSION);
10924
10925 if (!(ird->ird_flags & IRD_REPORT_ALL)) {
10926 if (ire->ire_testhidden)
10927 return;
10928 if (ire->ire_type & IRE_IF_CLONE)
10929 return;
10930 }
10931
10932 if ((re = kmem_zalloc(sizeof (*re), KM_NOSLEEP)) == NULL)
10933 return;
10934
10935 if ((attrp = ire->ire_gw_secattr) != NULL) {
10936 mutex_enter(&attrp->igsa_lock);
10937 if ((gc = attrp->igsa_gc) != NULL) {
10938 gcgrp = gc->gc_grp;
10939 ASSERT(gcgrp != NULL);
10940 rw_enter(&gcgrp->gcgrp_rwlock, RW_READER);
10941 }
10942 mutex_exit(&attrp->igsa_lock);
10943 }
10944 /*
10945 * Return all IRE types for route table... let caller pick and choose
10946 */
10947 re->ipRouteDest = ire->ire_addr;
10948 ill = ire->ire_ill;
10949 re->ipRouteIfIndex.o_length = 0;
10950 if (ill != NULL) {
10951 ill_get_name(ill, re->ipRouteIfIndex.o_bytes, OCTET_LENGTH);
10952 re->ipRouteIfIndex.o_length =
10953 mi_strlen(re->ipRouteIfIndex.o_bytes);
10954 }
10955 re->ipRouteMetric1 = -1;
10956 re->ipRouteMetric2 = -1;
10957 re->ipRouteMetric3 = -1;
10958 re->ipRouteMetric4 = -1;
10959
10960 re->ipRouteNextHop = ire->ire_gateway_addr;
10961 /* indirect(4), direct(3), or invalid(2) */
10962 if (ire->ire_flags & (RTF_REJECT | RTF_BLACKHOLE))
10963 re->ipRouteType = 2;
10964 else if (ire->ire_type & IRE_ONLINK)
10965 re->ipRouteType = 3;
10966 else
10967 re->ipRouteType = 4;
10968
10969 re->ipRouteProto = -1;
10970 re->ipRouteAge = gethrestime_sec() - ire->ire_create_time;
10971 re->ipRouteMask = ire->ire_mask;
10972 re->ipRouteMetric5 = -1;
10973 re->ipRouteInfo.re_max_frag = ire->ire_metrics.iulp_mtu;
10974 if (ire->ire_ill != NULL && re->ipRouteInfo.re_max_frag == 0)
10975 re->ipRouteInfo.re_max_frag = ire->ire_ill->ill_mtu;
10976
10977 re->ipRouteInfo.re_frag_flag = 0;
10978 re->ipRouteInfo.re_rtt = 0;
10979 re->ipRouteInfo.re_src_addr = 0;
10980 re->ipRouteInfo.re_ref = ire->ire_refcnt;
10981 re->ipRouteInfo.re_obpkt = ire->ire_ob_pkt_count;
10982 re->ipRouteInfo.re_ibpkt = ire->ire_ib_pkt_count;
10983 re->ipRouteInfo.re_flags = ire->ire_flags;
10984
10985 /* Add the IRE_IF_CLONE's counters to their parent IRE_INTERFACE */
10986 if (ire->ire_type & IRE_INTERFACE) {
10987 ire_t *child;
10988
10989 rw_enter(&ipst->ips_ire_dep_lock, RW_READER);
10990 child = ire->ire_dep_children;
10991 while (child != NULL) {
10992 re->ipRouteInfo.re_obpkt += child->ire_ob_pkt_count;
10993 re->ipRouteInfo.re_ibpkt += child->ire_ib_pkt_count;
10994 child = child->ire_dep_sib_next;
10995 }
10996 rw_exit(&ipst->ips_ire_dep_lock);
10997 }
10998
10999 if (ire->ire_flags & RTF_DYNAMIC) {
11000 re->ipRouteInfo.re_ire_type = IRE_HOST_REDIRECT;
11001 } else {
11002 re->ipRouteInfo.re_ire_type = ire->ire_type;
11003 }
11004
11005 if (!snmp_append_data2(ird->ird_route.lp_head, &ird->ird_route.lp_tail,
11006 (char *)re, (int)sizeof (*re))) {
11007 ip1dbg(("ip_snmp_get2_v4: failed to allocate %u bytes\n",
11008 (uint_t)sizeof (*re)));
11009 }
11010
11011 if (gc != NULL) {
11012 iaes.iae_routeidx = ird->ird_idx;
11013 iaes.iae_doi = gc->gc_db->gcdb_doi;
11014 iaes.iae_slrange = gc->gc_db->gcdb_slrange;
11015
11016 if (!snmp_append_data2(ird->ird_attrs.lp_head,
11017 &ird->ird_attrs.lp_tail, (char *)&iaes, sizeof (iaes))) {
11018 ip1dbg(("ip_snmp_get2_v4: failed to allocate %u "
11019 "bytes\n", (uint_t)sizeof (iaes)));
11020 }
11021 }
11022
11023 /* bump route index for next pass */
11024 ird->ird_idx++;
11025
11026 kmem_free(re, sizeof (*re));
11027 if (gcgrp != NULL)
11028 rw_exit(&gcgrp->gcgrp_rwlock);
11029 }
11030
11031 /*
11032 * ire_walk routine to create ipv6RouteEntryTable and ipRouteEntryTable.
11033 */
11034 static void
11035 ip_snmp_get2_v6_route(ire_t *ire, iproutedata_t *ird)
11036 {
11037 ill_t *ill;
11038 mib2_ipv6RouteEntry_t *re;
11039 mib2_ipAttributeEntry_t iaes;
11040 tsol_ire_gw_secattr_t *attrp;
11041 tsol_gc_t *gc = NULL;
11042 tsol_gcgrp_t *gcgrp = NULL;
11043 ip_stack_t *ipst = ire->ire_ipst;
11044
11045 ASSERT(ire->ire_ipversion == IPV6_VERSION);
11046
11047 if (!(ird->ird_flags & IRD_REPORT_ALL)) {
11048 if (ire->ire_testhidden)
11049 return;
11050 if (ire->ire_type & IRE_IF_CLONE)
11051 return;
11052 }
11053
11054 if ((re = kmem_zalloc(sizeof (*re), KM_NOSLEEP)) == NULL)
11055 return;
11056
11057 if ((attrp = ire->ire_gw_secattr) != NULL) {
11058 mutex_enter(&attrp->igsa_lock);
11059 if ((gc = attrp->igsa_gc) != NULL) {
11060 gcgrp = gc->gc_grp;
11061 ASSERT(gcgrp != NULL);
11062 rw_enter(&gcgrp->gcgrp_rwlock, RW_READER);
11063 }
11064 mutex_exit(&attrp->igsa_lock);
11065 }
11066 /*
11067 * Return all IRE types for route table... let caller pick and choose
11068 */
11069 re->ipv6RouteDest = ire->ire_addr_v6;
11070 re->ipv6RoutePfxLength = ip_mask_to_plen_v6(&ire->ire_mask_v6);
11071 re->ipv6RouteIndex = 0; /* Unique when multiple with same dest/plen */
11072 re->ipv6RouteIfIndex.o_length = 0;
11073 ill = ire->ire_ill;
11074 if (ill != NULL) {
11075 ill_get_name(ill, re->ipv6RouteIfIndex.o_bytes, OCTET_LENGTH);
11076 re->ipv6RouteIfIndex.o_length =
11077 mi_strlen(re->ipv6RouteIfIndex.o_bytes);
11078 }
11079
11080 ASSERT(!(ire->ire_type & IRE_BROADCAST));
11081
11082 mutex_enter(&ire->ire_lock);
11083 re->ipv6RouteNextHop = ire->ire_gateway_addr_v6;
11084 mutex_exit(&ire->ire_lock);
11085
11086 /* remote(4), local(3), or discard(2) */
11087 if (ire->ire_flags & (RTF_REJECT | RTF_BLACKHOLE))
11088 re->ipv6RouteType = 2;
11089 else if (ire->ire_type & IRE_ONLINK)
11090 re->ipv6RouteType = 3;
11091 else
11092 re->ipv6RouteType = 4;
11093
11094 re->ipv6RouteProtocol = -1;
11095 re->ipv6RoutePolicy = 0;
11096 re->ipv6RouteAge = gethrestime_sec() - ire->ire_create_time;
11097 re->ipv6RouteNextHopRDI = 0;
11098 re->ipv6RouteWeight = 0;
11099 re->ipv6RouteMetric = 0;
11100 re->ipv6RouteInfo.re_max_frag = ire->ire_metrics.iulp_mtu;
11101 if (ire->ire_ill != NULL && re->ipv6RouteInfo.re_max_frag == 0)
11102 re->ipv6RouteInfo.re_max_frag = ire->ire_ill->ill_mtu;
11103
11104 re->ipv6RouteInfo.re_frag_flag = 0;
11105 re->ipv6RouteInfo.re_rtt = 0;
11106 re->ipv6RouteInfo.re_src_addr = ipv6_all_zeros;
11107 re->ipv6RouteInfo.re_obpkt = ire->ire_ob_pkt_count;
11108 re->ipv6RouteInfo.re_ibpkt = ire->ire_ib_pkt_count;
11109 re->ipv6RouteInfo.re_ref = ire->ire_refcnt;
11110 re->ipv6RouteInfo.re_flags = ire->ire_flags;
11111
11112 /* Add the IRE_IF_CLONE's counters to their parent IRE_INTERFACE */
11113 if (ire->ire_type & IRE_INTERFACE) {
11114 ire_t *child;
11115
11116 rw_enter(&ipst->ips_ire_dep_lock, RW_READER);
11117 child = ire->ire_dep_children;
11118 while (child != NULL) {
11119 re->ipv6RouteInfo.re_obpkt += child->ire_ob_pkt_count;
11120 re->ipv6RouteInfo.re_ibpkt += child->ire_ib_pkt_count;
11121 child = child->ire_dep_sib_next;
11122 }
11123 rw_exit(&ipst->ips_ire_dep_lock);
11124 }
11125 if (ire->ire_flags & RTF_DYNAMIC) {
11126 re->ipv6RouteInfo.re_ire_type = IRE_HOST_REDIRECT;
11127 } else {
11128 re->ipv6RouteInfo.re_ire_type = ire->ire_type;
11129 }
11130
11131 if (!snmp_append_data2(ird->ird_route.lp_head, &ird->ird_route.lp_tail,
11132 (char *)re, (int)sizeof (*re))) {
11133 ip1dbg(("ip_snmp_get2_v6: failed to allocate %u bytes\n",
11134 (uint_t)sizeof (*re)));
11135 }
11136
11137 if (gc != NULL) {
11138 iaes.iae_routeidx = ird->ird_idx;
11139 iaes.iae_doi = gc->gc_db->gcdb_doi;
11140 iaes.iae_slrange = gc->gc_db->gcdb_slrange;
11141
11142 if (!snmp_append_data2(ird->ird_attrs.lp_head,
11143 &ird->ird_attrs.lp_tail, (char *)&iaes, sizeof (iaes))) {
11144 ip1dbg(("ip_snmp_get2_v6: failed to allocate %u "
11145 "bytes\n", (uint_t)sizeof (iaes)));
11146 }
11147 }
11148
11149 /* bump route index for next pass */
11150 ird->ird_idx++;
11151
11152 kmem_free(re, sizeof (*re));
11153 if (gcgrp != NULL)
11154 rw_exit(&gcgrp->gcgrp_rwlock);
11155 }
11156
11157 /*
11158 * ncec_walk routine to create ipv6NetToMediaEntryTable
11159 */
11160 static int
11161 ip_snmp_get2_v6_media(ncec_t *ncec, iproutedata_t *ird)
11162 {
11163 ill_t *ill;
11164 mib2_ipv6NetToMediaEntry_t ntme;
11165
11166 ill = ncec->ncec_ill;
11167 /* skip arpce entries, and loopback ncec entries */
11168 if (ill->ill_isv6 == B_FALSE || ill->ill_net_type == IRE_LOOPBACK)
11169 return (0);
11170 /*
11171 * Neighbor cache entry attached to IRE with on-link
11172 * destination.
11173 * We report all IPMP groups on ncec_ill which is normally the upper.
11174 */
11175 ntme.ipv6NetToMediaIfIndex = ill->ill_phyint->phyint_ifindex;
11176 ntme.ipv6NetToMediaNetAddress = ncec->ncec_addr;
11177 ntme.ipv6NetToMediaPhysAddress.o_length = ill->ill_phys_addr_length;
11178 if (ncec->ncec_lladdr != NULL) {
11179 bcopy(ncec->ncec_lladdr, ntme.ipv6NetToMediaPhysAddress.o_bytes,
11180 ntme.ipv6NetToMediaPhysAddress.o_length);
11181 }
11182 /*
11183 * Note: Returns ND_* states. Should be:
11184 * reachable(1), stale(2), delay(3), probe(4),
11185 * invalid(5), unknown(6)
11186 */
11187 ntme.ipv6NetToMediaState = ncec->ncec_state;
11188 ntme.ipv6NetToMediaLastUpdated = 0;
11189
11190 /* other(1), dynamic(2), static(3), local(4) */
11191 if (NCE_MYADDR(ncec)) {
11192 ntme.ipv6NetToMediaType = 4;
11193 } else if (ncec->ncec_flags & NCE_F_PUBLISH) {
11194 ntme.ipv6NetToMediaType = 1; /* proxy */
11195 } else if (ncec->ncec_flags & NCE_F_STATIC) {
11196 ntme.ipv6NetToMediaType = 3;
11197 } else if (ncec->ncec_flags & (NCE_F_MCAST|NCE_F_BCAST)) {
11198 ntme.ipv6NetToMediaType = 1;
11199 } else {
11200 ntme.ipv6NetToMediaType = 2;
11201 }
11202
11203 if (!snmp_append_data2(ird->ird_netmedia.lp_head,
11204 &ird->ird_netmedia.lp_tail, (char *)&ntme, sizeof (ntme))) {
11205 ip1dbg(("ip_snmp_get2_v6_media: failed to allocate %u bytes\n",
11206 (uint_t)sizeof (ntme)));
11207 }
11208 return (0);
11209 }
11210
11211 int
11212 nce2ace(ncec_t *ncec)
11213 {
11214 int flags = 0;
11215
11216 if (NCE_ISREACHABLE(ncec))
11217 flags |= ACE_F_RESOLVED;
11218 if (ncec->ncec_flags & NCE_F_AUTHORITY)
11219 flags |= ACE_F_AUTHORITY;
11220 if (ncec->ncec_flags & NCE_F_PUBLISH)
11221 flags |= ACE_F_PUBLISH;
11222 if ((ncec->ncec_flags & NCE_F_NONUD) != 0)
11223 flags |= ACE_F_PERMANENT;
11224 if (NCE_MYADDR(ncec))
11225 flags |= (ACE_F_MYADDR | ACE_F_AUTHORITY);
11226 if (ncec->ncec_flags & NCE_F_UNVERIFIED)
11227 flags |= ACE_F_UNVERIFIED;
11228 if (ncec->ncec_flags & NCE_F_AUTHORITY)
11229 flags |= ACE_F_AUTHORITY;
11230 if (ncec->ncec_flags & NCE_F_DELAYED)
11231 flags |= ACE_F_DELAYED;
11232 return (flags);
11233 }
11234
11235 /*
11236 * ncec_walk routine to create ipNetToMediaEntryTable
11237 */
11238 static int
11239 ip_snmp_get2_v4_media(ncec_t *ncec, iproutedata_t *ird)
11240 {
11241 ill_t *ill;
11242 mib2_ipNetToMediaEntry_t ntme;
11243 const char *name = "unknown";
11244 ipaddr_t ncec_addr;
11245
11246 ill = ncec->ncec_ill;
11247 if (ill->ill_isv6 || (ncec->ncec_flags & NCE_F_BCAST) ||
11248 ill->ill_net_type == IRE_LOOPBACK)
11249 return (0);
11250
11251 /* We report all IPMP groups on ncec_ill which is normally the upper. */
11252 name = ill->ill_name;
11253 /* Based on RFC 4293: other(1), inval(2), dyn(3), stat(4) */
11254 if (NCE_MYADDR(ncec)) {
11255 ntme.ipNetToMediaType = 4;
11256 } else if (ncec->ncec_flags & (NCE_F_MCAST|NCE_F_BCAST|NCE_F_PUBLISH)) {
11257 ntme.ipNetToMediaType = 1;
11258 } else {
11259 ntme.ipNetToMediaType = 3;
11260 }
11261 ntme.ipNetToMediaIfIndex.o_length = MIN(OCTET_LENGTH, strlen(name));
11262 bcopy(name, ntme.ipNetToMediaIfIndex.o_bytes,
11263 ntme.ipNetToMediaIfIndex.o_length);
11264
11265 IN6_V4MAPPED_TO_IPADDR(&ncec->ncec_addr, ncec_addr);
11266 bcopy(&ncec_addr, &ntme.ipNetToMediaNetAddress, sizeof (ncec_addr));
11267
11268 ntme.ipNetToMediaInfo.ntm_mask.o_length = sizeof (ipaddr_t);
11269 ncec_addr = INADDR_BROADCAST;
11270 bcopy(&ncec_addr, ntme.ipNetToMediaInfo.ntm_mask.o_bytes,
11271 sizeof (ncec_addr));
11272 /*
11273 * map all the flags to the ACE counterpart.
11274 */
11275 ntme.ipNetToMediaInfo.ntm_flags = nce2ace(ncec);
11276
11277 ntme.ipNetToMediaPhysAddress.o_length =
11278 MIN(OCTET_LENGTH, ill->ill_phys_addr_length);
11279
11280 if (!NCE_ISREACHABLE(ncec))
11281 ntme.ipNetToMediaPhysAddress.o_length = 0;
11282 else {
11283 if (ncec->ncec_lladdr != NULL) {
11284 bcopy(ncec->ncec_lladdr,
11285 ntme.ipNetToMediaPhysAddress.o_bytes,
11286 ntme.ipNetToMediaPhysAddress.o_length);
11287 }
11288 }
11289
11290 if (!snmp_append_data2(ird->ird_netmedia.lp_head,
11291 &ird->ird_netmedia.lp_tail, (char *)&ntme, sizeof (ntme))) {
11292 ip1dbg(("ip_snmp_get2_v4_media: failed to allocate %u bytes\n",
11293 (uint_t)sizeof (ntme)));
11294 }
11295 return (0);
11296 }
11297
11298 /*
11299 * return (0) if invalid set request, 1 otherwise, including non-tcp requests
11300 */
11301 /* ARGSUSED */
11302 int
11303 ip_snmp_set(queue_t *q, int level, int name, uchar_t *ptr, int len)
11304 {
11305 switch (level) {
11306 case MIB2_IP:
11307 case MIB2_ICMP:
11308 switch (name) {
11309 default:
11310 break;
11311 }
11312 return (1);
11313 default:
11314 return (1);
11315 }
11316 }
11317
11318 /*
11319 * When there exists both a 64- and 32-bit counter of a particular type
11320 * (i.e., InReceives), only the 64-bit counters are added.
11321 */
11322 void
11323 ip_mib2_add_ip_stats(mib2_ipIfStatsEntry_t *o1, mib2_ipIfStatsEntry_t *o2)
11324 {
11325 UPDATE_MIB(o1, ipIfStatsInHdrErrors, o2->ipIfStatsInHdrErrors);
11326 UPDATE_MIB(o1, ipIfStatsInTooBigErrors, o2->ipIfStatsInTooBigErrors);
11327 UPDATE_MIB(o1, ipIfStatsInNoRoutes, o2->ipIfStatsInNoRoutes);
11328 UPDATE_MIB(o1, ipIfStatsInAddrErrors, o2->ipIfStatsInAddrErrors);
11329 UPDATE_MIB(o1, ipIfStatsInUnknownProtos, o2->ipIfStatsInUnknownProtos);
11330 UPDATE_MIB(o1, ipIfStatsInTruncatedPkts, o2->ipIfStatsInTruncatedPkts);
11331 UPDATE_MIB(o1, ipIfStatsInDiscards, o2->ipIfStatsInDiscards);
11332 UPDATE_MIB(o1, ipIfStatsOutDiscards, o2->ipIfStatsOutDiscards);
11333 UPDATE_MIB(o1, ipIfStatsOutFragOKs, o2->ipIfStatsOutFragOKs);
11334 UPDATE_MIB(o1, ipIfStatsOutFragFails, o2->ipIfStatsOutFragFails);
11335 UPDATE_MIB(o1, ipIfStatsOutFragCreates, o2->ipIfStatsOutFragCreates);
11336 UPDATE_MIB(o1, ipIfStatsReasmReqds, o2->ipIfStatsReasmReqds);
11337 UPDATE_MIB(o1, ipIfStatsReasmOKs, o2->ipIfStatsReasmOKs);
11338 UPDATE_MIB(o1, ipIfStatsReasmFails, o2->ipIfStatsReasmFails);
11339 UPDATE_MIB(o1, ipIfStatsOutNoRoutes, o2->ipIfStatsOutNoRoutes);
11340 UPDATE_MIB(o1, ipIfStatsReasmDuplicates, o2->ipIfStatsReasmDuplicates);
11341 UPDATE_MIB(o1, ipIfStatsReasmPartDups, o2->ipIfStatsReasmPartDups);
11342 UPDATE_MIB(o1, ipIfStatsForwProhibits, o2->ipIfStatsForwProhibits);
11343 UPDATE_MIB(o1, udpInCksumErrs, o2->udpInCksumErrs);
11344 UPDATE_MIB(o1, udpInOverflows, o2->udpInOverflows);
11345 UPDATE_MIB(o1, rawipInOverflows, o2->rawipInOverflows);
11346 UPDATE_MIB(o1, ipIfStatsInWrongIPVersion,
11347 o2->ipIfStatsInWrongIPVersion);
11348 UPDATE_MIB(o1, ipIfStatsOutWrongIPVersion,
11349 o2->ipIfStatsInWrongIPVersion);
11350 UPDATE_MIB(o1, ipIfStatsOutSwitchIPVersion,
11351 o2->ipIfStatsOutSwitchIPVersion);
11352 UPDATE_MIB(o1, ipIfStatsHCInReceives, o2->ipIfStatsHCInReceives);
11353 UPDATE_MIB(o1, ipIfStatsHCInOctets, o2->ipIfStatsHCInOctets);
11354 UPDATE_MIB(o1, ipIfStatsHCInForwDatagrams,
11355 o2->ipIfStatsHCInForwDatagrams);
11356 UPDATE_MIB(o1, ipIfStatsHCInDelivers, o2->ipIfStatsHCInDelivers);
11357 UPDATE_MIB(o1, ipIfStatsHCOutRequests, o2->ipIfStatsHCOutRequests);
11358 UPDATE_MIB(o1, ipIfStatsHCOutForwDatagrams,
11359 o2->ipIfStatsHCOutForwDatagrams);
11360 UPDATE_MIB(o1, ipIfStatsOutFragReqds, o2->ipIfStatsOutFragReqds);
11361 UPDATE_MIB(o1, ipIfStatsHCOutTransmits, o2->ipIfStatsHCOutTransmits);
11362 UPDATE_MIB(o1, ipIfStatsHCOutOctets, o2->ipIfStatsHCOutOctets);
11363 UPDATE_MIB(o1, ipIfStatsHCInMcastPkts, o2->ipIfStatsHCInMcastPkts);
11364 UPDATE_MIB(o1, ipIfStatsHCInMcastOctets, o2->ipIfStatsHCInMcastOctets);
11365 UPDATE_MIB(o1, ipIfStatsHCOutMcastPkts, o2->ipIfStatsHCOutMcastPkts);
11366 UPDATE_MIB(o1, ipIfStatsHCOutMcastOctets,
11367 o2->ipIfStatsHCOutMcastOctets);
11368 UPDATE_MIB(o1, ipIfStatsHCInBcastPkts, o2->ipIfStatsHCInBcastPkts);
11369 UPDATE_MIB(o1, ipIfStatsHCOutBcastPkts, o2->ipIfStatsHCOutBcastPkts);
11370 UPDATE_MIB(o1, ipsecInSucceeded, o2->ipsecInSucceeded);
11371 UPDATE_MIB(o1, ipsecInFailed, o2->ipsecInFailed);
11372 UPDATE_MIB(o1, ipInCksumErrs, o2->ipInCksumErrs);
11373 UPDATE_MIB(o1, tcpInErrs, o2->tcpInErrs);
11374 UPDATE_MIB(o1, udpNoPorts, o2->udpNoPorts);
11375 }
11376
11377 void
11378 ip_mib2_add_icmp6_stats(mib2_ipv6IfIcmpEntry_t *o1, mib2_ipv6IfIcmpEntry_t *o2)
11379 {
11380 UPDATE_MIB(o1, ipv6IfIcmpInMsgs, o2->ipv6IfIcmpInMsgs);
11381 UPDATE_MIB(o1, ipv6IfIcmpInErrors, o2->ipv6IfIcmpInErrors);
11382 UPDATE_MIB(o1, ipv6IfIcmpInDestUnreachs, o2->ipv6IfIcmpInDestUnreachs);
11383 UPDATE_MIB(o1, ipv6IfIcmpInAdminProhibs, o2->ipv6IfIcmpInAdminProhibs);
11384 UPDATE_MIB(o1, ipv6IfIcmpInTimeExcds, o2->ipv6IfIcmpInTimeExcds);
11385 UPDATE_MIB(o1, ipv6IfIcmpInParmProblems, o2->ipv6IfIcmpInParmProblems);
11386 UPDATE_MIB(o1, ipv6IfIcmpInPktTooBigs, o2->ipv6IfIcmpInPktTooBigs);
11387 UPDATE_MIB(o1, ipv6IfIcmpInEchos, o2->ipv6IfIcmpInEchos);
11388 UPDATE_MIB(o1, ipv6IfIcmpInEchoReplies, o2->ipv6IfIcmpInEchoReplies);
11389 UPDATE_MIB(o1, ipv6IfIcmpInRouterSolicits,
11390 o2->ipv6IfIcmpInRouterSolicits);
11391 UPDATE_MIB(o1, ipv6IfIcmpInRouterAdvertisements,
11392 o2->ipv6IfIcmpInRouterAdvertisements);
11393 UPDATE_MIB(o1, ipv6IfIcmpInNeighborSolicits,
11394 o2->ipv6IfIcmpInNeighborSolicits);
11395 UPDATE_MIB(o1, ipv6IfIcmpInNeighborAdvertisements,
11396 o2->ipv6IfIcmpInNeighborAdvertisements);
11397 UPDATE_MIB(o1, ipv6IfIcmpInRedirects, o2->ipv6IfIcmpInRedirects);
11398 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembQueries,
11399 o2->ipv6IfIcmpInGroupMembQueries);
11400 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembResponses,
11401 o2->ipv6IfIcmpInGroupMembResponses);
11402 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembReductions,
11403 o2->ipv6IfIcmpInGroupMembReductions);
11404 UPDATE_MIB(o1, ipv6IfIcmpOutMsgs, o2->ipv6IfIcmpOutMsgs);
11405 UPDATE_MIB(o1, ipv6IfIcmpOutErrors, o2->ipv6IfIcmpOutErrors);
11406 UPDATE_MIB(o1, ipv6IfIcmpOutDestUnreachs,
11407 o2->ipv6IfIcmpOutDestUnreachs);
11408 UPDATE_MIB(o1, ipv6IfIcmpOutAdminProhibs,
11409 o2->ipv6IfIcmpOutAdminProhibs);
11410 UPDATE_MIB(o1, ipv6IfIcmpOutTimeExcds, o2->ipv6IfIcmpOutTimeExcds);
11411 UPDATE_MIB(o1, ipv6IfIcmpOutParmProblems,
11412 o2->ipv6IfIcmpOutParmProblems);
11413 UPDATE_MIB(o1, ipv6IfIcmpOutPktTooBigs, o2->ipv6IfIcmpOutPktTooBigs);
11414 UPDATE_MIB(o1, ipv6IfIcmpOutEchos, o2->ipv6IfIcmpOutEchos);
11415 UPDATE_MIB(o1, ipv6IfIcmpOutEchoReplies, o2->ipv6IfIcmpOutEchoReplies);
11416 UPDATE_MIB(o1, ipv6IfIcmpOutRouterSolicits,
11417 o2->ipv6IfIcmpOutRouterSolicits);
11418 UPDATE_MIB(o1, ipv6IfIcmpOutRouterAdvertisements,
11419 o2->ipv6IfIcmpOutRouterAdvertisements);
11420 UPDATE_MIB(o1, ipv6IfIcmpOutNeighborSolicits,
11421 o2->ipv6IfIcmpOutNeighborSolicits);
11422 UPDATE_MIB(o1, ipv6IfIcmpOutNeighborAdvertisements,
11423 o2->ipv6IfIcmpOutNeighborAdvertisements);
11424 UPDATE_MIB(o1, ipv6IfIcmpOutRedirects, o2->ipv6IfIcmpOutRedirects);
11425 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembQueries,
11426 o2->ipv6IfIcmpOutGroupMembQueries);
11427 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembResponses,
11428 o2->ipv6IfIcmpOutGroupMembResponses);
11429 UPDATE_MIB(o1, ipv6IfIcmpOutGroupMembReductions,
11430 o2->ipv6IfIcmpOutGroupMembReductions);
11431 UPDATE_MIB(o1, ipv6IfIcmpInOverflows, o2->ipv6IfIcmpInOverflows);
11432 UPDATE_MIB(o1, ipv6IfIcmpBadHoplimit, o2->ipv6IfIcmpBadHoplimit);
11433 UPDATE_MIB(o1, ipv6IfIcmpInBadNeighborAdvertisements,
11434 o2->ipv6IfIcmpInBadNeighborAdvertisements);
11435 UPDATE_MIB(o1, ipv6IfIcmpInBadNeighborSolicitations,
11436 o2->ipv6IfIcmpInBadNeighborSolicitations);
11437 UPDATE_MIB(o1, ipv6IfIcmpInBadRedirects, o2->ipv6IfIcmpInBadRedirects);
11438 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembTotal,
11439 o2->ipv6IfIcmpInGroupMembTotal);
11440 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembBadQueries,
11441 o2->ipv6IfIcmpInGroupMembBadQueries);
11442 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembBadReports,
11443 o2->ipv6IfIcmpInGroupMembBadReports);
11444 UPDATE_MIB(o1, ipv6IfIcmpInGroupMembOurReports,
11445 o2->ipv6IfIcmpInGroupMembOurReports);
11446 }
11447
11448 /*
11449 * Called before the options are updated to check if this packet will
11450 * be source routed from here.
11451 * This routine assumes that the options are well formed i.e. that they
11452 * have already been checked.
11453 */
11454 boolean_t
11455 ip_source_routed(ipha_t *ipha, ip_stack_t *ipst)
11456 {
11457 ipoptp_t opts;
11458 uchar_t *opt;
11459 uint8_t optval;
11460 uint8_t optlen;
11461 ipaddr_t dst;
11462
11463 if (IS_SIMPLE_IPH(ipha)) {
11464 ip2dbg(("not source routed\n"));
11465 return (B_FALSE);
11466 }
11467 dst = ipha->ipha_dst;
11468 for (optval = ipoptp_first(&opts, ipha);
11469 optval != IPOPT_EOL;
11470 optval = ipoptp_next(&opts)) {
11471 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11472 opt = opts.ipoptp_cur;
11473 optlen = opts.ipoptp_len;
11474 ip2dbg(("ip_source_routed: opt %d, len %d\n",
11475 optval, optlen));
11476 switch (optval) {
11477 uint32_t off;
11478 case IPOPT_SSRR:
11479 case IPOPT_LSRR:
11480 /*
11481 * If dst is one of our addresses and there are some
11482 * entries left in the source route return (true).
11483 */
11484 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
11485 ip2dbg(("ip_source_routed: not next"
11486 " source route 0x%x\n",
11487 ntohl(dst)));
11488 return (B_FALSE);
11489 }
11490 off = opt[IPOPT_OFFSET];
11491 off--;
11492 if (optlen < IP_ADDR_LEN ||
11493 off > optlen - IP_ADDR_LEN) {
11494 /* End of source route */
11495 ip1dbg(("ip_source_routed: end of SR\n"));
11496 return (B_FALSE);
11497 }
11498 return (B_TRUE);
11499 }
11500 }
11501 ip2dbg(("not source routed\n"));
11502 return (B_FALSE);
11503 }
11504
11505 /*
11506 * ip_unbind is called by the transports to remove a conn from
11507 * the fanout table.
11508 */
11509 void
11510 ip_unbind(conn_t *connp)
11511 {
11512
11513 ASSERT(!MUTEX_HELD(&connp->conn_lock));
11514
11515 if (is_system_labeled() && connp->conn_anon_port) {
11516 (void) tsol_mlp_anon(crgetzone(connp->conn_cred),
11517 connp->conn_mlp_type, connp->conn_proto,
11518 ntohs(connp->conn_lport), B_FALSE);
11519 connp->conn_anon_port = 0;
11520 }
11521 connp->conn_mlp_type = mlptSingle;
11522
11523 ipcl_hash_remove(connp);
11524 }
11525
11526 /*
11527 * Used for deciding the MSS size for the upper layer. Thus
11528 * we need to check the outbound policy values in the conn.
11529 */
11530 int
11531 conn_ipsec_length(conn_t *connp)
11532 {
11533 ipsec_latch_t *ipl;
11534
11535 ipl = connp->conn_latch;
11536 if (ipl == NULL)
11537 return (0);
11538
11539 if (connp->conn_ixa->ixa_ipsec_policy == NULL)
11540 return (0);
11541
11542 return (connp->conn_ixa->ixa_ipsec_policy->ipsp_act->ipa_ovhd);
11543 }
11544
11545 /*
11546 * Returns an estimate of the IPsec headers size. This is used if
11547 * we don't want to call into IPsec to get the exact size.
11548 */
11549 int
11550 ipsec_out_extra_length(ip_xmit_attr_t *ixa)
11551 {
11552 ipsec_action_t *a;
11553
11554 if (!(ixa->ixa_flags & IXAF_IPSEC_SECURE))
11555 return (0);
11556
11557 a = ixa->ixa_ipsec_action;
11558 if (a == NULL) {
11559 ASSERT(ixa->ixa_ipsec_policy != NULL);
11560 a = ixa->ixa_ipsec_policy->ipsp_act;
11561 }
11562 ASSERT(a != NULL);
11563
11564 return (a->ipa_ovhd);
11565 }
11566
11567 /*
11568 * If there are any source route options, return the true final
11569 * destination. Otherwise, return the destination.
11570 */
11571 ipaddr_t
11572 ip_get_dst(ipha_t *ipha)
11573 {
11574 ipoptp_t opts;
11575 uchar_t *opt;
11576 uint8_t optval;
11577 uint8_t optlen;
11578 ipaddr_t dst;
11579 uint32_t off;
11580
11581 dst = ipha->ipha_dst;
11582
11583 if (IS_SIMPLE_IPH(ipha))
11584 return (dst);
11585
11586 for (optval = ipoptp_first(&opts, ipha);
11587 optval != IPOPT_EOL;
11588 optval = ipoptp_next(&opts)) {
11589 opt = opts.ipoptp_cur;
11590 optlen = opts.ipoptp_len;
11591 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11592 switch (optval) {
11593 case IPOPT_SSRR:
11594 case IPOPT_LSRR:
11595 off = opt[IPOPT_OFFSET];
11596 /*
11597 * If one of the conditions is true, it means
11598 * end of options and dst already has the right
11599 * value.
11600 */
11601 if (!(optlen < IP_ADDR_LEN || off > optlen - 3)) {
11602 off = optlen - IP_ADDR_LEN;
11603 bcopy(&opt[off], &dst, IP_ADDR_LEN);
11604 }
11605 return (dst);
11606 default:
11607 break;
11608 }
11609 }
11610
11611 return (dst);
11612 }
11613
11614 /*
11615 * Outbound IP fragmentation routine.
11616 * Assumes the caller has checked whether or not fragmentation should
11617 * be allowed. Here we copy the DF bit from the header to all the generated
11618 * fragments.
11619 */
11620 int
11621 ip_fragment_v4(mblk_t *mp_orig, nce_t *nce, iaflags_t ixaflags,
11622 uint_t pkt_len, uint32_t max_frag, uint32_t xmit_hint, zoneid_t szone,
11623 zoneid_t nolzid, pfirepostfrag_t postfragfn, uintptr_t *ixa_cookie)
11624 {
11625 int i1;
11626 int hdr_len;
11627 mblk_t *hdr_mp;
11628 ipha_t *ipha;
11629 int ip_data_end;
11630 int len;
11631 mblk_t *mp = mp_orig;
11632 int offset;
11633 ill_t *ill = nce->nce_ill;
11634 ip_stack_t *ipst = ill->ill_ipst;
11635 mblk_t *carve_mp;
11636 uint32_t frag_flag;
11637 uint_t priority = mp->b_band;
11638 int error = 0;
11639
11640 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragReqds);
11641
11642 if (pkt_len != msgdsize(mp)) {
11643 ip0dbg(("Packet length mismatch: %d, %ld\n",
11644 pkt_len, msgdsize(mp)));
11645 freemsg(mp);
11646 return (EINVAL);
11647 }
11648
11649 if (max_frag == 0) {
11650 ip1dbg(("ip_fragment_v4: max_frag is zero. Dropping packet\n"));
11651 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11652 ip_drop_output("FragFails: zero max_frag", mp, ill);
11653 freemsg(mp);
11654 return (EINVAL);
11655 }
11656
11657 ASSERT(MBLKL(mp) >= sizeof (ipha_t));
11658 ipha = (ipha_t *)mp->b_rptr;
11659 ASSERT(ntohs(ipha->ipha_length) == pkt_len);
11660 frag_flag = ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_DF;
11661
11662 /*
11663 * Establish the starting offset. May not be zero if we are fragging
11664 * a fragment that is being forwarded.
11665 */
11666 offset = ntohs(ipha->ipha_fragment_offset_and_flags) & IPH_OFFSET;
11667
11668 /* TODO why is this test needed? */
11669 if (((max_frag - ntohs(ipha->ipha_length)) & ~7) < 8) {
11670 /* TODO: notify ulp somehow */
11671 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11672 ip_drop_output("FragFails: bad starting offset", mp, ill);
11673 freemsg(mp);
11674 return (EINVAL);
11675 }
11676
11677 hdr_len = IPH_HDR_LENGTH(ipha);
11678 ipha->ipha_hdr_checksum = 0;
11679
11680 /*
11681 * Establish the number of bytes maximum per frag, after putting
11682 * in the header.
11683 */
11684 len = (max_frag - hdr_len) & ~7;
11685
11686 /* Get a copy of the header for the trailing frags */
11687 hdr_mp = ip_fragment_copyhdr((uchar_t *)ipha, hdr_len, offset, ipst,
11688 mp);
11689 if (hdr_mp == NULL) {
11690 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11691 ip_drop_output("FragFails: no hdr_mp", mp, ill);
11692 freemsg(mp);
11693 return (ENOBUFS);
11694 }
11695
11696 /* Store the starting offset, with the MoreFrags flag. */
11697 i1 = offset | IPH_MF | frag_flag;
11698 ipha->ipha_fragment_offset_and_flags = htons((uint16_t)i1);
11699
11700 /* Establish the ending byte offset, based on the starting offset. */
11701 offset <<= 3;
11702 ip_data_end = offset + ntohs(ipha->ipha_length) - hdr_len;
11703
11704 /* Store the length of the first fragment in the IP header. */
11705 i1 = len + hdr_len;
11706 ASSERT(i1 <= IP_MAXPACKET);
11707 ipha->ipha_length = htons((uint16_t)i1);
11708
11709 /*
11710 * Compute the IP header checksum for the first frag. We have to
11711 * watch out that we stop at the end of the header.
11712 */
11713 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
11714
11715 /*
11716 * Now carve off the first frag. Note that this will include the
11717 * original IP header.
11718 */
11719 if (!(mp = ip_carve_mp(&mp_orig, i1))) {
11720 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11721 ip_drop_output("FragFails: could not carve mp", mp_orig, ill);
11722 freeb(hdr_mp);
11723 freemsg(mp_orig);
11724 return (ENOBUFS);
11725 }
11726
11727 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragCreates);
11728
11729 error = postfragfn(mp, nce, ixaflags, i1, xmit_hint, szone, nolzid,
11730 ixa_cookie);
11731 if (error != 0 && error != EWOULDBLOCK) {
11732 /* No point in sending the other fragments */
11733 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11734 ip_drop_output("FragFails: postfragfn failed", mp_orig, ill);
11735 freeb(hdr_mp);
11736 freemsg(mp_orig);
11737 return (error);
11738 }
11739
11740 /* No need to redo state machine in loop */
11741 ixaflags &= ~IXAF_REACH_CONF;
11742
11743 /* Advance the offset to the second frag starting point. */
11744 offset += len;
11745 /*
11746 * Update hdr_len from the copied header - there might be less options
11747 * in the later fragments.
11748 */
11749 hdr_len = IPH_HDR_LENGTH(hdr_mp->b_rptr);
11750 /* Loop until done. */
11751 for (;;) {
11752 uint16_t offset_and_flags;
11753 uint16_t ip_len;
11754
11755 if (ip_data_end - offset > len) {
11756 /*
11757 * Carve off the appropriate amount from the original
11758 * datagram.
11759 */
11760 if (!(carve_mp = ip_carve_mp(&mp_orig, len))) {
11761 mp = NULL;
11762 break;
11763 }
11764 /*
11765 * More frags after this one. Get another copy
11766 * of the header.
11767 */
11768 if (carve_mp->b_datap->db_ref == 1 &&
11769 hdr_mp->b_wptr - hdr_mp->b_rptr <
11770 carve_mp->b_rptr - carve_mp->b_datap->db_base) {
11771 /* Inline IP header */
11772 carve_mp->b_rptr -= hdr_mp->b_wptr -
11773 hdr_mp->b_rptr;
11774 bcopy(hdr_mp->b_rptr, carve_mp->b_rptr,
11775 hdr_mp->b_wptr - hdr_mp->b_rptr);
11776 mp = carve_mp;
11777 } else {
11778 if (!(mp = copyb(hdr_mp))) {
11779 freemsg(carve_mp);
11780 break;
11781 }
11782 /* Get priority marking, if any. */
11783 mp->b_band = priority;
11784 mp->b_cont = carve_mp;
11785 }
11786 ipha = (ipha_t *)mp->b_rptr;
11787 offset_and_flags = IPH_MF;
11788 } else {
11789 /*
11790 * Last frag. Consume the header. Set len to
11791 * the length of this last piece.
11792 */
11793 len = ip_data_end - offset;
11794
11795 /*
11796 * Carve off the appropriate amount from the original
11797 * datagram.
11798 */
11799 if (!(carve_mp = ip_carve_mp(&mp_orig, len))) {
11800 mp = NULL;
11801 break;
11802 }
11803 if (carve_mp->b_datap->db_ref == 1 &&
11804 hdr_mp->b_wptr - hdr_mp->b_rptr <
11805 carve_mp->b_rptr - carve_mp->b_datap->db_base) {
11806 /* Inline IP header */
11807 carve_mp->b_rptr -= hdr_mp->b_wptr -
11808 hdr_mp->b_rptr;
11809 bcopy(hdr_mp->b_rptr, carve_mp->b_rptr,
11810 hdr_mp->b_wptr - hdr_mp->b_rptr);
11811 mp = carve_mp;
11812 freeb(hdr_mp);
11813 hdr_mp = mp;
11814 } else {
11815 mp = hdr_mp;
11816 /* Get priority marking, if any. */
11817 mp->b_band = priority;
11818 mp->b_cont = carve_mp;
11819 }
11820 ipha = (ipha_t *)mp->b_rptr;
11821 /* A frag of a frag might have IPH_MF non-zero */
11822 offset_and_flags =
11823 ntohs(ipha->ipha_fragment_offset_and_flags) &
11824 IPH_MF;
11825 }
11826 offset_and_flags |= (uint16_t)(offset >> 3);
11827 offset_and_flags |= (uint16_t)frag_flag;
11828 /* Store the offset and flags in the IP header. */
11829 ipha->ipha_fragment_offset_and_flags = htons(offset_and_flags);
11830
11831 /* Store the length in the IP header. */
11832 ip_len = (uint16_t)(len + hdr_len);
11833 ipha->ipha_length = htons(ip_len);
11834
11835 /*
11836 * Set the IP header checksum. Note that mp is just
11837 * the header, so this is easy to pass to ip_csum.
11838 */
11839 ipha->ipha_hdr_checksum = ip_csum_hdr(ipha);
11840
11841 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragCreates);
11842
11843 error = postfragfn(mp, nce, ixaflags, ip_len, xmit_hint, szone,
11844 nolzid, ixa_cookie);
11845 /* All done if we just consumed the hdr_mp. */
11846 if (mp == hdr_mp) {
11847 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragOKs);
11848 return (error);
11849 }
11850 if (error != 0 && error != EWOULDBLOCK) {
11851 DTRACE_PROBE2(ip__xmit__frag__fail, ill_t *, ill,
11852 mblk_t *, hdr_mp);
11853 /* No point in sending the other fragments */
11854 break;
11855 }
11856
11857 /* Otherwise, advance and loop. */
11858 offset += len;
11859 }
11860 /* Clean up following allocation failure. */
11861 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutFragFails);
11862 ip_drop_output("FragFails: loop ended", NULL, ill);
11863 if (mp != hdr_mp)
11864 freeb(hdr_mp);
11865 if (mp != mp_orig)
11866 freemsg(mp_orig);
11867 return (error);
11868 }
11869
11870 /*
11871 * Copy the header plus those options which have the copy bit set
11872 */
11873 static mblk_t *
11874 ip_fragment_copyhdr(uchar_t *rptr, int hdr_len, int offset, ip_stack_t *ipst,
11875 mblk_t *src)
11876 {
11877 mblk_t *mp;
11878 uchar_t *up;
11879
11880 /*
11881 * Quick check if we need to look for options without the copy bit
11882 * set
11883 */
11884 mp = allocb_tmpl(ipst->ips_ip_wroff_extra + hdr_len, src);
11885 if (!mp)
11886 return (mp);
11887 mp->b_rptr += ipst->ips_ip_wroff_extra;
11888 if (hdr_len == IP_SIMPLE_HDR_LENGTH || offset != 0) {
11889 bcopy(rptr, mp->b_rptr, hdr_len);
11890 mp->b_wptr += hdr_len + ipst->ips_ip_wroff_extra;
11891 return (mp);
11892 }
11893 up = mp->b_rptr;
11894 bcopy(rptr, up, IP_SIMPLE_HDR_LENGTH);
11895 up += IP_SIMPLE_HDR_LENGTH;
11896 rptr += IP_SIMPLE_HDR_LENGTH;
11897 hdr_len -= IP_SIMPLE_HDR_LENGTH;
11898 while (hdr_len > 0) {
11899 uint32_t optval;
11900 uint32_t optlen;
11901
11902 optval = *rptr;
11903 if (optval == IPOPT_EOL)
11904 break;
11905 if (optval == IPOPT_NOP)
11906 optlen = 1;
11907 else
11908 optlen = rptr[1];
11909 if (optval & IPOPT_COPY) {
11910 bcopy(rptr, up, optlen);
11911 up += optlen;
11912 }
11913 rptr += optlen;
11914 hdr_len -= optlen;
11915 }
11916 /*
11917 * Make sure that we drop an even number of words by filling
11918 * with EOL to the next word boundary.
11919 */
11920 for (hdr_len = up - (mp->b_rptr + IP_SIMPLE_HDR_LENGTH);
11921 hdr_len & 0x3; hdr_len++)
11922 *up++ = IPOPT_EOL;
11923 mp->b_wptr = up;
11924 /* Update header length */
11925 mp->b_rptr[0] = (uint8_t)((IP_VERSION << 4) | ((up - mp->b_rptr) >> 2));
11926 return (mp);
11927 }
11928
11929 /*
11930 * Update any source route, record route, or timestamp options when
11931 * sending a packet back to ourselves.
11932 * Check that we are at end of strict source route.
11933 * The options have been sanity checked by ip_output_options().
11934 */
11935 void
11936 ip_output_local_options(ipha_t *ipha, ip_stack_t *ipst)
11937 {
11938 ipoptp_t opts;
11939 uchar_t *opt;
11940 uint8_t optval;
11941 uint8_t optlen;
11942 ipaddr_t dst;
11943 uint32_t ts;
11944 timestruc_t now;
11945
11946 for (optval = ipoptp_first(&opts, ipha);
11947 optval != IPOPT_EOL;
11948 optval = ipoptp_next(&opts)) {
11949 opt = opts.ipoptp_cur;
11950 optlen = opts.ipoptp_len;
11951 ASSERT((opts.ipoptp_flags & IPOPTP_ERROR) == 0);
11952 switch (optval) {
11953 uint32_t off;
11954 case IPOPT_SSRR:
11955 case IPOPT_LSRR:
11956 off = opt[IPOPT_OFFSET];
11957 off--;
11958 if (optlen < IP_ADDR_LEN ||
11959 off > optlen - IP_ADDR_LEN) {
11960 /* End of source route */
11961 break;
11962 }
11963 /*
11964 * This will only happen if two consecutive entries
11965 * in the source route contains our address or if
11966 * it is a packet with a loose source route which
11967 * reaches us before consuming the whole source route
11968 */
11969
11970 if (optval == IPOPT_SSRR) {
11971 return;
11972 }
11973 /*
11974 * Hack: instead of dropping the packet truncate the
11975 * source route to what has been used by filling the
11976 * rest with IPOPT_NOP.
11977 */
11978 opt[IPOPT_OLEN] = (uint8_t)off;
11979 while (off < optlen) {
11980 opt[off++] = IPOPT_NOP;
11981 }
11982 break;
11983 case IPOPT_RR:
11984 off = opt[IPOPT_OFFSET];
11985 off--;
11986 if (optlen < IP_ADDR_LEN ||
11987 off > optlen - IP_ADDR_LEN) {
11988 /* No more room - ignore */
11989 ip1dbg((
11990 "ip_output_local_options: end of RR\n"));
11991 break;
11992 }
11993 dst = htonl(INADDR_LOOPBACK);
11994 bcopy(&dst, (char *)opt + off, IP_ADDR_LEN);
11995 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
11996 break;
11997 case IPOPT_TS:
11998 /* Insert timestamp if there is romm */
11999 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
12000 case IPOPT_TS_TSONLY:
12001 off = IPOPT_TS_TIMELEN;
12002 break;
12003 case IPOPT_TS_PRESPEC:
12004 case IPOPT_TS_PRESPEC_RFC791:
12005 /* Verify that the address matched */
12006 off = opt[IPOPT_OFFSET] - 1;
12007 bcopy((char *)opt + off, &dst, IP_ADDR_LEN);
12008 if (ip_type_v4(dst, ipst) != IRE_LOCAL) {
12009 /* Not for us */
12010 break;
12011 }
12012 /* FALLTHRU */
12013 case IPOPT_TS_TSANDADDR:
12014 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
12015 break;
12016 default:
12017 /*
12018 * ip_*put_options should have already
12019 * dropped this packet.
12020 */
12021 cmn_err(CE_PANIC, "ip_output_local_options: "
12022 "unknown IT - bug in ip_output_options?\n");
12023 return; /* Keep "lint" happy */
12024 }
12025 if (opt[IPOPT_OFFSET] - 1 + off > optlen) {
12026 /* Increase overflow counter */
12027 off = (opt[IPOPT_POS_OV_FLG] >> 4) + 1;
12028 opt[IPOPT_POS_OV_FLG] = (uint8_t)
12029 (opt[IPOPT_POS_OV_FLG] & 0x0F) |
12030 (off << 4);
12031 break;
12032 }
12033 off = opt[IPOPT_OFFSET] - 1;
12034 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
12035 case IPOPT_TS_PRESPEC:
12036 case IPOPT_TS_PRESPEC_RFC791:
12037 case IPOPT_TS_TSANDADDR:
12038 dst = htonl(INADDR_LOOPBACK);
12039 bcopy(&dst, (char *)opt + off, IP_ADDR_LEN);
12040 opt[IPOPT_OFFSET] += IP_ADDR_LEN;
12041 /* FALLTHRU */
12042 case IPOPT_TS_TSONLY:
12043 off = opt[IPOPT_OFFSET] - 1;
12044 /* Compute # of milliseconds since midnight */
12045 gethrestime(&now);
12046 ts = (now.tv_sec % (24 * 60 * 60)) * 1000 +
12047 NSEC2MSEC(now.tv_nsec);
12048 bcopy(&ts, (char *)opt + off, IPOPT_TS_TIMELEN);
12049 opt[IPOPT_OFFSET] += IPOPT_TS_TIMELEN;
12050 break;
12051 }
12052 break;
12053 }
12054 }
12055 }
12056
12057 /*
12058 * Prepend an M_DATA fastpath header, and if none present prepend a
12059 * DL_UNITDATA_REQ. Frees the mblk on failure.
12060 *
12061 * nce_dlur_mp and nce_fp_mp can not disappear once they have been set.
12062 * If there is a change to them, the nce will be deleted (condemned) and
12063 * a new nce_t will be created when packets are sent. Thus we need no locks
12064 * to access those fields.
12065 *
12066 * We preserve b_band to support IPQoS. If a DL_UNITDATA_REQ is prepended
12067 * we place b_band in dl_priority.dl_max.
12068 */
12069 static mblk_t *
12070 ip_xmit_attach_llhdr(mblk_t *mp, nce_t *nce)
12071 {
12072 uint_t hlen;
12073 mblk_t *mp1;
12074 uint_t priority;
12075 uchar_t *rptr;
12076
12077 rptr = mp->b_rptr;
12078
12079 ASSERT(DB_TYPE(mp) == M_DATA);
12080 priority = mp->b_band;
12081
12082 ASSERT(nce != NULL);
12083 if ((mp1 = nce->nce_fp_mp) != NULL) {
12084 hlen = MBLKL(mp1);
12085 /*
12086 * Check if we have enough room to prepend fastpath
12087 * header
12088 */
12089 if (hlen != 0 && (rptr - mp->b_datap->db_base) >= hlen) {
12090 rptr -= hlen;
12091 bcopy(mp1->b_rptr, rptr, hlen);
12092 /*
12093 * Set the b_rptr to the start of the link layer
12094 * header
12095 */
12096 mp->b_rptr = rptr;
12097 return (mp);
12098 }
12099 mp1 = copyb(mp1);
12100 if (mp1 == NULL) {
12101 ill_t *ill = nce->nce_ill;
12102
12103 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12104 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12105 freemsg(mp);
12106 return (NULL);
12107 }
12108 mp1->b_band = priority;
12109 mp1->b_cont = mp;
12110 DB_CKSUMSTART(mp1) = DB_CKSUMSTART(mp);
12111 DB_CKSUMSTUFF(mp1) = DB_CKSUMSTUFF(mp);
12112 DB_CKSUMEND(mp1) = DB_CKSUMEND(mp);
12113 DB_CKSUMFLAGS(mp1) = DB_CKSUMFLAGS(mp);
12114 DB_LSOMSS(mp1) = DB_LSOMSS(mp);
12115 DTRACE_PROBE1(ip__xmit__copyb, (mblk_t *), mp1);
12116 /*
12117 * XXX disable ICK_VALID and compute checksum
12118 * here; can happen if nce_fp_mp changes and
12119 * it can't be copied now due to insufficient
12120 * space. (unlikely, fp mp can change, but it
12121 * does not increase in length)
12122 */
12123 return (mp1);
12124 }
12125 mp1 = copyb(nce->nce_dlur_mp);
12126
12127 if (mp1 == NULL) {
12128 ill_t *ill = nce->nce_ill;
12129
12130 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12131 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12132 freemsg(mp);
12133 return (NULL);
12134 }
12135 mp1->b_cont = mp;
12136 if (priority != 0) {
12137 mp1->b_band = priority;
12138 ((dl_unitdata_req_t *)(mp1->b_rptr))->dl_priority.dl_max =
12139 priority;
12140 }
12141 return (mp1);
12142 }
12143
12144 /*
12145 * Finish the outbound IPsec processing. This function is called from
12146 * ipsec_out_process() if the IPsec packet was processed
12147 * synchronously, or from {ah,esp}_kcf_callback_outbound() if it was processed
12148 * asynchronously.
12149 *
12150 * This is common to IPv4 and IPv6.
12151 */
12152 int
12153 ip_output_post_ipsec(mblk_t *mp, ip_xmit_attr_t *ixa)
12154 {
12155 iaflags_t ixaflags = ixa->ixa_flags;
12156 uint_t pktlen;
12157
12158
12159 /* AH/ESP don't update ixa_pktlen when they modify the packet */
12160 if (ixaflags & IXAF_IS_IPV4) {
12161 ipha_t *ipha = (ipha_t *)mp->b_rptr;
12162
12163 ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
12164 pktlen = ntohs(ipha->ipha_length);
12165 } else {
12166 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
12167
12168 ASSERT(IPH_HDR_VERSION(mp->b_rptr) == IPV6_VERSION);
12169 pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
12170 }
12171
12172 /*
12173 * We release any hard reference on the SAs here to make
12174 * sure the SAs can be garbage collected. ipsr_sa has a soft reference
12175 * on the SAs.
12176 * If in the future we want the hard latching of the SAs in the
12177 * ip_xmit_attr_t then we should remove this.
12178 */
12179 if (ixa->ixa_ipsec_esp_sa != NULL) {
12180 IPSA_REFRELE(ixa->ixa_ipsec_esp_sa);
12181 ixa->ixa_ipsec_esp_sa = NULL;
12182 }
12183 if (ixa->ixa_ipsec_ah_sa != NULL) {
12184 IPSA_REFRELE(ixa->ixa_ipsec_ah_sa);
12185 ixa->ixa_ipsec_ah_sa = NULL;
12186 }
12187
12188 /* Do we need to fragment? */
12189 if ((ixa->ixa_flags & IXAF_IPV6_ADD_FRAGHDR) ||
12190 pktlen > ixa->ixa_fragsize) {
12191 if (ixaflags & IXAF_IS_IPV4) {
12192 ASSERT(!(ixa->ixa_flags & IXAF_IPV6_ADD_FRAGHDR));
12193 /*
12194 * We check for the DF case in ipsec_out_process
12195 * hence this only handles the non-DF case.
12196 */
12197 return (ip_fragment_v4(mp, ixa->ixa_nce, ixa->ixa_flags,
12198 pktlen, ixa->ixa_fragsize,
12199 ixa->ixa_xmit_hint, ixa->ixa_zoneid,
12200 ixa->ixa_no_loop_zoneid, ixa->ixa_postfragfn,
12201 &ixa->ixa_cookie));
12202 } else {
12203 mp = ip_fraghdr_add_v6(mp, ixa->ixa_ident, ixa);
12204 if (mp == NULL) {
12205 /* MIB and ip_drop_output already done */
12206 return (ENOMEM);
12207 }
12208 pktlen += sizeof (ip6_frag_t);
12209 if (pktlen > ixa->ixa_fragsize) {
12210 return (ip_fragment_v6(mp, ixa->ixa_nce,
12211 ixa->ixa_flags, pktlen,
12212 ixa->ixa_fragsize, ixa->ixa_xmit_hint,
12213 ixa->ixa_zoneid, ixa->ixa_no_loop_zoneid,
12214 ixa->ixa_postfragfn, &ixa->ixa_cookie));
12215 }
12216 }
12217 }
12218 return ((ixa->ixa_postfragfn)(mp, ixa->ixa_nce, ixa->ixa_flags,
12219 pktlen, ixa->ixa_xmit_hint, ixa->ixa_zoneid,
12220 ixa->ixa_no_loop_zoneid, NULL));
12221 }
12222
12223 /*
12224 * Finish the inbound IPsec processing. This function is called from
12225 * ipsec_out_process() if the IPsec packet was processed
12226 * synchronously, or from {ah,esp}_kcf_callback_outbound() if it was processed
12227 * asynchronously.
12228 *
12229 * This is common to IPv4 and IPv6.
12230 */
12231 void
12232 ip_input_post_ipsec(mblk_t *mp, ip_recv_attr_t *ira)
12233 {
12234 iaflags_t iraflags = ira->ira_flags;
12235
12236 /* Length might have changed */
12237 if (iraflags & IRAF_IS_IPV4) {
12238 ipha_t *ipha = (ipha_t *)mp->b_rptr;
12239
12240 ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION);
12241 ira->ira_pktlen = ntohs(ipha->ipha_length);
12242 ira->ira_ip_hdr_length = IPH_HDR_LENGTH(ipha);
12243 ira->ira_protocol = ipha->ipha_protocol;
12244
12245 ip_fanout_v4(mp, ipha, ira);
12246 } else {
12247 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
12248 uint8_t *nexthdrp;
12249
12250 ASSERT(IPH_HDR_VERSION(mp->b_rptr) == IPV6_VERSION);
12251 ira->ira_pktlen = ntohs(ip6h->ip6_plen) + IPV6_HDR_LEN;
12252 if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &ira->ira_ip_hdr_length,
12253 &nexthdrp)) {
12254 /* Malformed packet */
12255 BUMP_MIB(ira->ira_ill->ill_ip_mib, ipIfStatsInDiscards);
12256 ip_drop_input("ipIfStatsInDiscards", mp, ira->ira_ill);
12257 freemsg(mp);
12258 return;
12259 }
12260 ira->ira_protocol = *nexthdrp;
12261 ip_fanout_v6(mp, ip6h, ira);
12262 }
12263 }
12264
12265 /*
12266 * Select which AH & ESP SA's to use (if any) for the outbound packet.
12267 *
12268 * If this function returns B_TRUE, the requested SA's have been filled
12269 * into the ixa_ipsec_*_sa pointers.
12270 *
12271 * If the function returns B_FALSE, the packet has been "consumed", most
12272 * likely by an ACQUIRE sent up via PF_KEY to a key management daemon.
12273 *
12274 * The SA references created by the protocol-specific "select"
12275 * function will be released in ip_output_post_ipsec.
12276 */
12277 static boolean_t
12278 ipsec_out_select_sa(mblk_t *mp, ip_xmit_attr_t *ixa)
12279 {
12280 boolean_t need_ah_acquire = B_FALSE, need_esp_acquire = B_FALSE;
12281 ipsec_policy_t *pp;
12282 ipsec_action_t *ap;
12283
12284 ASSERT(ixa->ixa_flags & IXAF_IPSEC_SECURE);
12285 ASSERT((ixa->ixa_ipsec_policy != NULL) ||
12286 (ixa->ixa_ipsec_action != NULL));
12287
12288 ap = ixa->ixa_ipsec_action;
12289 if (ap == NULL) {
12290 pp = ixa->ixa_ipsec_policy;
12291 ASSERT(pp != NULL);
12292 ap = pp->ipsp_act;
12293 ASSERT(ap != NULL);
12294 }
12295
12296 /*
12297 * We have an action. now, let's select SA's.
12298 * A side effect of setting ixa_ipsec_*_sa is that it will
12299 * be cached in the conn_t.
12300 */
12301 if (ap->ipa_want_esp) {
12302 if (ixa->ixa_ipsec_esp_sa == NULL) {
12303 need_esp_acquire = !ipsec_outbound_sa(mp, ixa,
12304 IPPROTO_ESP);
12305 }
12306 ASSERT(need_esp_acquire || ixa->ixa_ipsec_esp_sa != NULL);
12307 }
12308
12309 if (ap->ipa_want_ah) {
12310 if (ixa->ixa_ipsec_ah_sa == NULL) {
12311 need_ah_acquire = !ipsec_outbound_sa(mp, ixa,
12312 IPPROTO_AH);
12313 }
12314 ASSERT(need_ah_acquire || ixa->ixa_ipsec_ah_sa != NULL);
12315 /*
12316 * The ESP and AH processing order needs to be preserved
12317 * when both protocols are required (ESP should be applied
12318 * before AH for an outbound packet). Force an ESP ACQUIRE
12319 * when both ESP and AH are required, and an AH ACQUIRE
12320 * is needed.
12321 */
12322 if (ap->ipa_want_esp && need_ah_acquire)
12323 need_esp_acquire = B_TRUE;
12324 }
12325
12326 /*
12327 * Send an ACQUIRE (extended, regular, or both) if we need one.
12328 * Release SAs that got referenced, but will not be used until we
12329 * acquire _all_ of the SAs we need.
12330 */
12331 if (need_ah_acquire || need_esp_acquire) {
12332 if (ixa->ixa_ipsec_ah_sa != NULL) {
12333 IPSA_REFRELE(ixa->ixa_ipsec_ah_sa);
12334 ixa->ixa_ipsec_ah_sa = NULL;
12335 }
12336 if (ixa->ixa_ipsec_esp_sa != NULL) {
12337 IPSA_REFRELE(ixa->ixa_ipsec_esp_sa);
12338 ixa->ixa_ipsec_esp_sa = NULL;
12339 }
12340
12341 sadb_acquire(mp, ixa, need_ah_acquire, need_esp_acquire);
12342 return (B_FALSE);
12343 }
12344
12345 return (B_TRUE);
12346 }
12347
12348 /*
12349 * Handle IPsec output processing.
12350 * This function is only entered once for a given packet.
12351 * We try to do things synchronously, but if we need to have user-level
12352 * set up SAs, or ESP or AH uses asynchronous kEF, then the operation
12353 * will be completed
12354 * - when the SAs are added in esp_add_sa_finish/ah_add_sa_finish
12355 * - when asynchronous ESP is done it will do AH
12356 *
12357 * In all cases we come back in ip_output_post_ipsec() to fragment and
12358 * send out the packet.
12359 */
12360 int
12361 ipsec_out_process(mblk_t *mp, ip_xmit_attr_t *ixa)
12362 {
12363 ill_t *ill = ixa->ixa_nce->nce_ill;
12364 ip_stack_t *ipst = ixa->ixa_ipst;
12365 ipsec_stack_t *ipss;
12366 ipsec_policy_t *pp;
12367 ipsec_action_t *ap;
12368
12369 ASSERT(ixa->ixa_flags & IXAF_IPSEC_SECURE);
12370
12371 ASSERT((ixa->ixa_ipsec_policy != NULL) ||
12372 (ixa->ixa_ipsec_action != NULL));
12373
12374 ipss = ipst->ips_netstack->netstack_ipsec;
12375 if (!ipsec_loaded(ipss)) {
12376 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12377 ip_drop_packet(mp, B_TRUE, ill,
12378 DROPPER(ipss, ipds_ip_ipsec_not_loaded),
12379 &ipss->ipsec_dropper);
12380 return (ENOTSUP);
12381 }
12382
12383 ap = ixa->ixa_ipsec_action;
12384 if (ap == NULL) {
12385 pp = ixa->ixa_ipsec_policy;
12386 ASSERT(pp != NULL);
12387 ap = pp->ipsp_act;
12388 ASSERT(ap != NULL);
12389 }
12390
12391 /* Handle explicit drop action and bypass. */
12392 switch (ap->ipa_act.ipa_type) {
12393 case IPSEC_ACT_DISCARD:
12394 case IPSEC_ACT_REJECT:
12395 ip_drop_packet(mp, B_FALSE, ill,
12396 DROPPER(ipss, ipds_spd_explicit), &ipss->ipsec_spd_dropper);
12397 return (EHOSTUNREACH); /* IPsec policy failure */
12398 case IPSEC_ACT_BYPASS:
12399 return (ip_output_post_ipsec(mp, ixa));
12400 }
12401
12402 /*
12403 * The order of processing is first insert a IP header if needed.
12404 * Then insert the ESP header and then the AH header.
12405 */
12406 if ((ixa->ixa_flags & IXAF_IS_IPV4) && ap->ipa_want_se) {
12407 /*
12408 * First get the outer IP header before sending
12409 * it to ESP.
12410 */
12411 ipha_t *oipha, *iipha;
12412 mblk_t *outer_mp, *inner_mp;
12413
12414 if ((outer_mp = allocb(sizeof (ipha_t), BPRI_HI)) == NULL) {
12415 (void) mi_strlog(ill->ill_rq, 0,
12416 SL_ERROR|SL_TRACE|SL_CONSOLE,
12417 "ipsec_out_process: "
12418 "Self-Encapsulation failed: Out of memory\n");
12419 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
12420 ip_drop_output("ipIfStatsOutDiscards", mp, ill);
12421 freemsg(mp);
12422 return (ENOBUFS);
12423 }
12424 inner_mp = mp;
12425 ASSERT(inner_mp->b_datap->db_type == M_DATA);
12426 oipha = (ipha_t *)outer_mp->b_rptr;
12427 iipha = (ipha_t *)inner_mp->b_rptr;
12428 *oipha = *iipha;
12429 outer_mp->b_wptr += sizeof (ipha_t);
12430 oipha->ipha_length = htons(ntohs(iipha->ipha_length) +
12431 sizeof (ipha_t));
12432 oipha->ipha_protocol = IPPROTO_ENCAP;
12433 oipha->ipha_version_and_hdr_length =
12434 IP_SIMPLE_HDR_VERSION;
12435 oipha->ipha_hdr_checksum = 0;
12436 oipha->ipha_hdr_checksum = ip_csum_hdr(oipha);
12437 outer_mp->b_cont = inner_mp;
12438 mp = outer_mp;
12439
12440 ixa->ixa_flags |= IXAF_IPSEC_TUNNEL;
12441 }
12442
12443 /* If we need to wait for a SA then we can't return any errno */
12444 if (((ap->ipa_want_ah && (ixa->ixa_ipsec_ah_sa == NULL)) ||
12445 (ap->ipa_want_esp && (ixa->ixa_ipsec_esp_sa == NULL))) &&
12446 !ipsec_out_select_sa(mp, ixa))
12447 return (0);
12448
12449 /*
12450 * By now, we know what SA's to use. Toss over to ESP & AH
12451 * to do the heavy lifting.
12452 */
12453 if (ap->ipa_want_esp) {
12454 ASSERT(ixa->ixa_ipsec_esp_sa != NULL);
12455
12456 mp = ixa->ixa_ipsec_esp_sa->ipsa_output_func(mp, ixa);
12457 if (mp == NULL) {
12458 /*
12459 * Either it failed or is pending. In the former case
12460 * ipIfStatsInDiscards was increased.
12461 */
12462 return (0);
12463 }
12464 }
12465
12466 if (ap->ipa_want_ah) {
12467 ASSERT(ixa->ixa_ipsec_ah_sa != NULL);
12468
12469 mp = ixa->ixa_ipsec_ah_sa->ipsa_output_func(mp, ixa);
12470 if (mp == NULL) {
12471 /*
12472 * Either it failed or is pending. In the former case
12473 * ipIfStatsInDiscards was increased.
12474 */
12475 return (0);
12476 }
12477 }
12478 /*
12479 * We are done with IPsec processing. Send it over
12480 * the wire.
12481 */
12482 return (ip_output_post_ipsec(mp, ixa));
12483 }
12484
12485 /*
12486 * ioctls that go through a down/up sequence may need to wait for the down
12487 * to complete. This involves waiting for the ire and ipif refcnts to go down
12488 * to zero. Subsequently the ioctl is restarted from ipif_ill_refrele_tail.
12489 */
12490 /* ARGSUSED */
12491 void
12492 ip_reprocess_ioctl(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *dummy_arg)
12493 {
12494 struct iocblk *iocp;
12495 mblk_t *mp1;
12496 ip_ioctl_cmd_t *ipip;
12497 int err;
12498 sin_t *sin;
12499 struct lifreq *lifr;
12500 struct ifreq *ifr;
12501
12502 iocp = (struct iocblk *)mp->b_rptr;
12503 ASSERT(ipsq != NULL);
12504 /* Existence of mp1 verified in ip_wput_nondata */
12505 mp1 = mp->b_cont->b_cont;
12506 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12507 if (ipip->ipi_cmd == SIOCSLIFNAME || ipip->ipi_cmd == IF_UNITSEL) {
12508 /*
12509 * Special case where ipx_current_ipif is not set:
12510 * ill_phyint_reinit merged the v4 and v6 into a single ipsq.
12511 * We are here as were not able to complete the operation in
12512 * ipif_set_values because we could not become exclusive on
12513 * the new ipsq.
12514 */
12515 ill_t *ill = q->q_ptr;
12516 ipsq_current_start(ipsq, ill->ill_ipif, ipip->ipi_cmd);
12517 }
12518 ASSERT(ipsq->ipsq_xop->ipx_current_ipif != NULL);
12519
12520 if (ipip->ipi_cmd_type == IF_CMD) {
12521 /* This a old style SIOC[GS]IF* command */
12522 ifr = (struct ifreq *)mp1->b_rptr;
12523 sin = (sin_t *)&ifr->ifr_addr;
12524 } else if (ipip->ipi_cmd_type == LIF_CMD) {
12525 /* This a new style SIOC[GS]LIF* command */
12526 lifr = (struct lifreq *)mp1->b_rptr;
12527 sin = (sin_t *)&lifr->lifr_addr;
12528 } else {
12529 sin = NULL;
12530 }
12531
12532 err = (*ipip->ipi_func_restart)(ipsq->ipsq_xop->ipx_current_ipif, sin,
12533 q, mp, ipip, mp1->b_rptr);
12534
12535 DTRACE_PROBE4(ipif__ioctl, char *, "ip_reprocess_ioctl finish",
12536 int, ipip->ipi_cmd,
12537 ill_t *, ipsq->ipsq_xop->ipx_current_ipif->ipif_ill,
12538 ipif_t *, ipsq->ipsq_xop->ipx_current_ipif);
12539
12540 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), ipsq);
12541 }
12542
12543 /*
12544 * ioctl processing
12545 *
12546 * ioctl processing starts with ip_sioctl_copyin_setup(), which looks up
12547 * the ioctl command in the ioctl tables, determines the copyin data size
12548 * from the ipi_copyin_size field, and does an mi_copyin() of that size.
12549 *
12550 * ioctl processing then continues when the M_IOCDATA makes its way down to
12551 * ip_wput_nondata(). The ioctl is looked up again in the ioctl table, its
12552 * associated 'conn' is refheld till the end of the ioctl and the general
12553 * ioctl processing function ip_process_ioctl() is called to extract the
12554 * arguments and process the ioctl. To simplify extraction, ioctl commands
12555 * are "typed" based on the arguments they take (e.g., LIF_CMD which takes a
12556 * `struct lifreq'), and a common extract function (e.g., ip_extract_lifreq())
12557 * is used to extract the ioctl's arguments.
12558 *
12559 * ip_process_ioctl determines if the ioctl needs to be serialized, and if
12560 * so goes thru the serialization primitive ipsq_try_enter. Then the
12561 * appropriate function to handle the ioctl is called based on the entry in
12562 * the ioctl table. ioctl completion is encapsulated in ip_ioctl_finish
12563 * which also refreleases the 'conn' that was refheld at the start of the
12564 * ioctl. Finally ipsq_exit is called if needed to exit the ipsq.
12565 *
12566 * Many exclusive ioctls go thru an internal down up sequence as part of
12567 * the operation. For example an attempt to change the IP address of an
12568 * ipif entails ipif_down, set address, ipif_up. Bringing down the interface
12569 * does all the cleanup such as deleting all ires that use this address.
12570 * Then we need to wait till all references to the interface go away.
12571 */
12572 void
12573 ip_process_ioctl(ipsq_t *ipsq, queue_t *q, mblk_t *mp, void *arg)
12574 {
12575 struct iocblk *iocp = (struct iocblk *)mp->b_rptr;
12576 ip_ioctl_cmd_t *ipip = arg;
12577 ip_extract_func_t *extract_funcp;
12578 cmd_info_t ci;
12579 int err;
12580 boolean_t entered_ipsq = B_FALSE;
12581
12582 ip3dbg(("ip_process_ioctl: ioctl %X\n", iocp->ioc_cmd));
12583
12584 if (ipip == NULL)
12585 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12586
12587 /*
12588 * SIOCLIFADDIF needs to go thru a special path since the
12589 * ill may not exist yet. This happens in the case of lo0
12590 * which is created using this ioctl.
12591 */
12592 if (ipip->ipi_cmd == SIOCLIFADDIF) {
12593 err = ip_sioctl_addif(NULL, NULL, q, mp, NULL, NULL);
12594 DTRACE_PROBE4(ipif__ioctl, char *, "ip_process_ioctl finish",
12595 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12596 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12597 return;
12598 }
12599
12600 ci.ci_ipif = NULL;
12601 switch (ipip->ipi_cmd_type) {
12602 case MISC_CMD:
12603 case MSFILT_CMD:
12604 /*
12605 * All MISC_CMD ioctls come in here -- e.g. SIOCGLIFCONF.
12606 */
12607 if (ipip->ipi_cmd == IF_UNITSEL) {
12608 /* ioctl comes down the ill */
12609 ci.ci_ipif = ((ill_t *)q->q_ptr)->ill_ipif;
12610 ipif_refhold(ci.ci_ipif);
12611 }
12612 err = 0;
12613 ci.ci_sin = NULL;
12614 ci.ci_sin6 = NULL;
12615 ci.ci_lifr = NULL;
12616 extract_funcp = NULL;
12617 break;
12618
12619 case IF_CMD:
12620 case LIF_CMD:
12621 extract_funcp = ip_extract_lifreq;
12622 break;
12623
12624 case ARP_CMD:
12625 case XARP_CMD:
12626 extract_funcp = ip_extract_arpreq;
12627 break;
12628
12629 default:
12630 ASSERT(0);
12631 }
12632
12633 if (extract_funcp != NULL) {
12634 err = (*extract_funcp)(q, mp, ipip, &ci);
12635 if (err != 0) {
12636 DTRACE_PROBE4(ipif__ioctl,
12637 char *, "ip_process_ioctl finish err",
12638 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12639 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12640 return;
12641 }
12642
12643 /*
12644 * All of the extraction functions return a refheld ipif.
12645 */
12646 ASSERT(ci.ci_ipif != NULL);
12647 }
12648
12649 if (!(ipip->ipi_flags & IPI_WR)) {
12650 /*
12651 * A return value of EINPROGRESS means the ioctl is
12652 * either queued and waiting for some reason or has
12653 * already completed.
12654 */
12655 err = (*ipip->ipi_func)(ci.ci_ipif, ci.ci_sin, q, mp, ipip,
12656 ci.ci_lifr);
12657 if (ci.ci_ipif != NULL) {
12658 DTRACE_PROBE4(ipif__ioctl,
12659 char *, "ip_process_ioctl finish RD",
12660 int, ipip->ipi_cmd, ill_t *, ci.ci_ipif->ipif_ill,
12661 ipif_t *, ci.ci_ipif);
12662 ipif_refrele(ci.ci_ipif);
12663 } else {
12664 DTRACE_PROBE4(ipif__ioctl,
12665 char *, "ip_process_ioctl finish RD",
12666 int, ipip->ipi_cmd, ill_t *, NULL, ipif_t *, NULL);
12667 }
12668 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), NULL);
12669 return;
12670 }
12671
12672 ASSERT(ci.ci_ipif != NULL);
12673
12674 /*
12675 * If ipsq is non-NULL, we are already being called exclusively
12676 */
12677 ASSERT(ipsq == NULL || IAM_WRITER_IPSQ(ipsq));
12678 if (ipsq == NULL) {
12679 ipsq = ipsq_try_enter(ci.ci_ipif, NULL, q, mp, ip_process_ioctl,
12680 NEW_OP, B_TRUE);
12681 if (ipsq == NULL) {
12682 ipif_refrele(ci.ci_ipif);
12683 return;
12684 }
12685 entered_ipsq = B_TRUE;
12686 }
12687 /*
12688 * Release the ipif so that ipif_down and friends that wait for
12689 * references to go away are not misled about the current ipif_refcnt
12690 * values. We are writer so we can access the ipif even after releasing
12691 * the ipif.
12692 */
12693 ipif_refrele(ci.ci_ipif);
12694
12695 ipsq_current_start(ipsq, ci.ci_ipif, ipip->ipi_cmd);
12696
12697 /*
12698 * A return value of EINPROGRESS means the ioctl is
12699 * either queued and waiting for some reason or has
12700 * already completed.
12701 */
12702 err = (*ipip->ipi_func)(ci.ci_ipif, ci.ci_sin, q, mp, ipip, ci.ci_lifr);
12703
12704 DTRACE_PROBE4(ipif__ioctl, char *, "ip_process_ioctl finish WR",
12705 int, ipip->ipi_cmd,
12706 ill_t *, ci.ci_ipif == NULL ? NULL : ci.ci_ipif->ipif_ill,
12707 ipif_t *, ci.ci_ipif);
12708 ip_ioctl_finish(q, mp, err, IPI2MODE(ipip), ipsq);
12709
12710 if (entered_ipsq)
12711 ipsq_exit(ipsq);
12712 }
12713
12714 /*
12715 * Complete the ioctl. Typically ioctls use the mi package and need to
12716 * do mi_copyout/mi_copy_done.
12717 */
12718 void
12719 ip_ioctl_finish(queue_t *q, mblk_t *mp, int err, int mode, ipsq_t *ipsq)
12720 {
12721 conn_t *connp = NULL;
12722
12723 if (err == EINPROGRESS)
12724 return;
12725
12726 if (CONN_Q(q)) {
12727 connp = Q_TO_CONN(q);
12728 ASSERT(connp->conn_ref >= 2);
12729 }
12730
12731 switch (mode) {
12732 case COPYOUT:
12733 if (err == 0)
12734 mi_copyout(q, mp);
12735 else
12736 mi_copy_done(q, mp, err);
12737 break;
12738
12739 case NO_COPYOUT:
12740 mi_copy_done(q, mp, err);
12741 break;
12742
12743 default:
12744 ASSERT(mode == CONN_CLOSE); /* aborted through CONN_CLOSE */
12745 break;
12746 }
12747
12748 /*
12749 * The conn refhold and ioctlref placed on the conn at the start of the
12750 * ioctl are released here.
12751 */
12752 if (connp != NULL) {
12753 CONN_DEC_IOCTLREF(connp);
12754 CONN_OPER_PENDING_DONE(connp);
12755 }
12756
12757 if (ipsq != NULL)
12758 ipsq_current_finish(ipsq);
12759 }
12760
12761 /* Handles all non data messages */
12762 void
12763 ip_wput_nondata(queue_t *q, mblk_t *mp)
12764 {
12765 mblk_t *mp1;
12766 struct iocblk *iocp;
12767 ip_ioctl_cmd_t *ipip;
12768 conn_t *connp;
12769 cred_t *cr;
12770 char *proto_str;
12771
12772 if (CONN_Q(q))
12773 connp = Q_TO_CONN(q);
12774 else
12775 connp = NULL;
12776
12777 switch (DB_TYPE(mp)) {
12778 case M_IOCTL:
12779 /*
12780 * IOCTL processing begins in ip_sioctl_copyin_setup which
12781 * will arrange to copy in associated control structures.
12782 */
12783 ip_sioctl_copyin_setup(q, mp);
12784 return;
12785 case M_IOCDATA:
12786 /*
12787 * Ensure that this is associated with one of our trans-
12788 * parent ioctls. If it's not ours, discard it if we're
12789 * running as a driver, or pass it on if we're a module.
12790 */
12791 iocp = (struct iocblk *)mp->b_rptr;
12792 ipip = ip_sioctl_lookup(iocp->ioc_cmd);
12793 if (ipip == NULL) {
12794 if (q->q_next == NULL) {
12795 goto nak;
12796 } else {
12797 putnext(q, mp);
12798 }
12799 return;
12800 }
12801 if ((q->q_next != NULL) && !(ipip->ipi_flags & IPI_MODOK)) {
12802 /*
12803 * The ioctl is one we recognise, but is not consumed
12804 * by IP as a module and we are a module, so we drop
12805 */
12806 goto nak;
12807 }
12808
12809 /* IOCTL continuation following copyin or copyout. */
12810 if (mi_copy_state(q, mp, NULL) == -1) {
12811 /*
12812 * The copy operation failed. mi_copy_state already
12813 * cleaned up, so we're out of here.
12814 */
12815 return;
12816 }
12817 /*
12818 * If we just completed a copy in, we become writer and
12819 * continue processing in ip_sioctl_copyin_done. If it
12820 * was a copy out, we call mi_copyout again. If there is
12821 * nothing more to copy out, it will complete the IOCTL.
12822 */
12823 if (MI_COPY_DIRECTION(mp) == MI_COPY_IN) {
12824 if (!(mp1 = mp->b_cont) || !(mp1 = mp1->b_cont)) {
12825 mi_copy_done(q, mp, EPROTO);
12826 return;
12827 }
12828 /*
12829 * Check for cases that need more copying. A return
12830 * value of 0 means a second copyin has been started,
12831 * so we return; a return value of 1 means no more
12832 * copying is needed, so we continue.
12833 */
12834 if (ipip->ipi_cmd_type == MSFILT_CMD &&
12835 MI_COPY_COUNT(mp) == 1) {
12836 if (ip_copyin_msfilter(q, mp) == 0)
12837 return;
12838 }
12839 /*
12840 * Refhold the conn, till the ioctl completes. This is
12841 * needed in case the ioctl ends up in the pending mp
12842 * list. Every mp in the ipx_pending_mp list must have
12843 * a refhold on the conn to resume processing. The
12844 * refhold is released when the ioctl completes
12845 * (whether normally or abnormally). An ioctlref is also
12846 * placed on the conn to prevent TCP from removing the
12847 * queue needed to send the ioctl reply back.
12848 * In all cases ip_ioctl_finish is called to finish
12849 * the ioctl and release the refholds.
12850 */
12851 if (connp != NULL) {
12852 /* This is not a reentry */
12853 CONN_INC_REF(connp);
12854 CONN_INC_IOCTLREF(connp);
12855 } else {
12856 if (!(ipip->ipi_flags & IPI_MODOK)) {
12857 mi_copy_done(q, mp, EINVAL);
12858 return;
12859 }
12860 }
12861
12862 ip_process_ioctl(NULL, q, mp, ipip);
12863
12864 } else {
12865 mi_copyout(q, mp);
12866 }
12867 return;
12868
12869 case M_IOCNAK:
12870 /*
12871 * The only way we could get here is if a resolver didn't like
12872 * an IOCTL we sent it. This shouldn't happen.
12873 */
12874 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
12875 "ip_wput_nondata: unexpected M_IOCNAK, ioc_cmd 0x%x",
12876 ((struct iocblk *)mp->b_rptr)->ioc_cmd);
12877 freemsg(mp);
12878 return;
12879 case M_IOCACK:
12880 /* /dev/ip shouldn't see this */
12881 goto nak;
12882 case M_FLUSH:
12883 if (*mp->b_rptr & FLUSHW)
12884 flushq(q, FLUSHALL);
12885 if (q->q_next) {
12886 putnext(q, mp);
12887 return;
12888 }
12889 if (*mp->b_rptr & FLUSHR) {
12890 *mp->b_rptr &= ~FLUSHW;
12891 qreply(q, mp);
12892 return;
12893 }
12894 freemsg(mp);
12895 return;
12896 case M_CTL:
12897 break;
12898 case M_PROTO:
12899 case M_PCPROTO:
12900 /*
12901 * The only PROTO messages we expect are SNMP-related.
12902 */
12903 switch (((union T_primitives *)mp->b_rptr)->type) {
12904 case T_SVR4_OPTMGMT_REQ:
12905 ip2dbg(("ip_wput_nondata: T_SVR4_OPTMGMT_REQ "
12906 "flags %x\n",
12907 ((struct T_optmgmt_req *)mp->b_rptr)->MGMT_flags));
12908
12909 if (connp == NULL) {
12910 proto_str = "T_SVR4_OPTMGMT_REQ";
12911 goto protonak;
12912 }
12913
12914 /*
12915 * All Solaris components should pass a db_credp
12916 * for this TPI message, hence we ASSERT.
12917 * But in case there is some other M_PROTO that looks
12918 * like a TPI message sent by some other kernel
12919 * component, we check and return an error.
12920 */
12921 cr = msg_getcred(mp, NULL);
12922 ASSERT(cr != NULL);
12923 if (cr == NULL) {
12924 mp = mi_tpi_err_ack_alloc(mp, TSYSERR, EINVAL);
12925 if (mp != NULL)
12926 qreply(q, mp);
12927 return;
12928 }
12929
12930 if (!snmpcom_req(q, mp, ip_snmp_set, ip_snmp_get, cr)) {
12931 proto_str = "Bad SNMPCOM request?";
12932 goto protonak;
12933 }
12934 return;
12935 default:
12936 ip1dbg(("ip_wput_nondata: dropping M_PROTO prim %u\n",
12937 (int)*(uint_t *)mp->b_rptr));
12938 freemsg(mp);
12939 return;
12940 }
12941 default:
12942 break;
12943 }
12944 if (q->q_next) {
12945 putnext(q, mp);
12946 } else
12947 freemsg(mp);
12948 return;
12949
12950 nak:
12951 iocp->ioc_error = EINVAL;
12952 mp->b_datap->db_type = M_IOCNAK;
12953 iocp->ioc_count = 0;
12954 qreply(q, mp);
12955 return;
12956
12957 protonak:
12958 cmn_err(CE_NOTE, "IP doesn't process %s as a module", proto_str);
12959 if ((mp = mi_tpi_err_ack_alloc(mp, TPROTO, EINVAL)) != NULL)
12960 qreply(q, mp);
12961 }
12962
12963 /*
12964 * Process IP options in an outbound packet. Verify that the nexthop in a
12965 * strict source route is onlink.
12966 * Returns non-zero if something fails in which case an ICMP error has been
12967 * sent and mp freed.
12968 *
12969 * Assumes the ULP has called ip_massage_options to move nexthop into ipha_dst.
12970 */
12971 int
12972 ip_output_options(mblk_t *mp, ipha_t *ipha, ip_xmit_attr_t *ixa, ill_t *ill)
12973 {
12974 ipoptp_t opts;
12975 uchar_t *opt;
12976 uint8_t optval;
12977 uint8_t optlen;
12978 ipaddr_t dst;
12979 intptr_t code = 0;
12980 ire_t *ire;
12981 ip_stack_t *ipst = ixa->ixa_ipst;
12982 ip_recv_attr_t iras;
12983
12984 ip2dbg(("ip_output_options\n"));
12985
12986 dst = ipha->ipha_dst;
12987 for (optval = ipoptp_first(&opts, ipha);
12988 optval != IPOPT_EOL;
12989 optval = ipoptp_next(&opts)) {
12990 opt = opts.ipoptp_cur;
12991 optlen = opts.ipoptp_len;
12992 ip2dbg(("ip_output_options: opt %d, len %d\n",
12993 optval, optlen));
12994 switch (optval) {
12995 uint32_t off;
12996 case IPOPT_SSRR:
12997 case IPOPT_LSRR:
12998 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
12999 ip1dbg((
13000 "ip_output_options: bad option offset\n"));
13001 code = (char *)&opt[IPOPT_OLEN] -
13002 (char *)ipha;
13003 goto param_prob;
13004 }
13005 off = opt[IPOPT_OFFSET];
13006 ip1dbg(("ip_output_options: next hop 0x%x\n",
13007 ntohl(dst)));
13008 /*
13009 * For strict: verify that dst is directly
13010 * reachable.
13011 */
13012 if (optval == IPOPT_SSRR) {
13013 ire = ire_ftable_lookup_v4(dst, 0, 0,
13014 IRE_INTERFACE, NULL, ALL_ZONES,
13015 ixa->ixa_tsl,
13016 MATCH_IRE_TYPE | MATCH_IRE_SECATTR, 0, ipst,
13017 NULL);
13018 if (ire == NULL) {
13019 ip1dbg(("ip_output_options: SSRR not"
13020 " directly reachable: 0x%x\n",
13021 ntohl(dst)));
13022 goto bad_src_route;
13023 }
13024 ire_refrele(ire);
13025 }
13026 break;
13027 case IPOPT_RR:
13028 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
13029 ip1dbg((
13030 "ip_output_options: bad option offset\n"));
13031 code = (char *)&opt[IPOPT_OLEN] -
13032 (char *)ipha;
13033 goto param_prob;
13034 }
13035 break;
13036 case IPOPT_TS:
13037 /*
13038 * Verify that length >=5 and that there is either
13039 * room for another timestamp or that the overflow
13040 * counter is not maxed out.
13041 */
13042 code = (char *)&opt[IPOPT_OLEN] - (char *)ipha;
13043 if (optlen < IPOPT_MINLEN_IT) {
13044 goto param_prob;
13045 }
13046 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
13047 ip1dbg((
13048 "ip_output_options: bad option offset\n"));
13049 code = (char *)&opt[IPOPT_OFFSET] -
13050 (char *)ipha;
13051 goto param_prob;
13052 }
13053 switch (opt[IPOPT_POS_OV_FLG] & 0x0F) {
13054 case IPOPT_TS_TSONLY:
13055 off = IPOPT_TS_TIMELEN;
13056 break;
13057 case IPOPT_TS_TSANDADDR:
13058 case IPOPT_TS_PRESPEC:
13059 case IPOPT_TS_PRESPEC_RFC791:
13060 off = IP_ADDR_LEN + IPOPT_TS_TIMELEN;
13061 break;
13062 default:
13063 code = (char *)&opt[IPOPT_POS_OV_FLG] -
13064 (char *)ipha;
13065 goto param_prob;
13066 }
13067 if (opt[IPOPT_OFFSET] - 1 + off > optlen &&
13068 (opt[IPOPT_POS_OV_FLG] & 0xF0) == 0xF0) {
13069 /*
13070 * No room and the overflow counter is 15
13071 * already.
13072 */
13073 goto param_prob;
13074 }
13075 break;
13076 }
13077 }
13078
13079 if ((opts.ipoptp_flags & IPOPTP_ERROR) == 0)
13080 return (0);
13081
13082 ip1dbg(("ip_output_options: error processing IP options."));
13083 code = (char *)&opt[IPOPT_OFFSET] - (char *)ipha;
13084
13085 param_prob:
13086 bzero(&iras, sizeof (iras));
13087 iras.ira_ill = iras.ira_rill = ill;
13088 iras.ira_ruifindex = ill->ill_phyint->phyint_ifindex;
13089 iras.ira_rifindex = iras.ira_ruifindex;
13090 iras.ira_flags = IRAF_IS_IPV4;
13091
13092 ip_drop_output("ip_output_options", mp, ill);
13093 icmp_param_problem(mp, (uint8_t)code, &iras);
13094 ASSERT(!(iras.ira_flags & IRAF_IPSEC_SECURE));
13095 return (-1);
13096
13097 bad_src_route:
13098 bzero(&iras, sizeof (iras));
13099 iras.ira_ill = iras.ira_rill = ill;
13100 iras.ira_ruifindex = ill->ill_phyint->phyint_ifindex;
13101 iras.ira_rifindex = iras.ira_ruifindex;
13102 iras.ira_flags = IRAF_IS_IPV4;
13103
13104 ip_drop_input("ICMP_SOURCE_ROUTE_FAILED", mp, ill);
13105 icmp_unreachable(mp, ICMP_SOURCE_ROUTE_FAILED, &iras);
13106 ASSERT(!(iras.ira_flags & IRAF_IPSEC_SECURE));
13107 return (-1);
13108 }
13109
13110 /*
13111 * The maximum value of conn_drain_list_cnt is CONN_MAXDRAINCNT.
13112 * conn_drain_list_cnt can be changed by setting conn_drain_nthreads
13113 * thru /etc/system.
13114 */
13115 #define CONN_MAXDRAINCNT 64
13116
13117 static void
13118 conn_drain_init(ip_stack_t *ipst)
13119 {
13120 int i, j;
13121 idl_tx_list_t *itl_tx;
13122
13123 ipst->ips_conn_drain_list_cnt = conn_drain_nthreads;
13124
13125 if ((ipst->ips_conn_drain_list_cnt == 0) ||
13126 (ipst->ips_conn_drain_list_cnt > CONN_MAXDRAINCNT)) {
13127 /*
13128 * Default value of the number of drainers is the
13129 * number of cpus, subject to maximum of 8 drainers.
13130 */
13131 if (boot_max_ncpus != -1)
13132 ipst->ips_conn_drain_list_cnt = MIN(boot_max_ncpus, 8);
13133 else
13134 ipst->ips_conn_drain_list_cnt = MIN(max_ncpus, 8);
13135 }
13136
13137 ipst->ips_idl_tx_list =
13138 kmem_zalloc(TX_FANOUT_SIZE * sizeof (idl_tx_list_t), KM_SLEEP);
13139 for (i = 0; i < TX_FANOUT_SIZE; i++) {
13140 itl_tx = &ipst->ips_idl_tx_list[i];
13141 itl_tx->txl_drain_list =
13142 kmem_zalloc(ipst->ips_conn_drain_list_cnt *
13143 sizeof (idl_t), KM_SLEEP);
13144 mutex_init(&itl_tx->txl_lock, NULL, MUTEX_DEFAULT, NULL);
13145 for (j = 0; j < ipst->ips_conn_drain_list_cnt; j++) {
13146 mutex_init(&itl_tx->txl_drain_list[j].idl_lock, NULL,
13147 MUTEX_DEFAULT, NULL);
13148 itl_tx->txl_drain_list[j].idl_itl = itl_tx;
13149 }
13150 }
13151 }
13152
13153 static void
13154 conn_drain_fini(ip_stack_t *ipst)
13155 {
13156 int i;
13157 idl_tx_list_t *itl_tx;
13158
13159 for (i = 0; i < TX_FANOUT_SIZE; i++) {
13160 itl_tx = &ipst->ips_idl_tx_list[i];
13161 kmem_free(itl_tx->txl_drain_list,
13162 ipst->ips_conn_drain_list_cnt * sizeof (idl_t));
13163 }
13164 kmem_free(ipst->ips_idl_tx_list,
13165 TX_FANOUT_SIZE * sizeof (idl_tx_list_t));
13166 ipst->ips_idl_tx_list = NULL;
13167 }
13168
13169 /*
13170 * Flow control has blocked us from proceeding. Insert the given conn in one
13171 * of the conn drain lists. When flow control is unblocked, either ip_wsrv()
13172 * (STREAMS) or ill_flow_enable() (direct) will be called back, which in turn
13173 * will call conn_walk_drain(). See the flow control notes at the top of this
13174 * file for more details.
13175 */
13176 void
13177 conn_drain_insert(conn_t *connp, idl_tx_list_t *tx_list)
13178 {
13179 idl_t *idl = tx_list->txl_drain_list;
13180 uint_t index;
13181 ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
13182
13183 mutex_enter(&connp->conn_lock);
13184 if (connp->conn_state_flags & CONN_CLOSING) {
13185 /*
13186 * The conn is closing as a result of which CONN_CLOSING
13187 * is set. Return.
13188 */
13189 mutex_exit(&connp->conn_lock);
13190 return;
13191 } else if (connp->conn_idl == NULL) {
13192 /*
13193 * Assign the next drain list round robin. We dont' use
13194 * a lock, and thus it may not be strictly round robin.
13195 * Atomicity of load/stores is enough to make sure that
13196 * conn_drain_list_index is always within bounds.
13197 */
13198 index = tx_list->txl_drain_index;
13199 ASSERT(index < ipst->ips_conn_drain_list_cnt);
13200 connp->conn_idl = &tx_list->txl_drain_list[index];
13201 index++;
13202 if (index == ipst->ips_conn_drain_list_cnt)
13203 index = 0;
13204 tx_list->txl_drain_index = index;
13205 } else {
13206 ASSERT(connp->conn_idl->idl_itl == tx_list);
13207 }
13208 mutex_exit(&connp->conn_lock);
13209
13210 idl = connp->conn_idl;
13211 mutex_enter(&idl->idl_lock);
13212 if ((connp->conn_drain_prev != NULL) ||
13213 (connp->conn_state_flags & CONN_CLOSING)) {
13214 /*
13215 * The conn is either already in the drain list or closing.
13216 * (We needed to check for CONN_CLOSING again since close can
13217 * sneak in between dropping conn_lock and acquiring idl_lock.)
13218 */
13219 mutex_exit(&idl->idl_lock);
13220 return;
13221 }
13222
13223 /*
13224 * The conn is not in the drain list. Insert it at the
13225 * tail of the drain list. The drain list is circular
13226 * and doubly linked. idl_conn points to the 1st element
13227 * in the list.
13228 */
13229 if (idl->idl_conn == NULL) {
13230 idl->idl_conn = connp;
13231 connp->conn_drain_next = connp;
13232 connp->conn_drain_prev = connp;
13233 } else {
13234 conn_t *head = idl->idl_conn;
13235
13236 connp->conn_drain_next = head;
13237 connp->conn_drain_prev = head->conn_drain_prev;
13238 head->conn_drain_prev->conn_drain_next = connp;
13239 head->conn_drain_prev = connp;
13240 }
13241 /*
13242 * For non streams based sockets assert flow control.
13243 */
13244 conn_setqfull(connp, NULL);
13245 mutex_exit(&idl->idl_lock);
13246 }
13247
13248 static void
13249 conn_drain_remove(conn_t *connp)
13250 {
13251 idl_t *idl = connp->conn_idl;
13252
13253 if (idl != NULL) {
13254 /*
13255 * Remove ourself from the drain list.
13256 */
13257 if (connp->conn_drain_next == connp) {
13258 /* Singleton in the list */
13259 ASSERT(connp->conn_drain_prev == connp);
13260 idl->idl_conn = NULL;
13261 } else {
13262 connp->conn_drain_prev->conn_drain_next =
13263 connp->conn_drain_next;
13264 connp->conn_drain_next->conn_drain_prev =
13265 connp->conn_drain_prev;
13266 if (idl->idl_conn == connp)
13267 idl->idl_conn = connp->conn_drain_next;
13268 }
13269
13270 /*
13271 * NOTE: because conn_idl is associated with a specific drain
13272 * list which in turn is tied to the index the TX ring
13273 * (txl_cookie) hashes to, and because the TX ring can change
13274 * over the lifetime of the conn_t, we must clear conn_idl so
13275 * a subsequent conn_drain_insert() will set conn_idl again
13276 * based on the latest txl_cookie.
13277 */
13278 connp->conn_idl = NULL;
13279 }
13280 connp->conn_drain_next = NULL;
13281 connp->conn_drain_prev = NULL;
13282
13283 conn_clrqfull(connp, NULL);
13284 /*
13285 * For streams based sockets open up flow control.
13286 */
13287 if (!IPCL_IS_NONSTR(connp))
13288 enableok(connp->conn_wq);
13289 }
13290
13291 /*
13292 * This conn is closing, and we are called from ip_close. OR
13293 * this conn is draining because flow-control on the ill has been relieved.
13294 *
13295 * We must also need to remove conn's on this idl from the list, and also
13296 * inform the sockfs upcalls about the change in flow-control.
13297 */
13298 static void
13299 conn_drain(conn_t *connp, boolean_t closing)
13300 {
13301 idl_t *idl;
13302 conn_t *next_connp;
13303
13304 /*
13305 * connp->conn_idl is stable at this point, and no lock is needed
13306 * to check it. If we are called from ip_close, close has already
13307 * set CONN_CLOSING, thus freezing the value of conn_idl, and
13308 * called us only because conn_idl is non-null. If we are called thru
13309 * service, conn_idl could be null, but it cannot change because
13310 * service is single-threaded per queue, and there cannot be another
13311 * instance of service trying to call conn_drain_insert on this conn
13312 * now.
13313 */
13314 ASSERT(!closing || connp == NULL || connp->conn_idl != NULL);
13315
13316 /*
13317 * If the conn doesn't exist or is not on a drain list, bail.
13318 */
13319 if (connp == NULL || connp->conn_idl == NULL ||
13320 connp->conn_drain_prev == NULL) {
13321 return;
13322 }
13323
13324 idl = connp->conn_idl;
13325 ASSERT(MUTEX_HELD(&idl->idl_lock));
13326
13327 if (!closing) {
13328 next_connp = connp->conn_drain_next;
13329 while (next_connp != connp) {
13330 conn_t *delconnp = next_connp;
13331
13332 next_connp = next_connp->conn_drain_next;
13333 conn_drain_remove(delconnp);
13334 }
13335 ASSERT(connp->conn_drain_next == idl->idl_conn);
13336 }
13337 conn_drain_remove(connp);
13338 }
13339
13340 /*
13341 * Write service routine. Shared perimeter entry point.
13342 * The device queue's messages has fallen below the low water mark and STREAMS
13343 * has backenabled the ill_wq. Send sockfs notification about flow-control on
13344 * each waiting conn.
13345 */
13346 void
13347 ip_wsrv(queue_t *q)
13348 {
13349 ill_t *ill;
13350
13351 ill = (ill_t *)q->q_ptr;
13352 if (ill->ill_state_flags == 0) {
13353 ip_stack_t *ipst = ill->ill_ipst;
13354
13355 /*
13356 * The device flow control has opened up.
13357 * Walk through conn drain lists and qenable the
13358 * first conn in each list. This makes sense only
13359 * if the stream is fully plumbed and setup.
13360 * Hence the ill_state_flags check above.
13361 */
13362 ip1dbg(("ip_wsrv: walking\n"));
13363 conn_walk_drain(ipst, &ipst->ips_idl_tx_list[0]);
13364 enableok(ill->ill_wq);
13365 }
13366 }
13367
13368 /*
13369 * Callback to disable flow control in IP.
13370 *
13371 * This is a mac client callback added when the DLD_CAPAB_DIRECT capability
13372 * is enabled.
13373 *
13374 * When MAC_TX() is not able to send any more packets, dld sets its queue
13375 * to QFULL and enable the STREAMS flow control. Later, when the underlying
13376 * driver is able to continue to send packets, it calls mac_tx_(ring_)update()
13377 * function and wakes up corresponding mac worker threads, which in turn
13378 * calls this callback function, and disables flow control.
13379 */
13380 void
13381 ill_flow_enable(void *arg, ip_mac_tx_cookie_t cookie)
13382 {
13383 ill_t *ill = (ill_t *)arg;
13384 ip_stack_t *ipst = ill->ill_ipst;
13385 idl_tx_list_t *idl_txl;
13386
13387 idl_txl = &ipst->ips_idl_tx_list[IDLHASHINDEX(cookie)];
13388 mutex_enter(&idl_txl->txl_lock);
13389 /* add code to to set a flag to indicate idl_txl is enabled */
13390 conn_walk_drain(ipst, idl_txl);
13391 mutex_exit(&idl_txl->txl_lock);
13392 }
13393
13394 /*
13395 * Flow control has been relieved and STREAMS has backenabled us; drain
13396 * all the conn lists on `tx_list'.
13397 */
13398 static void
13399 conn_walk_drain(ip_stack_t *ipst, idl_tx_list_t *tx_list)
13400 {
13401 int i;
13402 idl_t *idl;
13403
13404 IP_STAT(ipst, ip_conn_walk_drain);
13405
13406 for (i = 0; i < ipst->ips_conn_drain_list_cnt; i++) {
13407 idl = &tx_list->txl_drain_list[i];
13408 mutex_enter(&idl->idl_lock);
13409 conn_drain(idl->idl_conn, B_FALSE);
13410 mutex_exit(&idl->idl_lock);
13411 }
13412 }
13413
13414 /*
13415 * Determine if the ill and multicast aspects of that packets
13416 * "matches" the conn.
13417 */
13418 boolean_t
13419 conn_wantpacket(conn_t *connp, ip_recv_attr_t *ira, ipha_t *ipha)
13420 {
13421 ill_t *ill = ira->ira_rill;
13422 zoneid_t zoneid = ira->ira_zoneid;
13423 uint_t in_ifindex;
13424 ipaddr_t dst, src;
13425
13426 dst = ipha->ipha_dst;
13427 src = ipha->ipha_src;
13428
13429 /*
13430 * conn_incoming_ifindex is set by IP_BOUND_IF which limits
13431 * unicast, broadcast and multicast reception to
13432 * conn_incoming_ifindex.
13433 * conn_wantpacket is called for unicast, broadcast and
13434 * multicast packets.
13435 */
13436 in_ifindex = connp->conn_incoming_ifindex;
13437
13438 /* mpathd can bind to the under IPMP interface, which we allow */
13439 if (in_ifindex != 0 && in_ifindex != ill->ill_phyint->phyint_ifindex) {
13440 if (!IS_UNDER_IPMP(ill))
13441 return (B_FALSE);
13442
13443 if (in_ifindex != ipmp_ill_get_ipmp_ifindex(ill))
13444 return (B_FALSE);
13445 }
13446
13447 if (!IPCL_ZONE_MATCH(connp, zoneid))
13448 return (B_FALSE);
13449
13450 if (!(ira->ira_flags & IRAF_MULTICAST))
13451 return (B_TRUE);
13452
13453 if (connp->conn_multi_router) {
13454 /* multicast packet and multicast router socket: send up */
13455 return (B_TRUE);
13456 }
13457
13458 if (ipha->ipha_protocol == IPPROTO_PIM ||
13459 ipha->ipha_protocol == IPPROTO_RSVP)
13460 return (B_TRUE);
13461
13462 return (conn_hasmembers_ill_withsrc_v4(connp, dst, src, ira->ira_ill));
13463 }
13464
13465 void
13466 conn_setqfull(conn_t *connp, boolean_t *flow_stopped)
13467 {
13468 if (IPCL_IS_NONSTR(connp)) {
13469 (*connp->conn_upcalls->su_txq_full)
13470 (connp->conn_upper_handle, B_TRUE);
13471 if (flow_stopped != NULL)
13472 *flow_stopped = B_TRUE;
13473 } else {
13474 queue_t *q = connp->conn_wq;
13475
13476 ASSERT(q != NULL);
13477 if (!(q->q_flag & QFULL)) {
13478 mutex_enter(QLOCK(q));
13479 if (!(q->q_flag & QFULL)) {
13480 /* still need to set QFULL */
13481 q->q_flag |= QFULL;
13482 /* set flow_stopped to true under QLOCK */
13483 if (flow_stopped != NULL)
13484 *flow_stopped = B_TRUE;
13485 mutex_exit(QLOCK(q));
13486 } else {
13487 /* flow_stopped is left unchanged */
13488 mutex_exit(QLOCK(q));
13489 }
13490 }
13491 }
13492 }
13493
13494 void
13495 conn_clrqfull(conn_t *connp, boolean_t *flow_stopped)
13496 {
13497 if (IPCL_IS_NONSTR(connp)) {
13498 (*connp->conn_upcalls->su_txq_full)
13499 (connp->conn_upper_handle, B_FALSE);
13500 if (flow_stopped != NULL)
13501 *flow_stopped = B_FALSE;
13502 } else {
13503 queue_t *q = connp->conn_wq;
13504
13505 ASSERT(q != NULL);
13506 if (q->q_flag & QFULL) {
13507 mutex_enter(QLOCK(q));
13508 if (q->q_flag & QFULL) {
13509 q->q_flag &= ~QFULL;
13510 /* set flow_stopped to false under QLOCK */
13511 if (flow_stopped != NULL)
13512 *flow_stopped = B_FALSE;
13513 mutex_exit(QLOCK(q));
13514 if (q->q_flag & QWANTW)
13515 qbackenable(q, 0);
13516 } else {
13517 /* flow_stopped is left unchanged */
13518 mutex_exit(QLOCK(q));
13519 }
13520 }
13521 }
13522
13523 mutex_enter(&connp->conn_lock);
13524 connp->conn_blocked = B_FALSE;
13525 mutex_exit(&connp->conn_lock);
13526 }
13527
13528 /*
13529 * Return the length in bytes of the IPv4 headers (base header, label, and
13530 * other IP options) that will be needed based on the
13531 * ip_pkt_t structure passed by the caller.
13532 *
13533 * The returned length does not include the length of the upper level
13534 * protocol (ULP) header.
13535 * The caller needs to check that the length doesn't exceed the max for IPv4.
13536 */
13537 int
13538 ip_total_hdrs_len_v4(const ip_pkt_t *ipp)
13539 {
13540 int len;
13541
13542 len = IP_SIMPLE_HDR_LENGTH;
13543 if (ipp->ipp_fields & IPPF_LABEL_V4) {
13544 ASSERT(ipp->ipp_label_len_v4 != 0);
13545 /* We need to round up here */
13546 len += (ipp->ipp_label_len_v4 + 3) & ~3;
13547 }
13548
13549 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
13550 ASSERT(ipp->ipp_ipv4_options_len != 0);
13551 ASSERT((ipp->ipp_ipv4_options_len & 3) == 0);
13552 len += ipp->ipp_ipv4_options_len;
13553 }
13554 return (len);
13555 }
13556
13557 /*
13558 * All-purpose routine to build an IPv4 header with options based
13559 * on the abstract ip_pkt_t.
13560 *
13561 * The caller has to set the source and destination address as well as
13562 * ipha_length. The caller has to massage any source route and compensate
13563 * for the ULP pseudo-header checksum due to the source route.
13564 */
13565 void
13566 ip_build_hdrs_v4(uchar_t *buf, uint_t buf_len, const ip_pkt_t *ipp,
13567 uint8_t protocol)
13568 {
13569 ipha_t *ipha = (ipha_t *)buf;
13570 uint8_t *cp;
13571
13572 /* Initialize IPv4 header */
13573 ipha->ipha_type_of_service = ipp->ipp_type_of_service;
13574 ipha->ipha_length = 0; /* Caller will set later */
13575 ipha->ipha_ident = 0;
13576 ipha->ipha_fragment_offset_and_flags = 0;
13577 ipha->ipha_ttl = ipp->ipp_unicast_hops;
13578 ipha->ipha_protocol = protocol;
13579 ipha->ipha_hdr_checksum = 0;
13580
13581 if ((ipp->ipp_fields & IPPF_ADDR) &&
13582 IN6_IS_ADDR_V4MAPPED(&ipp->ipp_addr))
13583 ipha->ipha_src = ipp->ipp_addr_v4;
13584
13585 cp = (uint8_t *)&ipha[1];
13586 if (ipp->ipp_fields & IPPF_LABEL_V4) {
13587 ASSERT(ipp->ipp_label_len_v4 != 0);
13588 bcopy(ipp->ipp_label_v4, cp, ipp->ipp_label_len_v4);
13589 cp += ipp->ipp_label_len_v4;
13590 /* We need to round up here */
13591 while ((uintptr_t)cp & 0x3) {
13592 *cp++ = IPOPT_NOP;
13593 }
13594 }
13595
13596 if (ipp->ipp_fields & IPPF_IPV4_OPTIONS) {
13597 ASSERT(ipp->ipp_ipv4_options_len != 0);
13598 ASSERT((ipp->ipp_ipv4_options_len & 3) == 0);
13599 bcopy(ipp->ipp_ipv4_options, cp, ipp->ipp_ipv4_options_len);
13600 cp += ipp->ipp_ipv4_options_len;
13601 }
13602 ipha->ipha_version_and_hdr_length =
13603 (uint8_t)((IP_VERSION << 4) + buf_len / 4);
13604
13605 ASSERT((int)(cp - buf) == buf_len);
13606 }
13607
13608 /* Allocate the private structure */
13609 static int
13610 ip_priv_alloc(void **bufp)
13611 {
13612 void *buf;
13613
13614 if ((buf = kmem_alloc(sizeof (ip_priv_t), KM_NOSLEEP)) == NULL)
13615 return (ENOMEM);
13616
13617 *bufp = buf;
13618 return (0);
13619 }
13620
13621 /* Function to delete the private structure */
13622 void
13623 ip_priv_free(void *buf)
13624 {
13625 ASSERT(buf != NULL);
13626 kmem_free(buf, sizeof (ip_priv_t));
13627 }
13628
13629 /*
13630 * The entry point for IPPF processing.
13631 * If the classifier (IPGPC_CLASSIFY) is not loaded and configured, the
13632 * routine just returns.
13633 *
13634 * When called, ip_process generates an ipp_packet_t structure
13635 * which holds the state information for this packet and invokes the
13636 * the classifier (via ipp_packet_process). The classification, depending on
13637 * configured filters, results in a list of actions for this packet. Invoking
13638 * an action may cause the packet to be dropped, in which case we return NULL.
13639 * proc indicates the callout position for
13640 * this packet and ill is the interface this packet arrived on or will leave
13641 * on (inbound and outbound resp.).
13642 *
13643 * We do the processing on the rill (mapped to the upper if ipmp), but MIB
13644 * on the ill corrsponding to the destination IP address.
13645 */
13646 mblk_t *
13647 ip_process(ip_proc_t proc, mblk_t *mp, ill_t *rill, ill_t *ill)
13648 {
13649 ip_priv_t *priv;
13650 ipp_action_id_t aid;
13651 int rc = 0;
13652 ipp_packet_t *pp;
13653
13654 /* If the classifier is not loaded, return */
13655 if ((aid = ipp_action_lookup(IPGPC_CLASSIFY)) == IPP_ACTION_INVAL) {
13656 return (mp);
13657 }
13658
13659 ASSERT(mp != NULL);
13660
13661 /* Allocate the packet structure */
13662 rc = ipp_packet_alloc(&pp, "ip", aid);
13663 if (rc != 0)
13664 goto drop;
13665
13666 /* Allocate the private structure */
13667 rc = ip_priv_alloc((void **)&priv);
13668 if (rc != 0) {
13669 ipp_packet_free(pp);
13670 goto drop;
13671 }
13672 priv->proc = proc;
13673 priv->ill_index = ill_get_upper_ifindex(rill);
13674
13675 ipp_packet_set_private(pp, priv, ip_priv_free);
13676 ipp_packet_set_data(pp, mp);
13677
13678 /* Invoke the classifier */
13679 rc = ipp_packet_process(&pp);
13680 if (pp != NULL) {
13681 mp = ipp_packet_get_data(pp);
13682 ipp_packet_free(pp);
13683 if (rc != 0)
13684 goto drop;
13685 return (mp);
13686 } else {
13687 /* No mp to trace in ip_drop_input/ip_drop_output */
13688 mp = NULL;
13689 }
13690 drop:
13691 if (proc == IPP_LOCAL_IN || proc == IPP_FWD_IN) {
13692 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
13693 ip_drop_input("ip_process", mp, ill);
13694 } else {
13695 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
13696 ip_drop_output("ip_process", mp, ill);
13697 }
13698 freemsg(mp);
13699 return (NULL);
13700 }
13701
13702 /*
13703 * Propagate a multicast group membership operation (add/drop) on
13704 * all the interfaces crossed by the related multirt routes.
13705 * The call is considered successful if the operation succeeds
13706 * on at least one interface.
13707 *
13708 * This assumes that a set of IRE_HOST/RTF_MULTIRT has been created for the
13709 * multicast addresses with the ire argument being the first one.
13710 * We walk the bucket to find all the of those.
13711 *
13712 * Common to IPv4 and IPv6.
13713 */
13714 static int
13715 ip_multirt_apply_membership(int (*fn)(conn_t *, boolean_t,
13716 const in6_addr_t *, ipaddr_t, uint_t, mcast_record_t, const in6_addr_t *),
13717 ire_t *ire, conn_t *connp, boolean_t checkonly, const in6_addr_t *v6group,
13718 mcast_record_t fmode, const in6_addr_t *v6src)
13719 {
13720 ire_t *ire_gw;
13721 irb_t *irb;
13722 int ifindex;
13723 int error = 0;
13724 int result;
13725 ip_stack_t *ipst = ire->ire_ipst;
13726 ipaddr_t group;
13727 boolean_t isv6;
13728 int match_flags;
13729
13730 if (IN6_IS_ADDR_V4MAPPED(v6group)) {
13731 IN6_V4MAPPED_TO_IPADDR(v6group, group);
13732 isv6 = B_FALSE;
13733 } else {
13734 isv6 = B_TRUE;
13735 }
13736
13737 irb = ire->ire_bucket;
13738 ASSERT(irb != NULL);
13739
13740 result = 0;
13741 irb_refhold(irb);
13742 for (; ire != NULL; ire = ire->ire_next) {
13743 if ((ire->ire_flags & RTF_MULTIRT) == 0)
13744 continue;
13745
13746 /* We handle -ifp routes by matching on the ill if set */
13747 match_flags = MATCH_IRE_TYPE;
13748 if (ire->ire_ill != NULL)
13749 match_flags |= MATCH_IRE_ILL;
13750
13751 if (isv6) {
13752 if (!IN6_ARE_ADDR_EQUAL(&ire->ire_addr_v6, v6group))
13753 continue;
13754
13755 ire_gw = ire_ftable_lookup_v6(&ire->ire_gateway_addr_v6,
13756 0, 0, IRE_INTERFACE, ire->ire_ill, ALL_ZONES, NULL,
13757 match_flags, 0, ipst, NULL);
13758 } else {
13759 if (ire->ire_addr != group)
13760 continue;
13761
13762 ire_gw = ire_ftable_lookup_v4(ire->ire_gateway_addr,
13763 0, 0, IRE_INTERFACE, ire->ire_ill, ALL_ZONES, NULL,
13764 match_flags, 0, ipst, NULL);
13765 }
13766 /* No interface route exists for the gateway; skip this ire. */
13767 if (ire_gw == NULL)
13768 continue;
13769 if (ire_gw->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
13770 ire_refrele(ire_gw);
13771 continue;
13772 }
13773 ASSERT(ire_gw->ire_ill != NULL); /* IRE_INTERFACE */
13774 ifindex = ire_gw->ire_ill->ill_phyint->phyint_ifindex;
13775
13776 /*
13777 * The operation is considered a success if
13778 * it succeeds at least once on any one interface.
13779 */
13780 error = fn(connp, checkonly, v6group, INADDR_ANY, ifindex,
13781 fmode, v6src);
13782 if (error == 0)
13783 result = CGTP_MCAST_SUCCESS;
13784
13785 ire_refrele(ire_gw);
13786 }
13787 irb_refrele(irb);
13788 /*
13789 * Consider the call as successful if we succeeded on at least
13790 * one interface. Otherwise, return the last encountered error.
13791 */
13792 return (result == CGTP_MCAST_SUCCESS ? 0 : error);
13793 }
13794
13795 /*
13796 * Return the expected CGTP hooks version number.
13797 */
13798 int
13799 ip_cgtp_filter_supported(void)
13800 {
13801 return (ip_cgtp_filter_rev);
13802 }
13803
13804 /*
13805 * CGTP hooks can be registered by invoking this function.
13806 * Checks that the version number matches.
13807 */
13808 int
13809 ip_cgtp_filter_register(netstackid_t stackid, cgtp_filter_ops_t *ops)
13810 {
13811 netstack_t *ns;
13812 ip_stack_t *ipst;
13813
13814 if (ops->cfo_filter_rev != CGTP_FILTER_REV)
13815 return (ENOTSUP);
13816
13817 ns = netstack_find_by_stackid(stackid);
13818 if (ns == NULL)
13819 return (EINVAL);
13820 ipst = ns->netstack_ip;
13821 ASSERT(ipst != NULL);
13822
13823 if (ipst->ips_ip_cgtp_filter_ops != NULL) {
13824 netstack_rele(ns);
13825 return (EALREADY);
13826 }
13827
13828 ipst->ips_ip_cgtp_filter_ops = ops;
13829
13830 ill_set_inputfn_all(ipst);
13831
13832 netstack_rele(ns);
13833 return (0);
13834 }
13835
13836 /*
13837 * CGTP hooks can be unregistered by invoking this function.
13838 * Returns ENXIO if there was no registration.
13839 * Returns EBUSY if the ndd variable has not been turned off.
13840 */
13841 int
13842 ip_cgtp_filter_unregister(netstackid_t stackid)
13843 {
13844 netstack_t *ns;
13845 ip_stack_t *ipst;
13846
13847 ns = netstack_find_by_stackid(stackid);
13848 if (ns == NULL)
13849 return (EINVAL);
13850 ipst = ns->netstack_ip;
13851 ASSERT(ipst != NULL);
13852
13853 if (ipst->ips_ip_cgtp_filter) {
13854 netstack_rele(ns);
13855 return (EBUSY);
13856 }
13857
13858 if (ipst->ips_ip_cgtp_filter_ops == NULL) {
13859 netstack_rele(ns);
13860 return (ENXIO);
13861 }
13862 ipst->ips_ip_cgtp_filter_ops = NULL;
13863
13864 ill_set_inputfn_all(ipst);
13865
13866 netstack_rele(ns);
13867 return (0);
13868 }
13869
13870 /*
13871 * Check whether there is a CGTP filter registration.
13872 * Returns non-zero if there is a registration, otherwise returns zero.
13873 * Note: returns zero if bad stackid.
13874 */
13875 int
13876 ip_cgtp_filter_is_registered(netstackid_t stackid)
13877 {
13878 netstack_t *ns;
13879 ip_stack_t *ipst;
13880 int ret;
13881
13882 ns = netstack_find_by_stackid(stackid);
13883 if (ns == NULL)
13884 return (0);
13885 ipst = ns->netstack_ip;
13886 ASSERT(ipst != NULL);
13887
13888 if (ipst->ips_ip_cgtp_filter_ops != NULL)
13889 ret = 1;
13890 else
13891 ret = 0;
13892
13893 netstack_rele(ns);
13894 return (ret);
13895 }
13896
13897 static int
13898 ip_squeue_switch(int val)
13899 {
13900 int rval;
13901
13902 switch (val) {
13903 case IP_SQUEUE_ENTER_NODRAIN:
13904 rval = SQ_NODRAIN;
13905 break;
13906 case IP_SQUEUE_ENTER:
13907 rval = SQ_PROCESS;
13908 break;
13909 case IP_SQUEUE_FILL:
13910 default:
13911 rval = SQ_FILL;
13912 break;
13913 }
13914 return (rval);
13915 }
13916
13917 static void *
13918 ip_kstat2_init(netstackid_t stackid, ip_stat_t *ip_statisticsp)
13919 {
13920 kstat_t *ksp;
13921
13922 ip_stat_t template = {
13923 { "ip_udp_fannorm", KSTAT_DATA_UINT64 },
13924 { "ip_udp_fanmb", KSTAT_DATA_UINT64 },
13925 { "ip_recv_pullup", KSTAT_DATA_UINT64 },
13926 { "ip_db_ref", KSTAT_DATA_UINT64 },
13927 { "ip_notaligned", KSTAT_DATA_UINT64 },
13928 { "ip_multimblk", KSTAT_DATA_UINT64 },
13929 { "ip_opt", KSTAT_DATA_UINT64 },
13930 { "ipsec_proto_ahesp", KSTAT_DATA_UINT64 },
13931 { "ip_conn_flputbq", KSTAT_DATA_UINT64 },
13932 { "ip_conn_walk_drain", KSTAT_DATA_UINT64 },
13933 { "ip_out_sw_cksum", KSTAT_DATA_UINT64 },
13934 { "ip_out_sw_cksum_bytes", KSTAT_DATA_UINT64 },
13935 { "ip_in_sw_cksum", KSTAT_DATA_UINT64 },
13936 { "ip_ire_reclaim_calls", KSTAT_DATA_UINT64 },
13937 { "ip_ire_reclaim_deleted", KSTAT_DATA_UINT64 },
13938 { "ip_nce_reclaim_calls", KSTAT_DATA_UINT64 },
13939 { "ip_nce_reclaim_deleted", KSTAT_DATA_UINT64 },
13940 { "ip_dce_reclaim_calls", KSTAT_DATA_UINT64 },
13941 { "ip_dce_reclaim_deleted", KSTAT_DATA_UINT64 },
13942 { "ip_tcp_in_full_hw_cksum_err", KSTAT_DATA_UINT64 },
13943 { "ip_tcp_in_part_hw_cksum_err", KSTAT_DATA_UINT64 },
13944 { "ip_tcp_in_sw_cksum_err", KSTAT_DATA_UINT64 },
13945 { "ip_udp_in_full_hw_cksum_err", KSTAT_DATA_UINT64 },
13946 { "ip_udp_in_part_hw_cksum_err", KSTAT_DATA_UINT64 },
13947 { "ip_udp_in_sw_cksum_err", KSTAT_DATA_UINT64 },
13948 { "conn_in_recvdstaddr", KSTAT_DATA_UINT64 },
13949 { "conn_in_recvopts", KSTAT_DATA_UINT64 },
13950 { "conn_in_recvif", KSTAT_DATA_UINT64 },
13951 { "conn_in_recvslla", KSTAT_DATA_UINT64 },
13952 { "conn_in_recvucred", KSTAT_DATA_UINT64 },
13953 { "conn_in_recvttl", KSTAT_DATA_UINT64 },
13954 { "conn_in_recvhopopts", KSTAT_DATA_UINT64 },
13955 { "conn_in_recvhoplimit", KSTAT_DATA_UINT64 },
13956 { "conn_in_recvdstopts", KSTAT_DATA_UINT64 },
13957 { "conn_in_recvrthdrdstopts", KSTAT_DATA_UINT64 },
13958 { "conn_in_recvrthdr", KSTAT_DATA_UINT64 },
13959 { "conn_in_recvpktinfo", KSTAT_DATA_UINT64 },
13960 { "conn_in_recvtclass", KSTAT_DATA_UINT64 },
13961 { "conn_in_timestamp", KSTAT_DATA_UINT64 },
13962 };
13963
13964 ksp = kstat_create_netstack("ip", 0, "ipstat", "net",
13965 KSTAT_TYPE_NAMED, sizeof (template) / sizeof (kstat_named_t),
13966 KSTAT_FLAG_VIRTUAL, stackid);
13967
13968 if (ksp == NULL)
13969 return (NULL);
13970
13971 bcopy(&template, ip_statisticsp, sizeof (template));
13972 ksp->ks_data = (void *)ip_statisticsp;
13973 ksp->ks_private = (void *)(uintptr_t)stackid;
13974
13975 kstat_install(ksp);
13976 return (ksp);
13977 }
13978
13979 static void
13980 ip_kstat2_fini(netstackid_t stackid, kstat_t *ksp)
13981 {
13982 if (ksp != NULL) {
13983 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
13984 kstat_delete_netstack(ksp, stackid);
13985 }
13986 }
13987
13988 static void *
13989 ip_kstat_init(netstackid_t stackid, ip_stack_t *ipst)
13990 {
13991 kstat_t *ksp;
13992
13993 ip_named_kstat_t template = {
13994 { "forwarding", KSTAT_DATA_UINT32, 0 },
13995 { "defaultTTL", KSTAT_DATA_UINT32, 0 },
13996 { "inReceives", KSTAT_DATA_UINT64, 0 },
13997 { "inHdrErrors", KSTAT_DATA_UINT32, 0 },
13998 { "inAddrErrors", KSTAT_DATA_UINT32, 0 },
13999 { "forwDatagrams", KSTAT_DATA_UINT64, 0 },
14000 { "inUnknownProtos", KSTAT_DATA_UINT32, 0 },
14001 { "inDiscards", KSTAT_DATA_UINT32, 0 },
14002 { "inDelivers", KSTAT_DATA_UINT64, 0 },
14003 { "outRequests", KSTAT_DATA_UINT64, 0 },
14004 { "outDiscards", KSTAT_DATA_UINT32, 0 },
14005 { "outNoRoutes", KSTAT_DATA_UINT32, 0 },
14006 { "reasmTimeout", KSTAT_DATA_UINT32, 0 },
14007 { "reasmReqds", KSTAT_DATA_UINT32, 0 },
14008 { "reasmOKs", KSTAT_DATA_UINT32, 0 },
14009 { "reasmFails", KSTAT_DATA_UINT32, 0 },
14010 { "fragOKs", KSTAT_DATA_UINT32, 0 },
14011 { "fragFails", KSTAT_DATA_UINT32, 0 },
14012 { "fragCreates", KSTAT_DATA_UINT32, 0 },
14013 { "addrEntrySize", KSTAT_DATA_INT32, 0 },
14014 { "routeEntrySize", KSTAT_DATA_INT32, 0 },
14015 { "netToMediaEntrySize", KSTAT_DATA_INT32, 0 },
14016 { "routingDiscards", KSTAT_DATA_UINT32, 0 },
14017 { "inErrs", KSTAT_DATA_UINT32, 0 },
14018 { "noPorts", KSTAT_DATA_UINT32, 0 },
14019 { "inCksumErrs", KSTAT_DATA_UINT32, 0 },
14020 { "reasmDuplicates", KSTAT_DATA_UINT32, 0 },
14021 { "reasmPartDups", KSTAT_DATA_UINT32, 0 },
14022 { "forwProhibits", KSTAT_DATA_UINT32, 0 },
14023 { "udpInCksumErrs", KSTAT_DATA_UINT32, 0 },
14024 { "udpInOverflows", KSTAT_DATA_UINT32, 0 },
14025 { "rawipInOverflows", KSTAT_DATA_UINT32, 0 },
14026 { "ipsecInSucceeded", KSTAT_DATA_UINT32, 0 },
14027 { "ipsecInFailed", KSTAT_DATA_INT32, 0 },
14028 { "memberEntrySize", KSTAT_DATA_INT32, 0 },
14029 { "inIPv6", KSTAT_DATA_UINT32, 0 },
14030 { "outIPv6", KSTAT_DATA_UINT32, 0 },
14031 { "outSwitchIPv6", KSTAT_DATA_UINT32, 0 },
14032 };
14033
14034 ksp = kstat_create_netstack("ip", 0, "ip", "mib2", KSTAT_TYPE_NAMED,
14035 NUM_OF_FIELDS(ip_named_kstat_t), 0, stackid);
14036 if (ksp == NULL || ksp->ks_data == NULL)
14037 return (NULL);
14038
14039 template.forwarding.value.ui32 = WE_ARE_FORWARDING(ipst) ? 1:2;
14040 template.defaultTTL.value.ui32 = (uint32_t)ipst->ips_ip_def_ttl;
14041 template.reasmTimeout.value.ui32 = ipst->ips_ip_reassembly_timeout;
14042 template.addrEntrySize.value.i32 = sizeof (mib2_ipAddrEntry_t);
14043 template.routeEntrySize.value.i32 = sizeof (mib2_ipRouteEntry_t);
14044
14045 template.netToMediaEntrySize.value.i32 =
14046 sizeof (mib2_ipNetToMediaEntry_t);
14047
14048 template.memberEntrySize.value.i32 = sizeof (ipv6_member_t);
14049
14050 bcopy(&template, ksp->ks_data, sizeof (template));
14051 ksp->ks_update = ip_kstat_update;
14052 ksp->ks_private = (void *)(uintptr_t)stackid;
14053
14054 kstat_install(ksp);
14055 return (ksp);
14056 }
14057
14058 static void
14059 ip_kstat_fini(netstackid_t stackid, kstat_t *ksp)
14060 {
14061 if (ksp != NULL) {
14062 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
14063 kstat_delete_netstack(ksp, stackid);
14064 }
14065 }
14066
14067 static int
14068 ip_kstat_update(kstat_t *kp, int rw)
14069 {
14070 ip_named_kstat_t *ipkp;
14071 mib2_ipIfStatsEntry_t ipmib;
14072 ill_walk_context_t ctx;
14073 ill_t *ill;
14074 netstackid_t stackid = (zoneid_t)(uintptr_t)kp->ks_private;
14075 netstack_t *ns;
14076 ip_stack_t *ipst;
14077
14078 if (kp == NULL || kp->ks_data == NULL)
14079 return (EIO);
14080
14081 if (rw == KSTAT_WRITE)
14082 return (EACCES);
14083
14084 ns = netstack_find_by_stackid(stackid);
14085 if (ns == NULL)
14086 return (-1);
14087 ipst = ns->netstack_ip;
14088 if (ipst == NULL) {
14089 netstack_rele(ns);
14090 return (-1);
14091 }
14092 ipkp = (ip_named_kstat_t *)kp->ks_data;
14093
14094 bcopy(&ipst->ips_ip_mib, &ipmib, sizeof (ipmib));
14095 rw_enter(&ipst->ips_ill_g_lock, RW_READER);
14096 ill = ILL_START_WALK_V4(&ctx, ipst);
14097 for (; ill != NULL; ill = ill_next(&ctx, ill))
14098 ip_mib2_add_ip_stats(&ipmib, ill->ill_ip_mib);
14099 rw_exit(&ipst->ips_ill_g_lock);
14100
14101 ipkp->forwarding.value.ui32 = ipmib.ipIfStatsForwarding;
14102 ipkp->defaultTTL.value.ui32 = ipmib.ipIfStatsDefaultTTL;
14103 ipkp->inReceives.value.ui64 = ipmib.ipIfStatsHCInReceives;
14104 ipkp->inHdrErrors.value.ui32 = ipmib.ipIfStatsInHdrErrors;
14105 ipkp->inAddrErrors.value.ui32 = ipmib.ipIfStatsInAddrErrors;
14106 ipkp->forwDatagrams.value.ui64 = ipmib.ipIfStatsHCOutForwDatagrams;
14107 ipkp->inUnknownProtos.value.ui32 = ipmib.ipIfStatsInUnknownProtos;
14108 ipkp->inDiscards.value.ui32 = ipmib.ipIfStatsInDiscards;
14109 ipkp->inDelivers.value.ui64 = ipmib.ipIfStatsHCInDelivers;
14110 ipkp->outRequests.value.ui64 = ipmib.ipIfStatsHCOutRequests;
14111 ipkp->outDiscards.value.ui32 = ipmib.ipIfStatsOutDiscards;
14112 ipkp->outNoRoutes.value.ui32 = ipmib.ipIfStatsOutNoRoutes;
14113 ipkp->reasmTimeout.value.ui32 = ipst->ips_ip_reassembly_timeout;
14114 ipkp->reasmReqds.value.ui32 = ipmib.ipIfStatsReasmReqds;
14115 ipkp->reasmOKs.value.ui32 = ipmib.ipIfStatsReasmOKs;
14116 ipkp->reasmFails.value.ui32 = ipmib.ipIfStatsReasmFails;
14117 ipkp->fragOKs.value.ui32 = ipmib.ipIfStatsOutFragOKs;
14118 ipkp->fragFails.value.ui32 = ipmib.ipIfStatsOutFragFails;
14119 ipkp->fragCreates.value.ui32 = ipmib.ipIfStatsOutFragCreates;
14120
14121 ipkp->routingDiscards.value.ui32 = 0;
14122 ipkp->inErrs.value.ui32 = ipmib.tcpIfStatsInErrs;
14123 ipkp->noPorts.value.ui32 = ipmib.udpIfStatsNoPorts;
14124 ipkp->inCksumErrs.value.ui32 = ipmib.ipIfStatsInCksumErrs;
14125 ipkp->reasmDuplicates.value.ui32 = ipmib.ipIfStatsReasmDuplicates;
14126 ipkp->reasmPartDups.value.ui32 = ipmib.ipIfStatsReasmPartDups;
14127 ipkp->forwProhibits.value.ui32 = ipmib.ipIfStatsForwProhibits;
14128 ipkp->udpInCksumErrs.value.ui32 = ipmib.udpIfStatsInCksumErrs;
14129 ipkp->udpInOverflows.value.ui32 = ipmib.udpIfStatsInOverflows;
14130 ipkp->rawipInOverflows.value.ui32 = ipmib.rawipIfStatsInOverflows;
14131 ipkp->ipsecInSucceeded.value.ui32 = ipmib.ipsecIfStatsInSucceeded;
14132 ipkp->ipsecInFailed.value.i32 = ipmib.ipsecIfStatsInFailed;
14133
14134 ipkp->inIPv6.value.ui32 = ipmib.ipIfStatsInWrongIPVersion;
14135 ipkp->outIPv6.value.ui32 = ipmib.ipIfStatsOutWrongIPVersion;
14136 ipkp->outSwitchIPv6.value.ui32 = ipmib.ipIfStatsOutSwitchIPVersion;
14137
14138 netstack_rele(ns);
14139
14140 return (0);
14141 }
14142
14143 static void *
14144 icmp_kstat_init(netstackid_t stackid)
14145 {
14146 kstat_t *ksp;
14147
14148 icmp_named_kstat_t template = {
14149 { "inMsgs", KSTAT_DATA_UINT32 },
14150 { "inErrors", KSTAT_DATA_UINT32 },
14151 { "inDestUnreachs", KSTAT_DATA_UINT32 },
14152 { "inTimeExcds", KSTAT_DATA_UINT32 },
14153 { "inParmProbs", KSTAT_DATA_UINT32 },
14154 { "inSrcQuenchs", KSTAT_DATA_UINT32 },
14155 { "inRedirects", KSTAT_DATA_UINT32 },
14156 { "inEchos", KSTAT_DATA_UINT32 },
14157 { "inEchoReps", KSTAT_DATA_UINT32 },
14158 { "inTimestamps", KSTAT_DATA_UINT32 },
14159 { "inTimestampReps", KSTAT_DATA_UINT32 },
14160 { "inAddrMasks", KSTAT_DATA_UINT32 },
14161 { "inAddrMaskReps", KSTAT_DATA_UINT32 },
14162 { "outMsgs", KSTAT_DATA_UINT32 },
14163 { "outErrors", KSTAT_DATA_UINT32 },
14164 { "outDestUnreachs", KSTAT_DATA_UINT32 },
14165 { "outTimeExcds", KSTAT_DATA_UINT32 },
14166 { "outParmProbs", KSTAT_DATA_UINT32 },
14167 { "outSrcQuenchs", KSTAT_DATA_UINT32 },
14168 { "outRedirects", KSTAT_DATA_UINT32 },
14169 { "outEchos", KSTAT_DATA_UINT32 },
14170 { "outEchoReps", KSTAT_DATA_UINT32 },
14171 { "outTimestamps", KSTAT_DATA_UINT32 },
14172 { "outTimestampReps", KSTAT_DATA_UINT32 },
14173 { "outAddrMasks", KSTAT_DATA_UINT32 },
14174 { "outAddrMaskReps", KSTAT_DATA_UINT32 },
14175 { "inChksumErrs", KSTAT_DATA_UINT32 },
14176 { "inUnknowns", KSTAT_DATA_UINT32 },
14177 { "inFragNeeded", KSTAT_DATA_UINT32 },
14178 { "outFragNeeded", KSTAT_DATA_UINT32 },
14179 { "outDrops", KSTAT_DATA_UINT32 },
14180 { "inOverFlows", KSTAT_DATA_UINT32 },
14181 { "inBadRedirects", KSTAT_DATA_UINT32 },
14182 };
14183
14184 ksp = kstat_create_netstack("ip", 0, "icmp", "mib2", KSTAT_TYPE_NAMED,
14185 NUM_OF_FIELDS(icmp_named_kstat_t), 0, stackid);
14186 if (ksp == NULL || ksp->ks_data == NULL)
14187 return (NULL);
14188
14189 bcopy(&template, ksp->ks_data, sizeof (template));
14190
14191 ksp->ks_update = icmp_kstat_update;
14192 ksp->ks_private = (void *)(uintptr_t)stackid;
14193
14194 kstat_install(ksp);
14195 return (ksp);
14196 }
14197
14198 static void
14199 icmp_kstat_fini(netstackid_t stackid, kstat_t *ksp)
14200 {
14201 if (ksp != NULL) {
14202 ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
14203 kstat_delete_netstack(ksp, stackid);
14204 }
14205 }
14206
14207 static int
14208 icmp_kstat_update(kstat_t *kp, int rw)
14209 {
14210 icmp_named_kstat_t *icmpkp;
14211 netstackid_t stackid = (zoneid_t)(uintptr_t)kp->ks_private;
14212 netstack_t *ns;
14213 ip_stack_t *ipst;
14214
14215 if ((kp == NULL) || (kp->ks_data == NULL))
14216 return (EIO);
14217
14218 if (rw == KSTAT_WRITE)
14219 return (EACCES);
14220
14221 ns = netstack_find_by_stackid(stackid);
14222 if (ns == NULL)
14223 return (-1);
14224 ipst = ns->netstack_ip;
14225 if (ipst == NULL) {
14226 netstack_rele(ns);
14227 return (-1);
14228 }
14229 icmpkp = (icmp_named_kstat_t *)kp->ks_data;
14230
14231 icmpkp->inMsgs.value.ui32 = ipst->ips_icmp_mib.icmpInMsgs;
14232 icmpkp->inErrors.value.ui32 = ipst->ips_icmp_mib.icmpInErrors;
14233 icmpkp->inDestUnreachs.value.ui32 =
14234 ipst->ips_icmp_mib.icmpInDestUnreachs;
14235 icmpkp->inTimeExcds.value.ui32 = ipst->ips_icmp_mib.icmpInTimeExcds;
14236 icmpkp->inParmProbs.value.ui32 = ipst->ips_icmp_mib.icmpInParmProbs;
14237 icmpkp->inSrcQuenchs.value.ui32 = ipst->ips_icmp_mib.icmpInSrcQuenchs;
14238 icmpkp->inRedirects.value.ui32 = ipst->ips_icmp_mib.icmpInRedirects;
14239 icmpkp->inEchos.value.ui32 = ipst->ips_icmp_mib.icmpInEchos;
14240 icmpkp->inEchoReps.value.ui32 = ipst->ips_icmp_mib.icmpInEchoReps;
14241 icmpkp->inTimestamps.value.ui32 = ipst->ips_icmp_mib.icmpInTimestamps;
14242 icmpkp->inTimestampReps.value.ui32 =
14243 ipst->ips_icmp_mib.icmpInTimestampReps;
14244 icmpkp->inAddrMasks.value.ui32 = ipst->ips_icmp_mib.icmpInAddrMasks;
14245 icmpkp->inAddrMaskReps.value.ui32 =
14246 ipst->ips_icmp_mib.icmpInAddrMaskReps;
14247 icmpkp->outMsgs.value.ui32 = ipst->ips_icmp_mib.icmpOutMsgs;
14248 icmpkp->outErrors.value.ui32 = ipst->ips_icmp_mib.icmpOutErrors;
14249 icmpkp->outDestUnreachs.value.ui32 =
14250 ipst->ips_icmp_mib.icmpOutDestUnreachs;
14251 icmpkp->outTimeExcds.value.ui32 = ipst->ips_icmp_mib.icmpOutTimeExcds;
14252 icmpkp->outParmProbs.value.ui32 = ipst->ips_icmp_mib.icmpOutParmProbs;
14253 icmpkp->outSrcQuenchs.value.ui32 =
14254 ipst->ips_icmp_mib.icmpOutSrcQuenchs;
14255 icmpkp->outRedirects.value.ui32 = ipst->ips_icmp_mib.icmpOutRedirects;
14256 icmpkp->outEchos.value.ui32 = ipst->ips_icmp_mib.icmpOutEchos;
14257 icmpkp->outEchoReps.value.ui32 = ipst->ips_icmp_mib.icmpOutEchoReps;
14258 icmpkp->outTimestamps.value.ui32 =
14259 ipst->ips_icmp_mib.icmpOutTimestamps;
14260 icmpkp->outTimestampReps.value.ui32 =
14261 ipst->ips_icmp_mib.icmpOutTimestampReps;
14262 icmpkp->outAddrMasks.value.ui32 =
14263 ipst->ips_icmp_mib.icmpOutAddrMasks;
14264 icmpkp->outAddrMaskReps.value.ui32 =
14265 ipst->ips_icmp_mib.icmpOutAddrMaskReps;
14266 icmpkp->inCksumErrs.value.ui32 = ipst->ips_icmp_mib.icmpInCksumErrs;
14267 icmpkp->inUnknowns.value.ui32 = ipst->ips_icmp_mib.icmpInUnknowns;
14268 icmpkp->inFragNeeded.value.ui32 = ipst->ips_icmp_mib.icmpInFragNeeded;
14269 icmpkp->outFragNeeded.value.ui32 =
14270 ipst->ips_icmp_mib.icmpOutFragNeeded;
14271 icmpkp->outDrops.value.ui32 = ipst->ips_icmp_mib.icmpOutDrops;
14272 icmpkp->inOverflows.value.ui32 = ipst->ips_icmp_mib.icmpInOverflows;
14273 icmpkp->inBadRedirects.value.ui32 =
14274 ipst->ips_icmp_mib.icmpInBadRedirects;
14275
14276 netstack_rele(ns);
14277 return (0);
14278 }
14279
14280 /*
14281 * This is the fanout function for raw socket opened for SCTP. Note
14282 * that it is called after SCTP checks that there is no socket which
14283 * wants a packet. Then before SCTP handles this out of the blue packet,
14284 * this function is called to see if there is any raw socket for SCTP.
14285 * If there is and it is bound to the correct address, the packet will
14286 * be sent to that socket. Note that only one raw socket can be bound to
14287 * a port. This is assured in ipcl_sctp_hash_insert();
14288 */
14289 void
14290 ip_fanout_sctp_raw(mblk_t *mp, ipha_t *ipha, ip6_t *ip6h, uint32_t ports,
14291 ip_recv_attr_t *ira)
14292 {
14293 conn_t *connp;
14294 queue_t *rq;
14295 boolean_t secure;
14296 ill_t *ill = ira->ira_ill;
14297 ip_stack_t *ipst = ill->ill_ipst;
14298 ipsec_stack_t *ipss = ipst->ips_netstack->netstack_ipsec;
14299 sctp_stack_t *sctps = ipst->ips_netstack->netstack_sctp;
14300 iaflags_t iraflags = ira->ira_flags;
14301 ill_t *rill = ira->ira_rill;
14302
14303 secure = iraflags & IRAF_IPSEC_SECURE;
14304
14305 connp = ipcl_classify_raw(mp, IPPROTO_SCTP, ports, ipha, ip6h,
14306 ira, ipst);
14307 if (connp == NULL) {
14308 /*
14309 * Although raw sctp is not summed, OOB chunks must be.
14310 * Drop the packet here if the sctp checksum failed.
14311 */
14312 if (iraflags & IRAF_SCTP_CSUM_ERR) {
14313 SCTPS_BUMP_MIB(sctps, sctpChecksumError);
14314 freemsg(mp);
14315 return;
14316 }
14317 ira->ira_ill = ira->ira_rill = NULL;
14318 sctp_ootb_input(mp, ira, ipst);
14319 ira->ira_ill = ill;
14320 ira->ira_rill = rill;
14321 return;
14322 }
14323 rq = connp->conn_rq;
14324 if (IPCL_IS_NONSTR(connp) ? connp->conn_flow_cntrld : !canputnext(rq)) {
14325 CONN_DEC_REF(connp);
14326 BUMP_MIB(ill->ill_ip_mib, rawipIfStatsInOverflows);
14327 freemsg(mp);
14328 return;
14329 }
14330 if (((iraflags & IRAF_IS_IPV4) ?
14331 CONN_INBOUND_POLICY_PRESENT(connp, ipss) :
14332 CONN_INBOUND_POLICY_PRESENT_V6(connp, ipss)) ||
14333 secure) {
14334 mp = ipsec_check_inbound_policy(mp, connp, ipha,
14335 ip6h, ira);
14336 if (mp == NULL) {
14337 BUMP_MIB(ill->ill_ip_mib, ipIfStatsInDiscards);
14338 /* Note that mp is NULL */
14339 ip_drop_input("ipIfStatsInDiscards", mp, ill);
14340 CONN_DEC_REF(connp);
14341 return;
14342 }
14343 }
14344
14345 if (iraflags & IRAF_ICMP_ERROR) {
14346 (connp->conn_recvicmp)(connp, mp, NULL, ira);
14347 } else {
14348 ill_t *rill = ira->ira_rill;
14349
14350 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCInDelivers);
14351 /* This is the SOCK_RAW, IPPROTO_SCTP case. */
14352 ira->ira_ill = ira->ira_rill = NULL;
14353 (connp->conn_recv)(connp, mp, NULL, ira);
14354 ira->ira_ill = ill;
14355 ira->ira_rill = rill;
14356 }
14357 CONN_DEC_REF(connp);
14358 }
14359
14360 /*
14361 * Free a packet that has the link-layer dl_unitdata_req_t or fast-path
14362 * header before the ip payload.
14363 */
14364 static void
14365 ip_xmit_flowctl_drop(ill_t *ill, mblk_t *mp, boolean_t is_fp_mp, int fp_mp_len)
14366 {
14367 int len = (mp->b_wptr - mp->b_rptr);
14368 mblk_t *ip_mp;
14369
14370 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14371 if (is_fp_mp || len != fp_mp_len) {
14372 if (len > fp_mp_len) {
14373 /*
14374 * fastpath header and ip header in the first mblk
14375 */
14376 mp->b_rptr += fp_mp_len;
14377 } else {
14378 /*
14379 * ip_xmit_attach_llhdr had to prepend an mblk to
14380 * attach the fastpath header before ip header.
14381 */
14382 ip_mp = mp->b_cont;
14383 freeb(mp);
14384 mp = ip_mp;
14385 mp->b_rptr += (fp_mp_len - len);
14386 }
14387 } else {
14388 ip_mp = mp->b_cont;
14389 freeb(mp);
14390 mp = ip_mp;
14391 }
14392 ip_drop_output("ipIfStatsOutDiscards - flow ctl", mp, ill);
14393 freemsg(mp);
14394 }
14395
14396 /*
14397 * Normal post fragmentation function.
14398 *
14399 * Send a packet using the passed in nce. This handles both IPv4 and IPv6
14400 * using the same state machine.
14401 *
14402 * We return an error on failure. In particular we return EWOULDBLOCK
14403 * when the driver flow controls. In that case this ensures that ip_wsrv runs
14404 * (currently by canputnext failure resulting in backenabling from GLD.)
14405 * This allows the callers of conn_ip_output() to use EWOULDBLOCK as an
14406 * indication that they can flow control until ip_wsrv() tells then to restart.
14407 *
14408 * If the nce passed by caller is incomplete, this function
14409 * queues the packet and if necessary, sends ARP request and bails.
14410 * If the Neighbor Cache passed is fully resolved, we simply prepend
14411 * the link-layer header to the packet, do ipsec hw acceleration
14412 * work if necessary, and send the packet out on the wire.
14413 */
14414 /* ARGSUSED6 */
14415 int
14416 ip_xmit(mblk_t *mp, nce_t *nce, iaflags_t ixaflags, uint_t pkt_len,
14417 uint32_t xmit_hint, zoneid_t szone, zoneid_t nolzid, uintptr_t *ixacookie)
14418 {
14419 queue_t *wq;
14420 ill_t *ill = nce->nce_ill;
14421 ip_stack_t *ipst = ill->ill_ipst;
14422 uint64_t delta;
14423 boolean_t isv6 = ill->ill_isv6;
14424 boolean_t fp_mp;
14425 ncec_t *ncec = nce->nce_common;
14426 int64_t now = LBOLT_FASTPATH64;
14427 boolean_t is_probe;
14428
14429 DTRACE_PROBE1(ip__xmit, nce_t *, nce);
14430
14431 ASSERT(mp != NULL);
14432 ASSERT(mp->b_datap->db_type == M_DATA);
14433 ASSERT(pkt_len == msgdsize(mp));
14434
14435 /*
14436 * If we have already been here and are coming back after ARP/ND.
14437 * the IXAF_NO_TRACE flag is set. We skip FW_HOOKS, DTRACE and ipobs
14438 * in that case since they have seen the packet when it came here
14439 * the first time.
14440 */
14441 if (ixaflags & IXAF_NO_TRACE)
14442 goto sendit;
14443
14444 if (ixaflags & IXAF_IS_IPV4) {
14445 ipha_t *ipha = (ipha_t *)mp->b_rptr;
14446
14447 ASSERT(!isv6);
14448 ASSERT(pkt_len == ntohs(((ipha_t *)mp->b_rptr)->ipha_length));
14449 if (HOOKS4_INTERESTED_PHYSICAL_OUT(ipst) &&
14450 !(ixaflags & IXAF_NO_PFHOOK)) {
14451 int error;
14452
14453 FW_HOOKS(ipst->ips_ip4_physical_out_event,
14454 ipst->ips_ipv4firewall_physical_out,
14455 NULL, ill, ipha, mp, mp, 0, ipst, error);
14456 DTRACE_PROBE1(ip4__physical__out__end,
14457 mblk_t *, mp);
14458 if (mp == NULL)
14459 return (error);
14460
14461 /* The length could have changed */
14462 pkt_len = msgdsize(mp);
14463 }
14464 if (ipst->ips_ip4_observe.he_interested) {
14465 /*
14466 * Note that for TX the zoneid is the sending
14467 * zone, whether or not MLP is in play.
14468 * Since the szone argument is the IP zoneid (i.e.,
14469 * zero for exclusive-IP zones) and ipobs wants
14470 * the system zoneid, we map it here.
14471 */
14472 szone = IP_REAL_ZONEID(szone, ipst);
14473
14474 /*
14475 * On the outbound path the destination zone will be
14476 * unknown as we're sending this packet out on the
14477 * wire.
14478 */
14479 ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone, ALL_ZONES,
14480 ill, ipst);
14481 }
14482 DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
14483 void_ip_t *, ipha, __dtrace_ipsr_ill_t *, ill,
14484 ipha_t *, ipha, ip6_t *, NULL, int, 0);
14485 } else {
14486 ip6_t *ip6h = (ip6_t *)mp->b_rptr;
14487
14488 ASSERT(isv6);
14489 ASSERT(pkt_len ==
14490 ntohs(((ip6_t *)mp->b_rptr)->ip6_plen) + IPV6_HDR_LEN);
14491 if (HOOKS6_INTERESTED_PHYSICAL_OUT(ipst) &&
14492 !(ixaflags & IXAF_NO_PFHOOK)) {
14493 int error;
14494
14495 FW_HOOKS6(ipst->ips_ip6_physical_out_event,
14496 ipst->ips_ipv6firewall_physical_out,
14497 NULL, ill, ip6h, mp, mp, 0, ipst, error);
14498 DTRACE_PROBE1(ip6__physical__out__end,
14499 mblk_t *, mp);
14500 if (mp == NULL)
14501 return (error);
14502
14503 /* The length could have changed */
14504 pkt_len = msgdsize(mp);
14505 }
14506 if (ipst->ips_ip6_observe.he_interested) {
14507 /* See above */
14508 szone = IP_REAL_ZONEID(szone, ipst);
14509
14510 ipobs_hook(mp, IPOBS_HOOK_OUTBOUND, szone, ALL_ZONES,
14511 ill, ipst);
14512 }
14513 DTRACE_IP7(send, mblk_t *, mp, conn_t *, NULL,
14514 void_ip_t *, ip6h, __dtrace_ipsr_ill_t *, ill,
14515 ipha_t *, NULL, ip6_t *, ip6h, int, 0);
14516 }
14517
14518 sendit:
14519 /*
14520 * We check the state without a lock because the state can never
14521 * move "backwards" to initial or incomplete.
14522 */
14523 switch (ncec->ncec_state) {
14524 case ND_REACHABLE:
14525 case ND_STALE:
14526 case ND_DELAY:
14527 case ND_PROBE:
14528 mp = ip_xmit_attach_llhdr(mp, nce);
14529 if (mp == NULL) {
14530 /*
14531 * ip_xmit_attach_llhdr has increased
14532 * ipIfStatsOutDiscards and called ip_drop_output()
14533 */
14534 return (ENOBUFS);
14535 }
14536 /*
14537 * check if nce_fastpath completed and we tagged on a
14538 * copy of nce_fp_mp in ip_xmit_attach_llhdr().
14539 */
14540 fp_mp = (mp->b_datap->db_type == M_DATA);
14541
14542 if (fp_mp &&
14543 (ill->ill_capabilities & ILL_CAPAB_DLD_DIRECT)) {
14544 ill_dld_direct_t *idd;
14545
14546 idd = &ill->ill_dld_capab->idc_direct;
14547 /*
14548 * Send the packet directly to DLD, where it
14549 * may be queued depending on the availability
14550 * of transmit resources at the media layer.
14551 * Return value should be taken into
14552 * account and flow control the TCP.
14553 */
14554 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
14555 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
14556 pkt_len);
14557
14558 if (ixaflags & IXAF_NO_DEV_FLOW_CTL) {
14559 (void) idd->idd_tx_df(idd->idd_tx_dh, mp,
14560 (uintptr_t)xmit_hint, IP_DROP_ON_NO_DESC);
14561 } else {
14562 uintptr_t cookie;
14563
14564 if ((cookie = idd->idd_tx_df(idd->idd_tx_dh,
14565 mp, (uintptr_t)xmit_hint, 0)) != 0) {
14566 if (ixacookie != NULL)
14567 *ixacookie = cookie;
14568 return (EWOULDBLOCK);
14569 }
14570 }
14571 } else {
14572 wq = ill->ill_wq;
14573
14574 if (!(ixaflags & IXAF_NO_DEV_FLOW_CTL) &&
14575 !canputnext(wq)) {
14576 if (ixacookie != NULL)
14577 *ixacookie = 0;
14578 ip_xmit_flowctl_drop(ill, mp, fp_mp,
14579 nce->nce_fp_mp != NULL ?
14580 MBLKL(nce->nce_fp_mp) : 0);
14581 return (EWOULDBLOCK);
14582 }
14583 BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
14584 UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
14585 pkt_len);
14586 putnext(wq, mp);
14587 }
14588
14589 /*
14590 * The rest of this function implements Neighbor Unreachability
14591 * detection. Determine if the ncec is eligible for NUD.
14592 */
14593 if (ncec->ncec_flags & NCE_F_NONUD)
14594 return (0);
14595
14596 ASSERT(ncec->ncec_state != ND_INCOMPLETE);
14597
14598 /*
14599 * Check for upper layer advice
14600 */
14601 if (ixaflags & IXAF_REACH_CONF) {
14602 timeout_id_t tid;
14603
14604 /*
14605 * It should be o.k. to check the state without
14606 * a lock here, at most we lose an advice.
14607 */
14608 ncec->ncec_last = TICK_TO_MSEC(now);
14609 if (ncec->ncec_state != ND_REACHABLE) {
14610 mutex_enter(&ncec->ncec_lock);
14611 ncec->ncec_state = ND_REACHABLE;
14612 tid = ncec->ncec_timeout_id;
14613 ncec->ncec_timeout_id = 0;
14614 mutex_exit(&ncec->ncec_lock);
14615 (void) untimeout(tid);
14616 if (ip_debug > 2) {
14617 /* ip1dbg */
14618 pr_addr_dbg("ip_xmit: state"
14619 " for %s changed to"
14620 " REACHABLE\n", AF_INET6,
14621 &ncec->ncec_addr);
14622 }
14623 }
14624 return (0);
14625 }
14626
14627 delta = TICK_TO_MSEC(now) - ncec->ncec_last;
14628 ip1dbg(("ip_xmit: delta = %" PRId64
14629 " ill_reachable_time = %d \n", delta,
14630 ill->ill_reachable_time));
14631 if (delta > (uint64_t)ill->ill_reachable_time) {
14632 mutex_enter(&ncec->ncec_lock);
14633 switch (ncec->ncec_state) {
14634 case ND_REACHABLE:
14635 ASSERT((ncec->ncec_flags & NCE_F_NONUD) == 0);
14636 /* FALLTHROUGH */
14637 case ND_STALE:
14638 /*
14639 * ND_REACHABLE is identical to
14640 * ND_STALE in this specific case. If
14641 * reachable time has expired for this
14642 * neighbor (delta is greater than
14643 * reachable time), conceptually, the
14644 * neighbor cache is no longer in
14645 * REACHABLE state, but already in
14646 * STALE state. So the correct
14647 * transition here is to ND_DELAY.
14648 */
14649 ncec->ncec_state = ND_DELAY;
14650 mutex_exit(&ncec->ncec_lock);
14651 nce_restart_timer(ncec,
14652 ipst->ips_delay_first_probe_time);
14653 if (ip_debug > 3) {
14654 /* ip2dbg */
14655 pr_addr_dbg("ip_xmit: state"
14656 " for %s changed to"
14657 " DELAY\n", AF_INET6,
14658 &ncec->ncec_addr);
14659 }
14660 break;
14661 case ND_DELAY:
14662 case ND_PROBE:
14663 mutex_exit(&ncec->ncec_lock);
14664 /* Timers have already started */
14665 break;
14666 case ND_UNREACHABLE:
14667 /*
14668 * nce_timer has detected that this ncec
14669 * is unreachable and initiated deleting
14670 * this ncec.
14671 * This is a harmless race where we found the
14672 * ncec before it was deleted and have
14673 * just sent out a packet using this
14674 * unreachable ncec.
14675 */
14676 mutex_exit(&ncec->ncec_lock);
14677 break;
14678 default:
14679 ASSERT(0);
14680 mutex_exit(&ncec->ncec_lock);
14681 }
14682 }
14683 return (0);
14684
14685 case ND_INCOMPLETE:
14686 /*
14687 * the state could have changed since we didn't hold the lock.
14688 * Re-verify state under lock.
14689 */
14690 is_probe = ipmp_packet_is_probe(mp, nce->nce_ill);
14691 mutex_enter(&ncec->ncec_lock);
14692 if (NCE_ISREACHABLE(ncec)) {
14693 mutex_exit(&ncec->ncec_lock);
14694 goto sendit;
14695 }
14696 /* queue the packet */
14697 nce_queue_mp(ncec, mp, is_probe);
14698 mutex_exit(&ncec->ncec_lock);
14699 DTRACE_PROBE2(ip__xmit__incomplete,
14700 (ncec_t *), ncec, (mblk_t *), mp);
14701 return (0);
14702
14703 case ND_INITIAL:
14704 /*
14705 * State could have changed since we didn't hold the lock, so
14706 * re-verify state.
14707 */
14708 is_probe = ipmp_packet_is_probe(mp, nce->nce_ill);
14709 mutex_enter(&ncec->ncec_lock);
14710 if (NCE_ISREACHABLE(ncec)) {
14711 mutex_exit(&ncec->ncec_lock);
14712 goto sendit;
14713 }
14714 nce_queue_mp(ncec, mp, is_probe);
14715 if (ncec->ncec_state == ND_INITIAL) {
14716 ncec->ncec_state = ND_INCOMPLETE;
14717 mutex_exit(&ncec->ncec_lock);
14718 /*
14719 * figure out the source we want to use
14720 * and resolve it.
14721 */
14722 ip_ndp_resolve(ncec);
14723 } else {
14724 mutex_exit(&ncec->ncec_lock);
14725 }
14726 return (0);
14727
14728 case ND_UNREACHABLE:
14729 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14730 ip_drop_output("ipIfStatsOutDiscards - ND_UNREACHABLE",
14731 mp, ill);
14732 freemsg(mp);
14733 return (0);
14734
14735 default:
14736 ASSERT(0);
14737 BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
14738 ip_drop_output("ipIfStatsOutDiscards - ND_other",
14739 mp, ill);
14740 freemsg(mp);
14741 return (ENETUNREACH);
14742 }
14743 }
14744
14745 /*
14746 * Return B_TRUE if the buffers differ in length or content.
14747 * This is used for comparing extension header buffers.
14748 * Note that an extension header would be declared different
14749 * even if all that changed was the next header value in that header i.e.
14750 * what really changed is the next extension header.
14751 */
14752 boolean_t
14753 ip_cmpbuf(const void *abuf, uint_t alen, boolean_t b_valid, const void *bbuf,
14754 uint_t blen)
14755 {
14756 if (!b_valid)
14757 blen = 0;
14758
14759 if (alen != blen)
14760 return (B_TRUE);
14761 if (alen == 0)
14762 return (B_FALSE); /* Both zero length */
14763 return (bcmp(abuf, bbuf, alen));
14764 }
14765
14766 /*
14767 * Preallocate memory for ip_savebuf(). Returns B_TRUE if ok.
14768 * Return B_FALSE if memory allocation fails - don't change any state!
14769 */
14770 boolean_t
14771 ip_allocbuf(void **dstp, uint_t *dstlenp, boolean_t src_valid,
14772 const void *src, uint_t srclen)
14773 {
14774 void *dst;
14775
14776 if (!src_valid)
14777 srclen = 0;
14778
14779 ASSERT(*dstlenp == 0);
14780 if (src != NULL && srclen != 0) {
14781 dst = mi_alloc(srclen, BPRI_MED);
14782 if (dst == NULL)
14783 return (B_FALSE);
14784 } else {
14785 dst = NULL;
14786 }
14787 if (*dstp != NULL)
14788 mi_free(*dstp);
14789 *dstp = dst;
14790 *dstlenp = dst == NULL ? 0 : srclen;
14791 return (B_TRUE);
14792 }
14793
14794 /*
14795 * Replace what is in *dst, *dstlen with the source.
14796 * Assumes ip_allocbuf has already been called.
14797 */
14798 void
14799 ip_savebuf(void **dstp, uint_t *dstlenp, boolean_t src_valid,
14800 const void *src, uint_t srclen)
14801 {
14802 if (!src_valid)
14803 srclen = 0;
14804
14805 ASSERT(*dstlenp == srclen);
14806 if (src != NULL && srclen != 0)
14807 bcopy(src, *dstp, srclen);
14808 }
14809
14810 /*
14811 * Free the storage pointed to by the members of an ip_pkt_t.
14812 */
14813 void
14814 ip_pkt_free(ip_pkt_t *ipp)
14815 {
14816 uint_t fields = ipp->ipp_fields;
14817
14818 if (fields & IPPF_HOPOPTS) {
14819 kmem_free(ipp->ipp_hopopts, ipp->ipp_hopoptslen);
14820 ipp->ipp_hopopts = NULL;
14821 ipp->ipp_hopoptslen = 0;
14822 }
14823 if (fields & IPPF_RTHDRDSTOPTS) {
14824 kmem_free(ipp->ipp_rthdrdstopts, ipp->ipp_rthdrdstoptslen);
14825 ipp->ipp_rthdrdstopts = NULL;
14826 ipp->ipp_rthdrdstoptslen = 0;
14827 }
14828 if (fields & IPPF_DSTOPTS) {
14829 kmem_free(ipp->ipp_dstopts, ipp->ipp_dstoptslen);
14830 ipp->ipp_dstopts = NULL;
14831 ipp->ipp_dstoptslen = 0;
14832 }
14833 if (fields & IPPF_RTHDR) {
14834 kmem_free(ipp->ipp_rthdr, ipp->ipp_rthdrlen);
14835 ipp->ipp_rthdr = NULL;
14836 ipp->ipp_rthdrlen = 0;
14837 }
14838 if (fields & IPPF_IPV4_OPTIONS) {
14839 kmem_free(ipp->ipp_ipv4_options, ipp->ipp_ipv4_options_len);
14840 ipp->ipp_ipv4_options = NULL;
14841 ipp->ipp_ipv4_options_len = 0;
14842 }
14843 if (fields & IPPF_LABEL_V4) {
14844 kmem_free(ipp->ipp_label_v4, ipp->ipp_label_len_v4);
14845 ipp->ipp_label_v4 = NULL;
14846 ipp->ipp_label_len_v4 = 0;
14847 }
14848 if (fields & IPPF_LABEL_V6) {
14849 kmem_free(ipp->ipp_label_v6, ipp->ipp_label_len_v6);
14850 ipp->ipp_label_v6 = NULL;
14851 ipp->ipp_label_len_v6 = 0;
14852 }
14853 ipp->ipp_fields &= ~(IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14854 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6);
14855 }
14856
14857 /*
14858 * Copy from src to dst and allocate as needed.
14859 * Returns zero or ENOMEM.
14860 *
14861 * The caller must initialize dst to zero.
14862 */
14863 int
14864 ip_pkt_copy(ip_pkt_t *src, ip_pkt_t *dst, int kmflag)
14865 {
14866 uint_t fields = src->ipp_fields;
14867
14868 /* Start with fields that don't require memory allocation */
14869 dst->ipp_fields = fields &
14870 ~(IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14871 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6);
14872
14873 dst->ipp_addr = src->ipp_addr;
14874 dst->ipp_unicast_hops = src->ipp_unicast_hops;
14875 dst->ipp_hoplimit = src->ipp_hoplimit;
14876 dst->ipp_tclass = src->ipp_tclass;
14877 dst->ipp_type_of_service = src->ipp_type_of_service;
14878
14879 if (!(fields & (IPPF_HOPOPTS | IPPF_RTHDRDSTOPTS | IPPF_DSTOPTS |
14880 IPPF_RTHDR | IPPF_IPV4_OPTIONS | IPPF_LABEL_V4 | IPPF_LABEL_V6)))
14881 return (0);
14882
14883 if (fields & IPPF_HOPOPTS) {
14884 dst->ipp_hopopts = kmem_alloc(src->ipp_hopoptslen, kmflag);
14885 if (dst->ipp_hopopts == NULL) {
14886 ip_pkt_free(dst);
14887 return (ENOMEM);
14888 }
14889 dst->ipp_fields |= IPPF_HOPOPTS;
14890 bcopy(src->ipp_hopopts, dst->ipp_hopopts,
14891 src->ipp_hopoptslen);
14892 dst->ipp_hopoptslen = src->ipp_hopoptslen;
14893 }
14894 if (fields & IPPF_RTHDRDSTOPTS) {
14895 dst->ipp_rthdrdstopts = kmem_alloc(src->ipp_rthdrdstoptslen,
14896 kmflag);
14897 if (dst->ipp_rthdrdstopts == NULL) {
14898 ip_pkt_free(dst);
14899 return (ENOMEM);
14900 }
14901 dst->ipp_fields |= IPPF_RTHDRDSTOPTS;
14902 bcopy(src->ipp_rthdrdstopts, dst->ipp_rthdrdstopts,
14903 src->ipp_rthdrdstoptslen);
14904 dst->ipp_rthdrdstoptslen = src->ipp_rthdrdstoptslen;
14905 }
14906 if (fields & IPPF_DSTOPTS) {
14907 dst->ipp_dstopts = kmem_alloc(src->ipp_dstoptslen, kmflag);
14908 if (dst->ipp_dstopts == NULL) {
14909 ip_pkt_free(dst);
14910 return (ENOMEM);
14911 }
14912 dst->ipp_fields |= IPPF_DSTOPTS;
14913 bcopy(src->ipp_dstopts, dst->ipp_dstopts,
14914 src->ipp_dstoptslen);
14915 dst->ipp_dstoptslen = src->ipp_dstoptslen;
14916 }
14917 if (fields & IPPF_RTHDR) {
14918 dst->ipp_rthdr = kmem_alloc(src->ipp_rthdrlen, kmflag);
14919 if (dst->ipp_rthdr == NULL) {
14920 ip_pkt_free(dst);
14921 return (ENOMEM);
14922 }
14923 dst->ipp_fields |= IPPF_RTHDR;
14924 bcopy(src->ipp_rthdr, dst->ipp_rthdr,
14925 src->ipp_rthdrlen);
14926 dst->ipp_rthdrlen = src->ipp_rthdrlen;
14927 }
14928 if (fields & IPPF_IPV4_OPTIONS) {
14929 dst->ipp_ipv4_options = kmem_alloc(src->ipp_ipv4_options_len,
14930 kmflag);
14931 if (dst->ipp_ipv4_options == NULL) {
14932 ip_pkt_free(dst);
14933 return (ENOMEM);
14934 }
14935 dst->ipp_fields |= IPPF_IPV4_OPTIONS;
14936 bcopy(src->ipp_ipv4_options, dst->ipp_ipv4_options,
14937 src->ipp_ipv4_options_len);
14938 dst->ipp_ipv4_options_len = src->ipp_ipv4_options_len;
14939 }
14940 if (fields & IPPF_LABEL_V4) {
14941 dst->ipp_label_v4 = kmem_alloc(src->ipp_label_len_v4, kmflag);
14942 if (dst->ipp_label_v4 == NULL) {
14943 ip_pkt_free(dst);
14944 return (ENOMEM);
14945 }
14946 dst->ipp_fields |= IPPF_LABEL_V4;
14947 bcopy(src->ipp_label_v4, dst->ipp_label_v4,
14948 src->ipp_label_len_v4);
14949 dst->ipp_label_len_v4 = src->ipp_label_len_v4;
14950 }
14951 if (fields & IPPF_LABEL_V6) {
14952 dst->ipp_label_v6 = kmem_alloc(src->ipp_label_len_v6, kmflag);
14953 if (dst->ipp_label_v6 == NULL) {
14954 ip_pkt_free(dst);
14955 return (ENOMEM);
14956 }
14957 dst->ipp_fields |= IPPF_LABEL_V6;
14958 bcopy(src->ipp_label_v6, dst->ipp_label_v6,
14959 src->ipp_label_len_v6);
14960 dst->ipp_label_len_v6 = src->ipp_label_len_v6;
14961 }
14962 if (fields & IPPF_FRAGHDR) {
14963 dst->ipp_fraghdr = kmem_alloc(src->ipp_fraghdrlen, kmflag);
14964 if (dst->ipp_fraghdr == NULL) {
14965 ip_pkt_free(dst);
14966 return (ENOMEM);
14967 }
14968 dst->ipp_fields |= IPPF_FRAGHDR;
14969 bcopy(src->ipp_fraghdr, dst->ipp_fraghdr,
14970 src->ipp_fraghdrlen);
14971 dst->ipp_fraghdrlen = src->ipp_fraghdrlen;
14972 }
14973 return (0);
14974 }
14975
14976 /*
14977 * Returns INADDR_ANY if no source route
14978 */
14979 ipaddr_t
14980 ip_pkt_source_route_v4(const ip_pkt_t *ipp)
14981 {
14982 ipaddr_t nexthop = INADDR_ANY;
14983 ipoptp_t opts;
14984 uchar_t *opt;
14985 uint8_t optval;
14986 uint8_t optlen;
14987 uint32_t totallen;
14988
14989 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
14990 return (INADDR_ANY);
14991
14992 totallen = ipp->ipp_ipv4_options_len;
14993 if (totallen & 0x3)
14994 return (INADDR_ANY);
14995
14996 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
14997 optval != IPOPT_EOL;
14998 optval = ipoptp_next(&opts)) {
14999 opt = opts.ipoptp_cur;
15000 switch (optval) {
15001 uint8_t off;
15002 case IPOPT_SSRR:
15003 case IPOPT_LSRR:
15004 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
15005 break;
15006 }
15007 optlen = opts.ipoptp_len;
15008 off = opt[IPOPT_OFFSET];
15009 off--;
15010 if (optlen < IP_ADDR_LEN ||
15011 off > optlen - IP_ADDR_LEN) {
15012 /* End of source route */
15013 break;
15014 }
15015 bcopy((char *)opt + off, &nexthop, IP_ADDR_LEN);
15016 if (nexthop == htonl(INADDR_LOOPBACK)) {
15017 /* Ignore */
15018 nexthop = INADDR_ANY;
15019 break;
15020 }
15021 break;
15022 }
15023 }
15024 return (nexthop);
15025 }
15026
15027 /*
15028 * Reverse a source route.
15029 */
15030 void
15031 ip_pkt_source_route_reverse_v4(ip_pkt_t *ipp)
15032 {
15033 ipaddr_t tmp;
15034 ipoptp_t opts;
15035 uchar_t *opt;
15036 uint8_t optval;
15037 uint32_t totallen;
15038
15039 if (!(ipp->ipp_fields & IPPF_IPV4_OPTIONS))
15040 return;
15041
15042 totallen = ipp->ipp_ipv4_options_len;
15043 if (totallen & 0x3)
15044 return;
15045
15046 for (optval = ipoptp_first2(&opts, totallen, ipp->ipp_ipv4_options);
15047 optval != IPOPT_EOL;
15048 optval = ipoptp_next(&opts)) {
15049 uint8_t off1, off2;
15050
15051 opt = opts.ipoptp_cur;
15052 switch (optval) {
15053 case IPOPT_SSRR:
15054 case IPOPT_LSRR:
15055 if ((opts.ipoptp_flags & IPOPTP_ERROR) != 0) {
15056 break;
15057 }
15058 off1 = IPOPT_MINOFF_SR - 1;
15059 off2 = opt[IPOPT_OFFSET] - IP_ADDR_LEN - 1;
15060 while (off2 > off1) {
15061 bcopy(opt + off2, &tmp, IP_ADDR_LEN);
15062 bcopy(opt + off1, opt + off2, IP_ADDR_LEN);
15063 bcopy(&tmp, opt + off2, IP_ADDR_LEN);
15064 off2 -= IP_ADDR_LEN;
15065 off1 += IP_ADDR_LEN;
15066 }
15067 opt[IPOPT_OFFSET] = IPOPT_MINOFF_SR;
15068 break;
15069 }
15070 }
15071 }
15072
15073 /*
15074 * Returns NULL if no routing header
15075 */
15076 in6_addr_t *
15077 ip_pkt_source_route_v6(const ip_pkt_t *ipp)
15078 {
15079 in6_addr_t *nexthop = NULL;
15080 ip6_rthdr0_t *rthdr;
15081
15082 if (!(ipp->ipp_fields & IPPF_RTHDR))
15083 return (NULL);
15084
15085 rthdr = (ip6_rthdr0_t *)ipp->ipp_rthdr;
15086 if (rthdr->ip6r0_segleft == 0)
15087 return (NULL);
15088
15089 nexthop = (in6_addr_t *)((char *)rthdr + sizeof (*rthdr));
15090 return (nexthop);
15091 }
15092
15093 zoneid_t
15094 ip_get_zoneid_v4(ipaddr_t addr, mblk_t *mp, ip_recv_attr_t *ira,
15095 zoneid_t lookup_zoneid)
15096 {
15097 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
15098 ire_t *ire;
15099 int ire_flags = MATCH_IRE_TYPE;
15100 zoneid_t zoneid = ALL_ZONES;
15101
15102 if (is_system_labeled() && !tsol_can_accept_raw(mp, ira, B_FALSE))
15103 return (ALL_ZONES);
15104
15105 if (lookup_zoneid != ALL_ZONES)
15106 ire_flags |= MATCH_IRE_ZONEONLY;
15107 ire = ire_ftable_lookup_v4(addr, NULL, NULL, IRE_LOCAL | IRE_LOOPBACK,
15108 NULL, lookup_zoneid, NULL, ire_flags, 0, ipst, NULL);
15109 if (ire != NULL) {
15110 zoneid = IP_REAL_ZONEID(ire->ire_zoneid, ipst);
15111 ire_refrele(ire);
15112 }
15113 return (zoneid);
15114 }
15115
15116 zoneid_t
15117 ip_get_zoneid_v6(in6_addr_t *addr, mblk_t *mp, const ill_t *ill,
15118 ip_recv_attr_t *ira, zoneid_t lookup_zoneid)
15119 {
15120 ip_stack_t *ipst = ira->ira_ill->ill_ipst;
15121 ire_t *ire;
15122 int ire_flags = MATCH_IRE_TYPE;
15123 zoneid_t zoneid = ALL_ZONES;
15124
15125 if (is_system_labeled() && !tsol_can_accept_raw(mp, ira, B_FALSE))
15126 return (ALL_ZONES);
15127
15128 if (IN6_IS_ADDR_LINKLOCAL(addr))
15129 ire_flags |= MATCH_IRE_ILL;
15130
15131 if (lookup_zoneid != ALL_ZONES)
15132 ire_flags |= MATCH_IRE_ZONEONLY;
15133 ire = ire_ftable_lookup_v6(addr, NULL, NULL, IRE_LOCAL | IRE_LOOPBACK,
15134 ill, lookup_zoneid, NULL, ire_flags, 0, ipst, NULL);
15135 if (ire != NULL) {
15136 zoneid = IP_REAL_ZONEID(ire->ire_zoneid, ipst);
15137 ire_refrele(ire);
15138 }
15139 return (zoneid);
15140 }
15141
15142 /*
15143 * IP obserability hook support functions.
15144 */
15145 static void
15146 ipobs_init(ip_stack_t *ipst)
15147 {
15148 netid_t id;
15149
15150 id = net_getnetidbynetstackid(ipst->ips_netstack->netstack_stackid);
15151
15152 ipst->ips_ip4_observe_pr = net_protocol_lookup(id, NHF_INET);
15153 VERIFY(ipst->ips_ip4_observe_pr != NULL);
15154
15155 ipst->ips_ip6_observe_pr = net_protocol_lookup(id, NHF_INET6);
15156 VERIFY(ipst->ips_ip6_observe_pr != NULL);
15157 }
15158
15159 static void
15160 ipobs_fini(ip_stack_t *ipst)
15161 {
15162
15163 VERIFY(net_protocol_release(ipst->ips_ip4_observe_pr) == 0);
15164 VERIFY(net_protocol_release(ipst->ips_ip6_observe_pr) == 0);
15165 }
15166
15167 /*
15168 * hook_pkt_observe_t is composed in network byte order so that the
15169 * entire mblk_t chain handed into hook_run can be used as-is.
15170 * The caveat is that use of the fields, such as the zone fields,
15171 * requires conversion into host byte order first.
15172 */
15173 void
15174 ipobs_hook(mblk_t *mp, int htype, zoneid_t zsrc, zoneid_t zdst,
15175 const ill_t *ill, ip_stack_t *ipst)
15176 {
15177 hook_pkt_observe_t *hdr;
15178 uint64_t grifindex;
15179 mblk_t *imp;
15180
15181 imp = allocb(sizeof (*hdr), BPRI_HI);
15182 if (imp == NULL)
15183 return;
15184
15185 hdr = (hook_pkt_observe_t *)imp->b_rptr;
15186 /*
15187 * b_wptr is set to make the apparent size of the data in the mblk_t
15188 * to exclude the pointers at the end of hook_pkt_observer_t.
15189 */
15190 imp->b_wptr = imp->b_rptr + sizeof (dl_ipnetinfo_t);
15191 imp->b_cont = mp;
15192
15193 ASSERT(DB_TYPE(mp) == M_DATA);
15194
15195 if (IS_UNDER_IPMP(ill))
15196 grifindex = ipmp_ill_get_ipmp_ifindex(ill);
15197 else
15198 grifindex = 0;
15199
15200 hdr->hpo_version = 1;
15201 hdr->hpo_htype = htons(htype);
15202 hdr->hpo_pktlen = htonl((ulong_t)msgdsize(mp));
15203 hdr->hpo_ifindex = htonl(ill->ill_phyint->phyint_ifindex);
15204 hdr->hpo_grifindex = htonl(grifindex);
15205 hdr->hpo_zsrc = htonl(zsrc);
15206 hdr->hpo_zdst = htonl(zdst);
15207 hdr->hpo_pkt = imp;
15208 hdr->hpo_ctx = ipst->ips_netstack;
15209
15210 if (ill->ill_isv6) {
15211 hdr->hpo_family = AF_INET6;
15212 (void) hook_run(ipst->ips_ipv6_net_data->netd_hooks,
15213 ipst->ips_ipv6observing, (hook_data_t)hdr);
15214 } else {
15215 hdr->hpo_family = AF_INET;
15216 (void) hook_run(ipst->ips_ipv4_net_data->netd_hooks,
15217 ipst->ips_ipv4observing, (hook_data_t)hdr);
15218 }
15219
15220 imp->b_cont = NULL;
15221 freemsg(imp);
15222 }
15223
15224 /*
15225 * Utility routine that checks if `v4srcp' is a valid address on underlying
15226 * interface `ill'. If `ipifp' is non-NULL, it's set to a held ipif
15227 * associated with `v4srcp' on success. NOTE: if this is not called from
15228 * inside the IPSQ (ill_g_lock is not held), `ill' may be removed from the
15229 * group during or after this lookup.
15230 */
15231 boolean_t
15232 ipif_lookup_testaddr_v4(ill_t *ill, const in_addr_t *v4srcp, ipif_t **ipifp)
15233 {
15234 ipif_t *ipif;
15235
15236 ipif = ipif_lookup_addr_exact(*v4srcp, ill, ill->ill_ipst);
15237 if (ipif != NULL) {
15238 if (ipifp != NULL)
15239 *ipifp = ipif;
15240 else
15241 ipif_refrele(ipif);
15242 return (B_TRUE);
15243 }
15244
15245 ip1dbg(("ipif_lookup_testaddr_v4: cannot find ipif for src %x\n",
15246 *v4srcp));
15247 return (B_FALSE);
15248 }
15249
15250 /*
15251 * Transport protocol call back function for CPU state change.
15252 */
15253 /* ARGSUSED */
15254 static int
15255 ip_tp_cpu_update(cpu_setup_t what, int id, void *arg)
15256 {
15257 processorid_t cpu_seqid;
15258 netstack_handle_t nh;
15259 netstack_t *ns;
15260
15261 ASSERT(MUTEX_HELD(&cpu_lock));
15262
15263 switch (what) {
15264 case CPU_CONFIG:
15265 case CPU_ON:
15266 case CPU_INIT:
15267 case CPU_CPUPART_IN:
15268 cpu_seqid = cpu[id]->cpu_seqid;
15269 netstack_next_init(&nh);
15270 while ((ns = netstack_next(&nh)) != NULL) {
15271 tcp_stack_cpu_add(ns->netstack_tcp, cpu_seqid);
15272 sctp_stack_cpu_add(ns->netstack_sctp, cpu_seqid);
15273 udp_stack_cpu_add(ns->netstack_udp, cpu_seqid);
15274 netstack_rele(ns);
15275 }
15276 netstack_next_fini(&nh);
15277 break;
15278 case CPU_UNCONFIG:
15279 case CPU_OFF:
15280 case CPU_CPUPART_OUT:
15281 /*
15282 * Nothing to do. We don't remove the per CPU stats from
15283 * the IP stack even when the CPU goes offline.
15284 */
15285 break;
15286 default:
15287 break;
15288 }
15289 return (0);
15290 }